Model Selection

Selection of a watershed model is an important decision not only because of the time and resources a modeling effort involves, but also because of the technical expertise required to maintain a model. Before selecting a model for a watershed modeling effort, it has to be seen whether the model that fulfills other selection criteria is a public domain model or a commercial model. Because the cost of current version of the selected model and the cost of its renewed versions that may be available in future is a major factor if it is to be used in a developing country. Furthermore, the selection of the model should encompass not only current level of detail, but should also consider the future needs. The selected model should provide the flexibility from simple to complex level of details. Hence, before selecting a model, watershed

can fulfill the modeling needs of a number of future projects, it may be advantageous to use this particular model for a current project even when the model is less than optimal for the current application. USEPA (1992) advises that it is desirable to select a model that meets the most application requirements and has demonstrated applications and continuous support from the developer and user communities. Even if the model is not ideal, USEPA (1992) recommends that the user allow for the development of in-house expertise, rather than switching models from application to application.

The current generation of watershed models is quite diverse and varies significantly in data and computational requirements (ESCAP-UN, 1997; Chen, 2001; Singh and Woolhiser, 2002). Watershed scale models may be classified as physically based or conceptual model. In the former, all physical processes are described mathematically and such models are constructed from physically based representations of processes and use parameters determined from known watershed characteristics through measurements or by estimation. In the latter, the physical processes are simplified on the basis of empirical rules (Al-Abed and Whiteley, 2002; Hayashi et al., 2004). For a watershed in a developing country with limited data resources, the selection of a conceptual or calibrated–parameter model will ease the data requirements for model parameterization (Mishra and Singh, 2004). Furthermore, the selected model should be widely used and be supported with continuous development and improvement from the original developers and users groups (USEPA, 1992). For a modeler in a developing country, a widely used continuously worked on model provides easier application and critical evaluation owing to support provided by developers and user groups. If a model is not widely used, it becomes more difficult to establish credibility and to interpret its results. The various stakeholders involved in a project study must be willing to accept model results. For the characterization of hydrology and NPS pollution in a watershed of a developing country under data stringent conditions a widely used, watershed scale, distributed and continuous model will be most suitable that can comprehensively simulate hydrology, NPS pollution and water quality processes.

Review of relevant literature reveals that Hydrological Simulation Program-FORTRAN (HSPF) fulfills all the criteria listed above. It is a conceptual, distributed, continuous watershed model that can simulate the continuous, dynamic event, or

steady-state behavior of both hydrologic/hydraulic and water quality processes in a watershed, with an integrated linkage of surface, soil, and stream processes (Al-Abed

& Whiteley, 2002; Singh and Woolhiser, 2002; Hayashi et al., 2004; Singh et al..

2005; Xu et al., 2007). This model is one of the most comprehensive, flexible and modular programs of watershed hydrology and water quality available. Because of its modular design and organized development, watershed simulations in HSPF can range from the simple to the complex, and utilize a variety of methods, processes, and functions (Skahill, 2004). HSPF can simulate urban and agricultural land use, surface and subsurface processes, runoff, sediment export, and the fate and transport of nutrients, pesticides, and other water quality constituents (Bicknell et al., 2001).

HSPF is a widely used model with an increasingly large user group available for support regarding model application. The applications since its initial release in 1980 have been worldwide and number in the hundreds (Donigian et al., 1999).

HSPF is supported by USEPA and is incorporated into Better Assessment Science Integrating Point and Non-point Sources (BASINS) as its core watershed model in the form of WinHSPF. BASINS is a multi functional watershed analysis and modeling system developed by USEPA for watershed and water quality based studies (USEPA, 2007). It integrates data acquisition, data preparation, watershed characterization, application of models, interpretation model results, and development of maps and tables in the form of its data processing tools WDMUtil and GenScn, core model in the form of WinHSPF and Map Window platform for delineation and characterization of watershed using its Geographic Information Systems (GIS) capabilities (Tong and Chen, 2002). HSPF has been used successfully in modeling the stream hydrology and loadings of sediment, nutrients, and pesticides from agricultural lands (Diaz-Ramirez et al., 2008; Hayashi et al., 2004; Laroche et al., 1996; Munson, 1998; Shirinian-Orlando et al., 2007). The selection of HSPF is further justified by the availability of an expert system for calibration of HSPF, HSPEXP (Lumb et al., 1994) and a database of HSPF parameters from past calibration studies HSPF-Parm (Donigian et al., 1999) developed by USEPA to assist watershed modelers in calibrating HSPF. Considering limited resources in developing countries for collection of data for estimation of model parameters the availability of these tools is a major advantage for the usage of BASINS/HSPF for solving watershed scale environmental problems.