In this study, we formulate a stochastic dynamic framework for pest control over the growing season taking into account forecasts of weather conditions and pest infestation expectations. Using stochastic envelope theorem and stochastic comparative dynamics, we analytically show how the stochastic correlation between the prediction errors should affect optimal pesticide usage path. As a case study, we apply the analytical results of the paper for pesticide use in the Palouse region of Washington where pea aphid is the primary threat for lentil production. By stochastic dynamic programming, our simulation shows the optimal dimethoate usage path, which illustrates our findings in the analytical part.