Usefulness of GCM data for predicting global hydrological

Usefulness of GCM data for predicting global hydrological changes Frederiek Sperna Weiland Rens van Beek Jaap Kwadijk Marc Bierkens 15 december 2009

Overview • Validating GCM produced climate datasets on their usability for hydrological studies • Modelling hydrological effects of climate change and distinguishing signal from noise • Validating bias-corrected GCM datasets on their usability for hydrological studies 15 december 2009

Hydrological impact studies GCM data hydrological model statistical / dynamical downscaling 15 december 2009 modelled discharges hydrological model Biascorrection hydrological model

Background - GCM General Circulation Model (GCM) Global Climate Model: • Energy balance • Resolution: 1. 875 – 3. 75 9 - 26 layers • Forcings: - Greenhouse gas - Aerosols • No predictions on day to day base Wikipedia, 2009 15 december 2009

What has been said about GCMs…. • GCM data can show large deviations from reality, especially for precipitation (Covey, 2003) • Differences between GCM results are large and can be larger than differences between emission scenarios (Arnell, 2003) • The model mean might show the best results (Murphy, 2004; Covey, 2003) 15 december 2009

Datasets - Climate model data • Intergovernmental Panel for Climate Change (IPCC): http: //www. ipcc-data. org/ Provides data on a monthly timestep • PCMDI data portal: Program for Climate Model Diagnosis and Intercomparison https: //esg. llnl. gov: 8443/index. jsp Provides data on a daily timestep 15 december 2009

Datasets - Multiple AOGCM’s Model Institute Country Acronym BCM 2. 0 Bjerknes Centre for Climate Research Norway BCCR CGCM 3. 1 Canadian Centre for Climate modelling and Analysis Canada CCCMA CGCM 2. 3. 2 Meteorological Research Institute Japan CGCM CSIRO-Mk 3. 0 Commonwealth Scientific and Industrial Research Organisation Australia CSIRO ECHAM 5 Max Planck Institute Germany ECHAM ECHO-G Freie Universität Berlin ECHO GFDLCM 2. 0 Geophysical Fluid Dynamics Centre USA GFDL GISS ER Goddard institute for Space Studies USA GISS IPSL CM 4 Institute Pierre Simon Laplace France IPSL MIROC 3. 2 Center of Climate System Research Japan MIROC NCAR PCMI National Center for Atmospheric Research USA NCAR HADGEM 1 Met Office’s Hadley Centre for Climate Prediction UK HADGEM 15 december 2009

Parameters - Precipitation - Temperature Calculation of potential reference evapotranspiration Penman-Monteith: - Incominging and outgoing shortwave radiation - Incoming and outgoing longwave radiation - Airpressure - Windspeed - Temperature and minimum temperature Calculation of potential reference evapotranspiration Blaney-Criddle: - Temperature 15 december 2009

Reference dataset - CRU / ERA 40 CRU: • Climate Reasearch Unit, University of East-Anglia • Timeseries with monthly values • 1901 -1995 ERA 40: • ECMWF • Daily values • 1957 – 2002 Validation period: 1961 - 1990 - Downscaling CRU data to daily values based on ERA 40 - Projection on 0. 5 degrees model grid 15 december 2009

Discharge data GRDC - Global Runoff Data Centre: - Monthly discharges for 19 large rivers 15 december 2009

PCR-GLOBWB (Beek, 2007) • Global distributed hydrological model • Daily time-step • 0. 5 degrees resolution (360*720) • Sub-grid cell parameterisation • Contains three soil layers, lakes, rivers, snow, vegetation • Solves water balance per cell • Direction of surface runoff calculated with drainage direction map • River discharge calculated with routing scheme based on kinematic wave • Natural water availability – little antropoghenic influences included 15 december 2009

FEWS • 12 x GCM input • CRU/ERA FEWS-World: • Spatial/temporal interpolation • Unit conversion • Calculation of evaporation • PCRGLOB-WB model run 15 december 2009 • 13 x calculated: - Channel flow - Soil moisture - Snow cover - Actual evaporation

FEWS-World system 15 december 2009

First step: Validate models • PCR-GLOBWB is run for period 1961 -1990 with: - data from all individual GCMs - reference meteo dataset (CRU/ERA-40) • 30 -year average statistics are derived for the GCM runs and reference run and observations (GRDC) • GCM statistics are compared with CRU/ERA-40 and observations 15 december 2009

Hydrological regime - Brahmaputra 15 december 2009

Hydrological regime - Mac. Kenzie 15 december 2009

Hydrological regime - Rhine 15 december 2009

GCM discharge compared with CRU Relative 30 year mean discharge = (QGCM – QCRU) / QCRU 15 december 2009

Top 5 per catchment - mean discharge 15 december 2009

Modelling hydrological effects of climate change and distinguishing signal from noise 15 december 2009

selected IPCC scenarios 20 CM 3: • Control experiment (IPCC, 2007) A 1 B: • Rapid economic growth with a peak in global population in mid 21 st century followed by a population decline • Fast introduction of efficient technologies • Decrease of social and regional differences A 2: • Heterogeneous world with fragmented technological developments and large regional differences • Continuous increase of CO 2 emission Relative negative scenarios 2000 -2006: observed emissions larger than estimated (Global Carbon Project, 2008) 15 december 2009

Modeling change Relative change for ensemble of 12 GCMs: Mean discharge control experiment, period 1971 -1990 Mean discharges future experiments A 1 B and A 2, period 2081 -2100 15 december 2009

Global changes and model consistency A 2 A 1 B Nr. of models significant and consistent change 15 december 2009

Changes in river regimes 15 december 2009

Continental change • Freshwater discharge increases for all continents • Freshwater inflow to oceans only decreases for Mediteranean see • Large uncertainty amongst models 15 december 2009

Conclusions • GCM derived discharges show large deviations from observations and each other • Multi-model ensembles provide a ‘relative good mean’ and give uncertainty information • By quantifying significance and consistency of change, regions and catchments with high potential of hydrological change can be detected 15 december 2009