South Sumatra, Indonesia · 124 coal mining permits · 2016 to 2025
Ten years of land being opened
What satellite pictures show when every
patch of mining land is traced by hand, one year at a time.
Area
124 permits, 21 regions
Years
2016 to 2025, once a year
Source
30 m satellite images, traced by hand
Controls
← → move · N notes
Speaker notesStart with a question, not a number: "If we could look down at
the mining areas once a year for ten years, what would we see?" Say clearly that everything here
comes from satellite pictures traced by a person, not from company reports and not from an
automatic count. Page 13 carries the method and the limits, so point at it if anyone asks early.
01How this was made
The satellite takes the picture, a person draws the line
One picture every year
Landsat and Sentinel satellites record the same ground every year at 30 metres. One pixel is
about the size of a basketball court.
Drawn by hand
The edge of every patch of open ground was drawn on screen by a person, then checked against
sharper 3 metre images and Google base maps.
Checked again
12,584 hectares had been drawn twice and were removed first, so no hectare is counted twice.
2015 was dropped because the images that year were too cloudy to read. Every number
here starts in 2016.
Speaker notesThis slide answers "where do the numbers come from?" before
anyone asks. Make two points: a human eye decided what counts as open ground, and each year was
checked against the year before and the year after. The tooling behind that tracing is on page 13.
02One mine, ten years
What the record looks like when it moves
Land area, hectares
ever opened still open recovered
Years the ground has been open
1 yr
open now · 5 years or more
Years since the plants came back
1 yr
recovered · only ground that was dug first
1 km
A separate case study: one open-cut coal mine, same method, same ten years. The chart
and both maps redraw for the year on screen.
Speaker notesLet it run one full loop before saying anything. Then pause on
2016 and step forward by hand: the pit spreads in every direction, and the green map shows recovery
appearing only on the edges. This mine is not in South Sumatra, it is a second site processed the
same way, which is the point: the method travels.
03Summary
Four numbers worth remembering
3.6×
Open land in 2025 compared with 2016
6,291 → 22,373 ha
29,178
Hectares opened at least once during the ten years
about 40,000 football fields
6,805
Hectares now covered by plants again
23% of all land ever opened
41%
Open land that sits within 1 km of housing
second source says 44%
If the audience remembers only one slide, make it this one. The rest of the deck is
the evidence behind these four numbers.
Speaker notesRead all four slowly, one sentence each. Be careful with the
second: 29,178 ha is land opened at least once at some point in the ten years, which is not
the same as land that stayed open for ten years. That second figure is 2,464 ha, and it appears on
the duration slide. Mixing them up is the easiest mistake in the deck.
04Opened and grown back, year by year
Recovery has never caught up with clearing
Red bars are land opened that year. Green bars are land that turned green again. The
black line is the difference. In nine years in a row, green never rose above red.
Speaker notesPoint at 2022: 4,828 hectares opened in a single year, the
biggest jump in the record. Then point at the net line, which never touches zero. 2020 came closest
to a balance, 1,694 opened against 1,284 grown back. In total: 25,589 ha opened, 9,363 ha grown back.
05The real footprint
Every hectare counted only once
The grey line is all the land ever disturbed. The blue band is what is still open today.
The green band is what has grown back. The two bands always add up to the grey line.
Speaker notesExplain why this chart exists: adding up the yearly bars would
count the same ground twice when it is opened, planted, then opened again. Here each hectare is
counted once. What to read: the gap between the grey line and the green band widens every year.
That gap is the reclamation backlog, and it is not shrinking.
06How long
Some ground has stayed bare since the very first year
Left: how long a patch of ground has stayed open without a break. Right: how long the
new plant cover has lasted. The "10+" bar on the left is 2,464 hectares that have been open for
the whole ten years.
Speaker notesSeparate two things. Bars at one and two years are a mine that is
working, which is normal. Bars at eight, nine and ten years are a different question: 7,541 ha, one
third of the land still open, has been open for five years or more. The chart on the right is the
good news, most recovered ground is young, so replanting does happen, it is just slower.
07Quality of recovery
Where the land did recover, plants really are growing
Each box shows the spread of the greenness score (NDVI) for recovered ground at that
age. The red line at 0.33 is the official mark for healthy plant cover; the dotted line at 0.66
is very healthy.
