Mondulkiri, in full

Verified Area and elevation are from the Ministry of Interior’s Information Book on Capitals and Provinces of Cambodia (2024) and the provincial GIS contour layer.

Land and terrain

New here? There is a three-minute walkthrough of this hub — captioned, with a full transcript — covering the source badges, the filter, the question box and the map layers.

Administrative Breakdown

Administrative division counts as of 2024 · Source: Ministry of Interior records, via IntoCambodia.org

Official Provincial Profile

Straight from the Ministry of Interior's Information Book on Capitals and Provinces of Cambodia (15 March 2024) — the government's own reference profile for Mondulkiri. Where these figures differ from other sources on this site, both are shown rather than quietly reconciled.

District & Commune Explorer

Population by commune across the 1998, 2008 and 2019 censuses (National Institute of Statistics), plus land use and economic potential for each district. Two official 2019 population totals exist for this province — the NIS census count and a separate Commune Database/PDoP count — and they don't quite match; both are shown where relevant rather than picking one silently.

Second profile: Andoung Kraloeng

A Bunong community in Sen Monorom commune, Ou Reang district — 23 km from the provincial town, reached from National Road 76 beside the Keo Seima Wildlife Sanctuary office. Documented for the same reason as Lam Meh: it holds a communal land title under LASED III.

Population by commune: NIS General Population Census 1998 / 2019, Final Results, via citypopulation.de. District area, land-use percentages and protected-area figures: PCLMUP of Mondul Kiri Province, Technical Report of the Mondul Kiri Provincial Spatial Plan 2020–2040 (WWF, 2021). A full list of all 92 village (phum) names was not located in official, machine-readable form during this research pass — the National Committee for Sub-National Democratic Development's commune database is the most likely source and is flagged here as a known gap rather than filled with invented names.

Demographics & Bunong Heritage

Four figures for 2019, and a fifth that replaces them all

This Atlas already prints two counts for 2019 and says they disagree. That was right and it was incomplete. The 2019 census final report, published in September 2025, carries three figures of its own, and this page was already drawing one of them on a chart without saying it was the same census.

So the disagreement this Atlas flagged was real but mostly definitional: three of the four come from one census at different stages and on different definitions, and only the commune database is genuinely a different exercise. That is a better answer than “two sources disagree”, and it is only available because the final report was published six years after the count.

And the province has stopped filling up

This is the single most consequential number added to this Atlas. Every other page here describes a province filling up: migration in, concessions granted, forest converted, a growth-and-tenure section built on it. Between 2019 and 2024 that stopped. Growth fell from 3.7% a year to 0.3% — from among the fastest-growing provinces in Cambodia to the slowest. The survey also records the largest average household in the country at 4.9 people and the highest sex ratio at 105.2 men per 100 women, both consistent with a place that had been receiving working-age migrants and has stopped.

Nothing published explains why, and this Atlas is not going to invent a reason. It is worth setting beside three things already on this page: the collapse in domestic eco-tourism zone visitors after 2024, the province's unresolved land constraint, and the fact that its 2025 budget carried no capital spending at all.

The projection was wrong, and by how much

The National Institute of Statistics published a provincial projection to 2033 in July 2023. For 2024 it gives Mondulkiri 102,791. The survey measured 93,657. The projection overshot by 9,134 people, or 9.8% — because it was built on the 2008–2019 growth rate that has since collapsed. Its figure for 2026 is 105,636, and it is the only 2026 number that exists, so it is printed here with the warning attached rather than left out: treat it as an upper bound, not an estimate. Verified

One further caveat for anyone using the district table in that projection: it lists four districts for Mondulkiri and omits Saen Meaknourum entirely. The four do not sum to a provincial total, and a reader who adds them will be short by the municipality.

Not only Bunong

Mondulkiri is usually described as a Bunong province, and by weight of numbers that is right. But the NCDD's district data book records at least eight distinct indigenous peoples living in Kaoh Nheaek district alone. Flattening them into one category erases communities of a few dozen families.

Who actually holds title to the land

Cambodia's 2001 Land Law and the 2009 Sub-Decree on Indigenous Communal Land Registration allow an indigenous community to hold land collectively — residential, agricultural, spirit forest and burial forest together, in one title that cannot be sold off parcel by parcel. It is the single strongest legal protection a Bunong village has. Nationally, fewer than forty communities have ever completed the process. The provincial GIS shows how few that is on the ground here.

How to read this table. A "parcel" is one mapped polygon inside a community's title, not one family's plot — a single community title is registered as many parcels covering different land uses. Parcel counts therefore show how finely a title was surveyed, not how much land it covers. Village geocodes are the standard Cambodian eight-digit code (province 11 · district · commune · village); district names are resolved from the 1101–1105 district codes. Andoung Kra Leung carries a blank geocode in the source table, so its district is left unstated here rather than guessed.

Source: Mondulkiri provincial GIS database, FeatureServer layer 40. Parcel counts queried per community and cross-checked — 63 + 61 + 60 + 41 + 27 + 22 + 13 = 287, matching the service's own count of communal records exactly. District codes per NCDD/Cambodian standard geocoding.

What the forest is worth

In 2016 WWF put an economic value on Mondulkiri's forest — carbon storage, non-timber forest products and water yield — and modelled three futures to 2030. It is the single most useful document for anyone arguing about land use here, because it converts "protect the forest" from a sentiment into a number.

Forty-one years of forest, and what the total hides

Verified Source: Hectare forest cover by province in Cambodia (1973–2014), Open Development Cambodia. Zonal statistics in ArcGIS over ODC's forest-cover raster, against Economic Census 2011 provincial boundaries. Licensed CC BY 4.0 — attribution, no share-alike.

This Atlas has carried two points: 1999 and 2019, from the Spatial Plan, with 303,464 ha of change between them. A start and an end and nothing in the middle. ODC derived forest cover for every province from satellite imagery at six dates and split it into dense and mixed. Mondulkiri now has a series.

A reader told only that forest fell a fifth would miss the important part. The total is down 19.9%, which is serious. Underneath it, two thirds of the dense forest is gone and mixed forest has grown by more than three quarters. The province did not simply lose forest; it thinned it.

Where these figures stop being trustworthy

Between 2004 and 2009 dense forest falls 503,953 ha while mixed rises 364,597. Then between 2009 and 2014 dense rises 50,646 ha while mixed falls 141,385. Dense forest does not regrow fifty thousand hectares in five years. That reversal is the signature of pixels moving between two classification categories, not of trees.

So the honest reading is layered: the total series is trustworthy, the long-run direction of the split is trustworthy — dense was already falling and mixed already rising from 1973, long before the unstable years — and the individual 2009 and 2014 dense figures are not. They are greyed in the table and drawn faint in the chart rather than presented as measurements.

Why this says 72% and the Spatial Plan says 89.8%

ODC’s 2014 total is 71.7% of the province as the GIS measures it. The Spatial Plan reports 89.8% forest cover for 2019. That is not a contradiction and neither figure is wrong: they count different things. The Plan folds rubber plantation, regrowth and “other forest” into its forest class; ODC’s dense-plus-mixed does not. A province can lose closed-canopy forest and gain rubber and report more forest either way, depending which definition you reach for. Both are stated here; neither is reconciled away.

Twenty-five years of forest loss, and who is counting

Start with what does not exist. There is no official Cambodian forest cover figure for Mondulkiri at any recent date. The Forestry Administration's national series still rests on its 2010 assessment; no ministry publishes a provincial number. Every figure below is satellite-derived, from the University of Maryland's global analysis distributed through Global Forest Watch. For a province whose whole argument is about forest, that is the gap worth naming first.

