GHSL — population projection to 2100 (GHS-WUP, R2025A)¶

The R2025A GHS-WUP (World Urbanization Prospects) family projects population and built-up forward to 2100. Here we pull GHS-WUP-POP at two epochs — 2025 and 2100 — over a Portugal-sized AOI and compare. WUP is whole-globe 1 km (no tiles), so each epoch downloads the global file once and crops.

Code cells are NBVAL_SKIP (whole-globe downloads); the saved outputs are real.

Setup¶

Imports up front, the same way the other examples open. pyramids provides Dataset (reading the GeoTIFFs that download() writes); earthlens provides the unified EarthLens entry point. tempfile gives us a scratch directory for the downloaded files.

In [ ]:

# NBVAL_SKIP
import tempfile

import matplotlib.pyplot as plt
from pyramids.dataset import Dataset

from earthlens.core import EarthLens

# NBVAL_SKIP import tempfile import matplotlib.pyplot as plt from pyramids.dataset import Dataset from earthlens.core import EarthLens

1 · Download GHS-WUP-POP at two epochs¶

GHS-WUP is whole-globe 1 km (no tiles), so each epoch downloads the global file once and crops to the AOI. We build the request first — source, variable, the 2025 and 2100 epochs, release, and a Portugal-sized bounding box — then download() writes one GeoTIFF per epoch into the scratch directory.

In [ ]:

# NBVAL_SKIP
out = tempfile.mkdtemp()
wup = EarthLens(
    data_source='ghsl',
    variables=['GHS_WUP_POP'],
    start='2025-01-01',
    end='2100-12-31',
    epochs=[2025, 2100],
    release='R2025A',
    aoi=[-9.6, 36.9, -6.0, 42.2],
    path=out,
)
paths = wup.download(progress_bar=False)

# NBVAL_SKIP out = tempfile.mkdtemp() wup = EarthLens( data_source='ghsl', variables=['GHS_WUP_POP'], start='2025-01-01', end='2100-12-31', epochs=[2025, 2100], release='R2025A', aoi=[-9.6, 36.9, -6.0, 42.2], path=out, ) paths = wup.download(progress_bar=False)

The two written files — one per epoch — carry the epoch in their names:

In [ ]:

# NBVAL_SKIP
[p.name for p in paths]

# NBVAL_SKIP [p.name for p in paths]

2 · Compare projected population, 2025 vs 2100¶

Read each GeoTIFF into a population array, masking the negative no-data fill to NaN, and key the arrays by epoch so the plot can pull them by year.

In [ ]:

# NBVAL_SKIP
# GHS-WUP-POP declares its absent-cell fill (-200) in the band, so pyramids
# masks it in the map and in the statistics alike — no sentinel comparison and
# no second copy of the array are needed.
rasters = {
    ('2025' if p.name.find('_E2025') >= 0 else '2100'): Dataset.read_file(p)
    for p in paths
}
for year, ds in rasters.items():
    print(f'{year}: declared nodata {ds.no_data_value}')

# NBVAL_SKIP # GHS-WUP-POP declares its absent-cell fill (-200) in the band, so pyramids # masks it in the map and in the statistics alike — no sentinel comparison and # no second copy of the array are needed. rasters = { ('2025' if p.name.find('_E2025') >= 0 else '2100'): Dataset.read_file(p) for p in paths } for year, ds in rasters.items(): print(f'{year}: declared nodata {ds.no_data_value}')

Side-by-side maps¶

Render the 2025 and 2100 population grids on a shared magma colour ramp so the spatial redistribution over the century is directly comparable.

In [ ]:

# NBVAL_SKIP
fig, axes = plt.subplots(1, 2, figsize=(11, 5))
for ax, year in zip(axes, ['2025', '2100']):
    glyph = rasters[year].plot(
        fig=fig, ax=ax, cmap='magma', title=f'GHS-WUP-POP {year} (1 km)'
    )
    glyph.cbar.set_label('people / cell')
plt.tight_layout()

# NBVAL_SKIP fig, axes = plt.subplots(1, 2, figsize=(11, 5)) for ax, year in zip(axes, ['2025', '2100']): glyph = rasters[year].plot( fig=fig, ax=ax, cmap='magma', title=f'GHS-WUP-POP {year} (1 km)' ) glyph.cbar.set_label('people / cell') plt.tight_layout()

3 · Projected change in total population¶

Summing each grid gives the AOI-wide population per epoch; the difference is the projected change across the AOI, 2025 → 2100.

In [ ]:

# NBVAL_SKIP
tot = {year: float(ds.read_array(masked=True).sum()) for year, ds in rasters.items()}
print(f"2025: {tot['2025']:,.0f}  ->  2100: {tot['2100']:,.0f}")
print(f"change: {tot['2100'] - tot['2025']:+,.0f}")

for ds in rasters.values():
    ds.close()

# NBVAL_SKIP tot = {year: float(ds.read_array(masked=True).sum()) for year, ds in rasters.items()} print(f"2025: {tot['2025']:,.0f} -> 2100: {tot['2100']:,.0f}") print(f"change: {tot['2100'] - tot['2025']:+,.0f}") for ds in rasters.values(): ds.close()