Advances and Applications in Statistics
Volume 28, Issue 1, Pages 1 - 22
(May 2012)
|
|
MULTI-LEVEL MIXED MODELLING FOR WEATHER-CROP-YIELD RELATIONSHIPS ON AGRO-CLIMATIC ZONE BASIS IN HARYANA (NORTHERN INDIA)
U. Verma, H. P. Piepho, J. O. Ogutu, M. H. Kalubarme and M. Goyal
|
Abstract: Forecasting of crop production is one of the most important applications of statistics in agriculture. Such predictions before harvest are needed by the national and state governments for various policy decisions relating to storage, distribution, pricing, marketing, import-export, etc. In this paper, therefore, a methodology for the estimation of wheat yield, ahead of harvest time, is developed specifically for wheat growing districts in Haryana (India). The Haryana state, having a total geographical area of 44212 sq. km, was divided into four zones for pre-harvest crop yield forecasts. Zonal yield models using time trend and agro-meteorological predictors were generated using regression based analyses and mixed model procedures. The district level yield forecasts, the percent deviations from the real-time data and root mean square error(s) at zonal level show a preference for using mixed models in almost all the districts of the state. The common weather-based approach to yield forecast is linear regression with constant coefficients over time. This may be restrictive and of limited prediction power since it does not account for the year-to-year dependence in the yield variable. The key idea followed in this paper is to borrow strength across districts by fitting a multivariate model to the time series vector which has elements equal to the district-wise yields and weather parameters. A mixed model procedure provides a flexible way to fit a multi-level model for crop yield prediction. Alternatively, the more classical multivariate techniques viz., principal component analysis and discriminant analysis have been used to develop the zonal yield models. The purpose of this article is to illustrate the usefulness of the mixed model framework for pre-harvest crop yield forecasting and to provide some empirical evidence for the superiority of mixed models over the classical regression-based analyses. |
Keywords and phrases: eigenvector, principal component score, discriminant score, linear mixed model, agromet trend yield model, percent deviation, root mean square error. |
|
Number of Downloads: 351 | Number of Views: 838 |
|