Evaluating the APEX model for simulating streamflow and water quality on ten agricultural watersheds in the U.S.

Authors: VAN LIEW, MICHAEL - University Of Nebraska; WORTMANN, CHARLES - University Of Nebraska; Moriasi, Daniel; King, Kevin; Flanagan, Dennis; Veith, Tameria - Tamie; McCarty, Gregory; Bosch, David - Dave; Tomer, Mark

Submitted to: Transactions of the ASABE
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/26/2016
Publication Date: 2/28/2017
Publication URL: https://handle.nal.usda.gov/10113/6558888
Citation: Van Liew, M.N., Wortmann, C.S., Moriasi, D.N., King, K.W., Flanagan, D.C., Veith, T.L., McCarty, G.W., Bosch, D.D., Tomer, M.D. 2017. Evaluating the APEX model for simulating streamflow and water quality on ten agricultural watersheds in the U.S. Transactions of the ASABE. 60(1):123-146. https://doi.org/10.13031/trans.11903.
DOI: https://doi.org/10.13031/trans.11903

Interpretive Summary: Simulation models are increasingly used to assess water quality constituent losses from agricultural systems. Mis-use often gives irrelevant or erroneous answers. The Agricultural Policy Environmental Extender (APEX) model is one of the premier models used for fields, farms, and agricultural watersheds. In this study, the performance of APEX was evaluated in seven USDA ARS watersheds covering variable climatic, topographic, soils, and cover and land management conditions, to better understand its strengths and limitations in simulating streamflow and water quality. The model simulations for each watershed were run using: 1) a “best professional judgment” (BPJ) parameter approach that required no calibration, and 2) calibrated parameters. Based on a preset statistical model performance criteria, parameters determined using the BJP approach satisfactorily simulated 0%, 0%, 15%, and 9% monthly streamflow, sediment, nitrogen (N), and phosphorus (P) data sets, respectively. However, parameters obtained by calibration satisfactorily simulated 73%, 43%, 38%, and 27% of monthly streamflow, sediment, N, and P, respectively. Results of this study suggest that when fully calibrated, APEX performs very well in estimating streamflow, moderately well in estimating sediment and nitrogen, and marginally well in estimating phosphorus for watersheds at the monthly time scale. However, APEX lacks an element of robustness in simulating streamflow and the water quality constituents when applied to the same watershed with a different period of record or to an adjacent or nearby watershed with a similar period of record. Therefore, it is recommended that users exercise caution when employing APEX in validation mode.

Technical Abstract: Simulation models are increasingly used to assess water quality constituent losses from agricultural systems. Mis-use often gives irrelevant or erroneous answers. The Agricultural Policy Environmental Extender (APEX) model is emerging as one of the premier modeling tools for fields, farms, and agricultural watersheds. APEX has been used by federal and state agencies to assess policy and quantify nutrient losses from a range of cropping and management systems throughout the United States. Additionally, APEX is the foundation of user-friendly tools developed to help extension personnel and stakeholders assess sediment, nitrogen (N), and phosphorus (P) losses from agricultural fields or pastures. To better understand APEX’s strengths and limitations in simulating streamflow and water quality constituents, we conducted a study that evaluated the model’s performance under a range of climatic, topographic, soils, and cover and land management conditions. Hydrologic and water quality responses were simulated on seven USDA Agricultural Research Service watersheds that included Mahantango Creek in Pennsylvania, Little River in Georgia, German Creek in Maryland, Upper Big Walnut Creek in Ohio, Matson Ditch in Indiana, Lake Creek in Oklahoma, and Treynor W1 in Iowa. Most of these watersheds consisted of heterogeneous forested, pasture/range, and corn-soybean cropping systems. Simulations were performed on a total of eight calibration and seven validation data sets that were obtained from available records of climatic, streamflow, and water quality data from the watersheds. Model simulations for each watershed were initially run using a “best professional judgment” (BPJ) parameter approach that required no calibration. A total of 16 parameters from the parameter file and 2 variables from the control file that govern streamflow, crop yield, sediment, nitrogen, and phosphorus were selected for calibration on the test watersheds. A combination of both manual and automated approaches was taken to calibrate each of the eight calibration data sets; calibrated parameters and variables were then extended to the appropriate validation data sets. Based on percent bias and Nash Sutcliffe coefficient of efficiency test statistics employed at the monthly time scale, 0%, 0%, 15%, and 9% of the data sets were considered satisfactory or better in simulating streamflow, sediment, nitrogen, and phosphorus, respectively, in the BPJ mode. When fully calibrated, 73%, 43%, 38%, and 27% of the calibration/validation data sets were considered satisfactory or better in the monthly simulation of streamflow, sediment, N, and P, respectively. Findings from this study indicate that for watershed applications, APEX does not perform well in estimating streamflow, sediment or nutrients when using the BPJ approach for model simulations at the monthly time step. It is therefore recommended that use of the model in the BPJ mode for developing watershed scale water quality assessments be exercised on a limited basis. Results of this study suggest that when fully calibrated, APEX performs very well in estimating streamflow, moderately well in estimating sediment and nitrogen, and marginally well in estimating phosphorus for watersheds at the monthly time scale. When used for model validation, however, APEX lacks an element of robustness in simulating streamflow and the water quality constituents when applied to the same watershed with a different period of record or to an adjacent or nearby watershed with a similar period of record. Based on these findings, it is recommended that users exercise caution when employing APEX in validation mode.