Tim Barnett

February, 2002


Question:What is the impact of global warming on water resources inthe Western United States?


Click here to go to the first slide


Slide Number     Description

1 ACPI title page showing PIs.Note 3 DOE labs, 3 Universities, 1 DOD lab, USGS and NCAR participated in the project.

2 Project schematic.(Top)Observed ocean state of the mid-1990s is(Middle)used to initialize CGCM forced by BAU scenario.This is gives data(precip, snow,temperature, etc.) that is (Bottom)downscaled to 40km grid over the western US for subsequent impact analysis.Impacts:The lower right panels shows the large decrease in Sierra snow pack in 2050.The lower left panel shows the large reduction in annual mean snow pack in the Pacific Northwest.

3 The ocean initialization of the climate model is necessary since the oceans have already begun to warm as illustrated in this slide which shows the temperature change to a depth of 2000m in each of the major oceans relative to 1900. Taking this warming into account avoids the ‘cold start’ problem.

4 Strength of the global ocean warming signal from PCM(Parallel Climate Model) forced with observed/BAU anthropogenic forcing beginning in 1870 relative the signal predicted in 2100.Three different realizations are shown(horizontal bars) for decadal average signal strength.Blue/greens are from observed/estimated GHG emissions, while yellow/reds are from BAU scenarios.Observed signal strength from ocean data assimilation for6 years in the mid-1990s is shown with green circles.The model captured the observed warming quite well.

5 Does the PCM produce realistic simulation of present climate?If not, how could we believe its predictions of the future climate?Slide shows the observed and PCM simulated winter temperature and precipitation fields over the US, especially the western 1/3 which is the subject of this study.Conclusionde:The model, while not perfect, does a good job of reproducing the observations.Most CGCMs do not do as well.

6 The need for downscaling:The PCM grid (280 km) overlaid on the geographic drainage basins of interest..Note California is only two grid points wide and so does not resolve the Sierra.

7 The high resolution grid (40 km) of the downscaling routines now does crudely resolve major geographic features.This is important for simulating orographic precip and snow effects.

8 Results show downscaling adds a lot of realism to the PCM results.The PNNL-MM5 and RSM results for summer and winter precipitation are compared with the PCM field used to run the two downscaling models.

9 Verification of the two down scaling models (RSM and PNNL-MM5) against observed (PRISM) data.The models were initialized with reanalysis products from NCEP and/or ECMWF.The results show the two models capture well the observed monthly, annual and interannual variations in the precip and temperature fields over the main basins of interest.Note that the models do better in winter than in summer and that neither model is capable of simulating the SW monsoon.But since the main precip period important for this study occurs during the winter/spring these are not serious flaws for this study.

10 Areal comparison of PNNL-MM5 and RSM with observations of mean daily rainfall and the 95% extrema of the daily rainfall pdf for the winter and summer seasons.Again, both models do a good job of reproducing the observations.Reanalysis data was used to forced the regional models.

11 Estimating anthropogenic impacts on water resources.The flow diagram shows the downscaled ‘weather’ data forcing hydrological models which in turn give information on snowpack, run off and streamflow.

12 Example of the validity of one of the hydrological models (VIC).Shown is thetime history of observed inflow to the Shasta Reservoir (black line) along with the VIC produced values (red) when the VIC is forced by observed weather variables.The main message here is that if the downscaled anthropogenic-impacted data are correct, our estimates of river flow, etc. are apt to be pretty good.

13 RESULTS:The snow water equivalent(SWE) for the Sierra (top) compared with the BAU simulation 1,2,…,5 decades into the future.Note the large reduction in future snow pack and the large decade-to-decade variability predictied by the model.

14 RESULTS:The reduced snow pack is reflected in the reduced Spring time run off in the California drainage.Reductions of 30-50% are typical.

15 RESULTS:The calendar day of maximum runoff is shown.Note it retreats nearly one month in the annual cycle in the next 40-50 years. This is particularly true for rivers fed by snow melt from intermediate elevations such as the Carson River.Rivers fed by higher elevation snows (American River) show less impact.Retreats of a few days to a week have been already observed relative to 1970 (not shown) over most of western N. America.The point is that this retreat is already in progress.

16 RESULTS:The stream flow problems lead to biological problems.Due to increased warming, the salmon of the PNW will have to delay their run upstream until the waters cool sufficiently with the onset of Fall.But the freshet (maximum stream flow associated with snow melt) needed to flush the young smolt out of the rivers in the Spring of 2050 comes a month earlier than present.Essentially the reproductive cycle of these endangered fish is compressed (lower time line arrow).Can they adapt to this change within 50 years?If not, they will become extinct in the PNW.

17 RESULTS:The Sierra snow melt (backdrop) feeds streams (blue lines) which inturn flow into the Delta where they eventually encounter the salinewater pumped into the San Francisco Bayfrom the ocean. The inter-mix of the fresh and saline waters creates an unique and richbiological environment.The reduced summer flow will increase salinity in the entire California delta region(right) compared to what it is today (left).The increase is so large that it will invoke a biological disaster, plus effect trans-California water distributions.

18 RESULTS:$ impacts and management….the Sacramento System natural inflows now and over the next 50 years.Note the reduction in water volume.

19 RESULTS:If the above water reduction is met with forced 50% reduction in water demand in California’s Central Valley in what are dry years, then current agricultural demands would experience a 730 thousand acre feet. short fall or 23% less than is needed, in spite of the mandatory reduction.The associated current metropolitan and industrial (MI)demand will exceed what is available by 17%.Population growth in the central valley is projected at approximately 8 million in the next 20 years. The increased water demand of this growth and projected supply reduction are on a collision course.This is a bleak economic picture for California’s future.

20 RESULTS:Assuming the 50% enforced reduction in water usage will still lead to a reduction in winter-time hydro power production of order 20% (blue line).The monthly $ loss associated with the water reduction (bars) is of order 20 million dollars/month during the winter.

21 RESULTS:The summer-time wildfire activity in the Sierra will increase by approximately 50% by 2050.This is due to reduction in summer soil moisture associated with the earlier seasonal snow melt. Basically the fire season will start sooner and last longer.



The impacts of anthropogenic warming will have profound impacts on the water resources of the Western United States and beyond by the year 2050.Indeed, some of these predicted changes are already present in observations of stream flow and snow pack, especially in California.

The impacts will manifest themselves in a number of ways.The snow pack will melt earlier.As a result, maximum Spring-time streamflows will happen about one month earlier in the calendar year.This causes numerous water resource problems.It also impacts the life cycle of salmon in the PNW, as well as the entire ecosystem of the Sacramento Delta.Additional impacts will be found in reduced ability to generate hydro electric power in the winter and large increases in wildfire activity in the summer.

The above results are particularly disturbing since the computer model from which they originally spring is known to have a relatively low sensitivity to GHGs.This means the above statements should be considered as conservative; if anything, they are under estimates of what will happen.