Computer Model Improves Real-Time Management of Water Quality

The San Joaquin River sometimes contains high levels of selenium, boron and salt from drainage water. Reservoir releases can be timed to dilute these concentrations.

Computer model improves real-time management of water quality

Nigel W.T. Quinn P Leslie F. Grober o Jo-Anne Kipps P Carl W. Chen o Earle Cummings

Members of the San Joaquin The San Joaquin River provides essen- The major water-quality problems River Management Program’s tial drainage for agricultural land and on the San Joaquin River are caused by Water Quality Subcommittee have managed wetlands in the San Joaquin high loadings of selenium, boron and developed a water-quality fore- Valley. The river has an average an- salt, much of which is contained in ag- casting model to help improve the nual flow of approximately 1.8 million ricultural subsurface drainage from timing, coordination and manage- acre-feet, measured at Vernalis. Dur- the west side of the San Joaquin Val- ment of agricultural drainage and ing the period 1985 to 1994, an average ley. The Mud and Salt sloughs are reservoir releases into the San of 70% of the annual flow in the San tributaries of the San Joaquin River Joaquin River. The goal of this Joaquin River was derived from the and contribute approximately 85% of major east-side tributaries (Stanislaus, the selenium load, 65% of the boron effort is to improve water quality Tuolumne, Merced and Upper San load and 45% of the salt load carried in the San Joaquin River and to Joaquin rivers), which receive flow by the river. Seasonal drainage from meet federal and state water- from New Melones Reservoir, New 90,000 acres of wetlands in the Grass- quality objectives for salt, boron Don Pedro Reservoir, Lake McClure land Water District and in state and and selenium. A graphical user and Millerton Reservoir. The remain- federal refuges contributes approxi- interface has been developed and der of the flow during this period mately 20% of the boron load and 30% features have been added to consisted of surface drainage (14%), of the salt load discharged from the make this computer software use- subsurface drainage (I %) and contri- Mud and Salt sloughs to the San ful and accessible to a wide range butions from Salt Slough (9%)and Joaquin River. The average annual salt of decision makers. Mud Slough (2%). load carried by the San Joaquin River