The shape is the finding. Loss ran under 2,000 ha a year until 2007, climbed through the concession years to a peak of 13,512 ha in 2014, stayed above 7,500 ha through 2021 with one dip, and has fallen every year since 2021 to 3,787 ha in 2025. Humid primary forest loss follows the same curve and reached 1,426 ha in 2025, the lowest since 2004. Whatever is driving it — the concession moratorium, the protected-area reclassifications, the carbon project, exhaustion of the accessible frontier, or all four — the last four years are a genuine inflection and not noise.

Two cautions before anyone uses this series. Tree cover loss is not deforestation: it counts rubber and plantation harvest alongside forest clearing, and this province has 43,912 ha of rubber. And the 2001–2025 total is loss without any offsetting gain, so it is not a net figure. Read it as the rate at which canopy came down, which is what it measures. Reported

Protected areas, by instrument

Layer 44 gives each protected area with the legal instrument that created it, the year it was signed, its managing agency and its measured hectares. It resolves a confusion this Atlas had been carrying, and adds a sanctuary I had never mentioned.

What "protected" actually permits

A wildlife sanctuary is not one uniform rule. Under Cambodia's Protected Area Law each one is divided into four zones with very different permissions, and the zoning map is where conservation policy meets a household's ability to farm. The provincial GIS holds the January 2017 zoning for Phnom Prich Wildlife Sanctuary — the only Mondulkiri sanctuary whose full zoning sits in this database.

Three sanctuaries, three completely different bargains

The provincial GIS holds full four-zone plans for three of Mondulkiri's sanctuaries. Set side by side they show that "protected area" describes a legal category, not a consistent deal with the people living inside one. The share of land where residence and farming are recognised outright ranges from 3% to 12% — a fourfold difference between neighbouring sanctuaries.

Two honest notes on this table. Lomphat straddles the Ratanakiri border — the source records its provinces as “Rattanakiri-Mondulkiri” — so its totals are the whole sanctuary, not Mondulkiri's share, which the GIS does not separate. And the Srepok layer carried no name field in the records we first read, so we identified it by summing its zones to 374,053 ha and matching that against every protected-area polygon in the database: Srepok was 0.37% away, the next nearest candidate 27% away. We published that as an inference from geometry rather than a label. The service’s own layer index has since confirmed it — the layer is titled “Cambodia Mondulkiri Srepok Wildlife Sanctuary Zoning (Aug 2018 final version)”. The inference was right, and it is now a citation.

The Core Zone is closed to all but research and monitoring. The Conservation Use Zone allows only small-scale non-timber forest product collection by customary users. The Sustainable Use Zone is where economic development can be authorised by the government. The Community Use Zone is the only category that recognises settlement, farming and residence outright — and it is 3% of the sanctuary. Whether that 3% matches where people actually live is the question this map does not answer, and the honest reason land conflict recurs inside protected areas.

Community Protected Areas: 23 agreements, 63,300 ha

A Community Protected Area (CPA) is a formal agreement letting a village manage and use a defined block inside a sanctuary. It is the main mechanism by which people who live in a protected area keep legal access to it. The provincial GIS holds 23 current community agreements covering 63,300 ha — and they are distributed with striking evenness.

Source: Mondulkiri provincial GIS database, FeatureServer layer 43. Every record carries Status “Current”. The 23 hectare figures sum to 63,299.7 ha, and the CPA subtotals (Srepok 34,998.0 + Phnom Prich 22,474.1 = 57,472.1 ha) reconcile with the grand total once Community Forestry (4,153.1 ha) and Phnom Rohak (1,674.5 ha) are added. Phnom Rohak has a blank Category field in the source, so it is shown as uncategorised rather than assumed to be a CPA. Sanctuary attribution follows the source's Project code (PPWS / SWS / LPWS).

The same sanctuaries, two different sizes

The database holds two protected-area layers. Layer 44 is the one this Atlas has been quoting; layer 51 is the Ministry of Environment's national set. They cover the same ground and they do not agree — on names, on legal instruments, or on how big these places are. Their stated hectare fields differ by 78,120 ha. Measured from the polygons instead, the real difference is 27,279 ha — and layer 51 turns out to be the 1993 and 2002 designations rather than a rival survey. The vintages panel works it out, and corrects the figure this Atlas had been publishing.

Source: Mondulkiri provincial GIS, FeatureServer layers 44 and 51. Layer 51 is national — 41 protected areas across Cambodia — of which six touch Mondulkiri. Anyone citing a protected-area figure for this province should say which layer it came from and on what date; the two differ by more than the size of a whole sanctuary.

Two proposals that are not on any public map

The protected-area disagreement, resolved

This Atlas has been carrying an unresolved contradiction: the provincial database holds two protected-area layers, and they disagree. Until now all that could be said was by how much. Reading layer 51 properly settles it, and in doing so corrects a figure published further down this page.

Layer 51 is not a rival survey. It is an older one, and a national one.

The layer is named Cambodia Mondulkiri Protected_Areas (MoE), which suggests it describes this province. It does not. Its extent runs from 228,174 to 782,017 metres easting — the width of Cambodia — and it holds 41 protected areas nationally, of which five touch Mondulkiri. Its own description text is copied verbatim from the biodiversity-corridor layer, describing a “Cambodia Biordiversity Corridor”, misspelling included. This is the fourth layer in this database found to carry national extent behind a provincial name.

Every one of the five carries a designation year of 1993 or 2002. They are the original Royal Decree and Sub-Decree designations. Layer 44 is the current configuration. The two layers were never measuring the same moment, which is why nothing in the database “adjudicates” between them — there is nothing to adjudicate.

Correcting our own figure: 27,279 ha, not 78,120

The 78,120 ha figure this Atlas published came from adding up the difference between the two layers’ stated hectare fields. Recomputed from the polygons themselves, the five boundaries have moved +27,279 ha in total. The gap between those two answers is almost entirely one record.

Mondul Kiri Protected Forest — the 2002 designation covering the ground now called Srepok Wildlife Sanctuary — states 428,377 ha in its hectare field. Its polygon measures 375,062 ha. That is 14.2% out. Every other record in the layer agrees with its own geometry to within 0.6%. Measured honestly, Srepok’s boundary has moved 2,403 ha since 2002, not the 55,719 ha the attribute implies — a difference between a rounding-scale revision and a major reduction.

This is the second time an attribute in this database has been caught disagreeing with its own map. The first was the mining layer’s Hectares field. In both cases the geometry was right and the number beside it was not, which is a good reason to recompute rather than to read off.

Where the boundaries actually moved

A table of five numbers cannot show you where a boundary went. Phnom Namlear gained 20% and Keo Seima gained 3%, and they did it in completely different ways. Switch Boundaries 1993/2002 on in the interactive map above to watch the old outline against the current fills.

!Boundaries are indicative only. The line shown between Mondulkiri and Viet Nam is drawn from a simplified provincial GIS layer and does not represent the international boundary. Delimitation and demarcation of that boundary are matters for the Royal Government of Cambodia and the Government of Viet Nam; only maps published under their authority are official.ព្រំដែនដែលបង្ហាញ ជាការប្រមាណប៉ុណ្ណោះ។ ខ្សែរវាងខេត្តមណ្ឌលគិរី និងប្រទេសវៀតណាម មិនមែនជាព្រំដែនអន្តរជាតិផ្លូវការទេ។ មានតែផែនទីដែលបោះពុម្ពដោយអាជ្ញាធររដ្ឋប៉ុណ្ណោះ ដែលមានលក្ខណៈផ្លូវការ។

Derived · layers 44 and 51 of MDKPSPDATABASE, areas recomputed from polygon geometry on the WGS84 sphere. Layer 51 designations dated 1993 and 2002; layer 44 is the current configuration.

