Using Linked Models for Basin Management: Challenges and Opportunities

For many years, some water resource managers have employed watershed models to simulate hydrological and water quality processes and quantify impacts of land management practices in large, complex watersheds. Other water resource professionals have applied lake water quality models to simulate hydrological and water quality processes and predict effects of various management practices for large lakes. For hydrologically related systems (i.e., upland watersheds whose runoff drains into a lake), it follows that these two types of models could be used in tandem to provide a better understanding of more complex systems and yield more predictive capabilities, thereby supporting more effective management decisions including those decisions that would be central to the implementation of a Total Maximum Daily Load (TMDL). However, linkage of these two model types (i.e., watershed to traditional water quality) is not always as seamless as it might first seem. There are typically challenges with such a linkage, such as matching up state variables between the two model structures, determining what output from the watershed model should be “fed into” the water quality model, and rectifying temporal and spatial differences in the model domains. This paper discusses how one such linkage was accomplished between Soil and Water Assessment Tool (SWAT) and CE-QUAL-W2, respective watershed and lake models both widely used around the world. It focuses on the challenges experienced with such a linkage and presents lessons learned for future applications. Although the focus is on two specific modeling packages, many of the challenges and lessons are universal for most watershed-water quality linkage frameworks.Examples of the challenges encountered included the deconvolution of SWAT state variables to CE-QUAL-W2 state variables. This deconvolution is crucial because several variables output by SWAT are not directly translatable to inputs to CE-QUAL-W2. In some cases, CE-QUAL-W2 requires input variables in discrete groups whereas SWAT combines these groups in its output variables; for example, sediment loads from SWAT need to be adjusted before being input into CE-QUAL-W2 as inorganic suspended solids without particulate organic matter. In other cases, SWAT output needs to be broken down into components to take advantage of CE-QUAL-W2's capability of simulating finer groups of variables; for example, organic matter groups predicted by SWAT can be separated into carbon, nitrogen, and phosphorus organic matter groups for input into CE-QUAL-W2. In addition, it is necessary to convert the units for some SWAT outputs by way of stoichoimetry prior to insertion into CE-QUAL-W2. Conversions were done via a custom intermediary program for the following state variables: inorganics (inorganic suspended solids, orthophosphorus), nitrogen (ammonia, nitrate and nitrite); organic matter (labile and refractory, dissolved and particulate); algae; and dissolved oxygen. Other challenges included temporal and spatial resolution differences between the two models, as well as the treatment of the hydrology calibration from SWAT in relation to the hydrodynamic “flow balance” required by CE-QUAL-W2.A case study is presented that linked these two models for Lake Travis, Texas as part of the Lower Colorado River Authority's Colorado River Environmental Models (CREMs) project. The linked Lake Travis model is one of several tools to be used by water resource managers to provide information for supporting policy decisions that proactively and effectively protect the integrity of the water resources in the Lower Colorado River basin. The case study reveals the challenges and opportunities from using linked watershed and water quality models for basin management. This type of modeling framework can be instrumental in supporting basin management decisions such as those necessary for the successful implementation of a TMDL.