Final Report: Vulnerability Assessment of San Joaquin Basin Water Supply, Ecological Resources, and Rural Economy Due to Climate Variability and Extreme Weather EventsEPA Grant Number: R827448
Title: Vulnerability Assessment of San Joaquin Basin Water Supply, Ecological Resources, and Rural Economy Due to Climate Variability and Extreme Weather Events
Investigators: Dracup, John A. , Grober, Leslie , Howitt, Richard , Brekke, L. D. , Bashford, K. E. , Hidalgo, H. G. , Miller, N. L. , Hatchett, Stephen P , Quinn, Nigel
Institution: University of California - Berkeley , California EPA Central Valley Regional Water Quality Control Board , University of California - Davis
EPA Project Officer: Packard, Benjamin H
Project Period: July 1, 1999 through June 30, 2002 (Extended to June 30, 2003)
Project Amount: $859,654
RFA: Integrated Assessment of the Consequences of Climate Change (1999) RFA Text | Recipients Lists
Research Category: Global Climate Change , Ecological Indicators/Assessment/Restoration , Water , Ecosystems , Climate Change
The objectives of this research project were to: (1) assess the vulnerability of water supply, water demand, water quality, ecosystem health, and socioeconomic welfare within the San Joaquin River (SJR) region as a function of climate variability and extreme weather events that might happen in association with climate change; and (2) deliver decision support system (DSS) information to project stakeholders that guides management strategies to mitigate the potential impacts of climate change.
To carry out these objectives, several assessments were conducted to study climate change impacts on the SJR region’s: (1) water supplies and reservoir operations; (2) agricultural production and agroeconomics; (3) water quality; (4) aquatic ecosystem and fisheries; and (5) socioeconomics. Environmental and economic models of these processes were selected from those used by federal and state water agencies, the University of California, and from private consultants.
Framing Assumptions: CO2, Global Climate, and Local Hydrologic Response
The impacts assessments of this project are framed by assumptions on the future climate in California. At the very least, those assumptions require the following elements: (1) a scenario for greenhouse gas development in the global atmosphere; (2) model selection(s) for simulating that global climate response to greenhouse gas changes; and (3) methods for relating global climate changes to local hydroclimate response.
The global greenhouse gas condition assumed for this assessment is a 1 percent per year increase in “effective CO2” starting from present levels. “Effective CO2” represents the collective effect of CO2 and other greenhouse gases (Intergovernmental Panel on Climate Change [IPCC], 2001). Numerous global climate models have been used to simulate this particular scenario (IPCC, 2001). Two models were selected for simulating global climate response under this CO2 increase scenario: HadCM2 of the U.K. Hadley Centre and the Parallel Climate Model (PCM) of the U.S. National Center for Atmospheric Research. Given the selections of the CO2 scenario and climate response models, specific model runs of HadCM2 and the PCM were chosen. Based on data availability, those model runs were selected to be “HadCM2 run 1” and “PCM run B06.06” (Miller, et al., 2003).
Statistical downscaling was performed to relate the simulated global climate responses to basin-scale climate responses over California at a 10 km spatial resolution. The statistical approach involves using historically derived regression equations based on the Parameter-elevation Regressions on Independent Slopes Model technique (Daly, et al., 1994) to relate large-scale atmospheric circulation patterns to basin-scale meteorology. These meteorological sensitivities then were used to simulate natural runoff sensitivity in five major tributaries to the joint Sacramento River and SJR Basin (see Figure 1): Sacramento (above Lake Shasta), Feather (above Lake Oroville), American (above Folsom Lake), Merced (above Lake McClure), and Kings (hydrologically connected to the SJR Basin during wet years).
Figure 1. Basins (Name, Elevation, Area) Chosen for Studying Natural Runoff Sensitivity in California’s Sacramento-SJR Region Because of Climate Change (Miller, et al., 2003). Runoff sensitivities in these basins served as framing input to subsequent assessments of Central Valley reservoir operations, San Joaquin region water allocations, etc.