Who holds the land

Two registers sit side by side here, because they are the two competing claims on Mondulkiri's land: community-managed forest, and economic land concessions. Both are reproduced in full from the Provincial Spatial Plan's annexes.

Who claims this land

Derived Computed here by putting every land claim in this Atlas against the same denominator. Component figures come from the provincial GIS (layers 39–44), the Spatial Plan, and Open Development Cambodia. The reconciliation and the two findings below are this Atlas’s own.

Each claim on Mondulkiri lives in its own section of this Atlas — protected areas here, concessions there, mining somewhere else, communal title newly arrived. Each is correct and each is alone. Nobody had put them side by side against the province’s actual area, because until the communal title figures arrived nobody could.

Ninety-eight percent, and why that is not true

Added up as published, the claims come to 1,348,568 ha against a province of 1,372,090 ha. That is 98.3%, which reads like a province entirely spoken for. It is not, because the registers overlap and nobody subtracts.

Removing what is counted twice brings the total to 1,115,640 ha, 81.3%. The difference — 232,928 ha, seventeen percent of Mondulkiri — is land that appears in two registers at once.

Eight and a half thousand hectares claimed three ways

This Atlas computed separately that 754,888 ha of protected area carries neither a mining licence nor a concession. So 161,059 ha of protected area carries one of them. But mining-inside-a-protected-area and concession-inside-a-protected-area add to 169,628 ha — 8,569 ha more than the union.

Those 8,569 hectares carry a mining licence, an economic land concession, and protected-area status, all at once. Three claims on one piece of ground. It follows by inclusion–exclusion from figures already published here, and it has not been stated anywhere before — including by this Atlas.

What this is not

It is not a complete deduplication, and the 81.3% figure should not be quoted as one. Only three overlaps are known and measured. Communal title sitting inside a protected area, a social land concession inside a concession, a mining block over an ELC outside a protected area — none of those has been computed. 81.3% is a ceiling on the unique footprint, not a measurement of it. The real figure is lower, and nobody knows by how much.

Communal title: the gap closed, and a sharper one opened

Verified Source: Registered indigenous communal land, Open Development Cambodia, compiled from MLMUPC sub-decrees, OHCHR Cambodia, and Development and Partnership in Action (DPA). 44 titles nationally, covering 2009–2025. Licensed CC BY 4.0.

This Atlas has been carrying an admission: seven titled communities were named with parcel counts, but without areas it was “impossible to say how much land is under communal title province-wide”. It is no longer impossible.

The chain, closed end to end

This Atlas already held the sub-decrees that cut land out of Seima to release it for communal title. This registry holds the titles that were actually registered. Two of the three match to the hectare and to the family — from two separate datasets compiled from different sources, the Royal Gazette on one side and MLMUPC with NGO reporting on the other. That is a complete audit trail from the instrument that released the land to the title that now holds it.

The one that does not match

Sub-decree 62, promulgated 13 February 2013, cut 711.31 ha from the Seima Biodiversity Protected Area to grant Pnong communal ownership to 72 families. No registered title of 711.31 ha appears anywhere in the registry, in any province.

Two of three released parcels can be traced to a title. The third cannot. Either it was registered at a different area, or it was never registered at all — and these two sources cannot distinguish between those. That is stated rather than resolved. But it is a far sharper question than the one it replaces: it names the instrument, the date, the area, and the seventy-two families waiting on the answer.

One caution about the registry itself: ODC records a known inconsistency in it — the La En title in Ratanakiri is 1,454 ha per the United Nations and 1,218.56 ha per DPA. That is another province and does not touch the figures above, but it is a reminder that this is a registry assembled from reports, not a single authoritative register.

Four cuts in one month, March 2026

The slowest process in this province moved. Four sub-decrees, all indexed in March 2026, take land out of two protected areas, reclassify it as State private land, and hand it to Bunong communities as collective ownership — the mechanism this section describes as normally taking a decade.

Read alongside what the rest of this Atlas records, this cuts both ways and should be printed as such. It is the state doing the thing indigenous communities have been asking for, at a scale that is small but real: about 7.2 hectares per family. It is also four more excisions from protected areas in a province where excision by sub-decree is the mechanism by which protected land has repeatedly become something else. The same instrument does both jobs, and the only way to tell which is happening is to read who the land went to.

Two more instruments from the same period belong on the record: Prakas No. 76 of 16 July 2026 on establishing community forestry in Mondulkiri, and Prakas No. 061 of 11 August 2026 establishing the Ping Prong Klong Pesh Tho community forest at Poule village, Dak Dam commune. Verified

How people actually live

The National Institute of Statistics published sub-national data for five north-eastern provinces — Kratié, Stung Treng, Ratanakiri, Mondulkiri and Preah Vihear — in December 2020. Reading Mondulkiri against its neighbours is more useful than reading it against the national average, because these five share the same geography and the same constraints.

Social Services & Public Safety

Where the services are, and how far

Seven more layers drawn: two referral hospitals, eighteen health facilities, 98 primary schools, 30 pre-schools, nine high schools, ten colleges, and the single power transmission line that reaches this province. Counted, they sound like a network. Drawn, they are a line: almost everything sits on NR76 or within a few kilometres of Sen Monorom, and the west and north are close to empty.

How far is a village from care

The GIS holds village centres in one layer and health facilities in another, and never relates them — so this question has no answer in the database. Every one of the 77 mapped village centres is matched here to its nearest built facility by great-circle distance. Straight-line distance flatters the real journey, because in this province the road usually goes somewhere else first; treat every figure below as a floor, not an estimate.

The contrast is the finding. Primary schooling is genuinely distributed — a median of 1.5 km, and only 6 villages beyond 10 km. Health care is not. The median village is 7.0 km from the nearest built facility and 27.7 km from a hospital, and those medians hide a long tail in the west and north where the roads run out. Cambodia has put a school in nearly every village here. It has not done the same with clinics.

What these layers do not say

1. Twelve of the eighteen health records carry no name, district or commune — only a point and a type. Six of those are typed Future, so the province has 12 built facilities in this layer and 6 proposals.
2. One of the nine high schools has no geometry at all. The record exists, the location field is empty, so it can be counted but not placed. Eight are drawn.
3. Distances here are straight lines. On this terrain, with one north–south road, real travel distance is materially longer — especially in the wet season, when several of these routes are not reliably passable.
4. The school layers date from 2012 and the health layers from 2008. They are the most recent in the database, not the most recent in existence.

Water & irrigation: the Srae Huy project

The single largest infrastructure intervention currently documented for Mondulkiri. A weir built in 1975–77 in Kaoh Nheaek breached around 1980 and has never worked properly since. Rebuilding it is now funded under the Cambodia Water Security Improvement Project (P176615), running 2024–2030.

When the province was cleared

The settlement layer is the most quietly powerful thing in this database. Every one of its 1,134 footprints can carry a clearance year, traced to a specific named Landsat scene — so it is not a snapshot of where people live, it is a record of when each patch of ground was opened. 532 polygons carry a date. Sorted by decade, they describe an acceleration.

Source: Mondulkiri provincial GIS, FeatureServer layer 17 (settlement areas, WWF updated 2017). Footprints are classified as village, village mosaic, buildings or construction site; 279 are tagged plainly as “village”. Each dated polygon names the Landsat scene the clearance was read from, so individual entries can be re-checked against the original imagery.

The five districts, measured

The boundary layer carries each district's official code, its Khmer and English names, its class, and — usefully — its own count of communes and villages. Together they total 21 communes and 90 villages across 1,372,090 ha, which is 96.0% of the province's cited 1,428,800 ha.

