L. Phil Graham - DiVA portal8668/FULLTEXT01.pdf · atmosphere and the land surface, and the land surface and the sea can be studied in detail. Understanding and modeling the large-scale

LARGE-SCALE HYDROLOGIC MODELING IN THEBALTIC BASIN

L. Phil Graham

Stockholm 2000

Doctoral ThesisDivision of Hydraulic EngineeringDepartment of Civil and Environmental EngineeringRoyal Institute of Technology

© 2000 L. Phil Graham

Cover illustration: Subbasin map for the Baltic Basin

TRITA-AMI PHD 1033ISSN 1400-1284ISRN KTH/AMI/PHD 1033-SEISBN 91-7170-518-X

Printed at KTH Tryck och KopieringStockholm 2000

i

Abstract

Efforts to understand and simulate the global climate in numerical models have led toregional studies of the energy and water balance. The Baltic Basin provides an optimalregional test basin, whereby interaction between the sea and the atmosphere, theatmosphere and the land surface, and the land surface and the sea can be studied indetail. Understanding and modeling the large-scale hydrology of the Baltic Basin is animportant component in regional climate studies as it is conducted at the continentalscale where meteorology, oceanography and hydrology all can meet. Moreover,freshwater flow to the Baltic Sea plays an important role in the delicate ecologicalbalance of the sea.

Using a simple conceptual approach, a large-scale hydrologic model was set up tomodel the water balance of the total Baltic Sea Drainage Basin covering some1 600 000 km2 (HBV-Baltic). HBV-Baltic was then used to simulate the basinwidewater balance components for the present climate, to update river dischargeobservations, to evaluate the land surface components of atmospheric climate models,and to estimate potential impacts to water resources from climate change scenarios. Ithas been used extensively in cooperative BALTEX (The Baltic Sea Experiment)research, and it has become a standard tool within SWECLIM (Swedish RegionalClimate Modelling Programme) to support continued regional climate modeldevelopment. It is currently in use at SMHI (Swedish Meteorological and HydrologicalInstitute) as part of an operational system to produce near real-time river runoff.

Through these activities, HBV-Baltic has greatly improved the dialogue betweenhydrologic and meteorological modelers within the Baltic Basin research community.It is concluded that conceptual hydrologic models, although far from being complete,play an important role in the realm of continental scale hydrologic modeling.Atmospheric models benefit from the experience of hydrologic modelers in developingsimpler, yet more effective land surface parameterization. This basic modeling tool forsimulating the large-scale water balance of the Baltic Sea Drainage Basin is the onlyexisting hydrologic model that covers the entire basin and will continue to be useduntil more detailed models can be successfully applied at this scale.

Keywords: large-scale, continental scale, hydrologic modeling, Baltic Basin,atmospheric climate models, runoff, climate change, HBV, BALTEX

ii

Acknowledgements

The road to a doctoral thesis is not always direct. For me it goes at least as far back as1983 when I took my Bachelor of Science in Civil Engineering at Colorado StateUniversity (CSU). A Master’s followed in 1985 with the thesis, “Allocation of RiverBasin Water Supply Under Complex Water Rights and Interstate Agreements.” Theroad then took an abrupt turn as I traveled to Africa and spent some 2.5 years workingwith much simpler, yet more pressing aspects of water. Return to the USA plunged meback into more advanced (well, more abstract anyway!) methodologies for bothallocation of scarce water resources and determining flooding extent under extremeprecipitation. Yet another abrupt turn occurred when I moved to Sweden in 1992, notfor science but for love. Aside from learning a new language and a somewhat newculture, I returned to academia and a further broadening of my perspectives. This ledfirst to a Licentiate Degree at the Royal Institute of Technology (KTH) with the thesis,“Safety Analysis of Swedish Dams: Risk Analysis for the Assessment andManagement of Dam Safety.” Not completely satisfied with limiting my scope to theSwedish perspective, I welcomed the opportunity to extend my horizons even furtherto work on the hydrology of the entire Baltic Basin—and thus this thesis.

I learned many lessons along the way. From my work at CSU, I learned aboutscarcity and the value of water. In Africa, I learned about the all-essential need forwater. Under both my consulting work in the USA and my licentiate work at KTH, Ilearned about the power and danger of water. And now, within the Baltic Basin, I havelearned much about the interaction of water with land, the atmosphere, the ocean andthe world climate.

As for all of you that have helped me through … the list is long as this work couldnot have been possible without a large group of diverse players:

Klas Cederwall – you have always been available with the broader perspective ofwater resource problems and engineering solutions.

Sten Bergström – “Mr. HBV,” a lifetime of hydrological modelling experience,philosophical enlightenment, a catalyst, a mentor and a friend.

Anders Omstedt – you have kept me in perspective with the larger life of BALTEX,not to mention the Baltic Sea itself.

BALTEX doktoranderna (Anna, Lars & Ulrika) – we started together and we’velearned a lot, not just science either.

Bengt Carlsson and Lars Meuller – you performed the painstaking tasks ofassembling hydrological and meteorological databases.

Rossby Centre folk – enthusiasm, knowledge, comraderie and a commitment tosucceed.

SMHI-If – a wealth of knowledge at my fingertips.

Daniela Jacob – I’ll bet you’ve got another run that you’re just dying for me tolook at.

Anders Gyllander – my GIS guy.

Jörgen Nilsson – BALTEX doktoranderna’s champion.

iii

This list doesn’t want to stop! … Lennart Bengtsson, Zdzislaw Kaczmarek,NEWBALTIC colleagues, BALTEX colleagues, the BALTEX Secretariat, colleaguesat the Russian State Hydrological Institute, all the hydrological agencies in the BalticBasin that contributed to the river discharge database, HADLEY Centre,Max-Planck-Insitute for Meteorology, the German Climate Computing Centre,ECMWF, the Baltic Drainage Basin Project, everyone that has contributed over manyyears of development to the HBV model, and I know I am forgetting someone (sorry!).

Financing came from SMHI, SWECLIM (that’s through MISTRA, theFoundation for Strategic Environmental Research), the European Union(NEWBALTIC and NEWBALTIC II projects) and Naturvårdsverket (the SwedishEnvironmental Protection Agency).

A long time ago in my Master’s acknowledgements I expressed appreciation tomy parents for instilling in me the philosophy of “constantly confronting andcompleting challenging tasks.” After all this time I think I just have to repeat the sameacknowledgement again, but I think I might adjust my philosophy and leave out theword “constantly” this time. As for the rest of the family—both the American and theSwedish sides—it means a lot to me that you have actually traveled all this way to bewith me on the day of reckoning. (Listen well, there’ll be a quiz at the party!)

In Africa I found not only challenges, but also a loving wife. This formed the linkthat ultimately led to life in Sweden and research in the Baltic Basin. So Anna, youdeserve a fair share of the credit that this thesis came to be—not to mention yourpatience and perseverance! Lastly, I must thank Sofia—you have shown so muchpatience and understanding for one of your years. It cannot be easy to live with a fatherwho goes around muttering and grumbling about this subbasin, that river, data formats,scenarios and so forth!

Thus endeth the soliloquy.

Phil Graham Norrköping, 3 February 2000

iv

Preface

The thesis is based on the five publications listed below. These publications arereferred to as PAPER 1 to PAPER 5 and are appended to the thesis.

PAPER 1:

Bergström, S. and Graham, L.P., 1998. On the scale problem in hydrologicalmodelling. Journal of Hydrology 211, 253-265.

PAPER 2:

Graham, L.P., 1999. Modeling runoff to the Baltic Sea. Ambio 28, 328-334.

PAPER 3:

Graham, L.P. and Jacob, D., 2000. Using large-scale hydrologic modeling to reviewrunoff generation processes in GCM climate models. MeteorologischeZeitschrift/Contributions to Atmospheric Physics 1, 43-51.

PAPER 4:

Graham, L.P., 1999. Simulating climate change impacts on the water resources of theBaltic Basin. Nordic Hydrology (submitted).

PAPER 5:

Graham, L.P. and Bergström, S., 2000. Land surface modeling in hydrology andmeteorology – lessons learned from the Baltic Basin. Hydrology & Earth SystemSciences (in press).

v

Table of Contents

Abstract ..........................................................................................................................i

Acknowledgements ...................................................................................................... ii

Preface ..........................................................................................................................iv

List of Abbreviations ...................................................................................................ix

1. Introduction ..........................................................................................................1

1.1 Background.........................................................................................................11.2 Objectives ...........................................................................................................21.3 Interplay Between Models and Databases..........................................................31.4 Thesis Organization............................................................................................4

2. The Baltic Basin ....................................................................................................5

2.1 Description..........................................................................................................52.2 Model Grid Perspective ......................................................................................6

3. Databases ...............................................................................................................9

3.1 Synoptic Observations ........................................................................................93.2 River Flow ..........................................................................................................93.3 Topography and Land Use................................................................................11

4. HBV-Baltic ..........................................................................................................13

4.1 Introduction to HBV.........................................................................................144.2 Snow .................................................................................................................164.3 Soil Moisture ....................................................................................................164.4 Evapotranspiration............................................................................................17

5. Water Balance Modeling....................................................................................19

5.1 Simulations with HBV-Baltic...........................................................................195.2 Water Balance Components .............................................................................19

6. Climate Model Evaluation .................................................................................23

6.1 Hydrologic and Meteorological Approaches....................................................236.2 Evaluation of Climate Models ..........................................................................256.3 Interpretation of Climate Model Evaluation.....................................................30

7. Water Resources Scenarios for Climate Change.............................................33

7.1 Scenario Summary............................................................................................337.2 Scenario Response and Considerations ............................................................35

8. Discussion ............................................................................................................39

9. Conclusions..........................................................................................................45

10. The Future...........................................................................................................47

References....................................................................................................................49

vi

vii

List of Figures

Figure 1. Interplay between atmospheric climate models, hydrologic models anddatabases in the Baltic Basin. ........................................................................ 3

Figure 2. The Baltic Sea Drainage Basin........................................................................ 5

Figure 3. Representative grid resolutions for climate models over the Baltic Basin for2.5° (typical for GCMs), 0.8°, 0.4° and 0.2°.. ............................................... 7

Figure 4. Synoptic observation stations for the Baltic Basin.. ..................................... 10

Figure 5. Basin boundaries for HBV-Baltic.. ............................................................... 13

Figure 6. Schematic view of the HBV model showing subbasin division, snowdistribution, elevation and vegetation zones, unsaturated and saturatedzones, and river and lake routing................................................................. 14

Figure 7. Soil moisture parameters of HBV for 56 catchments in Sweden and theBaltic Basin plotted against basin size from 7.3 to 144 000 km2. ............... 18

Figure 8. HBV-Baltic model performance.. ................................................................. 20

Figure 9. The water balance of the Baltic Sea Drainage Basin – inputs and outputsfrom HBV-Baltic. ........................................................................................ 21

Figure 10. Schematic view of typical hydrologic and meteorological approaches tosurface parameterization, shown for one subbasin and one grid square,respectively.. ................................................................................................ 24

Figure 11. Modeled snow water equivalent (mm) over the total Baltic Sea DrainageBasin, where direct climate model output (ECHAM4, RCA0, RCA88-H andRCA88-E) is compared to output from HBV-Baltic with climate modelforcing (HBV-ECHAM4, HBV-RCA0, HBV-RCA88-H andHBV-RCA88-E).. ........................................................................................ 26

Figure 12. Modeled soil moisture deficit (mm) over the total Baltic Sea DrainageBasin, where direct climate model output (ECHAM4, RCA0, RCA88-H andRCA88-E) is compared to output from HBV-Baltic with climate modelforcing (HBV-ECHAM4, HBV-RCA0, HBV-RCA88-H andHBV-RCA88-E).. ........................................................................................ 27

Figure 13. Modeled evapotranspiration (mm/month) over the total Baltic Sea DrainageBasin, where direct climate model output (ECHAM4, RCA0, RCA88-H andRCA88-E) is compared to output from HBV-Baltic with climate modelforcing (HBV-ECHAM4, HBV-RCA0, HBV-RCA88-H andHBV-RCA88-E). ......................................................................................... 28

viii

Figure 14. Modeled runoff generation (mm/month) over the total Baltic Sea DrainageBasin, where direct climate model output (ECHAM4, RCA0, RCA88-H andRCA88-E) is compared to output from HBV-Baltic with climate modelforcing (HBV-ECHAM4, HBV-RCA0, HBV-RCA88-H andHBV-RCA88-E). ......................................................................................... 29

Figure 15. Average precipitation and temperature for the total Baltic Sea DrainageBasin from four atmospheric model runs (10-year present climatesimulations) and from synoptic data (HBV-Baltic base condition1981-1998)................................................................................................... 32

Figure 16. Total average annual river discharge to the Baltic Sea for observations,HBV-Baltic base condition and HBV-Baltic climate change scenarios...... 35

Figure 17. Average daily modeled river discharge for the Baltic Basin fromHBV-Baltic base condition (1981-1998), Today, and HBV-Baltic withRCA88-H Scenario perturbed forcing over 18 years.. ............................... 36

Figure 18. Average daily modeled river discharge for the Baltic Basin fromHBV-Baltic base condition (1981-1998), Today, and HBV-Baltic withRCA88-E Scenario perturbed forcing over 18 years. ................................. 37

ix

List of Abbreviations

BALTEX – Baltic Sea Experiment

Baltic HOME – Hydrology, Oceanography and Meteorology for theEnvironment - an interdisciplinary systems approach at SMHI

DCW – Digital Chart of the World

ERA – ECMWF Re-Analysis Project

ECHAM4 – European Climate Model - Hamburg (GCM) version 4 (at MPI)

GCM – General Circulation Model

GEWEX – Global Energy and Water Cycle Experiments

HADCM2 – Hadley Centre Climate Model (GCM) version 2(part of UKMO)

HBV – Swedish Hydrologic Model(the name emanates from “Hydrologiska ByrånsVattenbalansavdelning” or “Hydrological Bureau Water balanceSection,” the research unit of SMHI where it was first developedin the 1970s)

HBV-96 – current version of HBV model(extensively tested and updated with new components in 1996)

HBV-Baltic – HBV model for the entire Baltic Sea Drainage Basin

MPI – Max-Planck-Institute for Meteorology

NEWBALTIC – Numerical Studies of the Energy and Water Cycle of the BalticRegion Project (an EU project)

PILPS – Project for Intercomparison of LandsurfaceParameterization Schemes

RCA – Rossby Centre Regional Atmospheric Climate Model(part of SWECLIM, located at SMHI)

RCA0 – RCM model run with the 1st version of RCA, 44 km gridresolution, driven by HADCM2 GCM results at the boundaries

RCA88-E – RCM model run with the 2nd version of RCA, 88 km gridresolution, driven by ECHAM4 GCM results at the boundaries

RCA88-H – RCM model run with the 2nd version of RCA, 88 km gridresolution, driven by HADCM2 GCM results at the boundaries

RCM – limited area regional atmospheric model

x

SMHI – Swedish Meteorological and Hydrological Institute

SMHI-If – Research and Development Section of SMHI

SWECLIM – Swedish Regional Climate Modelling Programme

UKMO – United Kingdom Meteorological Office

UNEP/GRID – United Nations Environment ProgrammeGlobal Resource Information Database

WCRP – World Climate Research Programme

WMO – World Meteorological Organization

1

1. Introduction

1.1 BackgroundUnderstanding the global climate is a task of daunting complexity that relies heavilyon the use of numerical models. In order to represent the important physical processesof both the energy and water budgets, elements of meteorology, oceanography andhydrology must be included. The three scientific disciplines have traditionallydeveloped their own specific models to address relevant questions of interest withinthe respective discipline. It is now commonly recognized that these different types ofmodels must be combined in a rational effort to resolve the global climate throughcoupled models (IPCC, 1996).

