Basic physical mechanisms for monsoon failure in past and future …schewe/publications/Dissertation... · 2011-05-09 · At the same time, paleoclimatic records show evidence of

Institut fur Physik und Astronomie

Arbeitsgruppe Prof. A. Levermann

Basic physical mechanisms for monsoon failure inpast and future climate

Kumulative Dissertation

zur Erlangung des akademischen Grades

“doctor rerum naturalium” (Dr. rer. nat.)

in der Wissenschaftsdisziplin “Klimaphysik”

eingereicht an der

Mathematisch-Naturwissenschaftlichen Fakultat

der Universitat Potsdam

von

Jacob Schewe

Potsdam, im Mai 2011

Abstract

In this work, I first present simulations with a coupled climate model of intermediate complexity,

CLIMBER-3α, that project monsoon rainfall around the world to increase quasi–linearly with global

warming in the coming centuries. While this is generally consistent with many other studies, the atmo-

spheric component of CLIMBER-3α is based on a simplified statistical–dynamical approach, and may

not sufficiently represent all processes that are relevant for the response of monsoon circulations to rapid

and intense climate change. Therefore, this study attempts to identify those physical mechanisms that

are of first–order importance for large–scale monsoon dynamics and their response to external changes.

I perform a scaling analysis of the heat and moisture budgets of the Earth’s major monsoon systems,

based on reanalysis data and theoretical considerations. I find that, during the monsoon season, a self–

amplifying feedback involving the advection and release of latent heat is essential for sustaining monsoon

rainfall after the surface land–sea thermal contrast has ceased. I frame this moisture–advection feedback

in a minimal conceptual model and show that it leads to a threshold behaviour with respect to changes in

the system’s energy budget. In particular, when either net radiation over land or specific humidity over

the adjacent ocean region fall short of a critical value, no conventional monsoon circulation can exist.

I thus define a domain of existence for continental monsoon rainfall, and estimate the threshold values

within the restrictions of the conceptual model. I demonstrate the applicability of this concept to abrupt

and persistent monsoon shifts observed in paleoclimatic records.

To understand monsoon failure occurring on shorter timescales and within the domain of existence,

I develop a minimal theory of intraseasonal monsoon dynamics. Supported by observations and by

results from a comprehensive global climate model, the core assumption of this theory is that the positive

moisture–advection feedback and its interaction with the smaller–scale eddy field render the circulation

inherently unstable. I apply this theory to an ensemble of millennial climate simulations and show that

both multi–decadal variability and projected future trends in Indian summer monsoon (ISM) rainfall

can be reproduced with a simple stochastic model of the inherent instability, modulated by ambient

climate conditions only during the onset period. A projected increase in ISM failure in response to a

global warming scenario can thus be readily explained by a shift in central Pacific mean spring–time

climate that persistently alters the initial conditions for internal ISM dynamics. I thereby propose a

novel perspective on monsoon variability as the result of internal instabilities modulated by pre-seasonal

ambient climate conditions.

In summary, the results in this thesis offer a simplified framework for the investigation of both long–

term (permanent) and short–term (seasonal) monsoon failure, including the basic physical mechanisms

that lead to a non–linear response of the monsoon system to external changes.

Zusammenfassung

In dieser Arbeit stelle ich zunachst Zukunftsprojektionen mit dem Erdsystemmodell mittlerer Kom-

plexitat CLIMBER-3α vor, die eine weltweite Zunahme des Monsunniederschlags uber die nachsten

Jahrhunderte zeigen, welche annahernd proportional zum Anstieg der globalen Mitteltemperatur verlauft.

Zwar legen auch viele andere Studien einen solchen Anstieg nahe, jedoch verwendet CLIMBER-3α

eine vereinfachte, statistisch–dynamische Atmospharenkomponente und gibt wahrscheinlich nicht alle

Prozesse, die fur die Reaktion von Monsunzirkulationen auf rasanten Klimawandel relevant sind, hinre-

ichend wieder.

Um die physikalischen Mechanismen zu identifizieren, die fur die großskalige Monsundynamik und

ihre Reaktion auf außere Veranderungen von zentraler Bedeutung sind, unterziehe ich die Warme-

und Feuchtebilanzen der wichtigsten Monsunsysteme anhand von Reanalysedaten und theoretischen

Uberlegungen einer Skalenanalyse. Es zeigt sich, dass der Monsun wahrend der Regenzeit in erster

Linie von einem selbstverstarkenden Ruckkopplungsmechanismus angetrieben wird, bei dem die Ad-

vektion und Freisetzung latenter Warme den atmospharischen Temperaturunterschied zwischen Land

und Ozean aufrechterhalt. Ich stelle diese Feuchte–Advektions–Ruckkopplung in einem minimalistischen

konzeptionellen Modell dar und zeige, dass sich aus ihr ein nichtlineares Verhalten des Monsunsystems

gegenuber Storungen der Energiebilanz ergibt: Wenn etwa die atmospharische Strahlungsbilanz uber

dem Kontinent oder die Luftfeuchtigkeit uber der benachbarten Ozeanregion einen kritischen Wert un-

terschreiten, kann sich keine konventionelle Monsunzirkulation entwickeln. Durch diesen kritischen Wert

wird entsprechend der Parameterbereich definiert, in dem Monsunniederschlag uber Land moglich ist. Ich

nehme eine Abschatzung des kritischen Wertes fur verschiedene Monsunregionen vor und zeige, dass sich

das Konzept auf abrupte Monsunveranderungen anwenden lasst, wie sie anhand von in palaoklimatischen

Rekonstruktionen dokumentiert sind.

Des Weiteren entwerfe ich eine minimalistische Theorie intrasaisonaler Monsundynamik mit dem Ziel,

Monsunausfalle zu verstehen, die auf kurzeren Zeitskalen und innerhalb des durch den kritischen Wert

definierten Existenzbereiches auftreten. Die zugrundeliegende Hypothese ist, dass die positive Feuchte–

Advektions–Ruckkopplung und ihre Wechselwirkung mit turbulenten Storungen auf synoptischer und

kleinerer Skala zu einer inharenten Instabilitat fuhren. Ich entwickle ein einfaches stochastisches Modell,

in dem die Monsundynamik von dieser Instabilitat bestimmt und lediglich zu Beginn der Monsunsaison

von außeren klimatischen Einflussen moduliert wird. Dieses Modell vergleiche ich mit einem Ensemble

von Langzeitsimulationen eines realistischen Klimamodells und zeige, dass sowohl die multidekadische

Variabilitat als auch die fur die Zukunft projizierten Entwicklungen reproduziert werden konnen. Eine

Haufung von Monsunausfallen unter einem Klimawandelszenario kann ich auf diese Weise mit einer

6

Veranderung des zentralpazifischen Fruhjahrsklimas erklaren. Die vorgestellte Theorie eroffnet somit eine

neue Sichtweise auf die Variabilitat des Monsunniederschlags als das Ergebnis einer instabilen internen

Dynamik, die zu Beginn der Saison durch das umgebende Klima moduliert wird.

Die in dieser Arbeit vorgestellten Ergebnisse bieten einen vereinfachten theoretischen Rahmen fur

die Untersuchung von Monsunausfallen auf langen (palaoklimatischen) wie auch kurzen (saisonalen)

Zeitskalen, sowie der grundlegenden physikalischen Prozesse, die zu einer nichtlinearen Reaktion von

Monsunsystemen auf außere Storungen fuhren konnen.

Contents

1 Introduction 9

1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

1.2 Scope and contents of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.3 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 Original manuscripts 17

2.1 Climate change under a scenario near 1.5◦C of global warming . . . . . . . . . . . . . . . 17

2.2 Basic mechanism for abrupt monsoon transitions . . . . . . . . . . . . . . . . . . . . . . . 31

2.3 A critical humidity threshold for monsoon failure . . . . . . . . . . . . . . . . . . . . . . . 49

2.4 More frequent future monsoon failure due to inherent instability . . . . . . . . . . . . . . 81

3 Discussion and Conclusions 125

References 131

7

1 Introduction

1.1 Motivation

Monsoon systems are among the most prominent large–scale phenomena in the Earth’s atmosphere.

They govern seasonal and year–to–year climate variability in many tropical and subtropical regions,

and the associated rainfall is the dominant natural variable that human economies in those regions are

built around. Changes in monsoon rainfall appear to have affected human societies throughout history,

as illustrated for example by the fate of ancient civilizations in the Indus Valley (Rashid et al., 2011)

or the rise and fall of dynasties and kingdoms in China (Zhang et al., 2008). Also today, agricultural

productivity in South Asia, East Asia, and Africa is closely tied to the magnitude and timing of monsoon

rainfall (Parthasarathy et al., 1988; Kumar et al., 2004; Tao et al., 2004; Haile, 2005; Gregory et al., 2005;

Auffhammer et al., 2006). Present–day monsoon rainfall, as observed over the past century, exhibits

significant intraseasonal and interannual variability (e.g. Webster, 1987b), including catastrophic floods

and droughts that pose risks to large parts of the population in those regions the affected regions (e.g.

Sikka, 2003; Webster et al., 2011).

While decadal–scale average monsoon rainfall has been relatively stable during the past century of

direct observations, rising trends have been observed in the annual number of extreme rain events in

India (Goswami et al., 2006a). The future evolution of the Indian summer monsoon, and other monsoon

systems, under a combination of anthropogenic forcing factors is unclear, according to an intercomparison

of comprehensive climate models (Meehl et al., 2007). Recent projections indicate that the response to

increased greenhouse gas (GHG) concentrations may differ in sign among major monsoon regions, and

reveal large uncertainties about the magnitude of the response (Kripalani et al., 2007a,b; Cherchi et al.,

2010). The effect of increased aerosol abundance is significant and may be counteracting that of GHGs

(Ramanathan et al., 2005; Lau & Kim, 2006), while human-induced vegetation changes feed back on

precipitation (Ganopolski et al., 1998; Claussen, 2009). Observations and modelling studies suggest a

recent regime shift in Asian monsoon convection (Turner & Hannachi, 2010) and its relation to northern

hemisphere thermal gradients (Li et al., 2009; Sun et al., 2010).

At the same time, paleoclimatic records show evidence of abrupt and strong monsoon shifts in India,

the Bay of Bengal, and East Asia, during the last two glacial cycles (Burns et al., 2003; Wang et al.,

2005a, 2008) and the Holocene (Gupta et al., 2003; Hong et al., 2003; Wang et al., 2005b; Rashid et al.,

2011). Some of these abrupt changes have been linked to climatic events in the North Atlantic for the

last glacial period (Overpeck et al., 1996; Burns et al., 2003) as well as for the Holocene (Gupta et al.,

2003; Wang et al., 2005b). A physical mechanism for this teleconnection has been suggested (Goswami

et al., 2006b), but relevant climatic signals of the North Atlantic events in Asia (such as temperature

10 1 INTRODUCTION

and moisture anomalies) are very small (Zhang & Delworth, 2005). East Asian monsoon shifts during

the last two glacial cycles have been roughly in phase with hemispheric insolation changes (Wang et al.,

2008), but the latter are far too slow to explain the rapidity of the transitions between different periods

of relatively stable monsoon strength. These observations indicate that internal feedbacks in monsoon

dynamics may have amplified the weak external forcing.

Both the unclear cause of paleoclimatic monsoon shifts and the uncertain future evolution of monsoon

rainfall in a changing climate call for an improved understanding of possible non–linearities in large–scale

monsoon dynamics. This thesis aims at contributing to such an improvement by identifying the primary

driving mechanisms of continental monsoon rainfall and examining their stability properties and their

implications for the response of the monsoon circulation to external forcing. It does not attempt to

account for all aspects of monsoon dynamics in general, but focuses on the possibility of large–scale

monsoon failure and the associated processes.

1.2 Scope and contents of the thesis

Monsoons are among the most complex components of the climate system, and they interact with various

other components on multiple timescales (Webster et al., 1998; Wang, 2005). Both spatial patterns and

temporal evolution of monsoon rainfall are influenced by a number of physical processes, such as the

El Nino–Southern Oscillation (ENSO) phenomenon (e.g. Krishnamurthy & Goswami, 2000; Goswami &

Xavier, 2005), regional (Clark et al., 2000; Kucharski et al., 2006; Yang et al., 2007) and remote sea

surface temperatures (Goswami et al., 2006b), atmospheric aerosol concentrations (Ramanathan et al.,

2005), or Eurasian snow cover (Hahn & Shukla, 1976; Dash et al., 2005), as well as by characteristics

of vegetation (Meehl, 1994; Claussen, 1997; Robock et al., 2003) and topography (Liu & Yin, 2002).

Direct observational records of monsoon rainfall, available for the last century, exhibit a high variability

on different timescales. For instance, in India (where data availability is highest among the monsoon

regions), seasonal mean rainfall amounts vary from year to year and, though correlated to some degree

with ENSO, are difficult to predict in advance (Goswami et al., 2006). Within the monsoon season,

rainfall does not occur uniformly, but is concentrated in periods of roughly a few weeks length, called

active spells, which are intercepted by drier periods, called break spells (Krishnamurthy & Shukla, 2000;

Rajeevan et al., 2010). These active and break spells vary in number and length from year to year, and

within them, daily rainfall amounts also vary greatly. In addition, there are large inhomogeneities in the

spatial distribution of rainfall over the monsoon regions.

Despite this complexity, however, there is one fundamental driving force behind large–scale conti-

nental monsoon rainfall: Namely, the atmospheric temperature contrast between land and ocean (e.g.

1.2 Scope and contents of the thesis 11

2 4 6 8 10 12−4

−2

0

2

4

6

month

Δ T

(K

)

surfacecolumn

Figure 1.1: Climatological temperature difference ∆T between land and ocean in the

Indian region, derived from NCEP/NCAR reanalysis data (Kistler et al., 2001). At

the surface (solid line), ∆T is largest is spring and then decreases due to cooling of

the land surface by precipitation. Averaged over the atmospheric column (dashed

line), ∆T is maintained throughout the rainy season.

Webster, 1987a). This temperature contrast develops in spring when the continent heats up faster

than the ocean, owing to the large differences in heat capacity; and it is maintained throughout the

rainy season, even when the rains have started to cool the land surface and sensible heating has ceased

(Fig. 1.1). It shapes the anomalous pressure system that draws strong, mainly ageostrophic winds

towards the continent in the lower troposphere, carrying the moisture that is then released in convection.

While this is far from a complete description of monsoon dynamics, it is a central and necessary condition

for monsoon rainfall to develop. Furthermore, the magnitude of the land–sea atmospheric temperature

difference is generally correlated with seasonal rainfall amounts: The stronger the difference during a

given monsoon season, the more rainfall can be expected, irrespective of where and when exactly that

rain falls. That also means that long–term changes in the temperature contrast can be expected to affect

long–term mean monsoon rainfall.

The seasonal development of the land–sea atmospheric temperature contrast, as a driving force of

monsoon rainfall, is captured by linear empirical models (e.g. Srinivasan, 2001) and also by coarse–

resolution climate models that otherwise may not have the spatial resolution and the degree of realism in

atmospheric physics necessary to capture all aspects of monsoon dynamics. Such models can therefore

12 1 INTRODUCTION

be used to make meaningful projections of the direct effect of temperature changes on large–scale

monsoon characteristics. In the first article of this thesis (Schewe et al., 2011a), I have applied the

Earth system model of intermediate complexity, CLIMBER-3α (Montoya et al., 2005), to project

the climatic consequences of the Representative Concentration Pathways, a set of recently developed

GHG concentration scenarios for use in the forthcoming assessment of the Intergovernmental Panel

on Climate Change (IPCC). I find that average monsoon rainfall in South Asia, East Asia and Africa

increases approximately linearly with the regional land–ocean surface temperature contrast, which

in turn is a direct consequence of global surface warming (due to the increase in GHG abundance.)

Depending on the region and the scenario, the projected rainfall increases are substantial; e.g. between

about 25% (South Asia) and 50% (East Asia) until the end of the 21st century under the highest scenario.

As discussed above, paleoclimatic records reveal large and abrupt shifts in monsoon intensity that

obviously cannot be explained as a linear response to external forcing. They therefore require different

concepts than a perturbation analysis around the present–day state. In the second and third articles of

this thesis, I take the approach of a non–linear empirical model to obtain a first–order understanding

of the processes that may have led to such abrupt events. The second article (Levermann et al., 2009)

sets up a minimal conceptual model of a monsoon circulation, comprising only the conservation of heat

and moisture and knowingly neglecting many other important physical processes, in order to distill the

fundamental non–linearity of monsoon dynamics. The model is based on a scaling analysis of the heat

and moisture budgets of major monsoon systems around the world, using present–day reanalysis data.

I find that sensible heating is important in establishing the atmospheric land–sea temperature contrast

prior to the rainy season, but becomes small after the onset of heavy rainfall. During the rainy season,

the temperature gradient is maintained by the release of latent heat over the continent. Thus, the

advection of moist air from the ocean and subsequent condensation of that moisture sustains the driving

force for the monsoon winds, and thereby constitutes a self–amplifying feedback (illustration in Fig. 1.2).

I show in the conceptual model that this moisture–advection feedback implies a threshold behaviour

with respect to quantities that affect the energy budget: E.g., if net radiative flux to the atmospheric

column falls below a critical value, no conventional monsoon circulation can develop. If the system was

close to the threshold, a small variation in external parameters could therefore lead to a transition from

a “normal” monsoon regime into a regime without continental monsoon rainfall.

However, considering the present–day state of the Earth’s monsoon systems, huge changes in net

radiation would be necessary to reach the threshold. In the third article (Schewe et al., 2011b), I

show that the minimal conceptual model also yields a threshold behaviour with respect to atmospheric

humidity over the ocean adjacent to the monsoon region. This quantity is more volatile than net

1.2 Scope and contents of the thesis 13

Figure 1.2: Geometry of the minimal conceptual monsoon model used in the second

and third articles of the thesis, with wind W, precipitation P, and net radiative flux R.

The tripartite loop illustrates the fundamental moisture–advection feedback; arrows

indicate the amplification of one physical process by another.

radiation, and the thresholds are relatively closer to the present–day situation. Based on reanalysis

data, I estimate the threshold values for four major monsoon regions. I apply this concept to a proxy

record of East Asian monsoon rainfall that exhibits a series of abrupt monsoon transitions during the

penultimate glacial period. Assuming that average humidity over the ocean was altered by orbital–scale

changes in hemispheric solar insolation, I show that the conceptual model can qualitatively explain

these transitions. As evaporation from the ocean surface can also be affected by a number of other

processes (e.g. wind speed, oceanic upwelling) on different timescales, the model could serve to improve

the understanding of other past monsoon events as well.

The basic dynamics captured in the conceptual model thus define a domain of existence for continental

monsoon rainfall; in the sense that a conventional monsoon can only develop within this domain, e.g.

above the humidity threshold. This however does not mean that within the domain of existence, monsoon

rainfall will be at full strength at all times. As I show in the fourth article of this thesis (submitted and

under review at Nature), even under present–day conditions, seasonal–mean Indian summer monsoon

(ISM) rainfall can fall short of its long–term average by 70% and more in individual years, according to

a comprehensive atmosphere–ocean general circulation model (AOGCM). Such dry monsoon years have

14 1 INTRODUCTION

not been observed in the last century, but an ensemble simulation that covers over 6,000 model years

shows a characteristic frequency distribution of seasonal–mean rainfall that continuously extends down

to these extremely low values, though with low frequency of occurrence.

Understanding such temporary monsoon failure again requires looking at the fundamental driving

mechanism of monsoon rainfall. Using AOGCM results and observational data, I show that the moisture–

advection feedback generally acts on a short sub–seasonal timescale (on the order of days rather than

months) and is frequently interrupted because of continuous perturbation by stochastic fluctuations from

the synoptic–scale eddy field. Moreover, during such interruptions, the large-scale monsoon circulation

does not simply slow down or come to a hold, but even tends to reverse, in the sense that convection is

replaced by subsidence of upper–tropospheric air over large parts of the subcontinent and the adjacent

ocean. Since this air is much drier than lower–tropospheric air, the subsidence effectively dries out

the monsoon winds and further inhibits a recommencement of the moisture–advection feedback. Thus,

another self–amplifying feedback is constituted that counteracts the moisture–advection feedback.

In the article, I develop a minimal theory of intraseasonal monsoon dynamics, based on the assumption

that the monsoon season is governed by a permanent interplay of those two counteracting feedbacks: Each

feedback itself tends to persist, and stochastic fluctuations tend to perturb the currently active feedback

and induce a flip into the other one. I frame this assumption in a simple, statistically predictive model of

seasonal–mean monsoon rainfall and show that it can reproduce the characteristic frequency distribution

found in the AOGCM, including the very dry years, or failures.

An important aspect of this simple model is that it is linked to external forcing only during the onset

period; for the rest of the monsoon season, only the idealized internal dynamics are at work and produce

rainy and dry periods that then add up to a seasonal average. I show that the model can reproduce a

large portion of multidecadal monsoon variability when forced by central–Pacific mean sea level pressure

(MSLP) anomalies in May, i.e. at the onset time. The central Pacific is known to have a distinct influence

on Indian monsoon climate, as is evident in the correlation between ISM rainfall and ENSO. Anomalously

low spring-time MSLP in the central Pacific is assumed to induce atmospheric conditions that favor more

subsidence over the Indian region and thus lead to more deficient monsoon onsets.

I then apply the simple model to global warming simulations using the same AOGCM, where ISM

failure is projected to become much more frequent until the end of the 22nd century. At the same time,

a shift towards lower spring-time MSLP in the central Pacific is projected. Again forcing the simple

model with central–Pacific MSLP anomalies in May, it successfully reproduces the projected trend in

average monsoon rainfall. The minimal theory presented in this work thereby offers a novel perspective

on monsoon variability as the result of internal instabilities modulated by pre-seasonal ambient climate

conditions.

1.3 Overview 15

1.3 Overview

This thesis is organized around four scientific articles which are either published or under review.

Each article provides their own introductory and concluding remarks, and references; some also carry

supplementary material. Here, a brief overview is given of the titles, contents, and author contributions

of the individual articles. The original manuscripts are included in the following section.

Article 1:

Climate change under a scenario near 1.5◦C of global warming: monsoon

intensification, ocean warming and steric sea level rise

Jacob Schewe, Anders Levermann, & Malte Meinshausen

The first article, published in Earth System Dynamics, explores the climatic consequences of the latest

set of greenhouse gas concentration scenarios for the coming centuries. The response of the most impor-

tant large–scale oceanic, atmospheric, and coupled processes is investigated in a coupled climate model of

intermediate complexity. Among other results, monsoon rainfall is projected to increase approximately

linearly with the regional land–sea temperature contrast due to global warming.

Jacob Schewe performed the climate model simulations, analyzed the results, and wrote the paper.

Malte Meinshausen provided the AOGCM emulations. All three authors participated in the interpreta-

tion of the results and the improvement of the manuscript.

Article 2:

Basic mechanism for abrupt monsoon transitions

Anders Levermann, Jacob Schewe, Vladimir Petoukhov, & Hermann Held

The second article, published in Proceedings of the National Academy of Sciences of the USA (PNAS),

presents a scaling analysis of the heat and moisture budgets of the Earth’s major monsoon systems, and

develops a minimal conceptual monsoon model capturing the essential moisture–advection feedback. It

shows that this feedback yields a threshold behaviour with respect to changes in net radiation, and

estimates the thresholds from reanalysis data.

Jacob Schewe analyzed the data; Anders Levermann, Vladimir Petoukhov and Jacob Schewe devised

the conceptual model; Hermann Held performed statistical tests of the robustness of the results; Anders

Levermann performed the computations and wrote the paper. All four authors participated in the

interpretation of the results and the improvement of the manuscript. Anders Levermann initiated the

research.

16 1 INTRODUCTION

Supporting information is included following the main manuscript.

Article 3:

A critical humidity threshold for monsoon failure

Jacob Schewe, Anders Levermann, & Hai Cheng

The third article, published in Climate of the Past Discussions, shows that the minimal conceptual

monsoon model implies a threshold behaviour with respect to specific humidity over the ocean, a quantity

which is more volatile than net radiation and exhibits threshold values closer to modern climate. The

threshold values are estimated from reanalysis data for four major monsoon regions, and the model is

applied to a paleoclimatic reconstruction of East Asian summer monsoon rainfall, yielding a series of

abrupt transitions in response to gradual insolation changes which is similar to those observed in the

proxy record.

Jacob Schewe performed the computations, analyzed data, and wrote the paper. Hai Cheng provided

proxy data. Jacob Schewe and Anders Levermann interpreted the results and improved the manuscript.

Anders Levermann initiated the research.

Article 4:

More frequent future monsoon failure due to inherent instability

Jacob Schewe & Anders Levermann

The fourth article (submitted and currently under review) projects Indian summer monsoon failure

to become considerably more frequent under a global warming scenario, using a comprehensive coupled

climate model. It presents a minimal theory of intraseasonal monsoon dynamics, based on an inherent

instability that is modulated by ambient climate merely during the onset period. Forced only by global

mean temperature and central–Pacific sea level pressure anomalies in May, this simple model reproduces

future trends as well as past multidecadal variability of monsoon rainfall as found in the comprehensive

climate model. The study proposes a novel perspective on monsoon variability as the result of internal

instabilities modulated by pre-seasonal ambient climate conditions.

Jacob Schewe analyzed data and climate model results and wrote the paper. Jacob Schewe and Anders

Levermann developed the minimal theory, interpreted the results, and improved the manuscript.

Supplementary Material, and the Matlab R© code of the simple “day-to-day model”, are included

following the main manuscript.

2 Original manuscripts

2.1 Climate change under a scenario near 1.5◦C of global warming

Earth Syst. Dynam., 2, 25–35, 2011www.earth-syst-dynam.net/2/25/2011/doi:10.5194/esd-2-25-2011© Author(s) 2011. CC Attribution 3.0 License.

Earth SystemDynamics

Climate change under a scenario near 1.5◦C of global warming:monsoon intensification, ocean warming and steric sea level rise

J. Schewe1,2, A. Levermann1,2, and M. Meinshausen1

1Earth System Analysis, Potsdam Institute for Climate Impact Research, Potsdam, Germany2Physics Institute, Potsdam University, Potsdam, Germany

Received: 30 September 2010 – Published in Earth Syst. Dynam. Discuss.: 14 October 2010Revised: 8 February 2011 – Accepted: 4 March 2011 – Published: 8 March 2011

Abstract. We present climatic consequences of the Repre-sentative Concentration Pathways (RCPs) using the coupledclimate model CLIMBER-3α, which contains a statistical-dynamical atmosphere and a three-dimensional ocean model.We compare those with emulations of 19 state-of-the-artatmosphere-ocean general circulation models (AOGCM) us-ing MAGICC6. The RCPs are designed as standard sce-narios for the forthcoming IPCC Fifth Assessment Reportto span the full range of future greenhouse gas (GHG) con-centrations pathways currently discussed. The lowest of theRCP scenarios, RCP3-PD, is projected in CLIMBER-3α toimply a maximal warming by the middle of the 21st cen-tury slightly above 1.5◦C and a slow decline of temperaturesthereafter, approaching today’s level by 2500. We identifytwo mechanisms that slow down global cooling after GHGconcentrations peak: The known inertia induced by mixing-related oceanic heat uptake; and a change in oceanic convec-tion that enhances ocean heat loss in high latitudes, reducingthe surface cooling rate by almost 50%. Steric sea level riseunder the RCP3-PD scenario continues for 200 years afterthe peak in surface air temperatures, stabilizing around 2250at 30 cm. This contrasts with around 1.3 m of steric sea levelrise by 2250, and 2 m by 2500, under the highest scenario,RCP8.5. Maximum oceanic warming at intermediate depth(300–800 m) is found to exceed that of the sea surface bythe second half of the 21st century under RCP3-PD. Thisintermediate-depth warming persists for centuries even af-ter surface temperatures have returned to present-day values,with potential consequences for marine ecosystems, oceanicmethane hydrates, and ice-shelf stability. Due to an enhancedland-ocean temperature contrast, all scenarios yield an inten-sification of monsoon rainfall under global warming.

Correspondence to:J. Schewe([email protected])

1 Introduction

In December 2010, the international community agreed,under the United Nations Framework Convention on Cli-mate Change, to limit global warming to below 2◦C(Cancun Agreements, see http://unfccc.int/files/meetings/cop 16/application/pdf/cop16lca.pdf). At the same time, itwas agreed that a review, to be concluded by 2015, shouldlook into a potential tightening of this target to 1.5◦C – inpart because climate change impacts associated with 2◦Care considered to exceed tolerable limits for some regions,e.g. Small Island States. So far, research into climate systemdynamics under strong mitigation scenarios that keep warm-ing below 2◦C or even 1.5◦C has been sparse. IndividualAOGCMs were run for scenarios stabilizing at 2◦C (May,2008) or below (Washington et al., 2009), or for idealizedCO2 rampdown experiments (Wu et al., 2010).

Here we investigate climate projections for the full rangeof Representative Concentration Pathways (RCPs;Mosset al., 2010) but focus in particular on the lowest scenarioRCP3-PD, which reflects a scenario that will peak globalmean temperatures slightly above, but close to, 1.5◦C abovepre-industrial levels in our model. The RCPs were recentlydeveloped in order to complement, and in part replace, theSpecial Report on Emissions Scenarios (SRES;Nakicen-ovic and Swart, 2000) scenarios, and will be used in the Cli-mate Model Intercomparison Project’s Phase 5 (CMIP5) thatis to be assessed in the forthcoming Intergovernmental Panelon Climate Change (IPCC) Fifth Assessment Report (AR5).The RCP3-PD scenario is characterized by a peak of atmo-spheric greenhouse gas (GHG) concentrations in 2040 anda subsequent decline in GHG abundance. After 2070, CO2emissions turn negative and remain at around−1 Gt CO2-eq yr−1 after 2100 (Meinshausen et al., 2011). Concentra-tions in the medium-low RCP4.5 and the medium-high RCP6

Published by Copernicus Publications on behalf of the European Geosciences Union.