Source: Mondulkiri provincial GIS, FeatureServer layer 5, “Admin District Boundaries (2014)”. The five districts sum to 1,372,090 ha against the 1,428,800 ha usually cited for the province — a 4% shortfall consistent with boundary generalisation at this scale. Saen Meaknourum is classed City (ក្រុង) rather than District (ស្រុក); the other four are districts.

Provincial Finance

What the province actually has to spend. Figures below are extracted from the Ministry of Economy and Finance's Budget in Brief for Sub-National Administration, Fiscal Year 2025 — the official published allocation, not an estimate.

Infrastructure & Connectivity

Checking the elevation range against the contours

This Atlas has stated since its first version that Mondulkiri runs from about 200 m in the lowlands to just over 1,000 m at Phnom Nam Lyr. The contour layer lets that be tested rather than repeated — but only after clearing a trap that would have produced a badly wrong answer.

Source: Mondulkiri provincial GIS, FeatureServer layer 29 (contour lines). This is the third layer in this database found to carry national rather than provincial extent — the others being the soil and geology tables, whose area fields describe national polygons. Any figure drawn from this service should be checked for whether the layer was actually clipped to Mondulkiri before it is quoted.

What the soils actually are

The productivity layer above says how good the soil is; this one says what it is. Crocker's classification — the 1962 survey still used as Cambodia's soil reference — divides the province into 58 polygons across nine soil types.

Why the good soil is where it is

The geology layer maps 381 polygons across nine geological eras, each carrying a rock description. Read alongside the soil layer above, it answers the question the soil map raises: where does the fertile third come from?

Source: Mondulkiri provincial GIS, FeatureServer layer 34 (geology). We are publishing polygon counts rather than areas here: as with the soil layer, this table's AREA field carries values from the national survey it was cut from, and the largest single “Water” polygon alone claims 17,239,684,857 m² — larger than the province. The clipped geometry would have to be summed per era to give real areas, and we have not yet done that.

A third of the province is good soil — and everything is built on that third

The soil fertility layer divides Mondulkiri into three productivity classes covering 1,366,378 ha, or 95.6% of the province. The split explains a great deal about where people, plantations and concessions actually are.

The rivers, ranked — and a surprise at the top

Fifteen rivers are named in the major-river layer, each with its Khmer name and its mapped length. Sorted by how much of each actually runs through Mondulkiri, the ordering is not the one the province's reputation would suggest.

390 named waters, and what their names tell you

One of the quietest layers in the database is also one of the most revealing: 625 label points carrying 390 distinct named water features, each with both its Khmer name and a romanisation. That is 2.7 named waters for every 100 km² of Mondulkiri — a density that says more about how the province is actually inhabited than any population figure does. People name what they use.

Source: Mondulkiri provincial GIS, FeatureServer layer 24 (hydrology stream names). Prefix counts are over all 625 label points; the same feature can carry more than one label, which is why the distinct-name count is 390. A named-feature density of 2.7 per 100 km² is our calculation against the province's 14,288 km².

The grid, as far as the database knows it

Two layers hold Mondulkiri's electricity transmission lines, and between them they contain three line segments totalling 218 km. That is the entire grid as this database records it — for a province of 14,288 km².

Source: Mondulkiri provincial GIS, FeatureServer layers 19 and 20. Read together with the 14,120 connected households figure elsewhere on this page, the picture is consistent: a small grid around Sen Monorom and the national road, and a large majority of the province served by off-grid solar, small diesel or nothing at all — but the GIS cannot confirm that, because it does not describe the lines it holds.

The road layer, and why we are not quoting a road length

The provincial GIS holds a road layer from the Provincial Department of Public Works and Transport, dated 2017: 240 line segments, each with a road name, a type and a status. It is the obvious source for a headline like “Mondulkiri has X km of road”. We are not going to give you that number, and the reason is worth more than the number would have been.

Source: Mondulkiri provincial GIS, FeatureServer layer 21 (PDPWTP 2017). We have deliberately published the composition of the layer rather than a derived kilometre total: with three quarters of the records flagged Deleted and the rural network essentially unmapped, any single length figure drawn from it would be confidently wrong, and would be quoted long after the caveat was lost.

Eleven dams on paper, none in the ground

The provincial GIS carries a hydropower layer with investor, capacity, expected annual generation, district, commune and project status for each site. Eleven Mondulkiri dams appear in it, together 104 MW and 630 GWh a year. Not one is built. Not one is under construction. Every single site is either Under study or Potential site, and every one traces back to the same origin: a 2003 identification exercise by the then Ministry of Industry, Mines and Energy.

Source: Mondulkiri provincial GIS, FeatureServer layer 23 (hydropower dams). This is now the complete Mondulkiri set — the service reports exactly 11 records for province = ‘Mondul Kiri’, and all 11 are listed above. Capacities sum to 104.0 MW and generation to 630 GWh/yr. Sites in the same layer belonging to other provinces, such as Prek Rwei II in Ratanakiri’s Lumphat district, are excluded.

A power station that never happened, and three dots with no names

Two of the layers this Atlas had read but never drawn. Neither is large — thirteen points and three points — but between them they correct a figure published here and put names on data that had none.

Thirteen sites, 114 MW, and not one of them exists

This Atlas has been reporting eleven hydropower sites totalling 104 MW, taking the count from the layer’s own province field. Testing each point against the district polygons instead finds thirteen sites inside Mondulkiri, totalling 114 MW. Two — Prek Rwei II and Prek Chbar — are filed under Ratanak Kiri but fall 9.3 km and 11.6 km inside Kaoh Nheaek. That is far too deep to be a boundary-generalisation artefact.

Both misfiled records give their district as “Lumphat” and their commune as “Chey Oddam”. Lumphat is a district of Ratanakiri and a wildlife sanctuary that straddles both provinces, so the likeliest explanation is that whoever populated the field reached for the sanctuary name. That is a hypothesis, not a finding; what is certain is where the coordinates fall.

Seven are “Under study”, six are “Potential site”, and every one traces back to a single 2003 Ministry of Industry, Mines and Energy list. No investor, no financier, no EIA date and no EIA status is recorded for any of them, and the whole layer was last touched in June 2015. Read plainly: this is a decade-old wish list, not a pipeline.

The border layer has no attributes at all

Layer 6 holds three points. Its entire attribute table is an OBJECTID and an Id column that is zero for every record. No name, no type, no status, no country. The Spatial Plan (p83) names four crossings along the province’s 312 km of border with Vietnam and states the district of each, but publishes no coordinates. Joining the two on the district each point actually falls in gives named coordinates for three, and identifies exactly which one the GIS is missing.

• La Pha Khe — Kaev Seima district, major. Sits 0.57 km outside the drawn district boundary, which is what a border post does.
• Dak Dam — Ou Reang district, major. Falls inside the district.
• Nam Lear — Pechreada district, under development. Sits 0.47 km outside the drawn district boundary, which is what a border post does.
• Kaoh Romyearl (Ji Meat) — Kaoh Nheaek district, unpaved road. In the Plan, absent from the GIS.

Derived · layers 23 and 6 of MDKPSPDATABASE, each point tested against the district polygons of layer 2. Crossing names from the Provincial Spatial Plan 2020–2040, p83, matched on district. Capacity and status are the layer’s own.

