APPLICATIONS OF DOWNSCALING: HYDROLOGY AND WATER RESOURCES EXAMPLES Dennis P. Lettenmaier Department of Civil and Environmental Engineering For presentation at Workshop on Regional Climate Research NCAR April 4, 2001
OUTLINE Hydrologic models macroscale (VIC) and explicit (DHSVM) Example 1: Seasonal ensemble forecasting Example 2: Flood forecasting Example 3: Climate change assessment Final comments where are the weak links?
Variable Infiltration Capacity (VIC) Macroscale Hydrology Model Key features: Parameterization of runoff via saturation excess and base flow Explicit vegetation, including subgrid variability in vegetation classes Energy balance snow model, elevation bands for increased vertical resolution Streamflow routing Subgrid parameterizations for spatial precipitation variability, frozen soils, lakes and wetlands (new!) Full surface energy balance Spatial resolution typically 1/8-2 degrees lat-long
Distributed Hydrology-Soil- Vegetation Model (DHSVM) Key features: Explicit representation of runoff production via saturation excess mechanism, and base flow Explicit vegetation Energy balance snow model Streamflow routing to any identified location on stream channels Optional representation of roads and road drainage networks High spatial resolution (typically 30-150 m)
DHSVM Snow Accumulation and Melt Model
Land surface characterization required by DHSVM Terrain - 150 m. aggregated from 10 m. resolution DEM Land Cover - 19 classes aggregated from over 200 GAP classes Soils - 3 layers aggregated from 13 layers (31 different classes); variable soil depth from 1-3 meters Stream Network - based on 0.25 km 2 source area
Example 1: Ensemble Seasonal streamflow forecasting
Overview of Modeling Linkages Global Climate Models downscaling temp precip wind Hydrology Models streamflow Water Resources Models Regional Climate Models water demand
SST Forecast Coupled Global CM Embedded Regional CM Water Management Decisions Macro-Scale Hydrologic Model Water Management Model
MM5 Meso-Scale Climate Model ColSim Reservoir Model VIC Hydrology Model Precipitation, Temperature, Wind, Humidity Evapotransporation (canopy and soil layers) Snow Accumulation and Melt Vegetation and Soil Characteristics Runoff Routing Model Base Flow Inter-layer Infiltration Surface Infiltration
Constructing the Driving Data for the Perpetual January Streamflow Forecasting Experiments Jan MM5 MM5 Jan MM5 Jan 24 MM5 MM5 Jan MM5 Jan MM5 Jan Forecasts MM5 Jan MM5 Jan MM5 Jan MM5 Jan MM5 Jan MM5 Jan MM5 Jan MM5 Jan Jan... Oct Nov Dec Feb Mar Apr May Jun Jul... Repeating 1960 Observed Temp and Precip Time Series with January s Replaced by MM5 Forecasts
Hydrologic Simulations Using RCM Normal January Ensembles Before and After Bias Correction B) Uncorrected RCM Simulated Ensemble Hydrograph for The Dalles RCM Normal Met Data for January 1960 Water Year Met Data for Other Months 600000 500000 Flow (cfs) A) 600000 500000 400000 300000 200000 100000 0 Retrospective Ensemble Hydrograph for The Dalles Observed Normal Met Data for January 1960 Water Year Met Data for Other Months oct nov dec jan feb mar apr may jun jul aug sep Month Flow (cfs) 400000 300000 200000 100000 C) 0 600000 500000 oct nov dec jan feb mar apr may jun jul aug sep Month Bias Corrected Ensemble Hydrograph for The Dalles RCM Simulated Normal Met Data for January 1960 Water Year Met Data for Other Months Flow (cfs) 400000 300000 200000 100000 0 oct nov dec jan feb mar apr may jun jul aug sep Month