2.1 Climate change under a scenario near 1.5◦C of global warming 19

26 J. Schewe et al.: Climate near 1.5◦C warming

stabilize by 2150, while concentrations in the high RCP8.5continue to rise until 2250.

In Sect.2, we describe the models and their experimen-tal setup for this study. Simulation results are presented inSect.3, in particular for global mean temperature (Sect.3.1),and changes in large scale climate components like oceanicmeridional overturning circulation (Sect.3.2), monsoon(Sect.3.3), global sea level (Sect.3.4), and deep ocean tem-perature (Sect.3.5). In Sect.4 we provide the physicalmechanisms responsible for an asymmetrically slower cool-ing than warming under RCP3-PD. Section5 concludes.

2 Models and experiments

Our primary model for investigating key large-scale aspectsof climate change under the RCP scenarios is the Earth sys-tem model of intermediate complexity CLIMBER-3α (Mon-toya et al., 2005). CLIMBER-3α combines a statistical-dynamical atmosphere model (Petoukhov et al., 2000) witha three-dimensional ocean general circulation model basedon the GFDL MOM-3 code (Pacanowski and Griffies, 1999)and a dynamic and thermodynamic sea-ice model (Fichefetand Maqueda, 1997). In this study, CLIMBER-3α is usedwithout a carbon cycle. The atmosphere model has a coarsehorizontal resolution of 22.5◦ in longitude and 7.5◦ in lat-itude, and employs parameterized vertical temperature andhumidity profiles. Oceanic wind stress anomalies are com-puted with respect to the control simulation and added to theclimatology of Trenberth et al.(1989). The oceanic hori-zontal resolution is 3.75◦ × 3.75◦ with 24 variably spacedvertical levels. The model’s sensitivity to vertical diffusiv-ity (Mignot et al., 2006) and wind stress forcing (Scheweand Levermann, 2010) has been investigated as well as themodel’s behaviour under glacial boundary conditions (Mon-toya and Levermann, 2008) and global warming (Levermannet al., 2007). When compared to AOGCMs of the third Cou-pled Model Intercomparison Project (CMIP3) and previousgenerations, the model shows qualitatively and quantitativelysimilar results with respect to large-scale quantities (Gregoryet al., 2005; Stouffer et al., 2006b). The model version usedhere features a low background value of oceanic vertical dif-fusivity (0.3× 10−4 m2 s−1) and an improved representationof the Indonesian throughflow as compared to the version de-scribed byMontoya et al.(2005).

We complement our CLIMBER-3α projections of globalmean temperature with emulations of 19 AOGCMs used inthe IPCC Fourth Assessment Report (AR4). These emu-lations were performed with MAGICC6, a reduced com-plexity model with an upwelling-diffusion ocean which hasbeen used in the past three IPCC assessment reports (Wigleyand Raper, 2001). MAGICC6 was shown to be able toclosely emulate the global and hemispheric mean tempera-ture evolution of AOGCMs (Meinshausen et al., 2008). OurAOGCM emulations use RCPs harmonized emission inputs

with default efficacies for the individual forcing agents, iden-tical to the model’s setup for creating the default RCP GHGconcentration recommendations for CMIP5 (Meinshausenet al., 2011). The only exception is that MAGICC6’s climatemodel is calibrated and run for the range of 19 individualAOGCMs, rather than a single median set of climate moduleparameters.

Our CLIMBER-3α experiments focus on the four newRCPs, namely RCP3-PD (van Vuuren et al., 2007), RCP4.5(Clarke et al., 2007; Smith and Wigley, 2006; Wise et al.,2009), RCP6 (Fujino et al., 2006), and RCP8.5 (Riahi et al.,2007). We use the historical, 21st century and long-term(until 2500) RCP forcing trajectories as provided onhttp://www.pik-potsdam.de/∼mmalte/rcps/and described inMein-shausen et al.(2011). These forcings arose from the pro-cess of harmonizing RCP emissions, and producing a sin-gle default set of GHG concentrations, which are the basisfor the CMIP5 intercomparison runs that extend from pre-industrial times to 2300 (CMIP5,http://cmip-pcmdi.llnl.gov/cmip5/forcing.html). The extension beyond 2300 follows thesame guiding principle as the extension up to 2300, i.e. a con-tinuation of constant emissions for the RCP3-PD scenario(and correspondingly dropping forcing levels) and a stabi-lization of GHG concentrations and forcing levels for the up-per three RCPs, RCP4.5, RCP6 and RCP8.5.

For being used in CLIMBER-3α, we group our forcings ona forcing-equivalence basis, i.e. we aggregate longwave ab-sorbers into a CO2-equivalence concentration (Fig.1a and d).The radiative forcing of agents that scatter or absorb short-wave radiation is aggregated and assumed to modulate theincoming solar irradiance, taking into account geometry andalbedo (Fig.1b and e). CLIMBER-3α’s climate sensitivityis about 3.4◦C, which is higher than the average climatesensitivity of the transient AOGCM emulations of 2.9◦C(Meinshausen et al., 2008, Table 4), very close to the av-erage of the slab–ocean GCMs of 3.26◦C and still close tothe IPCC AR4 best estimate of 3◦C (Meehl et al., 2007a,Box 10.2). The transient climate response is about 1.9◦C forCLIMBER-3α, compared to about 1.8◦C for the average ofIPCC AR4 AOGCMs (Meehl et al., 2007b, Table 8.2).

3 Results

3.1 Global mean temperature

Global mean surface air temperatures, normalized to the pe-riod 1980–1999, are shown in Fig.1c and f relative to pre-industrial (1860–1890) using the median observed tempera-ture increase of 0.52◦C (Brohan et al., 2006). The warmingprojected by CLIMBER-3α lies well within the emulationof the AOGCMs (Fig.1c and f). For the highest scenario,RCP8.5, the simulation yields a temperature increase of upto 8.5◦C, while the lowest scenario, RCP3-PD, reaches upto 1.6◦C of global warming compared to pre-industrial and

Earth Syst. Dynam., 2, 25–35, 2011 www.earth-syst-dynam.net/2/25/2011/

20 2 ORIGINAL MANUSCRIPTS

J. Schewe et al.: Climate near 1.5◦C warming 27


ppm

va

0

1000

2000

3000W

/m2

b

1350135513601365

RCP8.5

RCP6

RCP4.5

RCP3−PD

° C

year

RCP extensionc

2000 2200 24000

2

4

6

8

10

12

ppm

v

d

500

1000

1500

W/m

2

e 1350

1355

1360

1365

° C

year

f

1950 2000 2050 2100

0

2

4

6

Fig. 1. Forcing and global mean temperature response of the CLIMBER-3α climate model under the RCP3-PD (blue), RCP4.5 (yellow),RCP6 (grey) and RCP8.5 (red) scenarios and their extensions until 2500. The grey vertical band marks the RCP period 2005 to 2100. (a)CO2-equivalence concentration (in ppmv) of longwave absorbers (Kyoto and Montreal protocol greenhouse gases as well as troposphericozone). (b) Incoming solar irradiance (W/m2), modified by the radiative forcing of agents active in the shortwave range (mainly volcanicand anthropogenic aerosols) and changes in surface albedo due to land-use change. (c) Global surface air temperature (SAT) differencein ◦C compared to pre–industrial (Brohan et al., 2006), for the CLIMBER-3α simulations (solid lines) and 19 AOGCM emulations usingMAGICC6 (the dashed line denotes the median, and dark and light shading denotes the 50% and 80% range, respectively). (d) to (f): As (a)to (c), but enlarged for the period 1950-2100.

Fig. 1. Forcing and global mean temperature response of the CLIMBER-3α climate model under the RCP3-PD (blue), RCP4.5 (yellow),RCP6 (grey) and RCP8.5 (red) scenarios and their extensions until 2500. The grey vertical band marks the RCP period 2005 to 2100.(a) CO2-equivalence concentration (in ppmv) of longwave absorbers (Kyoto and Montreal protocol greenhouse gases as well as troposphericozone).(b) Incoming solar irradiance (W m−2), modified by the radiative forcing of agents active in the shortwave range (mainly volcanicand anthropogenic aerosols) and changes in surface albedo due to land-use change.(c) Global surface air temperature (SAT) differencein ◦C compared to pre-industrial (Brohan et al., 2006), for the CLIMBER-3α simulations (solid lines) and 19 AOGCM emulations usingMAGICC6 (the dashed line denotes the median, and dark and light shading denotes the 50% and 80% range, respectively).(d) to (f): As (a)to (c), but enlarged for the period 1950–2100.

then drops at an average rate of about−0.16◦C per century.This is about ten times slower than the currently observedtemperature rise of 0.16 to 0.18◦C per decade (Trenberthet al., 2007, section 3.4). Although the reduction in GHGconcentrations in the RCP3-PD is generally slower than theincrease before the peak, this explains only part of the warm-ing/cooling asymmetry: The average cooling rate during thefirst 100 years after the peak is 12% of the warming rate inthe 100 years before the peak; over the same period, the GHGreduction rate is 35% of the increase rate prior to the peak.The mechanisms responsible for this asymmetry will be dis-cussed in Sect.4.

3.2 Spatial warming pattern and oceanic overturning

The spatial distribution of temperature change in 2100 re-flects the pattern of polar amplification (Winton, 2006),i.e. above-average surface warming in high latitudes (Fig.2).In the low RCP3-PD scenario (Fig.2a), warming in thenorthern North Atlantic region is offset by the cooling effect

of a 20% reduction of the Atlantic meridional overturningcirculation (AMOC; Fig.3a) and the associated reductionin oceanic convection and heat release (compare Sect.4).As the AMOC recovers over the course of the 22nd and23rd century, this offsetting effect will disappear. In theRCP8.5 scenario (Fig.2b), the AMOC reduction is relativelysmaller compared to the warming, and has no large offset-ting effect. The recovery of the AMOC beyond 2200 is facil-itated by the retreat of sea ice cover in the North Atlantic(Levermann et al., 2007), which in the case of RCP3-PDeven leaves the AMOC stronger in the long-term than un-der pre-industrial conditions. The behaviour of the AMOCunder global warming in CLIMBER-3α is a robust featureof most CMIP3 AOGCMs (Gregory et al., 2005), and themechanisms at play are in qualitative agreement across themodels (Levermann et al., 2007). Quantitatively, AOGCMsdiffer significantly in their response. With respect to thepre-industrial overturning strength, CLIMBER-3α is compa-rable to the IPCC AR4 model average and consistent with

www.earth-syst-dynam.net/2/25/2011/ Earth Syst. Dynam., 2, 25–35, 2011




80ºN

40ºN

40ºS

80ºS

0º

80ºN

40ºN

40ºS

80ºS

0º

100ºE100ºW 0º

10

8

6

4

2

0

3

2

1

0

a

b

Fig. 2. (a) Surface air temperature anomaly for the year 2100, in◦C, for RCP3-PD. Average warming south of 60◦S is 1.60 timeshigher than the global mean. Average warming north of 60◦N isonly 0.83 times the global mean (1.4◦C), because the cooling ef-fect of a reduction in the Atlantic Meridional Overturning Circula-tion (AMOC) counteracts polar amplification. (b) Same for RCP8.5(with a global mean of 4.8◦C). The polar amplification factors are1.48 in the south and 1.53 in the north.

Sv

aAMOC

10

12

14

16

18

year

Sv

SPG

b

2000 2200 2400 10

15

20

25

30

Fig. 3. (a) Maximum AMOC strength of the Atlantic Merid-ional Overturning Circulation (AMOC) in Sv (106m3s−1), forRCP3-PD (blue), RCP4.5 (yellow), RCP6 (grey), and RCP8.5 (red).(b) North Atlantic subpolar gyre strength, in Sv, computed frommeridional velocities at 55◦N between 33.8◦W and the Labradorcoast (62◦W).

Fig. 2. (a) Surface air temperature anomaly for the year 2100, in◦C, for RCP3-PD. Average warming south of 60◦ S is 1.60 timeshigher than the global mean. Average warming north of 60◦N isonly 0.83 times the global mean (1.4◦C), because the cooling ef-fect of a reduction in the Atlantic Meridional Overturning Circula-tion (AMOC) counteracts polar amplification.(b) Same for RCP8.5(with a global mean of 4.8◦C). The polar amplification factors are1.48 in the south and 1.53 in the north.

observations (cf. Fig. 10.15 inMeehl et al., 2007a). AMOCchanges in response to global warming in CLIMBER-3α aredominated by changes in heat flux, as in most other CMIP3models, while hydrological changes tend to have a minor,strengthening effect (Gregory et al., 2005). Further possibleAMOC reduction due to Greenland ice sheet melting is notaccounted for in these simulations.

3.3 Monsoon intensification

Directly influenced by atmospheric temperature patterns,large-scale monsoon circulations are arguably among themost societally relevant atmospheric systems. Within thelimitations of the statistical-dynamical atmosphere modeland its coarse resolution, CLIMBER-3α simulates the prin-cipal patterns of monsoon dynamics and precipitation rea-sonably well (Fig.4a), and its seasonal rainfall cycle com-pares favourably with reanalysis data (Fig.4b) and IPCCAR4 models (cf.Kripalani et al., 2007, Fig. 1). We find thataverage monsoon rainfall in Asia and Africa intensifies un-der global warming (Fig.5), consistent with many studies us-ing more complex models (e.g.Kripalani et al., 2007). Sea-sonal (June–August, JJA) mean rainfall associated with theSouth Asian summer monsoon (including India and the Bayof Bengal) strengthens by 10% (RCP3-PD) to 20% (RCP8.5)until the middle of the 21st century and, for RCP8.5, by up


80ºN

40ºN

40ºS

80ºS

0º

80ºN

40ºN

40ºS

80ºS

0º

100ºE100ºW 0º

10

8

6

4

2

0

3

2

1

0

a

b

Fig. 2. (a) Surface air temperature anomaly for the year 2100, in◦C, for RCP3-PD. Average warming south of 60◦S is 1.60 timeshigher than the global mean. Average warming north of 60◦N isonly 0.83 times the global mean (1.4◦C), because the cooling ef-fect of a reduction in the Atlantic Meridional Overturning Circula-tion (AMOC) counteracts polar amplification. (b) Same for RCP8.5(with a global mean of 4.8◦C). The polar amplification factors are1.48 in the south and 1.53 in the north.

Sv

aAMOC

10

12

14

16

18

year

Sv

SPG

b

2000 2200 2400 10

15

20

25

30

Fig. 3. (a) Maximum AMOC strength of the Atlantic Merid-ional Overturning Circulation (AMOC) in Sv (106m3s−1), forRCP3-PD (blue), RCP4.5 (yellow), RCP6 (grey), and RCP8.5 (red).(b) North Atlantic subpolar gyre strength, in Sv, computed frommeridional velocities at 55◦N between 33.8◦W and the Labradorcoast (62◦W).

Fig. 3. (a) Maximum AMOC strength of the Atlantic Merid-ional Overturning Circulation (AMOC) in Sv (106 m3 s−1), forRCP3-PD (blue), RCP4.5 (yellow), RCP6 (grey), and RCP8.5 (red).(b) North Atlantic subpolar gyre strength, in Sv, computed frommeridional velocities at 55◦ N between 33.8◦ W and the Labradorcoast (62◦ W).

to 30% during the 22nd century (Fig.5a). Similar resultsare found for the East Asian (including China, Fig.5b) andWest African (Fig.5c) monsoon, which both increase by upto 50% for RCP8.5. In absolute terms, this means increasesin JJA rainfall by up to 3–5 mm day−1 for RCP8.5. The de-cline of the South Asian monsoon for RCP8.5 after 2150 isdue to a shift of the center of maximum precipitation out ofthe South Asian region towards South China. While the mag-nitude and timing of this shift must be viewed in the contextof our intermediate-complexity model, observations suggestthat a displacement of the center of precipitation may be pos-sible under global warming (Wang et al., 2009). In all regionswe find a strong quasi-linear correlation of monsoon rainfallwith the regional temperature difference between land andocean (Fig.5d–f). Note that changes due to direct and in-direct aerosol effects are not captured by simulations withCLIMBER-3α and may have significant influence on mon-soon rainfall and circulation which is likely to counter-actthat of global warming (Lau and Kim, 2006; Rosenfeld et al.,2008).





2 4 6 8 10 120

5

10

month

mm

/day

b CLIMBER−3α

NCEP

80ºN

40ºN

40ºS

80ºS

0º

a

100ºE100ºW 0º

14

10

6

2

-2

-6

-10

-14

Fig. 4. (a) Difference between average boreal summer (JJA)and winter (DJF) precipitation (shading, in mm/day), and av-erage summer (JJA) near-surface winds (vectors) in the control(pre–industrial) climate of CLIMBER-3α. (b) Seasonal cycle ofmonthly average precipitation in the South Asian monsoon regionin CLIMBER-3α’s control climate (solid line) and in the NCEP-NCAR reanalysis (Kistler et al., 2001), averaged over the period1948-2007 (dashed line).

mm

/day a

South Asia

12

14

16

mm

/day b East Asia

5

10

yearm

m/d

ay c Africa

2000 2200 2400

6

8

d

e

°C

f

1 2 3 4

Fig. 5. Average seasonal (JJA) precipitation of (a) SouthAsian (67.5-112.5◦E, 15-22.5◦N), (b) East Asian (90-135◦E, 22.5-37.5◦N), and (c) African (22.5◦W-22.5◦E, 0-15◦N) summer mon-soon (mm/day). Panels (d-f) show the respective regional monsoonprecipitation versus the difference in JJA regional surface air tem-perature over land and the adjacent ocean. Generally this relationshows a clear linear trend. A shift of precipitation from the southAsian monsoon region towards the east Asian region leads to devia-tions for strong warming and does not represent a qualitative changein this relation.

Fig. 4. (a) Difference between average boreal summer (JJA)and winter (DJF) precipitation (shading, in mm day−1), and av-erage summer (JJA) near-surface winds (vectors) in the control(pre-industrial) climate of CLIMBER-3α. (b) Seasonal cycle ofmonthly average precipitation in the South Asian monsoon regionin CLIMBER-3α’s control climate (solid line) and in the NCEP-NCAR reanalysis (Kistler et al., 2001), averaged over the period1948–2007 (dashed line).

3.4 Steric sea level rise

Oceanic warming yields a steric sea level rise (SLR) of nearly0.5 m for RCP8.5 by 2100 compared to the 1980–1999 aver-age (Fig.6). Thus, thermal oceanic expansion under RCP8.5in our CLIMBER-3α simulations is about 20% higher thanthe upper 95% percentile (0.41 m by 2100) for the highestSRES scenario A1FI (see Table 10.7 inMeehl et al., 2007a)– in part because of slightly stronger anthropogenic forcingin RCP8.5. For RCP4.5 and RCP6, steric SLR is about 0.3 mby 2100 and thereby close to the upper 95% percentile pro-vided in IPCC AR4 for the similar SRES B1 scenario. Whilefor the upper three RCPs, steric SLR continues beyond 2500,the declining temperatures in RCP3-PD lead to a decelerationof steric SLR, a peaking at∼0.3 m and a gradual reversal inthe second half of the 23rd century, about 200 years afterthe peak in global temperatures. Other contributions to to-tal sea level rise, in particular from melting of the Greenlandand West Antarctic Ice Sheets, are beyond the scope of thisstudy.

During an initial phase, we find a quasi-linear relationshipbetween the rate of steric sea level rise and the global mean


2 4 6 8 10 120

5

10

month

mm

/day

b CLIMBER−3α

NCEP

80ºN

40ºN

40ºS

80ºS

0º

a

100ºE100ºW 0º

14

10

6

2

-2

-6

-10

-14

Fig. 4. (a) Difference between average boreal summer (JJA)and winter (DJF) precipitation (shading, in mm/day), and av-erage summer (JJA) near-surface winds (vectors) in the control(pre–industrial) climate of CLIMBER-3α. (b) Seasonal cycle ofmonthly average precipitation in the South Asian monsoon regionin CLIMBER-3α’s control climate (solid line) and in the NCEP-NCAR reanalysis (Kistler et al., 2001), averaged over the period1948-2007 (dashed line).

mm

/day a

South Asia

12

14

16

mm

/day b East Asia

5

10

year

mm

/day c Africa

2000 2200 2400

6

8

d

e

°C

f

1 2 3 4

Fig. 5. Average seasonal (JJA) precipitation of (a) SouthAsian (67.5-112.5◦E, 15-22.5◦N), (b) East Asian (90-135◦E, 22.5-37.5◦N), and (c) African (22.5◦W-22.5◦E, 0-15◦N) summer mon-soon (mm/day). Panels (d-f) show the respective regional monsoonprecipitation versus the difference in JJA regional surface air tem-perature over land and the adjacent ocean. Generally this relationshows a clear linear trend. A shift of precipitation from the southAsian monsoon region towards the east Asian region leads to devia-tions for strong warming and does not represent a qualitative changein this relation.

Fig. 5. Average seasonal (JJA) precipitation of(a) South Asian(67.5–112.5◦ E, 15–22.5◦ N), (b) East Asian (90–135◦ E, 22.5–37.5◦ N), and (c) African (22.5◦ W–22.5◦ E, 0–15◦ N) summermonsoon (mm/day). Panels(d-f) show the respective regional mon-soon precipitation versus the difference in JJA regional surface airtemperature over land and the adjacent ocean. Generally this re-lation shows a clear linear trend. A shift of precipitation from thesouth Asian monsoon region towards the east Asian region leads todeviations for strong warming and does not represent a qualitativechange in this relation.

surface warming (Fig.6, inset; cf.Rahmstorf, 2007). How-ever, the quasi-linear relation fails as soon as global warmingstarts to decelerate, i.e. around 2100 for RCP8.5, and sometime earlier for the lower scenarios. As suggested byVer-meer and Rahmstorf(2009), validity of semi-empirical pro-jections of sea level change based on this relation might beextended by taking rapid adjustment processes into account.

The horizontal distribution of steric SLR, shown in Fig.7for RCP3-PD, is qualitatively similar under different scenar-ios. By 2100 (Fig.7a), the weakening of the AMOC max-imum (cf. Fig. 3a) and of the North Atlantic current pro-duces a southeast-to-northwest SLR gradient in the NorthAtlantic via geostrophic adjustment (Levermann et al., 2005;Yin et al., 2010). Small shifts in the northern subpolar andsubtropical gyre systems induce smaller-scale variations ofSLR. The interhemispheric sea level pattern found byLever-mann et al.(2005) for an AMOC shutdown is not reflectedhere because the AMOC change is largely confined to theNorth Atlantic; Southern Ocean outflow, i.e. the AMOC fluxacross 30◦S, is only reduced by about 10% (not shown). By2200, the AMOC has partly recovered, and the most promi-nent feature in the North Atlantic is a negative SLR anomaly(Fig. 7b) due to a 60% increase in the subpolar gyre (Fig.3b;





m

year

°C

2000 2200 2400

0

0.5

1

1.5

2

2.5

mm

/yr

a=1.66 mm/yr/°C

0 3 0

6

Fig. 6. Globally averaged steric sea level change (in m) rela-tive to 1980-1999, under the RCP3-PD (blue), RCP4.5 (yellow),RCP6 (grey) and RCP8.5 (red) scenarios and their extensions inCLIMBER-3α. The inset shows the rate of steric sea level rise(in mm/yr, smoothed with a 15-year moving average) between1800 and 2100 as a function of global surface warming abovethe 1980-1999 mean (in ◦C). The slope of the quasi-linear part is1.66 mm yr−1 ◦C−1 (black line; cf. Rahmstorf, 2007). Circles markthe timing of peak GHG emissions.

80ºN

40ºN

40ºS

80ºS

0º

80ºN

40ºN

40ºS

80ºS

0º

160ºW100ºE100ºW 0º

a

b

6961534537292113

5-3

-118676665646362616

6-4

-14

Fig. 7. Horizontal pattern of steric sea level change (in cm), relativeto pre–industrial, under RCP3-PD: (a) Year 2100, (b) year 2200.The shading emphasizes the anomalies relative to the global averagesteric SLR (about 29 cm in 2100 and 36 cm in 2200).

Fig. 6. Globally averaged steric sea level change (in m) rela-tive to 1980–1999, under the RCP3-PD (blue), RCP4.5 (yellow),RCP6 (grey) and RCP8.5 (red) scenarios and their extensions inCLIMBER-3α. The inset shows the rate of steric sea level rise(in mm yr−1, smoothed with a 15-year moving average) between1800 and 2100 as a function of global surface warming abovethe 1980–1999 mean (in◦C). The slope of the quasi-linear part is1.66 mm yr−1◦C−1 (black line; cf.Rahmstorf, 2007). Circles markthe timing of peak GHG emissions.

Hakkinen and Rhines, 2004; Levermann and Born, 2007).In the Southern Ocean, SLR patterns in 2200 are similarto those in 2100: A strengthening of the Antarctic Circum-polar Current above the level of no motion by about 4 Svleads to below-average SLR around Antarctica (Fig.7). Ontop of that, strengthening of the Ross and Weddell gyresby 5 Sv and 6 Sv, respectively, induces large horizontal SLRanomalies.Hattermann and Levermann(2010) found that astrengthening of those gyres may significantly enhance basalice shelf melting around Antarctica.

Yin et al. (2010) showed by comparison of simulated andobserved present-day dynamic sea level patterns in twelveIPCC AR4 AOGCMs that their ensemble mean performsbetter than any of the individual models. The SLR patternfound in our analysis is in good qualitative agreement withthe ensemble mean projection of those models under theSRES A1B scenario (Yin et al., 2010).

3.5 Deep ocean warming

In contrast to the sea surface, deep ocean temperatures re-spond to atmospheric warming on centennial time scales.Due to its peaking characteristic, the RCP3-PD scenario iswell suited to study the propagation of the warming signalinto the deep ocean. Global average temperatures at 500 mand 1000 m depth exhibit delayed peaks around the years2200 and 2300, respectively, compared to a surface warmingpeak in the middle of the 21st century (Fig.8a). In the year


m

year

°C

2000 2200 2400

0

0.5

1

1.5

2

2.5

mm

/yr

a=1.66 mm/yr/°C

0 3 0

6

Fig. 6. Globally averaged steric sea level change (in m) rela-tive to 1980-1999, under the RCP3-PD (blue), RCP4.5 (yellow),RCP6 (grey) and RCP8.5 (red) scenarios and their extensions inCLIMBER-3α. The inset shows the rate of steric sea level rise(in mm/yr, smoothed with a 15-year moving average) between1800 and 2100 as a function of global surface warming abovethe 1980-1999 mean (in ◦C). The slope of the quasi-linear part is1.66 mm yr−1 ◦C−1 (black line; cf. Rahmstorf, 2007). Circles markthe timing of peak GHG emissions.

80ºN

40ºN

40ºS

80ºS

0º

80ºN

40ºN

40ºS

80ºS

0º

160ºW100ºE100ºW 0º

a

b

6961534537292113

5-3

-118676665646362616

6-4

-14

Fig. 7. Horizontal pattern of steric sea level change (in cm), relativeto pre–industrial, under RCP3-PD: (a) Year 2100, (b) year 2200.The shading emphasizes the anomalies relative to the global averagesteric SLR (about 29 cm in 2100 and 36 cm in 2200).

Fig. 7. Horizontal pattern of steric sea level change (in cm), relativeto pre-industrial, under RCP3-PD:(a) Year 2100,(b) year 2200.The shading emphasizes the anomalies relative to the global averagesteric SLR (about 29 cm in 2100 and 36 cm in 2200).

2370, about 300 years after the peak in global surface temper-atures, major anomalies of up to 2◦C are found in the upper1000 m of the North Atlantic and Southern Ocean (Fig.8b).In the North Atlantic, substantial warming is observed evenbelow 2000 m depth. Despite the weakening of the AMOCnoted earlier, the northern oceanic warming pattern clearlyreflects the structure of the overturning cell.

In general, the strong deep oceanic warming signal re-sults from outcropping of isopycnals (black lines in Fig.8b)at high latitudes, i.e. a lack of density stratification, whichis a characteristic and robust feature of the modern oceancirculation. Mixing along these surfaces of constant den-sity is strongly enhanced compared to diapycnal mixingacross these surfaces. In combination with the observed po-lar warming amplification, isopycnal mixing facilitates en-hanced heat uptake as also observed in AOGCMs (e.g.Stouf-fer et al., 2006a) and is the reason for the observed deepocean warming. These heat anomalies spread at intermediatedepths around 500 m, with the effect that peak global-averagewarming at those depths exceeds that of the ocean surface(Fig. 8a). After surface temperatures have relaxed, oceanicheat uptake is reduced and, after 2300, the ocean eventuallybecomes a very weak heat source, further damping the de-cline of surface atmospheric temperatures (compare Fig.9b).This weak heat exchange between ocean and atmosphere





1800 2000 2200 2400

0

0.5

1

1.5

2

year

° C

a

RCP3−PD

surface500m1000m

2.4

2.0

1.6

1.2

0.8

0.4

080ºN40ºN40ºS80ºS 0º

400

800

1200

1600

2000

2200

Depth

b

0.4ºC

2ºC

Fig. 8. Ocean response to the RCP3-PD scenario: (a) Global aver-age ocean temperature difference relative to pre–industrial levels, atthe ocean surface (black) and at 500 m (dark blue) and 1000 m (lightblue) depth. Due to polar amplification and outcropping oceanicisopycnals at high latitudes, peak warming is stronger at intermedi-ate depth around 500 m than at the surface. (b) Zonal average oceanwarming in the year 2370, compared to pre–industrial levels (shad-ing, in ◦C; ocean depth in m). Overlaid are contours of constantdensity (isopycnals; in kg/m3).