Hydrology is a critical link between meteorology and oceanography. Allhydrologic textbooks begin with a diagram showing the hydrologic cycle. In thesimplest of terms, water that evaporates from the oceans to the atmosphere eventuallyprecipitates to hydrologic drainage basins to form the runoff that again flows back tothe ocean. This direct coupling to the ocean is a one-way event as there is no directlink from the ocean back to the land. In contrast, the interaction between the landsurface and the atmosphere is a two-way event of both water and heat exchange, as isthe coupling between the ocean and the atmosphere.

Rigorous representation of the global climate with coupled models constitutes amajor effort that present research and computer resources cannot fully satisfy. Oneway to reduce the problem to manageable levels is to first look closer at the physicalprocesses on regional scales. This is the approach adopted for the WCRP GlobalEnergy and Water Cycle Experiment (GEWEX) (WCRP, 1990). Under this program,five specific regions of the globe have been identified for detailed study andinternational research collaboration among leading scientists (IGPO, 2000). Focus onthese regions should deepen our scientific knowledge and create modeling solutions onthe continental scale that can be used for global modeling.

The Baltic Basin is the focus of one GEWEX sub-progamme—the Baltic SeaExperiment (BALTEX). The region is unique from the other GEWEX regions in that itcontains such a large inland sea (BALTEX, 1995). Outflow and inflow from the BalticSea to the world ocean occurs only at the southern channels of Öresund and TheDanish Straits. This provides a natural test basin of limited extent where interactionand exchange between the sea and the atmosphere, and between the atmosphere andthe land surface can be studied. As an ultimate long-term goal, BALTEX will improvethe capability to analyze both environmental problems and potential climate changeover the Baltic Basin by incorporating modeling from all three scientific disciplines.

Freshwater inflows of river runoff to the Baltic Sea play a pivotal role in the waterbalance of this water body (Gustafsson, 1997; Matthäus and Schinke, 1999; Omstedtand Rutgersson, 2000). It has long been noted that variations in salinity andtemperature can be related to both variations in river runoff and variations in the waterexchange with the North Sea (Hela, 1966). According to Håkansson et al. (1996), thefirst documented need for total river runoff data to the Baltic Sea is found in Pettersson(1893), where he presented the results of the first hydrographic survey of the Balticfrom 1879. As river runoff affects both the salinity distribution in the sea and transportof nutrients from land to the sea (Arheimer and Brandt, 1998; Omstedt and Axell,1998; Wulff et al., 1990), changes in river runoff can have significant impacts on thefragile ecological balance in the Baltic Sea. Hydrologic modeling is thus a critical

Large-Scale Hydrologic Modeling in the Baltic Basin

2

component for the successful modeling of environmental impacts for both the presentclimate and a changed climate.

Ongoing parallel to BALTEX is the Swedish Regional Climate ModellingProgramme (SWECLIM). This research program has a more specific mandate toproduce future climate scenarios (SWECLIM, 1998; SWECLIM, 2000). Although theprogram is primarily intended to serve Sweden, and secondarily the Nordic region, itmust also cover the whole Baltic Basin in order to include the strong effects that theBaltic Sea plays on the regional climate. An example is the ice cover on the Baltic Sea,which heavily influences the energy exchange between the sea and atmosphere. Todate, several different climate change scenarios have been produced.

Recent research has concentrated on improving the two-way interaction betweenthe atmosphere and the land surface in meteorological climate models, as documentedby numerous articles evaluating different land surface schemes (Koster and Milly,1997; Lohmann et al., 1998; Robock et al., 1998; Wood et al., 1998). Contrary toprevious research efforts, focus is no longer steered simply by improvements tometeorological outputs, but real improvement to the representation of hydrology inthese models is also sought. This in turn helps to close the one-way link between theland and the ocean. This implies that hydrologic, meteorological and oceanographicmodels must work together to resolve both the water and energy budgets.

Both meteorological and oceanographic models operate at scales that are muchlarger than those used within traditional hydrologic applications. Meteorologicalmodels have their foundation in representing the large-scale processes, covering notonly whole continents, but also the entire globe. Lack of adequate computing powerhas always been a limitation for reducing grid square size, their horizontal unit of area.This is steadily becoming less of a technological hindrance, and together with theincreased use of regional climate models, grid square size has recently decreasedsignificantly. Hydrologic models have their origin in operational applications whereadequate representation on the basin and subbasin scale defined model dimensions.Thus, the two modeling cultures have previously co-existed on two quite differentspatial scales. As a start to bridging the scale gap between disciplines, large-scalehydrologic modeling has an important role within both BALTEX and SWECLIM.

1.2 ObjectivesAn important aim of this research was to model the water balance of the Baltic SeaDrainage Basin at a scale where both hydrology and meteorology can meet. Workingat this scale, the specific objectives were as follows:

� Set up and validate a large-scale hydrologic model for the Baltic Basin andcarry out water balance simulations,

� Use the water balance model to evaluate runoff generation processes inatmospheric climate models and provide feedback for model development,

� Use the water balance model to provide river runoff inputs to ocean models,particularly for periods where river discharge observations are not available,

� Use the water balance model to simulate regional impacts of climate change inthe Baltic Basin,

� Use the water balance model and knowledge gained from simulations forqualified discussion and recommendations on the harmonization of hydrologicand meteorological models.

Introduction

3

These objectives are limited to a one-way coupling between hydrology andmeteorology and do not include development of a two-way coupled modeling system,although some improved parameterization in atmospheric modeling has resulted.

An existing, well-established hydrologic model—HBV—that is widely used in theNordic regions for traditional hydrologic applications was used to carry out theseobjectives. The resulting HBV Baltic Basin Water Balance Model is henceforthreferred to as HBV-Baltic.

1.3 Interplay Between Models and DatabasesAs a precursor to harmonized, fully coupled models operating together on-line,hydrologic and meteorological models are presently linked to each other through anoff-line network of different models and inputs. This network relies heavily on existingdatabases of observations and physiographical data. Figure 1 shows the network ofinterplay between atmospheric climate models, hydrologic models and databases in theBaltic Basin. The organization of the figure is as follows:

� pointers – initial conditions for climate model runs,� boxes – models and model runs,� ovals – databases,� stars – important outputs,� arrows – directional links.

GCMControl

Run

GCMScenario

Run

RCMControl

Run

RCMScenario

Run

ModifiedClimate

Database

Calibrationof HBV

ClimateDatabase

Runoff Databases

ClimateModel

Evaluation

WaterResourcesScenarios

Present Climate

Future Climate

WaterBalance Modeling

HBVRuns

PhysiographicalDatabases

Figure 1. Interplay between atmospheric climate models, hydrologic models and databases inthe Baltic Basin.


4

This figure may take some study before the paths of its arrangement becomeapparent. As an example, starting with present climate conditions, initial conditions areintroduced to a global atmospheric general circulation model (GCM) of coarseresolution to produce a control run of the present global climate. Results from this arein turn used as boundary conditions for a limited area regional atmospheric climatemodel (RCM; i.e. limited to a specific region of the globe) of finer resolution (Papers 4and 5). RCM results are then fed into a hydrologic model for further evaluation, eitherdirectly, or as part of a perturbed climate database for climate change scenarios. Finalresults ultimately end up in the yellow stars, for this case either as evaluation studies(Paper 3), or in combination with other links, as water resource scenarios (Paper 4).

Important elements of the basinwide runoff generation processes are simulated bythe large-scale hydrologic model driven with present-day climatological data andcalibrated against runoff observations (Papers 1 and 2). More detail and clarity on thisinterplay between models and databases is given in forthcoming chapters of the thesisand the reader is encouraged to refer back to this figure as the links are furtherexplored.

1.4 Thesis OrganizationFigure 1 presents a schematic overview of many of the important elements of ongoingclimate and climate change research, and shows where large-scale hydrology fits intothe picture. The rest of the thesis is organized around this central figure as more detailon how large-scale hydrology interacts with databases and climate models is described.First, more background on the Baltic Basin is presented. Then, the different databasesare presented together in one chapter. An introduction to the HBV model applied to theentire Baltic Sea Drainage Basin (HBV-Baltic) follows. The three output types—waterbalance (Papers 1 and 2), climate model evaluation (Papers 3 and 5) and waterresources scenarios (Paper 4)—are presented in separate chapters of their own. This isthen summed up with discussion, conclusions and some words about the future.

5

2. The Baltic Basin

2.1 DescriptionRunoff to the Baltic Sea originates under varied geographic and land use patterns. Atotal land area of 1 729 000 km2 (including Kattegat, Öresund and the Danish Straits)contributes to flow generation—on average 15 310 m3s-1 or 480 km3yr-1 (Bergströmand Carlsson, 1994)—at the outflow to the North Sea.1 This does not include thesurface area of the Baltic Sea itself, 377 400 km2, which receives net precipitationcontributing to flow—on average 1990 m3s-1 or 60 km3yr-1 (Omstedt et al., 1997).2

Whereas the annual volume of runoff places this basin in the same class as large riverssuch as the Mississippi and the Mekong, its flow path through the largest brackishwater body in the world makes it unique. The delicate ecological balance of the BalticSea is highly influenced by the volume and quality of runoff flowing into it.

As shown in Figure 2, the Baltic Sea Drainage Basin includes catchment areas in14 nations—Belarus, Czech Republic, Denmark, Estonia, Finland, Germany, Latvia,Lithuania, Norway, Poland, Russia, Slovakia, Sweden and Ukraine. In total, 85 millionpeople live in this region, with the highest concentrations in the south; 64% live within

Figure 2. The Baltic Sea Drainage Basin. (Used with permission (BDBP, 2000))

1 The runoff average is based on the years 1950-1990.2 The Baltic Sea net precipitation average is based on the years 1981-1994.


6

the Baltic Proper Drainage Basin alone (Sweitzer et al., 1996). Figure 2 liberallyextends the basin boundary out to and including Kattegat, although in strict geographicterms, it ends just south and east of the Danish isles at the entrance to Öresund and theDanish Straits (Nordstedts, 1997; Oxford, 1995). This is also the boundary from anestuarine circulation point of view.

The north-south elongated basin stretches from latitudes above the Arctic Circleof more than 69° N to Central European latitudes less than 49° N. It is characterized byboreal forests in the north and agriculture in the south. The majority of forest lands liein Sweden and Finland, with most of the agricultural land in Poland. Some 6% of thebasin is covered by lakes, most of which are in Sweden, Finland and Russia. Thisincludes the two largest lakes in Europe, Lakes Ladoga and Onega, both in Russia. Thefive largest rivers in descending order are Neva, Vistula, Daugava, Neman, and Oder.There are mountains in the northwest (the Scandinavian Mountains) and in the south(the Carpathian Mountains).

The inland Baltic Sea is the largest brackish water body in the world, with avolume of some 21 200 km3. Nine of the watershed countries have coastlines along theBaltic. Over many years, this sea has received large quantities of pollutants(HELCOM, 1993; Stålnacke, 1996). River runoff from both heavily industrialized andagricultural countries poses a particular ecological threat—this brings with it oldpollutants and nutrients originating from deposits in both river sediments and the soilsof floodplains (Wulff and Niemi, 1992). Episodic saltwater inflows through Kattegatand the Danish Straits are a dynamic feature of the estuarine circulation within theBaltic Sea that is critical for the ecosystem, as they help maintain salinity and oxygenlevels (Sjöberg, 1992).

Hydropower production is highly utilized on many rivers in the Baltic Basin,particularly in northern Sweden where plentiful snowmelt from the mountainsseasonally flows eastward to the sea. Power plants are also common between theabundant lakes in Finland and downstream of the large lakes of Russia. Althoughhydropower production does not typically affect the annual flows on these rivers, it canchange the seasonal distribution of flows due to storage in times of high flow forrelease in times of low flow. This is particularly pronounced in Sweden with itsextensive network of manmade reservoirs (Carlsson and Sanner, 1996). It is lessnoticeable for the natural lakes in Finland and Russia where regulated storage issmaller and the power plants are primarily operated as run-of-river production. Forexample, the large Lake Saimaa in Finland is subjected to daily regulation, but theoperational rules are aimed at keeping the weekly total flows in accordance withnatural conditions as determined by model forecasts of the coming week’s naturaldischarge (Kuusisto, 1999).

2.2 Model Grid PerspectiveFrom a hydrologic modeling perspective, the continental scale of the Baltic Basin islarge. From a global perspective, it is only one region of many. This is apparent whenone examines the different horizontal resolutions in use for various models over thebasin. Figure 3 shows the Baltic Basin divided into different hydrologic subbasins andoverlain by grid systems of typical size for different atmospheric models.