ENSO Neutral Simulated Ensemble Hydrograph for The Dalles Observed Normal Met Data for January 1960 Water Year Met Data for Other Months ENSO Warm Simulated Ensemble Hydrograph for The Dalles Observed El Nino Met Data for January 1960 Water Year Met Data for Other Months 600000 600000 500000 500000 Flow (cfs) 400000 300000 200000 Flow (cfs) 400000 300000 200000 100000 100000 0 oct nov dec jan feb mar apr may jun jul aug sep 0 oct nov dec jan feb mar apr may jun jul aug sep Month Month Simulated Ensemble Hydrograph for The Dalles Normal SST Meso-Scale Met Simulations for January 1960 Water Year Met Data for Other Months Simulated Ensemble Hydrograph for The Dalles Warm SST Meso-Scale Met Simulations for January 1960 Water Year Met Data for Other Months 600000 600000 500000 500000 Flow (cfs) 400000 300000 200000 Flow (cfs) 400000 300000 200000 100000 100000 0 oct nov dec jan feb mar apr may jun jul aug sep Month 0 oct nov dec jan feb mar apr may jun jul aug sep Month
Experimental Hydrologic Forecasting for East Coast US River Basins during Summer 2000 Andrew Wood Ed Maurer Dennis Lettenmaier University of Washington Arun Kumar NCEP/EMC/CMB
Model forecasting domain
General Method Climate model forecast outputs 2.81 degree resolution monthly total P, avg T hydrologic model inputs 1/8 degree resolution daily P, Tmin, Tmax streamflow, hydrologic variables Use 3 step approach: a) statistical bias correction b) downscaling c) hydrologic simulation of ensembles
Coupling GSM to VIC Step 1: Statistical Bias Correction at the GSM scale (~2.8 deg), use a quantile to quantile mapping from GSM climatology to observed historical climatology, for precip & temperature separately, e.g., http://maximus.ce.washington.edu/~aww/east_fcast/ds_bias_meth.htm
GSM ensemble forecasts from NCEP/EMC rolling climatology based on 1979-1999 historical SSTs roughly 2.8 degree resolution ensembles available around 10 th of month, extend 6 months beginning in following month each month: new set of 210 climatology ensemble members for monthly total precip & average temperature derived from 10 initial condition perturbations for each year of the 21 year climatology period 20 forecast ensemble members
Kanawha River Basin Simulations Using Observed Data and Bias Corrected GCM Output Mean Observations Bias-Corrected GCM Variance
Soil Moisture example: April `00 forecast for summer forecast ensemble median shown as percentile of climatology ensemble
Streamflow example: April `00 forecast for summer
Example 2: Flood forecasting
2000/2001 Real-time Streamflow Forecast System 26 basins 48,896 km 2 2,173,155 pixels @ 150 m resolution
Calibration-Validation with all available meteorological observations (50 sites) Calibration (Snohomish River) From 1987-1991 (USGS gauges at Gold Bar and Carnation only ) Validation 1991-1996
DHSVM Calibration (Snoqualmie at Carnation) Flood of record Principal calibration locations were the Skykomish at Gold Bar and the Snoqualmie at Carnation
DHSVM Parameter Transferability Application of Snohomish River parameters to Bull Run Reservoir (Portland, OR Water Supply) (124 km 2 )
Sauk Accurate precipitation forecasts are key!!! Snoqualmie Observed MM5-DHSVM NWRFC
Deschutes Nisqually MM5-DHSVM Observed NWRFC
Example 3: Climate change assessment
April 1 Average Snow Cover Extent Hadley Centre Model Base Case ~2025 ~2095
April 1 Average Snow Cover Extent MPI Model Base Case ~2025 ~2045
600000 Average Virgin Streamflow at The Dalles Hadley Centre (CO2 plus aerosols) 1/8 Degree Model Streamflow (cfs) 500000 400000 300000 200000 100000 Base Decade 2 Decade 9 0 10 11 12 1 2 3 4 5 6 7 8 9 Calendar Month
Average Virgin Streamflow at The Dalles MPI (CO2 plus aerosols) 1/8 Degree Model 600000 Streamflow (cfs) 500000 400000 300000 200000 100000 Base MPI Dec2 MPI Dec4 0 10 11 12 1 2 3 4 5 6 7 8 9 Calendar Month
Final comments where are the weak links? Downscaling problem isn t likely to go away, especially for local (explicit) model applications Bias (in the surface forcings) is the zero order problem in linking weather/climate and hydrology models For S/I forecasting, outstanding question is whether there is value added in downscaling of climate signal (e.g., via RCM) v direct downscaling (e.g., via PDF approach) to hydrology model For climate change applications, downscaling remains problematic esp. failure to preserve properties of large scale forcings