Malwa · Madhya Pradesh · measured Jan 2023 – Dec 2025

Two check dams.
Only one is banking water.

Both structures were built to hold monsoon runoff and let it soak into the aquifer. Thirty-six months of satellite measurement say one of them does exactly that — and the other has not held water once.

Where the water stands

Three years, month by month

Sentinel-2 at 10 m, one frame per month. Turbid Malwa pond water is the same brown-green as a wet field in plain colour, so the measured water is painted blue. The second layer reads the ground itself: bare rock and soil, vegetation, water. Blank months are monsoon cloud.

The ledger

Three years of water through Barkheda Kamaliya. Recharge is the water that soaked through the bed into the aquifer — the part a farmer gets back from a borewell. Evaporation is the part the sky took.

One fill a year, then five months of drawdown

Water surface measured from Sentinel-2. The dam fills at the end of the monsoon and is empty again by February. July and August have no bar because cloud blocks the optical sensor.

Water surface (m²) Open-water evaporation (mm) No cloud-free scene

The soil beneath, from two radiometers

SMAP (NASA, L-band, 9 km) and AMSR2 (JAXA, C-band, 10 km) both measure moisture in the top few centimetres, through cloud, every day. Both are district-scale readings around each dam, never the dam itself.

Over the 33 months where both answered at Barkheda, AMSR2 reads about 3.5× lower than SMAP (median ratio 3.48) and the two correlate only moderately (r = 0.44). L-band penetrates a crop canopy far better than C-band, which is the usual explanation for the offset — but these two instruments plainly do not substitute for one another, so both lines are drawn and neither is averaged into the other.

SMAP L-band (% volumetric) AMSR2 C-band (% volumetric)

Seen from above

Esri World Imagery at 1.09 m per pixel, every distinct capture since 2022 — sharp enough to see the wall, the apron and the bed, but roughly one capture a year.

How the balance closes

Storage change is measured; evaporation is physics; seepage is an assumed rate. What is left over is water a person took. In 23 of 26 month-pairs nothing was left over — the dam emptied at exactly the speed sun and soil explain.

MonthStorageChange EvaporationRechargeUnexplained

Unexplained loss is a residual: it absorbs every error in the other terms. Read it as an upper bound on pumping, not a meter reading.

What this is built from

Sentinel-2 L2A · 10 mWater surface and land structure, monthly. Public COGs on AWS.
Sentinel-1 RTC · 10 mRadar through monsoon cloud. Microsoft Planetary Computer.
SMAP SPL3SMP_E · 9 kmL-band soil moisture, daily. NASA Earthdata / NSIDC.
AMSR2 L3 SMC · 10 kmC-band soil moisture, daily. JAXA G-Portal.
GPM IMERG · 0.1°Rainfall. Separates the two dams, which a 0.5° grid cannot.
NASA POWERTemperature, humidity, wind, solar — the inputs to Penman evaporation.
Copernicus GLO-30Terrain. Turns a measured water surface into a stored volume.
Esri World ImageryDated sub-metre stills, 2022 to 2026.

Seepage is assumed at 8 mm/day — mid-range for the silty beds of Malwa check dams, and the one term no satellite measures. Every figure that depends on it scales linearly.