The first grid scale shown in the figure—2.5° (~275 km)—represents theresolution of current GCMs. The remaining grid scales shown—0.8° (~88 km),0.4° (~44 km) and 0.2° (~22 km)—represent those in use for different resolutions ofRCM models. Considering that the modeled meteorological variables on the surfacefor each of these grid squares are constant over the whole grid square, one gets an idea

The Baltic Basin

7

Figure 3. Representative grid resolutions for climate models over the Baltic Basin for 2.5°(typical for GCMs), 0.8°, 0.4° and 0.2°. The grids show standard latitude/longitudecoordinates. Actual model domains would have different orientation and projection.


8

of how coarse or fine the Baltic Basin climate is represented in the respective models.The climate in global models is thus coarsely represented. The very idea behindregional models is to improve detail with improved resolution. The finer resolutions ofthe regional models should thus provide more surface variation and better climaterepresentation for applications such as hydrology.

9

3. Databases

3.1 Synoptic ObservationsSynoptic precipitation and temperature observations for the period 1979 through

1998 from the entire Baltic Basin are available in an interpolated 1 × 1 degree griddatabase at SMHI (Omstedt et al., 1997). Figure 4 shows both the synoptic stationdistribution and the interpolated synoptic station grid over the basin. This includesdaily observations from 700-800 stations. The number of stations is not exact as allstations do not reflect 100% recording over the period. A much more detailedprecipitation network exists in the Baltic Basin, but observations are not available forlong time periods. Rubel and Hantel (1999) used some 4200 gauge stations in theirwork, but this covered only a short 4-month period.

The synoptic data were gridded with a two-dimensional univariate optimuminterpolation scheme (Gustafsson, 1981), whereby the degree of spatial filtering foroptimum interpolation is determined by an isotropic autocorrelation function estimatedfrom the database. A subjective form of quality control from an experiencedmeteorologist was used to reject observations that appeared erroneous. Theinterpolated grid values were reduced to average values for each of the 25 subbasinsused in HBV-Baltic.

Parts of the former Soviet Union exhibit patterns of data inhomogeneity in theform of an increasing trend in measured precipitation amounts after 1986. While notinvestigated in depth, this coincides with a reported change in precipitation gagingmethods at that time (personal communication with H. Alexandersson, January 1997,SMHI, Norrköping). To compensate for this, correction factors were applied toprecipitation in the Neva River Basin for the period 1981-1986. Thus, these subbasinsare calibrated to gage measurements of the more recent period, assumed to reflectcurrent sampling, which should be compatible with incoming new synoptic data.

Similar data inhomogeneity patterns were exhibited in Polish subbasins. Asdescribed above, correction factors were applied to adjust calibration of basinparameters to be in tune with the most recent period measurements. Even with theseadjustments, precipitation data from these basins were suspect. Independent checks bycolleagues in Warsaw confirmed suspicions that our available synoptic data set forPoland does not appear very reliable (personal communication with Professor Z.Kaczmarek, February 1997, Polish Institute of Geophysics, Warsaw). As newprecipitation and temperature data are scheduled to come in from the Polish authoritiesunder future BALTEX studies, further effort toward identifying the error sources wasnot pursued. Until then, the model will be used with the realization that results fromthese basins are not as reliable as for the rest of the Baltic Basin.

3.2 River FlowAs outlined in Paper 2, 11 years of observed monthly river discharge were availablefor model calibration and verification for the years 1981-1991. These river runoffrecords are from all available measurement stations on rivers flowing into the BalticSea. This accounts for 86% of the total drainage area. The remaining 14% of drainagearea outside the network of flow measurements consists of coastal zones locatedbetween river mouths. Estimation of runoff from these areas came from specific runoffcalculations using representative neighboring stations (Bergström and Carlsson, 1994).


10

Figure 4. Synoptic observation stations for the Baltic Basin. Shown are the stations on line for7 December 1999.

Databases

11

From the viewpoint of climate modeling and coupling to atmospheric models,natural river discharge is desired over observed regulated river discharge. This is anartificial discharge record that closely resembles river flow that would occur if thereservoirs were not in place. It is obtained by adding and subtracting reservoir storageand release records from actual river discharge observations to take away the effects ofreservoirs from the record. This is necessary for rivers where reservoir storage isconsiderable, as atmospheric models have no provisions for reservoir storage routing.

For Sweden and northern Finland, artificial records of natural river discharge wereavailable up through 1991 for the rivers in the north where the effects of hydropowerare most predominant (Carlsson and Sanner, 1996). They are not available for othercountries in the basin, but they may not be as important there. As mentioned inChapter 2, the effect of regulation on the large Lake Saimaa system in southernFinland is constrained to follow weekly natural discharge, so it does not generallyshow up in the monthly observations. No specific information on regulation effectswas available for Lake Onega in Russia; Lake Ladoga is not regulated (personalcommunication with the State Hydrological Institute, St. Petersburg). Likewise, damsare known to exist on the Vistula River in Poland, but no specific information as to theextent of their effect on river flow was available.

The modeling presented in this research is based on natural river discharge forsubbasins in the Bothnian Bay and Bothnian Sea drainage basins, and flows of recordfor the rest of the Baltic Basin. There is a slight inconsistency introduced by this, but itshould not be significant, as the variations introduced by run-of-river power productionare not typically noticeable in the monthly flow records used for calibration (ascompared to daily flows, where available). An exception is Lake Onega in Russia,where the effects of regulation were apparent in the record, but no additionalinformation was available to calculate natural discharge.

From the viewpoint of oceanographic modeling, simulation of actual recordeddischarge—including the effects of regulation—to the Baltic Sea is of primary interest.An altered version of the model is needed for this case, whereby the Swedish andFinnish natural river discharge records are replaced by actual recorded discharge andreservoir storage routing operations are included for these subbasins. This affects onlysubbasins in the Bothnian Bay and Bothnian Sea drainage basins as mentioned above.This modification is not included in the results presented here.

As a final note on river flows, it should be pointed out that observed records ofdaily river discharge are simply not available for the entire Baltic Sea Drainage Basin.Even up-to-date monthly flows are not available for the whole basin, although thissituation seems to be improving. To date, full basin coverage of monthly river flowdata does not extend beyond 1993.

3.3 Topography and Land UseTopography for the entire basin was summarized using the Digital Chart of the World(DCW) database available through the worldwide web (EROS, 1997). Hypsographiccurves for each subbasin were compiled with a vertical resolution of 100 m. The HBVmodel adjusts temperature and precipitation inputs in each subbasin according to thesehypsographic curves. Land use data came from GRID Arendal’s GIS (geographicinformation system) database over the Baltic Region (BDBP, 2000; Sweitzer et al.,1996). Land use classifications from this database were simplified to categories offorest, open land and water.


12

13

4. HBV-Baltic

Paper 2 introduces the Baltic Basin Water Balance Model—HBV-Baltic—includingdetails of calibration/verification. The paper presents model results up through 1994;this has since been extended to include up to the end of 1998. Figure 5 shows the 25subbasins used in the model. They range in size from 21 000 to 144 000 km2. The fivemain drainage basins are outlined—Bothnian Bay, Bothnian Sea, Gulf of Finland, Gulfof Riga and Baltic Proper. The total area of about 1 600 000 km2 is slightly smallerthan listed in Chapter 2 and shown in Figures 2, 3 and 4, as it does not include thedrainage basins to Kattegat, Öresund, or the Danish Straits. This more geographicallycorrect definition of the Baltic Basin (Nordstedts, 1997; Oxford, 1995) was adopted forHBV-Baltic at the outset, due to both a lack of available runoff data and the desire toavoid detailed hydrologic modeling for the relatively small area of Denmark.

Skagerak

BalticProper

Gulf of Finland

BothnianBay

BothnianSea

DanishStraits

Kattegat

Gulf ofRiga

Figure 5. Basin boundaries for HBV-Baltic. The five main Baltic Sea drainage basins areoutlined in black.


14

4.1 Introduction to HBVUse of the conceptual HBV Model (Bergström, 1976; Bergström, 1995) plays a keyrole in this research. As Papers 1, 2 and 5 present many of the details of HBV-Baltic,only an overview and some additional detail is given here. A schematic view showingdifferent components of the HBV model is presented in Figure 6. Additionaldiscussion in the context to comparisons with atmospheric models is presented inChapter 6.

Figure 6. Schematic view of the HBV model showing subbasin division, snow distribution,elevation and vegetation zones, unsaturated and saturated zones, and river and lake routing.

HBV-Baltic

15

One question that might be asked is, “why HBV?” The answer is that: 1) it hasshown to perform well in representing runoff processes in the Baltic Basin as well asother regions of the world, 2) it can be run with sparse data, 3) calibration can belimited to a small number of parameters, 4) it is computationally efficient, and 5) it isrelatively insensitive to scales. It has been applied to a wide range of applicationsincluding analysis of extreme floods, effects of land-use change, effects of climatechange, acidification of groundwater, nutrient transport and sediment transport(Arheimer, 1998; Brandt, 1990; Harlin, 1992; Lidén, 1999; Saelthun et al., 1998;Sandén and Warfvinge, 1992; Vehviläinen and Huttunen, 1997). To the abovearguments should be added the practical side that the HBV model was readilyavailable, together with the knowledge base of a group of experienced users, whichgreatly reduced initial startup and setup time.

There are other conceptual hydrologic models with similar characteristics to HBVthat could also have been used for this type of analysis. Some of them may well havegiven equally satisfactory results. However, a more physically-based model such asMIKE SHE (Refsgaard and Storm, 1995) was thought to require too much data andcomputational resources for the applications at hand. As the study objectives includednot only the setup of a water balance model for the entire Baltic Basin, but use of thismodel in several different applications, a more physically-based type of model wasdeemed as too time and data intensive. Such a model is more appropriate if a primaryobjective is, for example, to study in detail the spatial and temporal movement ofgroundwater. Although they offer other advantages over conceptual models, such asmore suitable feedback for direct coupling to atmospheric models, there are still manyissues to be resolved before physically-based models can be satisfactorily applied tosuch large scales as the Baltic (Beven, 1996; Refsgaard, 1998).

HBV-96 v4.4.1 is the most current version of HBV (Lindström et al., 1997). It hasbeen upgraded to make it more “distributed” than previous versions. The termdistributed refers to the degree of discretization that describes the terrain in the basin; itcan be applied equally to either physically-based or conceptual models. It is importantto point out that the differences between conceptual and physically-based hydrologicmodels are becoming more and more fuzzy as the two approaches tend to convergetoward each other. Conceptual models are becoming more physical at the same timethat physically-based models use conceptual approaches as a means to overcome lackof fully-distributed physical data (Beven, 1996; Refsgaard, 1996).

For the large subbasins of HBV-Baltic, it is probably more correct to use thedescription “semi-distributed.” This refers to the fact that the distribution of eachsubbasin into different elevations and land categories—forest, open land and water—isnot spatially fixed. That is, geographical information is taken from actual physicaldata, but it is represented in each subbasin as a percentage of the whole for thatsubbasin without keeping track of exactly where the percentage is located in space.Thus, model results, although integrated on their physical characteristics (e.g. elevationand land use), cannot be placed at a certain location within a subbasin. At present, ifone wants to present more detailed spatial results, the only option is to use moresubbasins. However, this doesn’t necessarily mean that runoff results at the outflow oflarge basins would be better with smaller subbasins.

It was known from the start that HBV could be used at relatively large scales(Carlsson and Sanner, 1996), but it then remained to be seen how well it wouldperform at continental scale. Paper 1, and to some extent Paper 2, discuss scaleproblems in hydrologic modeling. Both papers present arguments that HBV-Balticperforms satisfactorily at the Baltic Basin scale for the purpose of resolving the water


16

balance and its components. Since these papers have been published, independentresearchers working on large scales for the Elbe River have also shown it to performsatisfactorily at large scale (Krysanova et al., 1999).

4.2 SnowSnow is an important process for runoff generation for the Baltic Basin. In HBV, the2 m air temperature inputs for each subbasin are first adjusted by a standard lapserate—usually -0.6°C per 100 m increase—according to elevation zones in thesubbasin. Precipitation inputs are then modeled as snow or rain according to theadjusted temperature and certain threshold values. Snow accumulation thus builds upduring subfreezing periods. An improvement with HBV-96 is that the accumulation ofsnow is not evenly distributed within an elevation zone (Lindström et al., 1997). Astatistical distribution divides the total accumulating snowfall for each zone intodifferent statistical classes. This accounts for naturally occurring spatial variability,often due to blowing and drifting, which is particularly pronounced above tree line.

As with many hydrologic models, the simple degree-day approach is used forsnowmelt, supplemented by a liquid water-holding capacity that has to be exceededbefore runoff is generated. Although atmospheric modelers have criticized thedegree-day approach for lack of information on proper energy fluxes, it is stillrecognized within the hydrologic modeling community as an effective means ofproducing runoff from snow. This is documented by the WMO hydrologic modelingintercomparison (WMO, 1986). This reference is admittedly somewhat dated, yetcontrary evidence to its findings has not appeared in the literature. Confirmation of itsfindings, however, is available (Braun, 1984; Ferguson, 1999; Johansson et al., 1998;Rango and Martinec, 1995; Vehviläinen, 1992). Ferguson (1999) concludes withpredictions about the future of snowmelt runoff models, “… no one model willdominate the field in ten years’ time … For climate change applications,energy-balance approximations will be used but there is likely to be much debate overhow to distribute the necessary inputs and surface parameters, and how to parameterizesubgrid variability in snow cover.” The biggest problem with implementing the moretheoretically superior energy balance approach is the excessive input data required(Rango and Martinec, 1995).

4.3 Soil MoistureA primary point from Paper 1 is that the HBV soil moisture routine, based onvariability parameters, provides a way to treat the great heterogeneity of soils,regardless of scale. The three parameters controlling modeled soil moisture content,SM, are introduced in Paper 1 (fig. 1), together with representative values for theseparameters in different parts of Sweden (fig. 2). They are FC, model field capacity (ormore correctly, the maximum model soil moisture storage); β, a descriptor for thebehavior of the runoff coefficient according to modeled soil moisture; and LP, the limitfor potential evapotranspiration. Runoff generation, R, in HBV is a function of FC, βand SM as follows:

R/IN = (SM/FC)β (eq. 1)

where IN is infiltration to the soil (rainfall + snowmelt - interception3).

