Role of crop condition based dummy regressor alongwith weather parameters for pre-harvest yield prediction of cotton crop in Western Agro-climatic zone of Haryana
Loading...
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
CCSHAU
Abstract
Crop yield models are abstract presentation of interaction of crop with its environment and can range from
simple correlation of yield with a finite number of variables to the complex statistical models with predictive end.
The pre-harvest forecasts are useful to farmers to decide in advance their future prospects and course of action. An
efficient crop predicting infrastructure is pre-requisite for information system about food supply, especially
export–import policies, procurement and price-fixation. Multiple Linear regression was used to develop zonal
yield models for obtaining cotton yield prediction in Hisar, Bhiwani, Sirsa and Fatehabad districts of Haryana.
Linear time-trend has been obtained using cotton yield data of the period 1980-81 to 2011-12. The fortnightly
weather data along with trend yield have been utilized for the same period for building the zonal weather-yield
models. Models have been validated for subsequent years i.e. 2012-13 to 2016-17, not included in the development
of the models. The zonal models were fitted by taking DOA yield as dependent variable and fortnightly weather
variables along with trend yield/CCT/dummy variables as regressors. The predictive performance(s) of the
contending models were observed in terms of average absolute percent deviations of cotton yield forecasts in
relation to the observed yield(s) and root mean square error(s). The adequacy of the fitted models was examined
through histogram, normal-probability plot for the residuals and residual plot against fitted values for the selected
models.
The yield(s) estimated by zonal weather-yield models had sometimes higher percent deviations from the
real-time yield(s) i.e. too high than considered to be tolerable for reliable yield prediction in the districts under
consideration. Consequent upon, an attempt was made to improve the predictive accuracy of the developed models
by adding trend yield based crop condition term to the zonal weather-yield model and that significantly improved
the predictive accuracy of forecast models. The CCT is an indicator variable generated by splitting the DOA crop
yield series into different non-overlapping classes. The level of accuracy achieved by zonal yield model(s) using
CCT as categorical covariate along with weather variables was considered adequate for estimating the district-level
cotton yield(s) at least 4-5 weeks in advance of the crop harvest. The average absolute percent deviations of postsample
period forecasts falling between 4-9 percent favour the use of developed models for cotton yield prediction
in western zone of Haryana. Zonal yield models incorporating CCT and weather variables consistently showed the
satisfactory results pertaining to cotton yield prediction and performed well with lower error metrics as compared
to the remaining models in all time regimes for the district under consideration.
Description
Keywords
null