Multisite Evaluation of APEX for Water Quality: II. Regional Parameterization

Authors

Nathan O. Nelson, Kansas State UniversityFollow
Claire Baffaut, USDA, Agricultural Research Service
John A. Lory, University of Missouri
G.M.M.M. Anomaa Senaviratne, University of Missouri
Ammar B. Bhandari, Fort Hays State University
Ranjith P. Udawatta, University of Missouri
Daniel W. Sweeney, Kansas State University
Matt J. Helmers, Iowa State UniversityFollow
Mike W. Van Liew, University of Nebraska-LincolnFollow
Antonio P. Mallarino, Iowa State University
Charles S. Wortmann, University of Nebraska-LincolnFollow

Document Type

Article

Date of this Version

2017

Citation

J. Environ. Qual. 46:1349–1356 (2017)

Comments

Copyright © American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America.

This is an open access article

doi:10.2134/jeq2016.07.0254

Abstract

Phosphorus (P) Index assessment requires independent estimates of long-term average annual P loss from fields, representing multiple climatic scenarios, management practices, and landscape positions. Because currently available measured data are insufficient to evaluate P Index performance, calibrated and validated process-based models have been proposed as tools to generate the required data. The objectives of this research were to develop a regional parameterization for the Agricultural Policy Environmental eXtender (APEX) model to estimate edgeof- field runoff, sediment, and P losses in restricted-layer soils of Missouri and Kansas and to assess the performance of this parameterization using monitoring data from multiple sites in this region. Five site-specific calibrated models (SSCM) from within the region were used to develop a regionally calibrated model (RCM), which was further calibrated and validated with measured data. Performance of the RCM was similar to that of the SSCMs for runoff simulation and had Nash–Sutcliffe efficiency (NSE) > 0.72 and absolute percent bias (|PBIAS|) < 18% for both calibration and validation. The RCM could not simulate sediment loss (NSE < 0, |PBIAS| > 90%) and was particularly ineffective at simulating sediment loss from locations with small sediment loads. The RCM had acceptable performance for simulation of total P loss (NSE > 0.74, |PBIAS| < 30%) but underperformed the SSCMs. Total P-loss estimates should be used with caution due to poor simulation of sediment loss. Although we did not attain our goal of a robust regional parameterization of APEX for estimating sediment and total P losses, runoff estimates with the RCM were acceptable for P Index evaluation.