3 A routine for interception modeling is available in HBV-96, but it was not used in HBV-Baltic.

HBV-Baltic

17

Figure 7 shows a sampling of these parameters plotted against basin size rangingfrom 7.3 to 144 000 km2. This is based on a set of 56 catchments in Sweden and theBaltic Basin. The fact that there is no clear trend to these plots is evidence that theseempirical factors are relatively independent of scale. Paper 1 further discusses that thefactors that are more dependent on basin scale are those determining recessionparameters governing storage and routing, which are more basin specific. Recentliterature on other conceptual hydrologic models using a variability approach have alsoshown benefits of the simpler, conceptual approach over more theoretically detailedphysical models for large scales (Arnell, 1999; Kite and Haberlandt, 1999; Lobmeyr etal., 1999; Nijssen et al., 1997). Furthermore, results from the Project forIntercomparison of Landsurface Parameterization Schemes (PILPS) show nosignificant advantage for complex soil moisture schemes over simpler approaches(Pitman and Henderson-Sellers, 1998).

4.4 EvapotranspirationActual evapotranspiration in HBV, EA, is a function of potential evapotranspiration,EP, and the variability of the modeled soil moisture content, SM. The limit forpotential evapotranspiration, LP, modifies this relationship as shown in equations 2and 3, as well as in Paper 1 (fig. 1). LP typically ranges between 70 and 90 percent ofFC.

EA/EP = SM/LP SM < LP (eq. 2)

EA/EP = 1.0 SM >= LP (eq. 3)

There is a choice of different methods for estimating EP in HBV. At present,these include Penman, Priestly-Taylor and Thornthwaite type calculations (Burmanand Pochop, 1994; Eriksson, 1981; Gardelin and Lindström, 1997; Lindström et al.,1994). The last one is referred to as a Thornthwaite “type” of calculation because itresembles Thornthwaite’s approach, but it is not exactly the same as Thornthwaite’sequation. HBV-Baltic uses the Thornthwaite type of calculation, which is essentially asimple temperature anomaly method. Using KT, a calibrated model parameter, STF(t),a seasonal variation coefficient and T, the daily temperature, EP is calculated as,

EP = KT · STF(t) · T (eq. 4)

This method was used for two reasons. The first was that it gives anapproximation of potential evapotranspiration without requiring an extensive amountof data. The second is that as it provides a method for evapotranspiration to be adjustedas a function of temperature, it can be used for climate change scenarios. Morediscussion on the validity of this approach is included in Chapter 7.


18

β

012345

0 25000 50000 75000 100000 125000 150000

Subbasin Area (km²)

FC (mm)

0

100

200

300

400

0 25000 50000 75000 100000 125000 150000


LP (mm)

0

100

200

300

400

0 25000 50000 75000 100000 125000 150000


Figure 7. Soil moisture parameters of HBV for 56 catchments in Sweden and the Baltic Basinplotted against basin size from 7.3 to 144 000 km2.

19

5. Water Balance Modeling

5.1 Simulations with HBV-BalticResults from the calibrated HBV-Baltic model are presented in Papers 1 and 2. Paper 2presents results from the five main drainage basins, as well as for the total Baltic SeaDrainage Basin for the period October 1980 through December 1994. Figure 8 showsan update with results through December 1998. These results will hereafter be referredto as the HBV-Baltic base condition. The figure indicates periods for calibration,verification and extension of record. At publication date, the observed river dischargerecords for recent years were still not available. This is due both to lack of availabilityof actual observations from all of the different countries in the basin and lack ofadditional natural river discharge calculations from the northernmost basins (asdiscussed in Chapter 3).

As presented in Paper 2, HBV-Baltic performed equally well for both thecalibration and verification periods and achieved an overall efficiency criterion value,R2, of 0.83 for the total Baltic Basin (R2 ranges from 0 to 1, with 1 representing aperfect match). This is the well-known Nash/Sutcliffe efficiency criterion (Nash andSutcliffe, 1970) that rates model performance as a function of the initial variance inriver discharge observations to the variance in computed river discharge. Regionalvariation of R2 values range from 0.85 for the Bothnian Bay to 0.67 for the BalticProper as further documented in the paper. These R2 values are not exactly comparableto other reference basins, as only monthly observations were available for thecalculation with daily computed river discharge. For comparison, a strictly monthlycalculation of R2 (i.e. monthly observations vs. monthly computed) yields values of0.91, 0.95 and 0.73 for the total Baltic, Bothnian Bay and Baltic Proper, respectively.

5.2 Water Balance ComponentsThe daily modeled water balance components for the total Baltic Sea Drainage Basinup through 1998 are shown in Figure 9. This figure gives a quick view over the waterbalance conditions during the entire modeling period. As discussed in the papers,output parameters from the model that are not typically measured or known fromactual conditions are particularly interesting. These include snow water equivalent, soilmoisture deficit and evapotranspiration. They can be used in operational studiescovering the entire Baltic Basin, as a climatological database to be compared toatmospheric climate models and as a basis for comparison for climate change impactstudies. River discharge (here expressed as runoff depth), which is measured, is alsoquite important as observations are slow in coming in from the different politicalentities of the Baltic Basin. Thus, as synoptic station data is available long before riverobservations, HBV-Baltic is used to get initial estimates for total river inflows to theBaltic Sea.

The same information shown in Figure 9 is presented for each of the five majordrainage basins in Paper 2 (up through 1994). Recognizing that the water balanceoutput variables from HBV-Baltic represent only index type values over large basins,one can use them as relative measures of the runoff generation processes. Given thatextensive measurements are not taken, this is perhaps the best we can do—and we cando it with some confidence as these model processes have been validated in smallerresearch basins in previous studies (Andersson, 1988; Andersson and Harding, 1991;Brandt and Bergström, 1994; Lindström et al., 1997; Sandén and Warfvinge, 1992).


20

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

a) Bothnian Bay Drainage Basin

0

3000

6000

9000

12000

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

b) Bothnian Sea Drainage Basin

0

3000

6000

9000

12000

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

c) G ulf of Finland Drainage Basin

0

3000

6000

Riv

er D

isch

arg

e (m

³/s)

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

d) G ulf of Riga Drainage Basin

0

3000

6000

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

e) Baltic Proper Drainage Basin

0

3000

6000

9000

O bserved

M ode led

1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998

f) Total Baltic Sea Drainage Basin

0

5000

10000

15000

20000

25000

30000

35000

Ver ification Per iodC a libration Period E xtens ion

Figure 8. HBV-Baltic model performance. Periods for calibration, verification and extensionof record are indicated.

Water Balance Modeling

21

1 981 19 83 198 5 1987 1989 1991 1993 19 95 19 97

f) Runoff Depth [m m ]

012345

Total Baltic Sea Drainage Basin

O bservedM ode led

e) Soil M oisture Deficit (m m )

0

50

100

150

200

d) Snow W ater Equivalent [m m ]

0

50

100

150

200

250

c) M ean Daily Tem perature [°C]

-30-20-10

0102030

b) Daily Evapotranspiration [m m ]

0

5

1 981 19 83 198 5 1987 1989 1991 1993 19 95 19 97

a) Daily Precipitation [m m ]

0

5

10

15

20

25

Figure 9. The water balance of the Baltic Sea Drainage Basin – inputs and outputs fromHBV-Baltic. Precipitation and temperature are model inputs; evapotranspiration, snow waterequivalent, soil moisture deficit and runoff are model outputs. Runoff is the only variable thatis verified. (These are mean values over the entire drainage basin area.)


22

23

6. Climate Model Evaluation

6.1 Hydrologic and Meteorological ApproachesPaper 5 addresses the question of the representation of runoff generation processes inatmospheric climate models and how hydrologic modeling experience can benefitmeteorological modeling. It outlines some of the differences between how hydrologicmodels and atmospheric climate models treat the land surface. Figure 10 presents aschematic view of the principle processes involved and how they are typicallyrepresented in the two types of models. Some important details are listed below:

Hydrologic Approach Meteorological Approach

• large lakes are modeled explicitly, smalllakes are integrated into the saturated zone

• lake storage not included in water balance,lake surface may be included in energybalance

• each subbasin is divided into elevationzones

• one elevation for each grid square

• each elevation zone is divided into openland and forestland

• one vegetation type for each grid square orfractions of land cover may be used

• snow accumulation can be distributedstatistically in each elevation zone

• no snow distribution

• water is stored as interception, snow,capillary water in snow, soil moisture,groundwater and lakes

• water is stored as interception, snow, soilmoisture and groundwater

• flow from the saturated zone is routedthrough lakes and rivers

• no lateral flow routing

There are of course variations to this general comparison as there are many differentmodels around. The ECHAM4 atmospheric model, for instance, includes aspects of amore hydrologic-oriented approach to soil moisture (Dümenil and Todini, 1992).

In summary, the hydrologic approach is not very detailed in the vertical sense, butit is quite detailed in the horizontal. Such a model can be used on large scales whiletaking into account subgrid variability. Land surface treatment in meteorologicalmodels concentrates on vertical processes and pays little attention to either subgridvariability or the lateral flow of water to downstream subbasins (or grid squares). Themain difference, however, is the lack of need for an explicit energy balance simulationin many hydrologic applications. This is a key reason why it is possible to keep suchhydrologic models within a simple vertical structure, particularly with regard to timescales. Energy balance calculations require time steps of minutes, compared to themore robust daily time step often used in hydrologic water balance applications. Thisfactor alone, increases the complexity required for energy balance calculationsmanifold.

Of minor difference is that hydrologic models have traditionally used naturalsubbasins for horizontal boundaries, whereas atmospheric models use an evenlyspaced grid network. This is more a question of practicality and application than a realdifference. Hydrologic models can easily use square subbasins, as some modelscurrently do, but then one must always deal with the issue of deciding how todistribute runoff in grid squares that lie on the divide between natural subbasins. (Thesmaller the grid square, the smaller the impact this has.) For purposes of discussion,


24

one can compare the processes that occur in the atmospheric grid square with thosethat occur in the hydrologic subbasin—as done in Figure 10.

Soil Moisture 2

Soil Moisture 3

Upper Saturated Zone

Lower Saturated Zone

Lakes RiverDischarge

EvapotranspirationInterception

Gravity

Gravity

Soil Moisture

EvapotranspirationInterception

Runoff Generation

Diffusivity

DiffusivityGravity

Gravity

Runoff Generation

Soil Moisture 1

Snow

Snow

Runoff Generation

METEOROLOGICAL APPROACH

HYDROLOGIC APPROACH

Figure 10. Schematic view of typical hydrologic and meteorological approaches to surfaceparameterization, shown for one subbasin and one grid square, respectively. The hydrologicapproach is represented by the HBV Model.

Climate Model Evaluation

25

It is appropriate to point out here that there is sometimes confusion overterminology, particularly in terms of runoff. Thus far the term “runoff generation” hasbeen used without any specific definition. This refers to the instantaneous excess waterper surface unit—grid square or subbasin—without any translation or transformationfor either groundwater, lake and channel storage, or transport time. As in Paper 5, itcan be simply expressed as,

Runoff Generation = P – EA – ∆S (eq. 5)

where,P = precipitationEA = actual evapotranspiration

∆S = change in storage (snowpack, interception, soil moisture).

In traditional hydrologic terminology, this is “effective precipitation,” which isavailable for routing to the subbasin outlet and includes snowmelt. Its dimension istypically millimeters per unit time. This is equivalent to what meteorological modelersoften refer to simply as “runoff,” which they often divide into two parts, “surfacerunoff” and “deep runoff.” This should not be confused with “river runoff” or “riverdischarge,” terms used synonymously for the measured streamflow in a river channelat some downstream point in the subbasin. River discharge is usually expressed inunits of cubic meters per second.

6.2 Evaluation of Climate ModelsAs introduced in Figure 1, regional climate modeling typically consists of a globalGCM that in turn drives a regional RCM over a limited area of the globe. HBV-Balticis used as a tool for evaluation of these climate models as presented in Papers 3, 4and 5. This provides a way to use the runoff record in model development (i.e. throughthe calibration of HBV-Baltic). We use the hydrologic model to transfer backwardfrom river discharge to runoff generation, and look at other important model processes,such as snow and soil moisture, along the way. These studies consist of using climatemodel results as forcing for HBV-Baltic. The climate-model-forced HBV-Balticresults are then compared to corresponding results from the atmospheric models.Variations on this approach have been performed by other researchers (Kite, 1997;Liston et al., 1994).

The occurrence of runoff generation in a grid square is typically the end of thewater cycle in a meteorological model. Combining flows from grid squares and furtherlateral routing is usually not attempted. Thus, this off-line coupling to a hydrologicmodel also provides a way to estimate river discharge from climate model results.

As a complement to results presented in Papers 3 and 5, Figures 11 through 14show additional evaluation results for snow water equivalent, soil moisture deficit,evapotranspiration and runoff generation, respectively. Each figure presents acomparison between a climate model run and HBV-Baltic forced with the respectiveclimate model run for four different cases. The first two cases, ECHAM4 and RCA0,are those described in Papers 3, 4 and 5. The RCA0 model run is the first version ofRossby Centre Regional Atmospheric Model (RCA) at 44 km grid resolution andlateral boundary forcing from the HADCM2 global model. The ECHAM4 model runis a global model (GCM), which for this case was run on a higher resolution thannormal—100 km. The other two cases are newer results from RCA – second version.


26

RCA88-E

HBV-RCA88-E

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA88-E

0

50

100

150

200

S im u la tion Y ear

RCA88-H

HBV-RCA88-H

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA88-H

0

50

100

150

200

RCA0

HBV-RCA0

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA0

0

50

100

150

200

ECHAM4

HBV-ECHAM 4

HBV-Climate

01 02 03 04 05 06 07 08 09 10

ECHAM4

0

50

100

150

200

Total Baltic Sea Drainage Basin Snow W ater Equivalent (mm)

Figure 11. Modeled snow water equivalent (mm) over the total Baltic Sea Drainage Basin,where direct climate model output (ECHAM4, RCA0, RCA88-H and RCA88-E) is compared tooutput from HBV-Baltic with climate model forcing (HBV-ECHAM4, HBV-RCA0,HBV-RCA88-H and HBV-RCA88-E). HBV-Climate is the long-term average daily values fromthe HBV-Baltic base condition results (1981-1998) repeated for each year.