14 CALIFORNIA AGRICULTURE, VOLUME 51, NUMBER 5 is approximately 920,000 tons, of load that the river can accept which the east-side tributaries contrib- without exceeding EC objec- ute 16%,Salt Slough 33% and Mud tives at Vernalis. Positive as- Slough 11%.The remainder of the salt similative capacity indicates load consists of groundwater (21%) that additional drainage dis- and surface drainage (16%)(fig. 1). charges can be made to the river without violating water- Assimilative capacity quality objectives. Negative as- There is a large year-to-year and similative capacity indicates a seasonal variation in water quality in need for additional dilution. the San Joaquin River (fig. 2). Down- Past analyses using mass-balance stream users are particularly sensitive models of the river, such as the State Fig. 1. Sources of total dissolved solids to elevated salinity during the spring Water Resources Control Board’s San (TDS) contributed to the San Joaquin months of March and April when Joaquin River Input-Output model River for water years 1985-1994. preirrigation and crop germination oc- (SJRIO), suggest that considerable op- cur. March and April are also the portunity exists for improved coordi- months when many seasonal wetlands nation of drainage discharges and res- are drained for reseeding to provide ervoir releases to more efficiently use overwintering habitat for migrating river assimilative capacity for salt, bo- wildfowl. Spring pulse flows for salmon ron and selenium while complying migration also occur during this time, with water-quality objectives. Comple- providing potential dilution flows. mentary studies by the US. Bureau of Electrical conductivity (EC) is di- Reclamation - using a model that rectly correlated with the concentra- combined irrigation, drainage and tion of dissolved salts. In the 1985-to- crop-production functions - sug- 1994 period, the electrical conductivity gested that successful long-term op- at Vernalis exceeded the State Water eration of such a system might be Resources Control Board (SWRCB) possible without impairing crop water-quality objective of 700 micro- yields and forcing irrigated land out Siemens per centimeter (pS/cm) 61% of production. Fig. 2. Variation in annual TDS loading by of the time during the irrigation sea- Recent actions initiated by local wa- water year, 1985-1 994. son (April l-Aug. 31). The non- ter districts to improve compliance irrigation season (Sept. 1-March 31) with water-quality objectives include currently made up of representatives EC objective at Vernalis of 1,000 pS/ increased drainage recycling; use of from the California Department of cm was exceeded 15% of the time. short-term, on-farm surface and sub- Water Resources (DRW), Lawrence These objectives were set as part of a surface storage; and the adoption of Berkeley National Laboratory and negotiated settlement between the water conservation and drainage re- California Regional Water Quality U.S. Bureau of Reclamation (USBR) duction technologies to increase con- Control Board. In 1994 the SJRMP- and the South Delta Water Agency. trol of drainage discharges and con- WQS was awarded a U.S. Bureau of Although the U.S. Bureau of Recla- taminant loads. In addition, real-time, Reclamation Challenge Grant to dem- mation tries to meet these water-qual- telemetered data-acquisition systems onstrate that improved management ity objectives by making releases from have been installed on the major tribu- and coordination of tributary releases New Melones Reservoir, water-quality taries to the San Joaquin River, at com- and agricultural drainage contaminant forecasting capability has been inad- pliance points on the river as well as at loads from west-side sources could equate to prevent frequent violations. drainage outlets from the agricultural significantly reduce the frequency of Water is sometimes wasted when re- water districts. violations of water-quality objectives leases exceed the amount required for for salinity, selenium and boron on the dilution. Precisely timing releases is San Joaquin River Management main stem of the San Joaquin River. difficult because of the fluctuating EC The California Legislature estab- The SJRMP-WQSidentified the at Vernalis and the uncertain travel lished the San Joaquin River Manage- need to develop a decision-support time between the reservoir and the ment Program (SJRMP)in 1990 to system. This system would provide main stem of the San Joaquin River. identify some of the pressing water- current flow and water-quality data Improved forecasting of river assimi- management and water-quality prob- and forecasts of river assimilative ca- lative capacity for salt load would lems in the San Joaquin River. A Water pacity for selenium, boron and salin- help to remedy this problem. Quality Subcommittee (SJRMP-WQS) ity to decision makers. The system Assimilative capacity for salt, as was formed to implement solutions to would interrogate existing monitoring used here, is defined as additional salt these problems. The SJRMP-WQSis stations that continuously measure