!Boundaries are indicative only. The line shown between Mondulkiri and Viet Nam is drawn from a simplified provincial GIS layer and does not represent the international boundary. Delimitation and demarcation of that boundary are matters for the Royal Government of Cambodia and the Government of Viet Nam; only maps published under their authority are official.ព្រំដែនដែលបង្ហាញ ជាការប្រមាណប៉ុណ្ណោះ។ ខ្សែរវាងខេត្តមណ្ឌលគិរី និងប្រទេសវៀតណាម មិនមែនជាព្រំដែនអន្តរជាតិផ្លូវការទេ។ មានតែផែនទីដែលបោះពុម្ពដោយអាជ្ញាធររដ្ឋប៉ុណ្ណោះ ដែលមានលក្ខណៈផ្លូវការ។

Mining, licence by licence

This Atlas previously recorded the mining footprint as a flagged gap — licensed area was known, actual extraction was not. Layer 41 of the WWF database resolves it, with company names, mineral, licence type, status and expiry for 27 licence blocks.

The licence map, transcribed

The Atlas has been carrying an open gap for the mining licence boundaries: the layer existed in the provincial GIS but had never been read out. It has now been. Twenty-seven licence polygons are recorded, each with a company name, mineral, licence type, status and expiry — and reading them together changes the headline figure considerably.

Three things this table will not let you do by accident. First, Poh Resources appears twice with byte-identical 25,200 ha polygons, so any naive sum double-counts 25,200 ha; de-duplicated, all records come to 286,305 ha rather than 311,505 ha. Second, eight records are empty shells — Riva Natural Resources, Sun Hill Minerals (×3), Maxum Metals (×3) and Southern Gold all carry no mineral, no area and mostly status Invalid; they are names retained in the register, not ground under licence. Third, one entry is not a company at all: Chong Plas Community Mining Enterprise, 41.2 ha of Valid gold licence — the only community-held mining title in the province, and roughly one six-hundredth the size of the largest corporate block.

Source: Mondulkiri provincial GIS, FeatureServer layer 41 (PDPE, Sep 2017). Status counts: Valid 12 · Invalid 12 · not yet granted 2 · blank 1 = 27. Note this GIS figure of 173,186 ha of live licence area sits below the 213,531 ha register total of January 2021 elsewhere on this page — the two are measuring different things (mapped boundaries against register entries) and at different dates, and we have not reconciled them.

The national register, and a province with two spellings

Verified Source: Mining concessions in Cambodia (1995–2022), Open Development Cambodia, 307 records nationally, licensed CC BY 4.0.

Every query this Atlas ran against that register for “Mondulkiri” came back empty. The register spells it “Mondul Kiri”, two words. That is worth recording on its own: a province with two spellings is a province that quietly goes missing from searches, including from the searches of anyone else trying to audit it.

Spelled correctly, the register holds 43 licence records touching this province between 1995 and 2022, against the 27 blocks the provincial GIS carries at September 2017. It names directors, addresses and company websites, none of which the GIS layer has.

647,185 hectares — and why you must not quote it

The 24 records that carry an area sum to 647,185 ha, forty-seven percent of the province. That number is real and it is not a mining footprint. It is a cumulative historical register: lapsed licences, superseded blocks, overlapping grants and at least one outright duplicate, each counted once. The figure to quote for live licensed area remains the GIS’s 163,586 ha. A further 17 records carry no area at all.

The register also disagrees with itself. Anging (Cambodia) Invest Company Co. Ltd. and Anquing Cambodia Investment Company both appear at exactly 21,600 ha — two transliterations of Anqing, almost certainly one licence entered twice.

Six holders the provincial GIS does not name

Among them BHP Billiton and an aluminium venture with a larger recorded area than any block in the provincial layer.

Activity after September 2017, at last

This Atlas has recorded, as an open gap, that nothing was known about licence activity after the provincial layer’s September 2017 date. One record closes part of it: Renaissance Minerals (Cambodia) Limited, area O'Ktung in Keo Seima district, 19,520 ha — status line dated 23 March 2020 — government selected as a candidate for an exploration licence. That is three years past the GIS snapshot, and it shows the ground in and around Keo Seima still moving through the licensing process.

Who is growing, and who holds the land

Two more layers drawn for the first time. The first records how each commune's population changed between 2006 and 2016 — the only decade-scale, commune-level population series that exists for this province. The second puts the three competing forms of land holding on one frame, which is the only way to see how unequal they are.

A commune whose people are missing from the layer

Layer 4 has 21 rows for 21 communes, so at a glance it looks complete. It is not. Two of the rows are byte-for-byte identical — same name, same 381.23 km², same 2,674 people in 2006, same 3,354 in 2016, same growth rate. Both belong to Ou Reang district, which has two communes. One of them has been given a copy of the other's record.

1. Dak Dam has no population figures in this layer at all. Its 46,875 ha are on the map; its people are not in the table.
2. Summing the layer as delivered gives 82,157 for 2016 — a figure that double-counts one commune and omits another. The honest total across the 20 unique rows is 78,803, and it is short by one commune.
3. The area on the duplicated row matches neither polygon. It records 381.23 km²; the two Ou Reang communes measure roughly 788 km² and 469 km² on the ground.
4. Four sangkats of Saen Meaknourum city record zero villages, which is why a village count summed from this layer comes to 73 rather than the province's 90.

Derived · 227 of the 287 titled parcels — 79% — are in Kaev Seima district. Four of Mondulkiri's five districts have either one titled community or none. Set against 23 concession blocks spread across the whole province, the map shows an asymmetry that the hectare figures alone understate: concession land is provincial, titled indigenous land is local to one corner.

The Master Plan set a destination. This is the road so far.

Verified Reported Derived Sources are given per row in the table. The 2035 targets are from the Master Plan for Mondulkiri Tourism Development 2021–2035; the 2017 baseline is from the Provincial Spatial Plan 2021–2040; the outturn figures are from the Mondulkiri Provincial Department of Tourism and from the deputy provincial governor, reported in the press; the national comparisons are from the Ministry of Tourism’s own statistics report.

This Atlas has been publishing the Master Plan’s targets — 900,000 international visitors and two million domestic by 2035 — since the Plan was extracted. Targets without an outturn are advertising. Here is every arrival figure that can be traced to a named authority, and what those targets therefore require.

Every figure, with its source and its working

Where the one derived row comes from

The 2022 half-year was never published as a number. It was published as three percentages: the 2023 half-year was reported as −23.3% overall, −24% domestic and +65% foreign. Reversing each one separately gives 224,656 total, 222,795 domestic and 1,807 foreign.

That reversal is only trustworthy because of what happens next: the two reversed parts add to 54 visitors away from the reversed total — 0.024% apart. Three independently rounded percentages had no obligation to agree that closely. They do, which is the evidence that the arithmetic is sound. The row is still marked Derived and drawn faint, because it is arithmetic and not a measurement.

Is 57.7% a year a lot?

A growth rate means nothing without a yardstick, so here is one. Cambodia took international arrivals from 118,183 in 1993 to 6,610,592 in 2019 — the entire post-war tourism boom, Angkor and the coast included, twenty-six years of it. That is 16.7% a year.

To reach 900,000 international visitors by 2035 from roughly 6,000 in 2024, Mondulkiri needs to grow international arrivals about 3.4 times faster than that, and hold it for eleven straight years. The domestic target, by contrast, needs 15.8% a year, which is demanding but not extraordinary — the province has been running at roughly that pace already.

The honest conclusion is not “the plan will fail”. It is narrower and firmer than that: the domestic half of the target is within reach and the international half is not, on any trajectory the record so far supports. The two should stop being quoted as one number.

The national market is going the other way

The Ministry of Tourism’s own figures for the period the plan is running through: Cambodia received 5,569,752 international visitors in 2025, -16.9% down on 2024. The first quarter of 2026 brought 1,014,147, -44.8% down on the same quarter of 2025 and -46% below the 2019 quarter. Mondulkiri’s international target requires the steepest sustained climb in the province’s history to begin during a national contraction.