Speaker notesHow to read it: age zero is the year the ground was first
recorded as green again. The line climbs steadily until age four or five, then flattens near 0.74.
Say the limit yourself, before anyone asks: this measures greenness, not what kind of plant is
growing. Page 13 spells out why that matters for a rehabilitation report.
08Who lives nearby
This is not empty land far from anyone
Left: public places that stand inside a mining permit area, including 43 schools, 45
places of worship and 232 village points. Right: how many of them sit less than 2 km from a pit.
Speaker notesThe easiest number to remember: 84 of the 124 permits have a
mapped village inside their boundary, and four schools sit less than 500 metres from open pit
ground. Be honest about the source: this map is made by volunteers and is only about 30% complete,
so every distance here is an upper limit. The real distance can be shorter, never longer.
09Rivers
Seventy seven permits sit right on a main river
Of the 124 permits, 77 have a main river within 500 m of their boundary. Of the land
open in 2025, 2,498 hectares sit less than 500 m from a river and 1,883 hectares between 500 m
and 1,000 m.
Speaker notesTwo things are easy to mix up: how far the permit area is from a
river (left chart) and how far the pit itself is (middle and right). Many permits are crossed by a
river, but most digging stays away from it, with 11,678 ha more than 2 km out. The part to watch is
the 2,498 ha right beside the water.
10Which rivers
Two rivers carry two thirds of the load
Air Enim (8,371 ha) and Air Lematang (7,128 ha) together carry about 69% of all land
open in 2025. The colours show how far each pit sits from that river.
Speaker notesKeep "largest" and "most exposed" apart. Air Lematang has a large
area of open land, but most of it is far from the water. The rivers with the thickest red segment,
meaning pits closer than 500 m, are Air Enim and Sungai Putih. Total hectares is not a measure of
risk. Distance is.
11The point
Land is opened almost three times faster than it is brought back, and much of it
lies in somebody's back yard.
25,589 ha
opened between 2017 and 2025
9,363 ha
grown back over the same years
1 : 2.7
ratio of land recovered to land opened
A study of 124 coal mining permits in South Sumatra, 2016 to
2025, using 30 m Harmonized Landsat and Sentinel images. Every number can be traced back to the
table it came from. Read the next slide before quoting any of them.
Speaker notesClose on the large sentence, then pause. If you want one call to
action: the question is not whether mines may open land, it is how fast that land is given back,
and that can now be measured every year by anyone. Then move straight to the disclaimer slide
rather than taking questions first.
12Disclaimer and method
What these numbers are, and what they are not
Nothing here has been validated on the ground. There has been no field visit,
no survey control and no comparison against any company reclamation report or regulator record.
Everything is measured from imagery, so it is an independent observation, not an audit finding.
The polygons are model-assisted, not purely manual. Digitising was done in
QGIS with the Geo-SAM plugin running the SAM 2.1 Large segmentation model:
the operator clicks inside a feature, the model proposes the boundary, and the operator accepts,
edits or rejects it. Boundaries therefore follow image texture and the model's idea of an edge,
which is not the same as a surveyed line.
The false colour is a stretch, not a measurement. Bands B8A, B4 and B3 are
mapped to red, green and blue, each scaled linearly and clipped: near infrared 0 to 0.5,
red 0 to 0.2, green 0 to 0.2, then rescaled to 0 to 255. Red pixels are simply high near
infrared, which is what living plants reflect. Change the stretch and the picture
changes; the underlying reflectance does not.
Green is not the same as rehabilitated. NDVI measures how much living plant
matter is present. A high value can be grass, weeds, a cover crop or regrowth of the wrong
species, and a mine rehabilitation plan usually asks for far more: agreed species, soil profile,
slope stability, drainage, land capability and a defined completion criterion. This work can say
that something green is growing and how vigorous it is. It cannot say that a rehabilitation
obligation has been met.
Areas depend on the boundary they are measured in, and nothing here is causal.
Mines sit where the coal is, not in random places, so two things moving together is not proof
that one caused the other.
Speaker notesDo not rush this slide, and do not apologise for it. Saying
plainly what the work cannot support is what makes the rest credible. If someone pushes on the
fourth point, the honest line is: greenness is a screening tool that tells you where to look, and
the rehabilitation judgement still needs the plan, the species list and a site visit.