Abstract
Precision agriculture harnesses data-driven techniques to optimize crop production, resource use, and sustainability. However, low-income countries like Bangladesh face a shortage of localized, high-quality datasets that reflect regional agroclimatic conditions and cropping practices. To address this gap, we present SPAS-Dataset-BD, a robust dataset compiled through a hybrid approach: secondary extraction from the Bangladesh Bureau of Statistics (BBS) 2022 Yearbook and primary on-field surveys of 223 farmers across ten diverse districts. The dataset comprises 4191 records over 73 crop types, with 12 agronomic and environmental features, including underrepresented species. Robustness is demonstrated via threshold-based missing-value handling (<5 % deletion, targeted imputation), hash-based deduplication, and cross-validation against official statistics. We illustrate potential applications, in machine learning (73-class crop classification, yield forecasting) and IoT-driven irrigation scheduling. SPAS-Dataset-BD's scale, methodological transparency, and contextual richness make it a valuable resource for precision agriculture research and policy-making in South Asia.