Hyper-Localized Pest Prediction and Mitigation via Spatiotemporal Deep Learning in Korean Online Farmer Communities

**Abstract:** This paper presents a novel approach to targeted pest management within Korean 온라인 농민 커뮤니티 (online farmer communities) by leveraging spatiotemporal deep learning models trained on community-sourced data (location, crop type, pest reports, localized weather). By integrating historical pest occurrence data with real-time environmental factors and farmer observations, our system provides hyper-localized, predictive pest alerts and recommends optimized mitigation strategies. This system promises a significant reduction in pesticide usage, improved crop yields, and a strengthened resilience of small-scale farming operations.