Furaha-Albert-Patrick/rusle-rwanda-gee

# RUSLE Soil Erosion Model for Rwanda (Google Earth Engine) A ready-to-run **Revised Universal Soil Loss Equation (RUSLE)** implementation in Google Earth Engine that estimates annual soil loss for any district in Rwanda — or any FAO GAUL level-2 district worldwide — at **10 m resolution**. > **Change one line — the district name — and run.** No other edits required. `A = R × K × LS × C × P` → soil loss in tonnes per hectare per year (t ha⁻¹ yr⁻¹) --- Run it live at "code.earthengine.google.com" ## Why this version is scientifically sound Most quick RUSLE scripts badly **over-estimate** erosion on steep terrain — and Rwanda is "the land of a thousand hills." The usual culprit is the old Wischmeier & Smith slope-steepness factor, which was never validated above ~50% slope and explodes on hillsides. This implementation fixes that and uses defensible, peer-reviewed methods for every factor: | Factor | Method | Source | |--------|--------|--------| | **R** – rainfall erosivity | Linear, `R = 38.5 + 0.35 × P` | Roose-type, calibrated for East Africa | | **K** – soil erodibility | EPIC equation from sand/silt/clay/organic carbon | Williams (1995) | | **LS** – topography | **Sine-based S factor** + RUSLE β slope-length | **McCool et al. (1987)** | | **C** – cover management | `C = (1 − NDVI) / 2`, full-year composite | Durigon et al. (2014) | | **P** – support practice | Land cover + slope-dependent terracing factor | ESA WorldCover 10 m | The **LS fix is the key one**: McCool's formula uses the *sine* of the slope angle (not the tangent) specifically to stop steep-slope over-estimation. On a 30° hillside this is roughly one-third of the old polynomial value, which is the difference between forest reading "moderate" instead of an impossible "severe." --- ## Data sources (all free, in the GEE catalog) - **Rainfall:** CHIRPS pentad precipitation - **Terrain:** SRTM 30 m DEM - **Vegetation:** Sentinel-2 Surface Reflectance …