The inversion
One choice flips the ranking
District mean scores with annual rainfall versus dry-season rainfall (Nov–May).
Annual rainfallDry-season rainfall
Loading districts…
Under reviewPaper submitted for publication
DOI10.5281/zenodo.21422759
Where is water hardest to reach in Sindh?
A drinking-water access score for 5,159 villages, built only from open satellite and geospatial data. The usual way of building such an index turns the ranking upside down, making Pakistan's driest district look well served.
Key findings
Annual rain hides the dry months
Most published water indices treat annual rainfall as supply. In Sindh the monsoon falls in a few weeks of July and August and is gone by March, so annual totals point the wrong way.
71.5→45.9
Mithi (Tharparkar), Pakistan's driest district: near the top with annual rainfall, last of 25 with dry-season rainfall.
Last in every one of 500 weight bootstraps
14.5%
Average share of a village's annual rain that falls between November and May. The range runs from 0.6% to 39.6%.
Annual vs dry-season indicators: r = −0.67
88.6%
Villages whose nearest perennial water is a canal, not a natural river. Natural hydrography alone misses nine villages in ten.
4,573 canals vs 586 rivers
ρ +0.72
Districts scored most accessible are among the least safe. Thatta scores 72.9, yet every published water sample there was unsafe.
6 districts · PCRWR 2021; Solangi et al. 2017
Accessibility is not potability. The index has no salinity data, so the coastal delta, where seawater has spoiled groundwater, scores well.
District explorer
Look up a district
Scores are relative within Sindh, from 0 to 100. Tap a village for its five indicators, or switch to annual rainfall to see the inversion.
Loading villages…
Groundwater sample, Thatta
Area of mapped villages
Dry-season rainfall · as published
Method
Five open indicators, one score
Each indicator is log-transformed where skewed, winsorised at the 1st and 99th percentiles, and rescaled to 0–100 so that higher always means better access.
Rin S · 45%
Distance to perennial water
HydroRIVERS (discharge over 1 m³/s) merged with OpenStreetMap canals.
closer = better
Win S · 45%
Dry-season rainfall
CHIRPS daily rainfall, November–May, averaged over 2020–2024.
more = better
N20%
Surface water
Sentinel-2 NDWI, 2023 cloud-masked median, mean within 1 km.
more = better
P20%
Population within 2 km
WorldPop 2024, constrained and UN-adjusted, 100 m grid.
fewer = better
E15%
Elevation
SRTM 1 arc-second. Lower ground sits nearer the water table.
lower = better
S = max(R, W)
Score = 0.45·S + 0.20·P
+ 0.20·N + 0.15·E // 0–100
Rivers or canals and rainfall are substitutes, not complements. A canal past the village makes rain matter less; in the Thar desert, with no canal, rain is everything. Taking the better of the two avoids penalising either kind of village. Bands are percentiles: Poor is the bottom 20%, Good the top 20%.
Answered
Accessible
Is water likely within reach?
A relative score per village. District rankings hold across weighting choices (Spearman ρ 0.84–0.96).
Not answered
Safe to drink
Is the water potable?
There is no salinity data. Thatta and Badin rank near the top, yet that is where seawater intrusion has made groundwater undrinkable.
Partly
Checked on the ground
Does the score match reality?
21 groundwater samples near Thatta all exceed 500 mg/L TDS beside villages rated Good. The checks show the access–safety gap, not score accuracy.
- Scores are relative. 70 means better served than most Sindhi villages, not adequate by any external standard.
- The population indicator barely moves the score (correlation 0.06), and its direction, more people meaning worse access, may be wrong.
- Rainfall is coarse. CHIRPS cells are about 5.5 km and undercount dry-season rain in hyper-arid areas, so Thar's figures are lower bounds.
- River discharge is naturalised. HydroRIVERS models flow before abstraction, so the lower Indus carries less than it shows.
Data and citation
Use and cite SWAI
The full dataset and pipeline are open. Code is MIT-licensed; derived data is for non-commercial research and education, following its sources.
Paper under review
The paper describing SWAI is in peer review. This page will link to it once it is published. Until then, please cite the dataset.
SWAI asks whether open satellite and geospatial data can show village-level water access where household surveys stop at the district. Its main lesson is methodological: the indicator choice that looks standard reverses the answer.
APA 7th edition
Citing the dataset for now. A citation for the paper will be added here when it is published.