CTWF Workshop on Development of Regional Earth System Model and Its Applications Kunming, China

Land surface hydrological prediction: The role of off-line, partially coupled, and fully

coupled models  

CTWF Workshop on Development of Regional Earth System Model and Its Applications

Kunming, China

September19, 2008

With thanks to:Joe Casola (Atm Sci, UW)

Ben Livneh (CEE, UW)Chunmei Zhu (CEE, UW)

Dennis P. LettenmaierDepartment of Civil and Environmental Engineering

University of Washington

Outline of this talk

• The role of the land surface in earth system models, and the heritage of Land Surface Parameterizations

• Three examples of coupled and uncoupled applications– Climate change – Projections of Colorado River runoff– Climate change – sensitivity of Pacific Northwest

snowpack to regional warming– Regional climate predictability – The North American

Monsoon System

• Concluding thoughts

The role of the land surface in earth system models (or, what is the difference between a

land surface scheme and a hydrologic model?)

Hydrologic Model: Predict Q (and perhaps other variables related to surface and subsurface moisture) given P and Ep

Land surface model: Predict partitioning of Rn given downward solar and longwave radiation

Typical LSM structure

Streamflow simulation as a measure of model performance

Hydrographs of routed runoff show good correspondence with observed and naturalized flows.

Point Evaluation of a Surface Hydrology Model for BOREAS – Summer 1994 - Mean Diurnal Cycle

Flu

x (W

/m2)

-100

100

300 Rnet

-50

50

150

250

H

0

60

120LE

0 3 6 9 12 15 18 21 24

SSA Mature Black Spruce

Rnet

H

LE

0 3 6 9 12 15 18 21 24

SSA Mature Jack Pine

Rnet

H

LE

0 3 6 9 12 15 18 21 24

Local time (hours)

NSA Mature Black Spruce

Observed Fluxes

Simulated Fluxes

Rnet Net Radiation

H Sensible Heat Flux

LE Latent Heat Flux

Surface energy flux simulation as a measure of model performance

General situation c. 1990

• Hydrologic model: Typically compute Ep given T (and perhaps other variables) but often w/o explicit representation of vegetation. The calibrate for parameters related to soil properties. Often implicit assumption (if needed) is that Ts = Ta

• L/S model: Given vegetation and associated parameters, iterate for Ts by closing water and energy balances simultaneously. Runoff is essentially residual of P, E, and storage change.

Trends last 10-15 years

• Hydrology models have been driven toward explicit vegetation representations by the need to predict land cover change implications, among other factors

• L/S models have been driven to do a better job wrt land surface hydrologic variables via realization that errors in water cycle representation affect surface energy balance, i.e., errors in runoff errors in ET errors in energy partitioning (and timing)

Coupled vs uncoupled (off-line) simulation

• Coupled land-atmosphere modeling– Represents full effect of land-atmosphere feedbacks– Biases at the land surface (e.g., in precipitation and other surface

variables) are inevitable– Biases can dominate the climate signal

• Uncoupled (prescribed surface forcing)– Allows opportunity to deal with effects of bias– No representation of feedback effects

• Partially coupled– Off-line simulations with output of coupled model simulations,

with bias correction– Partially (but only partially) accounts for feedback effects

Example 1: Colorado River climate change assessments

NCAR/DOE Parallel Climate Model at ~ 2.8 degrees lat-long) – global grid mesh, representation of western U.S.

Bias Correction

from NCDC observations

from PCM historical runraw climate scenario

bias-corrected climate scenario

month mmonth m

Note: future scenario temperature trend (relative to control run) removed before, and replaced after, bias-correction step.

