Policy-Guiding Framework for Sustainable Urban and Agricultural Systems

While the impact of urban expansion on climate is well-studied in existing literature, its effect on crop yields remains relatively unexplored. In developing regions, where both metropolitan growth and agricultural activity are vigorous, the expansion of cities leads to a significant loss of crop yields. Although past investigations have utilized machine learning to predict crop yield loss, the topic has not been thoroughly examined in the context of urban expansion. To address these concerns, this paper introduces URBAN, a novel hybrid deep learning model combining U-Net and Bi-LSTM architectures, providing spatial-temporal predictions of global crop yields from segmented geospatial urban extent maps. These predictions achieved an MAE of 0.3689, an RMSE of 0.6887, and an MSE of 0.4743, indicating high-accuracy analyses of adverse urban expansion. Currently, metropolitan authorities implement various policies to influence city growth, attempting to preserve rural practices. With the predictive capabilities of URBAN, such policy-makers may recognize areas of harmful urban land growth, developing more informed, focused, and effective limitations to sustain fundamental agrarian institutions.