Clindatta-1/GIS

Forest Cover Change Detection & Land-Cover Classification (Kenya AOI) # GIS Forest Cover Change Detection & Land-Cover Classification (Kenya AOI) A complete, reproducible GIS/remote-sensing pipeline for **land-cover classification**, **multi-temporal change detection**, and **accuracy assessment** — built around a nature-based carbon (REDD+/forest monitoring) use case for a simulated Area of Interest (AOI) near the Mau Forest Complex, Kenya. This project demonstrates the core outputs commonly required in GIS/remote-sensing and forest carbon project roles: NDVI-based land-cover classification, change-detection mapping, shapefile/GeoPackage exports with attribute metadata, statistical accuracy assessment (confusion matrix, Kappa), and publication-quality cartographic map layouts. ## Why this project Forest monitoring for carbon projects (REDD+, ARR) depends on being able to answer three questions reliably: *What land cover exists now? What changed? How confident are we in the answer?* This pipeline answers all three, end-to-end, using an open-source Python geospatial stack — the same core workflow used with real Sentinel-2/Landsat imagery pulled from Google Earth Engine or QGIS. ## Methodology | Step | Script | What it does | |---|---|---| | 1 | `01_generate_synthetic_data.py` | Generates two NDVI raster surfaces (2015 baseline, 2024) simulating a forest AOI with realistic disturbance patches | | 2 | `02_land_cover_classification.py` | Threshold-based NDVI classification into Water / Bare-Cleared / Cropland / Forest, with area (ha) summaries | | 3 | `03_change_detection.py` | Post-classification change detection; vectorizes change polygons into a GeoPackage + Shapefile with attributes | | 4 | `04_accuracy_assessment.py` | Stratified reference-point sampling, confusion matrix, overall accuracy, Cohen's Kappa, producer's/user's accuracy | | 5 | `05_generate_map_layouts.py` | Publication-ready cartographic maps (title, legend, north arrow, scale bar, coordinate grid) | **Note on data:** This project uses procedurally generated …