phoebe-slight/nile-valley-landscape-change

Research Question: How can AI algorithms and machine-learning be applied to assess changing archaeological landscapes in Egypt, and in what ways can these methods enhance archaeological interpretation and predictive modelling? # Nile Valley Landscape Change Detection (Google Earth Engine) This repository contains two Google Earth Engine (GEE) scripts developed for an undergraduate dissertation examining land-cover change and archaeological visibility in Egypt's Nile Valley, centred on Luxor. ## Research Question *How can multi-temporal satellite imagery, classified using Random Forest in Google Earth Engine, detect land-cover change in the Nile Valley, and what do these changes reveal about archaeological visibility and vulnerability?* ## Scripts **Script A: Land-cover Classification** Preprocesses Sentinel-2 imagery, builds a multi-predictor stack (spectral bands, NDVI, NDWI, NDBI, BSI, IBI, VIIRS night-time lights, SRTM elevation and slope), trains a Random Forest classifier and exports classification outputs and validation results. **Script B: Change Detection** Applies the trained classifier to before and after composites, generates a land-cover transition matrix, runs EAMENA archaeological site buffer analysis and includes a custom interactive GEE panel for selecting dates and transition types. ## How to Use 1. Open the GEE Code Editor at code.earthengine.google.com 2. Upload your Area of Interest as a FeatureCollection 3. Upload training points as a FeatureCollection with a `landcover` property (0 = water, 1 = vegetation, 2 = bare ground, 3 = built-up, 4 = mountainous) 4. Replace all asset paths marked `users/[username]/` with your own 5. Click Run ## Requirements - Google Earth Engine account - Sentinel-2 Level-2A (`COPERNICUS/S2_SR_HARMONIZED`) - VIIRS night-time lights (`NOAA/VIIRS/DNB/MONTHLY_V1/VCMSLCFG`) - SRTM DEM (`USGS/SRTMGL1_003`) - Landsat 8/9 Collection 2 (`LANDSAT/LC08/C02/T1_L2`, `LANDSAT/LC09/C02/T1_L2`) ## Attribution Developed by Phoebe Slight (2026), Newcastle University Supervised by Dr Louise Rayne, Newcastle University