The work involving climate projection and natural runoff modeling was completed through collaboration with Lawrence Berkeley National Laboratory and the National Weather Service River Forecast Center in Sacramento, CA. Natural runoff sensitivities are mean monthly runoff under climate change relative to the baseline present climate, defined to occur from 1963 to 1992. Natural runoff sensitivities were evaluated for three future climatological periods: 2010-2039, 2050-2079, and 2080-2099. Only the first two periods were included consistently in subsequent impacts assessments. Details on the development methods and interpretation of these sensitivities are summarized in Miller, et al. (2003). In summary, the HadCM2 projection is warm and wet over California relative to the PCM projection. Both projections grow warmer in the future, but the rate of warming is slower in the PCM projection. Most importantly, the HadCM2 projection involves an annual average increase in precipitation, whereas the PCM projection involves an annual average decrease in precipitation.
Reservoir Operations and Water Allocation . The agricultural economy of the SJR region depends on irrigation practices and surface water supplies (see Figure 2). The origin of these supplies varies for irrigators on the west or east side of the region. East-side irrigators are supplied by Sierra Nevada runoff on the east side of the SJR region via local water systems along each of the SJR’s main tributaries: the Stanislaus River, the Tuolumne River, and the Merced River. West-side irrigators receive supplies from the Sacramento River region via the federal Central Valley Project (CVP). The CVP includes instream reservoirs in the Sacramento River region (i.e., Lake Shasta, Trinity Reservoir, and Folsom Lake) and conveyance facilities to deliver these supplies to contractors in the SJR region. The state system, referred to as the State Water Project (SWP), contains north-of-Delta storage, Delta export pumps, and south-of-Delta conveyance facilities to serve a community of state water contractors primarily located in Southern California. The coordinated Delta operations of CVP and SWP systems play a large role in determining the reliability of surface water deliveries for west-side irrigators in the SJR region.
Figure 2. Schema of Hydrologic and Water System Features in the SJR Region
To assess the impacts of climate change on water allocation and supplies to the SJR region, it was necessary to simulate changes in reservoir operations for the combined CVP and SWP systems and for the local water systems located on the east side of the SJR region. The model chosen to handle this simulation task is CALSIM II, developed jointly by the California Department of Water Resources (CDWR) and the U.S. Bureau of Reclamation’s (USBR) Mid-Pacific Region office. CALSIM II generates monthly decisions for reservoir storage, releases, and water deliveries for the CVP and SWP systems. Deliveries to the east- and west-side irrigators of the SJR region are simulated by CALSIM II. The model is designed to simulate system conditions given: (1) climate-dependent reservoir inflow assumptions; (2) static agricultural cropping and water demands; (3) static regulatory requirements; and (4) static physical infrastructure. In essence, it is a model that is designed to permit long-term planning analysis (e.g., reservoir enlargements, changes in regulatory regimes, changes in hydrologic regimes, etc.).
The application of CALSIM II for this assessment first involved mapping the results of Miller, et al.’s (2003) assessment on natural runoff sensitivities to the reservoir inflow assumptions of CALSIM II. Details of these responses in the SJR region are described in Brekke, et al. (2004). The following is a summary of impacts and key findings:
- In the HadCM2 case, there would be increased reservoir inflows, increased stored water volumes limited by physical capacity, and increased releases for deliveries and river flows.
- In the PCM case, there would be decreased reservoir inflows, decreased storage and releases, and decreased deliveries.
- Impacts under either projection case cannot be regarded as more likely than the other.
- Most of the impacts uncertainty is attributable to the divergence in the HadCM2 and PCM precipitation projections over California.
- The range of assessed impacts is too broad to guide selection of mitigation projects.
- Regional planning agencies can respond by developing contingency strategies for these cases and applying a methodology therein to evaluate a broader set of CO2 scenarios, land use projections, and operational assumptions.
- Improved agency access to climate projection information is necessary to support this effort.
