Upper Cisokan Pumped Storage · Biodiversity Management Plan

Where the line went

Since 2014 the working-zone boundary of the Upper Cisokan biodiversity and reforestation programme has been redrawn at least seven times. This page follows those changes file by file, then follows the work it would take to turn today's scattered tree cover into one continuous forest of 3,800 hectares.

BMP Advisor Team (2026) Updated September 2026 Watch 90 seconds · How the numbers add up · The five films · Get the data
3,814.6ha
2014 target map, the BMP's 3,800 ha
1,515–2,160ha
range of the buffer zone across three redraws
3,844.2ha
effective working zones, Dec 2025
3,581.9ha
WorkingZone26, Aug 2026
≈ 350ha
still to find to reach 3,800 ha
Start here · Video · 90 seconds · narrated

The whole argument in ninety seconds

If you watch only one of these, watch this one. 2,445 ha of forest already stands inside the working zones — but in 183 separate blocks. Planting 1,214 ha in the order that joins them, rather than the order that is convenient, leaves 3,529 ha in a single continuous block by the end of 2028, and 271 ha still to find beyond the boundary.

Cut from part 5 and narrated afresh; every figure in it is produced by the same scripts, so the short and the long film cannot disagree. The five longer films below take the same material apart step by step.

2014201720212022 2024–25Dec 25Aug 26
Map of the Upper Cisokan working-zone boundaries for the current chapter
UTM 48S · north up
Before we start

One target, many lines

The Biodiversity Management Plan commits the project to 3,800 hectares of wooded cover around the two reservoirs of the Upper Cisokan pumped-storage scheme in West Java. Where that cover is counted depends on a boundary — and the boundary has been redrawn by different teams, for different purposes, for twelve years.

Scroll to follow it. The map keeps the reservoir high-water line fixed as a reference; each chapter swaps in the boundary of its year.

Oct 2014BMP Disclosure · eForest

Four zones designed; the target drawn as one polygon

The BMP Disclosure, prepared by eForest, set out four working zones. The biodiversity important areas (BIAs) and 21 corridors were printed on maps. The buffer was defined as a rule — 500 m from the reservoir's high-water line. And the reforestation target was drawn as a single polygon. That polygon is the 3,800 ha the BMP commits to.

3,814.6hatarget map
3,231.9haof it: buffer-reservoir parts
582.8haof it: restoration-target parts
21corridors, printed on maps only

BIA outlines shown here were reconstructed in 2026 from the Disclosure maps; 11 of the 15 close inside their map frames.

Jun 2017PPSDAL · Universitas Padjadjaran

The oldest vectors that survive

No digital boundary files from 2014 survive. The oldest are from June 2017: 15 BIAs forming Working Zone 1. Two things changed on the way from the printed maps.

450.9ha15 BIAs — Working Zone 1
15.5 → 26.9BIA-05 absorbed its corridor (ha)
21 → 9corridors dissolved into shapes — individual corridors can no longer be reported
2021–2025Working Zone 3 · buffer

The buffer moved the most

Working Zone 3 was redrawn three times in four years, each time for a different purpose. Its area ranged from 1,515 to 2,160 ha.

Feb 2021 · cut against the BIAs and corridors1,515.3 ha
Mar 2022 · plain 500 m buffer, forest-use permit (IPPKH) package2,127.8 ha
2024–25 · recut along Perhutani compartment lines2,160.2 ha
Mar 2022Working Zone 3 · permit package

For the forest-use permit application the buffer went back to its 2014 rule — a plain 500 m band from the high-water line — and grew to 2,127.8 ha. The dashed line is the 2021 version.

2024–25Working Zone 3 · compartments

Recut along Perhutani's forest compartment (anak petak) lines so that it can be managed on the ground: 2,160.2 ha, the largest version.

Dec 2025Consolidated working zones

Net of the permit area and the reservoir

In December 2025 the zones were consolidated. Together they span 4,428.3 ha. After removing the IPPKH forest-use permit area and the reservoir, the effective working zones cover 3,844.2 ha — and, unlike earlier versions, they no longer overlap, so zone areas add up.

