SSA-LULC Systematic Review — Extraction Matrix and Supplementary Data

Extraction matrix and supplementary data for the systematic review of land use and land cover (LULC) classification methodologies in Sub-Saharan Africa (1991-2026), conducted following the PRISMA 2020 protocol. The main dataset comprises 164 eligible peer-reviewed articles with 22 extracted variables, covering radiometric preprocessing (TOA vs ARD), algorithm generation (parametric vs machine learning vs deep learning), spatial validation protocols, sampling strategies, and OA plausibility assessments. The meta-regression OLS model (N=155; logit-transformed Overall Accuracy; EPV=31:1) identified is_MLC as the only statistically significant predictor (β=−0.369; p=0.024), robust to HC3 errors, bootstrap (B=5000) and Cook's Distance sensitivity analysis. Fisher's exact test for GEE_Native × Spatial_CV association: OR=1.57; p=0.544 (not significant). Contents: extraction matrix (164×22), Codebook V2.1, extraction protocol V2.1, diagnostic tables (VIF, Shapiro-Wilk, Breusch-Pagan), full article list, Fisher exact 2×2, plausibility assessment (Risco D8×F6), OLS diagnostic figures (300 DPI PNG + vector EPS). Computational pipeline: https://github.com/norisfrancisco/ssa-lulc-systematic-review OSF project: https://osf.io/cg8wp/