tillytales92/urbanafrica

Urban Africa project # urbanafrica Measuring urban built-up expansion across the **100 largest African agglomerations** (by Africapolis 2020 population), 2000–2025, and characterising *how* they grew — sprawl vs. densification, lit (formal) vs. unlit (informal) settlement. Two satellite inputs: - **GHSL-BUILT-S R2023** — built-up surface, m² per 100 m cell, six epochs (2000, 2005, 2010, 2015, 2020, 2025), total and non-residential layers. - **Nighttime lights** — harmonised DMSP-OLS + VIIRS annual series 2000–2024, via Google Earth Engine; used to classify each built-up pixel lit / unlit. The main product is a **Shiny app** (`app/`) with interactive growth maps, NTL / lit-unlit maps, multi-city time series, a rankings table, and sprawl scatterplots. ## Repository layout ``` scripts/ pipeline scripts — run from the repo root: Rscript scripts/ .R scripts/exploratory/ one-off / prototype scripts, not part of the pipeline app/ the Shiny app (app.R + a small app/data/ bundle) data/ source + derived data — NOT in git (see below) output/ generated plots / maps — NOT in git CLAUDE.md detailed pipeline, data architecture, conventions, planned work ``` ## Data The `data/` tree (~100 GB of rasters) is **not** version-controlled. Source data must be downloaded manually: - **GHSL** (BUILT-S, and optionally POP / BUILT-V): - **Africapolis 2020** agglomeration boundaries: - **Nighttime lights** via `scripts/0_loadntl.R` (needs an Earth Engine account). See the *Data architecture* and *Pipeline & run order* sections of `CLAUDE.md` for exact paths and the order scripts must run in. The five small `.Rds` files in `app/data/` **are** tracked, so the app can run without regenerating the full pipeline. ## Running the app ```r # from the repo root shiny::runApp("app", launch.browser = TRUE) ``` Package dependencies load via `pacman::p_load(...)` at the top of each script. ## Environment variables `scripts/0_loadntl …