Azizur Rahman, Prachi Rini Saha, Kyra Hudson / Earth & Environmental Sciences / Faculty Mentor: Yike Shen

Understanding regional variations in pesticide application is essential for developing data-driven environmental policies and optimizing agricultural sustainability. This study employs unsupervised machine learning algorithms to assess pesticide usage across land cover types in France and Poland at the NUTS3 regional level. Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP) are applied for dimensionality reduction, enabling the visualization of complex, high-dimensional pesticide application data. Additionally, K-means clustering is utilized to cluster regions based on pesticide usage intensity. Results show UMAP excels in grouping local structures, while PCA preserves global data relationships for better interpretability.  Judging from pesticide emission patterns colored by different land covers, Arable Land and Grassland were identified as the primary contributors to pesticide variability. A heatmap highlights Poland as the highest pesticide consumer, while France demonstrates more distributed usage across land types, reflecting differences in agricultural practices and regulatory policies. This study utilizes the EU-wide pesticide dataset (Udias et al., 2023) to present a scalable analytical framework for assessing pesticide application patterns across diverse agricultural landscapes, offering insights for precision agriculture, policy-making, and sustainable pesticide management strategies aimed at mitigating environmental and human health risks.
Keywords: Precision Agroecology, Environmental Policy Optimization, Data-Driven Agricultural Analytics.


Poster

Video Presentation