CORN YIELD PREDICTION WITH A DAILY ENERGY-CROP GROWTH VARIABLE FOR COUNTIES IN INDIANA (METEOROLOGY)
Abstract
Accurate and early assessment of weather and climate impact on crop yields would allow for timely agricultural extension advisories to help reduce farm management costs and improve marketing decisions. A method to predict county corn (Zea mays L.) yields with weather information was developed with data from four counties and tested on 15 other counties in Indiana. Daily average maximum and minimum temperatures, precipitation, class A pan evaporation, and solar radiation were calculated for each county from the nearest 3 to 5 weather stations, using a reciprocal distance weighting method after setting appropriate station data to the proper calendar day. The county weather data were used to calculate a daily energy-crop growth (ECG) variable, which is a product of the solar radiation intercepted by the corn leaf area (FLAI*SR/L), a corn growth temperature function (FT), and a moisture stress ratio, actual to potential evapotranspiration (ET/PET). The daily ECG was summed for critical corn growth periods and used with an applied nitrogen (N) fertilizer * ECG interaction variable to predict county yields. The N variable was used as a marker for all technology in the period of study, 1960-1984. The ECG variable also was modified by weighting its factors with individual exponents. Estimates of these exponents were obtained numerically for Tippecanoe and Rush County data with a least squares error criterion, and showed the ET/PET factor to be dominant over the FLAI*SR/L and FT factors in reducing observed corn yield variability. Best results in yield prediction with the numerically estimated weights were found for a 36-day summation period, 16 days before to 20 days after silking. Regression models were developed with 36- and 90-day sums of the weighted and unweighted ECG to estimate corn yields for a four-county pooled data set, and these models were tested on data from 14 counties in Indiana Crop Reporting District 5 not used to estimate the regression coefficients. Excellent overall results were obtained, with r$\sp2$ values of predicted vs. observed average corn yields in individual counties in the range 0.64-0.85 and root mean square errors 0.49-0.96 ton/ha.
Degree
Ph.D.
Subject Area
Atmosphere|Plant propagation
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.