Skip to main content
ARS Home » Southeast Area » Oxford, Mississippi » National Sedimentation Laboratory » Watershed Physical Processes Research » Research » Publications at this Location » Publication #304673
Title: Combined PEST and Trial-Error approach to improve APEX calibration

Author
item MBONIMPA, ERIC - South Dakota State University
item GAUTAM, SAGAR - South Dakota State University
item KUMAR, SANDEEP - South Dakota State University
item LAI, LAI - South Dakota State University
item Bonta, James - Jim
item WANG, XIUYING - Blackland Research And Extension Center
item RAFIQUE, RASHID - Pacific Northwest National Laboratory

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/29/2015
Publication Date: 6/1/2015
Citation: Mbonimpa, E., Gautam, S., Kumar, S., Lai, L., Bonta, J.V., Wang, X., Rafique, R. 2015. Combined PEST and Trial-Error approach to improve APEX calibration. Computers and Electronics in Agriculture. 114:296-303. doi:10.1016/j.compag.2015.04.014.

Interpretive Summary: The Agricultural Policy Environmental eXtender (APEX), a physically-based hydrologic model simulating management impacts on the environment for small watersheds, requires an improved method for determining the input parameters to give the good simulations. This study analyzed the use of the automated program, Parameter Estimation (PEST), to calibrate APEX- simulated daily runoff from a small agricultural watershed managed with continuous corn and corn-soybean-rye cropping systems for the whole year (Jan-Dec) and for the growing season (Apr-Nov). These results were compared with those from the often-used trial and error method. The PEST-APEX approach improved runoff simulation for both cropping systems compared to the trial and error procedure. The PEST tool is also computationally efficient; it can calibrate the APEX runoff in minutes or hours compared with days or weeks for the trial and error method. The calibration performance was better for the growing season compared to the whole year for both cropping systems. This was likely due to the difficulty in accurately measuring precipitation and flow in winter and modeling complex watershed processes in winter. Overall, The PEST-APEX approach was the superior method compared to the trial and error method. Furthermore, PEST determined all APEX input parameters to which the runoff was sensitive, including those that indirectly impact runoff, whereas, the trial and error method only considered parameters that directly impact runoff. Scientists and conservation practitioners can use the results of this study to reduce the time required for calibration of parameters in watershed studies.

Technical Abstract: The Agricultural Policy Environmental eXtender (APEX), a physically-based hydrologic model that simulates management impacts on the environment for small watersheds, requires improved understanding of the input parameters for improved simulations. However, most previously published studies used the trial and error method for calibrating the APEX model. This method is tedious, time consuming, and required experience with the model. This study presented the application of Parameter Estimation (PEST), a parameter estimation and optimization tool based on inverse modeling approach, to calibrate and validate APEX simulated runoff for an agricultural land seeded with continuous corn (Zea mays L.; 2000-2005) and corn-soybean (Glycine max L.)-rye (Secale Cereale L.) (CSR; 2006-2011) systems. PEST and trial and error methods were compared for the CSR and CC systems for growing (April-November) and whole year (January-December) periods. The results indicated that PEST calibration had higher R2 (0.82) and NSE (0.82) than the model calibrated using trial and error (R2 = 0.65 and NSE = 0.57) for the CSR (April-November), and also had higher R2 (0.70) and NSE (0.69) than trial and error (0.46 and 0.45) for the CC system. The percent bias (PBIAS), root mean square error (RMSE), and the slope of the linear regression between observed and measured runoff indicated that APEX coupled with PEST calibration performed better compared with that of APEX alone. PEST-calibrated APEX model performance for the whole year (January-December) was lower compared to that of the growing period, but was still better than in trial and error method. This was likely due to errors in winter weather data and because the model better captured the runoff trends during the growing season compared with the winter season when runoff is also caused by snow melt. It can be concluded from this study that APEX calibration using PEST was more robust and efficient particularly because it is automated, and utilizes the entire model parameter space to get the best-fit parameters. Furthermore, it determined all APEX input parameters to which the runoff was sensitive, including those that indirectly impact runoff, whereas, the trial and error method only considered parameters that directly impact runoff.