Wheat yield forecasting using statistical models
Loading...
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Punjab Agricultural University, Ludhiana
Abstract
The present study had been conducted to develop models based on weather variables for
forecasting the wheat productivity of Amritsar, Ludhiana and Patiala districts of Punjab. The
forty years data: (1970-71 to 2009-10) on wheat productivity and weather variables were used
for model development and seven years data (2010-11 to 2016-17) for validation. The linear
and non-linear models; simple linear, quadratic, cubic, fourth degree polynomial,
monomolecular, logistic and gompertz were developed to remove the effect of technological
factors over time. On the basis of goodness of fit statistics, the logistic model came best from
linear and non-linear models. The detrended wheat productivity obtained after fitting the
logistic model was used for forecasting the wheat productivity on the basis of weather
variables; maximum temperature, minimum temperature, rainfall, relative humidity morning,
relative humidity evening and bright sunshine hours. The fourteen weeks weather data of
vegetative period of wheat crop had been utilized to forecast the productivity. Three weather
indices based models; model I (un-weighted), model II (weighted) and model III (combined)
were developed for each district using weather indices and detrended wheat productivity. The
stepwise regression technique was applied and the results revealed that weighted model
(model II) declared as best model for Amritsar, Ludhiana and Patiala districts explaining
60%, 67% and 52% variation in the detrended wheat productivity and had RMSPE 8.57%,
6.93% and 6.20% respectively. The weighted interaction of maximum and minimum
temperature played crucial role in wheat productivity. The selected models followed the
assumptions of residuals.
Description
Keywords
null