Pre-harvest forecasting of sugarcane yield

Abstract

Several yield prediction models were tried to examine their suitability for the pre-harvest prediction of yield of two varieties of sugarcane namely CO-997 and CO-62175 in different months of plant growth using biometric characters based on the data collected from the Sugarcane Research Station, Thiruvalla. The methods of multiple regression analysis, path coefficient analysis and principal component analysis were used for the above purpose. Multiple regression analysis using plant biometric characters revealed that cane yield could be predicted on the basis of observations on height of the cane, girth of the cane find estimated total leaf area per cane or area of third leaf from the seventh month after planting onwards with an accuracy in the range of 59*5 to 81.9 per cent. The estimated cane yield when multiplied by the number of canes in the plot will give an advance estimate of the plot yield Linear models with five biometric characters viz., height of the cane, girth of the cane, width of the third leaf determined from the selected plants of each plot and number of canes/tillers and number of leaves determined on a whole plot basis were sufficient to predict the plot yield of the crop as early as in the fifth month of plant growth with an accuracy in the range 68 to 90 per cent. Path analysis revealed that height of the cane and girth of the cane were the t wo important characters contributing towards cane yield in all stages of plant growth. Using the forecasting models fitted with principal components as explanatory variables, yield could effectively be predicted with 81.4. per cent accuracy for variety CO-997 and with 76 per cent accuracy for variety 00-62175 in the Sixth month of plant growth.