Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Sensor data driven parameter estimation for Agricultural Model using Coordinate Descent

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  2018 ASABE Annual International Meeting  1801400.(doi:10.13031/aim.201801400)
Authors:   Anupam Bhar, Zhiming Qi, Robert W Malone, Ratnesh Kumar
Keywords:   Calibration, Coordinate descent, Optimization, RZWQM.

Abstract. In this study, Root Zone Water Quality Model (RZWQM) was calibrated against measured field data for full irrigation in terms of yield, plant height, leaf area index (LAI), evapotranspiration (ET), and soil water contents (SWC) of seven soil layers, which was experimentally collected in May-Oct 2010 in Limited Irrigation Research Farm near Greeley, Colorado. The crop grown was Dekalb brand 52-59 maize. A total of 77 soil hydraulic parameters, with eleven parameters for each of the seven soil layers and three crop parameters were calibrated using the measured data. In a previous work, calibration of RZWQM‘s hydraulic parameters was done manually using the same measurement data. Tuning a large number of parameters for a nonlinear complex system is hard because of the vastness of solution space. In this study, calibration was done automatically using a so called Coordinate-Descent (CD) optimization algorithm which successively finds the parameter values, one at a time, so that the goodness of fit between the field-measured vs. the model-predicted output values is maximized. Coefficient of determination (R2) and Model Efficiency (ME) was used to evaluate the goodness of fit. Average ME for manual and CD calibration was 0.67 and 0.74 (more than 10% improvement) respectively. Average R2 for manual and CD calibration was 0.77 and 0.78 respectively. The CD calibrated model was validated against data from adjacent field which was deficit irrigated. In this case, Average R2 for manual and CD calibration was 0.77 and 0.78 respectively. Average ME values were slightly better for CD..

(Download PDF)    (Export to EndNotes)