Predicting Vertical Profiles of Aerial Insect Population from Entomological Radar Utilizing the GRU Encoder-Decoder Model

Monitoring and predicting aerial insect population (AP) are crucial for effective pest control strategies. Although traditional entomological radars have contributed to monitoring insect migration, their low resolution limits detailed observations, making accurate predictions of AP challenging. The new generation of high-resolution phased array radars, with their superior range resolution and rapid data update rates, can precisely monitor and quantify AP, capturing the dynamic changes in populations as they respond to meteorological factors over time and space. These dynamics necessitate more accurate predictions of AP. In this study, we introduce deep learning techniques for predicting AP by proposing a Gated Recurrent Unit (GRU)-based encoder-decoder model designed for shortterm predictions of AP vertical profiles. Our approach integrates historical AP data and meteorological factors, where the encoder captures past states, and the decoder combines this information with the current meteorological conditions to predict AP. Experimental results demonstrate that the proposed method effectively predicts vertical profiles of AP.