Downscaling

observed mean fields

(1/8-1/4 degree)

monthly PCManomaly (T42)

VIC-scale monthly simulation

interpolated to VIC scale

Bias Correction and Downscaling Approach

climate model scenariometeorological outputs

hydrologic model inputs

snowpackrunoffstreamflow

• 1/8-1/4 degree resolution• daily P, Tmin, Tmax

•2.8 (T42)/0.5 degree resolution•monthly total P, avg. T

Christensen and Lettenmaier (HESSD, 2007) – multimodel ensemble analysis with 11 IPCC AR4 models

(downscaled as in C&L, 2004)

Magnitude and Consistency of Model-Projected Changesin Annual Runoff by Water Resources Region, 2041-2060

Median change in annual runoff from 24 numerical experiments (color scale)and fraction of 24 experiments producing common direction of change (inset numerical values).

+25%

+10%

+5%

+2%

-2%

-5%

-10%

-25%

Dec

reas

eIn

crea

se

(After Milly, P.C.D., K.A. Dunne, A.V. Vecchia, Global pattern of trends in streamflow andwater availability in a changing climate, Nature, 438, 347-350, 2005.)

96%

75%67%

62%87%

87%

71%

67%62%

58%

67%

62%58%

67%100%

from Seager et al, Science, 2007

Question: Why such a large discrepancy in projected Colorado River flow changes?

• ~6% annual flow reduction in Christensen and Lettenmaier (2007)

• 10-25% by Milly et al (2005)

• > 35% by Seager et al (2007)

VIC NOAH SAC

Unconditional histograms of 1/8 degree grid cell precipitation elasticities from model runs for 20 years, ~1985-2005

Model Precipitation-Elasticity

Temp-sensitivity (Tmin & Tmax ) %/ 0C

Temp-sensitivity ( Tmax) %/ 0C

Flow @ Lees Ferry(MACF)

VIC 1.9 -2.2 -3.3 15.43

NOAH 1.81 -2.85 -3.93 17.43

SAC 1.77 -2.65 -4.10 15.76

Summary of precipitation elasticities and temperatures sensitivities for Colorado River at Lees Ferry for VIC, NOAH, and SAC models

b) On the other hand, from Seager et al (2007), very roughly, mid-century ΔP -18%, so for = 1.5-1.9, and temperature sensitivity -0.02 - -0.03, and ΔT 2 oC, ΔQ 35% (vs > 50% + from GCM multimodel average)

a) Lowest mid-century estimate (Christensen and Lettenmaier, 2007) is based on a precipitation downscaling method that yields smaller mid-century precipitation changes (by about a factor of two on multimodel average). Adjusting for this difference doubles the projected change to around 12% by mid century – not far from Milly et al (2005), but still well below Seager et al (2007)

Example 2: Sensitivity of Pacific Northwest snowpack to regional warming

•Off-line land surface models:

NOAH 2.0 (problems with albedo, lack of re-freeze in pack; surface turbulent exchange (generally biases snow water equivalent (SWE) down

NOAH 2.8 (resolves most problems with V 2.0)

•Coupled model simulations: MM5 nested to ~15 km spatial resolution; with boundary forcings for decade of 1990s from ECHAM 5 GCM

•Off-line model forcing data sets:Maurer et al 2002 (1/8 degree hourly; based on daily gridded gage observations with precipitation and temperature adjusted for orography using PRISM

MM5 15 km output, aggregated to daily, then disaggregated using Maurer et al methods – for decade of 1990s, and 2020s

Maurer et al 2002 with temperature (only) adjusted to have same monthly (grid cell) means as MM5 for 1990s and 2020s

Winter daily minima 1916-2003 Winter daily maxima 1916-2003

What have been (and will be) the implications of a warming climate on PNW snowpacks?

