Impacts of uncertainty in efficient management of California s water resources

Transcription

1 Impacts of uncertainty in efficient management of California s water resources Newsha Ajami, George Hornberger, David Sunding Berkeley Water Center and David Purkey, David Yates Stockholm Environment Institute-US Center California Water and Environmental Modeling Forum, Asilomar CA February th,2007

2 Competing Demands in California Domestic Power & Industry Wildlife Agriculture Recreation Navigation

3 CAL2030 project: Overview In the near future California will be dealing with Population growth California Neighboring states (Arizona-Nevada) Growing Water Demand Water quality regulations and environmental requirements Shrinking Water Resources Climate Change Shrinking snow pack Time shift in water supply California s water crisis This calls for an integrated, efficient and sustainable plan to management of water resources which is one of focuses of the Berkeley Water Center

4 CAL2030 project: system integration This integrated system will include an open and flexible platform that connects all California s water systems. Why an open platform? More efficient resource management by allowing decision makers to test different possible coordinated operation scenarios. Improved system reliability in term of water supply by considering: uncertainty in future hydrologic events (e.g. impacts of climate change). uncertainty within the management system (e.g. modeling processes). disaster management issues such as exploring alternative water supply sources in case of a catastrophe.

5 CAL2030 project: characteristics of the platform A few major characteristics of such a platform are that it: contains multiple modeling components (e.g. surface water component, ground water component) that can be attached and detached from the main platform without causing any operational challenges for the user is flexible enough to accept different kind of inputs as source of water supply (including historical hydrologic records or direct output of any hydrological model). can be operated on various time scales (e.g. daily, monthly, and seasonally). This way the platform can be operated for both water supply and flood control purposes.

6 CAL2030 project: pilot project Conduct a pilot project to focus on some of the raised issues in a smaller scale by: Identifying and assessing end-to-end uncertainty in a water resources management system with multiple users such as Sacramento River Basin. Evaluating the reliability of water supplies by analyzing the estimated uncertainty. Evaluating if more accurate estimation of the uncertainty leads to sustainable management of our limited water resources in the state of California.

7 Impacts of reliable forecast More reliable forecast can: Change our perspective of expected future extreme events such as Floods and Droughts. Enables the decision makers to better handle risk in decision making Therefore mitigate some of social, economical and environmental impacts. Also manage our limited water resources more efficiently under new adaptation plans.

8 Methodology Study area: Upstream of Shasta reservoir in Sacramento basin (including all the catchments that contribute to the inflows to Shasta). Monthly precipitation and temperature data from Single aggregated demand which represents the water demand south of Shasta. Two simple hydrologic models to generate streamflow ensembles. Hydrologic MODel (HYMOD,5 par) Simple Water Balance model (SWB, 5 par) Using the Water Evaluation and Planning (WEAP) model for resource management and planning (developed by the Stockholm Environment Institute).

9 Methodology First step (accounting for hydrological uncertainty) Second step (uncertainty propagation) Hydrologic inputs (e.g. precipitation ensembles) Hydrologic model 1 Hydrologic model 2 Streamflow Ensembles WEAP Future Water supply uncertainty Sustainable planning Input Uncertainty Model Uncertainty Increase in the Water demand

10 Why uncertainty is important? Hoshin Gupta, 2004

11 Accounting for hydrological uncertainty Integrated Bayesian Uncertainty Estimator (IBUNE). Framework that accounts for uncertainty in input forcings, model parameters and model structure. Parameters HYM HYM Output Parameters SWB SWB Output Optimization + MCMC + Model combination Here we use it just to account for model parameter and model structural uncertainty. Model Combination Output time Final ensembles of streamflow

12 Streamflow (CMS) Performance of IBUNE s probabilistic and Deterministic simulations Streamflow Ensemble Model Predictions Uncertainty associated with parameters and model structure IBUNE predictive mean Observed Months Streamflow (CMS) Months