° C

a

0

0.5

1

1.5

yearm

c

2000 2200 2400

250

300

W m

−2

b

0

1

2

Fig. 9. Slow-down of global cooling under the RCP3-PD sce-nario: (a) Global surface air temperature anomaly as in Fig. 1c (blueline), compared to the result of the simple energy-balance equation(1) that only takes into account diffusive oceanic mixing (dashedblack line). Thin grey lines represent modified scenarios that areidentical to RCP3-PD until 2070, and after that have zero emis-sions or two, three, four or five times as large negative emissions asRCP3-PD, respectively. All curves are smoothed with an 11-yearrunning mean to remove short-term variability from solar and vol-canic sources. The vertical dashed line marks the year 2110. (b)Globally averaged heat flux from atmosphere to ocean. IncreasingGHG concentration results in enhanced oceanic heat uptake whichdeclines after the peak in atmospheric warming and vanishes aroundthe year 2300 after which the ocean becomes a source for atmo-spheric warming. The solid line is the CLIMBER-3α simulation,while the 19 AOGCM emulations using MAGICC6 are representedby the dashed line (median) and shading (50% and 80% range). Theonset of convection in the southern North Atlantic appears here as adistinct drop in ocean heat uptake after 2110 (vertical dashed line).All curves are smoothed as in (a). (c) Average depth of the NorthAtlantic ocean mixed layer in winter (January-April) south of thelatitudes of Iceland (40◦W-0◦, 50-65◦N). Starting around the year2110 (vertical dashed line), an abrupt increase in mixed layer depthmarks the onset of enhanced convection.

Fig. 8. Ocean response to the RCP3-PD scenario:(a) gobal averageocean temperature difference relative to pre-industrial levels, at theocean surface (black) and at 500 m (dark blue) and 1000 m (lightblue) depth. Due to polar amplification and outcropping oceanicisopycnals at high latitudes, peak warming is stronger at intermedi-ate depth around 500 m than at the surface.(b) Zonal average oceanwarming in the year 2370, compared to pre-industrial levels (shad-ing, in ◦C; ocean depth in m). Overlaid are contours of constantdensity (isopycnals; in kg m−3).

eventually cools deeper oceanic layers, but this cooling is soslow that the intermediate-depth warming persists for cen-turies even after surface temperatures have reached present-day levels of approximately 0.8◦C relative to pre-industrial.Conversely, these oceanic heat anomalies serve as a long-term reservoir that slowly discharges into the atmosphere anddelays surface cooling, as discussed in the following section.

4 Slow cooling under RCP3-PD

As mentioned in Sect.3.1, global cooling after the tempera-ture peak in RCP3-PD is much slower, relative to the rate ofGHG emissions, than the warming before the peak (Fig.9a,blue line). We find that two processes are responsible for thisasymmetry.

Generally oceanic heat uptake by vertical mixing createsthermal inertia that delays any temperature change at the sur-face (Fig.9b). In order to identify additional effects, we


1800 2000 2200 2400

0

0.5

1

1.5

2

year

° C

a

RCP3−PD

surface500m1000m

2.4

2.0

1.6

1.2

0.8

0.4

080ºN40ºN40ºS80ºS 0º

400

800

1200

1600

2000

2200

Depth

b

0.4ºC

2ºC

Fig. 8. Ocean response to the RCP3-PD scenario: (a) Global aver-age ocean temperature difference relative to pre–industrial levels, atthe ocean surface (black) and at 500 m (dark blue) and 1000 m (lightblue) depth. Due to polar amplification and outcropping oceanicisopycnals at high latitudes, peak warming is stronger at intermedi-ate depth around 500 m than at the surface. (b) Zonal average oceanwarming in the year 2370, compared to pre–industrial levels (shad-ing, in ◦C; ocean depth in m). Overlaid are contours of constantdensity (isopycnals; in kg/m3).

° C

a

0

0.5

1

1.5

year

m

c

2000 2200 2400

250

300

W m

−2

b

0

1

2

Fig. 9. Slow-down of global cooling under the RCP3-PD sce-nario: (a) Global surface air temperature anomaly as in Fig. 1c (blueline), compared to the result of the simple energy-balance equation(1) that only takes into account diffusive oceanic mixing (dashedblack line). Thin grey lines represent modified scenarios that areidentical to RCP3-PD until 2070, and after that have zero emis-sions or two, three, four or five times as large negative emissions asRCP3-PD, respectively. All curves are smoothed with an 11-yearrunning mean to remove short-term variability from solar and vol-canic sources. The vertical dashed line marks the year 2110. (b)Globally averaged heat flux from atmosphere to ocean. IncreasingGHG concentration results in enhanced oceanic heat uptake whichdeclines after the peak in atmospheric warming and vanishes aroundthe year 2300 after which the ocean becomes a source for atmo-spheric warming. The solid line is the CLIMBER-3α simulation,while the 19 AOGCM emulations using MAGICC6 are representedby the dashed line (median) and shading (50% and 80% range). Theonset of convection in the southern North Atlantic appears here as adistinct drop in ocean heat uptake after 2110 (vertical dashed line).All curves are smoothed as in (a). (c) Average depth of the NorthAtlantic ocean mixed layer in winter (January-April) south of thelatitudes of Iceland (40◦W-0◦, 50-65◦N). Starting around the year2110 (vertical dashed line), an abrupt increase in mixed layer depthmarks the onset of enhanced convection.

Fig. 9. Slow-down of global cooling under the RCP3-PD sce-nario: (a) global surface air temperature anomaly as in Fig.1c (blueline), compared to the result of the simple energy-balance Eq. (1)that only takes into account diffusive oceanic mixing (dashed blackline). Thin grey lines represent modified scenarios that are identicalto RCP3-PD until 2070, and after that have zero emissions or two,three, four or five times as large negative emissions as RCP3-PD, re-spectively. All curves are smoothed with an 11-year running meanto remove short-term variability from solar and volcanic sources.The vertical dashed line marks the year 2110.(b) Globally aver-aged heat flux from atmosphere to ocean. Increasing GHG con-centration results in enhanced oceanic heat uptake which declinesafter the peak in atmospheric warming and vanishes around theyear 2300 after which the ocean becomes a source for atmosphericwarming. The solid line is the CLIMBER-3α simulation, while the19 AOGCM emulations using MAGICC6 are represented by thedashed line (median) and shading (50% and 80% range). The on-set of convection in the southern North Atlantic appears here as adistinct drop in ocean heat uptake after 2110 (vertical dashed line).All curves are smoothed as in(a). (c) Average depth of the NorthAtlantic ocean mixed layer in winter (January–April) south of thelatitudes of Iceland (40◦ W–0◦, 50–65◦ N). Starting around the year2110 (vertical dashed line), an abrupt increase in mixed layer depthmarks the onset of enhanced convection.




isolate this ocean mixing effect with an intentionally simpleenergy-balance equation for global mean surface temperatureanomalyT (t), assuming a diffusive ocean (followingAllenet al., 2009; Hansen et al., 1985):

a1dT

dt= a3 log2

(C

C0

)− a0 T − a2

∫ t

0

dT (t ′)

dt ′

dt ′√

t − t ′(1)

whereC(t) is CO2 concentration;C0 = 280 ppm is the initialconcentration att = 0; a1 is the heat capacity of the oceanicmixed layer;a2 is ocean vertical diffusivity;a3 ' 1.3◦ C isclimate sensitivity not accounting for any feedbacks; and1/a0 is the climate feedback factor, such thata3/a0 is the fullclimate sensitivity, which is∼3.4◦C for CLIMBER-3α.

This model, with parametersa0−2 calibrated to matchCLIMBER-3α, reproduces the global mean temperature sim-ulated by CLIMBER-3α very well until about 2100 (blackdashed line in Fig.9a). However, at the beginning of the22nd century, the CLIMBER-3α result deviates from the sim-ple diffusive ocean heat uptake relationship: While the lat-ter projects a steady cooling trend all the way until 2500,CLIMBER-3α projects a substantial slow-down of the cool-ing around the year 2110 (vertical dashed line in Fig.9). Thecooling rate thereafter remains almost 50% lower than sug-gested by Eq. (1) for about two centuries, consequently ar-riving at a significantly higher temperature. Plotted versusCO2-equivalent GHG concentration, this is visible as a clearexcursion from the smooth hysteresis projected according toEq. (1) (Fig. 10).

To test the robustness of this behaviour, we have conductedadditional simulations using a set of scenarios that are iden-tical to RCP3-PD until 2070. Thereafter, we set CO2 emis-sions in RCP3-PD equal to zero or two, three, four or fivetimes as large negative emissions as in the original RCP3-PD,respectively. Using these modified RCP3-PD scenarios, wethen computed radiative forcings following the same processas in generating the recommended CMIP5 GHG concentra-tions of the RCPs (for details, seeMeinshausen et al., 2011).Under all these modified RCP3-PD scenarios, CLIMBER-3α projects a drop in the cooling rate at the same time, nearthe year 2110, i.e., some decades after global mean tem-perature started to decline (thin grey lines in Fig.9a). Forzero emissions after 2070 (top grey line), this even leads toa slow global warming until the early 24th century, despitethe net decrease in radiative forcing. Again, viewed rela-tive to CO2-equivalent GHG concentration, Eq. (1) yieldsessentially the same hysteresis for all the scenarios (Fig.10,dashed grey lines), while the CLIMBER-3α projections forthe modified scenarios depart from that hysteresis soon afterthe peak (solid grey lines).

This result suggests that, on the one hand, the global meantemperature response of the coupled climate model to a peak-and-decline scenario such as RCP3-PD is, up until about70 years after the peak in GHG concentrations, mainly gov-erned by the heat capacity of the oceanic mixed layer andheat exchange with the deep ocean due to mixing. The inertia


CO2 concentration (ppm)

° C

300 400 500−0.5

0

0.5

1

1.5

2500

2400

23002200 2100

2050

2000

1975

1925

1900

2025

Fig. 10. As Fig. 9a, but plotted versus CO2-equivalence concen-tration (sum of longwave absorbers) instead of time, and with theresults of eq. (1) for the modified scenarios shown as dashed greylines. This figure represents the transient “hysteresis” of globalwarming in RCP3-PD (blue line, marked every 25 years) and themodified peak-and-decline scenarios, i.e. how much GHG reduc-tion it takes to cool the surface back to a given temperature that ithad during the warming phase. The dashed lines show the hysteresisexpected from the processes represented by eq. (1), while the solidlines show the hysteresis behaviour observed in CLIMBER-3α. Theconvection-related slow-down of the cooling rate (marked by a bluecircle for the RCP3-PD scenario) translates into a widening of thehysteresis. The slow-down occurs at the same time under differentscenarios (at the beginning of the 21st century, see thin grey lines inFig. 9a), and at different CO2 concentrations.

Fig. 10. As Fig. 9a, but plotted versus CO2-equivalence concen-tration (sum of longwave absorbers) instead of time, and with theresults of Eq. (1) for the modified scenarios shown as dashed greylines. This figure represents the transient “hysteresis” of globalwarming in RCP3-PD (blue line, marked every 25 years) and themodified peak-and-decline scenarios, i.e. how much GHG reduc-tion it takes to cool the surface back to a given temperature that ithad during the warming phase. The dashed lines show the hysteresisexpected from the processes represented by Eq. (1), while the solidlines show the hysteresis behaviour observed in CLIMBER-3α. Theconvection-related slow-down of the cooling rate (marked by a bluecircle for the RCP3-PD scenario) translates into a widening of thehysteresis. The slow-down occurs at the same time under differentscenarios (at the beginning of the 21st century, see thin grey lines inFig. 9a), and at different CO2 concentrations.

induced by these processes delays the cooling that resultsfrom the decline in GHG concentrations (Stouffer, 2004). Onthe other hand, another mechanism comes into play aroundthe year 2110 that further reduces the cooling rate, over aperiod of two centuries, by almost 50%.

We find that a relatively rapid change in oceanic convec-tion is responsible for this reduction. The depth of the winter-time oceanic mixed layer in the North Atlantic is a directindicator of the strength of convection associated with theAMOC. This mixed layer depth shrinks during the warmingphase in the 21st century, but then extends strongly betweenthe years 2110 and 2150, which coincides with the changein the surface cooling rate (Fig.9c). Enhanced convection inthese latitudes results in enhanced heat loss of the ocean tothe atmosphere; thus, globally, net ocean heat uptake is re-duced by this effect (Fig.9b, solid blue line), slowing downatmospheric cooling.




5 Discussion and conclusions

We have presented large-scale climatic consequences of thenew RCP scenarios, which are designed for the forthcom-ing IPCC AR5 to span the full range of future pathwaysof anthropogenic GHG emissions currently discussed in theliterature (Moss et al., 2008, page i). CLIMBER-3α atmo-spheric temperature projections and AOGCM emulations us-ing MAGICC6 are qualitatively and quantitatively similarfor the 21st century. CLIMBER-3α temperatures tend to beslightly higher than the median of the AOGCM emulations(cf. Fig. 1), owing to the difference in climate sensitivity.While the CLIMBER-3α simulations are based on the stan-dard settings presented inMontoya et al.(2005), the widerrange of possible climate responses is covered by the emu-lation ensemble with MAGICC6, spanning climate sensitiv-ities from 1.9◦C (emulation of the NCAR PCM model) to5.7◦C (emulation of the MIROC3.2 high resolution model,seeMeinshausen et al., 2008, Table 4). With respect to at-mospheric quantities, the coarse resolution of CLIMBER-3α

and the limitations of the statistical-dynamical representationmust be kept in mind. On the other hand, large-scale oceanicquantities have been shown to be in good agreement with re-cent AOGCM results.

Our evaluation of the peak-and-decline scenario RCP3-PDreveals that global maximal temperatures can be expectedclose to 1.5◦C warming relative to pre-industrial levels. Ow-ing to negative CO2 emissions, concentrations under thisscenario are projected to drop markedly after peaking in2070, and induce a slow cooling. This finding is consis-tent with recent studies using other models of varying com-plexity (e.g.Solomon et al., 2009), which showed that un-der zero-emission scenarios temperatures are projected notto drop substantially for several centuries. Our work goesbeyond those studies by demonstrating that in a physical cli-mate model, cooling is not only delayed by mixing-relatedheat exchange with the ocean, but that dynamical effects cansignificantly add to the delay. The abrupt strengthening ofconvection in the North Atlantic indicates an important roleof internal dynamical processes in the oceans, especially be-cause the timing of the convection change seems to be inde-pendent of the rate of (negative) GHG emissions, once atmo-spheric temperatures have started to fall. Although the exacttiming will probably differ across models, the onset of strongconvection is likely to be a robust feature, because decliningatmospheric temperatures lead to stronger cooling of surfacewaters and thus reduce the stability of the water column.

The projections of steric sea level rise presented here aregenerally consistent with previous simulations. The high-est scenario, RCP8.5, being warmer than the highest SRESscenario, yields enhanced steric sea level rise of up to 2 mby 2500. According to our simulations, thermal oceanic ex-pansion can be halted only for emission trajectories corre-sponding to, or below, RCP3-PD. In this scenario we observean enhanced oceanic warming of intermediate depth due to

polar amplification in combination with the lack of oceanicdensity stratification in high latitudes. The associated heatcontent persists for centuries. Thus, these results will al-low future studies to quantify the risk of such a mid-oceanwarming for marine ecosystems (Sarmiento et al., 2004) andenvironments. For example, prolonged deep ocean warm-ing could be sufficient to trigger the dissociation of shallowmethane hydrates trapped in ocean sediments, and therebyrelease additional amounts of greenhouse gases into the at-mosphere (Reagan and Moridis, 2008; Archer et al., 2009).Furthermore, melting of Antarctic ice shelves (Holland et al.,2008) and the initiation of oceanic anoxic events (Hofmannand Schellnhuber, 2009; Stramma et al., 2009) could be fa-cilitated.

Acknowledgements.This work was supported by the Heinrich BollFoundation, the German National Academic Foundation, and theBMBF PROGRESS project (support code 03IS2191B). MM re-ceived support from the UFOPLAN project FKZ 370841103 by theGerman Federal Environment Agency. NCEP Reanalysis Deriveddata was provided by the NOAA/OAR/ESRL PSD, Boulder, Col-orado, USA, from their Web site athttp://www.esrl.noaa.gov/psd/.We thank two anonymous referees for their helpful comments.

Edited by: K. Keller

References

Allen, M. R., Frame, D. J., Huntingford, C., Jones, C. D., Lowe,J. A., Meinshausen, M., and Meinshausen, N.: Warming causedby cumulative carbon emissions towards the trillionth tonne, Na-ture, 458, 1163–1166,doi:10.1038/nature08019, 2009.

Archer, D., Buffett, B., and Brovkin, V.: Ocean methane hydrates asa slow tipping point in the global carbon cycle, P. Natl. Acad. Sci.USA, 106, 20596–20601,doi:10.1073/pnas.0800885105, 2009.

Brohan, P., Kennedy, J., Harris, I., Tett, S., and Jones, P.: Un-certainty estimates in regional and global observed temperaturechanges: A new data set from 1850, J. Geophys. Res.-Atmos.,111, D12106,doi:10.1029/2005JD006548, 2006.

Clarke, L., Edmonds, J., Jacoby, H., Pitcher, H., Reilly, J., andRichels, R.: Scenarios of Greenhouse Gas Emissions and Atmo-spheric Concentrations, Sub-report 2.1A of Synthesis and As-sessment Product 2.1 by the U.S. Climate Change Science Pro-gram and the Subcommittee on Global Change Research, US De-partment of Energy, p.154, 2007.

Fichefet, T. and Maqueda, M. A. M.: Sensitivity of a global sea icemodel to the treatment of ice thermodynamics and dynamics, J.Geophys. Res., 102, 12609–12646, 1997.

Fujino, J., Nair, R., Kainuma, M., Masui, T., and Matsuoka, Y.:Multi-gas Mitigation Analysis on Stabilization Scenarios UsingAim Global Model, Energy J., Special Issue 3, 343–354, 2006.

Gregory, J. M., Dixon, K. W., Stouffer, R. J., Weaver, A. J.,Driesschaert, E., Eby, M., Fichefet, T., Hasumi, H., Hu, A.,Jungclaus, J. H., Kamenkovich, I. V., Levermann, A., Mon-toya, M., Murakami, S., Nawrath, S., Oka, A., Sokolov,A. P., and Thorpe, R. B.: A model intercomparison ofchanges in the Atlantic thermohaline circulation in response to




increasing atmospheric CO2 concentration, Geophys. Res. Lett.,32, L12703,doi:10.1029/2005GL023209, 2005.

Hakkinen, S. and Rhines, P. B.: Decline of Subpolar North AtlanticCirculation During the 1990s, Science, 304, 555–559, 2004.

Hansen, J., Russell, G., Lacis, A., Fung, I., Rind, D., and Stone, P.:Climate response-times - dependence on climate sensitivity andocean mixing, Science, 229, 857–859, 1985.

Hattermann, T. and Levermann, A.: Response of SouthernOcean circulation to global warming may enhance basal iceshelf melting around Antarctica, Clim. Dynam., 35, 741–756,doi:10.1007/s00382-009-0643-3, 2010.

Hofmann, M. and Schellnhuber, H.-J.: Oceanic acidificationaffects marine carbon pump and triggers extended marineoxygen holes, P. Natl. Acad. Sci. USA, 106, 3017–3022,doi:10.1073/pnas.0813384106, 2009.

Holland, P. R., Jenkins, A., and Holland, D. M.: The response of iceshelf basal melting to variations in ocean temperature, J. Climate,21, 2558–2572,doi:10.1175/2007JCLI1909.1, 2008.

Kistler, R., Kalnay, E., Saha, S., White, G., Woollen, J., Chelliah,M., Ebisuzaki, W., Kanamitsu, M., Kousky, V., van den Dool, H.,Jenne, R., and Fiorino, M.: The NCEP/NCAR 50-year reanaly-sis, B. Am. Meteorol. Soc., 82, 247–267, 2001.

Kripalani, R. H., Oh, J. H., and Chaudhari, H. S.: Response ofthe East Asian summer monsoon to doubled atmospheric CO2:Coupled climate model simulations and projections under IPCCAR4, Theor. Appl. Climatol., 87, 1–28,doi:10.1007/s00704-006-0238-4, 2007.

Lau, K. M. and Kim, K. M.: Observational relationships betweenaerosol and Asian monsoon rainfall, and circulation, Geophys.Res. Lett., 33, L21810,doi:10.1029/2006GL027546, 2006.

Levermann, A. and Born, A.: Bistability of the subpolar gyrein a coarse resolution climate model, Geophys. Res. Lett., 34,L24605,doi:10.1029/2007GL031732, 2007.

Levermann, A., Griesel, A., Hofmann, M., Montoya, M., andRahmstorf, S.: Dynamic sea level changes following changes inthe thermohaline circulation, Clim. Dynam., 24, 347–354, 2005.

Levermann, A., Mignot, J., Nawrath, S., and Rahmstorf, S.: Therole of northern sea ice cover for the weakening of the thermo-haline circulation under global warming, J. Climate, 20, 4160–4171, 2007.

May, W.: Climatic changes associated with a global “2 de-grees C-stabilization” scenario simulated by the ECHAM5/MPI-OM coupled climate model, Clim. Dynam., 31, 283–313,doi:10.1007/s00382-007-0352-8, 2008.

Meehl, G. A., Stocker, T. F., Collins, W. D., Friedlingstein, P., Gaye,A. T., Gregory, J. M., Kitoh, A., Knutti, R., Murphy, J. M., Noda,A., Raper, S. C. B., Watterson, I. G., Weaver, A. J., and Zhao, Z.-C.: Climate Change 2007: The Physical Science Basis, Contribu-tion of Working Group I to the Fourth Assessment Report of theIntergovernmental Panel on Climate Change, chap. Global Cli-mate Projections, Cambridge University Press, Cambridge, Ukand New York, NY, USA, 2007a.

Meehl, G. A., Stocker, T. F., Collins, W. D., Friedlingstein, P., Gaye,A. T., Gregory, J. M., Kitoh, A., Knutti, R., Murphy, J. M., Noda,A., Raper, S. C. B., Watterson, I. G., Weaver, A. J., and Zhao,Z.-C.: Climate Change 2007: The Physical Science Basis, Con-tribution of Working Group I to theFourth Assessment Reportof the Intergovernmental Panel on Climate Change, chap. Cli-mate Models and their Evaluation, Cambridge University Press,

Cambridge, UK and New York, NY, USA, 2007b.Meinshausen, M., Raper, S. C. B., and Wigley, T. M. L.: Emulat-

ing coupled atmosphere-ocean and carbon cycle models with asimpler model, MAGICC6 - Part 1: Model description and cal-ibration, Atmos. Chem. Phys., 11, 1417–1456,doi:10.5194/acp-11-1417-2011, 2011.

Meinshausen, M., Smith, S., Calvin, K., Daniel, J. S., Kainuma,M., Lamarque, J.-F., Matsumoto, K., Montzka, S. A.,Raper, S. C. B., Riahi, K., Thomson, A. M., Velders,G. J. M., and van Vuuren, D.: The RCP GreenhouseGas Concentrations and their Extension from 1765 to 2300,Climatic Change, http://www.pik-potsdam.de/∼mmalte/pubs/09 GHG Concentrations&Extension1Sep2010.pdf, last access:7 March 2011, submitted, 2011.

Mignot, J., Levermann, A., and Griesel, A.: A decomposition ofthe Atlantic Meridional Overturning Circulation into physicalcomponents using its sensitivity to vertical diffusivity, J. Phys.Oceanogr., 36, 636–650, 2006.

Montoya, M. and Levermann, A.: Surface wind-stress threshold forglacial Atlantic overturning, Geophys. Res. Lett., 35, L03608,doi:10.1029/2007GL032560, 2008.

Montoya, M., Griesel, A., Levermann, A., Mignot, J., Hofmann,M., Ganopolski, A., and Rahmstorf, S.: The Earth System Modelof Intermediate Complexity CLIMBER-3α, Part I: descriptionand performance for present day conditions, Clim. Dynam., 25,237–263, 2005.

Moss, R., Babiker, M., Brinkman, S., Calvo, E., Carter, T., Ed-monds, J., Elgizouli, I., Emori, S., Erda, L., Hibbard, K.,Jones, R., Kainuma, M., Kelleher, J., Lamarque, J. F., Man-ning, M., Matthews, B., Meehl, J., Meyer, L., Mitchell, J.,Nakicenovic, N., O’Neill, B., Pichs, R., Riahi, K., Rose, S.,Runci, P., Stouffer, R., van Vuuren, D., Weyant, J., Wilbanks,T., van Ypersele, J. P., and Zurek, M.: Towards New Scenariosfor Analysis of Emissions, Climate Change, Impacts, and Re-sponse Strategies,http://www.ipcc.ch/pdf/supporting-material/expert-meeting-report-scenarios.pdf(last access: 7March 2011),2008.

Moss, R., Edmonds, J., Hibbard, K., Manning, M. R., Rose, S., vanVuuren, D., Carter, T., Emori, S., Kainuma, M., Kram, T., Meehl,G., Mitchell, J., Nakicenovic, N., Riahi, K., Smith, S., Stouffer,R., Thomson, A., Weyant, J., and Wilbanks, T.: The next gen-eration of scenarios for climate change research and assessment,Nature, 463, 747–756,doi:10.1038/nature08823, 2010.

Nakicenovic, N. and Swart, R.: IPCC Special Report on EmissionsScenarios, Cambridge University Press, Cambridge, 2000.

Pacanowski, R. C. and Griffies, S. M.: The MOM-3 manual, Tech.Rep. Tech. Rep. 4, NOAA/Geophyical Fluid Dynamics Labora-tory, Princeton, NJ, USA, 1999.

Petoukhov, V., Ganopolski, A., Brovkin, V., Claussen, M., Eliseev,A., Kubatzki, C., and Rahmstorf, S.: CLIMBER-2: a climatesystem model of intermediate complexity, Part I: model descrip-tion and performance for present climate, Clim. Dynam., 16, 1,2000.

Rahmstorf, S.: A Semi-Empirical Approach to Project-ing Future Sea-Level Rise, Science, 315, 368–370,doi:10.1126/science.1135456, 2007.

Reagan, M. T. and Moridis, G. J.: Dynamic response of oceanichydrate deposits to ocean temperature change, J. Geophys. Res.-Oceans, 113, C12023,doi:10.1029/2008JC004938, 2008.




Riahi, K., Gruebler, A., and Nakicenovic, N.: Scenarios oflong-term socio-economic and environmental development un-der climate stabilization, Technol. Forecast. Soc., 74, 887–935,doi:10.1016/j.techfore.2006.05.026, 2007.

Rosenfeld, D., Lohmann, U., Raga, G. B., O’Dowd, C. D., Kul-mala, M., Fuzzi, S., Reissell, A., and Andreae, M. O.: Floodor drought: How do aerosols affect precipitation?, Science, 321,1309–1313,doi:10.1126/science.1160606, 2008.

Sarmiento, J., Slater, R., Barber, R., Bopp, L., Doney, S., Hirst, A.,Kleypas, J., Matear, R., Mikolajewicz, U., Monfray, P., Solda-tov, V., Spall, S., and Stouffer, R.: Response of ocean ecosys-tems to climate warming, Global Biogeochem. Cy., 18, GB3003,doi:10.1029/2003GB002134, 2004.

Schewe, J. and Levermann, A.: The role of meridional density dif-ferences for a wind-driven overturning circulation, Clim. Dy-nam., 34, 547–556,doi:10.1007/s00382-009-0572-1, 2010.

Smith, S. J. and Wigley, T. M. L.: Multi-gas forcing stabilizationwith Minicam, Energ. J., Special Issue 3, 373–391, 2006.

Solomon, S., Plattner, G.-K., Knutti, R., and Friedling-stein, P.: Irreversible climate change due to carbon diox-ide emissions, P. Natl. Acad. Sci. USA, 106, 1704–1709,doi:10.1073/pnas.0812721106, 2009.

Stouffer, R.: Time scales of climate response, J. Climate, 17, 209–217, 2004.

Stouffer, R., Broccoli, A., Delworth, T., Dixon, K., Gudgel, R.,Held, I., Hemler, R., Knutson, T., Lee, H., Schwarzkopf, M.,Soden, B., Spelman, M., Winton, M., and Zeng, F.: GFDL’sCM2 global coupled climate models, Part IV: Idealized climateresponse, J. Climate, 19, 723–740, 2006a.

Stouffer, R. J., Yin, J., Gregory, J. M., Dixon, K. W., Spelman,M. J., Hurlin, W., Weaver, A. J., Eby, M., Flato, G. M., Hasumi,H., Hu, A., Jungclaus, J. H., Kamenkovich, I. V., Levermann, A.,Montoya, M., Murakami, S., Nawrath, S., Oka, A., Peltier, W. R.,Robitaille, D. Y., Sokolov, A. P., Vettoretti, G., and Weber, S. L.:Investigating the Causes of the Response of the ThermohalineCirculation to Past and Future Climate Changes, J. Climate, 19,1365–1387, 2006b.

Stramma, L., Visbeck, M., Brandt, P., Tanhua, T., and Wallace,D.: Deoxygenation in the oxygen minimum zone of the east-ern tropical North Atlantic, Geophys. Res. Lett., 36, L20607,doi:10.1029/2009GL039593, 2009.

Trenberth, K., Olson, J., and Large, W.: A Global OceanWind Stress Climatology based on ECMWF Analyses, Tech.Rep. NCAR/TN-338+STR, National Center for Atmospheric Re-search, Boulder, Colorado, USA, 1989.

