Hydrologic issues

Walter Collischonn IPH/UFRGS – Porto Alegre, Brazil

First Meeting of the GEWEX/CLIVAR/VAMOSSteering Group

18-19 September 2006 Clube dos 500 – Guaratinguetá – SP – Brasil

Scientific questions• What role do soil processes play in the basin?• Do the large variations in the flooded area of the

Pantanal impact and are themselves influenced by the variations in the region’s climatology?

• Does water evaporated from the Pantanal wetland fall as precipitation elsewhere within the Plata Basin? If so, how much is recycled, and where does it fall?

• What developments and improvements in hydrological models are required to better represent the relationships among model parameters and changes in soil use?

Hydrology of LPB

NaturalStorage

ReservoirStorage

Hydrologic modelling

• Macro-scale hydrologic model• Several parts of the LPB already modelled• Attempts to evaluate land use change effects• Climatic variability effects• Streamflow forecasts based on Precipitation

Forecasts• Verification of satellite estimates of

precipitation (TRMM and Hidroestimador)• Use of reanalysis data

Soil

+

BlocksVegetation andland use

DEM

SimilarHydrologicbehaviour

Drainage network, sub-basins

Free regional/global availableinformation

• Soil data:– FAO (1974)– Soil and Terrain Digital Database for Latin America and the Caribbean

(SOTERLAC, 1993-1997) – RADAM Brasil project;

• Topography data:– Shuttle Radar Topographic Mission (SRTM) – Global 30 Arc-Second Elevation Dataset (GTOPO30);

• Land cover:– Global Land Cover Characteristics Data Base (USGS)– LANDSAT images from Earth Science Data Interface (ESDI) at the

Global Land Cover Facility; • Climate:

– NOAA– NCEP – METeorological Aerodrome Report– National Water Resources Agency of Brazil (ANA);

Grande river basin drainage network

22o

21o

20o

51o

52o

53o

54o

55o

56o

57o

58o

51o

52o

53o

54o

55o

56o

57o

58o

22o

21o

20o

Coarse-resolutiondrainage network

River networkavailable from ANA

Ended and ongoing applications of the hydrological model

• 1) Taquari-Antas• 2) Uruguay*• 3) Paraguay*• 4) São Francisco• 5) Corumbá*• 6) Paraná to Itaipu*• 7) Paranaíba*• 8) Tapajós• 9) Madeira

• * (inside LPB)

Parameters are calibrated but are related to the blocks, or patches, whichmeans they have the same value for the same land use + soil types.

Application with parameters calibrated on neighbour basin give good results.Example: Calculated versus observed hydrograph in river Uruguay with

model parameters calibrated for the Taquari Antas basin (the basins havesimilar geological characteristics and climate)

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

jan-87 fev-87 mar-87 abr-87 mai-87 jun-87 jul-87 ago-87 set-87 out-87 nov-87 dez-87

Vaz

ão (m

3/s)

calculadoobservado

Streamflow forecasting using mediumrange ETA/CPTEC forecasts

Comparison with forecasts currently in use

Previsão de vazões de curto prazo – ÁGUA VERMELHA

ÁGUA VERMELHA - Previsão de vazão com dados previsão ETA

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

7/10/1

998

14/10

/1998

21/10

/1998

28/10

/1998

4/11/1

998

11/11

/1998

18/11

/1998

25/11

/1998

2/12/1

998

9/12/1

998

16/12

/1998

23/12

/1998

30/12

/1998

6/1/19

9913

/1/19

9920

/1/19

9927

/1/19

993/2

/1999

10/2/

1999

data

vazã

o (m

3/s)

previsões

obs

PREVISÃO COM CHUVA ETA

Mean error compared to stochastic model that is in operational use today

first week second week

Stochasticmodel

Comparison for years 2001 and 2002(dryer than normal)

Long range forecasts

Forecasting errors

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

15/4/

1998

15/5/

1998

15/6/

1998

15/7/

1998

15/8/

1998

15/9/

1998

15/10

/1998

15/11

/1998

15/12

/1998

15/1/

1999

15/2/

1999

15/3/

1999

15/4/

1999

15/5/

1999

15/6/

1999

15/7/

1999

15/8/

1999

15/9/

1999

15/10

/1999

15/11

/1999

vazã

o (m

3/s)

Previsão Global (máx e mín)observadoPrevisão Global média

Previsão de vazão de longo prazo – ÁGUA VERMELHA

- rodada do modelo global de jun/98, com previsão de precipitação de set/98 a fev/99 (até seis meses de antecedência).

Previsão de vazões em Água Vermelha – COM correção do Global

Use of sattellite estimates of precipitation

Challenges and opportunities in hydrologic model development

• Parameter values for hydrological models can be hardly known a priori, as shown one more time recently by Lohmann et al. (2004).

• The usual approach in the case of distributed models of large basins is to relate parameter values to readily obtainable information, such as land use and vegetation classes, or soil types.

• However, unlike regional atmospheric models, hydrological models are seldom used without parameter calibration. Calibration of such models can use a criterion of fit in the form of a multi-objective function, minimized by manual iterations or using automatic optimization methods based on evolution algorithms (Sorooshian et al., 1993; Collischonn and Tucci 2003; Gupta et al., 1998; Boyle et al., 2000; Vrugt et al., 2003).

• Multi-objective methods will be even more important as data from flux measurement towers become available, leading to a problem of which data should be given more confidence: latent heat fluxes for different land use and vegetation, measured by flux towers at spatially restricted spots, or streamflow, as an integrator of the hydrological processes occurring over the whole basin.

SVATs in hydrology and meteorology

• One could say that hydrological models have good results when integrated over large areas for the wrong reasons, i.e. a sum of locally doubtful results is able to generate regionally reasonable results, while atmospheric models SVATs are a collection of good reasons that result in a rather poor performance when integrated over larger areas. Therefore the coupling of hydrologic and atmospheric modeling needs a move to more similarity in both kinds of models.

• Development of coupled atmospheric-hydrologic models could be accomplished in well identifiable steps. The first step could be to implement the SVATs used in atmospheric models in the structure of an hydrological model and force it with observed rainfall and surface temperature, incoming short wave radiation, wind speed, humidity and pressure data. Calculated runoff could then be compared to observed streamflow at gauging stations, in order to analyze SVAT resultsintegrated over large areas. The second step would be the identification of shortcomings of the SVAT performance, and identify the reasons for poor performance. At the same time evaporation calculated by distributed hydrological models well fitted to reproduce only streamflow should be compared to latent heat fluxes at the locations where measurement towers were installed. So the fourth step would be analyze these results and identify shortcomings of the hydrological models SVATs. A fifth step would be to propose common SVATs that could be used both in hydrological and atmospheric models, maybe with a flexible, user settable, complexity, and taking into account the spatial variability that cannot be entirely represented by the models spatial resolution (subgrid variability). After this steps the new common hydrologic-atmospheric SVATs could be tested as the land module of atmospheric models.

Variability in hydrologic time series

River Paraguay level at Ladario (Pantanal)Return to “normal” range? Unpublished work by Allasia