1) Calculating SWE using a simple sensitivity approach and hypsometric relationship for Cascade Mountains, WA

Ele

vati

on

(m

)

red area

area under outer curve

SWE

SWE

=23% mean Apr 1 SWE lost for +1°C

warming

SWE Volume (S(z) x A(z))

SW

E V

olu

me (

km3)

Winter Temp (deg C)

=27% mean Apr 1 SWE lost for +1°C warming, but range

is 12-42%

Regression of SWE vs. Mean Winter Temperature

2) Regression results

0.0710% (+2-22%)Nearby HCN, just East

0.2325% (9-41%)Nearby HCN, just West

0.1518% (3-33%)Nearby HCN stations, All

0.2821% (9-33%)Clim. Div. 5

0.2827% (12-42%)Clim. Div. 4

r2Temperature Data Used

•95% confidence limits are LARGE

•Regressions explain small portion of variance

3) Model-based results (coupled and uncoupled)

All model results (coupled and uncoupled) use NOAH land scheme, earlier versions (2.7 and prior) have substantial downward bias in SWE, especially for deep mountain snowpacks

Description of Forcings

• Maurer et al. 1990s Forcings: set of regridded daily observed station data (P,Tmax,Tmin,W) from 1949-2000 following the procedure outlined in Maurer el al., 2002 (only 1990s used in this study)

• MM5-ECHAM-5 Forcings: GCM computed fields of Tmax,Tmin,Wind,etc…at 6-hour intervals used to force Noah model, for 1990s and 2020s* (*warming scenario).

• MM5 delta-T applied to Maurer et al: The long-term monthly mean temperature departures between the MM5-ECHAM-5 Tmax, Tmin and Maurer et al, were applied to the Maurer et al. set to yield a forcing set with Maurer et al P and W, that has the same long-term monthly mean Tmax,Tmin as the MM5 GCM-output forcings.

Noah 2.0

Noah 2.8

Δ: Noah 2.8 – Noah 2.0

Seasonal Snow water storage

0

10

20

30

40

50

60

Nov Dec Jan Feb Mar Apr May Jun Jul

Sno

w W

ater

Sto

rage

(km

**3

/ WA

)

Noah 2.8

Noah 2.0

Noah 2.0

Noah 2.8

Δ: Noah 2.8 – Noah 2.0

Seasonal Snow water storage

0

10

20

30

40

50

60

70

Nov Dec Jan Feb Mar Apr May Jun Jul

Sn

ow

Wat

er S

tora

ge

(km

**3

/ WA

) Noah 2.8

Noah 2.0

Noah 2.0

Noah 2.8 Seasonal Snow water storage

Δ: Noah 2.8 – Noah 2.0

0

10

20

30

40

50

60

70

Nov Dec Jan Feb Mar Apr May Jun Jul

Sn

ow

Wat

er S

tora

ge

(km

**3

/ WA

) Noah 2.8

Noah 2.0

Noah 2.0

Noah 2.8

Δ: Noah 2.8 – Noah 2.0

Seasonal Snow water storage

0

10

20

30

40

50

60

Nov Dec Jan Feb Mar Apr May Jun Jul

Sn

ow

Wat

er S

tora

ge

(km

**3

/ WA

) Noah 2.8

Noah 2.0

MM5 SWE 2020s

MM5 – SWE 1990s

Δ: SWEMM5-1990s - SWEMM5-2020s

Seasonal Snow water storage

0

10

20

30

40

50

60

Nov Dec Jan Feb Mar Apr May Jun Jul

Sn

ow

Wat

er

Sto

rag

e (k

m**

3 / W

A) MM5 SWE 1990s

MM5 SWE 2020s

Model sensitivities

Forcings Maurer: 1990sMM5 GCM output

forcing: 1990s

Mean air temp over snow season (Nov – Mar, °C)

1.0 -1.5

Sensitivity of Mean Apr1. Snow Water Storage to 1° C change

Noah 2.8 (over WA)5.1 km3 / °C (9.5 %) 12.8 km3 / °C (20.6 %)

Sensitivity of Mean Apr1. Snow Water Storage to 1° C change Noah 2.0 (over WA)

6.2 km3 / °C (18.8 %) 4.4 km3 / °C (27.0 %)

Sensitivity of Mean Apr1. Snow Water Storage to 1° C change MM5 Snow Output (over WA)

