Cambodia: Housing and Flood Maps: What Can You Actually Verify?

A map can verify a relationship to a layer; it cannot verify a building in the engineering sense. If the property point is trustworthy and the product has a clear date, event type and scale, you may be able to say that the point falls inside, outside or near a mapped extent. That is a useful finding, but it remains a finding about the dataset. The same map does not tell you whether water entered the lobby, reached a basement ramp, disabled lifts, overwhelmed pumps or blocked the road for several hours.

The first task is to identify what kind of evidence is on screen. MRC’s near-real-time service is station based: it presents hydrometeorological observations and water-level charts. MRC river forecasting is a different product, using station data to forecast water levels at forecast points on the Mekong mainstream. Its flash-flood guidance is separate again and is designed around a faster, smaller-scale hazard. Cambodia’s NFFC publishes flood-extent maps by selected periods, while CamDi records disaster events and losses. The WFP/NCDM assessment adds another category: modelled climate-disaster risk and its drivers at commune level. Similar colours on different websites do not make those products interchangeable.

Scale matters as much as the label. A commune classified in a risk map does not turn every parcel in that commune into the same risk case. A river gauge reading does not become water depth at a condominium. A historical event record does not establish that a particular building flooded, and it certainly does not provide a future probability for that building. A flood-extent image may be much closer to the question a buyer is asking, but only if its date, coverage, legend and resolution are good enough for the point being tested.

The reverse mistake is to treat a blank area as reassurance. A layer can miss the mechanism that matters to the property. River flooding, rapid flash flooding and urban rainwater ponding are different problems; a building can sit outside a river-flood extent and still have a vulnerable basement entrance or a road that becomes impassable after intense rain. A map can also be older than the drainage works, land filling or surrounding development that now shapes local runoff. The correct reading is therefore “no relevant signal in this particular product” rather than “no flood risk.”

Location itself needs provenance. MLMUPC’s land-information service can help anchor a parcel or property document when the certificate details are available. OpenStreetMap is useful for finding roads and buildings, but it is community maintained and is not an official cadastral or hazard source. A marketing pin deserves even less authority unless it can be matched to the actual tower, entrance or parcel. Before any map comparison, the buyer should know what point is being compared and why that point represents the property.

Used this way, public maps are valuable because they narrow the uncertainty instead of pretending to remove it. They can establish river conditions, a dated flood footprint, an administrative history or a modelled risk context. They cannot certify the drainage design, maintenance or flood performance of a particular building. That second question requires evidence from the property itself.

No single source answers every property question. Match the source to the claim you actually need to test.

A map passport is a guardrail against over-reading the picture. For every product actually used, record the publisher, exact product name and direct page, the date or observation period, whether it is observed, historical, forecast or modelled, the flood mechanism if stated, geographic coverage, spatial or administrative scale, legend, units and thresholds. Downloadable GIS data also needs resolution, coordinate reference system, metadata and reuse terms. An undisclosed parameter should remain “not publicly established,” not be replaced with a plausible guess.

NFFC shows why this matters. As checked on 29 September 2026, its public Flood Extent Maps page lets a user select a year and month and displays mapped products for the chosen period. The text interface does not itself expose the coordinate system, spatial resolution, production method or a complete machine-readable legend for the specific image. The page can establish that a flood-extent product exists for the selected period, but it does not by itself justify parcel-scale measurements. If metadata for the exact map cannot be found, record the missing fields and stop the precision there.

MRC products need a different passport because the service type changes the meaning of every number. Near Real-Time Monitoring is station based and presents hydrometeorological observations and water-level charts. During the wet season, MRC river forecasting uses a broad station network to issue forecasts for points on the Mekong mainstream. Flash-flood guidance is a separate operational service for a faster, smaller-scale hazard. A station value is not interchangeable with a mapped flood footprint, and a forecast point is not a proxy for water depth at an apartment building. If the legend says water level, call it water level.

CamDi belongs in the historical-evidence column. Its current public profile covers 1996–2023, and the system is designed to analyse registered disasters and losses across administrative levels. It can show whether floods and impacts were recorded in an area, depending on the query, but it does not resolve the history of an individual basement or entrance. Absence in the record is also only an absence in that record, not proof that no relevant event occurred.

The WFP/NCDM assessment is neither a loss register nor a building inspection. WFP’s 2024 Cambodia climate-disaster risk assessment, published on its site in February 2025, is accompanied by maps presenting flood and drought risk and the main drivers—hazard, exposure and socioeconomic vulnerability—at commune level. For a buyer, that administrative resolution is decisive: a commune-level classification remains commune-level even when the map can be zoomed to a street.