27

RCA88-E

HBV-RCA88-E

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA88-E

0

50

100

150

200


RCA88-H

HBV-RCA88-H

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA88-H

0

50

100

150

200

RCA0

HBV-RCA0

HBV-Climate

01 02 03 04 05 06 07 08 09 10

RCA0

0

50

100

150

200

ECHAM4

HBV-ECHAM 4

HBV-Climate

01 02 03 04 05 06 07 08 09 10

ECHAM4

0

50

100

150

200

Total Baltic Sea Drainage Basin Soil Moisture Deficit (mm)

Figure 12. Modeled soil moisture deficit (mm) over the total Baltic Sea Drainage Basin, wheredirect climate model output (ECHAM4, RCA0, RCA88-H and RCA88-E) is compared to outputfrom HBV-Baltic with climate model forcing (HBV-ECHAM4, HBV-RCA0, HBV-RCA88-Hand HBV-RCA88-E). HBV-Climate is the long-term average daily values from the HBV-Balticbase condition results (1981-1998) repeated for each year.


28

RCA88-E

HBV-RCA88-E

01 02 03 04 05 06 07 08 09 10

RCA88-E

0

25

50

75

100

125

150


RCA88-H

HBV-RCA88-H

01 02 03 04 05 06 07 08 09 10

RCA88-H

0

25

50

75

100

125

150

RCA0

HBV-RCA0

01 02 03 04 05 06 07 08 09 10

RCA0

0

25

50

75

100

125

150

ECHAM4

HBV-ECHAM 4

01 02 03 04 05 06 07 08 09 10

ECHAM4

0

25

50

75

100

125

150

Total Baltic Sea Drainage Basin Evapotranspiration (mm/mon)

Figure 13. Modeled evapotranspiration (mm/month) over the total Baltic Sea Drainage Basin,where direct climate model output (ECHAM4, RCA0, RCA88-H and RCA88-E) is compared tooutput from HBV-Baltic with climate model forcing (HBV-ECHAM4, HBV-RCA0,HBV-RCA88-H and HBV-RCA88-E).


29

RCA88-E

HBV-RCA88-E

01 02 03 04 05 06 07 08 09 10

RCA88-E

0

25

50

75

100

125

150


RCA88-H

HBV-RCA88-H

01 02 03 04 05 06 07 08 09 10

RCA88-H

0

25

50

75

100

125

150

RCA0

HBV-RCA0

01 02 03 04 05 06 07 08 09 10

RCA0

0

25

50

75

100

125

150

ECHAM4

HBV-ECHAM 4

01 02 03 04 05 06 07 08 09 10

ECHAM4

0

25

50

75

100

125

150

Total Baltic Sea Drainage Basin Runoff Generation (mm/mon)

Figure 14. Modeled runoff generation (mm/month) over the total Baltic Sea Drainage Basin,where direct climate model output (ECHAM4, RCA0, RCA88-H and RCA88-E) is compared tooutput from HBV-Baltic with climate model forcing (HBV-ECHAM4, HBV-RCA0,HBV-RCA88-H and HBV-RCA88-E).


30

The second version of RCA included many parameterization improvements overthe first version (personal communication with Rossby Centre). Lateral boundaryforcing from two different GCMs—HADCM2 and ECHAM4—was used to producetwo sets of model runs, RCA88-H and RCA88-E. The GCM forcing for RCA88-H isexactly the same as used for RCA0. The GCM forcing used for RCA88-E, however, isnot from the same ECHAM4 run referred to above (but rather from a more typicalcoarse resolution around 300 km). The model domain consisted of two nested gridresolutions, first at 88 km resolution and then at 22 km resolution. Presented here areonly the results from the 88 km runs. These are control runs for the present climate,which implies that they can be compared against recent climatological data.

6.3 Interpretation of Climate Model EvaluationCharacteristics of the different climate models can be seen from the plotted resultsshown in Figures 11 through 14. All of these control climate runs are said to representthe present climate. This means that they have been driven with boundary conditionsderived from the present climate (e.g. sea surface temperatures, typically from the1980s), but they do not simulate a specific time period that can be directly compared toobservations or each other. For this reason, they cannot be compared directly to theHBV-Baltic base condition, other than to compare long-term average values. As a typeof climatic reference, Figures 11 and 12 include light dashed lines—HBV-Climate—that represent HBV-Baltic model runs with the base condition synoptic data; these aresimply the long-term average daily values from 1981-1998 repeated for each year.

Figure 11 shows daily results for snow water equivalent. Apparent from the plotsis that all the model cases show modeled snow values much lower than in HBV-Balticwith the same forcing. The snow season appears to get off to a good start early in thecold season, but the snowpack never reaches the same magnitude as in HBV-Baltic.This implies that snowmelt must be much higher for all cases.

Figure 12 shows daily results for soil moisture, expressed in terms of soil moisturedeficit (smd), which is the most relevant quantity to compare as it reflects theamplitude of change for soil moisture. Some models have absolute values for soilmoisture that are disproportionately large. Although this may be physically wrong, itmakes little difference if the volume is never used; it equates to a type of dead storage.Therefore, it is only the dynamical part of the soil moisture that is of interest,particularly when comparing model outputs.

Interpreting the smd plots with descriptive terms, ECHAM4 shows trends ofwetter soil conditions (low soil moisture deficit), whereas RCA0 shows considerablydrier conditions (high soil moisture deficit). Smd for both RCA88-H and RCA88-Ereflects soil moisture conditions that are very similar to HBV-Baltic results. Thegeneral seasonality in all cases appears to be reasonable, that is highest soil moisture inwinter and lowest in summer.

Figure 13 shows monthly results for evapotranspiration. This variable is perhapsthe most difficult to evaluate as there is so much uncertainty regarding its true value.Widespread regional measurements are not readily available and there is noestablished basis to adequately determine that one model does a better job than theother in simulating evapotranspiration. Common to all four comparisons is thatwintertime evapotranspiration is higher in the atmospheric models than in theHBV-Baltic. These differences in winter are not surprising as wintertimeevapotranspiration from snow in the HBV model is included in a general snowfallcorrection factor when the ground is covered by snow. It is thus not modeled explicitlyunder such periods and does not show up as evapotranspiration in model results.


31

Figure 14 shows monthly results for runoff generation. For all comparisons, theatmospheric models show considerably lower runoff than the hydrologic model. Innone of the cases do they reach the magnitudes of runoff apparent in HBV-Baltic. Thisis direct evidence that the land surface routines of climate models have a long way togo before runoff processes can be directly used for further analyses, such as climatechange impact studies to water resources. For the RCA0 model run, this was anexpected result as it was known that water was not conserved in the first version ofRCA. This was corrected for the RCA88 model runs. Some improvement can be seenfor both RCA88-H and RCA88-E. The ECHAM4 results show runoff generation thatcomes closest to the HBV-Baltic values, but the peaks occur a little earlier.

Of particular interest for this type of comparison between models is to see howchanges to model parameterizations affect results. Taking the results from an earlierversion of a model as a first iteration in continued model development, similaranalyses made after model changes can be compared back to the previous version orversions. As they were driven by the same lateral boundary conditions (i.e.HADCM2), this can be done for RCA0 and RCA88-H. Looking again at figures 11through 14 with this in mind, several observations can be made. There are no dramaticchanges in snow results, which is not surprising as the snow parameterization isexactly the same for the two model versions. Both smd and evapotranspiration,however, changed considerably, reflecting changes made to the land surface scheme.Furthermore, runoff increased for the RCA88-H case.

A final evaluation concerns the skill of the atmospheric models at representing thepresent climate. Figure 15 shows long-term precipitation and temperature averagesfrom the total Baltic Sea Drainage Basin for the four model cases, together with thesynoptic data used to drive the HBV-Baltic base condition (1981-1998). Apparentfrom the figure is that the climate models do not do a particularly good job atreproducing the seasonality of precipitation. In some cases the annual precipitation isquite good, but the seasonal patterns are wrong. All of the atmospheric model runsshow too much wintertime precipitation and the typical summertime peaks are mostlymissing. One should keep in mind that the base condition is a specific period of recordand model results should not be expected to match exactly. However, the trends shouldbe similar.

Annual temperature results from the climate models all lie either similar to thesynoptic data or colder. Seasonal results vary, but all of the climate models exhibit acold bias during winter. The combination of high winter precipitation and cold wintertemperatures implies that snow amounts should be higher than the present climate forall cases. This is however not the fact, as shown by HBV-Climate—the light dashedlines in Figure 11. All of the HBV-Baltic results with climate model forcing show theexpected trend of snow amounts that are considerably higher than HBV-Climate. Incontrast, all of the climate model results show snow amounts that are closer toHBV-Climate values, despite considerably higher precipitation and lowertemperatures. For RCA88-E, which has the highest annual precipitation of all themodel cases shown in Figure 15, snow amounts are consistently lower thanHBV-Climate. These results point to an inconsistency in the atmospheric snowroutines, which is further discussed in Chapter 8.


32

HBV-Base

ECHAM4

RCA0

RCA88-H

RCA88-E

1 2 3 4 5 6 7 8 9 10 11 12

M o n th

-10

-5

0

5

10

15

20 A vg. Seasona l (°C )

HB

V-B

ase

EC

HA

M4

RC

A0

RC

A88

-HR

CA

88-E

0

1

2

3

4

5

A vg. A nnua l (°C )

Tem perature

HBV-Base

HBV-Base (corrected)

ECHAM4

RCA0

RCA88-H

RCA88-E

1 2 3 4 5 6 7 8 9 10 11 12

M o n th

0

20

40

60

80

100 A vg. Seasona l (m m /m onth)

HB

V-B

ase

EC

HA

M4

RC

A0

RC

A88

-HR

CA

88-E

0

100

200

300

400

500

600

700

800

900

A vg. A nnual (m m /year)

Precipitation

Figure 15. Average precipitation and temperature for the total Baltic Sea Drainage Basinfrom four atmospheric model runs (10-year present climate simulations) and from synopticdata (HBV-Baltic base condition 1981-1998). The HBV-Base seasonal values shown areuncorrected; the HBV-Base annual bar shows both the uncorrected value (636 mm) and thevalue including corrections for aerodynamically induced undercatch (666 mm).

33

7. Water Resources Scenarios forClimate Change

7.1 Scenario SummaryProducing climate change impact scenarios for water resources can be a fairlystraightforward operation to perform. However, this does not mean that interpretationof the results is straightforward! There are many unanswered questions that arise. Firstand foremost are the questions of uncertainty. Paper 4 describes how scenarios wereapplied to the entire Baltic Basin with HBV-Baltic.

As described in the paper, perturbations to both 2 m temperature and precipitationfrom regional climate model scenarios were applied on a subcatchment basis to thefive main drainage basins in the Baltic Basin. Direct input of climate scenario forcingfrom atmospheric climate models is not currently advisable as the control runs forthese models still have a hard time matching climatic observations, primarilyprecipitation. This is shown in Chapter 6, Figure 15. In addition, Paper 4 shows thelarge discrepancy in modeled river discharge from direct forcing with climate modelresults for the present climate.

Use of perturbations in temperature and precipitation is currently the de factostandard treatment of climate change scenarios from climate models (Arnell, 1998;Kaczmarek et al., 1996; Lemmelä and Helenius, 1998; Lettenmaier et al., 1999;Saelthun et al., 1999; Vehviläinen and Huttunen, 1997). The long-term averagemonthly changes in 2 m temperatures were added directly to daily temperatures in theclimate database for the base condition (1981-1998). The long-term averageprecipitation changes were applied as percentage change from the present climate—these were filtered to 3-month seasonal values as there was large variation from monthto month. Values for these changes for RCA0 are presented in figures 4 and 5 ofPaper 4.

Two additional RCA climate model scenarios were evaluated in exactly the sameway and are presented below. They are from future climate scenario runs associatedwith the RCA88-H and RCA88-E model control runs described in Chapter 6. Thescenarios originate from different GCMs with different forcing, but they both representsimulation of an approximate doubling of the CO2 content in the atmosphere,compared to the present climate.

The scenario seasonal and average annual changes (temperature and precipitation)for the total Baltic Sea Drainage Basin are presented in Table 1 for the three modelcases. Table 2 shows how these perturbed climate changes affect annual river inflowsto the Baltic Sea, as modeled by HBV-Baltic. This is further illustrated in Figure 16,which shows the total average annual river discharge to the Baltic Sea for recordedflows, HBV-Baltic base condition flows and the three HBV-Baltic climate scenarios.Figures 17 and 18 show average daily modeled river discharge for the five main BalticSea drainage basins, together with extreme maximum and minimum values for theRCA88-H and RCA88-E scenarios, respectively. Although smoothed out byaveraging, these figures illustrate both the effects of the climate change scenarios onseasonal variation and give an overview of how extremes are affected. Flows to theGulf of Finland are heavily influenced by the many lakes in the region and particularlyby annual ice damming at the outflow from Lake Ladoga to the Baltic Sea. AlthoughHBV-Baltic does a reasonable job at representing this unique feature for the present


34

climate, changes to modify it for a warmer future climate were not made. Thus, riverdischarge from Lake Ladoga is likely misrepresented in these results. One wouldexpect to see minimized effects of ice damming as both the ice freezeup date shouldoccur later in the winter and the ice breakup date should occur earlier in the spring.

Table 1. 2 m temperature change (°C) and precipitation change (%) for the total Baltic SeaDrainage Basin summarized by season for RCA0, RCA88-H and RCA88-E climate changescenarios. The values reflect 10-year averages from climate model runs.

Scenario D,J,F M,A,M J,J,A S,O,N Annual

Temperature °C:RCA0 5.6° 3.8° 2.2° 3.7° 3.8°RCA88-H 4.2° 3.2° 2.2° 3.2° 3.2°RCA88-E 5.1° 2.6° 3.2° 4.2° 3.8°

Precipitation %:RCA0 16.3% 15.7% 36.2% 22.7% 22.7%RCA88-H 8.6% 12.0% 28.8% 20.4% 17.4%RCA88-E 1.4% 8.6% -10.1% 12.5% 3.1%

(D,J,F - December, January and February; M,A,M - March, April and May)(J,J,A - June, July and August; S,O,N - September, October and November)

Table 2. Modeled percent change (%) in annual freshwater inflow to the Baltic Sea for threeclimate change scenarios. Results shown are drawn from the difference between HBV-Balticbase condition and HBV-Baltic scenarios over the 18-year model period.