CALIFORNIA AGRICULTURE, SEPTEMBER-OCTOBER 1997 15 San Joaquin River at Vernalis Bridge (EC, flow, temp) Grasslands Bypass opposite Pond 7, Kesterson Reservoir (EC, flow, temp) Orestimba Creek (EC, flow) Secondary (not yet installed) San Joaquin River at Newman Bridge (flow) San Joaquin River at Lander Av- enue Bridge (EC, flow, temp) Patterson Irrigation District diversion from San Joaquin River (flow) West Stanislaus Irrigation District diversion from San Joaquin River (flow) The first 10 stations on the list are located at primary sites that are essen- tial for water-quality forecasting (fig. 3). The last four stations are at secondary sites that will help improve the reli- ability of flow and water-quality forecasts once they are installed. The sensors deployed at each site collect stage, EC and temperature data. At each station, discharge is cal- culated from stage readings with the application of a rating curve, which specifies the stage-discharge relation- ship and applies any curve shifts to take account of changes in river hy- flow and water-quality parameters, trations. The U.S. Bureau of Reclama- draulics or bed configuration. using telemetry equipment to allow tion Challenge Grant required that the Dataloggers and cellular telephones rapid access to current data and using project demonstrate the utility of the are used at each station to record data computer models to make reliable final product. at 15-minute intervals and to transmit forecasts of future flow and water- the recorded data to computers at the quality conditions. These forecasts Real-time monitoring USBR, DWR and the U.S. Geological would be useful for timing and coordi- The SJRMP-WQSidentified 14 sites Survey (USGS). nating west-side flows and contami- for continuous monitoring of flow, EC Real-time management principles nant loads from agricultural fields, and temperature along the San wetlands and wildlife refuges with Joaquin River and its tributaries. These Using the hydrologic year 1993 as east-side reservoir releases for salmon monitoring sites, with the type of data an example, the San Joaquin River as- migration, recreation, power and wa- collected at each site, are listed below: similative capacity for salts was plot- ter quality. The SJRMP-WQS hired ted as an area graph (fig. 4). The river Systech Engineering Inc. to design a Primary (existing) assimilative capacity depicted by the Microsoft Windows-based user inter- Mud Slough in Kesterson Wildlife graph is shown as positive from Octo- face for the forecasting model. The Refuge, near Gustine (EC, flow, ber 1992 through February 1993, nega- interface was to be simple and under- temp) tive from March through April, posi- standable to the novice user and ca- Salt Slough at Highway 165 Bridge tive from May until June, negative in pable of uploading the data necessary to (EC, flow, temp) July and positive in August and Sep- run the model. Merced River near Stevinson (EC, tember of 1993. Eliminating the nega- The SJRMP-WQSlimited the water- flow, temp) tive peaks requires a shift of the excess quality parameter of concern to electri- Tuolumne River at Modesto (flow) salt loading to periods when river as- cal conductivity because EC can be Stanislaus River at Ripon (EC, flow) similative capacity is positive. Dis- monitored continuously with commer- Stanislaus River at Orange Blossom charging salts earlier in the non- cially available sensors, whereas the Bridge (flow) irrigation season during water year technology is not yet available to San Joaquin River at Crows Land- 1993 might have allowed compliance monitor selenium and boron concen- ing Bridge (EC, flow, temp) with EC objectives (fig. 5). A practical

16 CALIFORNIA AGRICULTURE, VOLUME 51, NUMBER 5 application of this strategy might be an early release of ponded water by refuges and duck clubs in the Grass- land Water District. Early release and refilling of refuges and duck clubs would increase the salt loading during February and reduce the average EC of the ponded water for release during March and April. This strategy might also benefit wildlife by making more food available to birds during February at the mar- gins of the draining ponds. To improve communica- tion among individuals in- volved in decision making related to water quality on Site H: The San Joaquin River upstream the San Joaquin River, the SJRMP- from the confluence with the Merced WQS set up an electronic listserv on a River. USBR DataGeneral workstation. The daily flows and concentrations of total listserv forwards e-mail sent by one dissolved solids (TDS), boron and sele- person to everyone who has sub- nium for a 60-mile reach of the San scribed to the service. The listserv cur- Joaquin River from Highway 165 to 5 miles of the Merced River below rently has more than 60 subscribers Vernalis. The original version of the the USGS gauging station near representing more than 15 agencies model, SJRIO, and the database cre- Stevinson and organizations. Information posted ated to run the model were used in de- 15 miles of the Tuolumne River be- on the listserv includes weekly up- veloping water-quality objectives and low the USGS gauging station at dates of past flow and water quality in evaluating agricultural drainage re- Modesto on the San Joaquin River, current duction strategies to meet these objec- 9 miles of the Stanislaus River be- water-quality forecasts, notices of tives. The SJRIO was modified to run low the DWR gauging station near meetings and experiments and notices on a daily time step so that it could be Ripon of major actions that might affect flow used with real-time flow and water- 6 miles of Salt Slough below the or water quality. quality data on the San Joaquin River. USGS gauging station near An extensive database has been as- Stevinson SJRIODAY model sembled, with data for water years 9 miles of Mud Slough below the The decision-support system devel- 1977 through 1996, to run the model. USGS gauging station near Gustine oped by the SJRh4P-WQScontains a The daily model, SJRIODAY, con- Several miles of three west-side graphical user interface and a daily tains the following tributary river tributaries: Del Puerto, Orestimba mass-balance model that calculates segments: and Hospital/Ingram creeks

Fig. 4. Assimilative capacity for daily TDS in the San Joaquin Fig. 5. Electrical conductivity (EC) for San Joaquin River at River near Vernalis, 1993. Vernalis showing 30-day running average EC objectives, 1993.

