Managing Climate Risks on the Colorado River Balaji Rajagopalan Department of Civil, Environmental and Architectural Engineering & CIRES University of Colorado, Boulder, CO Symposium on Climate Risk Management Oct 10-12, 2011 Guayaquil, Ecuador
Acknowledgments Kevin Werner & Andy Wood CBRFC, NOAA/NWS, Salt Lake City, UT Cameron Bracken, University of Colorado, Boulder, CO Kenneth Nowak, University of Colorado, Boulder, CO & USBR Katrina Grantz USBR, Salt Lake City, UT Seasonal to Year-Two Colorado River Streamflow Prediction Workshop, Mar 21-22, 2011, Salt Lake City, UT http://www.cbrfc.noaa.gov/present/2011/s1y2/agenda.htm
Colorado River Basin Overview 7 States, 2 Nations Upper Basin: CO, UT, WY, NM Lower Basin: AZ, CA, NV Fastest Growing Part of the U.S. Over 1,450 miles in length Basin makes up about 8% of total U.S. lands Highly variable Natural Flow which averages 15 MAF 60 MAF of total storage 4x Annual Flow 50 MAF in Powell + Mead Irrigates 3.5 million acres Serves 30 million people Very Complicated Legal Environment Law of the River Denver, Albuquerque, Phoenix, Tucson, Las Vegas, Los Angeles, San Diego all use CRB water DOI Reclamation Operates 1 acre-foot = 325,000 gals, 1 maf = 325 * 10 9 gals Mead/Powell 1 maf = 1.23 km 3 = 1.23*10 9 m 3
Below normal flows into Lake Powell 2000-2004 62%, 59%, 25%, 51%, 51%, respectively 2002 at 25% lowest inflow recorded since completion of Glen Canyon Dam Woodhouse et al., WRR, 2007 Motivation Recent conditions in the Paleo Context Some relief in 2005 105% of normal inflows Not in 2006! 73% of normal inflows 2007 at 68% of Normal inflows 2008 at 111% of Normal inflows 2009 at 88% and 2010 at 72.5% 5 year running average Decadal Variability!
Needs Given the stress on the system (socio-economic & climate via droughts) skillful long-lead streamflow forecasts on the Colorado River Basin are crucial for efficient system management. Forecasts of spring (Apr-Jul) flow, the key inflow period are desired at lead times starting previous November at each month. Forecasts are required at a number of decision point locations on the River Basin. Need to Manage Climate Risk at short (seasonal to interannual) and long (multidecadal) time scales
Water Supply Forecast Overview (River Forecast Centers) (Water and Climate Center) SWS (Statistical Prediction) ESP (Hydrologic Model Prediction) VIPER (Statistical Prediction) Forecast Coordination Official Coordinate d Forecast Decisions Water Managers and Users Other Inputs....
Water Supply Forecast Methods Statistical Forecasting Statistical Regression Equations Primary NOAA/RFC forecast method from 1940 s to mid 1990 s. Primary NRCS/NWCC forecast method Historical Relationships between flow, snow, & precipitation (1971-2000+) Tied to a fixed runoff period (inflexible) Ensemble Simulation Model Forecasting A component of a continuous conceptual model (NWSRFS) Continuous real time inputs (temperature, precipitation, forecasts) Accounts for soil moisture states (SAC-SMA) - drives runoff efficiency Builds and melts snowpack (Snow-17) output feeds SAC-SMA Flexible run date, forecast period, forecast parameters. Evolving toward ESP as primary forecast tool at NOAA/RFCs
13 River Forecast Centers Established in the 1940s around major river basins for water supply forecasting Three primary missions: 1. Seasonal Water supply forecasts for water management 2. Daily forecasts for flood, recreation, water management 3. Flash flood warning support 4. Each RFC develops hydrologic models relevant for their basin/region River Forecast Centers They integrate short term and seasonal Forecasts from NWS/NOAA Colorado Basin River Forecasting Center Example www.cbrfc.noaa.gov
Weather and Climate Forecasts RFC forecast system incorporates both weather and climate forecasts: Weather forecasts integrated into daily operations with forecaster control over point and basin average values Water supply forecasts typically only use QPF during late season or in lower basin When QPF is used, it is used in a deterministic manner Climate forecasts integrated into seasonal water supply forecasts through probability shifts of forcing ensemble Climate forecasts are typically only considered in lower basin and only in ENSO years
General RFC Model Hydrologic Model Analysis Weather and Climate Forecasts Weather and Climate Forecasts Forecast precip / temp hydrologic expertise & judgment Analysis & Quality Control Observed Data model guidance River Forecast Outputs System Graphics parameters Calibration River Forecasts
RFC Models RFC forecast uses a snow model and a rainfall-runoff model: SNOW-17: Temperature index model for simulating snowpack accumulation and melt Snow Model: SNOW-17 Temperature Index Snow model Sacramento Soil Moisture Accounting Model: Conceptual hydrologic model used to generate runoff
Key Steps in Model Building River basin is divided into homogeneous sub-basins based on topographical attributes (using GIS, Remote-sensing data etc.) For each sub-basin historic 6-hour weather (Precip., Tmax, Tmin) are developed for three regions higher, middle and lower elevations. All historic data is used in this process. Natural / unregulated streamflow data is also compiled The Hydrologic model is calibrated These are constantly updated