Agricultural Production, Agro-Economics, and Irrigation Drainage. To simulate more comprehensively the impact of reduced water deliveries on land use in the San Joaquin Basin, a model called APSIDE—Agricultural-Production-Salinity-Irrigation-Drainage-Economics—was developed. APSIDE is an outgrowth of two previous USBR models—the Westside Agricultural Drainage Economics model (WADE) and the Irrigation Drainage Operations model (IRDROP)—and a University of California at Davis model, the Statewide Agricultural Production Model. The model estimates agricultural yield and productivity response to reductions in water supply, irrigation water quality, root zone and groundwater salinity and predicts future agricultural drainage flows and water quality. Contaminated subsurface drainage flows from agricultural lands on the west side of the San Joaquin Basin have a significant impact on SJR water quality. In the absence of a means to export these contaminants, levels of salt and boron build up in the crop root zone and shallow groundwater aquifer, reducing crop yield and leading to reduced agricultural income and eventual retirement of the land when production costs exceed farm income. The crop production function contained within the APSIDE objective function attempts to maximize land productivity by minimizing the costs on agricultural production and salinity impacts on crop yield. Agricultural demand for water is directly proportional to crop production; however, different crops have different water requirements and cultural practices. Cultural practices and the type of irrigation technology employed also affect irrigation efficiency and management practices that, in turn, affect how these technologies are deployed in the field.
The APSIDE model can be thought of as three interacting models that are updated monthly. The first is an agricultural production submodel that projects cropping decisions and estimates the total crop production for a given year based on the economics of production and yield response to soil salinity. The second submodel of ASPIDE computes the water balance for each drainage zone within each subarea and for each of the four vertical layers into which the groundwater aquifer underlying each subarea is subdivided. The third submodel is a salinity mass balance model, which determines mass transport of salts between all four vertical layers in each APSIDE model subarea. Salts introduced to the imported irrigation water supply can evapo-concentrate in the soil root zone unless leached by downward-moving percolating water. Salts transported vertically in the aquifer eventually reach deep layers, where they can be returned to the surface through groundwater pumpage. In west-side areas affected by soil salinity, reliance on groundwater pumping can accelerate salinity buildup in surface soils, leading to potential reductions in crop yield. Salinity-related impacts on yield reduce revenue, which provides feedback to the agricultural production submodel. The APSIDE model, like its predecessor, the IRDROP model, solves the agricultural production model once and computes water and salinity mass balances on a monthly basis. These flows and salt mass loadings are combined from the individual districts to provide input to the Delta Simulation Model 2 (DSM2)-SJR hydrodynamic flow and salinity model of the mainstem SJR.
The APSIDE model was coded in the Generalized Algebraic Modeling System, (GAMS) which is a high-level computer language that aids the formulation of the system of algebraic equations that include the model objective function and system constraints. The CONOPT2 nonlinear solver was determined to be the most appropriate for the model. Although the original plan was to develop the APSIDE model for the entire San Joaquin Basin, this goal was predicated on the availability of the surface groundwater simulation model WESTSIM, a highly detailed simulation model of the entire west-side of the San Joaquin Basin with individual water districts as the unit of analysis. A peer review of the basic computer code used to build the WESTSIM application revealed flaws that delayed calibration of this model by 18 months; therefore an APSIDE application was formulated for just five subareas of the SJR’s west side. Because the water districts included in this initial version of the model were those most impacted by shallow water tables and salinity, they were considered the most likely to respond to changes in water deliveries, and the return flows from these districts the most likely to show changes in water quality.