One further reference point, offered with a caveat. The Ministry reports internal tourism by destination region, and Mondulkiri sits in its Eco-tourism Zone. That whole zone drew 15,604 foreign visitors in the first quarter of 2026, -4.3% on the year before, against 137,861 Cambodian visitors, +9.9%. The report does not state which provinces the zone contains, so this is not a Mondulkiri figure — but it caps how large one plausibly is, and it points the same way: domestic up, foreign down.

What the one visible decline was blamed on

The series contains one clear fall — the 2023 half-year, -23.3% down on 2022. It is worth recording why, because the reasons are structural rather than one bad season, and none of them has been reversed.

• The Phnom Penh-Sihanoukville Expressway opened — Day trips to nearer places replaced overnight trips east.
• Economic downturn — Households chose closer destinations over costlier long trips.
• Road condition — The national road into the province was cut for a period by flooding.

Sim Dimanche, general manager, Avocado Guesthouse, Sen Monorom recorded a 50% year-on-year drop in guests over the same period (Phnom Penh Post, 27 Aug 2023). Seang Soeun, director, Mondulkiri Provincial Department of Tourism attributed the fall to the Russia-Ukraine war and the 2023 national election. Both explanations are recorded; neither is endorsed here. Note that foreign arrivals rose 65% in that same half-year while domestic fell 24% — whatever drove the fall, it was a domestic-market effect.

How many tourism sites does Mondulkiri have? Three official answers.

Verified Reported Sources: Mondulkiri Provincial Spatial Plan 2021–2040 (text and GIS layer 7); Ngin Sovimean, deputy provincial governor, interviewed in the Phnom Penh Post, 14 February 2025.

The Visitor Guide already carries two of these counts and the gap between them. A third has since appeared, from the provincial administration itself, and it is seven times the first. All three are printed here rather than averaged, because the spread is the finding: nobody is maintaining a register.

These are not three attempts at the same measurement. 15 counts sites the province formally recognises, with an operator attached to each. 29 counts points somebody drew on a map. 106 counts places worth visiting, on a definition nobody has written down. A visitor wants the first, a planner needs the second, and the third is the only one that has ever been used in public to describe the province’s potential.

One of the eleven blanks can now be filled

GIS layer 7 carries eleven points with a Khmer name and no English one, which is why they never reach guidebooks. The Atlas listed them anyway. One of them — នាមលៀរ, transliterated Nam Lear — turns out to be the site the deputy governor calls Namlear Mountain, and he gives it a size: 24 hectares, named as one of the developments that lifted visitor numbers before the pandemic. Ten blanks left.

What is actually built

Set the last two cards against each other and the province’s position is legible in one line: roughly two thousand places to sleep, one developed attraction, and about six thousand foreign visitors a year. The constraint is not beds.

The airport the whole forecast rests on

The Master Plan’s international growth curve assumes an airport opening in 2023. The current state of that project: a site of 300–600 ha in Sen Monorom commune, Ou Reang district, costed at over $60 million, awarded to the FOCUS Consortium under Notification No. 41 of the NCC, dated 9 January 2024, with a feasibility study started in March 2024. A previous Chinese partner withdrew, citing COVID-era finances. A feasibility study is not a runway. Every international figure in the Master Plan should be read against this line.

A smaller discrepancy, resolved

The same interview describes Mondulkiri as “one town, four districts, 17 communes and 92 villages”. This Atlas publishes 21 communes from the GIS boundaries, and both are right. Sen Monorom is a krong — a municipality — and its four urban quarters are sangkat, not khum. 17 rural communes plus 4 urban quarters is 21 commune-level units. The counts differ by exactly the number of sangkat, which is what a real reconciliation looks like as opposed to a coincidence.

Legal Register

The instruments that govern land, boundaries, protected areas and administration in Mondulkiri — and, more usefully, what each one actually establishes. The register is a finding aid, not the substance: where a document carries figures or rules that matter, those have been pulled out into the sections above and the summaries below. For the instruments that move a boundary rather than describe one — every sub-decree that has cut hectares out of the protected estate, and who received them — see The protected estate, cut by decree.

Foundational chronology

The instruments that actually constitute Mondulkiri as a jurisdiction — where its borders came from, when it became a province, and what has been carved out of it since.

Source documents

All four district-level emblem Prakas were issued the same day (17 Jul 2026) — a coordinated province-wide rollout. Sen Monorom's own municipal emblem instrument was not found in this pass; the NCDD e-Library's search filter did not return targeted results at time of writing, so this list was assembled by direct inspection of the library's homepage feed rather than a full-text search — treat the absence of a fifth entry as a research gap, not confirmation none exists.

The protected estate, cut by decree

Verified Source: Open Development Cambodia, catalogued from the Royal Gazette and published through its open DataHub. ODC holds 71 legal records for this province; the 35 below are the ones that move a boundary. Retrieved 2026-08-11.

Everything else in this Atlas describes where Mondulkiri’s protected areas are. This describes them being cut. Cambodia excises land from protected forest one sub-decree at a time, and each instrument names its hectares, the forest it comes out of and who receives the result. Sorted by date, those instruments become a record that exists nowhere as a single document.

It is not only a story of loss

The obvious reading is a list of losses, and that reading is wrong. Sub-decrees 62, 239 and 250 cut a combined 1,582 ha out of Seima specifically to grant Pnong indigenous communal ownership to three communities totalling 196 families. Sub-decree 393 cut 3,521 ha and passed 3,134 ha of it to 750 families in 1,335 titled parcels. The instrument that hands forest to a rubber company is the same instrument that hands it to the people already living on it. Publishing only the total would misrepresent the record, so the chart is coloured by recipient.

Forty-eight hours in June 2011

Four sub-decrees dated 15 and 16 June 2011 reclassified 9,773, 9,656, 9,014 and 9,614 hectares — all in Rayor commune, Kaoh Nheaek district. That is 38,057 ha in one commune inside two days, and roughly 17% of the commune this Atlas measures at 220,853 ha. Each is a distinct instrument with a distinct area; none is a correction of another. None of the four names a recipient. It is visible only when the whole series is sorted by date, which is what this panel does.

Every instrument, in order

Instruments that establish protection rather than remove it — Sub-decree 75 (2002, Mondulkiri Protected Forest) and Sub-decree 143 (2009, Keo Seima) — carry no hectare figure in the catalogue and are shown for context. Note Sub-decree 206 of 2007: it downsizes the protected forest created by Sub-decree 75 five years earlier.

The zoning, dated at last — and one figure confirmed exactly

Verified Source: Management zones of natural protected areas (2019–2026), Open Development Cambodia, digitised from Royal Gazette sub-decrees by ODC’s mapping team using QGIS. 63 rows nationally, updated 2026-07-24. Licensed CC BY 4.0 — attribution only, not share-alike, unlike ODC’s legal records above.

This Atlas already draws sanctuary zoning from the provincial GIS. What it never had was the date each zoning was issued. Mondulkiri’s three zoned sanctuaries: Lomphat, 5 January 2017 (Sub-decree 04); Srepok, 7 February 2019; Phnom Prich, 15 February 2019.

And one figure now stands confirmed from two independent chains. Phnom Prich’s conservation zone is 78,850 ha in ODC’s table, digitised from the national gazette — and 78,850 ha in Table 5 of the Provincial Spatial Plan, compiled separately by the provincial planning committee. Nothing else in this Atlas is corroborated that precisely.

Lomphat Wildlife Sanctuary, all fourteen zones

Lomphat is the only one of the three ODC sizes completely. This is the whole sanctuary across Ratanakiri and Mondulkiri, not the Mondulkiri share — the Spatial Plan’s 49,210 ha figure counts only the part inside this province. Srepok’s sixteen zones and twelve of Phnom Prich’s thirteen carry no area in the source at all.