% BothnianBay

BothnianSea

Gulf ofFinland

Gulf ofRiga

BalticProper

TotalBaltic

RCA0:Maximum 34% 26% 36% 53% 4% 29%Minimum 10% -2% 3% 1% -16% 0%Average 21% 13% 19% 28% -5% 14%

RCA88-H:Maximum 28% 19% 21% 42% 2% 22%Minimum 8% -8% -4% 3% -13% -3%Average 19% 8% 9% 22% -5% 8%

RCA88-E:Maximum 10% -13% 12% -20% -35% -12%Minimum -9% -30% -16% -51% -61% -28%Average 1% -21% -7% -39% -54% -21%

Water Resources Scenarios for Climate Change

35

1920 1930 1940 1950 1960 1970 1980 1990 2000Ye ar

0

5000

10000

15000

20000

25000

(m3

/s)

RC A0 R CA88-H RC A88-E ----------------------------------------------------------

Fu ture C lim ate

Observations

HBV-Baltic

1921-1990 average

HBV-base 1981-1998 average

RCA0 average

RCA88-H average

RCA88-E average

Total Baltic Sea Drainage BasinAverage Annual River Discharge

Figure 16. Total average annual river discharge to the Baltic Sea for observations,HBV-Baltic base condition and HBV-Baltic climate change scenarios.

7.2 Scenario Response and ConsiderationsResults from the water resources scenarios show a clear trend of reduction of thespringtime runoff peak from snowmelt, as seen in Figures 17 and 18. This is mostpronounced in the northern basins of Bothnian Bay and Bothnian Sea where snowplays a dominant role in the hydrologic cycle. River discharge is thus generally moreuniform over the year for all basins, with higher flows in winter and lower flows insummer. Another feature apparent from the maximum extremes in Figure 17 is thechange in season for peak flows. Instead of occurring in the spring, they have beenshifted to fall for much of the basin. This can have a significant impact on localflooding, as during this period reservoirs for hydropower production are typically fulland would have less of a dampening effect than during the spring when they aretypically low. For the Baltic Proper, historic peaks in both late summer and wintershow higher values with the scenario, which could be interpreted as an increased riskfor flooding.


36

Jan Feb M ar A pr M ay Jun Ju l A ug S ep O ct N ov D ec


0

2500

5000

7500

10000

12500TodayScenario R C A88-H



0

2500

5000

7500

10000

Jan Feb M ar A pr M ay Jun Jul A ug S ep O ct N ov D ec

c) Gulf of Finland Drainage Basin

0

2500

5000

7500

10000

Riv

er D

isch

arg

e (m

³/s)


d) Gulf of R iga Drainage Basin

0

2500

5000

7500

10000



0

2500

5000

7500

10000

12500

Figure 17. Average daily modeled river discharge for the Baltic Basin from HBV-Baltic basecondition (1981-1998), Today, and HBV-Baltic with RCA88-H Scenario perturbed forcingover 18 years. Daily maximum and minimum values over the period are shown with blueshading for Today and pink shading for the Scenario.

Water Resources Scenarios for Climate Change

37



0

2500

5000

7500

10000

12500Today

Scenario RC A88-E



0

2500

5000

7500

10000


c) Gulf of Finland Drainage Basin

0

2500

5000

7500

10000

Riv

er D

isch

arg

e (m

³/s)


d) Gulf of R iga Drainage Basin

0

2500

5000

7500

10000



0

2500

5000

7500

10000

12500

Figure 18. Average daily modeled river discharge for the Baltic Basin from HBV-Baltic basecondition (1981-1998), Today, and HBV-Baltic with RCA88-E Scenario perturbed forcingover 18 years. Daily maximum and minimum values over the period are shown with blueshading for Today and pink shading for the Scenario.


38

Both Figure 18 and Table 2 show that the effects from the RCA88-E scenario areconsiderably different than those from RCA88-H, particularly in the southern basins.River discharge decreases significantly under this scenario, up to -54% annually in theBaltic Proper. With river discharge from the Bothnian Bay little changed at +1%, thischanges the distribution of freshwater inflow sources to the Baltic Sea. The RCA88-Hscenario also shows an increase in the contribution of freshwater from the north incomparison to the contribution from the south. Although there are many other aspectsthat should be considered, one outcome of such changes might be a reduction of theeffects of pollution loads from the more heavily industrialized south due to increasedmixing with purer waters from the north.

On the other hand, an increase in freshwater on the magnitude of the RCA88-Hscenario will also affect salinity in the Baltic Sea (Omstedt and Axell, 1998), whichcould have potentially threatening water quality consequences. Aside from simplyadding more freshwater volume, changes in river runoff may affect the major inflowsof salt rich waters to the Baltic Sea from the Atlantic through the Danish Straits(Matthäus and Schinke, 1999). Matthäus and Schinke (1999) suggest that increasedriver runoff during winter has a particularly strong effect.

The annual flows in Figure 16 clearly show the contrast between the threedifferent climate model scenarios, two indicate increased flow while the third shows asignificant decrease. From Table 2 these effects range from an average annual increasein total freshwater inflow to the Baltic Sea from 14% to a decrease of -21%. This widerange reflects the wide range of uncertainties for such studies.

One source of uncertainty is that between different climate models themselves. Inthis case, we have two different global models driving the same regional model withtheir respective different climate change emission scenarios. Another uncertainty isassociated with different versions of a particular climate model. In this case, the RCA0version of the RCA model was driven with exactly the same boundary conditions asthe RCA88-H version. Aside from other changes to the model, they were run on twodifferent spatial resolutions, which can have its own significant impact. In this case, itis assumed that the RCA88 version is an improvement over the RCA0 version as theimprovements included, among other things, conservation of the water cycle.However, this assumption is not a clear given when comparing one model version toanother—there will almost always be discrepancies between different versions of thesame model.

There is also uncertainty associated with the interface between the climate modelresults and the hydrologic impact assessment model (for this case, HBV-Baltic). A lotof smoothing occurs in the transfer of information between the models. More adequatetreatment of variability is needed. Using three-month running averages instead ofseasonal values for precipitation change would be one refinement. Differentprecipitation changes could also be derived from different classes of hydrologicconditions—for instance, dry, average and wet years—instead of applying the sameaveraged changes every year. Consideration of things like storm frequency andintensity should be factored in. Evapotranspiration is another weak factor. Thetemperature index method for evapotranspiration does not take into account increasedhumidity, which is likely under the combined conditions of increased temperature andincreased precipitation. Thus, this approach is admittedly simple and would benefitfrom enhanced representation of variability, frequency and extremes. This is furtherdiscussed in the next chapter.

39

8. Discussion

As outlined above, HBV-Baltic has been used to estimate water balance componentsfor the present climate, to update river discharge observations, to evaluate the landsurface components of atmospheric climate models, and to estimate potential impactsto water resources from climate change scenarios. It is currently in use at SMHI as partof an operational system, where it is driven by analyzed meteorological fields(Häggmark et al., 2000) to produce near real-time river runoff. This is part of aninterdisciplinary expert system to assess the state of the environment in the BalticBasin (SMHI, 2000). HBV-Baltic has been used extensively in cooperative BALTEXresearch, and it has become a standard tool within SWECLIM for support in continuedregional climate model development.

Moreover, HBV-Baltic has served to improve the dialogue between hydrologicand meteorological modelers within the Baltic Basin research community. Regardlessof the accuracy of the generated results, the analysis process itself served as a catalystto get both disciplines looking at land surface interface problems together. This is hardto measure in real terms, but the experience gained is invaluable.

The analysis also lead to other somewhat intangible results. Compensating errorsare a key obstacle in modeling, a dilemma that all complex models are subject to.Results from the climate model analyses showed evidence of compensating errorsbetween snow and precipitation for all model cases. That is, values for snow waterequivalent in the climate models lie closer to climatological values than the respectiveHBV-Baltic results with the same driving forces. At the same time, wintertimeprecipitation for the climate models is overestimated. Thus, the snow routines appeartuned to the present climate, regardless of the fact that the corresponding precipitationis mismatched with the present climate. This leads to a strong suspicion ofcompensating errors in the snow and precipitation components of the climate models.Experiments with better controlled boundary conditions, such as re-analysis products(Gibson et al., 1999), should lead to further internal model validation. This is still onlya hypothesis, yet it can be followed up if we proceed in a structured manner. (Thiswork is in progress.)

Although not presented in detail here, many of the changes made in the RCAversion 2 land surface parameterization scheme brought the representation of water atthe land surface/atmosphere interface into much closer agreement with methods usedby hydrologic modelers. This utilized the concept of variability parameters and wasdone with little added complexity to the atmospheric model. If one accepts the fact thatHBV performs reasonably well at representing the water balance, then changes to thewater balance components of climate models that come closer to HBV results with thesame driving forces ought to be improvements for the better. The change in resultsfrom RCA0 to RCA88-H, which used the same lateral boundary forcing, is positive. Inparticular, the modeling of soil moisture follows much closer the results fromHBV-Baltic with the same forcing (Figure 12). Evapotranspiration also showedimprovements in the right direction, as results from RCA0 were shown to be too lowwhen compared to both climatic calculations (Rummukainen et al., 1998) and theHBV-Baltic simulations. Runoff generation, although still far from the valuesproduced by HBV-Baltic did show improvement. Part of this is attributed to the factthat RCA0 did not close the water budget properly and water was actually lost fromthe system in that model version.


40

One might argue that the initial treatment of land surface processes in RCA0 wasso coarse that it could only be improved. This may be true, but if so, this is also truefor many other atmospheric climate models in use at present, even those that havehighly complex soil moisture schemes. This is shown in recent papers describing thework from the Project for Intercomparison of Landsurface Parameterization Schemes(PILPS) (Lohmann et al., 1998; Pitman and Henderson-Sellers, 1998). Anothercritique would be the argument that RCA88-H should be expected to respond morelike HBV-Baltic as soil moisture modeling concepts from HBV were introduced intoRCA. There is truth to this statement, although given the complex nature of climatemodels, there was no way to know from the outset that improvements would beimmediately obvious. This critique should be seen in a positive light. With RCA soilmoisture responding in a way that closely mimics a proven hydrologic model, we cannow focus on trying to localize other errors in the atmospheric model. For instance, wecan come back to the hypothesis of compensating errors and look more closely at snowprocesses.

Results from this work show that large-scale hydrologic modeling on the scale ofthe Baltic Basin can be done with relatively simple models such as HBV. This isconfirmed by (Krysanova et al., 1999). As recommended by (Refsgaard, 1998), wehave to choose a strategy that is proper for the problem we want to solve. The strategyof using HBV-Baltic as an integral part of the development process can take us quite along way toward eventually creating a new generation of harmonizedland-sea-atmospheric models that provide a balanced approach to climate modeling.As pointed out in Paper 5, we also need to keep in mind the concept of a complexitychain when putting model components together. We should avoid the irrational processof going from high levels of sophistication to lower levels of sophistication, and thenback to high levels of sophistication again. The details that are lost in the first step(high to low) can never be recovered in the second step (low to high), unless importantnew information is added.

Looking to the question of modeling water resources scenarios for climate change,we have to accept the fact that it may be some time before atmospheric climate modelscan rigorously represent the hydrologic variables that drive such analyses.Precipitation is the most important variable as the hydrologic cycle is so sensitive to it.Temperature (2 m) is also important, but climate models already do better at getting itright. Thus, the interface between climate models and hydrologic impacts assessmentmodels needs further development to improve the quality of climate change impactsmodeling. If we continue to extract the climate change signal as perturbations betweencontrol and scenario runs, we must improve our representation of variability,frequency and intensity from the scenarios. Direct use of outputs from climate modelsshould be an ultimate goal, even if precipitation continues to be problematic. Forexample, if the seasonality of precipitation in regional climate models improves, it maybe possible to use results directly with some adjustment to magnitude. Lastly,uncertainty between different climate models should be addressed by using the resultsof many different scenarios produced by different models and different drivingassumptions (e.g. different emission scenarios).

Coming back to basic hydrologic modeling, there are uncertainties from theHBV-Baltic base condition that must be considered. First, there is the issue of inputclimate data. The base condition is founded on databases from the specific period1980-1998. All subsequent comparisons were made to this period as if it adequatelyrepresents the present climate. Given the range of climate variability shown by long-term records, this is a relatively short period. However, this is the only period for

Discussion

41

which full data coverage for the entire Baltic Basin is available. Thus this period isused, recognizing that it includes natural biases. From annual runoff data presented inFigure 16, it is seen that this period is wetter than the long-term average. However,even this can be misleading, as it shows the overall trend for the full Baltic Basin, andsub-regional trends within the basin may be the opposite.

Data quality is always a concern. Regional differences in the quality of thesynoptic database were noted during the original calibration/verification procedure.These are discussed in Paper 2. In short, the precipitation records available for thisstudy, primarily for Polish regions, exhibit inconsistencies that affect the quality of thecalibration. Additional data from other precipitation stations have recently beenreceived that confirm this initial suspicion and pinpoint certain years where stronganomalies in the synoptic data occur. During the same period, significant errors in theriver discharge records for the Vistula River were recently discovered. It is thus nocoincidence that the poorest HBV-Baltic model performance occurs in just theseregions. Notations of these errors have been made, but a re-calibration of the modelwith the new data has yet to be performed.

Calculations for evapotranspiration are a definite source of error for this approach.The level of uncertainty for just this variable is high and efforts are underway forimprovements. For example, the effects of wintertime evapotranspiration are currentlybeing further investigated in ongoing research at SMHI (personal communication withthe hydrology research group at SMHI). This should lead to better representation ofwinter conditions in the HBV model and help reduce some of the uncertainty inevapotranspiration.