Trenberth, K., Jones, P., Ambenje, P., Bojariu, R., Easterling, D.,Tank, A. K., Parker, D., Rahimzadeh, F., Renwick, J., Rusticucci,M., Soden, B., and Zhai, P.: Climate Change 2007: The PhysicalScience Basis, Contribution of Working Group I to the FourthAssessment Report of the Intergovernmental Panel on ClimateChange, chap. Observations: Surface and Atmospheric ClimateChange, Cambridge University Press, Cambridge, UK and NewYork, NY, USA, 2007.

van Vuuren, D. P., Den Elzen, M. G. J., Lucas, P. L., Eickhout, B.,Strengers, B. J., van Ruijven, B., Wonink, S., and van Houdt,R.: Stabilizing greenhouse gas concentrations at low levels: anassessment of reduction strategies and costs, Climatic Change,81, 119–159,doi:10.1007/s10584-006-9172-9, 2007.

Vermeer, M. and Rahmstorf, S.: Global sea level linked toglobal temperature, P. Natl. Acad. Sci. USA, 106, 21527–21532,doi:10.1073/pnas.0907765106, 2009.

Wang, C., Kim, D., Ekman, A. M. L., Barth, M. C., and Rasch, P. J.:Impact of anthropogenic aerosols on Indian summer monsoon,Geophys. Res. Lett., 36, L21704,doi:10.1029/2009GL040114,2009.

Washington, W. M., Knutti, R., Meehl, G. A., Teng, H., Tebaldi, C.,Lawrence, D., Buja, L., and Strand, W. G.: How much climatechange can be avoided by mitigation?, Geophys. Res. Lett., 36,L08703,doi:10.1029/2008GL037074, 2009.

Wigley, T. M. L. and Raper, S. C. B.: Interpretation of highprojections for global-mean warming, Science, 293, 451–454,doi:10.1126/science.1061604, 2001.

Winton, M.: Amplified Arctic climate change: What does surfacealbedo feedback have to do with it?, Geophys. Res. Lett., 33,L03701,doi:10.1029/2005GL025244, 2006.

Wise, M., Calvin, K., Thomson, A., Clarke, L., Bond-Lamberty, B.,Sands, R., Smith, S. J., Janetos, A., and Edmonds, J.: Implica-tions of Limiting CO2 Concentrations for Land Use and Energy,Science, 324, 1183–1186,doi:10.1126/science.1168475, 2009.

Wu, P., Wood, R., Ridley, J., and Lowe, J.: Temporary accelera-tion of the hydrological cycle in response to a CO2 rampdown,Geophys. Res. Lett., 37, L12705,doi:10.1029/2010GL043730,2010.

Yin, J., Griffies, S. M., and Stouffer, R. J.: Spatial Variability of SeaLevel Rise in Twenty-First Century Projections, J. Climate, 23,4585–4607,doi:10.1175/2010JCLI3533.1, 2010.



2.2 Basic mechanism for abrupt monsoon transitions

SPEC

IAL

FEAT

URE

GEO

PHYS

ICS

Basic mechanism for abrupt monsoon transitionsAnders Levermanna,b,1, Jacob Schewea,b, Vladimir Petoukhova, and Hermann Helda

aEarth System Analysis, Potsdam Institute for Climate Impact Research, 14473 Potsdam, Germany; and bInstitute of Physics, Potsdam University, 14473 Potsdam,Germany

Edited by Hans Joachim Schellnhuber, Potsdam Institute for Climate Impact Research, Potsdam, Germany and approved August 18, 2009 (received for reviewFebruary 11, 2009)

Monsoon systems influence the livelihood of hundreds of millionsof people. During the Holocene and last glacial period, rainfallin India and China has undergone strong and abrupt changes.Though details of monsoon circulations are complicated, obser-vations reveal a defining moisture-advection feedback that domi-nates the seasonal heat balance and might act as an internal ampli-fier, leading to abrupt changes in response to relatively weak exter-nal perturbations. Here we present a minimal conceptual modelcapturing this positive feedback. The basic equations, motivated byobserved relations, yield a threshold behavior, robust with respectto addition of other physical processes. Below this threshold in netradiative influx, Rc , no conventional monsoon can develop; aboveRc , two stable regimes exist. We identify a nondimensional para-meter l that defines the threshold and makes monsoon systemscomparable with respect to the character of their abrupt transition.This dynamic similitude may be helpful in understanding past andfuture variations in monsoon circulation. Within the restrictions ofthe model, we compute Rc for current monsoon systems in India,China, the Bay of Bengal, West Africa, North America, and Australia,where moisture advection is the main driver of the circulation.

Earth system | tipping element | abrupt climate change | atmosphericcirculation | nonlinear dynamics

M onsoon rainfall shapes regional culture and the livelihoodsof hundreds of millions of people (e.g., 1, 2). The future evo-

lution of monsoon rainfall under increasing levels of atmosphericCO2 and aerosol pollution is highly uncertain (3). Although green-house gas abundance tends to increase monsoon rainfall strength(4–6), the situation is more complex with changing aerosol distri-bution (7, 8). Given this large uncertainty in the future forcing ofmonsoons, it is crucial to understand internal monsoon dynam-ics, especially with respect to self-amplifying feedbacks, whichmight result in potentially strong responses to small perturba-tions. Zickfeld et al. (2005) found two stable states in a simplemodel of the Indian summer monsoon, which in principle allowsfor rapid transition between radically different monsoon circu-lations (9, 10) and thereby identified the Indian monsoon as apotential tipping element of the climate system (11). Evidence forsuch behavior is found in paleodata that show rapid and strongvariations in Indian and East Asian monsoon rainfall (12, 13).These abrupt changes have been linked to climatic events in theNorth Atlantic for the last glacial period (14, 15) as well as forthe Holocene (16, 17). Though a physical mechanism for thisteleconnection has been suggested (18), relevant climatic signalsof the North Atlantic events in Asia (such as temperature andmoisture anomalies) are very small (19) indicating that internalfeedbacks in monsoon dynamics may have amplified the weakexternal forcing.

Both spatial patterns and temporal evolution of monsoon rain-fall are influenced by a number of physical processes (7, 18, 20–28)as well as characteristics of vegetation (29–31) and topography(32). Though these details are crucial for the specific behaviorof different monsoon systems and their significance will vary fromregion to region, there exist defining processes fundamental to anymonsoon dynamics (e.g. 33, 34). These processes are the advectionof heat and moisture during monsoon season and the associatedrainfall and release of latent heat. In accordance with Zickfeld

et al. (9), we suggest the positive moisture-advection feedback (21)as a candidate for the main cause of abrupt changes in monsoondynamics.

We derive a minimal conceptual model of a monsoon circulation(Fig. 1A), comprising merely conservation of heat and moisture,knowingly neglecting a large number of relevant physical processesin order to distill the fundamental nonlinearity of monsoon circu-lations. The resulting governing equation exhibits the necessarysolution structure to explain qualitatively both strong, persistentchanges in monsoon rainfall, as observed in paleorecords, andabrupt variablity within one rainy season. This equation’s dynamicsimilitude, expressed through a single dimensionless number l,which defines the threshold behavior and makes different mon-soon systems comparable with respect to their transition, mayserve as a building block for understanding past and future abruptchanges in monsoon dynamics.

ResultsMoisture-Advection Feedback in Monsoon Dynamics. The seasonalevolution of the continental heat budget for different monsoonsystems (Fig. 2) shows that sensible heat flux from the land surfaceincreases during spring and heats up the atmospheric column priorto the rainy season. The onset of heavy rainfall (red vertical linesin Fig. 2) is associated with a drop in surface temperature on land,and consequently, sensible heat flux reduces drastically. Duringthe monsoon season, latent heat release dominates the atmos-pheric heat content, whereas net radiative fluxes are relativelyconstant throughout the year, reflecting the stabilizing long-waveradiative feedback. In response to the latent heat release, thermalenergy is transported out of the region through large-scale advec-tion and synoptic processes. The main dynamical driver of themonsoon is therefore the positive moisture-advection feedback(Fig. 1A): The release of latent heat from precipitation over landadds to the temperature difference between land and ocean, thusdriving stronger winds from ocean to land and increasing in thisway landward advection of moisture, which leads to enhanced pre-cipitation and associated release of latent heat. In the following,we seek to capture this feedback in a minimal conceptual model.

Minimal Conceptual Model for Abrupt Monsoon Transitions. For thispurpose, consider the heat-balance equation of the monsoonseason (Fig. 2, for example at blue vertical line).

L · P − εCpW · ΔT + R = 0, [1]

where latent heat release and net radiation into the atmosphericcolumn, R, balance heat divergence, and the relatively weak con-tribution from sensible heat transport from the land surface to theatmospheric column has been neglected. ΔT is the atmospheric

Author contributions: A.L. designed research; A.L. and V.P. performed research; A.L., J.S.,and H.H. analyzed data; and A.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/cgi/content/full/0901414106/DCSupplemental.

www.pnas.org / cgi / doi / 10.1073 / pnas.0901414106 Early Edition 1 of 6

2.2 Basic mechanism for abrupt monsoon transitions 33

Fig. 1. Basic mechanism of abrupt monsoon transitions. (A) Geometry ofconceptual model and fundamental moisture-advection feedback. The samenotation as in the text is used for wind W , precipitation P, net radiative influxR, vertical scale H and horizontal scale L. Arrows in the feedback loop indi-cate the amplification of one physical processes by another. (B) Mechanismof the abrupt transition. Heating by latent heat release and cooling throughheat advection compensate each other, and both decrease with decreasingwinds (or equivalently, land–ocean temperature difference ΔT ; see Eq. 2).The resultant heating balances the negative net radiative flux as long as it isabove a threshold RC , below which no conventional monsoon exists.

temperature difference between land and ocean. Latent heat ofcondensation is L = 2.6 · 106J/kg and volumetric heat capacityof air at constant pressure Cp = 1, 295 J/m3/K . P is the meanprecipitation over land (in kg/m2/s). The ratio ε = H/L betweenvertical extent H of the lower troposphere and the horizontal scaleL of the region of precipitation (Fig. 1) enters because of the bal-ance of the horizontal advective heat transport and the verticalfluxes of net radiative influx R and precipitation P. A length scalefor the coastline drops out. Note that no annual cycle is includedin the model. Only budgets for the rainy season are considered.Consequently, this model does not capture any interseasonal orany interannual dynamics. Equations are only valid for landwardwinds, W ≥ 0.

Assuming dominance of ageostrophic flow in low latitudes,the landward mean wind W is taken to be proportional to thetemperature difference between land and ocean (33, 36, 37):

W = α · ΔT . [2]

This assumption of a linear relation between the two quanti-ties is supported by National Centers for Environmental Predic-tion/National Center for Atmospheric Research (NCEP/NCAR)reanalysis data (Fig. 3) with correlation coefficients above 50% forall regions. There is significant scatter in some plots, reflecting thefact that other processes may be relevant for the monsoon dynam-ics in the corresponding regions. A possible offset, as observed in

some regions, does not alter the model behavior qualitatively. Thisoffset is discussed together with other possibly relevant processesin the SI Appendix. Here we seek to capture only processes rel-evant to the self-amplification feedback. Neglecting the effect ofevaporation over land and associated soil-moisture processes inthe continental moisture budget, precipitation has to be balancedby the net landward flow of moisture

εW · ρ(qO − qL) − P = 0, [3]

where qO and qL are specific humidity over ocean and land, andρ = 1.3 kg/m3 is mean air density.

Note that evaporation is clearly an important process for themoisture budget (e.g. (38)) and is omitted in Eq. 3 only for thesake of clarity. Including evaporation does not change the modelbehavior qualitatively (see SI Appendix). It does, however, shift thevalue of the critical threshold, as we will show in the next sectionwhen applying our model to data. In the minimalistic spirit ofthis section, we omit the effect of evaporation here because it isnot of first order to the problem. Consistent with reanalysis data(Fig. 4) and theoretical considerations (36, 39), continental rain-fall is assumed to be proportional to the mean specific humiditywithin the atmospheric column

P = βqL. [4]

The effect of an offset between these quantities does not changethe model behavior qualitatively (see SI Appendix). This set ofassumptions (Eqs. 1–4) yields the dimensional governing equationof the model

W 3 + β

ερW 2 − α

εCp(LqOβ + R) · W − αβ

ε2ρCp· R = 0. [5]

Note that through the linear relation of Eq. 2, this equation canequally be understood as an expression for the temperature dif-ference beween land and ocean ΔT , which might be more usefulfor some applications. Introduction of nondimensional variablesw ≡ W ερ/β and p = P/(qOβ) results in the nondimensionalequation

w3 + w2 − (l + r)w − r = 0, [6]

which depends on two parameters only: The dimensionless netradiative influx r ≡ R · εαρ2/(Cpβ

2) and a measure for the relativerole of latent and advective heat transport

l ≡ (εαρ2LqO)/(Cpβ) = (LqOβ)/(Cpβ2/(εαρ2)). [7]

Large l corresponds to a strong influence of moisture advec-tion (scaling as LqOβp) on the continental heat budget comparedwith heat advection by large-scale and synoptic processes (scalingas Cpβ

2w2/(αερ2)). The nondimensional precipitation is directlyrelated to the wind through p = w/(1 + w).

Solutions w(r) of Eq. 6 are determined entirely by a choice ofthe only free parameter l, which can be expressed in terms of acritical threshold of net radiative flux rc, below which no physicalsolution exists (Fig. 5). The critical point (rc, wc) will vary for dif-ferent monsoon systems. It is directly linked to the only remainingparameter l, through

wc(wc + 1)2 = l/2. [8]

and therefore uniquely defines the solution w(r) of the model. Thecritical radiation can be computed from

rc = −w2c(2wc + 1) [9]

Thus for large l (as observed in some monsoon systems) the criticalthreshold is well approximated by rc ≈ −l. Note that l is scalinglike qOα/β where α and β have clear-cut physical meaning (39).α is essentially a function of the near-surface cross-isobar angle andthereby a function of surface roughness and static stability of the

2 of 6 www.pnas.org / cgi / doi / 10.1073 / pnas.0901414106 Levermann et al.


SPEC

IAL

FEAT

URE

GEO

PHYS

ICS

Fig. 2. Seasonal heat contributions to the atmospheric column over different continental monsoon regions in NCEP/NCAR reanalysis data (35). Radiativeheating of the land surface in spring enhances sensible heat flux from the ground (’Sensible’). During the rainy season, latent heat release dominates the heatbudget (’Latent’). Radiative heat flux comprises all radiative fluxes in and out of the atmospheric column (’Radiative’). The excess heat is transported out ofthe continental monsoon region though large-scale advective and synoptic processes (’Convergence’). Error bars give the standard deviation from the 60 yearsfor which data is available (1948–2007). Regions from which values were taken are defined in the (SI Appendix). The red and blue vertical lines emphasize themonths of maximum sensible heat flux and latent heat flux, respectively.

planetary boundary layer (PBL). β is governed by the characteris-tic turnover (recycling) time of liquid water in the atmosphere andthereby determined by static stability and vertical velocity in thePBL. Any physical solution for r > rc is characterized by landwardwinds w > 0 and positive precipitation p > 0.

Let us now try to understand the physical mechanism behindthe threshold behavior observed in Fig. 5. In the tropics netradiative influx is negative, i.e. radiation cools the atmosphericcolumn. During monsoon season the same is true for the advec-tion of heat by the winds because winds blow predominantlyfrom the colder oceanic surrounding. The release of latent heatcompensates for both of these heat-loss processes. If monsoonwinds get weaker, condensation and therefore latent heat releasethrough precipitation are reduced (moisture-advection feedback,Fig. 1A). The abruptness of the transition emerges through anadditional stabilizing effect of the direct heat advection which iscooling the atmospheric column and is also reduced for reducedmonsoon winds. Thus both advection-related processes, precipi-tative warming and thermal cooling, are simultaneously reducedand partly compensate until a threshold is reached at whichcondensation/precipitation cannot provide the necessary latentheat to sustain a circulation. As a consequence, land-ocean

temperature difference ΔT and therewith monsoon winds breakdown (Fig. 1B).

Estimate of Critical Threshold for Current Monsoon Systems. In orderto estimate the critical threshold of different monsoon systemswithin the limitation of this very simple model, we use time seriesof precipitation P, radiation R, temperature difference ΔT , andspecific humidity qO from the NCEP/NCAR reanalysis data (35)to compute time series for α(t) = (LP + R)/(εCpΔT2) andβ(t) = ((LP+R)·ρP)/((LP+R)qOρ−CpΔTP), assuming applica-bility of the model and stationary statistics within the observationalperiod (1948-2007). Via α(t) and β(t), the parameter l(t) is knownand the system is estimated for each year. As a simple test for themodel, we calculate the remaining quantity that is not used for thecomputation of α(t) and β(t), the specific humidity over land

q′L(t) = qO(t) − CPΔT(t)P(t)

ρ(LP(t) + R(t)). [10]

The resulting model estimate of the specific humidity q′L compares

reasonably well (Fig. S2 in the SI Appendix) with the indepen-dently observed qL that was used in Fig. 4 to motivate the relationbetween specific humidity and precipitation (Eq. 4).

Levermann et al. Early Edition 3 of 6


Fig. 3. Landward zonal wind versus temperature difference between landand ocean during monsoon season [NCEP/NCAR reanalysis data (35)]. The linesshow best linear regression with correlation r.

Via the definition of l, we compute Rc from the time seriesα(t) and β(t) for each year between 1948-2007. Note that the onlyquantity that is not constrained by data in this computation isthe parameter ε, which defines the ratio of vertical and horizon-tal scale. However, the critical threshold RC is independent of ε,and thus the calculation depends only on relatively robust aver-aged values of precipitation, net radiation, average temperaturedifference between land and ocean, specific humidity over ocean,and the natural constants ρ, L, and Cp. We interpret the resultingdistribution of the critical threshold Rc (Fig. 6, blue) as a noisyestimate of a stationary critical threshold.

Within the limitations of the model, the observed net radiationis higher than the critical threshold in the Bay of Bengal, WestAfrica, and China. In India, North America, and Australia, the dis-tributions have significant overlap. Incorporating evaporation intothe model shifts the distribution toward lower thresholds (Fig. 6,red), while at the same time increasing the precipitation thresh-old Pc. Standard bootstrapping (see SI Appendix) reveals that theestimates in Fig. 6 are already relative robust distributions, in viewof the simplicity of the model approach.

DiscussionA minimal conceptual model for monsoon circulations that cap-tures the moisture-advection feedback is presented. The model isunlikely to describe details of monsoon circulations quantitatively,nor is it meant to capture all dynamical processes of a monsooncirculation. Following a minimalistic philosophy, the model com-prises the necessary processes for a positive feedback and therebydemonstrates the possibility of an abrupt transition of monsooncirculations from a state with strong rainfall to a weak precipi-tation state. All model equations are backed by relations foundin NCEP/NCAR reanalysis data. For India, it has been shownthat this data-set properly represents the statistics of precipita-tion when compared with regional observations with higher spatialresolution (40).

Because the processes represented in our model are fundamen-tal to monsoon systems, we believe that the results strongly suggestthe possibility of abrupt monsoon transitions. Because the domi-nant driving process is captured, it is not impossible that the modelcan provide a reasonable estimate for the critical threshold, Rc,once all necessary processes are incorporated. The bifurcationstructure of the model is robust with respect to incorporation ofother physical processes (see SI Appendix) and only changes quali-tatively when either of these perturbations dominate the dynamics.Thus, the applicability of our model is based on the assumptionthat moisture advection is the dominant process in the heat budgetof a monsoon system.

The possibility of abrupt transition is due to the competitionof the main heat transport processes during the rainy season.Although latent heat release through precipitation warms theatmospheric column, direct advection of heat is cooling it. Bothprocesses decrease with decreasing monsoon winds and therebycompensate each other with respect to the net heat injection intothe atmospheric column. The threshold of this stabilizing effect isset by the radiative cooling, which is characteristic to low-latitudesand is strongly influenced by aerosol distribution in the region.

According to our model, abrupt transitions may occur in two dif-ferent ways. For net radiation above the critical threshold R > RC,the system is bistable. Because the model only describes the rainyseason and does not capture the annual monsoon cycle, abrupttransitions in the bistable regime can only be interpreted intrasea-sonally, e.g., a month of heavy rain followed by a month of extra-ordinarily weak precipitation. An example could be the extremelyweak rainfall in July and September observed in India in the year2002, in which the rest of the season exhibited average rainfall (41).

Our model does not capture the dynamics of a decline orincrease in monsoon strength over several years. Thus, paleo-data in which strong variation in monsoon rainfall have beenrecorded cannot be explained by the bistable regime because theserecordings show monsoon changes over several years, decades, or

Fig. 4. Precipitation versus specific humidity over land during monsoon sea-son [NCEP/NCAR reanalysis data (35)]. The lines show best linear regressionwith correlation r.



SPEC

IAL

FEAT

URE

GEO

PHYS

ICS

Fig. 5. Solution of the nondimensional governing Eq. 6. (Top) Nondimen-sional landward wind for two values of the only parameter l = 0 and l = 1.(Middle) Corresponding nondimensional precipitation. (Bottom) Nondimen-sional precipitation for higher values of l as observed in some monsoonsystems. The functional form of the solution is not changed qualitatively.The critical threshold (wc(l), pc(l)) is given as the black curve in each frame.

even centuries. Such behavior would correspond to a shift of thesystem across the critical threshold into the monostable regimeR < RC without a conventional monsoon. If persistent, such ashift would be visible in paleorecords.

The reduction of the full set of model parameters to a sin-gle scaling number l, which determines the system and therebythe critical threshold, testifies to a remarkable dynamic similitudewith respect to the atmospheric quantities α, β, and q0. Differ-ent monsoon systems with the same l will have the same transitionbehavior. As illustrated in Fig. 5, l provides a measure for the posi-tion and the sharpness of the transition, i.e., for the point (rc, wc) instate space. This means, in particular, that a decrease in inflowinghumidity q0 associated with, e.g., colder climate conditions (whichwould decrease the threshold Rc and shift the system closer to acollapse) could be compensated by decreasing β, representing alower turnover (recycling) time of moisture in the atmosphere,which is influenced by, e.g., aerosols.

Fig. 6. State of current monsoon systems with respect to the criticalpoint computed from the conceptual model. The black distribution reflectsobserved fluctuations in net radiation in the different monsoon systems. Theblue distribution provides the computed critical threshold from the concep-tual model. The red distribution includes the effect of evaporation. Linesgive a Gaussian function with the same mean and standard deviation as thecorresponding discrete distribution.

The parameter q0 in our model can be interpreted in a ratherbroad sense as a specific humidity of the vicinity influencing amonsoon region. As an example, the years with anomalously highsnow cover over the Tibetan Plateau in spring and early summer(20, 26) could be characterized by a decrease in q0 during mid-summer, which would shift the threshold value Rc for the Indianmonsoon closer to the observed precipitation over the region, thusincreasing a possibility of monsoon breakdown in those years. Sim-ilarly, a colder climate with generally decreased humidity qO couldbe closer to the critical threshold, which might be the reason forless-stable monsoon circulations during glacial periods.

In the future, net radiation may be reduced through aerosol pol-lution, which will push the system qualitatively closer to the criticalthreshold (7). On the Indian subcontinent, in China, and in partsof sub-Saharan Africa, agricultural productivity is closely linked toand limited by monsoon rainfall. Food security in these regions isparticularly sensitive to monsoon variability (42–45). Studies withcomprehensive models are necessary to confirm or reject the ideaof the existence of a threshold as well as its position.

ACKNOWLEDGMENTS. We thank B.N. Goswami, R. Krishnan, and J. Srini-vasan for helpful hints and discussions; and T. Lenton for useful comments onthe manuscript. This work was funded by the Heinrich Böll Foundation, theGerman National Academic Foundation, and the German Federal Ministry ofEducation and Research.

1. Auffhammer M, Ramanathan V, Vincent JR (2006) Integrated model shows that atmos-pheric brown clouds and greenhouse gases have reduced rice harvests in India. ProcNatl Acad Sci USA 103:19668–19672.

2. Zhang P, et al. (2008) A test of climate, sun, and culture relationships from an 1810-yearchinese cave record. Science 322:940–942.

3. Patra PK, Behera SK, Herman JR, Akimoto S, Yamagata T (2005) The indian summermonsoon rainfall: Interplay of coupled dynamics, radiation and cloud microphysics.Atmos Chem Phys Discuss 5:2879–2895.

4. Meehl GA, Washington WM (1993) South Asian summer monsoon variability in amodel with doubled atmospheric carbon dioxide concentration. Science 260:1101–1104.

5. Zwiers FW, Kharin VV (1998) Changes in the extremes of the climate simulated by CCCGCM2 under CO2 doubling. J Climate 11:2200–2222.

6. May W (2002) Simulated changes of the Indian summer monsoon under enhancedgreenhouse gas conditions in a global time-slice experiment. Geophys Res Lett29:1118.

7. Ramanathan V, et al. (2005) Atmospheric brown clouds: Impacts on South Asianclimate and hydrological cycle. Proc Natl Acad Sci USA 102:5326–5333.

8. Lau KM, Kim KM (2006) Observational relationships between aerosol and asianmonsoon rainfall, and circulation. Geophys Res Lett 33:L21810.

9. Zickfeld K, Knopf B, Petoukhov V, Schellnhuber HJ (2005) Is the indian summermonsoon stable against global change? Geophys Res Lett 32:L15707.

10. Knopf B, Zickfeld K, Flechsig M, Petoukhov V (2008) Sensitivity of the Indian monsoonto human activities. Adv Atmos Sci 25:932–945.

11. Lenton TM, et al. (2008) Tipping elements in the earth’s climate system. Proc Natl AcadSci USA 105:1786–1793.

12. Wang P, et al. (2005) Evolution and variability of the Asian monsoon system: State ofthe art and outstanding issues. Quaternary Sci Rev 24:595–629.

13. Wang Y, et al. (2008) Millennial- and orbital-scale changes in the East Asian monsoonover the past 224,000 years. Nature 451:1090–1093.

14. Overpeck JT, Anderson D, Trumbore S, Prell W (1996) The southwest Indian monsoonover the last 18000 years. Climate Dynamics 12:213–225.

15. Burns SJ, Fleitmann D, Matter A, Kramers J, Al-Subbary AA (2003) Indian ocean cli-mate and an absolute chronology over Dansgaard/Oeschger events 9 to 13. Science301:1365–1367.

Levermann et al. Early Edition 5 of 6


16. Gupta AK, Anderson DM, Overpeck JT (2003) Abrupt changes in the Asian southwestmonsoon during the Holocene and their links to the North Atlantic ocean. Nature421:354–357.

17. Wang Y, et al. (2005) The Holocene Asian Monsoon: Links to solar changes and NorthAtlantic climate. Science 308:854–857.

18. Goswami BN, Madhusoodanan MS, Neema CP, Sengupta D (2006) A physical mecha-nism for North Atlantic SST influence on the Indian summer monsoon. Geophys ResLett 33:L02706.

19. Zhang R, Delworth TL (2005) Simulated tropical response to a substantial weakeningof the Atlantic Thermohaline Circulation. J Climate 18:1853–1860.

20. Hahn DG, Shukla J (1976) An apparent relationship between Eurasian snow cover andIndian monsoon rainfall. J Atmos Sci 33:2461–2462.

21. Webster PJ, et al. (1998) Monsoons: Processes, predictability, and the prospects forprediction. J Geophys Res 103:14451–14510.

22. Krishnamurthy V, Goswami BN (2000) Indian monsoon-ENSO relationship on inter-decadal timescale. J Climate 13:579–595.

23. Clark CO, Cole JE, Webster PJ (2000) Indian ocean SST and Indian summer rainfall:Predictive relationships and their decadal variability. J Climate 13:2503–2519.

24. Kucharski F, Molteni F, Yoo JH (2006) SST forcing of decadal Indian monsoon rainfallvariability. Geophys Res Lett 33:L03709.

25. Goswami BN, Xavier PK (2005) ENSO control on the south Asian monsoon throughthe length of the rainy season. Geophys Res Lett 32:L18717.

26. Dash SK, Singh GP, Shekhar MS, Vernekar AD (2005) Response of the Indian sum-mer monsoon circulation and rainfall to seasonal snow depth anomaly over Eurasia.Climate Dynamics 24:1–10.

27. Wang B (2005) The Asian Monsoon (Springer, Berlin).28. Yang J, Liu Q, Xie SP, Liu Z, Wu L (2007) Impact of the Indian ocean SST basin mode

of the Asian summer monsoon. Geophys Res Lett 34:L02708.29. Meehl GA (1994) Influence of the land surface in the Asian summer monsoon: External

conditions versus internal feedbacks. J Climate 7:1033–1049.30. Claussen M (1997) Modeling bio-geophysical feedback in the African and Indian

monsoon region. Climate Dynamics 54:247–257.31. Robock A, Mu M, Vinnikov K, Robinson D (2003) Land surface conditions over Eurasia

and Indian summer monsoon rainfall. J Geophys Res 108:4131.

32. Liu X, Yin Z (2002) Sensitivity of East Asian monsoon climate to the uplift of the TibetanPlateau. Palaeogeogr Palaeoclimatol Palaeoecol 183:223–245.

33. Webster PJ (1987) The elementary monsoon. In Monsoons, eds Fein JS, Stephens PL(Wiley, New York), pp. 3-32.

34. Webster PJ (1987) The variable and interactive monsoon. In Monsoons, eds Fein JS,Stephens PL (Wiley, New York), pp. 269–330.

35. Kistler R, et al. (2001) The NCEP/NCAR 50-year reanalysis. Bull Amer Meteor Soc82:247–267.

36. Petoukhov VK (1982) Two mechanisms of temperature oscillations in a thermo-dynamical model of the troposphere-stratosphere system. Atmos Ocean Phys 18:126–137.

37. Brovkin V, Claussen M, Petoukhov V, Ganopolski A (1998) On the stability of theatmosphere- vegetation system in the Sahara/Sahel region. J Geophys Res 103:31613–31624.

38. Eltahir EAB (1998) A soil moisture-rainfall feedback mechanism: 1. theory andobservations. Water Resour Research 34:765–776.

