Utilization of Spatially Precise Measurements to Autocalibrate the EPIC Agroecosystem Model
Corn nitrogen recommendations for individual fields must improve to minimize the negative influence that agriculture has on the environment and society. Two adaptive N management approaches for making in-season N fertilizer recommendations are remote sensing and crop systems modeling. Remote sensing has the advantage of characterizing the spatial variability at a high spatial resolution, and crop models are prognostic and can assess expected additions and losses that are not yet reflected by the plant (e.g., due to recent management, weather, etc.). Remote sensing can be used to estimate crop biophysical parameters such as leaf area index or biomass, and can be used to calibrate crop systems models for making more accurate N fertilizer recommendations. A challenge in implementing this, however, is that an independent model calibration is required for each spatial area to be modeled. This study aims to test an autocalibration method at the sub-field scale for use in calibration of the EPIC (Environmental Policy Integrated Climate) model so it can be used more reliably for precision agriculture applications. EPIC is capable of simulating crop growth, nutrient transport and demand, and water movement on a daily time step. Following an initial calibration, an independent calibration is performed for each area within the study area using spatially precise outputs derived from remote sensing data (i.e., leaf area index and biomass). A field experiment was conducted in 2017 with four nitrogen rates and two timings of N fertilizer (i.e., preplant and V5 leaf stage). Tissue N and aerial imagery were collected at several days during the early growth stages to use as a basis for implementing and testing the autocalibration approach. The Monte Carlo algorithm was used to generate samples for the autocalibration step, and the Nash-Sutcliff efficiency was used as an objective function to analyze and interpret the results. The techniques utilized in this study serve as a framework for being able to use crop systems models for making more accurate N fertilizer recommendations.