Key Steps in Model Building Process to assign parameter values to the runoff and snow modules within the model. Unique set for each basin (and sub-basin) Quality of calibration can vary greatly from basin to basin depending on data availability, period or record, quality of data, hydrology of the basin, etc. Archiving and Maintaining Data *Very Important*
Ensemble Streamflow Prediction
CBRFC currently provides raw ensemble time series forecasts to several user groups: Denver Water Utility Pacificorps (Bear River, UT) USBR (Gunnison, Utah, and MTOM) Forecasts updated daily in winter/spring Available via CBRFC webpage ESP applications http://www.cbrfc.noaa.gov/ Hydrodata 15
Web Reference: www.cbrfc.noaa.gov/gmap/gmapm.php?wcon=checked
Dissemination / Verification The forecasts are disseminated online Ensembles provided as raw data http://www.cbrfc.noaa.gov / Hydrodata Very good visualization tools showing a suite of forecast and observed variables. http://www.cbrfc.noaa.gov / (links on the left column) Forecast Verification http://www.cbrfc.noaa. gov/ (Verification)
CBRFC Products RFCs (CBRFC in particular) are a source of wealth of information for hydrology and water supply. They provide current, forecast and historic information. Western US wide all aspects of water supply Water Supply Outlook Current Snow Conditions 0-14 day Outlook Reservoir Outlook NRCS - products
La Nina Winter 2011 Snowpack Maps / Data NRCS El Nina Winter 2010
Proposed Improvements Newer statistical models for multi-side ensemble streamflow forecasting are being tested and developed for integration with the CBRFC forecasting system (e.g., Bracken et al., 2011, Regonda et al., 2006) Using large-scale climate information with multi-model statistical ensemble forecasting techniques Current ESP limited (by the length of the historical data) in its ability to generate ensembles Stochastic weather generators can provide a rich variety of input daily weather ensembles Easy to condition them on probabilistic seasonal forecast (e.g., Apipattanavis, et al., 2007; Apipattanavis, 2008) Optimal combination of statistical and watershed model based forecasts
Bracken et al. 2010, WRR Using Climate Information for long lead Streamflow Forecast
Water Resources Management (Interannual time scale) Reservoir Operations 12 major reservoirs (9 Upper, 3 Lower Basin) Seasonal/Monthly forecasts are input to the systems model In RiverWare contains all the operating rules and constraints Obtain mid-term reservoir conditions (storage, elevation, release, hydropower, etc.) Close coordination between
Seasonal to Two Year Simulation and Forecasting Hidden Markov model for simulation and forecasting Seasonal Forecast Historical Data 1000 32 Mo. Traces HMM Forecast Midterm Operations Model Probabilistic Reservoir Outlook Pool Elevation [feet] 3540 3560 3580 3600 3620 3640 3660 Western Water Assessment 0 5 10 15 20 25 30 Month
Water Resources Data/Information USBR archives and disseminates historical streamflow and reservoir data; current status; operating plans; real time releases etc. The stakeholders and general public can get a very good idea of the state of the resources Current Reservoir Levels Current snowpack in the sub-basins Operating Plans, Historic Data Current Reservoir Status
Decadal Scale Stochastic Streamflow Simulation and Water Resources Management Analysis of natural and paleo flow variability Identify spatially persistent modes of flow variability Wavelet spectral methods and principal component analysis Identify potential climatic drivers of variability Spectral coherence with climate indices; correlation with climate variables Develop simulation methods that reflect variance analysis Capture historic/simulate future spectrum; condition on climate projection Colorado River Water Resources Management Analysis of CRSS model to identify system components sensitive to certain flow or variance regimes Develop adaptive management policies to demonstrate improved system outcomes associated with variability understanding Western Water Assessment
Wavelet Power Spectrum of Lees Ferry Flow Features of interest 1) decadal (active past 30 years) 2) Low frequency (more persistent) Raw WY flow and precipitation correlation = 0.77 Raw WY flow and WY temperature correlation = 0.32 UCRB PRISM Temperature Spectrum Western Water Assessment UCRB PRISM Precipitation Spectrum
Decadal Flow Projection Given Application persistent to 2 periods low frequency of paleo variability data (natural flow and paleo data) mode, what can previous conditions inform about future characteristics? Right, 1925 to 1974 20 year mean flow projections Below, 1634 to 1683 20 year mean flow projections? Western Water Assessment
Thoughts/Ideas for Transfer to Other Context RFC structure should be seriously considered for Hydrologic Forecast Effective interface with weather/climate forecasts Data collection / management Robust collection and archiving of meteorologic and, hydrologic data in particular, is critical for skillful forecasts Resources from Met services, and State agencies can be pooled Close coordination and collaboration with Water Management and state holders in the basin Effective outreach online dissemination, weekly/monthly outlook briefings, workshops Model Development/Improvement This is always an on-going process. The forecasting framework need to be constantly improved, calibrated and validated. Need to have a good capacity and human resources building strategy