39. Petoukhov V, et al. (2000) CLIMBER-2: A climate system model of intermediatecomplexity. Part I: model description and performance for present climate. ClimateDynamics 16:1.

40. Goswami BN, Ramesh KV (2006) A comparison of interpolated NCEP (I-NCEP) rainfallwith high-resolution satellite observations. Geophysl Res Lett 33:L19821.

41. Fasullo J (2005) Atmospheric hydrology of the anomalous 2002 Indian summermonsoon. Monthly Weather Rev 133:2996–3014.

42. Kumar KK, Kumar KR, Ashrit RG, Deshpande NR, Hansen JW (2004) Climate impactsin Indian agriculture. Int J Climatol 24:1375–1393.

43. Gregory PJ, Ingram JSI, Brklacich M (2005) Climate change and food security. PhilosTrans R Soc London Ser B 360:2139–2148.

44. Haile M (2005) Weather patterns, food security and humanitarian response insub-Saharan Africa. Philos Trans R Soc London Ser B 360:2169–2182.

45. Tao F, et al. (2004) Variability in climatology and agricultural production in China inassociation with the East Asian summer monsoon and El Niño Southern Oscillation.Climate Res 28:23–30.

46. Hansen J, et al. (1983) Efficient three-dimensional global models for climate studies:Models I and II. Monthly Weather Rev 111:609–662.



Basic mechanism for abrupt monsoontransitionsAnders Levermann ∗ †, Jacob Schewe ∗ † , Vladimir Petoukhov ∗ , and Hermann Held ∗

∗Earth System Analysis, Potsdam Institute for Climate Impact Research, Potsdam, Germany, and †Institute of Physics, Potsdam University, Potsdam, Germany

Submitted to Proceedings of the National Academy of Sciences of the United States of America

Supporting Information

Monsoon regions and definitionsNCEP/NCARreanalysisdatawasobtained fromhttp://www.cdc.noaa.gov/as 60-year monthly mean time series, starting January 1948.Heat fluxand precipitation data are averaged over land in each monsoon region.∆T is the difference between the average temperatures over land andocean. HumiditiesqL andqO refer to the same land and ocean re-gions, respectively. The near-surface, landward zonal wind velocityW is averaged over a third region. All three regions are given intable 1 and illustrated in figures S1 and S2, together with therespec-tive definitions of the monsoon season that are used for the temporalaverages shown in figures 3 and 4.W is averaged vertically between850hPa and1000hPa.qO is averaged vertically between600hPa and1000hPa. All other vertical averages are over the entire atmosphericcolumn.

Robustness of R c estimateIn order to determine the statistical stability of the estimate of thedistribution ofRC , we proceeded in two steps. (1) From the timeseries’ autocorrelation we decided to treat the time seriesof α andβas containing virtually no memory, i.e. values from different yearscan be treated as statistically independent. Note that thisassumptiondoes not contradict the existence of interannual to decadalvariabilitythat is forced externally. (2) Via bootstrapping we generated surro-gate time series of length 60. From this time series ensemblewefound that the standard deviation of mean and standard deviation oftheRC -distribution are one order of magnitude below the standarddeviation of the shown distribution ofRC-estimates. Hence the redcurves in figure 6 are already relatively robust estimates, in view ofthe simplicity of the model approach.

Structural sensitivity of conceptual modelIn order to analyse the structural robustness of the governing equa-tion [6] to inclusion of further physical processes we startfrom thenon-dimensional forms of the unperturbed equations [1] and[3]

lp − w2 + r = 0 [S1]

w (1 − p) − p = 0 [S2]

using the same definitions of parametersl andr and non-dimensionalvariablesw andp as in the main text. Note that the parameterl ascomputed from observations is of the order104. Its qualitative in-fluence on the solution structure can, however, already beenseen forl = 1. Since 1 is the only other scale in the non-dimensional gov-erning equation, we will usel = 1 as an example. Similarly we willshow the qualitative influence of other paramters by settingthem to0.5 without claiming this to be an observed value. Note that for somecases the critical precipitation reduces and could in principle becomezero or negative. This would change the model behaviour qualita-tively. However this can only be the case when the correspondingprocess dominates the dynamics and is not merely a perturbation tothe dynamics described in the core model. Non of the processes in-cluded eliminates the bifurcation for small parameter values. In thissense the model behaviour is robust.

Addition of evaporation. To our understanding, the strongest miss-ing process is the effect of evaporation over land. In order to estimatethe critical threshold of different monsoon systems we generalize themodel by adding evaporation to the moisture budget (equation [3]). Inthe reanalysis data evaporation provides a very weak feedback withinthe dynamics and is well approximated byP −E ≈ γP −EO, withregion-specific constantsE0 andγ which is close to unity (figure S4).For this purpose we replace equation [S2] byw (1 − p) − (γp − e)with e ≡ E0/ (βqO). This equation can also be derived from the ap-proach by Hansen et al. [1] and an additional assumption of constanttotal soil moisture within a rainy season. We obtain

p =w + e

w + γ[S3]

Accordingly, the governing equation transforms to

w3 + γw2 − (l + r)w − (el + γr) = 0 [S4]

Both constant and precipitaton-dependent evaporation shift criticalradiation to lower and critical precipitation towards higher values (fig-ure S5). As in the minimal model set-up, the critical threshold can becomputed analytically from

wc (wc + γ)2 = (γ − e) l/2 [S5]

rc = 3w2c + 2γwc − l

By additional use of the evaporation time seriesE(t) fromNCEP/NCAR reanalysis, the parametere can be computed.

e = E(t)/ (βqO(t)) [S6]

By assumingγ to be constant (taken from the regression in figure S4),the critical threshold of this generalized model can be computed (fig-ure 6).

Addition of cloud-albedo feedback. Assuming that cloud-albedoover land increases with the atmospheric moisture content we adda term−a′qL to equation [1] wherea′ is a constant. Consequentlythe nondimensional heat equation is transformed into

(l − a) p − w2 + r = 0 [S7]

wherea ≡ a′qOǫα/`

Cpβ2´

. The governing equation

w3 + w2 − (l + r − a)w − r = 0 [S8]

shows the same functional form with an effective shift of theorigi-nal l-parameter towards lower values (figure S6). This reduces thesignificance of the moisture-advection feedback for the monsoon cir-culation by lowering the threshold precipitation value. Onthe otherhand the threshold is reached at higher net radiationrc.

Reserved for Publication Footnotes

www.pnas.org/cgi/doi/10.1073/pnas.0709640104 PNAS Issue Date Volume Issue Number 1–10


Tab. S1: Regional definitions used for data analysis

Monsoon system INDIA BAY OF BENGAL CHINA W.AFRICA N.AMERICA AUSTRALIALand region 70 − 90◦E 80 − 100◦E 100 − 110◦E 15◦W − 10◦E 110 − 100◦W 120 − 150◦E

5 − 30◦N 15 − 30◦N 25 − 30◦N 2 − 14◦N 20 − 30◦N 18 − 10◦SOcean region 65 − 78◦E 80 − 100◦E 80 − 100◦E 28◦W − 10◦E 120 − 110◦W 100 − 130◦E

5 − 30◦N 10 − 20◦N 10 − 20◦N 5◦S − 14◦N 20 − 30◦N 10 − 0◦SWind region 65 − 78◦E 80 − 100◦E 90 − 105◦E 15◦W − 10◦E 111 − 109◦W 100 − 130◦E

5 − 30◦N 15 − 30◦N 15 − 25◦N 2 − 9◦N 20 − 30◦N 10◦S − 0◦

Monsoon season JJA JJA JJA JAS JJA JFM

Addition of constant equatorial easterlies. The effect of a constantinflow of moist air leads to an addition of a constantwt to the windsin the heat balance and moisture balance equations

lp − w (w +wt) + r = 0 [S9]

(w +wt) (1 − p) − p = 0 [S10]

yielding the governing equation

w3+(2wt + 1)w2−(l + r − wt (wt + 1))w−((1 +wt) r + lwt) = 0[S11]

and a shift of the critical threshold towards lower radiation and pre-cipitation values (figure S7).

Addition of stabilizing radiative feedback. Adding a negative con-tribution −σ′TL to the heat balance may be used to parameterize astabilizing temperature feedback due to changes in long wave radia-tion. This addition transforms the non-dimensional heat balance into

lp − (w + σ)w − σtO + r = 0 [S12]

whereσ ≡ σ′/ (Cpβ) and tO ≡ TOαǫ/β is the nondimensionalatmospheric temperature over the ocean. The resulting governingequation

w3+(σ + 1)w2−(l + r − σ (σ + 1))w−(r + σtO) = 0 [S13]

resembles the corresponding relation with additional trade winds (fig-ure S8).

Addition of threshold for precipitation. Adding a threshold mois-ture valueqth to equation [4] above which precipitation is initi-ated does not change the governing equation after redefiningp ≡P/ ((qO − qth)β). It however changes the physical quantities. Crit-ical precipitation then reduces to zero when the threshold value ap-proachesqO.

1. Hansen J, et al. (1983) Efficient three-dimensional global models for clima te studies:Models I and II. Monthly Weather Review 111:609–662.

2 www.pnas.org/cgi/doi/10.1073/pnas.0709640104 Footline Author


Fig. S1. Difference in precipitation between seasons (JJA-DJF) and the differentmonsoon regions studied (black boxes).

Fig. S2. Different ocean (blue), land (dark gray) and wind (red box) regions forthe different monsoon systems as used for computation of the different quanti-ties used to motivate the conceptual model. Flow lines represent summer windsconnecting the ocean with the land region.

Footline Author PNAS Issue Date Volume Issue Number 3


8.5 9 9.5

6

7

8

9

q′ L (g/

kg)

India

10 10.5 11

9

10

11 Bay of Bengal

7.5 8 8.5

7

8

q′ L (g/

kg)

W. Africa

7.5 8 8.5 9

7

8

9 N. America

10 11 12

8

10

12

qL (g/kg)

q′ L (g/

kg)

China

6.5 7 7.5 8 8.5

8

9

10

11

qL (g/kg)

Australia

Fig. S3. Specific humidity q′L over land as computed by the model from time se-ries for precipitation, radiation, temperature difference and specific humidity overthe ocean versus observed mean specific humidity over land qL. The line givesthe unit function q′L = qL



6 81

2

3

4

P−

E (

mm

/day

)

INDIAr=0.98

γ=0.84

E0=−2.5

8 9 10 114

5

6

7 BAY OF BENGALr=0.99

γ=0.98

E0=−3.8

6 8 101

2

3

4

5

P−

E (

mm

/day

)

W.AFRICAr=0.99

γ=0.99

E0=−4.4

4 6 8

1

2

3

4 N.AMERICAr=0.98

γ=0.68

E0=−1.2

5 10

0

2

4

6

P (mm/day)

P−

E (

mm

/day

)

CHINAr=0.99

γ=1.2

E0=−6.2

5 10

0

2

4

6

P (mm/day)

AUSTRALIAr=0.99

γ=0.83

E0=−3.6

Fig. S4. Scaling of precipitation minus evaporation with precipitation in NCEP-NCAR reanalysis data.



−1

0

1γ=0

γ=0.9

Win

d w

−0.5 0 0.50

0.2

0.4

γ=0

γ=0.9

Radiation r

Pre

cipi

tatio

n p

e=0.2e=0

−0.5 0 0.5

e=0.2

e=0

Radiation rFig. S5. Change in solution structure due to inclusion of evaporation. Left: con-stant offset e = 0.2 without linear dependence on precipitation γ = 1 (equa-tion [S4]). Right: linearly dependent evaporation γ = 0.9 without constant offsete = 0.



−1

0

1

a=0.5a=0

Win

d w

−0.5 0 0.50

0.2

0.4

a=0.5

a=0

Radiation r

Pre

cipi

tatio

n p

Fig. S6. Change in solution structure due to inclusion of the cloud-albedo feed-back



−1

0

1

wT=0.5

wT=0

Win

d w

−0.5 0 0.50

0.2

0.4

wT=0.5

wT=0

Radiation r

Pre

cipi

tatio

n p

Fig. S7. Change in solution structure due to inclusion of an inflow of moistureand heat by constant trade winds



−1

0

1

σ=0

σ=0.5

Win

d w

−0.5 0 0.50

0.2

0.4

σ=0.5σ=0

Radiation r

Pre

cipi

tatio

n p

Fig. S8. Change in solution structure due to inclusion of a stabilizing long waveradiation feedback



2.3 A critical humidity threshold for monsoon failure

Manuscript prepared for Clim. Past Discuss.with version 2.2 of the LATEX class copernicus discussions.cls.Date: 4 May 2011

A critical humidity threshold for monsoontransitionsJacob Schewe1,2, Anders Levermann1,2, and Hai Cheng3,4

1Earth System Analysis, Potsdam Institute for Climate Impact Research, 14473 Potsdam,Germany2Institute of Physics, University of Potsdam, 14476 Potsdam, Germany3Institute of Global Environmental Change, Xi’an Jiaotong University, Xi’an 710049, China4Department of Geology and Geophysics, University of Minnesota, Minneapolis 55455, USA

Correspondence to: J. Schewe([email protected])

Abstract

Monsoon systems around the world are governed by the so–called moisture–advectionfeedback. Here we show that, in a minimal conceptual model, this feedback implies acritical threshold with respect to the atmospheric specific humidity qo over the oceanadjacent to the monsoon region. If qo falls short of this critical value qco, monsoon rainfallover land cannot be sustained. Such a case could occur if evaporation from the oceanwas reduced, e. g. due to low sea surface temperatures. Within the restrictions ofthe conceptual model, we estimate qco from present–day reanalysis data for four majormonsoon systems, and demonstrate how this concept can help understand abruptvariations in monsoon strength on orbital timescales as found in proxy records.

1

2.3 A critical humidity threshold for monsoon failure 51

1 Introduction

Monsoon rainfall is the major prerequisite of agricultural productivity in many tropicaland subtropical regions of the world, and its variability has been affecting the livelihoodsof a large share of the world’s population from ancient civilizations until today (e.g.Parthasarathy et al., 1988; Kumar et al., 2004; Auffhammer et al., 2006; Zhang et al.,2008; Rashid et al., 2011). Proxy records show evidence of abrupt and strong monsoonshifts during the last two glacial cycles (Burns et al., 2003; Wang et al., 2005a, 2008)and the Holocene (Gupta et al., 2003; Hong et al., 2003; Wang et al., 2005b; Rashidet al., 2011) in India, the Bay of Bengal, and East Asia. In many instances in thepast, periods of strong monsoon rainfall thus appear to have alternated with periods ofprolonged drought, with comparatively rapid transitions between the two.

Both spatial patterns and temporal evolution of continental monsoon rainfall are in-fluenced by a number of physical processes (Hahn and Shukla, 1976; Webster et al.,1998; Krishnamurthy and Goswami, 2000; Clark et al., 2000; Kucharski et al., 2006;Goswami et al., 2006; Goswami and Xavier, 2005; Dash et al., 2005; Ramanathanet al., 2005; Wang, 2005; Yang et al., 2007) as well as characteristics of vegetation(Meehl, 1994; Claussen, 1997; Robock et al., 2003) and topography (Liu and Yin,2002). Though these details are crucial for the specific behavior of different mon-soon systems, and their significance will vary from region to region, there exist definingprocesses fundamental to any monsoon dynamics (e.g. Webster, 1987a,b). Theseare the advection of heat and moisture during the monsoon season and the associ-ated rainfall and release of latent heat. While differential heating of land and oceanin spring is important for the initiation of the monsoon season, land surface tempera-tures drop substantially after the onset of heavy precipitation, diminishing the surfacetemperature gradient. The monsoon circulation over the continent is thereafter pre-dominantly sustained by the release of latent heat and subsequent warming of theatmospheric column over land (Webster et al., 1998). Using a complex conceptualmodel, Zickfeld et al. (2005) found that a monsoon circulation that is sustained by the

2


moisture-advection feedback can undergo abrupt changes in response to changes inthe land surface albedo. Knopf et al. (2006) however showed that the threshold albedoin this model is very far away from modern conditions and is highly uncertain due tothe dependence on various model parameters.

Here we apply a minimal conceptual model that comprises the heat and moisturebudgets of an idealized monsoon circulation. It reflects the dominant role of the self–amplifying moisture–advection feedback during the monsoon season, which is sup-ported by observations Levermann et al. (2009). We find that the model yields athreshold behaviour with respect to the atmospheric humidity over the ocean adja-cent to the monsoon region. Below the threshold, the advection and release of latentheat is not sufficient to sustain monsoon rainfall over land. It is important to note thatglobally, rainfall associated with the intertropical convergence zone will naturally besustained even without continental monsoon rainfall. Furthermore the seasonal rever-sal of cross-equatorial winds, driven by the seasonal change in hemispheric insolation,is not affected by our analysis. The question addressed here is which conditions arenecessary in order to sustain a rainy season over land which goes beyond the zonal–mean dynamics of the intertropical convergence zone. The basic dynamics capturedin our model thus form a necessary condition for continental monsoon rainfall to exist.

We describe the conceptual model in section 2 and analyze its implications in section3. In section 4, the critical threshold is estimated for four major monsoon regions.In section 5, we apply the model to abrupt monsoon changes on orbital timescales.Section 6 concludes.

2 Conceptual model

We use the minimal conceptual model presented by Levermann et al. (2009) (seeillustration in Fig. 1). It is based on the observation that, during the rainy season, theregional-scale moist static energy balance of the atmospheric column is dominated bylatent heating due to precipitation, which is balanced by advective as well as radiative

3


cooling. According to NCEP/NCAR reanalysis data (1948-2007; Kistler et al., 2001),this holds for all major monsoon regions (Levermann et al., 2009). The moist staticenergy balance can thus be approximated by

L·P −εCpW ·∆T +R= 0, (1)

where ∆T is the atmospheric temperature difference between land and ocean, andP is the mean precipitation over land (in kg/m2s). Latent heat of condensation isL =2.6·106 J/kg and volumetric heat capacity of air Cp =1,295 J/m3/K. The ratioε=H/L between vertical extent H of the lower troposphere and the horizontal scale Lof the region of precipitation enters due to the balance of the horizontal advective heattransport and the vertical fluxes of net radiative forcing R and precipitation P . Note thatthis is a model of the monsoon season only, and includes no annual cycle. The abovebalance therefore neglects the contribution from sensible heating, which is importantat the onset stage but relatively weak once heavy rainfall has started to cool the landsurface (Fig. 2). Consequently, this model does not capture any interseasonal or in-terannual dynamics. Equations are only valid for landward winds, W ≥ 0. That meansthat situations in which no solution of the model with positive W can be found will beconsidered as parameter and forcing combinations for which no monsoon rainfall canbe sustained.

Assuming dominance of ageostrophic flow in low latitudes, the landward mean windW is taken to be proportional to the temperature difference between land and ocean(Petoukhov, 1982; Webster, 1987a; Brovkin et al., 1998):

W =α ·∆T. (2)

Reanalysis data confirm that this is a valid first–order approximation (Fig. 3). Neglect-ing the effect of evaporation over land (which will be discussed below) and associatedsoil moisture processes, precipitation has to be balanced by the net landward flow ofmoisture

εW ·(qo−qL)−P = 0, (3)4


where qo and qL are specific humidity over ocean and over land, respectively. Con-sistent with reanalysis data (Fig. 4) and theoretical considerations (Petoukhov, 1982;Petoukhov et al., 2000), continental rainfall is assumed to be proportional to the meanspecific humidity within the atmospheric column:

P =βqL. (4)

Note that the proportionality constants α and β have explicit physical interpretations. αis essentially a function of the near-surface cross-isobar angle and thereby a function ofsurface roughness and static stability of the planetary boundary layer (PBL); β is gov-erned by the characteristic turnover (recycling) time of liquid water in the atmosphereand thereby determined by static stability and vertical velocity in the PBL (Petoukhovet al., 2000). While equations (1) to (4) are the basic relations necessary to capturethe moisture–advection feedback, eq. (4) can be made more realistic by consideringan offsets in qL, which will be discussed further below.

3 Critical qo threshold for monsoon existence

From equations (1), (3) and (4) it follows that

Lβqo =

(1+

β

ερW

)εCpW ·∆T −

(1+

β

ερW

)·R (5)

This equation represents the heat budget of the conceptual monsoon circulation interms of latent heat. The two terms on the r. h. s. represent the loss of heat from theland region by advection of warm air and by radiation, respectively (note that R< 0).Their sum must be balanced by latent heat as provided by the inflow of humid air fromthe ocean, namely Lβqo. The term (1+β/ερW ) incorporates the fact that the latentheat has to be transported from ocean to land by means of advection; the lower theadvective velocity W , the higher the specific humidity qo that is necessary to providethe required amount of latent heat to the atmospheric column over land.

5


Using equation (2) and the relations w≡Wερ/β; r≡R ·εαρ/(Cpβ2), and

l≡(εαρ2Lqo

)/(Cpβ), the non-dimensional form of eq. (5) is obtained:

l=(1+w−1

)·w2−

(1+w−1

)·r (6)

where l is proportional to qo. This is the governing equation of the conceptual model.Its solution is determined entirely by the only free parameter r. The physical part (l≥ 0,w≥ 0) of the solution of eq. (6) is shown in Fig. 5 for the case r=−0.05, where the thickred line denotes a stable solution and the thin red line an unstable one. The advective(dashed line) and radiative (dotted line) terms are also plotted separately to show howthe solution is obtained as the sum of these two contributions (to illustrate this, thefigure is organized with the control parameter l on the y-axis). In the case r≡ 0, onlythe advective part of the solution remains (i. e. the red line would collapse onto thedashed line). The y-axis in Fig. 5 can be interpreted as the demand in latent heatingthat results from the loss of heat from the land region due to radiation and advection.

It turns out that no physical solution exists below a critical threshold lc (horizontalline in Fig. 5), which corresponds to a critical value of specific humidity over the ocean,qco. When qo falls short of this value, the supply of moisture is not sufficient to maintainthe monsoon circulation driven by the moisture–advection feedback. No conventionalmonsoon circulation can thus develop in a climate where qo<qco.

Equation (6) can also be expressed in terms of non–dimensional precipitation p≡P/(qoβ), which is directly related to the wind through

p=w/(1+w). (7)

The solution in terms of p has a similar shape as in terms of w (Fig. 6), while dimen-sional precipitation P scales approximately linearly with l as long as l is sufficientlyabove the threshold lc (Fig. 7). While we do not expect to find this quasi-linear relationperfectly reflected in observations, NCEP/NCAR reanalysis data show that seasonalmean precipitation and specific humidity over ocean are correlated to some extent(Fig. 8).

6


4 Estimation of the critical threshold qco

The critical point [lc, wc] (or [qco, wc]) will vary for different monsoon systems. It isdetermined by the non–dimensional radiation r via

−w2c (2wc+1) = r. (8)

The critical l can then be computed from

lc = 2wc(wc+1)2 , (9)

and the critical humidity threshold qco via the definition of l. Within the limitations of thisminimal conceptual model, we estimate qco for four different monsoon regions. We useseasonal mean precipitation P , radiation R, land–ocean temperature difference ∆T ,and specific humidity over ocean qo from the NCEP/NCAR reanalysis to compute timeseries for α(t) = (LP +R)/

(εCp∆T

2), β(t) = ((LP +R) ·ρP )/((LP +R)qoρ−Cp∆T P ),

and r(t) =R·εα(t)/(Cpβ(t)2

), assuming applicability of the model and stationary statis-

tics within the observational period (1948-2007). Because the observational period issubject to significant anthropogenic global warming, we remove a linear trend from allreanalysis data to get a closer approximation of a stationary climate. From α(t), β(t)and r(t), qco can then be obtained for each year via equations (8), (9) and the defini-tion of l. Note that qco is independent of ε, the only quantity that is not constrained bydata. The resulting qco distribution (Fig. 9, blue) is much lower than the observed dis-tribution of qo (black) in the Bay of Bengal, West Africa and China, while in India thedistributions are closer. The blue pin marks the qco estimate that is obtained from thetime–mean values α(t), β(t) and R(t).

The spread in the distribution of qco is due to substantial variability in α(t), β(t) andr(t) throughout the reanalysis period. The interannual variability in the dimensional netradiation R is about 15–20% during the reanalysis period, depending on the region. Onlonger timescales, R can be expected to be rather stable because of the negative long–wave radiation feedback according to the Stefan–Boltzmann law. However, the factorsα and β may also vary over time. Moreover, in reality the basic relations of eq. (2) and

7


eq. (4) are blurred by higher–order physical processes that are not represented in ouridealized model, limiting our ability to determine α and β (cf. Fig. 3 and 4). Dependingon the relative importance of actual variability and observational uncertainty, the dis-tribution of qco can be interpreted either (i) as a noisy estimate of a stationary criticalthreshold, or (ii) as a probability distribution of an interannually varying threshold.

In order to obtain more realistic estimates of qco, we extend eq. (4) by an offset qoL thatterrestrial humidity qL needs to exceed before precipitation is initiated (as suggestedby the correlation in Fig. 4):

P =β(qL−qoL). (10)

After replacing qo by (qo−qoL) in the definitions of l and p, the non–dimensional equa-tions (6)–(9) remain unchanged.

As above, we estimate qco for this refined version of the model, obtaining the param-eter qoL from linear regression of the corresponding reanalysis data (Fig. 4). Note that,due to eq. (10), qco now also depends on ε. Since we have no direct observation of ε,we choose ε such that α(t) matches the α that we observe as the slope of the linearregression between W and ∆T (Fig. 3). The results for qco are shown in Fig. 9 (red),where again the pin marks the estimate from mean quantities. The consideration of theoffset qoL generally yields a distribution of qco which is narrower and closer to, while stillclearly seperate from, the present-day range of qo. Only for the Indian region, the distri-bution of qco overlaps with that of the observed qo; however, when considered pointwise,qco(t) is still lower than qo(t) for all years.

5 Application to past abrupt monsoon changes

Wang et al. (2008) presented a speleothem δ18O record from central China that testi-fies to several large and persistent changes in the strength of the East Asian summermonsoon (EASM) during the penultimate glacial period. These changes are in phase

8


with, but much more abrupt than, precession–dominated oscillations in northern hemi-sphere summer insolation (NHSI): While the latter follow a quasi-sinusoidal cycle, theform of the monsoon changes rather resembles that of a step–function, with variationsaround either a strong or a weak mean state, followed by a comparatively rapid transi-tion into the other state (cf. Fig. 2b in Wang et al. (2008)). This behaviour is especiallyapparent before about 160 kyr BP (Fig. 10, grey line) and suggests that non–linearprocesses inherent to the monsoon system might have amplified changes in externalforcing. In particular, the abrupt transitions might have been triggered by the mean in-solation crossing a certain threshold that separated two different states of the monsooncirculation.

Our conceptual monsoon model offers a simplified but robust mechanism to explainsuch sort of behaviour. It shows that the moisture–advection feedback implies a thresh-old qco that seperates two regimes: One where a conventional monsoon circulation canexist, and one where it cannot. We therefore speculate that orbital–timescale varia-tions in NHSI and the associated surface temperature changes might have affectedevaporation at the ocean surface such that average humidity over the ocean persis-tently crossed the threshold, thus critically altering the moisture supply for the adjacentmonsoon region and triggering a transition between the two regimes. In the followingwe apply our model to demonstrate how such variations in qo could have led to mon-soon variations consistent with those observed in the proxy record. In doing so, weassume that the values of α, β, ε, and qoL estimated for modern climate from reanalysisdata also hold for the penultimate glacial period; and that R was also comparable dur-ing that period to its modern value. In reality, R might have also varied in phase withNHSI, however this variation would have been damped by the stabilizing long–wave ra-diation feedback. Moreover, a variation of R along with NHSI would act to exacerbatethe threshold effect, moving the threshold towards higher values when insolation, andthus the inferred qo, is low (cf. Fig. 11). Therefore, neglecting variations in R yields aconservative result with respect to the volatility of the system.

The solution of the conceptual model depends on the non-dimensional net radiation

9


r. We choose r= r(t)−σ/2 = -106.5, where σ denotes a standard deviation (Fig. 11).This corresponds to a critical threshold qco = 8.0 g/kg, which is in the upper part of theestimated qco distribution for the China region (cf. Fig. 9). We further assume qo to varylinearly with NHSI (Fig. 12). The linear relation is chosen such that the maximum qois close to the range of present–day observations. Finally, we assume that when thethreshold qco is crossed from below (i. e., coming from a no–monsoon regime), it takesan additional increase ∆q to trigger the transition into the monsoon regime (Fig. 11).The EASM precipitation thus resulting from the conceptual model is shown in Fig. 10(red line). We set P to zero during periods when the model yields no physical solution,to illustrate the idea that no conventional monsoon circulation can exist during thoseperiods, and no rainfall associated with that circulation would occur. However we wouldexpect sources of rainfall other than the large–scale monsoon circulation to play a role,too, so that actual rainfall would not completely cease during such periods. Note thatneither the hysteresis width ∆q nor the second degree of freedom in the relation qo∝NHSI are constrained by data; instead they are chosen such that the result of theconceptual model matches the transition behaviour found in the proxy record, takinginto account dating errors in the latter (grey bars in Fig. 10).

6 Discussion and conclusions

We have shown that, in a minimal conceptual model of large–scale monsoon circula-tion, a critical threshold qco exists with respect to specific humidity over the ocean regionupwind of the continental monsoon region. This threshold follows from the central roleof the self–amplifying moisture–advection feedback, which governs the atmosphericMSE balance during the monsoon season. If qo falls short of the threshold qco, no con-ventional monsoon circulation can exist over land. The model neglects any processesthat are not crucial to the moisture–advection feedback, in order to isolate the conse-quences of this feedback. The basic dynamics captured in the model therefore forma necessary condition for the existence of continental monsoon rainfall beyond what

10


is accounted for by the zonal-mean dynamics of the intertropical convergence zone.Our results complement those of Levermann et al. (2009), who found a threshold withrespect to the net radiation R. While qo can generally be expected to be more volatilethan R, the model allows for a superposition of changes in both quantities, with theone either damping or amplifying the effect of the other, depending on the direction ofchange.

