N rate and transport under variable cropping history and fertilizer rate on loamy sand and clay loam soils: II. Performance of LEACHMN using different calibration scenarios

Testing of existing agronomic models is needed to ensure their validity and applicability to different soils, cropping systems and environments. Data collected from a 3-year field experiment of maize (zea mays L.) on a loamy sand and a clay loam soil were used to validate the research version of the LEACHMN model for water flow and N fate and transport. Three calibration scenarios with increasing levels of generalization for transformation rate coefficients were used based on: (i) each year, treatment and soil type (ii) 3-year average values for each treatment and soil type, and (iii) average over years and soil types. Model accuracy was tested using both graphical and statistical methods including 1:1 scale plot, root mean square error and normalized root mean square error, and correlation coefficient values. The model accurately predicted drainage water flow rate and volume under both sites. Calibrated N transformation rate constants for each treatment, year and soil type provided satisfactory predictions of growing season cumulative NO₃–N leaching losses, and accurate predictions of growing season cumulative maize N uptake at both sites. The use of 3-year average rate constant values for each site resulted in fairly satisfactory predictions of NO₃–N leaching losses on the clay loam site, but inaccurate predictions on the loamy sand site. The model provided accurate predictions of cumulative maize N uptake for both sites. Using the rate constant values averaged over years and soil types resulted mostly in inaccurate predictions. Use of year and soil type-specific N rate coefficients results in accurate LEACHMN predictions of N leaching and maize N uptake. When rate coefficients are generalized over years for each soil type, satisfactory model predictions may be expected when N dynamics are not strongly affected by yearly variations in organic N inputs.