XX.X 13.5 km3 / °C (64.6 %)

Summary of results

Forcings Maurer: 1990sMM5 ΔT applied to Maurer: 1990s

MM5: 1990s MM5: 2020s

Mean Seasonal Temperature (Nov-Jul / Mean air temp over snow season (Nov-Mar °C)

5.9 / 1.0 3.3 / -1.5 3.3 / -1.5 4.3 / -0.4

Mean Apr1. Snow Water Storage

Noah 2.8 (km3/WA)40.5 53.2 61.8 47.7

Mean Apr1. Snow Water Storage

Noah 2.0 (km3/WA)17.4 32.8 16.3 11.9

Mean Apr1 Snow Water Storage

MM5 Output (km3/WA)XX.X XX.X 24.7 11.2

Example 3: Role of soil moisture in the onset of the North American Monsoon

North American monsoon is experienced as a pronounced increase in rainfall from extremely dry May to rainy June.

North American Monsoon Experiment (NAME): Tier 1,2,3. (http://www.cpc.ncep.noaa.gov/products/precip/monsoon/NAME.html)( Comrie & Glenn, 1998 )

The NAMS concept --- thermal contrast between land and adjacent oceanic regions

( http://www.ifm.uni-kiel.de )

The importance to explore possible links between NAMS and antecedent surface conditions.

Winter Precipitation - Monsoon Rainfall

feedback hypothesis

Higher (lower) winter precipitation & spring snowpack

More (less) spring & early summer soil moisture

Weak (strong) monsoon Lower (higher) spring & early summer surface temperature

2) Relationship between Antecedent Land Surface Conditions with Summer Monsoon Rainfall over Southwestern US

Zhu C. M., D. P. Lettenmaier, and Tereza Cavazos, 2005: Role of Antecedent Land Surface Conditions on North American Monsoon Rainfall Variability. J. Climate, 18, 2824-2841.

JFM Precipitation in extreme monsoon

years

DRY WET

DRY WET

Apr-May Soil Moisture in extreme monsoon years

Dry MonsoonWet Monsoon

WetDry

WarmCold

June Ts in extreme monsoon years

June Sm in extreme monsoon years

× ×

PNNL UWvegetation type: Single Multiple elevation band: Single Multiple Parameters: Soil, veg type dependent cell dependent initialization: Spin up 3 months Offline VIC

MM5-VIC coupled model system

Precipitation PressureRadiationWind HumidityAir temperature

Sensible heat fluxLatent heat fluxes…

First coupled by Drs. Ruby Leung at PNNL and Xu Liang at University of California, Berkeley

Modification of coupled MM5/VIC modeling system by UW

Experimental Design

Initial soil moisture prescribed at

OctSepAugJulyJune

SM free runningMay 15

Field capacity Wilting point

► Simulations performed on wet and dry monsoon years to represent different atmospheric circulation conditions

2

► The initial soil wetness condition on May 15 is a surrogate for previous winter precipitation condition.

► Control simulation s. moisture prescribed from offline VIC (1 year spin-up).

Positive Soil Moisture-Monsoon Rainfall

Feedback ?

1984-wet minus 1984-dry 1989-wet minus 1989-dry

June July

Aug Sep

June July

Aug Sep

mean monthly precipitation difference

Summary and Conclusions

• Main purpose of land surface representation to the atmosphere in coupled simulations is partitioning of net radiation into latent, sensible, and ground heat flux

• As coupled simulations have begun to focus on a range of issues beyond prediction of free atmosphere variables, demands on their performance have increased

• A key diagnostic of coupled model performance with respect to land surface predictions (e.g., soil moisture, snow, runoff, ET) is performance of the land model in both a) off-line, and b) partially coupled configurations

• Hence, it is important to have access to off-line simulations with the same version of the land model as is used in the coupled simulations

• Off-line simulations have to potential to allow synthesis of land variables (e.g., soil moisture) that are not directly observed, but off-line simulations need to be carefully verified