As the model contains the physical feedback that causes the threshold behaviour,it can be used to produce meaningful first–order estimates of the threshold values.Within the framework of the minimal model, we have estimated the critical threshold qcofor four major monsoon regions, using seasonally averaged reanalyses of regional pre-cipitation, net radiation, specific humidity, and temperature for the past sixty years. Theresulting distribution of qco can be interpreted either as a noisy estimate of a stationarycritical threshold, or as a probability distribution of an interannually varying threshold.The degree to which either of these interpretations is valid depends chiefly on the rel-ative importance of actual variability and observational uncertainty in the parametersα and β (see equations (2) and (4)), the assessment of which is beyond the scopeof this study. However we have seen that the consideration of an offset in terrestrialspecific humidity in eq. (4) leads to a qco distribution which is significantly narrower thanwith the basic version of the model, suggesting that at least some of the spread in qcocan be eliminated by making the model more realistic, and thus does not reflect actualyear–to–year variability in qco.

Consequently, relevant second–order physical processes would have to be includedinto the model in order to obtain more robust results for qco. Probably one of the mostimportant missing processes is evaporation over land (e. g. Eltahir, 1998). Its effecton the heat budget would be mainly to reduce sensible heat flux to the atmosphere,which we have already neglected (eq. (1)) because it is comparatively small during therainy season; on the other hand, its effect on the moisture budget (eq. (3)) would beto stabilize the monsoon regime by recycling a part of the atmospheric humidity that islost by precipitation. Therefore considering evaporation would tend to move the critical

11


threshold towards lower values of qo.While the estimation of qco could profit from a refinement of the model, the aim of

this study is to demonstrate how the simple concept that the model is based on canhelp understanding past monsoon variations. The non–linear solution structure of themodel can lead to abrupt changes in the modelled monsoon rainfall in response tosmooth changes in the control parameter, qo. Changes in qo could be brought aboutby various factors acting on different timescales. For instance, as wind speed over theoceans increases due to global warming (Young et al., 2011), evaporation e.g. in theArabian Sea could be affected both directly and via the amount of upwelling of coldwaters at the continental margins, and thereby alter the moisture supply for the Indiansummer monsoon. For the East Asian summer monsoon (EASM), we have shownthat, assuming variations in qo along orbital–timescale insolation changes, the modelyields a series of abrupt monsoon transitions similar to that observed in a proxy recordof the penultimate glacial period. While the additional assumption of a hysteresis is notcrucial for the transition behaviour, it changes the timing of the individual transitionssuch that they are all consistent, within dating errors, with those found in the proxyrecord (physically, a hysteresis might be induced by inert climate components suchas e.g. large–scale oceanic circulation or Himalayan glaciation, rather than by atmo-spheric processes). The idea of a threshold behaviour in monsoon circulations due tothe defining mechanism of the moisture–advection feedback may thus be a useful first–order concept for understanding past large–scale monsoon changes. The conceptualmodel investigated here may also serve as a basic building block that can be mademore realistic by the inclusion of other relevant processes and by a more detailed esti-mation of the model parameters. For a complete understanding of monsoon variationson multiple timescales, of course, more complex models will have to be invoked.

12


Appendix A

Methods

NCEP/NCAR reanalysis data has been obtained as monthly–mean time series (Jan-uary 1948 – December 2007), and regionally aggregated as indicated in Table 1. W isaveraged vertically between 850hPa and 1000hPa; qo between 600hPa and 1000hPa;qL between 400hPa and 1000hPa; and ∆T over the entire atmospheric column, asrepresented in the reanalysis data.

Acknowledgements. This work was funded by the Heinrich Boll Foundation, the German Na-tional Academic Foundation, and the BMBF PROGRESS project (support code 03IS2191B).NCEP Reanalysis Derived data was provided by the NOAA/OAR/ESRL PSD, Boulder, Col-orado, USA, from their Web site at http://www.esrl.noaa.gov/psd/.

References

Auffhammer, M., Ramanathan, V., and Vincent, J. R.: Integrated model shows that atmosphericbrown clouds and greenhouse gases have reduced rice harvests in India, Proceedings of theNational Academy of Sciences, 103, 19 668–19 672, 2006.

Brovkin, V., Claussen, M., Petoukhov, V., and Ganopolski, A.: On the stability of theatmosphere- vegetation system in the Sahara/Sahel region, Journal of Geophysical Re-search, 103, 31 613–31 624, 1998.

Burns, S. J., Fleitmann, D., Matter, A., Kramers, J., and Al-Subbary, A. A.: Indian Ocean Climateand an Absolute Chronology over Dansgaard/Oeschger Events 9 to 13, Science, 301, 1365–1367, 2003.

Clark, C. O., Cole, J. E., and Webster, P. J.: Indian Ocean SST and Indian summer rainfall:Predictive relationships and their decadal variability, Journal of Climate, 13, 2503–2519,2000.

Claussen, M.: Modeling bio-geophysical feedback in the African and Indian monsoon region,Climate Dynamics, 54, 247–257, 1997.

13


Dash, S. K., Singh, G. P., Shekhar, M. S., and Vernekar, A. D.: Response of the Indian summermonsoon circulation and rainfall to seasonal snow depth anomaly over Eurasia, ClimateDynamics, 24, 1–10, 2005.

Eltahir, E. A. B.: A soil moisture-rainfall feedback mechanism: 1. Theory and observations,Water Resources Research, 34, 765–776, 1998.

Goswami, B. N. and Xavier, P. K.: ENSO control on the south Asian monsoon through thelength of the rainy season, Geophysical Research Letters, 32, L18 717, 2005.

Goswami, B. N., Madhusoodanan, M. S., Neema, C. P., and Sengupta, D.: A physical mech-anism for North Atlantic SST influence on the Indian summer monsoon, Geophysical Re-search Letters, 33, L02 706, 2006.

Gupta, A. K., Anderson, D. M., and Overpeck, J. T.: Abrupt changes in the Asian southwestmonsoon during the Holocene and their links to the North Atlantic Ocean, Nature, 421, 354–357, 2003.

Hahn, D. G. and Shukla, J.: An apparent relationship between Eurasian snow cover and Indianmonsoon rainfall, Journal of Atmospheric Sciences, 33, 2461–2462, 1976.

Hong, Y., Hong, B., Lin, Q., Zhu, Y., Shibata, Y., Hirota, M., Uchida, M., Leng, X., Jiang, H., Xu,H., Wang, H., and Yi, L.: Correlation between Indian Ocean summer monsoon and NorthAtlantic climate during the Holocene, Earth and Planetary Science Letters, 211, 371–380,doi:10.1016/S0012-821X(03)00207-3, 2003.

Kistler, R., Kalnay, E., Saha, S., White, G., Woollen, J., Chelliah, M., Ebisuzaki, W., Kana-mitsu, M., Kousky, V., den Dool, H. V., Jenne, R., and Fiorino, M.: The NCEP/NCAR 50-yearreanalysis, Bull. Amer. Meteor. Soc., 82, 247 – 267, 2001.

Knopf, B., Flechsig, M., and Zickfeld, K.: Multi-parameter uncertainty analysis of a bifurcationpoint, Nonlinear Processes in Geophysics, 13, 531–540, 2006.

Krishnamurthy, V. and Goswami, B. N.: Indian monsoon-ENSO relationship on inter-decadaltimescale, Journal of Climate, 13, 579–595, 2000.

Kucharski, F., Molteni, F., and Yoo, J. H.: SST forcing of decadal Indian monsoon rainfall vari-ability, Geophysical Research Letters, 33, L03 709, 2006.

Kumar, K. K., Kumar, K. R., Ashrit, R. G., Deshpande, N. R., and Hansen, J. W.: Climateimpacts on Indian agriculture, International Journal of Climatology, 24, 1375–1393, doi:10.1002/joc.1081, 2004.

Levermann, A., Schewe, J., Petoukhov, V., and Held, H.: Basic mechanism for abrupt monsoontransitions, Proceedings of the National Academy of Sciences, 106, 20 572–20 577, 2009.

14


Liu, X. and Yin, Z.: Sensitivity of East Asian monsoon climate to the uplift of the Tibetan Plateau,Palaeogeography, Palaeoclimatology, Palaeoecology, 183, 223–245, 2002.

Meehl, G. A.: Influence of the Land Surface in the Asian Summer Monsoon: External Condi-tions versus Internal Feedbacks, Journal of Climate, 7, 1033–1049, 1994.

Parthasarathy, B., Munot, A., and Kothawale, D.: Regression model for estimation of Indianfoodgrain production from summer monsoon rainfall, Agricultural and Forest Meteorology,42, 167 – 182, doi:10.1016/0168-1923(88)90075-5, 1988.

Petoukhov, V., Ganopolski, A., Brovkin, V., Claussen, M., Eliseev, A., Kubatzki, C., and Rahm-storf, S.: CLIMBER-2: a climate system model of intermediate complexity. Part I: modeldescription and performance for present climate, Climate Dynamics, 16, 1, 2000.

Petoukhov, V. K.: Two mechanisms of temperature oscillations in a thermodynamical modelof the troposphere-stratosphere system, Izvestiya, Atmospheric and Oceanic Physics, 18,126–137, 1982.

Ramanathan, V., Chung, C., Kim, D., Bettge, T., Kiehl, J. T., Washington, W. M., Fu, Q., Sikka,D. R., and Wild, M.: Atmospheric brown clouds: Impacts on South Asian climate and hydro-logical cycle, Proceedings of the National Academy of Sciences, 102, 5326–5333, 2005.

Rashid, H., England, E., Thompson, L., and Polyak, L.: Late Glacial to Holocene Indian Sum-mer Monsoon Variability Based upon Sediment Records Taken from the Bay of Bengal,Terrestrial Atmospheric and Oceanic Sciences, 22, 215–228, doi:10.3319/TAO.2010.09.17.02(TibXS), 2011.

Robock, A., Mu, M., Vinnikov, K., and Robinson, D.: Land surface conditions over Eurasia andIndian summer monsoon rainfall, Journal of Geophysical Research, 108, 4131, 2003.

Wang, B.: The Asian monsoon, Springer-Verlag, 2005.Wang, P., Clemens, S., Beaufort, L., Braconnot, P., Ganssen, G., Jian, Z., Kershaw, P., and

Sarnthein, M.: Evolution and variability of the Asian monsoon system: state of the art andoutstanding issues, Quaternary Science Reviews, 24, 595–629, 2005a.

Wang, Y., Cheng, H., Edwards, R. L., He, Y., Kong, X., An, Z., Wu, J., Kelly, M. J., Dykoski,C. A., and Li, X.: The Holocene Asian Monsoon: Links to Solar Changes and North AtlanticClimate, Science, 308, 854–857, 2005b.

Wang, Y., Cheng, H., Edwards, R. L., Kong, X., Shao, X., Chen, S., Wu, J., Jiang, X., Wang, X.,and An, Z.: Millennial- and orbital-scale changes in the East Asian monsoon over the past224,000 years, Nature, 451, 1090–1093, 2008.

Webster, P. J.: The Elementary Monsoon, in: Monsoons, edited by Fein, J. S. and Stephens,

15


P. L., pp. 3–32, John Wiley, New York, N.Y., 1987a.Webster, P. J.: The Variable and Interactive Monsoon, in: Monsoons, edited by Fein, J. S. and

Stephens, P. L., pp. 269–330, John Wiley, New York, N.Y., 1987b.Webster, P. J., Magana, V. O., Palmer, T. N., Shukla, J., Tomas, R. A., Yanai, M., and Ya-

sunari, T.: Monsoons: Processes, predictability, and the prospects for prediction, Journal ofGeophysical Research, 103, 14,451–14,510, 1998.

Yang, J., Liu, Q., Xie, S.-P., Liu, Z., and Wu, L.: Impact of the Indian Ocean SST basin modeon the Asian summer monsoon, Geophysical Research Letters, 34, L02 708, 2007.

Young, I. R., Zieger, S., and Babanin, A. V.: Global Trends in Wind Speed and Wave Height,Science, 332, 451–455, doi:10.1126/science.1197219, 2011.

Zhang, P., Cheng, H., Edwards, R. L., Chen, F., Wang, Y., Yang, X., Liu, J., Tan, M., Wang, X.,Liu, J., An, C., Dai, Z., Zhou, J., Zhang, D., Jia, J., and Johnson, K. R.: A Test of Climate,Sun, and Culture Relationships from an 1810-Year Chinese Cave Record, Science, 322,940–942, 2008.

Zickfeld, K., Knopf, B., Petoukhov, V., and Schellnhuber, H. J.: Is the Indian summer monsoonstable against global change?, Geophysical Research Letters, 32, L15 707, 2005.

16


W

P

R

Ocean Land

W

qLqo

Fig. 1. Geometry of the conceptual model, illustrating wind W , precipitation P , net radiativeflux R, and atmospheric specific humidity over land (qL) and ocean (qo).

17


2 4 6 8 10 12

−200

−100

0

100

200

300

W/m

2

Latent

Sensible

RadiativeConvergence

INDIA

2 4 6 8 10 12

−200

−100

0

100

200

300 Latent

Sensible

RadiativeConverg.

BAY OF BENGAL

2 4 6 8 10 12

−200

−100

0

100

200

300

Month

W/m

2

Latent

Sensible

Radiative Convergence

W.AFRICA

2 4 6 8 10 12

−200

−100

0

100

200

300

Month

Sensible

Latent


CHINA

Fig. 2. Seasonal heat flux contributions to the atmospheric column over four major continentalmonsoon regions in NCEP/NCAR reanalysis data (Kistler et al., 2001). Radiative heating ofthe land surface in spring enhances sensible heat flux from the ground (’Sensible’). Duringthe rainy season, latent heat release dominates the heat budget (’Latent’). Radiative heat fluxcomprises all radiative fluxes in and out of the atmospheric column (’Radiative’). The excessheat is transported out of the continental monsoon region through large–scale advective andsynoptic processes (’Convergence’). Error bars give the standard deviation from the reanalysisperiod (1948-2007). See Table 1 for the geographical definitions of the monsoon regions. Thered and blue vertical lines emphasize the months of maximum sensible heat flux and latentheat flux, respectively.

18


1.8 2 2.2 2.4

7

7.5

8

8.5

W (

m/s

)

INDIA

r=0.49

α=3.62

1.6 1.8 2 2.22.5

3

3.5

4

4.5BAY OF BENGAL

r=0.57

α=1.85

0 0.2 0.4 0.60

2

4

6

Δ T (K)

W (

m/s

)

W.AFRICAr=0.44

α=9.64

0.5 1 1.5

2

3

4

Δ T (K)

CHINAr=0.63

α=3.00

Fig. 3. Wind W versus temperature difference between land and ocean region, ∆T , fromNCEP/NCAR reanalysis data, for the major monsoon regions of India, the Bay of Bengal, WestAfrica, and China (East Asia; see Table 1). The correlation coefficient r is indicated, as well asthe slope α of a linear fit through the origin (black line).

19


8 9 10

5

6

7

8

P (

mm

/day

) INDIA

r=0.83q

Lo=6.27

β=0.0265

10 10.5 11

8

9

10

11B. OF BENGAL

r=0.65

qLo=5.52

β=0.0218

7.5 8 8.5

6

7

8

9

qL (g/kg)

P (

mm

/day

) W.AFRICAr=0.63

qLo=4.58

β=0.0271

9 107

8

9

10

11

qL (g/kg)

CHINAr=0.61q

Lo=4.73

β=0.0205

Fig. 4. Precipitation P versus specific humidity over land, qL, from NCEP/NCAR reanalysisdata. The black line shows the result of a linear regression, the correlation coefficient r isindicated, as well as the slope β (in kg/m2s) and the offset in terrestrial humidity, qoL (in g/kg).

20


0 0.2 0.4 0.6 0.80

0.2

0.4

0.6

0.8

1

lc

Wind w

para

met

er l

∝ q

o

advectionradiationresulting latent heat demand

Fig. 5. Solution structure of the conceptual model as a function of the non–dimensional param-eter l, which is proportional to specific humidity over the ocean, qo. For illustrative purposes, lis plotted on the y-axis. The latent heat demand (red line; bold part indicates the stable branch)results from heat loss due to net radiative flux (dotted) and advection of warm air out of the landregion (dashed).

21


0 0.2 0.4 0.6 0.8 10

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

lc

Pre

cipi

tatio

n p

parameter l ∝ qo

Fig. 6. As Fig. 5 (red line), but in terms of non–dimensional precipitation p, and organized withthe control parameter l on the x-axis.

22


0 0.2 0.4 0.6 0.8 1

lc

P (

mm

/day

)

parameter l ∝ qo

Fig. 7. As Fig. 6, but in terms of dimensional precipitation P . The shape of the solution isdifferent than in terms of p because the relation between p and P depends on l. Units on they-axis are arbitrary.

23


9 9.5 10 10.5

5

6

7

8

P (

mm

/day

) INDIAr=0.63

10 10.5 118

9

10

11BAY OF BENGALr=0.47

8 8.5 9 9.55

6

7

8

9

qO

(g/kg)

P (

mm

/day

) W.AFRICAr=0.46

10 11 127

8

9

10

11

qO

(g/kg)

CHINAr=0.51

Fig. 8. Correlation between precipitation and specific humidity over the ocean fromNCEP/NCAR reanalysis data. Black lines show the result of a linear regression, the corre-lation coefficients are indicated.

24


0 0.005 0.010

10

20

30

40

qo (kg/kg)

India

0 0.005 0.010

10

20

30

40

50

qo (kg/kg)

Bay of Bengal

0 0.005 0.010

10

20

30

40

qo (kg/kg)

W.Africa

0 0.005 0.010

10

20

30

40

qo (kg/kg)

China

Fig. 9. Estimate of critical specific humidity value over the ocean, qco, from NCEP/NCAR reanal-ysis data for the basic minimal model (blue) and including the effect of a minimum terrestrialhumidity qoL required for the onset of precipitation (red). The black histogram shows the ob-served distribution of qo. Pins mark the estimates obtained from time–mean parameter values.

25


160 170 180 190 200 210 220

0

5

P (

mm

/day

)

time (kyr BP)160 170 180 190 200 210 220

−11

−9

−7

δ 18

O (

0 / 00, V

PD

B)

Fig. 10. Grey: Speleothem δ18O record from central China, used as EASM proxy for thepenultimate glacial period, where more negative values indicate stronger rainfall (Wang et al.,2008). The record is smoothed with a 5–point running average, and dating errors (± 2σ)are shown for selected dates (grey horizontal bars). Red: Result of the conceptual modelfor EASM precipitation P in response to qo variations driven by northern hemisphere (65◦N)summer insolation, assuming a hysteresis of 0.5 g/kg width. For illustration, we set P to zeroduring periods when no monsoon circulation exists according to the model.

26


6 8 10 12

x 10−3

0

2

4

6

8

10

r=−106.5

r=−83.8

r=−61.0

humidity over ocean q0 (kg/kg)

P (

mm

/day

)

Fig. 11. Physical solution for EASM precipitation P from the conceptual model, for three differ-ent values of the non–dimensional parameter r: r(t) (dashed), r(t)+σ/2 (dot–and–dash), andr(t)−σ/2 (solid), where r(t) is the estimate from NCEP/NCAR reanalysis data, and σ denotes astandard deviation. Bold lines indicate the upper, stable branch. The value r(t)−σ/2 = -106.5,which corresponds to a critical threshold qco = 8.0 g/kg, is used for the comparison of the modelwith EASM proxy data. Vertical dashed lines mark qco and qco +∆q, where we choose ∆q = 0.5g/kg as the width of an assumed hysteresis that is thought to appear when the threshold iscrossed from either side (illustrated by arrows).

27


160 170 180 190 200 210 2206.5

7

7.5

8

8.5

9

9.5

10x 10

−3

time (kyr BP)

q o (kg

/kg)

Fig. 12. Time series of qo used for the application of the conceptual model. qo is assumed tovary linearly with northern hemisphere (65◦N) summer insolation. The horizontal dashed linemarks qco.

28


Table 1. Regional definitions used for data analysis. Monthly–mean NCEP/NCAR reanalysisdata has been averaged over the indicated regions and seasons; Land and Ocean indicate thatonly terrestrial or oceanic grid points have been considered, respectively; and ∆T = TL−To.The bottom row lists the values for the dimensionless parameter ε that have been used in theestimation of the critical threshold (see section 4).

Quantity INDIA BAY OF BENGAL W.AFRICA CHINA (EASM)P , R, qL, TL (Land) 70-90◦E 80-100◦E 15◦W-10◦E 100-120◦E

5-30◦N 15-30◦N 2-14◦N 20-32◦Nqo, To (Ocean) 65-78◦E 80-100◦E 35-15◦W 105-115◦E

5-30◦N 10-20◦N 2-14◦N 15-25◦NW 65-78◦E 80-100◦E 15◦W-10◦E 100-120◦E

5-30◦N 15-30◦N 2-9◦N 20-32◦NMonsoon season June–Aug. June–Aug. July–Sep. June–Aug.ε 4.5·10−3 2.3·10−2 2.0·10−1 6.7·10−2

29


2.4 More frequent future monsoon failure due to inherent instability

More frequent future monsoon failure due to inherent in-stability

Jacob Schewe1,2 & Anders Levermann1,2

1Earth System Analysis, Potsdam Institute for Climate Impact Research, Potsdam, Germany

2Institute of Physics, Potsdam University, Potsdam, Germany

Indian summer monsoon rainfall is vital for a large share of the world’s population. Both

reliably projecting India’s future rainfall and unraveling abrupt monsoon shifts found in

paleo–records require improved understanding of its stability properties. Here we project

monsoon failure to become considerably more frequent due to global warming, and provide

a simple dynamical explanation for this trend as well as for multi–decadal rainfall variability.

Based on fundamental properties of observed monsoon dynamics and an associated inherent

instability that is modulated by ambient climate merely during the onset period, we develop

a statistically predictive model of seasonal rainfall. Forced only by global mean temperature

and central-Pacific sea level pressure anomalies in May, this simple model reproduces past

and future monsoon trends as found in a comprehensive climate model. We thereby propose

a novel perspective on monsoon variability as the result of internal instabilities modulated by

pre–seasonal ambient climate conditions.

Indian summer monsoon (ISM) rainfall is the major prerequisite of agricultural productiv-

ity in the region, and its variability severely affects the livelihoods of a large share of the world’s

population1, 2. While average ISM rainfall has been relatively stable during the past century of di-

1

2.4 More frequent future monsoon failure due to inherent instability 83

rect observations, rising trends have been observed in the annual number of extreme rain events3.

The future evolution of the ISM, and other monsoon systems, under a combination of anthro-

pogenic forcing factors is unclear4: Recent projections indicate that the response to increased

greenhouse gas (GHG) concentrations may differ in sign among major monsoon regions, and re-

veal large uncertainties about the magnitude of the response5–7. The effect of increased aerosol

abundance is significant and may be counteracting that of GHGs8, 9, while human-induced veg-

etation changes feed back on precipitation10, 11. Observations and modelling studies suggest a

recent regime shift in Asian monsoon convection12 and its relation to northern hemisphere thermal

gradients13, 14. At the same time, paleodata testify to abrupt and strong shifts in both the Indian and

the East Asian monsoon during the last two glacial cycles15–17 and the Holocene18, 19.

Fundamental monsoon dynamics: Persistence and self-amplification

These uncertainties call for an understanding of the internal monsoon dynamics to project their

apparently non-linear response to external perturbations. In order to assess large-scale failure of

seasonal rainfall, we focus here on fundamental monsoon dynamics which are universal across the

different regions: The onset of monsoon precipitation in spring is initiated by the development of

a tropospheric temperature contrast between land and ocean, and associated convergence of moist

air over the continent. During the rainy season, the atmospheric heat budget in the monsoon region

(cf. Supplementary Fig. S1) is then dominated by the so-called moisture-advection feedback: The

release of latent heat by precipitation enhances the sea-level pressure gradient between land and

ocean, thereby stabilizing the circulation that brings in more moist air from the ocean, which in

2


turn maintains precipitation (Fig. 1A and B, top). In conceptual models, this self-amplification

yields a threshold behaviour which can lead to a permanent cessation of the monsoon circulation

in response to shifts in external forcing20, 21.

Unless the threshold is crossed, however, the moisture-advection feedback stabilizes the

monsoon circulation by adding inertia to the system: Rainfall over a certain period within the

rainy season releases latent heat, thereby reinforces the circulation and increases the probability

for rainfall in subsequent days. This “memory effect”, or persistence, can be seen in direct ob-

servations, where it is reflected as a quasi-proportionality between the expectation value of daily

rainfall and the amount of latent heating accumulated during some period prior to the specific day.

Here we use daily rainfall data for 1951-2003 on a 1◦ grid from India Meteorological Department

(IMD)22, averaged over all available data within 5-30◦N, 70-90◦E. For the summer months (May-

September), we plot each day’s precipitation against the normalized average rainfall p during a

memory period τ prior to that day, such that at day t,

pt ≡〈P 〉t−1t−τ − P−P+ − P−

, (1)

where P is the daily precipitation rate; 〈...〉 indicate temporal averaging; and P+ and P− are max-

imum and minimum precipitation rates, respectively (see Methods section). The memory effect is

reflected in the correlation between Pt and pt for all days within the season, shown in Fig. 2A for

τ = 17 days. The correlation depends on the choice of τ , and a sensitivity test (Supplementary

Fig. S2) reveals that the rainfall memory is effective on a timescale of 2-3 weeks; after that time,

the information on the rainfall history is lost from the system. In particular, the signal-to-noise

ratio of the correlation is greater than 1 for τ ≤ 18 days. We interpret p as a probability for daily

3


precipitation.

Although the memory effect can be seen clearly in the IMD data, the observational period is

not long enough to probe the full influence of this effect on the statistics of seasonal mean precip-

itation, especially with respect to extreme events. Instead, we use an ensemble of five millennial

climate simulations performed with the comprehensive Earth system model COSMOS developed

at MPI-M Hamburg23. The model comprises the atmospheric general circulation model (GCM)

ECHAM5 at T31 (∼ 3.75◦) horizontal resolution, as well as an oceanic GCM, MPI-OM, and mod-

ules for terrestrial vegetation and ocean biogeochemistry. Comparison with other GCMs and with

observational data shows that ECHAM5 reproduces the Indian summer monsoon realistically24, 25.

The five simulations each span the period 800-2005 CE and do not differ in forcing. While they

include the past century of anthropogenic GHG emissions, we consider the total of 1206 years

sufficiently homogenous to be treated as belonging to the same climate state. Using the same value

τ = 17 days as above, and averaging daily rainfall over 5-30◦N, 70-90◦E, we find the memory ef-

fect well represented in the COSMOS output (Fig. 2B, inset). This confirms not only the presence

of the moisture-advection feedback in the model, but also the approximate timescale on which the

feedback operates.

Additional feedback mechanism for sustained dry–state

At the same time, seasonal (June-August) mean rainfall over the total of 6030 model years shows a

characteristic frequency distribution (Fig. 2B, gray bars) which extends far towards the lower end

4


of the spectrum, featuring very weak monsoon years with seasonal mean rainfall below 2 mm/day.

Latent–classes analysis suggests that this characteristic distribution represents a superposition of

two monsoon modes (Supplementary Fig. S3). While the rainy mode is dominated by the moisture-

advection feedback, the dry mode is determined by a second dynamical feedback. In order to

identify this second feedback, we investigate year 1158 of the first ensemble member as one of the

driest monsoon years within the entire simulation ensemble (due to significant internal variability

in this class of climate models, this year cannot be expected to coincide with any specific historical

year, and will henceforth be referred to as the ‘dry year’). Average seasonal precipitation over India

in this year is ∼ 70% below the long-term mean of 5.7 mm/day (Fig. 3A, circles). The dynamical

South Asian monsoon index26 indicates an exceptionally large deviation of the zonal tropospheric

wind shear from the long-term mean (Fig. 3A, bars).

The dry year’s rainfall cycle (Fig. 3B) shows that large-scale monsoon precipitation is erratic

and largely suppressed during the entire summer, caused by a qualitatively different circulation

pattern compared to normal monsoon years: The initial lack of latent heating over India leaves

the regional-scale sea level pressure anomalously high throughout the summer (Fig. 3C and map

in Supplementary Fig. S4). This causes subsidence of upper-tropospheric air both over the sub-

continent and over the Arabian sea (Fig. 3D and map in Supplementary Fig. S5). Since specific

humidity is lower in the upper troposphere, this large-scale subsidence effectively dries out the

monsoon region: Specific humidity in the lower troposphere (i.e. in the lower, landward branch

of the ISM circulation, Fig. 1A) substantially decreases over India and the eastern Arabian Sea,

and remains exceptionally low through June-August (Fig. 3E and map in Supplementary Fig. S6;

5


also note the spatial congruence of subsidence and moisture anomalies in Fig. S6). This acts to

further suppress precipitation and sustain the dry state, as illustrated in Fig. 1A (bottom). Note that

the dry regime is associated with a qualitatively different circulation pattern compared to the wet

regime: Instead of convective upwelling, the monsoon region is characterized by subsidence of

dry air from the upper troposphere; also, the flow direction in the upper troposphere changes from

generally westward to northward or even eastward, which is consistent with an anomalously low

zonal wind shear as reflected by the dynamical monsoon index (Fig. 3A, bars). Once initiated, this

self-amplifying feedback sustains the dry state in a similar way as the wet state is sustained by the

moisture-advection feedback (Fig. 1B).

A model of inherent instability

The dry year examined here is an extreme case in the sense that the dry feedback is dominant

throughout the season. Most years in the spectrum of the climate model (Fig. 2B, gray bars) are

less extreme, i.e. their seasonal precipitation averages somewhere between the dry state and an

upper limit of about 9 mm/day. We propose a simple statistically predictive model for seasonal

mean monsoon precipitation that explains this characteristic spectrum by an interplay between

the wet moisture-advection feedback and the dry subsidence feedback on a quasi-daily timescale.

