Unmapped Dataworks

Unmapped Search

Dam · North West · on the Maretlwane River

How full · 9 November 2026

Middelkraal Dam

94.8% full

A week before: 95.6%. A year before: 97.8% (the ring on the bar).

Last week
95.6%
A year ago
97.8%
Holds when full
0.8 million m³
Holds now
0.8 million m³

Save the image

01

The level, week by week

Full

25 Aug 20259 Nov 2026

This yearA year earlier53 weekly readings, 40% to 110%
Show the numbers
How full it was in each weekly report: every figure
When or whatFigure
9 November 202694.8% full, a year earlier 97.8%
2 November 202695.6% full, a year earlier 97.8%
26 October 2026100% full, a year earlier 97.8%
19 October 2026100% full, a year earlier 97.8%
12 October 2026100% full, a year earlier 97.8%
5 October 2026100% full, a year earlier 97.8%
31 August 2026100% full, a year earlier 97.8%
24 August 2026100% full, a year earlier 97.8%
17 August 2026100% full, a year earlier 98.1%
10 August 2026100% full, a year earlier 100%
3 August 2026100% full, a year earlier 100%
29 June 2026100% full, a year earlier 100%
22 June 2026100% full, a year earlier 100%
15 June 2026100% full, a year earlier 100%
8 June 2026100% full, a year earlier 100%
1 June 2026100% full, a year earlier 100%
25 May 2026100% full, a year earlier 100%
18 May 2026100% full, a year earlier 100%
11 May 2026100% full, a year earlier 100%
4 May 2026100% full, a year earlier 100%
20 April 2026100% full, a year earlier 100%
13 April 2026100% full, a year earlier 100%
6 April 2026100% full, a year earlier 100%
30 March 2026100% full, a year earlier 100%
16 March 2026100% full, a year earlier 100%
9 March 2026100% full, a year earlier 100%
2 March 202699.1% full, a year earlier 100%
23 February 202699.1% full, a year earlier 100%
16 February 2026100% full, a year earlier 100%
9 February 202698.7% full, a year earlier 100%
2 February 2026100% full, a year earlier 100%
26 January 2026100% full, a year earlier 100%
19 January 2026100% full, a year earlier 96.1%
12 January 2026100% full, a year earlier 50.5%
5 January 202697.8% full, a year earlier 50.5%
29 December 202597.8% full, a year earlier 50.5%
15 December 202597.8% full, a year earlier 50.5%
8 December 202597.8% full, a year earlier 50.5%
1 December 202597.8% full, a year earlier 50.5%
24 November 202597.8% full, a year earlier 50.5%
17 November 202597.8% full, a year earlier 50.5%
10 November 202597.8% full, a year earlier 50.5%
3 November 202597.8% full, a year earlier 50.8%
27 October 202597.8% full, a year earlier 53%
20 October 202597.8% full, a year earlier 53.9%
13 October 202597.8% full, a year earlier 56.8%
6 October 202597.8% full, a year earlier 58.5%
29 September 202597.8% full, a year earlier 60.9%
22 September 202597.8% full, a year earlier 63.8%
15 September 202597.8% full, a year earlier 64.1%
8 September 202597.8% full, a year earlier 66.1%
1 September 202597.8% full, a year earlier 67.8%
25 August 202597.8% full, a year earlier 69.4%

Source: Weekly State of Dams, 2026-11-09, Department of Water and Sanitation, station A2R018

02

Rain on the dam’s ward

Tallest 179 mm

Nov 2025: 143 mm, normal 85 mmDec 2025: 171 mm, normal 120 mmJan 2026: 96 mm, normal 100 mmFeb 2026: 52 mm, normal 92 mmMar 2026: 179 mm, normal 77 mmApr 2026: 75 mm, normal 39 mmMay 2026: 42 mm, normal 18 mmJun 2026: 9 mm, normal 7 mmJul 2026: 0 mm, normal 3 mmAug 2026: 13 mm, normal 7 mmSep 2026: 62 mm, normal 16 mmOct 2026: 5 mm, normal 55 mm