In the PCM climate scenario, in which a reduction in mean monthly precipitation produces a decline in reservoir storage in the CALSIM II model, the results suggest an average annual reduction in water deliveries of about 50 percent to agriculture on the west side of the San Joaquin Basin (Brekke, et al., 2004) in the worst case scenario. Surprisingly, these reductions in water deliveries did not appear to affect the loading of salts that return to the SJR as predicted by the APSIDE model. Although the concentration of return flows was seen to increase over time, this appeared to be complemented by a decrease in drainage volume leading to very little impact on salt loads. In Figure 3, results are shown for a 10-year APSIDE simulation for a base 25 percent supply reduction scenario, which is equivalent to current south-of-Delta water allocations. The model shows an upward trend in root zone salinity over time in both the Panoche and Broadview water districts. Drainage salinity declines initially in quality in the Panoche water district and then slowly begins to increase after year 7. This reversal in trend is associated with shifts in cropping and drainage management as the model moves from one optimal combination of crops and technology to another.
Figure 3. APSIDE Model Output for a 10-Year Simulation for Two West-Side Water Districts Under Reduced Agricultural Water Supply
The APSIDE model estimates production costs associated with more than 20 combinations of irrigation and drainage practices ranging from low technology, low cost options such as half-mile furrow irrigation without drainage recycling to high cost subsurface drip irrigation. The model modifies irrigation and drainage practices in circumstances in which there is economic advantage. Drainage volumes predicted by the APSIDE model are dynamic because they are the result of potential changes in agricultural land use, irrigation and drainage technology adoption, and land retirement decisions over time. The model suggests that the agricultural system on the west side of the Basin has resilience and is able to adapt to the change in water availability, at least in the intermediate term, without significantly affecting water quality in the SJR.
Model data-linkages between the APSIDE model and CALSIM II and between the APSIDE model and DSM2-SJR (i.e., the DSM2 “rivers” module applied for the mainstem SJR upstream of the Delta) can be performed relatively easily owing to the tabular structure of the GAMS, its ability to export Excel or ASCII text files, and the common DSS flat file format of both CALSIM II and DSM2-SJR. Linkage between the APSIDE and the CALSIM2 and DSM2-SJR models, considered essential at the start of the project, took lower priority and is a task that may be completed by the agencies who have already adopted many of the methodologies developed during the course of this project.
The USBR decided to include APSIDE in its reevaluation of solutions to the drainage problem in the San Joaquin Basin. The model will be used to estimate long-term land use changes in drainage-affected areas and guide land retirement planning. A training workshop was conducted in the fall of 2003 to explain the use of the model to analysts at the USBR. Future work with the APSIDE model would expand the model coverage to the entire salinity-affected west side of the San Joaquin Basin and would rely on a fully coupled surface and groundwater simulation model to simulate the interchange of water between shallow and deep groundwater in the regional aquifer as well as lateral flows between adjacent water.
San Joaquin River Water Quality. The hydrodynamic model DSM2 applied to the Delta region simulates tidal flows and mass transport of salt and various constituents affecting water quality using the continuity equation. The DSM2-SJR model was extended for use in this study to produce a distribution of the water quality along a reach of the SJR from its confluence with Bear Creek to just beyond its confluence with the Stanislaus River, where the Vernalis monitoring station is located. This work was performed largely by the CDWR with assistance from the project team to develop a realistic flow and salinity calibration. The calibrated model CALSIM II provides an approximate estimation of the electro-conductivity in the river at Vernalis, but it is based on an empirical relationship between streamflow and electro-conductivity.
The DSM2-SJR model has been calibrated with the baseline scenario from CALSIM II using a DSS linkage program created within Excel. This spreadsheet linkage aids transfer of data between the two models. Work began on developing a methodology for CALSIM II/DSM2-SJR feedback that can be incorporated as an additional arc in CALSIM II, representing releases from New Melones Reservoir to meet water quality requirements at Vernalis. During the last year of project, however, the issue of the flow/electro-conductivity relationship at Vernalis was given priority. The USBR devoted internal staff and a team of consultants to understand the issue more fully and develop a methodology to improve the simulation of Vernalis water quality that would be acceptable to both water agencies. At the time of the writing of this report, a decision had been made to develop a water quality mass balance module within CALSIM II that emulated DSM2-SJR. The U.S. Environmental Protection Agency (EPA) Science To Achieve Results (STAR) project can be credited for doing some of the preliminary analysis that brought this issue to light and paved the way for its ultimate resolution.