This “day-to-day” model (illustration Fig. 1C) is based only on the two feedbacks and the memory

effect. It employs the idealized assumption of two discrete circulation states: One with weak

precipitation (we choose P− ≡ 0 mm/day in our example, Supplementary Table S1), corresponding

to the dry feedback loop in Fig. 1B, and one with maximum precipitation (P+, here chosen to be 9

6


mm/day), corresponding to the wet feedback loop in Fig. 1B. It is further assumed that the system

can flip between these two states on a daily timescale, and that the probability pt for ending up

in the wet state at time step (day) t depends on the ratio of dry and rainy days during a certain

period τ prior to t - this corresponds to the precipitation probability p of equation (1): That is, the

more rainy days there were in the period τ before t, the more likely it is to have rainfall at time

t. Thus here the memory effect is represented by an exact proportionality between rainfall and

rainfall history. It acts to impede a flip between the two circulation states. Still, a flip can always

be triggered by stochastic atmospheric fluctuations, e.g. regional and local weather, that perturb

the large-scale dynamical flows comprised in the two feedback mechanisms. In the day-to-day

model, these perturbations are represented by an unbiased random process that, together with pt,

ultimately determines the system state at each time step.

This very basic auto-regression model of internal monsoon dynamics needs to be comple-

mented by information on the onset of the rainy season, prior to the dominance of the two positive

feedback loops. During the period of length τ at the beginning of the season, a constant initial

probability for rainfall, pinit, replaces the dynamical rainfall probability p. pinit represents external

factors that crucially influence the development of the monsoon circulation at a time when it is

most sensitive to external perturbations26. In the case of India this could for example comprise the

role of El Nino/Southern Oscillation (ENSO), Indian Ocean sea surface temperatures, or Eurasian

snow cover27, as well as the current phase of the Madden-Julian Oscillation (MJO)28. Finally,

we introduce a maximum probability pm, such that pt cannot be greater than pm or smaller than

(1− pm). This prevents locking into either P− or P+ when pt becomes zero or one, which oth-

7


erwise could occur as an artifact of the discrete nature of the model. In this form, the day-to-day

model can reproduce the spectrum of the comprehensive climate model (Fig. 2B, blue line). The

Matlab R© code is provided as Supplementary Information.

Monsoon failure under global warming

The COSMOS climate model simulations of the past, as used above, have been complemented

with IPCC SRES A1B scenario29 simulations until 2100, followed by constant CO2 concentrations

until 220030. This results in an increase in global mean surface air temperature of 4.6◦C relative to

pre–industrial by the end of the 22nd century (Fig. 4A, black line). We find that in the warm climate

of 2151-2200 CE, the frequency distribution of seasonal mean ISM rainfall is completely inverted,

in the sense that now dry years are much more frequent than wet years (Fig. 2C, gray bars). This

inversion occurs gradually over the course of the warming scenario and involves a sign change in

the distribution’s skewness at the end of the 21st century (Fig. 4B, red line). The associated change

in the expected seasonal mean rainfall, however, is not monotonic: An increase throughout much

of the 21st century is followed by a rapid decrease, falling short of the pre–industrial long–term

mean roughly by the turn of the 22nd century (Fig. 4C, red line).

These changes can be understood in the simple framework of the day-to-day model. A

warmer tropical troposphere can hold more water vapor, which tends to enhance rainfall in the

absence of counter–acting processes31. We translate this into a linear shift of the precipitation

range [P−, P+] towards higher values with increasing global mean temperature. In addition to this

8


thermodynamical effect, dynamical changes that affect the monsoon development during the onset

period can be represented by the day-to-day model via the parameter pinit. In order to understand

which process dominates the initiation of monsoon rainfall in the comprehensive climate model,

consider the dry year investigated above. Both during April and May, mean sea level pressure

(MSLP) was extremely low over the central Pacific Nino3.4 region (170-120◦W, 5◦S-5◦N; Supple-

mentary Fig. S7), which tends to suppress the development of the low pressure system over India,

represented by a low pinit value in the day-to-day-model. This relation can be found throughout

the 6030 years of historical climate simulations: There is a clear correlation between the two mon-

soon modes present in the distribution of seasonal mean rainfall and anomalous spring–time (May)

Nino3.4 MSLP (Fig. 2B, red bars; see also Supplementary Fig. S8). As a simple approximation,

we scale pinit linearly with May Nino3.4 MSLP.

Forced with decadal averages of global mean temperature and May Nino3.4 MSLP, the day-

to-day model reproduces the long–term trends in mean and skewness of the ISM rainfall frequency

distribution observed in COSMOS (Fig. 4B and C, gray lines and shading). The non–trivial peak–

and–decline response in ISM rainfall is captured due to the delay in the Pacific MSLP signal

compared to the warming signal (Fig. 4A). Moreover, a significant portion of multidecadal rainfall

variability is reproduced (Fig. 4C). Similar to the historical period, the model can explain the

frequency distribution of the future period 2151-2200 CE (Fig. 2C). Note that the variation of pinit

along with decadally–averaged spring–time Nino3.4 MSLP represents the influence of a slowly

varying central Pacific mean climate on the ISM onset, rather than a direct forcing of the ISM

season by ENSO.

9


Discussion and conclusions

We conclude that major characteristics of seasonal–mean monsoon rainfall can be explained on the

basis of only two fundamental assumptions: (i) An inherent instability of the monsoon circulation

on a daily timescale, in the form of a competition between two different circulation regimes, each

of which is associated with a self-amplifying feedback. (ii) A memory effect that is induced by

those feedbacks, and determines, in a probabilistic manner, the transition between the wet and

the dry periods. Both assumptions are backed by observations and comprehensive climate model

results. The subsidence feedback not only complements the picture of the atmospheric circulation

in the temporary absence of monsoon rainfall, but constitutes an active antagonist of the moisture-

advection feedback due to its self-amplifying nature. The memory effect is a logical consequence

of the dominant role of these two feedbacks, and is found in observational data as well as in climate

model results.

Monsoon rainfall exhibits intraseasonal variability on various time scales which influences

the seasonal mean. While some of this variability, in particular its slower components, is related to

interaction with larger-scale atmospheric phenomena such as the MJO28, internally generated vari-

ability is an equally important factor in shaping the monsoon cycle and determining the seasonal

mean rainfall32, 33. The notion of an internal instability of the large-scale monsoon circulation on

a quasi–daily timescale, incorporated here in a slim and fundamental framework, offers a first–

order explanation for the driving mechanism of internally generated intraseasonal variability. We

have demonstrated that this mechanism alone can explain the non–trivial, long–term frequency

10


distribution of seasonal mean monsoon rainfall in a complex climate model.

Moreover, it can explain ISM changes on decadal to centennial time scales with a minimal

amount of external information. We have shown that the “day-to-day model” can reproduce past

ISM multidecadal variability as well as projected future changes, including a substantial increase

in monsoon failure, when driven by the decadal changes in global mean temperature and tropical

Pacific MSLP in May. While global warming is assumed to elevate the baseline of monsoon rain-

fall due to increased atmospheric moisture content, a shift towards lower spring–time MSLP in

the tropical Pacific is assumed to induce atmospheric conditions that favor more subsidence over

the Indian region and thus lead to more deficient monsoon onsets. In the day-to-day model, this

translates into a lower initial probability pinit. This parameter summarizes the influence of ambient

climatic factors that, to some extent, predispose the monsoon system during the onset season34. It

does not determine the rainfall amount of an individual monsoon season, but it has strong influ-

ence on the probability distribution of seasonal mean rainfall. While in the comprehensive climate

model applied here, pinit is dominated by the influence of central Pacific climate variations, other

factors such as changing spring–time Indian Ocean sea surface temperatures might play a signif-

icant role, too35, and could be translated into monsoon statistics using the statistically predictive

day-to-day model.

11


Methods summary

We used a gridded observational data set of daily ISM rainfall to demonstrate the memory ef-

fect. Seasonal mean rainfall statistics as well as the feedback mechanism sustaining a dry circula-

tion regime were established using millennial historical simulations with a comprehensive climate

model. Based on those results, we developed a simple, statistically predictive model for seasonal

mean monsoon rainfall and applied it to the characteristics and trends found in the complex climate

model, including future projections under a global warming scenario.

1. Parthasarathy, B., Munot, A. & Kothawale, D. Regression model for estimation of Indian

foodgrain production from summer monsoon rainfall. Agricultural and Forest Meteorology

42, 167 – 182 (1988).

2. Auffhammer, M., Ramanathan, V. & Vincent, J. R. Integrated model shows that atmospheric

brown clouds and greenhouse gases have reduced rice harvests in India. Proceedings of the

National Academy of Sciences 103, 19668–19672 (2006).

3. Goswami, B. N., Venugopal, V., Sengupta, D., Madhusoodanan, M. S. & Xavier, P. K. In-

creasing trend of extreme rain events over India in a warming environment. Science 314,

1442–1445 (2006).

4. Meehl, G. A. et al. Climate Change 2007: The Physical Science Basis. Contribution of

Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate

12


Change, chap. Global Climate Projections (Cambridge University Press, Cambridge, United

Kingdom and New York, NY, USA, 2007).

5. Kripalani, R. H., Oh, J. H. & Chaudhari, H. S. Response of the East Asian summer monsoon

to doubled atmospheric CO2: Coupled climate model simulations and projections under IPCC

AR4. Theor. Appl. Climatol. 87, 1–28 (2007).

6. Kripalani, R. H., Oh, J. H., Kulkarni, A., Sabade, S. S. & Chaudhari, H. S. South Asian sum-

mer monsoon precipitation variability: Coupled climate model simulations and projections

under IPCC AR4. Theor. Appl. Climatol. 90, 133–159 (2007).

7. Cherchi, A., Alessandri, A., Masina, S. & Navarra, A. Effects of increased CO2 levels on

monsoons. Climate Dynamics (online) (2010).

8. Ramanathan, V. et al. Atmospheric brown clouds: Impacts on South Asian climate and hydro-

logical cycle. Proceedings of the National Academy of Sciences 102, 5326–5333 (2005).

9. Lau, K. M. & Kim, K. M. Observational relationships between aerosol and asian monsoon

rainfall, and circulation. Geophysical Research Letters 33, L21810 (2006).

10. Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V. & Petoukhov, V. The influence of

vegetation-atmosphere-ocean interaction on climate during the mid-holocene. Science 280,

1916–1919 (1998).

11. Claussen, M. Late quaternary vegetation-climate feedbacks. Clim. Past 5, 203–216 (2009).

13


12. Turner, A. G. & Hannachi, A. Is there regime behavior in monsoon convection in the late 20th

century? Geophysical Research Letters 37, L16706 (2010).

13. Li, J. et al. Summer monsoon moisture variability over China and Mongolia during the past

four centuries. Geophysical Research Letters 36, L22705 (2009).

14. Sun, Y., Ding, Y. & Da, A. Changing links between South Asian summer monsoon circulation

and tropospheric land-sea thermal contrasts under a warming scenario. Geophysical Research

Letters 37, L02704 (2010).

15. Burns, S. J., Fleitmann, D., Matter, A., Kramers, J. & Al-Subbary, A. A. Indian ocean climate

and an absolute chronology over Dansgaard/Oeschger events 9 to 13. Science 301, 1365–1367

(2003).

16. Wang, P. et al. Evolution and variability of the Asian monsoon system: state of the art and

outstanding issues. Quaternary Science Reviews 24, 595–629 (2005).

17. Wang, Y. et al. Millennial- and orbital-scale changes in the East Asian monsoon over the past

224,000 years. Nature 451, 1090–1093 (2008).

18. Gupta, A. K., Anderson, D. M. & Overpeck, J. T. Abrupt changes in the Asian southwest

monsoon during the Holocene and their links to the North Atlantic ocean. Nature 421, 354–

357 (2003).

19. Wang, Y. et al. The Holocene Asian Monsoon: Links to solar changes and North Atlantic

climate. Science 308, 854–857 (2005).

14


20. Zickfeld, K., Knopf, B., Petoukhov, V. & Schellnhuber, H. J. Is the indian summer monsoon

stable against global change? Geophysical Research Letters 32, L15707 (2005).

21. Levermann, A., Schewe, J., Petoukhov, V. & Held, H. Basic mechanism for abrupt monsoon

transitions. Proceedings of the National Academy of Sciences 106, 20572–20577 (2009).

22. Rajeevan, M., Bhate, J., Kale, J. A. & Lal, B. High resolution daily gridded rainfall data for

the Indian region: Analysis of break and active monsoon spells. Current Science 91, 296–306

(2006).

23. Haak, H. MPI-M Earth System Modelling Framework: millennium full forcing experiment

(ensemble members 1–5) (2008). World Data Center for Climate. CERA-DB ”mil0010”–

”mil0015”.

24. Annamalai, H., Hamilton, K. & Sperber, K. R. The South Asian summer monsoon and its

relationship with ENSO in the IPCC AR4 simulations. Journal of Climate 20, 1071–1092

(2007).

25. Li, J. & Zhang, L. Wind onset and withdrawal of Asian summer monsoon and their simulated

performance in AMIP models. Climate Dynamics 32, 935–968 (2009).

26. Webster, P. J. & Yang, S. Monsoon and ENSO: Selectively interactive systems. Quarterly

Journal of the Royal Meteorological Society 118, 877–926 (1992).

27. Charney, J. G. & Shukla, J. Monsoon Dynamics, chap. Predictability of Monsoons (Cambridge

University Press, Cambridge, United Kingdom, 1981).

15


28. Wheeler, M. & Hendon, H. An all-season real-time multivariate MJO index: Development of

an index for monitoring and prediction. Monthly Weather Review 132, 1917–1932 (2004).

29. Nakicenovic, N. & Swart, R. (eds.) IPCC Special Report on Emissions Scenarios (Cambridge

University Press, 2000).

30. Jungclaus, J. MPI-M Earth System Modelling Framework: millennium scenarios (contin-

uation of full forcing experiment ensemble members 1–5) (2009). World Data Center for

Climate. CERA-DB ”mil0016”–”mil0020”.

31. Allen, M. & Ingram, W. Constraints on future changes in climate and the hydrologic cycle.

Nature 419, 224–232 (2002).

32. Krishnamurthy, V. & Shukla, J. Intraseasonal and interannual variability of rainfall over India.

Journal of Climate 13, 4366–4377 (2000).

33. Goswami, B. N., Wu, G. & Yasunari, T. The annual cycle, intraseasonal oscillations, and

roadblock to seasonal predictability of the Asian summer monsoon. Journal of Climate 19,

5078–5099 (2006).

34. Palmer, T. N. Chaos and predictability in forecasting the monsoons. Proc. Indian Nat. Sci.

Acad. 60, 57–66 (1994).

35. Li, T., Zhang, Y., Chang, C. & Wang, B. On the relationship between Indian Ocean sea

surface temperature and Asian summer monsoon. Geophysical Research Letters 28, 2843–

2846 (2001).

16


36. Roeckner, E. et al. The atmospheric general circulation model ECHAM5 - Part I: Model de-

scription. Report No. 349, Max Planck Institute for Meteorology, Hamburg, Germany (2003).

37. Hagemann, S., Arpe, K. & Roeckner, E. Evaluation of the hydrological cycle in the ECHAM5

model. Journal of Climate 19, 3810–3827 (2006).

38. Raddatz, T. J. et al. Will the tropical land biosphere dominate the climate-carbon cycle feed-

back during the twenty-first century? Climate Dynamics 29, 565–574 (2007).

39. Marsland, S., Haak, H., Jungclaus, J., Latif, M. & Roske, F. The Max-Planck-Institute global

ocean/sea ice model with orthogonal curvilinear coordinates. Ocean Modelling 5, 91–127

(2003).

40. Wetzel, P. et al. Effects of ocean biology on the penetrative radiation in a coupled climate

model. Journal of Climate 19, 3973–3987 (2006).

41. Crowley, T. J. et al. Volcanism and the Little Ice Age. PAGES Newsletter 16, 22–23 (2008).

42. Krivova, N. A. & Solanki, S. K. Models of Solar Irradiance Variations: Current Status. J.

Astrophys. Astr. 29, 151–158 (2008).

43. Pongratz, J., Reick, C., Raddatz, T. & Claussen, M. A reconstruction of global agricultural

areas and land cover for the last millennium. Global Biogeochemical Cycles 22 (2008).

44. http://www.mpimet.mpg.de/fileadmin/ozean/drg/web page millennium nov09 model details.pdf.

45. Leisch, F. FlexMix: A general framework for finite mixture models and latent class regression

in R. Journal of Statistical Software 11, 1–18 (2004).

17


46. Schwarz, G. Estimating dimension of a model. Annals of statistics 6, 461–464 (1978).

Supplementary Information is linked to the online version of the paper at www.nature.com/nature.

Acknowledgements This work was funded by the Heinrich Boll Foundation, the German National Aca-

demic Foundation, and the BMBF PROGRESS project (support code 03IS2191B). The climate model simu-

lations were performed in the framework of the MPI-M project MILLENNIUM and have been partly funded

by the German Earth System Research Partnership Program ENIGMA. We thank J. Jungclaus for access to

the model output; V. Petoukhov, V. Brovkin, and R. Krishnan for helpful comments on the manuscript; and

K. Frieler for advice on statistical methods.

Author Information The authors declare that they have no competing financial interests. Correspondence

and requests for materials should be addressed to J.S. (email: [email protected]).

18


A B C

WET

DRY

WET

DRY

pinit

1-pinit

pt

1-pt

pt1-pt

Figure 1: Illustration of idealized ISM circulation regimes (A), associated positive feedback mech-

anisms (B), and the simple ‘day-to-day model’ (C). In the wet regime, latent heating due to precip-

itation (P) creates a regional sea level pressure (SLP) low, which leads to atmospheric upwelling

and associated inflow of moist air towards the ISM region. In turn, a lack of latent heating causes a

regional positive SLP anomaly, which leads to subsidence of dry upper-troposphere air and lowers

the humidity of the monsoon winds, thereby inhibiting precipitation and sustaining a dry regime.

In the ‘day-to-day model’, the probability for a wet or dry day depends on the amount of latent

heating accumulated during some previous period, except during the onset when it is determined

by an initial probability pinit.


A

p~P

(m

m/d

ay)

0 0.2 0.4 0.6 0.8 10

5

10

B

year

s ou

t of 6

030

0

200

400

p~P

(m

m/d

ay)

0 0.5 10

5

10

C

mm/day

year

s ou

t of 2

50

COSMOS

day−to−daymodel

0 2 4 6 80

10

20

Figure 2: ISM rainfall statistics. (A) IMD observed daily rainfall during May-September versus

precipitation probability p (see main text). The thick line connects the mean values for each of

50 bins on the p-axis; shading denotes ±1 standard deviation. Days with p < 0.2 stem mainly

from before the monsoon onset. (B) Distribution of seasonal (JJA) mean ISM precipitation from

the COSMOS ensemble over the period 800–2005 CE (gray bars; 5 simulations, 1206 years each)

and from 6030 runs of the stochastic day-to-day model (blue line shows mean value, and error

bars show ±1 standard deviation, from 100 realizations of the model). Dark red (light red) bars

show the distribution of those COSMOS years where the Nino3.4 mean sea level pressure in May

exceeds (falls short of) the long-term mean by more than one standard deviation. Inset: As A, but

with P and p diagnosed from one of the COSMOS simulations. (C) Gray bars as in B, but over the

period 2151–2200 CE in a global warming scenario (see main text). Blue line and errorbars as in

B, but from 250 runs and with a different set of model parameters (see Supplementary Table S1).


1140 1150 1160

−5

0

5

yearm

m/d

ay, i

ndex

precipitationmonsoon index

Bprecipitation

J F M A M J J A S O N D

mm

/day

0

5

10

C

sea level pressure

mb

1005

1010

Dsubsidence10

−2 P

a/s

−6

0

6

E

humidity

g/kg

May Jun Jul Aug Sep

4

8

12

Figure 3: The dry year. (A) Seasonal mean precipitation over India (circles, in mm/day; dashed

line shows long-term mean) and South Asian monsoon index26 (bars) in a 19-year interval of the

simulation; the dry year is marked in black. (B) Annual cycle of average daily precipitation over

India during the dry year (thick) and averaged over the previous 17 years (thin; shading denotes

the 16% and 84% quantiles). See Supplementary Fig. S9 for a comparison covering the entire

simulation period. (C-E) Characteristics of the monsoon circulation over western India and the

Arabian Sea during May-September: (C) Mean sea level pressure (MSLP, in hPa); (D) vertical

velocity ω (in 10−2 Pa/s) at 500hPa, with positive values indicating subsidence of air; and (E)

near-surface (850-1000hPa) specific humidity, in g/kg. In B-E, a 3-day running mean was applied;

thick lines are for the dry year; and thin lines give the average over the previous 17 years, where

the shading denotes ± 1 standard deviation.


precip skewnessB

−1

0

1

precip mean

C

year

mm

/day

800 1000 1200 1400 1600 1800 2000 22004

5

6

D2D modelCOSMOS

° C

Δ Tglobal

Nino3.4 SLP MayA

0

2.5

5

1009

1011

mb

Figure 4: Application of the ‘day-to-day model’ to past and future ISM variability. (A) Global

mean surface temperature (black, in ◦C relative to the 1980-1999 mean) and May MSLP over the

Nino3.4 region (blue, in mb) over the full period of the millennial COSMOS simulations and their

continuations under a global warming scenario. Data have been averaged decadally (801–810, etc.)

and over the 5 ensemble members, such that each point represents an average over 50 model years.

(B) Skewness of the decadal frequency distribution of seasonal mean ISM rainfall in COSMOS

(red). Each point is computed from a set of 50 model years (10 years, 5 ensemble members). The

skewness computed from 50 runs each of the day-to-day model, with parameters set according to

global mean temperature and May Nino3.4 MSLP as in A, is shown in gray (thin line shows mean

value, and shading shows ±1 standard deviation, from 100 realizations of the model). (C) As B,

but for the mean of the frequency distribution (in mm/day).


Methods

Memory effect in observations. For demonstrating the memory effect in the IMD observational

data set, we choose P+ = 12 mm/day and P− = 0 mm/day. P+ is chosen to approximate the

maximum observed 17-day average daily precipitation. We test the sensitivity of the memory effect

to the length of the memory period τ by computing the ‘signal-to-noise ratio’ of the correlation

for a range of τ values (Fig. S2): A linear regression of the mean values between p = 0.4 and 0.6

is evaluated at p = 0.2 and p = 0.8, and the difference in P is divided by the average standard

deviation over the same interval. As the maximum observed average daily precipitation is higher

for shorter averaging periods and vice versa, we have varied P+ along with τ – namely, between

14 and 10 mm/day – in producing Fig. S2. This variation however only has a minor effect on the

signal-to-noise ratio, which is much more sensitive to the choice of τ .

Climate model analysis. The climate model simulations used in this study belong to the

“Millennium” simulation ensemble23. They were performed with the comprehensive MPI-M Earth

System Model COSMOS, comprising the atmospheric GCM ECHAM536, 37 Version 5.4.01 at

T31L19 resolution; the land surface scheme JSBACH38 at T31 resolution, one soil layer, 13 vege-

tation types and four vegetation tiles per grid box; and the oceanic GCM MPI-OM39 Version 1.3.0,

including the ocean biogeochemistry module HAMOCC40, at GR3.0L40 resolution. The model

is run for the years 800 to 2005 CE with time-dependent external forcing including the effects of

volcanic stratospheric sulphate aerosols41, variation of total solar irradiance42, land-use change43,

and changes in orbital parameters. Following the historical period, the simulations are continued

23


with forcing according to the IPCC SRES A1B scenario29 until 2100, and with CO2 emissions cor-

responding to constant CO2 concentrations from 2101 until 220030. Further details can be found

on the MPI-M website44.

In analyzing the climate model results, precipitation is averaged over the land area in 5-

30◦N, 70-90◦E. The dynamical monsoon index26 is computed as the zonal wind shear difference

over 0-20◦N, 60-100◦E according to (U850hPa − 〈U850hPa〉) − (U200hPa − 〈U200hPa〉), where U is

the seasonal (May-September) average zonal wind speed, and angle brackets indicate the long term

mean over the historical simulation period. Sea level pressure is averaged over 0-25◦N, 60-80◦E;

vertical velocity ω in pressure coordinates is averaged over 5-30◦N, 55-80◦E; and specific humidity

over 5-25◦N, 60-80◦E. For demonstrating the memory effect in COSMOS, the same procedure is

applied as for the IMD observational data, except that we now choose P+ = 9 mm/day, which

is closer to the maximum 17-day average daily precipitation observed in the climate model. The

latent–classes analysis is performed using the flexmix routine45, fitting a model of one, two or three

superposed Gaussian distributions to the data. According to the Bayesian Information Criterion46

(BIC), the assumption of two Gaussians improves the model substantially (from a BIC value of

21364 to 20034) as compared to only one Gaussian, while a third Gaussian yields no significant

improvement, and is not physically motivated.

Statistically predictive model. The precipitation probability used in the simple day-to-day

24


model is defined as

pt =

pm if pt ≥ pm,

(1− pm) if pt ≤ (1− pm),

pt else,

(2)

where pm is a maximum probability, and pt is defined according to equation (1) in the main text;

except that t is now the index of model time steps, not days. However, the length of the season is

chosen such that a time step corresponds approximately to one day (see Supplementary Table S1).

During an initial period of length τ in the beginning of the season, pt is not (and cannot be)

determined by equation (2). Instead, we introduce an initial probability pinit which is treated as

a model parameter. This day-to-day model is integrated over l time steps, corresponding to the

length of the model season, and the seasonal mean precipitation P , P− ≤ P ≤ P+, is saved. See

Supplementary Table S1 for sets of parameters used in the model integration, and the attached

Matlab R© script for the model code.

For reproducing multidecadal trends in the historical as well as the future climate simulations

(Fig. 4 in the main paper), we run the day-to-day model for each decade, varying the parameter

pinit according to pinit = p′ · (m − m0) + p0, where m is the decadally-averaged May Nino3.4

MSLP in mb; m0 = 1008.9 mb; p′ = 0.39 mb−1; and p0 = 0.2. The parameters P+ and P− are

varied according to P± = P ′ ·∆T +P±0 , where ∆T is the decadally-averaged global mean surface

temperature anomaly in ◦C; P ′ = 0.42 mm day−1 ◦C−1; P+0 = 9 mm day−1; and P−0 = 0 mm

day−1.

25


Supplementary Material forMore frequent future monsoon failure due to inherent in-stability

Jacob Schewe1,2, Anders Levermann1,2

1Earth System Analysis, Potsdam Institute for Climate Impact Research, Potsdam, Germany

2Institute of Physics, Potsdam University, Potsdam, Germany

1


2 4 6 8 10 12

−200

−100

0

100

200

300

Month

W/m

2

Latent

Sensible


INDIA NCEP/NCAR

Figure S1: Net contributions (climatology) to the annual moist static energy budget of the atmo-

spheric column over India (5-30◦N, 70-90◦E), from NCEP/NCAR reanalysis data1 (1948-2007;

errorbars denote one standard deviation): Radiative fluxes (black solid line), sensible heat flux

from the ground (red), latent heat flux due to precipitation (blue), and convergence of heat due to

advection by the large-scale time-mean circulation and eddies (black dashed line). Sensible heat

flux is strongest just before the monsoon onset, but during the rainy season, latent heat release

dominates the energy budget, balanced by advective processes that transport the excess heat out of

the region. This dominance is found in all major monsoon regions2.

2


0 10 20 30 400

0.5

1

1.5

2

2.5

3

τ (days)

sign

al−

to−

nois

e ra

tio

Figure S2: Sensitivity of the memory effect to the memory period τ . The ‘signal-to-noise ratio’,

i.e. the ratio of the slope of the correlation vs. the standard deviation (cf. Fig. 2A in main paper),

is greater than one for τ ≤ 18 days, and becomes very small for τ greater than 25 days. The filled

circle marks τ = 17 days, the value used for the analysis in the main paper.

3


mm/day

num

ber

of y

ears

0 2 4 6 80

100

200

300

400

500

Figure S3: Gray bars as in Fig. 2B in the main paper. Black dashed lines show the characteristics of

the two modes obtained from latent–classes analysis, and the solid line shows their superposition.

4


30oS

15oS

0o

15oN

30oN

A dry

May

B dry

June

40oE 70oE 100oE 30oS

15oS

0o

15oN

30oN

C average

40oE 70oE 100oE

D average990

1000

1010

1020

1030

Figure S4: (A) Monthly mean sea level pressure (in hPa) in May of the dry year; (B) same for June;

(C) as (A), but averaged over the 17 previous years; (D) same for June. Contour spacing is 4 hPa.

Due to the absence of convection and associated latent heat release, the Eurasian surface Low is

diminished over India in the dry year, which favors a northward expansion of the Indian Ocean

surface High and subsequent subsidence over India and the northwestern Indian ocean region (cf.

Fig. S5).

5


30oS

15oS

0o

15oN

30oN

A dry

May

B dry

June


15oS

0o

15oN

30oN

C average

40oE 70oE 100oE

D average−0.1

0

0.1

Figure S5: (A) Monthly mean vertical velocity ω (in Pa/s) at 500hPa in May of the dry year; (B)

same for June; (C) as (A), but averaged over the 17 previous years; (D) same for June. Positive

ω (red shading) indicates downward motion of air. Contour spacing is 0.02 Pa/s. In the dry year,

central India and the Arabian Sea are characterized by subsidence instead of upwelling.