329.8haWZ1 · biodiversity important areas
231.6haWZ2 · corridors
1,124.2haWZ3 · buffer
2,158.6haWZ4 · reforestation
Aug 2026WorkingZone26

The newest boundary, over the 2014 map

The latest update covers 3,581.9 ha (WZ1 252.6 · WZ2 180.1 · WZ3 974.7 · WZ4 2,174.5). Laid over the 2014 target map, it leaves part of that map outside.

416.6haof the 2014 target map outside WorkingZone26
141.7haexplained by the permit area or the reservoir
274.9hanot the permit area or reservoir — mostly compartments planted in 2023–2024 that the new boundary left out
DraftReconciled WorkingZone26

Most of the gap was planted land

WorkingZone26 keeps 96 % of the 2025 planting but only about 1 % of the planting from 2023 and 2024 — it was drawn around the work still to do. A reconciled draft puts the left-out land back, piece by piece, with a reason recorded for each change.

234.5hacompartments planted 2023–2025, restored
19.5haforest or plantable land, restored
11.1hathin edge slivers, snapped to the agreed line
5.9haon land PLN acquired — to confirm

With the draft, 2023–2024 planting inside the boundary rises from 0.9 % to 88.7 %, and the boundary grows from 3,581.9 to 3,850.9 ha.

A draft awaiting agreement. Part of the restored land overlaps the construction plan and needs confirming; until the draft is adopted, figures on this page use the December 2025 effective zones.

≈ 350 haWhat it means for the target

The real gap to 3,800 ha

Inside the 2014 target map, net of the permit area and the reservoir, 3,644.8 ha remain. A land-cover map built from twelve PlanetScope dates (March 2024 – August 2026) shows that 3,451.8 ha of it is already wooded or plantable.

Gap to 3,800 ha: 348.2 ha — to be found beyond the current zones.

The aide-memoire put this gap at 600 ha, by subtracting the whole of the land the project takes — 448.7 ha of inundation and 178.9 ha of facilities, 627.6 ha together, or 630.9 ha measured from the layers. That subtraction is wrong, and eForest's own 2014 map shows why.

The larger of the target's two parts is named, in eForest's file, Buffer Reservoir — a buffer drawn around the lake. The measurement agrees with the name: of the reservoir's 378.7 ha, only 18.8 ha falls inside the target, and that 18.8 ha is 131 separate slivers averaging 29 m wide — two draughtsmen disagreeing about a shoreline, not land. The lake was never inside the 3,800 ha, so it cannot be taken out of it.

What the project genuinely took from the target is 156.1 ha of facilities, and that is compact ground averaging 80 m across. With the 18.8 ha of shoreline slivers, 169.9 ha in total — which is exactly why 3,644.8 ha of the 3,814.6 ha target still remains in the bar above. That 170 ha was promised as forest and is now built on, so it has to be replaced somewhere else; but it is 170 ha, not 628.

Within the working zones, 1,198.3 ha of plantable land — sparse vegetation and ladang — carries no planting record at all; counting only records from 2016 onward it is 1,522.8 ha. Read both as the upper end until they are checked in the field.

Part 2What it takes

One target, 88 pieces of forest

The boundary settles where the forest is counted. This second half asks what it takes to make that forest one continuous 3,800 ha block — the form that matters for wildlife moving between the reservoirs.

Start with what stands today inside the December 2025 zones: 1,366 ha is already wooded, but broken into 88 blocks of a hectare or more.

1,366.1hawooded land inside the zones
88separate blocks of 1 ha or more
169.7hathe largest single block

Blocks are joined across gaps of up to 30 m — tracks and streams — but never across the reservoir.

Works 1–2Protect · Nurture

The cheapest hectares are already planted

Protect the standing forest, and look after the planting already in the ground: 361.9 ha planted in 2023–2025, plus 706.1 ha that Perhutani's compartment records show as planted earlier. Most of it still reads as open land from above, because young trees have no canopy yet.

1,067.8haplanting to nurture to canopy (scenario B)
1,048.8halargest block once that canopy closes

Under scenario C — counting only records from 2016 onward — 743.4 ha is nurtured and the block reaches 850.1 ha. See "how the numbers add up" below.