Aquatic Ecosystem and Fisheries . The effects of climate change and/or extreme hydrologic responses to climate change upon the health of aquatic ecosystems in the San Joaquin Basin were assessed using fisheries populations as a proxy for the health of aquatic ecosystems. Historical spawning runs of chinook salmon in the Sacramento-San Joaquin Basin were on the order of one million fish, but major habitat alterations have limited current San Joaquin populations to four major tributaries draining the east side of the basin: the Merced, Tuolumne, Stanislaus, and Mokelumne Rivers (Healey, 1991; Kano, 2002). The causes for the diminished chinook runs are numerous, including impassable dams, diminished streamflow, and concomitant elevated water temperatures (Kjelson, et al., 1982). Because of the inherent difficulties in physically modeling ecosystems, this study used chinook salmon populations as a proxy for the vitality of riverine habitats.
Using the flow conditions in the SJR’s major tributaries simulated by CALSIM II, a modeling study was conducted on the impacts on chinook salmon populations. To address the formidable nonlinearities in biological systems, salmon population models were developed using an artificial neural network (ANN) approach. ANNs have been employed in a wide variety of research and practical applications because of their ability to discern complex numerical patterns within data, even including contributions that are presently not understood (Silverman and Dracup, 2000; Nelson and Illingworth, 1991). Consequently, an ANN model is well suited to the complex task of describing the time-dependent response of chinook salmon populations to the degradation of their natural habitat, which is superimposed upon natural population variability.
Annual salmon escapements (i.e., counts of spawning salmon) were used with streamflow discharge data to develop salmon population models for three San Joaquin tributaries: the Merced, Tuolumne, and Stanislaus Rivers. Key findings include:
- Salmon populations varied according to the year of the projection and the general circulation model used to produce the streamflow estimates.
- The HadCM2 model presents a wetter future climate, with significantly more (five to seven times) chinook salmon based on response to flow conditions. In contrast, the PCM foresees a drier climate than today, with an slight reduction (6% drop) in the median number of spawning salmon.
- Effectively, the HadCM2-based and PCM-based conditions represent an envelope for the range of conditions that future chinook salmon populations could experience in the SJR Basin.
Socioeconomics . Socioeconomic impacts of climate change in the San Joaquin Valley mostly derive from changes in agricultural production and agricultural profitability in the San Joaquin Basin. Salinization of agricultural land leads to poorer crop yields and a reduction in farm profit. A multiplier effect is typically assumed, whereby a dollar of production at the farm gate stimulates economic activity worth several dollars in the local community. .
As previously described, we were surprised by the resilience of the agricultural system in the San Joaquin Basin, as simulated by the APSIDE model, to potential climate change. The HadCM2 climate projections signaled wetter conditions, and any economic losses under this set of climate scenarios were associated more closely with flooding and land inundation in the Merced River watershed and lower SJR. Under the much drier PCM climate projections, which forecast up to a 50 percent reduction in water deliveries south of the Sacramento-San Joaquin Delta, the magnitude of crop revenue reduction did not reflect initial expectations. Figure 4 shows the crop revenue impacts for the same two water districts, Broadview and Panoche, for the same time period and water supply restrictions. Cotton, a somewhat salt-tolerant crop, is shown to decline over time in the Panoche water district and perhaps was replaced with a more profitable crop. Cotton acreage remains largely unaffected in the Broadview water district. Gross farm profits (combining all crops grown) remain somewhat constant in both districts.
At the beginning of the project, we intended to use the IMPLAN software to estimate economic multipliers for different sectors of the economy within the San Joaquin Basin. The socioeconomic impacts forecasted by the APSIDE model under climate change scenarios were of insufficient magnitude to warrant a full impacts analysis using IMPLAN. Our analysis showed that crop substitution, adoption of improved irrigation technologies, and limited drainage reuse could make up for reductions in long-term water supply forecasts stemming from the PCM climate projection.