6


30oS

15oS

0o

15oN

30oN

A ω anomaly

May

B ω anomaly

June


15oS

0o

15oN

30oN

C q anomaly

40oE 70oE 100oE

D q anomaly

−0.1

0

0.1

−3

0

3

Figure S6: (A) Anomaly of monthly mean vertical velocity ω (in Pa/s, contour spacing 0.02) at

500hPa in May of the dry year, with respect to the average over the 17 previous years; (B) same

for June; (C) anomaly of monthly mean lower troposphere (500-1000hPa) specific humidity (g/kg,

contour spacing 1 g/kg) in May of the dry year, with respect to the average over the 17 previous

years; (D) same for June. Anomalous subsidence of dry air over India and the Northern Indian

Ocean leads to moisture depletion in the monsoon region.

7


Nino3.4 MSLP

Feb Mar Apr May Jun

1005

1009

1013

Figure S7: Mean sea level pressure (MSLP) over the Nino3.4 region (170-120◦W, 5◦S-5◦N) during

spring in the COSMOS simulation, for the dry year (thick line) and averaged over the previous 17

years (thin line; shading denotes ± 1 standard deviation).

8


O N D J F M A M J J

previous year

corr

elat

ion

0

0.2

0.4

0.6

0.8

1

Figure S8: Coefficient of the correlation between seasonal (JJA) mean ISM rainfall and monthly-

averaged mean sea level pressure (MSLP) over the Nino3.4 region (170-120◦W, 5◦S-5◦N) in the

COSMOS simulation. The correlation is strongest when MSLP is taken in May just before the

monsoon season (indicated by the gray shading).

9


Precipitation

long−term mean

dry years

J F M A M J J A S O N D

mm

/day

0

2

4

6

8

10

Figure S9: Annual cycle of daily precipitation over India as a long-term mean over the first 1206-

year COSMOS simulation (thin line; shading denotes the 16% and 84% quantiles), and averaged

over all dry years within this simulation (JJA mean precipitation lower than 2.5 mm/day, thick

line).

10


Additional references

1. Kistler, R. et al. The NCEP/NCAR 50-year reanalysis. Bull. Amer. Meteor. Soc. 82, 247 – 267

(2001).

2. Levermann, A., Schewe, J., Petoukhov, V. & Held, H. Basic mechanism for abrupt monsoon

transitions. Proceedings of the National Academy of Sciences 106, 20572–20577 (2009).

11


Table S1: Parameters of the day-to-day model

Parameter Symbol Value used for

Fig. 2B

Value used for

Fig. 2C

Value used for Fig. 4

length of season l 135 time steps

length of memory period τ 17 time steps

precipitation in wet state P+ 9 mm/day 10.9 mm/day varying linearly with

global temperature,

see Methods text

precipitation in dry state P− 0 mm/day 1.9 mm/day varying linearly with

global temperature,

see Methods text

initial probability pinit 0.75 0.2 varying linearly with

Nino3.4 MSLP, see

Methods text

maximum probability pm 0.8 0.82 0.82

12




3 Discussion and Conclusions

In this thesis, I have offered a novel perspective on the possibility of large–scale monsoon failure, motivated

by the observation of two different types of events: The dry regimes, observed in paleoclimatic proxy

records, that persist on millennial and longer timescales; and individual years of extremely deficient

monsoon rainfall as found in present–day climate simulations with a comprehensive global climate model.

Previous studies have linked paleoclimatic monsoon failure to external influences such as remote forcing

from the North Atlantic region (Burns et al., 2003; Wang et al., 2005b), or orbital–scale changes in solar

insolation (Wang et al., 2008), but no consistent theory has been developed so far that can explain the

strong and abrupt response of monsoon rainfall to these either weak or gradual external perturbations.

Seasonal monsoon failure under present–day climate conditions, on the other hand, has been investigated

here in a realistic climate model, as a possible scenario that goes far beyond the interannual variability

found in direct observations.

In the first part of the thesis (section 2.1), I have presented simulations with a coupled climate

model of intermediate complexity, CLIMBER-3α, that project monsoon rainfall around the world to

increase quasi–linearly with global warming in the coming centuries. While this is generally consistent

with many other studies (e.g. Kripalani et al., 2007a), the atmospheric component of CLIMBER-3α is

based on a simplified statistical–dynamical approach, and may not sufficiently represent all processes that

are relevant for the response of monsoon circulations to rapid and intense climate change. Therefore,

in the subsequent work I have attempted to identify those physical mechanisms that are of first–order

importance for large–scale monsoon dynamics and their response to external changes.

The second and third part of my work (sections 2.2 and 2.3) have focused on the conditions that

are necessary in order to sustain a rainy season over land. It is important to note that globally, the

seasonal reversal of cross-equatorial winds is driven by the seasonal change in hemispheric insolation,

and in general the rainfall associated with the intertropical convergence zone will naturally be sustained

even without continental monsoon rainfall. However, the excursion of the tropical rain belt towards

high latitudes and the intense continental precipitation in monsoon regions goes beyond the zonal-mean

dynamics of the intertropical convergence zone.

I have identified the so–called moisture–advection feedback as the primary driving force of large–scale

continental monsoon rainfall (section 2.2). While differential heating of the land and ocean surfaces

is important in establishing the atmospheric land–sea thermal contrast during the onset period, the

moisture–advection feedback is crucial in maintaining this contrast after the rainy season has started.

From an energetic point of view, the advection of latent heat in the form of moisture from the ocean

to the continent is the main source of energy that keeps the monsoon circulation going during the rainy

126 3 DISCUSSION AND CONCLUSIONS

season.

I have shown in a minimal conceptual model that the self–amplifying nature of the moisture–advection

feedback implies a threshold behaviour with respect to changes in the energy budget of the monsoon sys-

tem. In particular, if specific humidity over the ocean falls below a critical value, no conventional monsoon

circulation can develop according to the basic dynamics captured in the minimal model (section 2.3).

These basic dynamics thus define a domain of existence for continental monsoon rainfall, which allows

for abrupt transitions between wet and dry regimes when the threshold is crossed. Physically, the possi-

bility for abrupt transition arises from the competition among the main heat transport processes during

the rainy season. Although latent heat release through precipitation warms the atmospheric column,

direct advection of heat is cooling it. Both processes become weaker when monsoon winds decrease, and

thereby compensate each other with respect to the net heat injection into the atmospheric column. The

threshold of this stabilizing effect is set by the net radiative cooling, which is a characteristic feature of

low latitudes.

I have demonstrated that this concept is qualitatively consistent with a series of abrupt monsoon

transitions found in a proxy record of the East Asian summer monsoon for the penultimate glacial

period. In this application, I have assumed a hysteresis that occurs when the threshold is crossed from

different directions. While this assumption is not crucial for the transition behaviour, it changes the

timing of the individual transitions such that they are all consistent, within dating errors, with those

found in the proxy record. Physically, a hysteresis on these paleoclimatic timescales might be induced

by inert climate components such as e.g. large-scale oceanic circulation or Himalayan glaciation, rather

than by atmospheric processes. While it is thus not inconsistent with the underlying physics, it remains

a hypothesis at this stage and should be further investigated in future research.

My findings suggest that the moisture–advection feedback could indeed be the main physical

mechanism within the dynamics of monsoon circulations that facilitates abrupt and persistent monsoon

failure on long timescales. Beyond these qualitative results, I have estimated the threshold humidity

values for major monsoon regions, using present–day reanalysis data. The resulting distribution of

threshold values can be interpreted either as an uncertain estimate of a stationary critical threshold, or as

a probability distribution of an interannually varying threshold. I have shown that the estimates become

more robust when further relevant processes are incorporated in the conceptual model. However, it must

be kept in mind that the conceptual model is designed in a minimalistic spirit, comprising only those

physical processes that are essential for the moisture–advection feedback, and neglecting many other

processes that are nevertheless important for a quantitative assessment. Therefore my estimates of the

critical threshold need to be considered a first attempt of quantification. For a more accurate estimation,

more complex models would be necessary that, for instance, take into account evaporation over land

and associated soil moisture processes, and regionally specific features such as orography. An extended

127

analysis of paleoclimatic reconstructions as well as detailed present–day observations could also be useful

to further constrain the results. However, the example of the Atlantic thermohaline circulation suggests

that it might be generally very difficult to quantify thresholds in the climate system with the accuracy

that would be necessary for future projection of a transition (Rahmstorf et al., 2005; Drijfhout et al., 2010).

I have also suggested a basic physical mechanism for seasonal monsoon failure under modern climate

conditions, which is again based on the moisture–advection feedback (section 2.4). Within the rainy

season, rainfall over a certain period tends to reinforce the circulation by adding heat to the continental

atmosphere, and thereby increases the probability for rainfall in subsequent days, as I have shown for

India in observational data as well as in a comprehensive AOGCM. On the other hand, the AOGCM

simulations feature years where average Indian summer monsoon rainfall is up to ∼70% below the long–

term mean; this is much lower than any drought year observed within the past century. Examining one

such dry year, I have identified a second self–amplifying feedback that counteracts the moisture–advection

feedback: Subsidence of dry air from the upper troposphere dries out the monsoon winds and thereby

further inhibits the onset of convection, leading to more subsidence, and so on. While this dry feedback is

most dominant and visible in those years with very weak rainfall, I suggest that it is a general dynamical

feature that, to a lesser extent, is also present in regular monsoon years. I have developed a minimal

theory of intraseasonal monsoon dynamics that is based on the idea of a permanent interplay between

these two, antagonistic feedbacks, where each of them tends to persist, while synoptic–scale weather

events can perturb the currently active feedback and induce a flip into the other one. The timescale

of this interplay is on the order of days, and follows from the timescale of synoptic and smaller–scale

tropospheric features and the time that the heating anomaly induced by the moisture–advection feedback

can persist in the atmosphere.

I have framed this minimal theory in a simple, statistically predictive model for seasonal mean monsoon

rainfall. The inherently dynamics of the two counteracting feedbacks produces a series of alternating

wet and dry periods of varying length and frequency, which finally add up to a total seasonal rainfall

amount. While this is a highly idealized representation of internally generated intraseasonal variability,

the fact that real monsoon systems also exhibit variability on similar timescales supports the idea that

an instability similar to the one captured in the simple model may indeed be at work in reality. This

would offer an explanation for some of the observed intraseasonal variability, on the basis of internal

monsoon dynamics. Note that these results do not contradict other sources of monsoon variability on

various timescales. For example, the Madden–Julian Oscillation (Wheeler & Hendon, 2004) plays an

important role for intraseasonal monsoon variability on a monthly timescale. A corresponding external

forcing could in principle be added to the statistically predictive model without qualitatively altering its

characteristics. On the other hand, it is a robust assumption that a major part of intraseasonal variability

128 3 DISCUSSION AND CONCLUSIONS

is generated by internal monsoon dynamics and not by external forcing (Krishnamurthy & Shukla, 2000;

Goswami et al., 2006).

I have shown that this simple conceptual model explains the characteristic, bimodal frequency

distribution of seasonal–mean Indian summer monsoon rainfall found in the AOGCM. Moreover, it can

also reproduce a large portion of multidecadal monsoon variability when forced with central–Pacific

mean sea level pressure (MSLP) anomalies in May, i.e. at the onset time; no external forcing is applied

during the rest of the rainy season. Anomalous spring-time conditions in the central Pacific are assumed

to induce atmospheric conditions that favor either convection or subsidence over the Indian region, thus

leading to either intense or deficient monsoon onsets. This initial condition is then propagated into

the seasonal–mean rainfall in a probabilistic sense; in principle however, even after a deficient onset an

overall strong monsoon season can develop according to the basic internal dynamics.

We can now delineate a common perspective for both seasonal and long–term monsoon failure. Under

present–day climate conditions, the monsoon season is governed by the inherent instability induced by

the self–amplifying moisture–advection feedback and the counteracting dry subsidence feedback. The

seasonal–mean rainfall is therefore determined mainly by stochastic internal dynamics, in conjunction with

an initial condition set by external forcing during the onset period. The resulting statistics of seasonal–

mean rainfall include very dry years, or monsoon failure, albeit with a low frequency of occurrence (or

low probability, if the frequency distribution is interpreted as a probability distribution). If however

overall climatic conditions change substantially, the system can shift into a regime where the moisture–

advection feedback cannot be established, and thus continental monsoon rainfall cannot be sustained.

Such a regime shift would go along with crossing a threshold that is associated with non–zero values of

expected seasonal–mean precipitation; in this sense, the regime shift would be abrupt, i.e. from “rain”

to “no rain”.

Of course, this is a highly simplified perspective. It cannot, and is not meant to, explain all relevant

aspects of monsoon circulations, nor to make quantitatively accurate assessments or predictions of mon-

soon characteristics. I have taken this simplified approach because for understanding large–scale monsoon

failure, whether seasonally or permanently, it might be a useful intermediate step to consider only the

most fundamental physical processes that are essential for the non–linearity which is dominant for large

temporal and spatial scales. My results are qualitatively consistent with abrupt monsoon transitions

found in paleo–records, as well as with monsoon variability under present–day conditions as observed in

a realistic climate model; and they offer a simple and readily understood explanation for both. This may

be seen as supporting the simplified approach.

It is important to note that, besides the quantitative weaknesses that result from the many simplifi-

cations, some of my conclusions may also depend on regional specifics. While in the first three articles

129

I have considered several major monsoon regions and set up a conceptual model that can be expected

to hold for all those regions to a similar degree, the fourth study has focused especially on India. There

is reason to believe that much of the fundamental dynamics I have extracted from the analysis of the

Indian summer monsoon are universal across different monsoon regions, but detailed analysis of those

other regions is needed to confirm this hypothesis.

The final part of this thesis lends further support to my conceptual results, and it also sheds light on a

possible risk associated with the response of unstable monsoon dynamics to future anthropogenic climate

change. Simulations with a realistic climate model project seasonal failure of Indian summer monsoon

rainfall to become much more frequent in response to global warming. The minimal, statistically predic-

tive model closely reproduces this trend, with spring-time MSLP in the central Pacific again determining

the initial condition at the monsoon onset. Within this simplified framework, the increase in monsoon

failure is thus caused by a strong but delayed response of spring–time central Pacific MSLP to global

warming. While in my analysis these central Pacific climate variations seem to be the dominant influence

on the initial condition, other factors such as changing springtime Indian Ocean sea surface temperatures

might play a significant role, too (Li et al., 2001), and could be translated into monsoon statistics using

the statistically predictive model.

Note that I have only considered climate simulations with a single AOGCM. Comparison with other

comprehensive models is needed to test the robustness of the projected trends in Indian summer monsoon

and central Pacific climate. The interrelation between the two needs to be investigated in detail using

complex models as well as observations, in order to confirm the link suggested by my results. Considering

the potential consequences of frequent large–scale monsoon failure in the near future, such research efforts

appear worthwhile.

References

Auffhammer, M., Ramanathan, V., & Vincent, J. R. 2006. Integrated model shows that atmo-

spheric brown clouds and greenhouse gases have reduced rice harvests in India. Proceedings of the

National Academy of Sciences, 103, 19668–19672.

Burns, S. J., Fleitmann, D., Matter, A., Kramers, J., & Al-Subbary, A. A. 2003. Indian

Ocean Climate and an Absolute Chronology over Dansgaard/Oeschger Events 9 to 13. Science, 301,

1365–1367.

Cherchi, Annalisa, Alessandri, Andrea, Masina, Simona, & Navarra, Antonio. 2010. Effects

of increased CO2 levels on monsoons. Climate Dynamics, (online), 1–19.

Clark, C. O., Cole, J. E., & Webster, P. J. 2000. Indian Ocean SST and Indian summer rainfall:

Predictive relationships and their decadal variability. Journal of Climate, 13(14), 2503–2519.

Claussen, M. 1997. Modeling bio-geophysical feedback in the African and Indian monsoon region.

Climate Dynamics, 54, 247–257.

Claussen, M. 2009. Late Quaternary vegetation-climate feedbacks. Climate of the Past, 5(2), 203–216.

Dash, S. K., Singh, G. P., Shekhar, M. S., & Vernekar, A. D. 2005. Response of the Indian

summer monsoon circulation and rainfall to seasonal snow depth anomaly over Eurasia. Climate

Dynamics, 24, 1–10.

Drijfhout, S. S, Weber, S. L, & van der Swaluw, E. 2010. The stability of the MOC as diagnosed

from model projections for pre-industrial, present and future climates. Climate Dynamics, 112.

Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V., & Petoukhov, V. 1998. The

Influence of Vegetation-Atmosphere-Ocean Interaction on Climate During the Mid-Holocene. Science,

280, 1916–1919.

Goswami, B. N., & Xavier, P. K. 2005. ENSO control on the south Asian monsoon through the

length of the rainy season. Geophysical Research Letters, 32, L18717.

Goswami, B. N., Venugopal, V., Sengupta, D., Madhusoodanan, M. S., & Xavier, P. K.

2006a. Increasing Trend of Extreme Rain Events Over India in a Warming Environment. Science, 314,

1442–1445.

Goswami, B. N., Madhusoodanan, M. S., Neema, C. P., & Sengupta, D. 2006b. A physical

mechanism for North Atlantic SST influence on the Indian summer monsoon. Geophysical Research

Letters, 33, L02706.

132 REFERENCES

Goswami, B. N., Wu, Guoxiong, & Yasunari, T. 2006. The annual cycle, intraseasonal oscillations,

and roadblock to seasonal predictability of the Asian summer monsoon. Journal of Climate, 19(20),

5078–5099.

Gregory, P. J., Ingram, J. S. I., & Brklacich, M. 2005. Climate change and food security. 360,

2139–2148.

Gupta, A. K., Anderson, D. M., & Overpeck, J. T. 2003. Abrupt changes in the Asian southwest

monsoon during the Holocene and their links to the North Atlantic Ocean. Nature, 421, 354–357.

Hahn, D. G., & Shukla, J. 1976. An apparent relationship between Eurasian snow cover and Indian

monsoon rainfall. Journal of Atmospheric Sciences, 33(12), 2461–2462.

Haile, M. 2005. Weather patterns, food security and humanitarian response in sub-Saharan Africa.

360, 2169–2182.

Hong, YT, Hong, B, Lin, QH, Zhu, YX, Shibata, Y, Hirota, M, Uchida, M, Leng, XT, Jiang,

HB, Xu, H, Wang, H, & Yi, L. 2003. Correlation between Indian Ocean summer monsoon and North

Atlantic climate during the Holocene. Earth and Planetary Science Letters, 211(3–4), 371–380.

Kistler, R., Kalnay, E., Saha, S., White, G., Woollen, J., Chelliah, M., Ebisuzaki, W.,

Kanamitsu, M., Kousky, V., den Dool, H. Van, Jenne, R., & Fiorino, M. 2001. The

NCEP/NCAR 50-year reanalysis. Bull. Amer. Meteor. Soc., 82, 247 – 267.

Kripalani, R. H., Oh, J. H., & Chaudhari, H. S. 2007a. Response of the East Asian summer

monsoon to doubled atmospheric CO2: Coupled climate model simulations and projections under

IPCC AR4. Theoretical and Applied Climatology, 87(1-4), 1–28.

Kripalani, R. H., Oh, J. H., Kulkarni, A., Sabade, S. S., & Chaudhari, H. S. 2007b. South

Asian summer monsoon precipitation variability: Coupled climate model simulations and projections

under IPCC AR4. Theoretical and Applied Climatology, 90(3-4), 133–159.

Krishnamurthy, V., & Goswami, B. N. 2000. Indian monsoon-ENSO relationship on inter-decadal

timescale. Journal of Climate, 13, 579–595.

Krishnamurthy, V, & Shukla, J. 2000. Intraseasonal and interannual variability of rainfall over

India. Journal of Climate, 13(24), 4366–4377.

Kucharski, F., Molteni, F., & Yoo, J. H. 2006. SST forcing of decadal Indian monsoon rainfall

variability. Geophysical Research Letters, 33(3), L03709.

REFERENCES 133

Kumar, K. Krishna, Kumar, K. Rupa, Ashrit, R. G., Deshpande, N. R., & Hansen, J. W.

2004. Climate impacts on Indian agriculture. International Journal of Climatology, 24, 1375–1393.

Lau, K. M., & Kim, K. M. 2006. Observational relationships between aerosol and Asian monsoon

rainfall, and circulation. Geophysical Research Letters, 33, L21810.

Levermann, A., Schewe, J., Petoukhov, V., & Held, H. 2009. Basic mechanism for abrupt

monsoon transitions. Proceedings of the National Academy of Sciences, 106(49), 20572–20577.

Li, J., Cook, E. R., Chen, F., Davi, N., D’Arrigo, R., Gou, X., Wright, W. E., Fang, K.,

Jin, L., Shi, J., & Yang, T. 2009. Summer monsoon moisture variability over China and Mongolia

during the past four centuries. Geophysical Research Letters, 36, L22705.

Li, T, Zhang, YS, Chang, CP, & Wang, B. 2001. On the relationship between Indian Ocean sea

surface temperature and Asian summer monsoon. Geophysical Research Letters, 28(14), 2843–2846.

Liu, X., & Yin, Z. 2002. Sensitivity of East Asian monsoon climate to the uplift of the Tibetan Plateau.

Palaeogeography, Palaeoclimatology, Palaeoecology, 183, 223–245.

Meehl, G. A. 1994. Influence of the Land Surface in the Asian Summer Monsoon: External Conditions

versus Internal Feedbacks. Journal of Climate, 7, 1033–1049.

Meehl, G. A., Stocker, T. F., Collins, W. D., Friedlingstein, P., Gaye, A. T., Gregory,

J. M., Kitoh, A., Knutti, R., Murphy, J. M., Noda, A., Raper, S. C. B., Watterson,

I. G., Weaver, A. J., & Zhao, Z.-C. 2007. Climate Change 2007: The Physical Science Basis.

Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on

Climate Change. Cambridge, United Kingdom and New York, NY, USA: Cambridge University Press.

Chap. Global Climate Projections.

Montoya, M., Griesel, A., Levermann, A., Mignot, J., Hofmann, M., Ganopolski, A., &

Rahmstorf, S. 2005. The Earth System Model of Intermediate Complexity CLIMBER-3α. Part I:

description and performance for present day conditions. Climate Dynamics, 25, 237–263.

Overpeck, J. T., Anderson, D., Trumbore, S., & Prell, W. 1996. The southwest Indian Monsoon

over the last 18000 years. Climate Dynamics, 12, 213–225.

Parthasarathy, B., Munot, A.A., & Kothawale, D.R. 1988. Regression model for estimation

of Indian foodgrain production from summer monsoon rainfall. Agricultural and Forest Meteorology,

42(2-3), 167 – 182.

134 REFERENCES

Rahmstorf, S., Crucifix, M., Ganopolski, A., Goosse, H., Kamenkovich, I., Knutti, R.,

Lohmann, G., Marsh, B., Mysak, L. A., Wang, Z., & Weaver, A. 2005. Thermohaline circu-

lation hysteresis: A model intercomparison. Geophysical Research Letters, 32, L23605.

Rajeevan, M, Gadgil, Sulochana, & Bhate, Jyoti. 2010. Active and break spells of the Indian

summer monsoon. Journal of Earth System Science, 119(3), 229–247.

Ramanathan, V., Chung, C., Kim, D., Bettge, T., Kiehl, J. T., Washington, W. M., Fu, Q.,

Sikka, D. R., & Wild, M. 2005. Atmospheric brown clouds: Impacts on South Asian climate and

hydrological cycle. Proceedings of the National Academy of Sciences, 102(15), 5326–5333.

Rashid, Harunur, England, Emily, Thompson, Lonnie, & Polyak, Leonid. 2011. Late Glacial

to Holocene Indian Summer Monsoon Variability Based upon Sediment Records Taken from the Bay

of Bengal. Terrestrial Atmospheric and Oceanic Sciences, 22(2, Sp. Iss. SI), 215–228.

Robock, A., Mu, M., Vinnikov, K., & Robinson, D. 2003. Land surface conditions over Eurasia

and Indian summer monsoon rainfall. Journal of Geophysical Research, 108(D4), 4131.

Schewe, J., Levermann, A., & Meinshausen, M. 2011a. Climate change under a scenario near

1.5◦C of global warming: monsoon intensification, ocean warming and steric sea level rise. Earth

System Dynamics, 2(1), 25–35.

Schewe, J., Levermann, A., & Cheng, H. 2011b. A critical humidity threshold for monsoon transi-

tions. Climate of the Past Discussions, 7.

Sikka, D. R. 2003. Evaluation of monitoring and forecasting of summer monsoon over India and a

review of monsoon drought of 2002. Proceedings of the Indian National Science Academy Part A,

69(5), 479–504.

Srinivasan, J. 2001. A simple thermodynamic model for seasonal variation of monsoon rainfall. Current

Science, 80(1), 73–77.

Sun, Ying, Ding, Yihui, & Da, Aiguo. 2010. Changing links between South Asian summer monsoon

circulation and tropospheric land-sea thermal contrasts under a warming scenario. Geophysical Research

Letters, 37, L02704.

Tao, F., Yokozawa, M., Zhang, Z., Hayashi, Y., Grassl, H., & Fu, C. 2004. Variability in

climatology and agricultural production in China in association with the East Asian summer monsoon

and El Nino Southern Oscillation. 28, 23–30.

Turner, A. G., & Hannachi, A. 2010. Is there regime behavior in monsoon convection in the late

20th century? Geophysical Research Letters, 37, L16706.

REFERENCES 135

Wang, B. 2005. The Asian monsoon. Springer-Verlag.

Wang, P., Clemens, S., Beaufort, L., Braconnot, P., Ganssen, G., Jian, Z., Kershaw, P.,

& Sarnthein, M. 2005a. Evolution and variability of the Asian monsoon system: state of the art

and outstanding issues. Quaternary Science Reviews, 24, 595–629.

Wang, Y., Cheng, H., Edwards, R. L., He, Y., Kong, X., An, Z., Wu, J., Kelly, M. J.,

Dykoski, C. A., & Li, X. 2005b. The Holocene Asian Monsoon: Links to Solar Changes and North

Atlantic Climate. Science, 308, 854–857.

Wang, Y., Cheng, H., Edwards, R. L., Kong, X., Shao, X., Chen, S., Wu, J., Jiang, X.,

Wang, X., & An, Z. 2008. Millennial- and orbital-scale changes in the East Asian monsoon over the

past 224,000 years. Nature, 451, 1090–1093.

Webster, P. J. 1987a. The Elementary Monsoon. Pages 3–32 of: Fein, J. S., & Stephens, P. L.

(eds), Monsoons. New York, N.Y.: John Wiley.

Webster, P. J. 1987b. The Variable and Interactive Monsoon. Pages 269–330 of: Fein, J. S., &

Stephens, P. L. (eds), Monsoons. New York, N.Y.: John Wiley.

Webster, P. J., Magana, V. O., Palmer, T. N., Shukla, J., Tomas, R. A., Yanai, M., &

Yasunari, T. 1998. Monsoons: Processes, predictability, and the prospects for prediction. Journal of

Geophysical Research, 103, 14,451–14,510.

Webster, P. J., Toma, V. E., & Kim, H.-M. 2011. Were the 2010 Pakistan floods predictable?

Geophysical Research Letters, 38, L04806.

Wheeler, MC, & Hendon, HH. 2004. An all-season real-time multivariate MJO index: Development

of an index for monitoring and prediction. Monthly Weather Review, 132(8), 1917–1932.

Yang, J., Liu, Q., Xie, S.-P., Liu, Z., & Wu, L. 2007. Impact of the Indian Ocean SST basin mode

on the Asian summer monsoon. Geophysical Research Letters, 34, L02708.

Zhang, P., Cheng, H., Edwards, R. L., Chen, F., Wang, Y., Yang, X., Liu, J., Tan, M.,

Wang, X., Liu, J., An, C., Dai, Z., Zhou, J., Zhang, D., Jia, J., & Johnson, K. R. 2008. A

Test of Climate, Sun, and Culture Relationships from an 1810-Year Chinese Cave Record. Science,

322, 940–942.

Zhang, R., & Delworth, T. L. 2005. Simulated Tropical Response to a Substantial Weakening of the

Atlantic Thermohaline Circulation. Journal of Climate, 18, 1853–1860.

Danksagung

Mein Dank gilt zuallererst Anders Levermann fur die hervorragende Betreuung, die fur den Erfolg

meiner Arbeit und nicht zuletzt die Freude an derselben von unschatzbarem Wert war und mich stets

sehr motiviert hat.

Herzlich bedanken mochte ich mich bei meinen Kolleginnen und Kollegen am PIK. Malte, Her-

mann, Vladimir, danke fur die gute Zusammenarbeit bei den jeweiligen Studien. Arathy, Alex,

Carl, Dana, Daria, Dim, Friederike, Hendrik, Johannes, Katja, Mahe, Maria, Marianne, Matthias,

Rica, Tore, Torsten... danke fur Rat und Tat, Obst und Kaffee, Wandern und Oper, fur schone drei Jahre!

Dank gebuhrt auch der Heinrich-Boll-Stiftung fur die Finanzierung meiner Promotion und fur ein

Begleitprogramm, das mir Ausblicke und Eindrucke jenseits der Naturwissenschaft ermoglicht hat; und

der Studienstiftung des deutschen Volkes fur die ideelle Forderung.

Schließlich danke ich meiner Familie fur alles.

“Siehst du! Jetzt fangt alles erst richtig an. Was du bisher erfahren

hast, das war doch nur eine Vorahnung vom Anfang und langst

noch nicht

alles.”

Hans Bemmann, Stein und Flote

Diese Arbeit ist bisher an keiner anderen Hochschule eingereicht worden. Sie wurde selbstandig und

ausschließlich mit den angegebenen Mitteln angefertigt.

Jacob Schewe

Potsdam, im Mai 2011

Diese Arbeit wurde durchgefuhrt am

Potsdam–Institut fur Klimafolgenforschung

Documents

Basic physical mechanisms for monsoon failure in past and future …schewe/publications/Dissertation... · 2011-05-09 · At the same time, paleoclimatic records show evidence of