Evaluation of Heterogeneous Data Fusion for Real-Time Agricultural Stress Monitoring in Small-Scale Farming Ecosystems

Smallholder agricultural systems represent a critical frontier for climate resilience strategies, yet they are often underserved by traditional remote sensing frameworks due to fragmented land parcels and low-frequency reporting. This research investigates an autonomous geospatial framework designed to synthesize multi-source satellite data, specifically Sentinel-2 MSI optical imagery and CHIRPS precipitation datasets, to detect and characterize crop stress at high spatial resolution. The study focuses on the development of a low-latency data pipeline that integrates NDVI anomaly analysis with rule-based logic to distinguish between moisture-driven deficits and biologically driven stress.

The methodology employs a multi-temporal approach to monitor canopy health across fragmented plots. Preliminary validation using field data from maize and legume crops demonstrates strong correlation between satellite-derived stress alerts and ground-truth observations, achieving a coefficient of determination (R²) of 0.81. Furthermore, the paper details the technical architecture required to bridge the operational gap between cloud-based geospatial processing and simplified, actionable alerts for remote end users.

The analysis also explores the integration of Sentinel-1 Synthetic Aperture Radar (SAR) data to improve radiometric consistency and mitigate the impact of cloud cover on temporal datasets. By quantifying the potential impact of satellite-driven agricultural intelligence on resource efficiency, this work provides a scalable template for applying heterogeneous data fusion to address food security challenges in regions lacking dense ground-based sensor networks. Ultimately, this research supports a shift from broad-scale estimation toward plot-specific, actionable diagnostics in resource-constrained environments.