Figure 4. APSIDE Results for Panoche and Broadview Water Districts Showing Cotton Acreage Over Time and Resulting Gross Farm Profit
Linking Project Work/Products to Stakeholder Efforts
Model integration was considered the highest priority for the impacts models CALSIM II, APSIDE, and DSM2-SJR. Collectively, these models contain hundreds of decision variables, many of which would need to be revised based on applicable policy and the level of development assumptions. The modeling toolbox we developed in this study involves the integration of state-of-the-art resource management models newly developed within the state and federal water agencies in California. The toolbox minimizes the time required for file manipulation and formulation of impact response scenarios, allowing the analyst to simulate the impacts of global climate change on important California resources such as water supply, water quality, agricultural production, and economic activity. These factors are key to the development of secondary and tertiary impacts assessments that address issues such as the California fisheries, endangered species, and socioeconomic welfare. Integration using a public domain toolbox modular modeling system/object user interface (MMS/OUI) failed owing to a lack of that software’s conformity to the monthly timestep required by both DSM2-SJR and CALSIM II applications. This was unfortunate because the MMS/OUI system has been in development for more than a decade and would have saved us considerable time and effort had it proved to be customizable to our applications. Instead, a less elegant, but more robust, data and analysis browsing system has been developed that satisfies the project goals.
California Department of Water Resources (CDWR). CALSIM II benchmark study released 12/14/2001.
Daly C, Neilson RP, Phillips DL. A statistical-topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology 1994;33(2):140-158.
Healey MC. Life history of chinook salmon. In: Groot C, Margolis L, eds. Pacific Salmon Life Histories. Vancouver: University of British Columbia Press, 1991, pp. 312-393.
Howitt RE, Mean P. Positive quadratic programming model. University of California at Davis, Department of Agricultural Economics, 1985.
Intergovernmental Panel on Climate Change (IPCC). Third assessment report: climate change 2001: the scientific basis. Cambridge: Cambridge University Press, 2001, 881 pp.
Kano B. Chinook salmon escapement data collected by the California Department of Fish and Game, Sacramento, CA, 2002.
Kjelson MA, Raquel PF, Fisher FW. Life history of fall-run juvenile chinook salmon, Oncorhychus tshawytscha, in the Sacramento-San Joaquin Estuary, California. In: Kennedy VS, ed. Estuarine Comparisons. New York: Academic Press, 1982, pp. 393-411.
Nelson MM, Illingworth WT. A Practical Guide to Neural Nets. New York: Addison-Wesley, 1991, 344 pp.
Silverman D, Dracup JA. Artificial neural networks and long-range precipitation prediction in California. Journal of Application Meteorogy 2000;39(1):57-66.
Journal Articles on this Report : 4 Displayed | Download in RIS Format
|Other project views:||All 21 publications||5 publications in selected types||All 4 journal articles|
||Brekke LD, Miller NL, Bashford KE, Quinn NWT, Dracup JA. Climate change impacts uncertainty for water resources in the San Joaquin River Basin, California. Journal of the American Water Resources Association 2004;40(1):149-164.||
||Miller NL, Bashford KE, Strem E. Potential impacts of climate change on California hydrology. Journal of the American Water Resources Association 2003;39(4):771-784.||
||Quinn NWT, Miller NL, Dracup JA, Brekke L, Croben LF. An integrated modeling system for environmental impact analysis of climate variability and extreme weather events in the San Joaquin Basin, California. Advances In Environmental Research 2001;5(4):309-317.||
||Quinn NWT, Brekke LD, Miller NL, Heinzer T, Hidalgo H, Dracup JA. Model integration for assessing future hydroclimate impacts on water resources, agricultural production, and environmental quality in the San Joaquin Basin, California. Environmental Modelling and Software 2004;19(3):305-316.||