*3.1. WARMF Water Quality Simulation Model*

The San Joaquin River Basin application of the public-domain, Watershed Analysis Risk Management Framework (WARMF) model [14,24] was developed in 2004 by Systech Water Resources Inc. as a TMDL decision support tool. The first application of the model was to assess options for control of dissolved oxygen sag in the SJR Deep Water Ship Channel [23–25]. The SJRB WARMF model application is a physically-based, data-intensive watershed model that simulates the hydrologic, chemical, and physical processes in the river and contributing waterbodies (Figure 2). The model was derived from the San Joaquin River Input–Output (SJRIO) model [26,27]. The model was updated and reconfigured as a salinity forecasting tool in 2014 [14,24] as the USBR's contribution to stakeholder-led real-time salinity management activities. The WARMF model application simulates flow and water quality in surface water diversions, groundwater pumping, and irrigation water supply, while keeping track of crop evapotranspiration, seepage, and irrigation surface and subsurface return flows [25]. Delineation of land catchments in WARMF conforms to both irrigation and drainage district boundaries and natural catchments, allowing the model to track salt loads from their points of diversion in delivery canals back to the river [25].

The data-intensive WARMF model is supplied with daily meteorology, diversion flows, and measured flow and electric conductivity (EC) at the upstream model boundaries [23,25]. The current upstream model boundaries are at gages where flow and EC are measured continuously in the SJR and along its major tributaries including the Merced River, Tuolumne River and Stanislaus River. Real-time data, tributary reservoir release forecasts, and meteorology forecasts are collected and imported into WARMF using an automated process consisting of custom scripts and web scraping tools that interact with agency web portals for hydrology and water quality monitoring [25,28]. WARMF model data acquisition accesses seven agency web portals and is accomplished as a separate data acquisition and pre-processing routine.

The combination of real-time monitoring, simulation modeling and forecasting of SJR assimilative capacity has the potential to optimize use of available river salt assimilative capacity, generated by releases of high quality Sierran water, which provides dilution to saline west-side agricultural and managed wetland return flows. However, there needs to be coordination and sufficient lead time to allow entities being asked to charge drainage practices or alter reservoir release patterns to be able to respond. Agricultural return flows and salt loads are highest during the summer irrigation season whereas return flows and salt loads from seasonally managed wetlands are highest during the spring months of March and April, when most seasonal wetland ponds are drained to promote establishment of moist soil plants and habitat for waterfowl [15]. These anticipated hydrologic patterns help to screen the array of practices on both the east and west sides of the basin that will be most effective at managing salinity.

**Figure 2.** Map of the SJR Basin represented as major contributing watersheds within the WARMF model. The WARMF model allows further disaggregation of these watersheds into small contributing subareas and allows the substitution of available data at the major outlets of these subareas for modelderived flow and water quality estimates.

Given the uncertainty associated with estimates of salt assimilative capacity, the need for adequate lead time for stakeholders to adjust tributary inflow and drainage return flow schedules and the fact that most weather forecasts provided by news organizations rarely extend beyond two weeks—a two-week forecast period and a one-week hindcast period was chosen for the real-time salinity management program. The one-week hindcast refers to the technique of beginning the simulation one week in arrears so that the first week of the forecast can be compared to observed flow and electrical conductivity (EC) data [14,16,22]. Model parameters affecting river and tributary inflow and water quality such as the partitioning coefficients that allocate watershed runoff and deep percolation to groundwater can be adjusted to recalibrate the model during periods when model output and river observations diverge. This activity is infrequently performed due to the significant effort involved and the fact that the WARMF model has exhibited excellent performance for simulation of flow EC and EC along Reach 83 of the SJR. Simulated flow and EC are compared to measured data along the SJR for model calibration including drainage return flows from east- and west-side catchments and direct diversions from the SJR to riparian water districts. Although agricultural and managed wetland stakeholders have yet to fully embrace the model as a decision support tool both have concurred that the suggested two-week forecast and one-week hindcast periods are a good compromise balancing the utility and credibility of the forecasts with the time stakeholders might need to adjust water management and drainage discharge operations.

The SJR WARMF model has a number of customized output visualization options designed to enhance user understanding of salinity fate and transport in the SJR Basin and the use of salt load assimilative capacity by river mile along the mainstem of the SJR [28]. The output visualization also allows users to estimate if and when the salinity concentration at the compliance monitoring sites will approach or exceed objectives. The model is also capable of showing the impact of potential salinity management changes in the watershed designed to comply with regulatory limits (Figure 3). For example, the SJR WARMF model can simulate the effect of increased irrigation water diversions from the river into riparian water districts, lowering salt loading in the river, which may help to improve compliance with salinity concentration objectives [28].

**Figure 3.** A unique feature of the WARMF model is the availability of customized model outputs such as the "Gowdy" output (named after its developer) shown here. This depicts a Lagrangian view of the SJR at any point in time showing the major inflow to and diversions from the river approximately every <sup>1</sup> <sup>2</sup> mile (800 m) along its main reach as well as the incremental flow and EC concentration from the origin at Lander Avenue to the EC compliance monitoring station at Vernalis [23,25].

The SJR WARMF model has been improved and customized over the past 15 years with the USBR and research grant support as a watershed-based simulation tool for flow and salinity forecasting in the SJR [14,25]. Updating time series data inputs and maintaining model calibration are expensive and time consuming. This constraint has restricted the stakeholders' pool and agency individuals able to run the model on a regular basis and has been an impediment for stakeholder entities such as the San Joaquin Valley Drainage Authority (SJDVA) to take over operation and maintenance of the model as a decision support tool. As a result, the USBR evaluated other approaches for providing flow and salinity forecasts of SJR at Vernalis and Crows Landing, the two salinity concentration compliance points for the TMDL. Although the WARMF model has been used for various decision support activities in the SJR for over 15 years, other less data-intensive and more easily understood approaches may be better received by stakeholders [28].