Work 3Plant

Plant the open ground

1,198.3 ha of sparse vegetation and ladang has no planting record at all. 885.9 ha of it lies on slopes of 45 % or less — the ground to plant first. About 155 ha of that is ladang still being farmed, so agreements with the farmers come before seedlings.

885.9haplantable, slope 45 % or less
155.1haof it ladang still farmed
2,834.5halargest block after planting it

Scenario C: 1,138.0 ha to plant on gentle slopes, 192.7 ha of it ladang.

Work 4Regenerate

Let the steep slopes come back on their own

312.4 ha lies above 45 % slope, where planting in rows is impractical and the soil is easily lost. Assisted natural regeneration and enrichment planting suit that ground better.

312.4haplantable land above 45 % slope (384.8 ha under C)
3,632 haThe ceiling

What planting cannot reach

With every plantable hectare planted, the zones hold 3,632 ha of forest, and 3,492 ha of it forms a single block. The rest is paddy, settlements and water — land that will not become forest.

3,492.0hathe main block, once works 1–4 are done
211.9hapaddy, settlements, water and other land
167.6hashort of 3,800 ha, even fully planted
Work 5Extend

The last stretch lies outside the line

To reach 3,800 ha in one piece, the main block needs about 308 ha more. Within 500 m of the zones there is plenty of candidate land: about 650 ha of forest already touches the main block, and about 1,700 ha of plantable land would join it.

≈ 308hastill needed for one 3,800 ha forest
647.3haexisting forest within 500 m that touches it
1,699.4haplantable land within 500 m that would join it

Who holds that land has not been verified.

FirstWhere to start

Tier 1 comes first

Taking the 1,198.3 ha with no planting record and setting aside 38.7 ha inside the contractor's construction plan leaves 1,159.6 ha to plant or regenerate. The draft order of work puts the biodiversity areas, the corridors and the ground next to existing forest first — that is where canopy closes fastest and blocks join soonest.

317.9hatier 1 · now — BIA, corridor or gap-fill beside existing canopy (557.4 ha under C)
827.0hatier 2 · next, open ground (895.6 ha under C)
14.4hatier 3 · after farmer agreements (24.8 ha under C)
38.7hakept clear — inside the construction plan

Draft for discussion. Every unit is a Perhutani anak petak from the May 2026 compartment map, which covers 98 % of the zones — so each line can carry a compartment number.

NextDecisions to take

So every report measures the same line

  1. Adopt the reconciled boundary once its open points are settled.
  2. Report the gap to 3,800 ha one way — on the 2014 map or as one continuous forest.
  3. Confirm who holds the land, with full Perhutani compartment maps.
  4. Start planting with tier 1, and find the ≈ 308 ha still needed beyond the zones.
How the numbers add up

Every figure on this page, in five steps

All of it comes from two inputs: the land-cover map and the December 2025 working zones. Nothing below is an estimate of future planting — it is what the ground looks like today, sorted.

1 · Sort every hectare of the zones

Kind of landArea (ha)What it is
Already wooded1,366.1production forest 1,236.9 · natural 123.1 · agroforest 6.1
Can be planted2,266.3sparse vegetation 1,974.9 · ladang 291.4
Never forest211.9paddy 132.6 · settlements 70.0 · water and other 9.3
Working zones3,844.30.1 ha is rounding

2 · Split the plantable land by who has already planted it

Plantable landArea (ha)
Planted by the project, 2023–2025361.9
Planted earlier — Perhutani compartment records (scenario B)706.1
No planting record at all — "still to plant"1,198.3
Total plantable2,266.3

This split uses Perhutani's compartment map of May 2026, which covers 98 % of the zones. How much counts as already planted depends on how far back the records are trusted — which is the one real choice in the whole calculation:

ScenarioRecords accepted as plantedCounted as plantedStill to plant
B · headlineevery Perhutani record, any year706.11,198.3
C · cautiousonly records from 2016 onward, where canopy can still be pending381.61,522.8

The 324.5 ha between them carries a planting record from before 2016 but still reads as open ground on 2024–2026 imagery. Across the whole compartment map only 47 % of recorded planting shows canopy today — 41 % for pine, 36 % for rimba campur — so that land was probably harvested, failed or encroached rather than being too young. A field check settles it compartment by compartment.