NDJFMAMJJASO

Rain that monthA normal month (1991 to 2020)Nov 2025 to Oct 2026
Show the numbers
Rain in Ward 26, Local Municipality of Madibeng, month by month, against a normal month: every figure
When or whatFigure
Nov 2025143 mm, normal 85 mm
Dec 2025171 mm, normal 120 mm
Jan 202696 mm, normal 100 mm
Feb 202652 mm, normal 92 mm
Mar 2026179 mm, normal 77 mm
Apr 202675 mm, normal 39 mm
May 202642 mm, normal 18 mm
Jun 20269 mm, normal 7 mm
Jul 20260 mm, normal 3 mm
Aug 202613 mm, normal 7 mm
Sep 202662 mm, normal 16 mm
Oct 20265 mm, normal 55 mm

7.2 mm of rain so far in November; a normal November has 85 mm. This is the rain on ward 26 of Local Municipality of Madibeng, the ward the dam lies in: not on its whole catchment, which we don’t hold. Rain upstream fills a dam, so the two don’t have to move together.

Modelled estimates from a weather reanalysis, averaged over the whole place: not readings from a weather station, and not a forecast. The weather model runs about five days behind the weather, so the newest day here is usually five to twelve days ago.

Source: ERA5-Land, Copernicus Climate Change Service. The weather there, day by day

03

Where it is

River
Maretlwane River
In
North West
Holds when full
0.8 million m³
Holds now
0.8 million m³
Ward
ward 26 of Local Municipality of Madibeng

Towns nearby

The nearest towns in a straight line from the dam. Being near a dam doesn’t mean a town gets its water from it: we hold no link yet between dams and the towns they supply.

Position and towns: © OpenStreetMap contributors, ODbL 1.0.

04

From space

Middelkraal Dam from space

From space on 6 Nov 2026, the water covered 74% of Middelkraal Dam’s full outline (0.20 of 0.27 km²). That fits a dam about 90% full (our estimate, usually within 14 points). The Department of Water and Sanitation’s own figure for the week of 9 Nov 2026 is 94.8%.

Two ways to say how full Middelkraal Dam is

DWS
94.8% full by volume
Weekly State of Dams, 9 Nov 2026
From space, by volume
about 90% full (76 to 104%)
Our estimate from the area, 6 Nov 2026, on a curve fitted to 41 of DWS’s own weekly figures; nine in ten held-out weeks were within 14 points. How we estimate % full from space
From space
73.8% of the full area covered
Sentinel-2, 6 Nov 2026

DWS measures how much water is stored (volume). From space we measure how much ground the water covers (area). A dam’s sides slope, so it is deep at the wall and shallow at the edges: as it empties, the deep water goes while the surface stays wide, so the area falls more slowly than the volume. The two numbers differ even when both are right. In the sketch the level has dropped to 60% of the depth (the dashed line is full): the water still spans 60% of the width and, in a bowl like this, covers about a third of the full area, yet holds only about a fifth of the water.

How we estimate % full by volume from space

For each month with a clear view we take the edge of the water seen by Sentinel-2 and read the height of the ground there from the Copernicus elevation model (GLO-30, 30 m). Heights against areas give the dam’s area-to-height curve, fitted as a smooth curve that only rises. The water held between two levels is the area added up over the height between them. Starting from full, with the full-supply capacity in our dam-levels records (0.8 million m³), the water lost down to a month’s level gives that month’s % full by volume. We show it only where the months spread over enough levels and the curve fits well (R² 0.8 or more); otherwise we say “not enough range yet”. The elevation model was made from radar passes in 2010 to 2015 and shows each lake as a flat surface at its level then, so below that level it has no heights: for many dams only the months near full can be used. This dam: 9 of 12 months read had a shoreline above the elevation model’s water line. An estimate, not a measurement: it can be several points off, and DWS’s figure comes from gauged levels.

Produced using Copernicus WorldDEM-30 © DLR e.V. 2010-2014 and © Airbus Defence and Space GmbH 2014-2018 provided under COPERNICUS by the European Union and ESA; all rights reserved. The organisations in charge of the Copernicus programme by law or by delegation do not incur any liability for any use of the Copernicus WorldDEM-30.

