Comparing the effects of inputs for NTT and ArcAPEX interfaces on model outputs and simulation performance

Authors: Nelson, Amanda; Moriasi, Daniel; TALEBIZADEH, MANSOUR - Orise Fellow; Tadesse, Haile; Steiner, Jean; Gowda, Prasanna; Starks, Patrick

Submitted to: Water
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/15/2019
Publication Date: 5/22/2019
Citation: Nelson, A.M., Moriasi, D.N., Talebizadeh, M., Tadesse, H.K., Steiner, J.L., Gowda, P.H., Starks, P.J. 2019. Comparing the effects of inputs for NTT and ArcAPEX interfaces on model outputs and simulation performance. Water. 11:554-580. https://doi.org/10.4236/jwarp.2019.115032.
DOI: https://doi.org/10.4236/jwarp.2019.115032

Interpretive Summary: The Agricultural Policy/Environmental eXtender (APEX) model is a commonly used to quantify the impacts of climate and land use change, land management, and conservation practices on water resources. APEX has five tools called interfaces used to process and build model projects. The different interfaces utilize different input databases that lead to different model input values, which can result in different outputs. In this study, we compared structure and input values associated with the ArcAPEX and Nutrient Tracking Tool (NTT) interfaces. Long-term, water quality data from the Rock Creek watershed, located in northern Ohio was used to determine the impact of the differences on computation time, sensitive model inputs, and the ability of the model to predict streamflow, nitrogen, and phosphorus. There were differences in the soil input structure between the two interfaces, in which NTT assigns three soil files per field, compared to one soil per field assigned by ArcAPEX. As a result, the project built using NTT took three times longer to run than the one built with ArcAPEX. In both cases, there were twelve sensitive model inputs, however the order of sensitivity was different. Both interfaces predicted streamflow well, but ARCAPEX predicted evapotranspiration, nitrogen, and phosphorus better than NTT, while NTT predicted crop yields better than ArcAPEX. However, none of the models met all of the performance criteria for either interface. Therefore, more work is needed to ensure models are well calibrated before they are used to determine impacts of management practices for water resources. Overall, this study shows that different interfaces lead to different model predictions and, therefore, it is critical that model users specify the interface used and any modifications made to the associated databases when reporting model results.

Technical Abstract: The Agricultural Policy/Environmental eXtender (APEX) model has five different interfaces used to process and build simulation projects. These interfaces utilize different input databases that lead to different model default values. These values can result in different hydrologic, crop growth, and nutrient flow model outputs. This study compared structural and input value differences of the ArcAPEX and Nutrient Tracking Tool (NTT) interfaces. Long-term, water quality data from the Rock Creek watershed, located in northern Ohio was used to determine the impact of the differences on computation time, parameter sensitivity, and streamflow, total nitrogen (TN), and total phosphorus (TP) simulation performance. The input structures were the same for both interfaces for all files except soils, where NTT assigns three soil files per field, rather than a single one in ArcAPEX. As a result, computation times were three times as long for NTT as for ArcAPEX. There were twelve sensitive parameters in both cases, but the order of sensitivity was different. Both interfaces simulated streamflow well, but ARCAPEX simulated evapotranspiration, TN, and TP better than NTT, while NTT simulated crop yields better than ArcAPEX. However, none of the models met all of the performance criteria for either interface. Therefore, more work is needed to ensure models are properly calibrated before being used for scenario analysis. Overall, the results indicate that different interfaces lead to different model simulation results and, therefore, the authors recommend users specify the interface used and any modifications made to the associated databases when reporting model results.