CALIFORNIA AGRICULTURE, SEPTEMBER-OCTOBER 1997 17 Site N: San Joaquin River at Crows Landing Bridge. The box contains a datalogger for the sensors, which measure electrical conductivity and height of the river. Daily flow cal- culations use the concept of mass balance and the assumption of fixed lag times between tributary nodes on the main stem of the San Joaquin River. Both his- torical and current real-time data are used to establish initial conditions for model runs and to generate a 2- week forecast of flow and EC. The validity of the first week can be judged based on real-time data available for jor model inputs and other model as- monitoring data and to run the fore- the San Joaquin River at Crows Land- sumptions are provided, together with casting model SJRIODAY. Weekly ing and Vernalis. Discrepancies be- model results. water-quality forecasts for the San tween modeled and observed data can Results from model runs were Joaquin River have been made by the then be used to update and correct in- posted 2 weeks in arrears during 1996, SJRMP-WQS since January 1996. put data for the next model run. providing a retrospective of model The GUI is typically used to make Forecasted rainfall can be used to performance and allowing the SJRMP- flow and water-quality forecasts ac- account for additional runoff in the ba- WQS to develop a working knowledge cording to the following sequence of sin. Real-time data is supplemented by of operations in the San Joaquin sys- operations: mean monthly flow and water-quality tem. To provide time for users to 1. Acquire operational schedule data for model components for which evaluate the merits of the model and from water managers, including reser- no real-time data is available, includ- to prevent them from acting in an un- voir operators, wetland managers, ag- ing groundwater, riparian and appro- coordinated manner on the informa- ricultural water district managers and priative diversions, surface and sub- tion provided, water-quality forecasts fishery biologists who are responsible surface agricultural return flows, have not been made available to users for setting fish flow schedules. riparian evapotranspiration, evapora- immediately. 2. Acquire real-time flow and water- tion and precipitation. These compo- quality data by telemetry to serve as ini- nents are estimated in the model Graphical user interface tial conditions for the water-quality based on seasonal variability and There are two versions of the model. water-year classification provided by graphical user interface (GUI). The 3. Forecast water-quality condi- the modeler. general version for water operators tions with SJRIODAY, using the fore- Model results for this 2-week pe- has the capabilities to edit and upload casted operational schedule obtained riod represent a model estimate of operational schedules of reservoir re- in (1) and the initial conditions ob- flow and water-quality conditions leases, to download the results of com- tained in (2). based on forecasted data. Validity of puter runs using the forecasting model 4. Display forecasted water-quality these results cannot therefore be and to display the output from these conditions to show pulses of water judged at the time the run is made; re- runs. This version does not allow the passing various points in the San sults can be judged only as real-time user to make a full model run. The full Joaquin River. data become available. For the end version of the GUI has the same capa- 5. Evaluate whether the water-qual- user to appraise the validity of the bilities as the operators’ version but ity objectives are met at various points model, flow and EC forecasts for ma- also allows the user to download and times.