Zone	Area	Share
Conservation zone	105,285 ha	42.1%
Sustainable use zone	79,125 ha	31.7%
Core zone	41,920 ha	16.8%
Community zone	23,669 ha	9.5%
Total	249,999 ha	100%

The total, 249,999 ha, is a third figure for Lomphat: this Atlas carries 248,014 ha from GIS layer 44 and 254,152 ha from summing the GIS zoning polygons. All three sit within 1.6% of each other, which for a boundary shared with another province is close agreement rather than a dispute.

Keo Seima is missing here too

This Atlas has recorded that the provincial GIS holds no zoning for Keo Seima. ODC’s national compilation — Lumphat, Srepok, Phnom Prich, Cardamom Mountains, Toul Pon Taley Boeung Snae, Phnom Krang Dei Meas, Phnom Kulen and Siem Pang — does not contain it either. Nor Phnom Namlear. So the gap is not a provincial filing failure: Mondulkiri’s second-largest protected area, the one with the most people living inside it, has no published management zoning in either source. That is a stronger and more troubling statement than the one this Atlas was making, and it is the reason the gap stays open rather than being quietly downgraded.

Licence. Open Development Cambodia does not use one licence for everything, and it matters. Its legal records — the sub-decree chronology above — carry CC BY-SA 4.0, which is share-alike: anything built from them must carry the same terms, so this panel is offered under CC BY-SA 4.0 and anyone reusing it inherits that obligation. Its zoning dataset, used in the section immediately above, carries CC BY 4.0 — attribution only, no share-alike. Applying the stricter licence to both by reflex would have been easier and wrong. The underlying sub-decrees are Royal Gazette instruments of the Royal Government of Cambodia.

The Spatial Plan, 2020–2040

Everything above this point is a snapshot: boundaries and attributes as they stood when somebody digitised them. The Provincial Spatial Plan is the other half of the picture. It measures the same province across twenty years, projects it twenty years forward under five different futures, and states what the administration intends to do about it. Adopted 2021, 164 pages, prepared for the Provincial Committee for Land Management and Urban Planning. Page numbers are given throughout so any figure here can be checked against the document.

A fifth of the province changed cover in twenty years

Between 1999 and 2019, 303,464 ha changed land cover — 21% of Mondulkiri. Two processes account for almost four fifths of it, and they run in opposite directions.

Five futures for 2040

The Plan does not forecast one population; it models five, and the difference between them is migration policy. That is the honest way to do it here, because in-migration is what drives growth in this province — net migration rates run from 2.2% to 23.3% by district against a national mean of 0.86% (p65).

Verified · The Plan then converts households into land: between 28,470 and 96,723 ha of new agricultural land will be needed by 2040, depending on how many households farm and how much each is allocated (Table 13, p100). The upper figure is larger than every economic land concession in the province combined.

Population and growth, district by district

Verified · The province grew at 4.56% a year against a national 1.2%. Pechreada nearly doubled. Meanwhile household size fell from 4.5 to 4.2 people, so the number of households grew faster still — which is what actually drives demand for land.

What each district farms

Verified · The districts farm differently. Kaoh Nheaek grows 70% of the province's paddy rice and holds the most rubber. Ou Reang and Saen Meaknourum are dominated by forest-crop mosaic — shifting cultivation and smallholdings rather than plantations. Pechreada carries 42% of all rubber.

Where the Plan and the GIS disagree

Having two official sources is more useful than having one, because the places they diverge are exactly the places a single figure would have misled you. All four gaps below are real, and none of them means one source is simply wrong — they were measured at different times, by different methods, for different purposes.

Contested · The mining gap is the one to watch: 49,945 ha, because the Plan counts licences surveyed in January 2021 while the GIS holds polygons from September 2017 and only those still valid. The area gap is definitional — 1,428,800 ha is the Ministry of Interior's administrative figure, while the GIS measures the boundary as drawn. This Atlas quotes both and says which is which; these checks also run live in The Atlas checks itself.

Roads, climate and work

Two futures, and what the province chose

The figures in the previous panel are the Plan's measurements. This one is its argument. It models two ways the next decade could go, finds that by Cambodia's own definition this province contains no towns at all, values six things the forest does for the economy, and commits to five goals. Everything here is read from the document, with pages given.

Ninety-two percent, or thirty-four

Working from a 2010 baseline of 92% forest cover, the Plan carries through two scenarios to 2030. The gap between them is the whole argument of the document.

A province with no towns in it

Cambodia declares a commune urban on three tests: population above 2,000, density above 200/km², and agriculture below half the labour force. By Cambodia's national definition, Mondulkiri has no urban areas at all. Sangkat Monourom meets the density test at 214 people/km² and clears the employment test with 13% in agriculture — and then fails on population, with 1,945 people. Not one commune or sangkat in the province passes all three.

So the Plan classifies centres by what they can actually do for you: 28 services — social, cultural, economic — each scoring one point.

Verified · Even the highest-ranked centre has neither a private hospital nor a university. This is the practical meaning of a population of 92,000 spread across 14,288 km²: the settlement hierarchy exists, but it is built out of health centres and secondary schools rather than anything a census would call a town.

Six things the forest does

Mapped with InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) on a 2010 baseline, each scored 0 to 1 across the province.

• Carbon storage (regulating) — How much carbon the vegetation holds.
• Non-timber forest products (provisioning) — Resin, honey, bamboo, rattan, wild food — a quarter of the province's adults collect these as a second occupation.
• Habitat quality (supporting) — How intact the habitat is for the species that need it.
• Water retention (regulating) — How much water the landscape holds rather than sheds.
• Nutrient retention (regulating) — Nutrients kept out of the watercourses.
• Sediment retention (regulating) — Soil held in place rather than washed downstream.

The combined index tracks carbon storage and non-timber forest products most closely — which is to say the two services people can most directly eat or sell are also the ones the map is most sensitive to.

Five goals to 2040

The vision the Plan sets out rests on four commitments: Forest conservation as Cambodia's contribution to climate change mitigation; Ecotourism that respects the environment and the province's ethnic diversity; Agriculture where agro-industry and smallholders co-evolve through agro-ecological intensification; An active north-east pole linking Cambodia into the CLV triangle. Those become five long-term development goals and 14 objectives, each tied to somewhere specific on the map (p106).

How much of the Plan is here

164 pages and 46 map sheets. This is what has been taken and what has not — kept current, because a claim of completeness is exactly the kind of thing the rest of this site exists to check.

Derived · All of the Plan's substantive data is now on this site. What remains uncopied is procedural text: committee membership, meeting records and monitoring templates.

Water, soil, people, risk

The rest of the Spatial Plan. Most of it fills in the physical picture — what the soil is, where the water goes, how the protected estate is actually divided. One part of it does something else: it shows a sentence this Atlas has been repeating quietly expiring.

“A Bunong province” is becoming untrue

In 2009 the Bunong were a majority of families in Mondulkiri. Ten years later they were not. This is the clearest single consequence of the in-migration the Plan measures elsewhere, and it happened inside one decade.

Verified · In-migration is almost entirely Khmer and Cham. The Plan is blunt about what that means: the influx puts pressure on indigenous groups already vulnerable to national forest and agricultural modernisation policy.

How the protected estate is actually divided

The Atlas has the zoning for Srepok and Lomphat from the GIS. The Plan has all four sanctuaries in one table — and the totals reframe what “protected” means here.

Verified · Two thirds of the protected estate — core plus conservation, 66% — is closed to settlement. Only 9.2% is community zone, the one category where people may unambiguously live and farm. And 16% of the province's population already lives inside a protected area, a share the Plan expects to keep rising. Co-management covers some of the gap: 21 community protected areas over 63,079 ha (6.7% of the protected estate, on 15-year agreements), plus 8 community forestry schemes over 9,799 ha involving 724 families.