Middelkraal Dam from space on 1993 (Landsat, whole year) Middelkraal Dam from space on 6 Nov 2026 1993 (Landsat, whole year)6 Nov 2026
Contains modified Copernicus Sentinel data 2026. Landsat imagery courtesy of the U.S. Geological Survey; Digital Earth Africa (CC BY 4.0), modified. True colour, the same frame each time. The earlier picture is from 1993, the year with the least water seen from space since 1991 (0.0% of full on average; Landsat’s cloud-free picture of the whole year).

One picture a year

Contains modified Copernicus Sentinel data 2017-2026. Landsat imagery courtesy of the U.S. Geological Survey; Digital Earth Africa (CC BY 4.0), modified. Before 2017 Landsat’s cloud-free picture of the whole year; from 2017 the clearest Sentinel-2 picture of the same season each year. Choose a picture to see it large.

Middelkraal Dam: water seen from space and DWS’s level

Water area from space (% of the full footprint)DWS level (% of full supply volume)Our estimate by volume, from space

The chart needs JavaScript.

The clearest picture of each month. Area is not volume: a dam’s sides slope, so its area falls more slowly than its volume. DWS: the department’s weekly State of Dams. Contains modified Copernicus Sentinel data 2017-2026.

Middelkraal Dam: every clear look from space since 1984

Water area, each clear Landsat look (% of the full outline)DWS level (% of full supply volume)

The chart needs JavaScript.

Landsat imagery courtesy of the U.S. Geological Survey; Digital Earth Africa (CC BY 4.0), modified. Water Observations from Space for each Landsat look (Digital Earth Africa), counted inside its Digital Earth Africa Waterbodies outline (0.28 km², CC BY 4.0). A look counts when 90% of the outline was seen clearly: 487 of the 971 looks since 1984. 13 looks are set aside: likely muddy water read as land (below). Against Sentinel-2 on 46 days both saw it (at most 2 days apart): 15 of 46 within 5 points. Area is not volume. DWS: weekly State of Dams.
13 looks set aside: likely muddy water read as land

Water Observations from Space can read muddy water as land, so a look can show the dam nearly empty between looks that show it well filled. A look is set aside when its water is under half of both the looks before it and the looks after it (the median of up to three on each side, within 180 days), at least 20 points of full below them, no other look within 30 days reads it as low, and neither Sentinel-2 (within 10 days) nor DWS’s weekly level (within 7 days) shows the dam that low. Such a look is kept in the record but left out of the chart, the yearly water and the driest year.

