Data on aphid (Lipaphis erysimi) counts available in Annual Reports of the All India Coordinated Research Project on Rapeseed-Mustard for multiple years and of field trials conducted for different locations were utilized for devising the forecast models. Weather data for the respective periods of the locations were also collected from the records of meteorological observatories of the centres.

Correlations of timing (days after sowing or d.a.s.) of attack of L. erysimi, peak number of aphids on the crop and crop age (d.a.s.) at peak aphid population with weather variables were determined. Linear prediction models based on the weather parameters as independent variables and crop age (d.a.s.) at time of first appearance of aphid on crop, peak number of aphids on the crop in the season and crop age at peak population of the pest at each week starting from week of sowing as dependent variables were fitted by multiple stepwise regression. Multiple regression is a powerful statistical technique that is most widely used by forecasters. The method is typically suited for the data available for the mustard aphid to estimate the average relationship between a dependent variable and two or more independent variables with regular data patterns. The stepwise regression took care of autocorrelation. Based on correlation coefficients between dependent variables under study with the respective weather parameter in different weeks, a composite weather variable was developed as the weighted sum of the weather variable in different weeks starting from pre-sowing week up to week of the prediction. Similarly, interaction terms were developed as weighted sums of product between two weather variables, weightings being correlation coefficients of dependent variable under study with products of weather variables in respective weeks. The important weather indices were selected through stepwise regression. Models were fitted for prediction of the dependent variables viz., highest aphid population or crop age at peak aphid population or crop age at time of first appearance of aphid on weekly basis starting from the time of sowing, 2nd week after sowing and so on. Weather indices based on summation of weightings of different meteorological factors as per correlation coefficients in different weeks after sowing until the forecast was provided, were taken into account.

Using SAS statistical software, models for forecasting Crop age at first appearance of aphid on crop, Crop age at highest aphid population (aphid counts on 10 cm of terminal shoot) in the season and Highest aphid population on the crop in the season for the locations were devised.