Weather Based Prediction Models for Forewarning Tobacco Caterpillar, Spodoptera Litura, Fabricius, Larval Population in Soybean

Abstract

The study is based on the daily village level data of tobacco caterpillar incidence and district level data on weather variables for Maharashtra collected from Crop Pest Surveillance and Advisory Project (CROPSAP). The Project conducted at Department of Agriculture, Govt. of Maharashtra for the period 2010-2015. The study was conducted to assess the influence of weather variables on tobacco Caterpillar (Spodoptera litura) infestation in soybean. The correlations of insect population with weather variables pertaining to current, first and second lag weeks were worked out. Gregarious phase larvae was significantly positively correlated with relative humidity of first lag week (RH-1) and significantly negatively correlated with rainfall of current and first lag weeks (RF0, RF-1 respectively); and Solitary phase was significantly and negatively correlated with relative humidity of current and first lag week (RH0 and RH-1), and rainfall of first and second lag week (RF-1, and RF-2). Forewarning models were developed using training dataset and validated using validation dataset. Mean regression models explained 61.58%, 72.08% and 46.48% variability in the population of S. litura egg mass, gregarious and solitary larva, respectively. The pre-disposing conditions favouring the tobacco caterpillar infestation for maximum temperature, minimum temperature, relative humidity and rainfall were in the range of 27.10-32.83 oC, 19.35-24.15 oC, 84.21-93.38% and 14.77- 92.95 mm with high or medium rainfall in previous weeks followed by low rainfall in current week (for S. litura egg mass); 27.29-31.94oC, 20.28- 25.63 oC, 86.00-93.75% and 11.07-112.65 mm with low rainfall in previous weeks followed by higher rainfall in current week (for S. litura gregarious larva); and 27.06-32.45 oC, 20.23-25.63 oC, 82.96-94.28% and 15.03-119.18 mm with high rainfall in RF-2 and slowed in RF-1 followed by rise in RF0 or continuously increasing pattern from RF-2 to RF0 (for S. litura solitary larva) respectively. The models were validated by cross-validation and independent dataset methodologies. Two sample t-test, RMSE, and other validation statistics revealed no significance difference between observed and predicted values of insect population. Hence, the models could be utilized to disseminate the insect advisories to the farmers.