Presence of natural enemies has a considerable impact on pest severity in a given geo-location. However, manually estimating pest severity or population of natural enemies is cumbersome, inaccurate and not scalable. Systems and methods of the present disclosure enable estimating effective pest severity index by receiving a first set of inputs pertaining to weather associated with a geo-location under consideration; receiving a second set of inputs pertaining to agronomic information; generating a pest forecasting model and a natural enemies forecasting model based on the received first set and the second set of inputs for each pest; and estimating the effective pest severity index based on the generated models. The timing and quantity of pesticide application can be optimized based on the estimated pest severity index. The generated models can be further enhanced continually based on one or more of historical data, participatory sensing inputs, crowdsourcing inputs and management practices.