Treatment of evapotranspiration carries over into the analysis of climate changeimpacts on water resources. The simple temperature anomaly method used is likelybiased to the present climate. This is a case where improvements need to be made tothe simple conceptual approach. More physically-based methods need to be introducedin the calculation of potential evapotranspiration. As mentioned in Chapter 6, lack ofmeasurements for validation compound the problem. However, given the right data,estimates can be made for present climate conditions, but these may not hold for futureclimates. More direct incorporation of results from the climate models are our onlyrecourse to solving this problem. Such work is in progress within SWECLIM.

Another issue of concern for climate change impacts analysis is vegetation.Potential changes in vegetation density and/or water use efficiency may have aprofound impact on the future climate water balance. Representation of changes invegetation for future climate scenarios was not attempted here. This needs to beaddressed in more detail for future studies. Krysanova et al. (1999) point this out as anarea where HBV could be improved to better represent diversity of land cover suchthat parameterization of potential changes can be made. Although true in principle, itwill yet be a while before both the atmospheric community can provide us with moreaccurate climate model results, and the biologists have decided just how changes toclimate, such as increased CO2, will affect evapotranspiration rates. Till that time wemay have to restrict ourselves to sensitivity studies of the wide range of changes thatare possible. For example, the uncertainty on how forests will change is not only aquestion of physical processes, but also one of economics—which trees will be plantedto give the best economic return?

As with all models that attempt to represent the complexity of Mother Nature,there are model limitations that should be considered for HBV-Baltic. One is that itdoes not include explicit resolution of the energy balance. A more specificrepresentation of the energy balance would allow for more specific treatment of


42

evapotranspiration and in turn more direct feedback to atmospheric models. Longertime scales is a related limitation, especially when comparing to atmospheric models.The daily time step is both a result of traditional application and the degree-dayapproach used for snowmelt modeling. Increased time scale resolution would allow forboth more detailed comparison with atmospheric models and better possibilities forincluding the energy balance. However, at the same time, it could create complicationsfor the degree-day approach to snowmelt modeling. This typically uses daily timesteps and even 12-hourly time steps, but is not well tested for lower time steps. Forother internal processes, such as soil moisture, results at increased time resolutions arenot particularly useful and therefore an unnecessary burden on model efficiency.

Model parameter uncertainty and calibration are also limitations. A recent studyby Uhlenbrook et al. (1999) looks specifically at parameter uncertainty for HBV. Theyfound a wide range of calibrated parameters that can lead to a “good fit” for modelingriver discharge and maintain that this can have implications for model predictions.Although the uncertainties pointed out by Uhlenbrook et al. are a real concern, theymay not be so grave as long as the model is run within its range of calibration.Extrapolation beyond this range, as may be the case for climate change impact studies,leads to greater uncertainty. This problem is hard to overcome in the absence ofhydrologic models that are so sophisticated (or well developed from a physical point ofview) that they perform well without calibration or tuning (which is in practice notpossible to achieve!). Uhlenbrook et al. further maintain that more physical models,although seemingly more realistic than conceptual models, have their ownuncertainties related to scale and basin heterogeneity. They recommend that all studiesevaluating land-use or climate change effects should be accompanied by estimates ofparameter uncertainty.

Components from models such as HBV have a structure that, in a simple fashion,mimics the natural processes in nature. This structure is shown in both Figures 6and 10. Subsurface runoff processes are shown as boxes in these figures. Theserepresent fast and slow components of runoff in the hydrologic cycle. Howevertempting it may be, one should show restraint in further definition of these modelcomponents. Although they represent different types of flow, one cannot directlyassume, for instance, that the upper box accurately represents all near-surface quickgroundwater flow and the bottom box baseflow just because river discharge is wellsimulated by the model. This may be established with more detailed internal validationfor a specific basin, but due to parameter uncertainty as discussed above, it should notbe taken for granted.

Spatial distribution in the model is another limiting factor. HBV-Baltic uses largesubbasins for modeling runoff to the Baltic Sea. These subbasins are furtherdiscretized into elevation zones and land-use categories, but as mentioned inChapter 4, the specific horizontal location of these zones is not identified by the model.Due to this constraint, presentation of subbasin results is limited to basinwideaverages. As the origin of the elevation and land-use data is a GIS database, combiningthe model output with GIS operations is a potential way to get better spatialdistribution from HBV-generated results.

Spatial distribution also plays a role for input of precipitation and temperaturedata (Koren et al., 1999). Haberlandt and Kite (1998) showed that advancedinterpolation techniques and the use of additional precipitation information fromatmospheric models can improve large-scale hydrologic simulations in areas withsparse data. For HBV-Baltic, precipitation and temperature are initially input on asubbasin basis as average values over the subbasin applied at the average subbasin

Discussion

43

elevation. Variation due to elevation is introduced to the initial values by modificationin each elevation zone according to standard lapse rates. Nonetheless, the naturalvariation of precipitation and temperature is limited in this type of approach.However, Haberlandt and Kite also noted that “smaller subbasins are more sensitive toimprovements in precipitation estimates than larger ones, at least when modeling at themacroscale.” This may mean that errors due to this simplification may be of lesssignificance for larger subbasins. Nonetheless, consistency is maintained between thebase condition run and climate model runs as both treatment of the synoptic stationdata and the climate model data is identical. That is, both are input as average subbasinvalues derived from gridded values. (The only difference, as mentioned in Paper 3, isthat gauge corrections were applied to the synoptic data and not to the climate modeldata.)

Falling back on the original argument that the HBV model is well tested andvalidated, there is no reason to expect that HBV-Baltic cannot continue to be used forfuture applications to climate modeling in the Baltic Basin, given that we take intoaccount the limitations associated with the model and this approach. Some of theselimitations will be addressed in future research, others may not be resolvable (or willtake longer to resolve) and therefore must be considered as part of the uncertainty ofthe results. There is a long way to go before atmospheric climate models can on theirown adequately represent even the large-scale hydrologic cycle in the detail needed forboth impact studies and operational studies, much less more detailed analyses.However, the long-term aim should be to get rid of the need to use an HBV-typemodel in one-way coupling to the atmospheric models applied at this scale. Assuggested in Paper 5, a better solution is to build the hydrologic model into theatmospheric model. This does not mean simply coupling the two together, but ratherincorporating the concepts from hydrologic modeling into atmospheric modeling in aharmonized fashion, while trying to both avoid over-parameterization and keeping inmind the complexity chain for coupling of different components. Off-line use ofHBV-Baltic as an analysis tool should continue under this development process.


44

45

9. Conclusions

HBV-Baltic provides a simple modeling tool for evaluation of the large-scale waterbalance of the Baltic Sea Drainage Basin. Although it provides only index-type ofresults over large subbasins, it is the only existing hydrologic model that covers theentire Baltic Basin and will continue to be used until more detailed models can besuccessfully applied at this scale. In operational applications, it is particularly usefulfor updating freshwater inflow records to the Baltic Sea in lieu of unavailableobservations.

Two main conclusions from this research are, 1) conceptual hydrologic modelsplay an important role in the realm of continental scale hydrologic modeling, and 2)atmospheric models can benefit from the experience of hydrologic modelers indeveloping simpler, yet effective land surface parameterization. Within these twothemes, there are many more conclusions that can be drawn from this work, the mostimportant of these are as follows:

� Scale is a minor problem for this type of hydrologic model and application.

� The lack of energy balance parameterization is the main limitation in theconceptual hydrologic approach.

� Meaningful studies of atmospheric models are conducted with large-scalehydrologic models—this helps to identify inconsistencies in models (bothtypes!).

� Incorporation of variability parameters in the soil moisture modeling routinesof atmospheric models can lead to better soil moisture performance withoutadding unneeded complexity.

� It's relatively easy to compute water resource scenarios for future climates, butthe uncertainty is great and the interface between climate models and impactmodels needs improvement.

� More physically based evapotranspiration routines are needed, particularly formodeling future climate impact studies; they should include vegetationdynamics

Finally, other model and data limitations mentioned under Discussion should alwaysbe kept in mind when analyzing model results. This is sound modeling practice thatshould be applied regardless of the model used.


46

47

10. The Future

The future holds potential for continuation of this research in many different areas.There are lots of things that could have been done, but never happened due to timeconstraints, data constraints, lack of expertise, lack of manpower, or simply that theyonly became apparent at a late date. And then there are those things of broader scope—far beyond what could be achieved in one doctor’s thesis—that should be done to keepthe science on track. A list of relevant future activities follows:

� Practical HBV-Baltic Updates:

� Explicit river regulation for runoff to the Bothnian Bay and Bothnian Sea,� Investigation of river regulation impacts for other basins,� Analysis and re-calibration with new input data (Poland),� Extension of basin boundaries to include Skagerak, Kattegat and the Danish

Straits,� Temperature based treatment of Lake Ladoga ice damming.

� Research oriented activities:

� Spatial distribution of HBV-Baltic results from elevation zones using GIS,� Additional climate change scenarios and accompanying sensitivity studies,� Better impact studies interface between hydrologic model and atmospheric

model to increase representation of climate variability, extremes and land-usechanges,

� Enhanced distribution of rainfall input; e.g. possibly from climate model toelevation zones.

� Research already underway:

� Continued incorporation of hydrologic modeling into atmospheric models(Rossby Centre/SMHI-If),

� Analysis of ERA-forced control climate runs (SMHI/MPI – NEWBALTIC)� Use of more physical evapotranspiration routines with better vegetation

dynamics in HBV (SMHI-If),� Environmental impacts modeling (Baltic HOME).

Internal model validation is an area of high priority that spans many of theactivities listed above. Wherever possible we should strive to create and use suitablevalidation datasets for both hydrologic and atmospheric models alike. Processes thatneed particular attention are snow, evapotranspiration, interception, soil moisture andin some areas, freezing ground. And further down the road, groundwater flows overlarge scales should be used for validation of more physically based models.

Finally, one of the real challenges for the future is adequate assessment of waterquality, particularly in the face of daunting uncertainty. There are many things thatmodels cannot take into account. For instance, there is little we can confidently sayabout future demographic development and the technological changes that will affectland use in the Baltic Basin. If we look back in time, we realize that a lot can happenover 100 years that can completely change the way things are done in certain sectors,such as agriculture, forestry and energy production.


48

49

References

Andersson, L., 1988. Hydological analysis of basin behaviour from soil moisture data.Nordic Hydrology 19, 1-18.

Andersson, L. and Harding, R.J., 1991. Soil-moisture deficit simulations with modelsof varying complexity for forest and grassland sites in Sweden and the U.K.Water Resources Management 5, 25-46.

Arheimer, B., 1998. Riverine Nitrogen - Analysis and Modelling under NordicConditions. Doctoral Thesis, Department of Water and Environmental Studies,Linköping University, Linköping, 64(+papers) pp.

Arheimer, B. and Brandt, M., 1998. Modelling nitrogen transport and retention in thecatchments of southern Sweden. Ambio 27, 471-480.

Arnell, N.W., 1998. The effect of climate change on hydrological regimes in Europe: acontinental perspective. Global Environmental Change 9, 5-23.

Arnell, N.W., 1999. A simple water balance model for the simulation of streamflowover a large geographic domain. Journal of Hydrology 217, 314-335.

BALTEX, 1995. Baltic Sea Experiment BALTEX – Initial Implementation Plan. 2,International BALTEX Secretariat, GKSS Research Center, Geesthacht,Germany, 84 pp.

BDBP, 2000. Baltic Sea Region GIS, Maps and Statistical Database. Baltic DrainageBasin Project (BDBP: Beijer Institute, Dept. of Systems Ecology at StockhomUniversity; UNEP/GRID-Arendal). (http://www.grida.no/baltic/).

Bergström, S., 1976. Development and Application of a Conceptual Runoff Model forScandinavian Catchments. Doctoral Thesis, Department of Water ResourcesEngineering, Lund Institute of Technology, Lund University, Lund, 134 pp.

Bergström, S., 1995. The HBV Model. In: V.P. Singh (Editor), Computer Models ofWatershed Hydrology. Water Resources Publications, Highlands Ranch,Colorado, pp. 443-476.

Bergström, S. and Carlsson, B., 1994. River runoff to the Baltic Sea: 1950-1990.Ambio 23, 280-287.

Beven, K.J., 1996. A Discussion of Distributed Hydrological Modelling. In: M.B.Abbot and J.C. Refsgaard (Editors), Distributed Hydrological Modelling. WaterScience and Technology Library. Kluwer Academic Publishers, Dordrecht, pp.255-278.

Brandt, M., 1990. Human Impacts and Weather-Dependent Effects on Water Balanceand Water Quality in some Swedish River Basins. Doctoral Thesis, Department ofLand and Water Resources, Royal Institute of Technology, Stockholm,37(+papers) pp.


50

Brandt, M. and Bergström, S., 1994. Integration of field data into operationalsnowmelt-runoff models. Nordic Hydrology 25, 101-112.

Braun, L.N., 1984. Simulation of Snowmelt-Runoff in Lowland and Lower AlpineRegions of Switzerland. Doctoral Thesis, Swiss Federal Institute of Technology,Zurich, 157 pp.

Burman, R. and Pochop, L.O., 1994. Evaporation, Evapotranspiration and ClimaticData. Elsevier Science B.V., Amsterdam, 278 pp.

Carlsson, B. and Sanner, H., 1996. Modelling influence of river regulation on runoff tothe Gulf of Bothnia. Nordic Hydrology 27, 337-350.

Dümenil, L. and Todini, E., 1992. A rainfall-runoff scheme for use in the Hamburgclimate model. In: J.P. O´Kane (Editor), Advances in Theoretical Hydrology - ATribute to James Dooge. Elsevier, Amsterdam, pp. 129-157.

Eriksson, B., 1981. Den Potentiella Evapotranspirationen i Sverige (The PotentialEvapotranspiration in Sweden, in Swedish). SMHI RMK, Nr. 28, SwedishMeteorological and Hydrological Institute, Norrköping, 40 pp.

EROS, 1997. Digital Chart of the World (DCW) Global 30 Arc Second Elevation DataSet. EROS Data Center. (http://edcwww.cr.usgs.gov/landdaac/).

Ferguson, R.I., 1999. Snowmelt runoff models. Progress in Physical Geography 23,205-227.

Gardelin, M. and Lindström, G., 1997. Priestly-Taylor evapotranspiration in HBVsimulations. Nordic Hydrology 28, 233-246.