18 CALIFORNIA AGRICULTURE, VOLUME 51, NUMBER 5 6. Display the assimilative capacity of the San Joaquin River for salt at various locations and times. 7. Provide information to water managers to help adjust operating schedules to meet water-quality objectives. Alert other managers to these operational changes and redo forecast of water-quality conditions. Model results and forecasts Forecasts of flow and water quality at Vernalis have been made weekly since November 1995 and a postaudit of forecast accuracy has been broad- cast on the electronic listserv, compar- ing the forecasts with observations obtained from the real-time monitoring system. Figures 6 and 7 show the performance of the forecasting model for predicting flow and TDS at Vernalis. As might be expected, the observed and model-simulated flows at Vernalis and the observed and simulated TDS concentrations and assimilative ca- pacities are in closer agreement for the 1-week forecast than for the 2-week forecast. The model performed well during most of 1996 and in particular during the summer months, when flows and water quality on the San Joaquin River were dominated by agricultural drainage from Mud and Salt sloughs (fig. 6). In general, the model tends to overestimate flow as well as EC. Between Dec. 25,1996, and Jan. 25, 1997, the San Joaquin Valley was sub- jected to severe winter storms, which produced an extraordinary volume of runoff from the east-side Sierran wa- tersheds. Figure 7 illustrates the problems encountered in making accurate flow forecasts during this period. Without an accurate watershed model, runoff forecasts were based on esti- mates of the flood hydrograph from each contributing watershed and real- time flow data. Although both model and forecast flows continued to overestimate real-time flows between Jan. 14 and Jan. 25,1997, levee breaks along the San Joaquin River diverted considerable flood flow and may account for much of the error. The domi- nance of east-side tributary flows on San Joaquin River water quality

CALIFORNIA AGRICULTURE, SEPTEMBER-OCTOBER 1997 19 DATE DATE Fig. 6. Comparison of model forecasts and real-time flow and Fig. 7. Comparison of model forecasts and real-time flow and salinity data, 1996. salinity data, 1997.

during this period made the EC some- time to make this release. A model proved with additional continuous- what easier to predict with accuracy. forecast made on Jan. 15,1996, sug- flow and water-quality monitoring Model and forecast EC are not signifi- gested that the combination of high stations in both agricultural water dis- cantly different from the real-time EC river flows and an imminent rainstorm tricts and wetland areas in the Grass- data (fig. 7). might provide the necessary assimila- lands Basin. A greater challenge will tive capacity. be to gain the cooperation of the east- Case study The peak wetland release was side reservoir operators, wetland wa- During early January 1996 the timed to coincide with the peak flow ter managers and agricultural water Grassland Water District, which had in the San Joaquin River. Wetland districts in manipulating their release acquired supplemental water under flushing began on Jan. 18 and ended schedules. Each institution has its own the Central Valley Project Improve- on Feb. 19, with the peak wetland dis- operating rules that are followed to ment Act, sought an early release of charge occurring on Jan. 27 (fig. 8). satisfy the needs of its clients and cus- ponded water to reduce the 1i.kelihood The flow at Vernalis was 2,600 cfs on tomers. At the present time, there are of downstream salinity impacts and Feb. 18, and the EC peaked at 806 pS/ few incentives or requirements for salinity objective violations at Vernalis cm on Jan. 20 before the arrival of the these institutions to assist in water- in the San Joaquin River. The water wetland discharge water. The peak quality management of the San Joaquin district requested that the SJRMP- discharge at Vernalis was 5,300 cfs on River. Policy changes and financial in- WQS forecast the most advantageous Feb. 1; the EC at Vernalis declined to centives may be needed to foster coop- 222 pS/cm on Feb. 19 (fig. 8). Because eration from these institutions. of the rainfall-runoff events in the upper watershed, assimilative capacity was positive in the river throughout N. W.T. Quinn is Staff Geological Scien- the simulation period. The EC objec- tist, Lawrence Berkeley National Labora- tive in the San Joaquin River was tory, and Water Resources Engineer, US. achieved during the trial period as Bureau of Reclamation; L.F. Grober is Wa- well as during the regular wetland dis- ter Quality Control Engineer, California charge period in March and April Regional Water Quality Control Board; J. 1996. Kipps is Associate Engineer, Department of Water Resources; C.W. Chen is Presi- Fig. 8. Flow and EC in the San Joaquin Future work dent, Systech lnc.; and E. Cummings is River near Vernalis showing period of peak wetlands discharge, Jan. 1CFeb. 29, The accuracy of San Joaquin River Wetlands Coordinator, Department of 1996. water-quality forecasts could be im- Water Resources.

20 CALIFORNIA AGRICULTURE, VOLUME 51, NUMBER 5