3 · The most forest the zones can ever hold

1,366.1 wooded + 2,266.3 plantable = 3,632.4 ha. The other 211.9 ha is paddy and villages. So even after planting everything: 3,800 − 3,632.4 = 167.6 ha short.

4 · From area to one continuous forest

Adding up hectares is not the same as having them in one piece. Counting blocks — pieces joined when the gap between them is 30 m or less, never across the reservoir:

After each workLargest single block (ha)
Today169.7
Nurture the 1,067.8 ha already planted or on record until canopy closes1,048.8
Plant the gentle ground (885.9 ha)2,834.5
Regenerate the steep ground (312.4 ha)3,492.0

3,492 is less than the 3,632 in step 3 because about 140 ha stays in separate small blocks. So for 3,800 ha in one piece: 3,800 − 3,492 = 308 ha must come from outside the zones.

5 · Why you see two different gaps — 350 and 308

They measure different ground:

Area measuredIts size (ha)Wooded + plantable in itGap to 3,800 ha
The 2014 BMP target map, minus permit area and reservoir3,644.83,451.8348.2 ≈ 350
The December 2025 working zones3,844.23,632.4167.6 — or 308 as one block

The 2014 map lies inside the zones, so the zones take in extra ground the 2014 map never covered: that is the 180 ha difference between 3,451.8 and 3,632.4. Both figures are right; they answer different questions, which is why picking one to report is a decision the project still has to take.

The three progress percentages

All three use the same 489.8 ha of planting delivered in 2023–2025, divided by different things:

  • 489.8 ÷ 3,800 = 12.9 % against the BMP target;
  • 489.8 ÷ (489.8 + 1,198.3) = 29.0 % against delivery plus what is left (24.3 % under C);
  • 361.9 ÷ (361.9 + 1,198.3) = 23.2 % using only the part of the planting the map can see (19.2 % under C).

And the planting plan's net figure

1,198.3 still to plant = 885.9 on gentle slopes + 312.4 on steep slopes. Take off 38.7 ha inside the construction plan and 1,159.6 ha is left: tier 1 317.9 + tier 2 827.0 + tier 3 14.4. Under scenario C the same arithmetic gives 1,478.4 ha net: 557.4 + 895.6 + 24.8.

Part 1 · Video · 5 min · narrated

The same story, told in five minutes

A narrated animation of the chronology above, for presentations and for readers who prefer to watch. It is the first of five longer films; the others are listed at the foot of this page.

Part 2 · Video · 6 min · narrated

One forest, 3,800 ha: the work it takes

The animated version of part 2 above: how the zones sort into wooded, plantable and never-forest land, then five works — protect, nurture, plant, regenerate, extend — how the largest continuous block of forest grows with each one, and where planting should start.

Largest continuous forest block after each work
WorkWhat it involvesArea of the work (ha)Largest block (ha)
TodayWooded land inside the working zones, in 88 blocks of 1 ha or more1,366.1169.7
1–2 · Protect, nurtureKeep standing forest; care for 361.9 ha planted in 2023–2025 and 706.1 ha on Perhutani's records until the canopy closes1,067.81,048.8
3 · PlantOpen ground on slopes of 45 % or less; about 155 ha is ladang still farmed, so agreements with farmers come first885.92,834.5
4 · RegeneratePlantable land on slopes above 45 %: assisted natural regeneration and enrichment planting312.43,492.0
5 · ExtendSecure about 308 ha outside the zones; within 500 m, about 650 ha of forest and 1,700 ha of plantable land join the main block≈ 3083,800

212 ha inside the zones is paddy, settlements, water or other land that will not become forest, so the zones can hold at most 3,632 ha. Blocks are joined across gaps of up to 30 m but never across the reservoir. The ≈ 350 ha gap earlier on this page is measured on the 2014 target map; the ≈ 308 ha here is measured on continuous forest across all four working zones. Land status beyond the zones has not been verified.