What the ground is made of

Verified · Read this against the GIS soil-productivity layer, which found 61% of the province low-productivity: the two agree. Acrisols on 64% of the province are the reason. The good ground — ferralsols on the south-eastern plateau — is 9%, and it is the same corner that carries the roads, the concessions and the fastest-growing communes.

The border economy

Mondulkiri shares 312 km of border with Vietnam — Dak Lak, Dak Nong and Binh Phuoc. The province exports timber, coffee, rubber, pepper and raw materials and imports agricultural products, vegetables, machinery, construction materials.

Verified · On tourism, the Plan records 114,349 visitors in 2017 — 102,590 Cambodian and 11,759 foreign, up 14% on 2013 — across 15 recognised attraction sites. Note the ratio: fewer than one visitor in ten is foreign, which is worth remembering against a Tourism Master Plan targeting 900,000 international arrivals by 2035.

What the climate is already doing

Climate change is felt here through drought, flood, fungal disease and insect infestation. Thuon and Sumaylo (2009) find high vulnerability to both flood and drought, on high exposure and high sensitivity. Both are increasing in frequency and duration, and both have already cut crop yield and quality in villages across the province, with consequences for food security. The general view in the province is that deforestation and land-use change are themselves driving the increase in intensity and frequency.

The Plan puts forward two responses:

• Agro-ecology — Conservation agriculture, home gardening, organic methods, the system of rice intensification, integrated pest management, agroforestry. Knowledge-intensive rather than input-intensive, and it values what farmers already know about their own ground. Makes farming more resilient to climate change while reducing its contribution to it.
• Forest conservation — Cambodia's contribution to mitigation, and the reason the conservation scenario matters beyond the province's own borders.

And one constraint that decides much of the rest. New settlement in the core and conservation zones is prohibited outright. With population rising under every scenario, that makes detailed land-use planning inside each sanctuary's community zone unavoidable rather than optional. The Plan argues for compact expansion around existing centres rather than linear sprawl along the roads — because sprawl is what pushes settlement into farmland and sanctuaries.

Verified · §3.1.3, p101. This is the point where the Plan's parts lock together: population rises under every scenario, the core and conservation zones are closed, only 9.2% of the protected estate is community zone, and 64% of the province is poor soil. The room to absorb 56,940–101,559 more people is not large, and it is not where most of them will want to go.

The Plan’s maps, made live

Derived Source: the seventeen map sheets of the Provincial Spatial Plan 2020–2040, extracted at native resolution. Placement is derived here by matching the province silhouette from the district geometry of GIS layer 2 against each sheet; overlap with the printed map is stated per sheet below the viewer.

The sheets below this one are pictures. These are not. Each has been located in longitude and latitude by matching the province’s own silhouette against the printed map, which means the Atlas’s vector layers can be laid straight over the top. Choose a sheet, fade it, add districts or mining licences or villages, and click anywhere to ask what is underneath. Two sheets can be wiped against each other — which is the only fair way to read the conservation scenario against business as usual, because it is the same frame both times.

Fifteen of the seventeen sheets are placed. Maps 7 and 9 are refused: both are wide regional sheets on which the province is a small part of the coloured area, and on both the fit landed somewhere plausible-looking and wrong. They stay in the gallery below as flat pictures, because a map that is out by a few hundred metres is worse than one that admits it has no coordinates. Map 10 and Map 23 print two panels side by side, and only the left panel is registered.

The map opens framed on the province, which puts each sheet’s printed legend outside the view. Legend zooms out to the whole sheet so you can read it; scroll to zoom and drag to pan at any time.

The Plan's own maps

Every other map on this site is drawn from GIS geometry, which is better than copying a picture. These seventeen are the exception. They show surfaces that were modelled for the Spatial Plan and never released as data — the InVEST ecosystem services, soil suitability, catchment boundaries, net migration, and the two 2030 scenarios. For these, the printed sheet is the only version there is, so it has been lifted from the document at native resolution and is reproduced here with its map number and page.

Tap any sheet to open it full size. Khmer and English labelling is the source's own.

The two futures, mapped

The Plan's central comparison. Land use first, then the ecosystem-service surface that follows from it. Read 34 against 36, and 35 against 37 — the same province, ten years out, under two policies.

Surfaces that exist nowhere else

Six ecosystem services modelled with InVEST on a 2010 baseline, then combined. These were produced for the Plan and never published as geometry, so the sheet is the data. They are printed small in the source and cannot be enlarged beyond what you see.

Physical ground

Catchments, soil suitability and the two-date land-cover comparison behind the 303,464 ha figure. None has a GIS equivalent the province publishes.

People and intent

Net migration from the 2008 census, the urban hierarchy the Plan intends by 2040, and every strategy on one sheet.

Verified · Extracted with pdfimages from Mondul Kiri Provincial Spatial Plan 2020–2040, adopted 2021, prepared for the Provincial Committee for Land Management and Urban Planning. Native resolution up to 1,990×1,407 at 220 ppi; resampled to 1,500 px and JPEG-compressed so the set totals 2.7 MB rather than 90. The ecosystem-service sheets are small in the original and have not been upscaled — enlarging them would invent detail the source does not contain. The remaining 29 sheets in the Plan duplicate features this Atlas already draws from the GIS and are not reproduced.

The provincial GIS database

The spatial data behind the Provincial Spatial Plan is published openly by WWF as a live ArcGIS feature service — 39 layers, 467 MB, licensed "No Restrictions". It is the single most important open dataset about this province, and it closes gaps this Atlas previously had to record as unfillable.

Commune growth, 2006 → 2016

Layer 4 carries what no published report does: population, area, density and a ten-year growth rate for every commune. It reframes several things this Atlas had only been able to state at district level.

VerifiedWWF PSP Database layer 4 (Settlement Population Growth Rate 2006–2016). Pu Chrey grew 9.89% a year — nearly tripling in a decade — while A Buon Leu managed 1.86%, the slowest, corroborating the population fall this Atlas noted from the census series. Roya is the extreme case of the province in miniature: 2,219 km² holding 2,647 people, a density of 1.19/km², against Rummeaknea's 61.7.

Service endpoint: MDKPSPDATABASE FeatureServer · item page · owner WWF_Globil · created Dec 2018, last modified Nov 2020 · attribution: "Collected by WWF from Multiple Provincial & National Sources". Exports are available as GeoJSON, shapefile, CSV, KML, File GDB and GeoPackage.

One map, eighteen layers

Pick one base map — districts, communes, or population growth — then stack areas, lines and places on top of it. The three bases are alternative pictures of the same ground, so only one shows at a time; everything else adds. Hover or tap any shape to read what it is: every feature carries its own figures, including the measured area, protected share and mining area for each commune from the cross-layer analysis. Keyboard users can tab through the features and the readout follows focus.

The province, drawn from its own database

Every figure in this Atlas came out of the provincial GIS; these are the same layers drawn rather than counted. The geometry was pulled from the FeatureServer as GeoJSON, generalised, projected with a cosine correction for this latitude, and written as inline SVG — so they need no tile server, no API key and no network connection, and they work in the offline copy of the site.

All four share one projection and one frame, so they can be read against each other: the concessions map and the protected-areas map are the same province at the same scale. Simplification is deliberate and stated — these are orientation maps, not survey documents, and should not be used to locate a boundary on the ground.

How much of this database we have actually read

The provincial GIS service publishes 39 feature layers. This Atlas has read and published all 39 — 100%. Every one has now been read. The table below is kept as the record of what was gone through, and as the place to mark any layer the service adds later.