Multisite Evaluation of APEX for Water Quality: I. Best Professional Judgment Parameterization

Abstract

The Agricultural Policy Environmental eXtender (APEX) model is capable of estimating edge-of-field water, nutrient, and sediment transport and is used to assess the environmental impacts of management practices. The current practice is to fully calibrate the model for each site simulation, a task that requires resources and data not always available. The objective of this study was to compare model performance for flow, sediment, and phosphorus transport under two parameterization schemes: a best professional judgment (BPJ) parameterization based on readily available data and a fully calibrated parameterization based on site-specific soil, weather, event flow, and water quality data. The analysis was conducted using 12 datasets at four locations representing poorly drained soils and row-crop production under different tillage systems. Model performance was based on the Nash-Sutcliffe efficiency (NSE), the coefficient of determination (r 2 ) and the regression slope between simulated and measured annualized loads across all site years. Although the BPJ model performance for flow was acceptable (NSE = 0.7) at the annual time step, calibration improved it (NSE = 0.9). Acceptable simulation of sediment and total phosphorus transport (NSE = 0.5 and 0.9, respectively) was obtained only after full calibration at each site. Given the unacceptable performance of the BPJ approach, uncalibrated use of APEX for planning or management purposes may be misleading. Model calibration with water quality data prior to using APEX for simulating sediment and total phosphorus loss is essential. (P) and sediment loss from agricultural fields continues to degrade fresh water quality despite decades of efforts to understand loss processes and implement management practices to reduce nonpoint pollution (e.g., Sharpley et al., 1994Sharpley et al., , 2015Sims and Kleinman, 2005;Jarvie et al., 2013). Failure of conservation practices to produce expected improvements in water quality has renewed appreciation for the complexity of P movement within landscapes and to waterbodies Sharpley et al., 2013). Processbased watershed-and field-scale models offer the prospect of integrating knowledge to assess and quantify impacts of conservation and management practices.The P Index was developed in the mid-1990s to assess risk of P loss from agricultural land (Lemunyon and Gilbert, 1993) and is now an integral part of the NRCS 590 Nutrient Management Standard and other state and federal programs (Sharpley et al., 2003(Sharpley et al., , 2017. However, the diversity in P Index ratings and P management recommendations for similar conditions led and Osmond et al. (2006) to emphasize the need for science-based assessment and improvement of existing P Indices.Extensive testing of P Indices requires water quality data from field-scale watersheds from a broad range of soils and management scenarios with a sufficient number of monitoring years to estimate long-term average annual losses. Such extensive datasets are rare. Hence, computer models must ...

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

“…Development of the RCM ensured that these sensitive parameters were adjusted within an appropriate range for restricted‐layer soils. For example, the value of P42 selected based on best professional judgement was 1.0 (Baffaut et al, 2017), which is less than the value of all SSCMs, whereas the value of P42 was 2.3 for the RCM, which was well within the range of values selected during calibration of the SSCMs.…”

“…The RCM was developed and tested on datasets that included four locations and management practices that spanned a wide range of tillage, conservation practices, and fertilizer management strategies in grain production systems (Supplemental Table S1). The RCM performed significantly better than the uncalibrated BPJ parameterization developed for and tested on these same datasets (Baffaut et al, 2017). The RCM parameterization successfully simulated runoff for more datasets (11 vs. 8) than the BPJ.…”

“…Therefore, it is well suited for simulation of nutrient and sediment losses at the field to small watershed scale (Wang et al, 2012). This study used the APEX 0806 version with modifications as described by Baffaut et al (2017). The APEX code and executable used to obtain the results presented herein are available on request from the corresponding author.…”

“…These are the same five models used by Baffaut et al (2017) to evaluate an uncalibrated best professional judgement (BPJ) model parameterization. The general watershed characteristics, data collection, and methods of calibration and validation are described in papers within this special issue (Baffaut et al, 2017; Bhandari et al, 2017) and in previously published works (Udawatta et al, 2002; Zeimen et al, 2006; Senaviratne et al, 2012, 2013, 2014; Sweeney et al, 2012). Calibration of each SSCM started with the BPJ control file options and parameters and resulted in five SSCMs that differed from the BPJ parameterization for three options in the control file (Supplemental Table S2) and 21 values in the parameter file (Supplemental Table S3).…”

Calibration of the APEX Model to Simulate Management Practice Effects on Runoff, Sediment, and Phosphorus Loss

“…However, models must be tested (calibrated and validated) to ensure they are capable of simulating loss accurately over a variety of management practices. Model calibration decreases margins of error and minimizes uncertainties related to model parameters (Wang et al, 2009; Winchell et al, 2011; Baffaut et al, 2016). …”