Which land is worth most

Under scenario B, the 1,198.3 ha with no planting record less the 38.7 ha inside the construction plan leaves 1,159.6 ha to plant or regenerate — 1,478.4 ha under scenario C. Each planning unit falls into one of three tiers by value. The tiers rank the land; the planting schedule below sets the season for each parcel:

317.9haTier 1 · now — biodiversity areas, corridors and open land beside existing forest (557.4 ha under scenario C)
827.0haTier 2 · next — open ground further from forest (895.6 ha under C)
14.4haTier 3 · after agreements with farmers still using the land (24.8 ha under C)

Draft for discussion. Every planning unit is a Perhutani anak petak from the May 2026 compartment map, so each line can carry a compartment number — where one is recorded. About two thirds of the compartments the plan touches have no anak petak in that delivery, so the proposal issued to Perhutani identifies those parcels by zone, land cover and coordinate instead, with the compartment number left for Perhutani to confirm.

Part 3 · Draft in consultation

The 2026–2028 planting schedule

The tiers above say which land is worth most. This says in which season to plant it. The ordering rule is connectivity, not convenience: parcels within 30 m of two or more separate blocks are planted first, because each of those hectares merges whole blocks and so buys far more continuous forest than its own area.

Three seasons, 1,214.1 ha, 470 parcels
SeasonArea (ha)ParcelsLargest block after (ha)Blocks remaining
Today——1,058.0183
2026499.7321,868.097
2027347.562,810.727
2028366.94323,529.319

2027 is the season to protect in any budget: six parcels, 347.5 ha, and the main block grows by 943 ha. The 2028 column is 432 small parcels filling gaps that only close once their neighbours are planted.

What is being planted, and where

1,058.4hasparse vegetation
155.7haladang still farmed — agreements with the farmer come first
897.1hagentle ground — rows of seedlings
317.0hasteeper than 45 % — assisted natural regeneration, not planting

By working zone: WZ4 703.3 ha, WZ3 432.9 ha, WZ1 43.9 ha, WZ2 34.0 ha.

Two parts, because a third of it cannot yet be named

Every parcel is proposed against a Perhutani anak petak where one exists. It often does not: 75 polygons in the May 2026 compartment delivery carry no attributes at all — no anak petak, and no KPH, BKPH, RPH or bagian hutan either, so there is no higher level to fall back on.

Part A includes 43 parcels whose numbers were recovered from Perhutani's fully attributed September 2025 delivery. That file is a year older than the current one and boundaries may have moved, so those numbers are carried as candidates for Perhutani to confirm, in their own column, never merged with the confirmed ones. Part B parcels are identified instead by parcel ID, zone, land cover and coordinate, so the schedule does not wait on the missing records.

Three limits, stated plainly

405ha/yrthe rate this schedule needs. The rate achieved in 2023–2025 was about 200 ha/yr — at that rate it is a six-year programme
271hashort of 3,800 ha contiguous even if every hectare is planted; the rest must come from beyond the zone boundary
155.7hais ladang still being farmed — nothing is planted there until there is an agreement

Status: draft, in consultation. This schedule has been sent to Perhutani for correction and agreement, because every hectare of it is on their land. It has not been submitted to PLN. Perhutani's response — the missing compartment numbers, the confirmed planting records, and their view on the steep-slope and ladang parcels — will be incorporated before it goes further. Figures here may change as a result.

Parts 3 to 5 · Video · narrated

Three further films

Part 1 covers how the boundary moved and part 2 the five works. These three go further: what the 3,800 ha actually counts, how every figure was produced, and the planting schedule that follows from them.

Part 3 · What counts toward 3,800 ha · 6 min

The working zones, the land the project takes, and the two definitions that decide whether the target is met: whether the 628 ha of reservoir and permit area comes off the target — it does not, as the section above shows — and whether pine counts as forest. Of the 1,366.1 ha standing today, 537.4 ha is Pinus and only 34.6 ha is Rimba Alam, true natural forest. The pine question is now settled: it counts. FAO and the EUDR-aligned typology both class planted forest as forest; the BMP's own monitoring method lists "monocultural planted forest" as a forest class, separate from non-woodland (p.132); and ForTy — an independent 10 m global map validated on 9,000+ random sites — classes 89.8 % of those hectares as forest, 59.7 % of them as naturally regenerating rather than plantation. The one clause that cuts the other way bars monocultures inside the BIAs, which touches 45.7 ha of the pine; there we propose native enrichment, as the BMP itself prescribes.