Gibson, J.K., Kållberg, P., Uppala, S., Hernandez, A., Nomura, A. and Serano, E.,1999. ECMWF Re-Analysis Project Report Series - 1. ERA description (Version2), European Centre for Medium-Range Weather Forecasts, Reading, UK, 72 pp.

Gustafsson, B., 1997. Dynamics of the Seas and Straits Between the Baltic and NorthSeas - a process-oriented oceanographic study. Doctoral Thesis, Department ofOceanography, Earth Sciences Centre, Göteborg University, Göteborg,22(+papers) pp.

Gustafsson, N., 1981. A Review of Methods for Objective Analysis. In: L. Bengtsson,M. Ghil and E. Källén (Editors), Dynamic Meteorology: Data AssimilationMethods. Applied Mathematical Sciences. Springer-Verlag, pp. 17-76.

Haberlandt, U. and Kite, G.W., 1998. Estimation of daily space-time precipitationseries for macroscale hydrological modelling. Hydrological Sciences 12,1419-1432.

Harlin, J., 1992. Hydrological Modelling of Extreme Floods in Sweden. DoctoralThesis, Division of Hydraulic Engineering, Department of Civil andEnvironmental Engineering, Royal Institute of Technology, Stockholm, 61(+papers) pp.

References

51

Hela, I., 1966. Secular changes in the salinity of the upper waters of the northern BalticSea. Commentat. Phys.-Math., Soc. Sci. Fenn. 31, 1-21.

HELCOM, 1993. Second Baltic Sea Pollution Load Compilation. Baltic SeaEnvironment Proceedings No. 45, Helsinki Commision, Helsinki, 161 pp.

Håkansson, B., Alenius, P. and Brydsten, L., 1996. Physical environment in the Gulfof Bothnia. Ambio Special Report 8, 5-12.

Häggmark, L., Ivarsson, K.-I., Gollvik, S. and Olofsson, P.-O., 2000. Mesan, anoperational mesoscale analysis system. Tellus 52A, 2-20.

IGPO, 2000. The GEWEX Homepage. International GEWEX Project Office (IGPO).(http://www.gewex.com/).

IPCC, 1996. Climate Change 1995 - The Science of Climate Change,Intergovernmental Panel on Climate Change, Cambridge University Press,Cambridge, UK, 572 pp.

Johansson, B., Edström, M., Losjö, K. and Bergström, S., 1998. Analys och beräkningav snösmältningsförlopp (Analysis and calculation of snowmelt). SMHI RapporterHydrologi, Nr. 75, Swedish Meteorological and Hydrological Institute,Norrköping, pp.

Kaczmarek, Z., Strzepek, K.M., Somlyódy, L. and Priazhinskaya, V., (eds.), 1996.Water Resources Management in the Face of Climatic/Hydrologic Uncertainties.Water Science and Technology Library. Kluwer Academic Publishers, Dordrecht,395 pp.

Kite, G.W., 1997. Simulating Columbia River flows with data from regional-scaleclimate models. Water Resources Research 33, 1275-285.

Kite, G.W. and Haberlandt, U., 1999. Atmospheric model data for macroscalehydrology. Journal of Hydrology 217, 303-313.

Koren, V.I., Finnerty, B.D., Schaake, J.C., Smith, M.B., Seo, D.-J. and Duan, Q.-Y.,1999. Scale dependencies of hydrologic models to spatial variability ofprecipitation. Journal of Hydrology 217, 285-302.

Koster, R.D. and Milly, P.C.D., 1997. The interplay between transpiration and runoffformulations in land surface schemes used with atmospheric models. Journal ofClimate 10, 1578-1591.

Krysanova, V., Bronstert, A. and Müller-Wohlfeil, D.-I., 1999. Modelling riverdischarge for large drainage basins: from lumped to distributed approach.Hydrological Sciences 44, 313-331.

Kuusisto, E. (Editor), 1999. Saimaa a Living Lake. Tammi Publishers, Helsinki,205 pp.


52

Lemmelä, R. and Helenius, N., (eds.), 1998. Proceedings of the Second InternationalConference on Climate and Water. Helsinki University of Technology, Espoo,Finland, August, 1998, 1676 pp.

Lettenmaier, D.P., Wood, A.W., Palmer, R.N., Wood, E.F. and Stakhiv, E.Z., 1999.Water resources implications of global warming: A U.S. regional perspective.Climatic Change 43, 537-579.

Lidén, R., 1999. A new approach for estimating suspended sediment yield. Hydrologyand Earth System Sciences 3, 285-294.

Lindström, G., Gardelin, M. and Persson, M., 1994. Conceptual Modelling ofEvapotranspiration for Simulations of Climate Change Effects. SMHI ReportsRH, No. 10, Swedish Meteorological and Hydrological Institute, Norrköping,25 pp.

Lindström, G., Johansson, B., Persson, M., Gardelin, M. and Bergström, S., 1997.Development and test of the distributed HBV-96 model. Journal of Hydrology201, 272-288.

Liston, G.E., Sud, Y.C. and Wood, E.F., 1994. Evaluating GCM land surfacehydrology parameterizations by computing river discharges using a runoff routingmodel: application to the Mississippi Basin. Journal of Applied Meteorology 33,394-405.

Lobmeyr, M., Lohmann, D. and Ruhe, C., 1999. An application of a large scaleconceptual hydrological model over the Elbe Region. Hydrology & Earth SystemSciences 3, 363-374.

Lohmann, D., Lettenmaier, D.P., Liang, X., Wood, E.F., Boone, A., Chang, S., Chen,F., Dai, Y., Desborough, C., Dickinson, R.E., Duan, Q., Ek, M., Gusev, Y.M.,Habets, P., Irannejad, P., Koster, R., Mitchel, K.E., Nasonova, O.N., Noilhan, J.,Schaake, J., Schlosser, A., Shao, Y., Shmakin, A.B., Verseghy, D., Warrach, K.,Wetzel, P., Xue, Y., Yang, Z.-L. and Zeng, Q.-C., 1998. The project forintercomparison of land-surface parameterization schemes (PILPS) phase 2(c)Red-Arkansas River Basin experiment : 3. Spatial and temporal analysis of waterfluxes. Global and Planetary Change 19, 161-179.

Matthäus, W. and Schinke, H., 1999. The influence of river runoff on deep waterconditions of the Baltic Sea. Hydrobiologia 393, 1-10.

Nash, J.E. and Sutcliffe, J.V., 1970. River flow forecasting through conceptual modelspart I – A discussion of principles. Journal of Hydrology 10, 282-290.

Nijssen, B., Lettenmaier, D.P., Liang, X., Wetzel, S. and Wood, E.F., 1997.Streamflow simulation for continental-scale river basins. Water ResourcesResearch 33, 711-724.

Nordstedts, 1997. Nordstedts Plus Svensk Ordbok + Uppslagsbok (NordstedtsReference Dictionary, in Swedish). Nordstedts Förlag AB, Göteborg, 1349 pp.

References

53

Omstedt, A. and Axell, L.B., 1998. Modeling the seasonal, interannual, and long-termvariations of salinity and temperature in the Baltic Proper. Tellus 50A, 637-652.

Omstedt, A., Meuller, L. and Nyberg, L., 1997. Interannual, seasonal and regionalvariations of precipitation and evaporation over the Baltic Sea. Ambio 26,484-492.

Omstedt, A. and Rutgersson, A., 2000. Closing the Water and Heat Cycles of theBaltic Sea. Meteorologische Zeitschrift/Cont. to Atmospheric Physics 1, 53-60.

Oxford, 1995. The Oxford English Reference Dictionary. Oxford University Press,Oxford, 1765 pp.

Pettersson, O., 1893. Den svenska hydrogafiska expeditionen 1877-II. Andraafdelningen (The Swedish hydrographic expedition of 1877, in Swedish). Kongl.Svenska Vetenskaps-Akadamiens Handlingar Band 25, N:o 1.

Pitman, A.J. and Henderson-Sellers, A., 1998. Recent progress and results from theproject for intercomparison of landsurface parameterization schemes. Journal ofHydrology 212-213, 128-135.

Rango, A. and Martinec, J., 1995. Revisiting the degree-day method for snowmeltcomputations. Water Resources Bulletin 31, 657-669.

Refsgaard, J.C., 1996. Distributed Physically-Based Modelling of the Entire LandPhase of the Hydrological Cycle. In: M.B. Abbot and J.C. Refsgaard (Editors),Distributed Hydrological Modelling. Water Science and Technology Library.Kluwer Academic Publishers, Dordrecht, pp. 55-67.

Refsgaard, J.C., 1998. Conceptual versus physically-based hydrological models: whichmodels to be used for BALTEX purposes? In: E. Raschke and H.-J. Isemer(Editors), Proceedings from the Second Study Conference on BALTEX. Rügen,Germany, May 25-29, pp. 180-184.

Refsgaard, J.C. and Storm, B., 1995. MIKE SHE. In: V.P. Singh (Editor), ComputerModels of Watershed Hydrology. Water Resources Publications, HighlandsRanch, Colorado, pp. 809-846.

Robock, A., Shlosser, C.A., Vinnikov, K.Y., Speranskaya, N.A. and Entin, J.K., 1998.Evaluation of AMIP soil moisture simulations. Global and Planetary Change 19,181-208.

Rubel, F. and Hantel, M., 1999. Correction of daily rain gauge measurements in theBaltic Sea Drainage Basin. Nordic Hydrology 30, 191-208.

Rummukainen, M., Räisänen, J., Ullerstig, A., Bringefelt, B., Hansson, U., Graham, P.and Willén, U., 1998. RCA - Rossby Centre Regional Atmospheric ClimateModel: Model Description and Results from the First Multi-Year Simulation.SMHI Reports RMK, No.83, Swedish Meteorological and Hydrological Institute,Norrköping, 76 pp.


54

Saelthun, N.R., Aittoniemi, P., Bergström, S., Einarsson, K., Jóhannesson, T.,Lindström, G., Ohlsson, P.-E., Thomsen, T., Vehviläinen, B. and Aamodt, K.O.,1998. Climate Change Impacts on Runoff and Hydropower in the NordicCountries. TemaNord 1998:522, Nordic Council of Ministers, Copenhagen,170 pp.

Saelthun, N.R., Bergström, S., Einarsson, K., Jóhannesson, T., Lindström, G.,Thomsen, T. and Vehviläinen, B., 1999. Potential impacts of climate change onfloods in Nordic hydrological regimes. In: P. Balabanis, A. Bronstert, R. Casaleand P. Samuels (Editors), Proceedings from the Ribamod - River BasinModelling, Management and Flood Mitigation Concerted Action - FinalWorkshop. Wallingford, UK, 26-27 February 1998, pp. 103-115.

Sandén, P. and Warfvinge, P., (eds.), 1992. Modelling Groundwater Response toAcidification. SMHI Reports RH, No. 5, Swedish Meteorological andHydrological Institute, Norrköping, 201 pp.

Sjöberg, B. (Editor), 1992. Sea and Coast. National Atlas of Sweden. SNA,Stockholm, 128 pp.

SMHI, 2000. Baltic Home - Hydrology, Oceanography and Meteorology for theEnvironment. Swedish Meteorological and Hydrological Institute.(http://www.smhi.se/sgn0102/n0205/baltichome/baltichome_en.htm).

Stålnacke, P., 1996. Nutrient Loads to the Baltic Sea. Doctoral Thesis, Department ofWater and Environmental Studies, Linköping University, Linköping,78(+papers) pp.

SWECLIM, 1998. Regional Climate Simulations for the Nordic Region - First Resultsfrom SWECLIM. SMHI Report, Swedish Meteorological and HydrologicalInstitute, Norrköping, 22 pp.

SWECLIM, 2000. Swedish Regional Climate Modelling Programme - Homepage.SMHI. (http://www.smhi.se/sgn0106/rossby/start1.htm).

Sweitzer, J., Langaas, S. and Folke, C., 1996. Land cover and population density in theBaltic Sea Drainage Basin: a GIS database. Ambio 25, 191-198.

Uhlenbrook, S., Seibert, J., Leibundgut, C. and Rodhe, A., 1999. Predictionuncertainty of conceptual rainfall-runoff models caused by problems inidentifying model parameters and structure. Hydrological Sciences 44, 779-797.

Vehviläinen, B., 1992. Snow Cover Models in Operational Watershed Forecasting.Doctoral Thesis, Faculty of Science, University of Helsinki, Helsinki, 112 pp.

Vehviläinen, B. and Huttunen, M., 1997. Climate change and water resources inFinland. Boreal Environment Research 2, 3-18.

WCRP, 1990. Scientific Plan for the Global Energy and Water Cycle Experiment(GEWEX). WCRP-40,WMO/TD-376, World Climate Research Program, Geneva,83 pp.

References

55

WMO, 1986. Intercomparison of Models of Snowmelt Runoff. WMO - No. 646,Secretariat of the World Meteorological Organization, Geneva, 36 & annexes pp.

Wood, E.F., Lettenmaier, D.P., Liang, X., Lohmann, D., Boone, A., Chang, S., Chen,F., Dai, Y., Dickinson, R.E., Duan, Q., Ek, M., Gusev, Y.M., Habets, P.,Irannejad, P., Koster, R., Mitchel, K.E., Nasonova, O.N., Noilhan, J., Schaake, J.,Schlosser, A., Shao, Y., Shmakin, A.B., Verseghy, D., Warrach, K., Wetzel, P.,Xue, Y., Yang, Z.-L. and Zeng, Q.-C., 1998. The project for intercomparison ofland-surface parameterization schemes (PILPS) phase 2(c) Red-Arkansas RiverBasin experiment : 1. Experiment description and summary intercomparisons.Global and Planetary Change 19, 115-135.

Wulff, F. and Niemi, Å., 1992. Priorities for the restoration of the Baltic Sea - ascientific perspective. Ambio 21, 193-195.

Wulff, F., Stigebrandt, A. and Rahm, L., 1990. Nutrient dynamics of the Baltic Sea.Ambio 19, 126-133.

Documents

L. Phil Graham - DiVA portal8668/FULLTEXT01.pdf · atmosphere and the land surface, and the land surface and the sea can be studied in detail. Understanding and modeling the large-scale