Weather Integrated Forecasting of Helicoverpa armigera using ARIMA and ANN Models Trained on a Decade of Pheromone Trap Data

<title>Abstract</title> <p> <italic>Helicoverpa armigera</italic> is a major polyphagous pest causing extensive damage across diverse cropping systems. Accurate longterm forecasting of adult moth activity is essential for timely interventions and improved integrated pest management (IPM). Time-series forecasting models were developed to predict the population dynamics of <italic>Helicoverpa armigera</italic> adults using ten years of weekly pheromone trap data and associated weather variables. Two approaches Autoregressive Integrated Moving Average (ARIMA) and Artificial Neural Network (ANN) were evaluated for their ability to capture temporal patterns and forecast moth catches. The ANN model (7-26-1 architecture) effectively represented nonlinear fluctuations and peak incidences, providing higher short-term predictive accuracy compared with the ARIMA (1, 0, 1) (1, 0, 2) model, which better characterized long-term seasonal trends. Model performance metrics (MAPE, RMSE, MAE, MASE, and R²) indicated superior accuracy for the ANN model (RMSE = 3.63, R² = 0.64) compared to ARIMA (RMSE = 6.62, R² = 0.76). While ARIMA achieved a marginally higher R², the ANN model yielded lower RMSE and MAE, indicating better predictive accuracy for short-term forecasting. Overall, ANN consistently generated more accurate predictions of adult moth activity. ANN based forecasting outperformed ARIMA for short term prediction of <italic>H. armigera</italic> adult populations and more effectively captured nonlinear population dynamics. These results highlight the value of ANN models as practical early warning systems for pest monitoring and decision support within IPM programs, offering improved precision for timely and targeted management interventions. </p>