Part 4 · The same numbers, made in QGIS · 11 min

Every frame is a real QGIS window, captured as the work ran, using standard Processing algorithms. Fourteen steps from the raw raster to a CSV, including the traps: the zones arrive in geographic coordinates where area reads as zero hectares, the permit boundary has a self-intersecting ring, and the two Perhutani deliveries overlap. The chain reproduces Pinus 537.4 ha — the same figure the separate Python analysis gives.

Part 5 · The 2026–2028 planting schedule · 7 min

The same hectares planted in a different order give a different forest. Parcels are ranked by what they connect, not by convenience: 499.7 ha in 2026 across 32 parcels, 347.5 ha in 2027 across just six, and 366.9 ha in 2028. The largest continuous block grows from 1,058 ha today to 3,529 ha — 271 ha short of 3,800, which has to come from beyond the zone boundary.

Part 5 supersedes the tier table above as the basis for scheduling: the tiers say which land is most valuable, the schedule says in which season to plant it. The 2026 figure of 499.7 ha is one season of a 1,214.1 ha programme, not the whole of it — the April 2026 workbook's 1,060.6 ha is the comparable total, and differs mainly because it excludes 376.8 ha of steep ground that this schedule retains for assisted regeneration.

Data

Every boundary on this page

Each layer is a GeoPackage in WGS 84 / UTM zone 48S (EPSG:32748). Areas are dissolved, planimetric hectares; the map above draws simplified outlines. Files 1–16 are the boundary versions; 17–22 add the Perhutani compartment map and this year's analysis — the reconciled boundary, the planting plan and the field-check list. Files 23–25 are the land-cover map itself: a 3 m raster with its legend, the same map as polygons clipped to the zones, and the seven-class version with dense vegetation undivided, which is the level the planting figures use and the one whose accuracy is quoted with them. Files marked Draft are this year's analysis by the BMP Advisor Team and have not been agreed with PLN or Perhutani — they are published so the reasoning can be checked, not as approved project data.

PeriodLayerOriginStatusFeaturesArea (ha)File
How the figures were made

Method notes

One frame for every version

All boundaries were reprojected to UTM zone 48S and drawn in a shared frame. Areas are computed on the full-resolution geometry after dissolving overlaps within each layer; outlines on the map are simplified for display.

"Effective" zones

The December 2025 zones exclude the IPPKH forest-use permit area and the reservoir high-water line. The four effective zones partition 3,844.2 ha with no overlap.

Outside the 2026 boundary

The 2014 target map minus WorkingZone26, split by whether the excluded ground falls inside the permit area or reservoir (explained) or not (unexplained).

Wooded, plantable, still to plant

Land cover comes from an object-based classification of twelve PlanetScope dates, March 2024 – August 2026, at 3 m. Plantable = sparse vegetation and ladang. "Still to plant" deducts planting recorded for 2023–2025 and earlier Perhutani planting records. The wooded-against-plantable distinction these figures rest on is cross-validated at 95.7 % overall accuracy, Kappa 0.93; the map scores 93.1 % and Kappa 0.92 across its seven classes. Splitting the wooded class into natural, production and agroforest is less certain — 87.9 % and Kappa 0.86, with natural forest the weakest class — but that split moves no land between wooded and plantable, so it changes none of the figures here. All of these are cross-validated on our own labelled samples, an upper bound rather than an independent field validation.

Reconciled boundary (draft)

Each piece of the 2014 target map left outside WorkingZone26, other than the permit area and reservoir, was classified by its planting record, holding, land cover and shape. Planted compartments, forest land and slivers are restored to their December 2025 zone; every change is logged with its reason and awaits sign-off.

Where to start

Planting land net of the construction plan is split into planning units — Perhutani anak petak from the May 2026 compartment map, which covers 98 % of the zones — and ordered in three tiers. Perhutani's own zone labels agree with the working zones exactly.

Two scenarios

How much land counts as already planted depends on how far back Perhutani's planting records are trusted: B accepts every record, C only those from 2016 onward, where canopy can still be pending. The difference is land with an older record that the imagery reads as open.