MSc

The study entitled “Pre-harvest forecasting models and Instability in production of cassava (Manihot esculenta Crantz.)” was conducted at Instructional Farm, College of Agriculture, Vellayani during 2015-2017 with the objectives to develop early forecasting models for yield of five major short duration varieties of cassava and also to carry out trend and instability analysis on area and production of cassava in Kerala.

The study was based on both primary and secondary data. The varieties Sree Jaya, Sree Vijaya, Sree Swarna, Vellayani Hraswa and Kantharipadarppan were grown in Randomized Block Design with three replication in a spacing of 90 cm x 90 cm. Twenty five plants were randomly selected and monthly observations were recorded for all the varieties on biometric parameters. Yield and yield parameters were recorded at harvest. Secondary data on area, production and productivity over a period of twenty five years (1992-2016) were collected from published sources of Directorate of Economics and Statistics, Government of Kerala and State Department of Agriculture.
In order to give an idea about the behavior of the biometric observations and yield of the plants, summary statistics including mean, standard deviation, minimum and maximum were worked out for all variety at each growth stage.
Inter correlations were worked out between growth parameters and yield and the results showed that the number of primary branches, height of branching and number of functional leaves had positive and significant correlation with yield while correlation between yield and yield attributes revealed that number of tubers and average tuber weight were positively correlated with yield.
Multiple linear regression and non linear regression analysis were carried out for all the varieties using yield as dependent variable and biometric observations as independent variables. Stepwise regression was performed and significantly contributing biometric characters were selected using R2, Mallow’s Cp and t-values for predicting the yield.
Among various linear regression equations the best model obtained for the prediction of yield in Sree Jaya was using inter nodal length at 2 and 3 MAP and number of primary branches at 4 and 5 MAP with R2 of 50 per cent and based on non linear equations the best model obtained was using number of functional leaves at 2 MAP, number of primary branches at 4 and 5 MAP and inter nodal length at 3 MAP with R2 of 56 per cent.
Best linear model obtained for the pre-harvest prediction of yield in Sree Vijaya was by using inter nodal length at 2, 4 and 5 MAP, number of functional leaves at 2 MAP and plant height at 5 MAP with R2 of 58 per cent. Non linear model obtained was using inter nodal length at 2, 3 and 5 MAP, number of functional leaves at 3 and 5 MAP with R2 of 59 per cent
Best linear model obtained for prediction of yield in Sree Swarna was using inter nodal length at 2 and 3 MAP and number of functional leaves at 5 MAP with R2 of 43 per cent and with non linear functions the best model obtained was with inter nodal length at 2 and 3 MAP and number of functional leaves at 5 MAP and leaf area index with R2 of 47 per cent.
Best linear model obtained for prediction of yield in Vellayani Hraswa was using number of functional leaves at 2 MAP and plant height at 4 MAP with R2 of 35 per cent and with non linear function the best model obtained was with plant height at 3 and 4 MAP and number of functional leaves at 2 MAP with R2 of 40 per cent.
Best linear model obtained for prediction of yield in Kantharipadarppan was using number of functional leaves at 4 MAP and plant height at 3 MAP with R2 of 34 per cent and with non linear equations the best model obtained was using plant height at 2 MAP and number of functional leaves at 4 MAP with R2 of 33 per cent .
The estimated trends in area, production and productivity of cassava using semilog function revealed that there was a significant decline in area (CAGR= -1.37 %), non significant decline in production (CAGR= -.02 %), and a significant increase in productivity (CAGR= 1.3 %).
Instability in area, production, productivity and nominal price of cassava was also worked out using various measures and the results of the analysis shown that Cuddy-Della Valle index provides best estimates and instability was found to be more in productivity (4.04) followed by area (3.98) and production (.80).
The present study concluded that non- linear model provides better yield prediction model of cassava as compared to linear prediction model on the basis of R2 and Mallow’s Cp. Moreover, the biometric characters such as number of functional leaves and inter nodal length were the most significant predictor variables in all short duration varieties included in this study.