Application of Seasonal Temperature and Rainfall Forecast for Wheat Yield Prediction for Palampur, Himachal Pradesh

Abstract

In this study, performance of seasonal temperature and rainfall forecast were evaluated for forecasting of wheat yield using CERES-wheat model for Himachal Pradesh region. Using the historical monthly total rainfall and mean maximum and minimum temperature values, 1000 daily weather realizations were generated using disaggregation technique. The moving average correlation between observed and generated weather sequences were calculated for different parameters viz., rainfall intensity, rainfall frequency and total monthly rainfall. The correlation values for all these parameters increased with the increase in number of weather realizations and attained its peak near 200 realizations. Thereafter, the correlation became almost constant which concluded that 200 weather realizations were optimum to know the behaviour of predicted weather sequences and its application in crop simulation models. The lead-1 temperature and rainfall forecast for November-April (6 months forecast), December-April (5 months forecast), January-April (4 months forecast), February-April (3 months forecast), March-April (2 months forecast) and monthly forecast of April month (1 month forecast) were generated for 1984-2008 in hindcast mode using multi-model ensemble technique. The monthly and seasonal forecasts were converted into daily weather sequences using the stochastic disaggregation technique. For each generated forecast, observed weather was merged with the respective forecast to make it entire season weather observations so that it can be used in crop models. CERES-wheat model was calibrated and validated for variety HPW-89 for Palampur region and the genetic coefficients thus generated were used to simulate the wheat yield in different years starting from 1984- 2008. It was found that six month forecast i.e. October start NovemberApril did not have significant skill and it failed to capture the variation in wheat yield for different years. However, with the advancement of season thereby reducing the period of forecast, the forecast skill improved progressively. It was also observed that 3 or 4 months seasonal forecast have almost similar results with a slight variation. However, forecast for three months or lesser period was able to simulate the wheat yield in an efficient manner.