Code modernization and modularization of APEX and SWAT watershed simulation models

SWAT(Soil and Water Assessment Tool)and APEX(Agricultural Policy/Environmental eXtender)are respectively large and small watershed simulation models derived from EPIC(Environmental Policy Integrated Climate),a field-scale agroecology simulation model.All three models are coded in Fortran and have evolved over several decades.They are widely used to analyze anthropogenic influences on soil and water quality and quantity.Much of the original Fortran code has been retained even though Fortran has been through several cycles of development.Fortran now provides functionality originally restricted to languages like C,designed to communicate directly with the operating system and hardware.One can now use an object-oriented style of programming in Fortran,including inheritance,run-time polymorphism and overloading.In order to enhance their utility in research and policy-making,the models are undergoing a major revision to use some of the new Fortran features.With these new programming paradigms the developers of SWAT,APEX,and EPIC are working to make communication between the two models seamless.This paper describes the ongoing revision of these models that will make them easier to use,maintain,modify and document.It is intended that they will converge as they continue to evolution,while maintaining their distinctive features,capabilities and identities.

Evaluation of alternative surface runoff accounting procedures using SWAT model

For surface runoff estimation in the Soil and Water Assessment Tool(SWAT)model,the curve number(CN)procedure is commonly adopted to calculate surface runoff by dynamically updating CN values based on antecedent soil moisture condition(SCSI)in field.From SWAT2005 and onward,an alternative approach has become available to apply the CN method by relating the runoff potential to daily evapotranspiration(SCSII).While improved runoff prediction with SCSII has been reported in several case studies,few investigations have been made on its influence to water quality output or on the model uncertainty associated with the SCSII method.The objectives of the research were:(1)to quantify the improvements in hydrologic and water quality predictions obtained through different surface runoff estimation techniques;and(2)to examine how model uncertainty is affected by combining different surface runoff estimation techniques within SWAT using Bayesian model averaging(BMA).Applications of BMA provide an alternative approach to investigate the nature of structural uncertainty associated with both CN methods.Results showed that SCSII and BMA associated approaches exhibit improved performance in both discharge and total NO3 predictions compared to SCSI.In addition,the application of BMA has a positive effect on finding well performed solutions in the multi-dimensional parameter space,but the predictive uncertainty is not evidently reduced or enhanced.Therefore,we recommend additional future SWAT calibration/validation research with an emphasis on the impact of SCSII on the prediction of other pollutants.