GeoTIFF Format and WorldPop Population Rasters¶

What is GeoTIFF?¶

GeoTIFF is a raster image format based on TIFF (Tagged Image File Format) with embedded georeferencing metadata. Each pixel in the image maps to a specific geographic location via:

Coordinate Reference System (CRS) -- defines the projection (e.g. EPSG:4326 for WGS 84 lat/lon).
Affine transform -- maps pixel row/column indices to geographic coordinates (origin, pixel size, rotation).
NoData value -- a sentinel value marking pixels with no valid data (e.g. ocean, uninhabited areas).

GeoTIFF files use the .tif extension and can be read by any GIS tool (QGIS, ArcGIS, GDAL) as well as Python libraries like rasterio and rioxarray.

How WorldPop encodes population¶

WorldPop publishes gridded population estimates as GeoTIFF rasters at ~100 m resolution. Each pixel value represents the estimated number of people living in that grid cell. Key properties:

CRS: EPSG:4326 (WGS 84 geographic coordinates)
Pixel size: ~100 m (approximately 0.000833 degrees at the equator)
Pixel values: Floating-point population count (can be fractional)
NoData: Typically -99999 or nan for ocean, lakes, and uninhabited areas

To get the total population of a region, you sum all valid pixel values that fall within its boundary -- this operation is called zonal statistics.

WorldPop Age/Sex Structures dataset: R2025A¶

The R2025A release provides age- and sex-disaggregated population estimates for 231 countries covering 2015--2030.

Coverage¶

231 countries and territories worldwide (full global coverage including Asia, unlike the wpgpas API dataset which is limited mostly to Africa)
Years: 2015, 2016, ..., 2030
Resolution: 100 m (constrained variant)

File naming convention¶

Files are organized under a consistent URL hierarchy:

https://data.worldpop.org/GIS/AgeSex_structures/Global_2015_2030/
  R2025A/{year}/{ISO3}/v1/100m/constrained/
    {iso3}_{group}_{year}_CN_100m_R2025A_v1.tif

Where: - {year} -- four-digit year (e.g. 2024) - {ISO3} -- uppercase ISO 3166-1 alpha-3 code (e.g. LAO) - {iso3} -- lowercase ISO code in the filename (e.g. lao) - {group} -- age/sex group identifier (see below)

Available files per country/year¶

Each country/year directory contains 42 individual rasters plus 2 summary rasters:

Individual age-group files (20 age groups x 2 sexes = 40 files):

Label	Age range	Male file	Female file
`0`	0--12 months	`{iso3}_M_0_{year}_...`	`{iso3}_F_0_{year}_...`
`1`	1--4 years	`{iso3}_M_1_{year}_...`	`{iso3}_F_1_{year}_...`
`5`	5--9 years	`{iso3}_M_5_{year}_...`	`{iso3}_F_5_{year}_...`
`10`	10--14 years	`{iso3}_M_10_{year}_...`	`{iso3}_F_10_{year}_...`
`15`	15--19 years	`{iso3}_M_15_{year}_...`	`{iso3}_F_15_{year}_...`
`20`	20--24 years	`{iso3}_M_20_{year}_...`	`{iso3}_F_20_{year}_...`
`25`	25--29 years	`{iso3}_M_25_{year}_...`	`{iso3}_F_25_{year}_...`
`30`	30--34 years	`{iso3}_M_30_{year}_...`	`{iso3}_F_30_{year}_...`
`35`	35--39 years	`{iso3}_M_35_{year}_...`	`{iso3}_F_35_{year}_...`
`40`	40--44 years	`{iso3}_M_40_{year}_...`	`{iso3}_F_40_{year}_...`
`45`	45--49 years	`{iso3}_M_45_{year}_...`	`{iso3}_F_45_{year}_...`
`50`	50--54 years	`{iso3}_M_50_{year}_...`	`{iso3}_F_50_{year}_...`
`55`	55--59 years	`{iso3}_M_55_{year}_...`	`{iso3}_F_55_{year}_...`
`60`	60--64 years	`{iso3}_M_60_{year}_...`	`{iso3}_F_60_{year}_...`
`65`	65--69 years	`{iso3}_M_65_{year}_...`	`{iso3}_F_65_{year}_...`
`70`	70--74 years	`{iso3}_M_70_{year}_...`	`{iso3}_F_70_{year}_...`
`75`	75--79 years	`{iso3}_M_75_{year}_...`	`{iso3}_F_75_{year}_...`
`80`	80--84 years	`{iso3}_M_80_{year}_...`	`{iso3}_F_80_{year}_...`
`85`	85--89 years	`{iso3}_M_85_{year}_...`	`{iso3}_F_85_{year}_...`
`90`	90+ years	`{iso3}_M_90_{year}_...`	`{iso3}_F_90_{year}_...`

Summary rasters (2 files -- these are what our flow downloads):

Label	Description	Filename
`T_M`	Total male (sum of all male age groups)	`{iso3}_T_M_{year}_CN_100m_R2025A_v1.tif`
`T_F`	Total female (sum of all female age groups)	`{iso3}_T_F_{year}_CN_100m_R2025A_v1.tif`

Example download URLs¶

# Total female, Laos, 2024
https://data.worldpop.org/GIS/AgeSex_structures/Global_2015_2030/R2025A/2024/LAO/v1/100m/constrained/lao_T_F_2024_CN_100m_R2025A_v1.tif

# Total male, Laos, 2024
https://data.worldpop.org/GIS/AgeSex_structures/Global_2015_2030/R2025A/2024/LAO/v1/100m/constrained/lao_T_M_2024_CN_100m_R2025A_v1.tif

Each file is approximately 15 MB.

Zonal statistics¶

Zonal statistics is the process of summarizing raster pixel values within vector polygon boundaries. For population estimation:

Load the raster -- open the GeoTIFF with rioxarray.open_rasterio()
Clip to polygon -- use .rio.clip([geometry]) to extract only the pixels that fall within the org unit boundary
Handle nodata -- mask out nodata pixels (ocean, lakes, etc.)
Sum values -- the sum of remaining pixel values gives the total population inside that boundary

This is done independently for the T_M (male) and T_F (female) rasters to produce sex-disaggregated population counts.

Comparison: API approach vs GeoTIFF approach¶

Aspect	API (wpgpas/wpgppop)	GeoTIFF (R2025A)
Coverage	wpgpas: mostly Africa; wpgppop: broader but no sex split	231 countries worldwide
Sex disaggregation	wpgpas: yes; wpgppop fallback: 50/50 split	Always available (T_M / T_F)
Data freshness	Varies by dataset	2015--2030 (annually)
Network calls	One HTTP POST per org unit per year	Two file downloads per country/year (shared across org units)
Latency	Async polling, 10--600 s per request	Seconds per zonal extraction (after download)
Offline capable	No -- requires live API	Yes -- once files are cached locally
Area limit	100,000 km^2 per request	No limit (entire country raster)
Accuracy	Summary statistics from server	Full pixel-level extraction