Norman L. Miller, Raj Singh, Charles Brush, Jay Famiglietti, Hans-Peter Plag

1. Groundwater Monitoring and Management for Sustainability:California Pilot Test and Transfer to the Nile Basin Norman L. Miller, Raj Singh, Charles Brush, Jay Famiglietti, Hans-Peter Plag University of California, Berkeley & Berkeley National Laboratory Modeling Support Branch, California Department of Water Resources Hydrologic Modeling Center, University of California, Irvine Nevada Geodetic Laboratory, University of Nevada, Reno IGCP 565 4rd Annual Meeting University of Witwatersrand Johannesburg, South Africa 22 November 2011

2. Intergovernmental Panel for Climate Change Special Report on Emissions Scenarios SRES IPCC 2001

4. SENSITIVITY OF SNOWFED HYDROCLIMATE TO A +3ºC WARMING … Rain? or Snow? • What fraction of each year’s precipitation historically fell on days with average temperatures just below freezing? Less vulnerable More vulnerable +3 Computed from UW’s VIC model daily INPUTS Courtesy Mike Dettinger.

5. Groundwater recharge—precipitation/elevation relationship Images from: http://education.usgs.gov/california

6. Statistically Downscaled Temperature and Precipitation

7. Analysis of the Hydrologic Response • Miller et al. 2003 • National Weather Service – River Forecast System Sacramento Soil Moisture Accounting Model (Burnash 1973) • Anderson Snow Model for computing snow accumulation and ablation (Anderson 1973) Sacramento-Delta, 1242m, 1181km2 Feather - Oroville, 1563m, 9989km2 NF American - NF Dam, 1402m, 950km2 Merced - Pohono Br, 2490m, 891km2 Kings - Pine Flat, 2274m, 4292km2

8. Diminishing Sierra Snowpack% Remaining, Relative to 1961-1990 Miller et al. 2003

9. Drought Experiments Approach: Recreate drought scenarios considering historic data Managed Surface Water Drought Scenarios 10 year spin-up; Duration: 10, 20, 30, 60 year managed droughts Intensity: Dry, Very Dry, Critical defined as 30, 50, 70 % effective reduction 30 year rebound period All simulations used fixed 1973-2003 precipitation, urban demands, cropping etc.

10. Analysis of Snowpack Reduction Impacts on California Groundwater Water Infrastructure Using DWR C2VSIM Model C2VSIM - California Central Valley Simulation Model Domain: ~ 20,000 square miles

11. Finite Element Grid 3 layers 1393 nodes 1392 elements Surface Water System 75 river reaches 2 lakes 97 surface water diversion points 6 bypasses Land Use Process 21 subregions 4 Land Use Types Agriculture Urban Native Riparian Simulation periods 10/1921-9/2003 (<8 min) 10/1972-9/2003 (<4 min) Framework

13. Hydraulic Conductivity

14. C2VSIM Performance – HeadsR305 – Initial Calibration

15. C2VSIM Performance - Flows

16. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation BASELINE • Baseline - no surface water reduction • Drought - 30 - 70 percent surface water reduction • All simulations used fixed 1973-2003 precipitation, urban demands, cropping etc. • Climate simulations using the IPCC SRES output indicates California Snowpack will be reduced by 60-90% by 2100. • Simulating drought scenarios acts as an analogue to climate warming and provides us with a means to analyze impacts. Relative WT Change (Feet)

17. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 10 YEARS Relative WT Change (Feet) DRY 30 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

18. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 20 YEARS Relative WT Change (Feet) DRY 30 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

19. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 30 YEARS Relative WT Change (Feet) DRY 30 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

20. Case II: Initial Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 60 YEARS Relative WT Change (Feet) DRY 30 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

21. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 10 YEARS Relative WT Change (Feet) VERY DRY 50 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

22. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 20 YEARS Relative WT Change (Feet) VERY DRY 50 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

23. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 30 YEARS Relative WT Change (Feet) VERY DRY 50 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

24. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 60 YEARS Relative WT Change (Feet) VERY DRY 50 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

25. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 10 YEARS Relative WT Change (Feet) CRITICAL 70 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

26. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 20 YEARS Relative WT Change (Feet) CRITICAL 70 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

27. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 30 YEARS Relative WT Change (Feet) CRITICAL 70 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

28. Central Valley Water Table ‘Relative’ Response Joint LBNL-CDWR Drought Simulation 60 YEARS Relative WT Change (Feet) CRITICAL 70 PERCENT EFFECTIVE REDUCTION IN MANAGED SURFACE FLOW. Miller et al. 2006

29. C2VSIM Sub-Regions

33. Qualifiers C2VSIM and ALL water allocation models are only partially verified. Many empirical parameters are tuned. The groundwater processes lack sufficient physical descriptions. Groundwater total mass and variation is not known. Pumping is based on a limited available demand record. Demand is fixed and agriculture does not shift with change in supply.

36. Assimilation of “DWR Groundwater depth measurements” and “GRACE” total change in terrestrial water storage with CLM4. • 1. CLM4 is run over the test region using custom-made high resolution 1km dataset containing high resolution DEM and Soil texture data. • 2. Assimilate the well measurement data and GRACE at monthly time step with the CLM4 simulation over the test region. The assimilation takes into account change in water table depth at well sites and the total TWS change over the whole region. • 3. The assimilation krigs a new watertable depth at the various cells using the calculated CLM4 data and the observation data. The method uses the method of simple kriging and kriging with external drift.

37. Simulated 1-km Water Table Depth for California 1 km resolution run over California. Meters

38. SFREC test site

39. Test Region (SFREC) Spatial Resolution of GRACE data ~ (100 km X 100 km) temporal scale - Monthly Well measurement data ( point measurements) Temporal Scale used - Monthly

40. Monthly time step Assimilation flowchart, Repeated every month • Detrending of Groundwater Measurement Data by calculating Height of WT from MSL. • Calculation of Experimental Variogram with the observed data for the month. • Fit an theoretical Variogram to the experimental Variogram. • Get the Measured GRACE ΔTWS for the month over the test region. • Krig the water table difference across all the subgrid. • New water table depth across the subgrid cells used for further calculations.

41. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater.

42. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater. • Groundwater acts as a water resource insurance during droughts.

43. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater. • Groundwater acts as a water resource insurance during droughts. • Direct monitoring is very sparse.

44. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater. • Groundwater acts as a water resource insurance during droughts. • Direct monitoring is very sparse. • Indirect monitoring requires new techniques that allow for bridging spatial gaps.

45. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater. • Groundwater acts as a water resource insurance during droughts. • Direct monitoring is very sparse. • Indirect monitoring requires new techniques that allow for bridging spatial gaps. • GRACE, GPS, and well data assimilations into dynamic surface-groundwater models are needed.

46. Conclusions • Groundwater is 0.8 percent of total water, but it is 2.8 percent of total freshwater. • Groundwater acts as a water resource insurance during droughts. • Direct monitoring is very sparse. • Indirect monitoring requires new techniques that allow for bridging spatial gaps. • GRACE, GPS, and well data assimilations into dynamic surface-groundwater models are needed. • Hindcast validation is required for advancing high-resolution groundwater monitoring.

47. THANK YOU !