A 100 m uncertainty-aware dataset of rooftop rainwater harvesting potential for South Asia derived from building footprints, remote sensing, and probabilistic rainfall modelling

The Rain_models_output.shp is a point shapefile summarizing modeled rainfall statistics at each location; the annual_rainfall_YYYY_chirps_native.nc files are CHIRPS-based annual precipitation grids; and the remaining GeoTIFFs report roof area, runoff coefficient, and rainwater-harvesting outputs (total and per capita) at ~100 m or 0.05° resolution. Key attributes and units are noted below, and all layers use standard GIS conventions for CRS and format.Rain_models_output.shpThis is an ESRI Shapefile of point features (likely in geographic coordinates, e.g. WGS84) covering each analysis pixel.

It contains rainfall-model summary statistics derived from CHIRPS data (1981–2025). Main attributes include pixel_id (int, unique ID), lat, lon (float, degrees), n_wet_days (int, count of rainy days), mean_precip, max_precip (float, mm), best_fit (string, name of best-fit distribution), AIC, RMSE (float, model-fit metrics), and key quantiles p10, p50, p90 (float, mm) among others. The shapefile was produced by fitting probability distributions to the local annual CHIRPS rainfall; its accuracy is limited by CHIRPS interpolation errors and sparse station data.

Primary uses include analyzing spatial rainfall variability, comparing regional precipitation behaviors, and informing rainwater-harvesting potential and hydrological modeling.annual_rainfall_YYYY_chirps_native.ncEach NetCDF file is a gridded raster of total annual precipitation (mm) from the CHIRPS v2 dataset. These files span 1981–2025 (one file per year) over 50°S–50°N at the native 0.05° (~5 km) lat/lon resolution.

The coordinate system is geographic (EPSG:4326). The main variable is precipitation (float, mm/year) indexed by latitude, longitude, and time (year). These annual rainfall surfaces are used for trend analysis and as inputs to modeling; they inherit CHIRPS strengths and limitations (e.g. high-resolution satellite blending with station data).100m_roofs.tifThis GeoTIFF raster has a 100 m spatial resolution (CRS likely World Mollweide, EPSG:54009) and represents the total rooftop area within each cell (m²).

It was derived from building footprint data (Rwanda’s 2020 cadaster supplemented by the GHSL built-up surface). Pixel values (unsigned int) give the sum of roof surface (in square meters) per 100 m cell. There is no temporal component (static map circa 2020/2023).

Accuracy depends on the completeness of the source building data; small or recent structures may be undercounted. This layer provides the catchment area available for rainwater harvesting on rooftops.Runoff_Coefficient.tifThis 100 m raster (same CRS as the roofs layer) stores dimensionless runoff coefficients (0–1) for each cell, representing the fraction of rainfall assumed to become runoff on a typical rooftop.

Values (float) were assigned from literature-based defaults (e.g. 0.8–0.9 for impervious roof materials). There is no time dimension (constant coefficients). As a simplification, it assumes uniform roof efficiency across each cell; actual runoff can vary with roof type and intensity.

This coefficient raster is used to convert rainfall depth into harvestable runoff volume.Annual_P90_RRWH.nc (.tif)These outputs (NetCDF and GeoTIFF) map the annual volume of harvestable rooftop rainwater under a conservative rainfall scenario (90th percentile rainfall) at each grid cell. They use the 1981–2025 CHIRPS rainfall, the 100m_roofs area, and Runoff_Coefficient to compute water yield. The spatial resolution matches the precipitation grid (~0.05°, lat/lon, EPSG:4326).

Main variables include harvest_volume (float, e.g. m³ per year per cell). The data cover each year’s precipitation with P90 thresholds, providing conservative (low-end) estimates of annual rainwater potential. Quality depends on the accuracy of the input rainfall and roof-area data.

Applications include water resource assessment and design of rainwater harvesting systems.Annual_P90_per_capita_RRWH.nc (.tif)These NetCDF/GeoTIFF files report per-capita annual rainwater availability under the same P90 rainfall scenario. They combine the Annual_P90_RRWH volume grids with a population grid (GHSL 2023A at 100 m resolution). The result is water_per_person (float, e.g. m³/person/year) on a ~0.05° grid (EPSG:4326).

Temporal coverage is annual (1981–2025 used for rainfall; population is a static 2023A estimate). This metric helps assess water supply equity and demand: higher values indicate more harvestable water per person. Uncertainty arises from population data accuracy and the assumption of static population distribution.File Format CRS Temporal Coverage Main variablesRain_models_output.shp Shapefile (vector) Geographic (WGS84, lat/lon) 1981–2025 (annual data) pixel_id; lat, lon; n_wet_days; mean_precip, max_precip; best_fit; AIC, RMSE; p10, p50, p90 (precip., mm)annual_rainfall_YYYY_chirps_native.nc NetCDF (raster) Geographic (WGS84, lat/lon) 1981–2025 (annual) precipitation (float, mm/year)100m_roofs.tif GeoTIFF (raster) World Mollweide (EPSG:54009) ~2020 (static) roof_area (unsigned int, m² per 100 m cell)Runoff_Coefficient.tif GeoTIFF (raster) World Mollweide (EPSG:54009) static runoff_coefficient (float, dimensionless 0–1)Annual_P90_RRWH.nc / .tif NetCDF / GeoTIFF Geographic (WGS84, lat/lon) 1981–2025 (annual) harvest_volume (float, m³ per cell per year)Annual_P90_per_capita_RRWH.nc / .tif NetCDF / GeoTIFF Geographic (WGS84, lat/lon) 1981–2025 (annual) water_per_person (float, m³ per person per year)Sources: CHIRPS precipitation data; GHSL population (2023A, 100 m); GHSL built-up surfaces (2023A, 100 m); and standard hydrologic modeling methods.