  • 21 Nov 1990: 0.0% of full, against 33.7% in the looks before and 84.2% in the looks after
  • 16 Apr 1992: 0.0% of full, against 65.4% in the looks before and 20.7% in the looks after
  • 12 Aug 1994: 0.0% of full, against 58.3% in the looks before and 31.4% in the looks after
  • 27 Apr 1996: 0.0% of full, against 82.9% in the looks before and 68.0% in the looks after
  • 30 Jun 1996: 1.3% of full, against 68.0% in the looks before and 63.1% in the looks after
  • 1 Aug 1996: 17.2% of full, against 63.1% in the looks before and 67.7% in the looks after
  • 1 Jun 1997: 0.6% of full, against 79.3% in the looks before and 73.2% in the looks after
  • 23 Oct 1997: 0.0% of full, against 73.2% in the looks before and 78.0% in the looks after
  • 19 May 1998: 7.1% of full, against 79.3% in the looks before and 61.8% in the looks after
  • 15 Oct 2000: 0.0% of full, against 80.0% in the looks before and 80.6% in the looks after
  • 17 Jan 2003: 4.9% of full, against 50.8% in the looks before and 81.6% in the looks after
  • 27 Mar 2008: 0.0% of full, against 72.8% in the looks before and 74.8% in the looks after
  • 5 Jan 2016: 9.7% of full, against 66.7% in the looks before and 66.7% in the looks after
How we estimate % full from space, and how well it does for Middelkraal Dam
  • The curve. 41 clear months since 26 Apr 2021 paired with DWS’s % full for the same days (read between the weekly reports either side). A rising curve, % full = c × (% of the area)0.50, used between 71% and 90% of the area.
  • Tested on weeks it didn’t see. Each year left out in turn and estimated from the others: typical error 8.2 points, on average 0.1 low; nine in ten within 14 points, the band we show. Fitted only on the months before 8 May 2026 and tested on the 4 after: typical error 5.8 points, 4 of 4 within the band.
  • Radar for cloudy weeks. Sentinel-1 sees through cloud. From 1 Jan 2025, 90 of its 150 passes saw the whole footprint with a clear split between water and land (wind roughens water and makes it look like land; those passes are left out, and of the passes within three days of a report the largest water area is used). Scaled to Sentinel-2 on 10 days both saw the dam (× 1.21). Of DWS’s 46 weekly reports since then, 9 had a clear Sentinel-2 view within three days and 26 more only radar. On radar weeks our estimate was off DWS by 8.5 points on average (26 weeks); on Sentinel-2 weeks by 6.4.
  • The footprint. 0.3 km² from Digital Earth Africa’s 40 years of water against 0.3 km² in OpenStreetMap’s outline; they overlap 73% (shared area over the area of either); 0.0 km² is in the footprint only and 0.1 km² in the outline only.
  • Each reading. A month is used when at most 5% of the footprint is under cloud or has no data: in the 113 months used, a median of 100.0% of it was clear (the worst 95.0%).
  • Shoreline pixels. A 20 m pixel at the edge is partly water and partly land; it counts as water when the satellite’s classification says water, else not at all. In a typical month 20.7% of the water area is such edge pixels; half of that is the most the edge can move the area either way.
  • Water outside the footprint. None worth noting in the months used.

Area is turned into % full by volume on this dam’s own curve, so the estimate is ours, not DWS’s. It can be off where the water’s edge is hard to see (shallow, muddy or weedy margins) or at levels the curve hasn’t seen. Where DWS and the estimate differ by more than the band, the page says so, without saying which is right.

How we measure a dam from space

Every month since January 2017 the European Space Agency’s Sentinel-2 satellites have pictured it; we measure the water in the clearest picture of each month. Before that, the Landsat satellites, through Digital Earth Africa, give the water in every clear look since 1984 (487 looks) and a picture a year.

Over the year to the newest DWS report, DWS’s level moved up 4.0 points and the water area seen from space up 0.8 points of full.

The chart counts the pixels of Sentinel-2’s water class at 20 m in the clearest scene of each month (at most 5% of the dam hidden by cloud), inside the dam’s footprint from Digital Earth Africa’s 40 years of water (Digital Earth Africa Water Observations from Space, all-time summary 1984 to 2025 (CC BY 4.0), modified).

Measured from space by Unmapped Dataworks, not by DWS: the water is the satellites’ own water class (Sentinel-2 Level-2A scene classification; for Landsat, Digital Earth Africa’s Water Observations from Space) inside the dam’s footprint widened by 40 m: the ground Digital Earth Africa saw under water in at least 10% of its clear Landsat looks from 1984 to 2025, joined to the dam (0.3 km²; OpenStreetMap’s outline, which we use for the dam’s name and place, covers 0.3 km²; Landsat’s yearly water is still measured in OpenStreetMap’s outline), so shallow or muddy edges, floating plants and shadow can count wrong. 113 of 119 months had a clear view. Footprint: Digital Earth Africa Water Observations from Space, all-time summary 1984 to 2025 (CC BY 4.0), modified. The outline is OpenStreetMap’s (© OpenStreetMap contributors, ODbL; the outlines and how we match them). Pictures: Copernicus Sentinel data, free, full and open; Landsat courtesy of the U.S. Geological Survey, through Digital Earth Africa (CC BY 4.0). Radar: Contains modified Copernicus Sentinel data 2025-2026 (Sentinel-1 radar, processed by Digital Earth Africa).

Unmapped Dataworks