chilispammm/crop-stress-intelligence-kenya

# Multi-Modal Agricultural Early Warning & Crop Health Intelligence Pipeline An end-to-end Earth Observation (EO) screening and spatial triage framework designed to identify anomalous co-occurrences of precipitation, root-zone soil moisture dynamics, and high-resolution optical canopy greenness across agricultural landscapes in Kenya. --- ## 1. Project Purpose & Problem Statement Smallholder agricultural monitoring requires balancing spatial resolution and hydrological context. Optical satellite data (Sentinel-2 at 30 m) detects fine-scale canopy changes but cannot explain their underlying drivers, while macro-climatic datasets (CHIRPS at ~5.5 km, NASA SMAP at ~9.0 km) provide moisture context but lack field-scale granularity. This system bridges that gap by establishing a **deterministic, multi-modal screening matrix** that enforces strict multi-sensor resolution integrity without artificial resampling. The pipeline identifies discrete, actionable candidate scouting polygons ($\ge 2.0\text{ ha}$ Minimum Mapping Unit) to guide ground extension officers on where to scout, rather than making unverified causal claims from space. --- ## 2. Study Area & Spatial Hierarchy * **Regional Pilot AOI (`ug_pilot_moiben_01`)**: $0.20^\circ \times 0.20^\circ \approx 490.5\text{ km}^2$ ($49,050\text{ ha}$) in Uasin Gishu County, Kenya (`EPSG:4326`: $[35.15^\circ\text{ E}, 0.55^\circ\text{ N}, 35.35^\circ\text{ E}, 0.75^\circ\text{ N}]$). * **Evaluated Focal Grid**: $86\text{ rows} \times 65\text{ columns}$ at $30.0\text{ m}$ in `EPSG:6933` ($503.10\text{ ha}$, $5,590\text{ cropland pixels}$). --- ## 3. Sensor Modalities & Core Workflow ``` [ CHIRPS Rainfall (5.5 km) ] [ Sentinel-2 Reflectance (30 m) ] [ NASA SMAP L4 SWI (9 km) ] │ │ │ ▼ ▼ ▼ Precipitation Z-Score Vegetation Z-Score Root-Zone SWI & ΔSWI_14d (Z_R …