13 IBUNE versus individual models Probability Density Estimates MRMS distributions HYM SWB IBUNE Probability Density Estimates MABS distributions HYM SWB IBUNE

14 IBUNE versus individual models Frequency of observation in 95% forecast ranges IBUNE Percentage HYM SWB Model

15 Propagation of hydrological uncertainty Propagate the estimated uncertainty through a water management and planning model such as WEAP. Looking at the inflow to Shasta Dam. Single aggregated demand which represents the water demand south of Shasta. Evaluating the reliability of water supplies by analyzing the estimated uncertainty. Two scenarios: Scenario 1: current supplies, current demand Scenario 2: current supply, future demand (considering population growth for year 2030)

16 WEAP model Sacramento River Shasta Res. Pit R. Pit R. Oroville Res. Feather R. $ ( Shasta Res. Big Giant Demand Integrated decision support system model that balances water supplies and multiple water demands Folsom Res. Climate driven hydrology

17 Reservoir Storage Volume-current supply, current demand 6 x 109 Reservoir Storage Volume (current supply, current Demand) Reservoir Storage Volume (Cubic Meter) Uncertainty bounds WEAP Observed IBUNE RMS-WEAP=6.1938e+008 RMS-IBUNE=2.8445e Months

18 Percent of annual unmet demand-current supply, current demand 80 Percent of Annual Unmet Demand RMS-IBUNE=13.96 RMS-WEAP=25.21 Percent of annual unmet deman Observed WEAP LB of uncertainty UB of unertainty IBUNE predictive mean Years

19 Reservoir Storage Volume-current supply, 2030 demand Reservoir Storage Volume (Cubic Meter) 6 x Reservoir Storage Volume (current supply, 2030 demand) Uncertainty bounds WEAP Observed IBUNE Months

20 Percent of annual unmet demand-current supply,2030 demand Percent of Annual Unmet Demand 80 Percent of annual unmet demand Observed WEAP LB of uncertainty UB of unertainty IBUNE predictive mean RMS-IBUNE=15.11 RMS-WEAP= Years

21 Preliminary conclusion IBUNE is an innovative steps toward higher accuracy and more reliable hydrological forecasts including floods and water supply. Accounting for model structural uncertainty (considering multiple models) considerably improves prediction of water management variables. More accurate prediction of these variables can lead to more efficient operation of reservoir consequently more efficient management of water resources. More precise assessment of uncertainty in the future events can help more reliable decision making and also provide an opportunity to consider alternative operational coordination strategy.

22 Thanks Questions? Man is a complex being; he makes the deserts bloom and lakes die. (Gil Stern ) e.mail: newshaajami@berkeley.edu

23 Future steps. Assessing the uncertainty in input forcing. Retrospective Climate change We are still looking for better ways of analyzing and interpreting the impacts of hydrological uncertainty on future water supply. Also we would like to answer the following questions in our pilot project: Will this accuracy lead to sustainable management of our limited water resources in the state of California? What kind of alternative operational coordination strategy, considering the assessed uncertainty within the system, will decrease the risks in supply reliability and facilitates more efficient management of water resources? What are the socio-economical impacts of more accurate quantification of uncertainty in the hydrologic component of the system?

24 Normalized width of 95% uncertainty bound from streamflow to Reservoir Normalized width of 95% uncertainty bounds Percent of time exceeded Reservoir storage capacity Streamflow

25 0.6 Reservoir storage capasity 0.5 Streamflow Probability Distribution Normalized width

26 Normalized width of 95% uncertainty bounds Reservoir storage capacity-current demand Reservoir storage capacity-2030 demand Percent of time exceeded

27 A closer look: an excerpt of the hydrograph Ensemble Model Predictions Uncertainty associated with paramters and model structure IBUNE predictive mean Observed Streamflow (CMS) Oct/1982 Dec Feb Apr Jun Aug Oct/1983 Dec Feb Apr Jun Aug Oct/1984 Months