The passport should also state what “flood” means in the product. River flooding, flash flooding and urban rainwater or drainage flooding are different mechanisms. Phnom Penh drainage studies by JICA help explain urban stormwater systems in the capital, but should not be projected onto the rest of Cambodia. If a source uses a generic flood label, the limitation belongs in the passport.

Coordinate systems matter when the analysis moves from description to measurement. A point-in-polygon test can show that a property coordinate intersects a polygon, provided the geometries are compatible and accurate enough; it does not create a probability or building-level depth. Distances to a boundary should not be reported with metre-level precision when resolution is unknown. Elevation comparisons are stricter still: site level, road level and modelled water level need a compatible vertical datum and traceable source. A phone altitude is not a substitute for surveyed or documented elevation data.

If NovAsia reproduces a map or layer, provenance and reuse rights belong in the passport too. Authority and fitness for purpose are separate tests. An official product can be reliable in origin and still be too coarse to answer a question about one condominium entrance. The passport makes that boundary visible before a coloured polygon becomes an unjustified conclusion.

Once the map has done its job, the investigation should move downstairs—literally. A completed building can be tested against records of what happened at the entrance, basement, car park, plant rooms, lift pits and the road immediately outside. Dated photographs after heavy rain, incident logs, maintenance records and documented access closures carry more weight than a generic assurance from a salesperson. They show how water interacted with the property, not merely how the surrounding area was classified.

Pumps are only useful evidence when their operating context is known. A buyer needs to understand whether drainage pumps are maintained, whether they have backup power and which systems remain live during an outage. Temporary flood barriers can be sensible engineering, but their existence says little without knowing when they are deployed and what opening they protect. The same applies to discharge: a drain grate near the ramp is not evidence that water can actually leave the site under the conditions that matter.

Access deserves its own line of inquiry. A tower can keep its lobby dry while a low section of the approach road holds water long enough to disrupt cars, deliveries or emergency access. That outcome is materially different from water entering the building, and it should not be hidden inside a single yes-or-no history of “flooding.” A visit during or shortly after heavy rain can reveal low points and drainage behaviour that are invisible in dry weather, although one visit is still only one event.

An off-plan project has no operating history, so the evidence shifts from records to design documents. Site and entrance elevations, stormwater routing, the discharge point, basement-ramp protection and the design of pumping and backup systems are more informative than a brochure phrase. Elevation figures also need a reference system. A claimed difference between the site and the road is not engineering evidence unless both numbers share a traceable datum and the locations being compared are clear.

Mechanism matters here as much as it does on a map. A high river, a rapid flash flood, intense rainfall that overwhelms urban drainage, water entering through a poorly protected ramp, and leakage through the building envelope are different failure modes. JICA’s drainage work is useful for understanding the Phnom Penh stormwater context, but it should not be treated as a nationwide description of Cambodian cities. A river-flood product may be relevant to one address while saying very little about the rainwater behaviour of the street outside it.

Evidence should also be internally consistent. A pump-maintenance log is stronger when the equipment can be identified and the service dates make sense. A photograph is stronger when the location and date are known. Resident accounts are more useful when several independent accounts refer to the same event rather than to an undefined memory of “bad flooding.” If the management office cannot provide a requested record, that absence is itself part of the evidence set, but it should not be turned automatically into a claim that the system failed.

This flood-focused review is narrower than full operating-building due diligence. The wider inspection still needs to cover legal status, management, lifts, power, leaks, fire safety, maintenance and other systems. For this page, the goal is more specific: connect the public flood context to the parts of the property that determine whether water becomes a real building problem.

The cleanest conclusion has three layers rather than one verdict. First, state what the public product establishes: the exact layer, its period, its event type and the scale at which it can be interpreted. Second, state what is known about the property itself: documented incidents, entrance and car-park performance, pumps, backup power, design elevations or access. Third, list the material unknowns that remain. Keeping those layers separate makes it clear which claim comes from which evidence.

Where the geometry is good enough, an address can sometimes be described as inside, outside or on the edge of a particular mapped polygon. The wording should remain tied to that product. “Outside the mapped extent for this dated layer” is defensible; “this building does not flood” is a much larger claim about a physical asset across future events. If the source is only commune- or district-level, zooming the web map does not create building-level resolution.

Property evidence also works better as facts than as a score. A documented basement incident, repeated pump failures and temporary loss of road access are three different findings. Their importance depends on the buyer’s use of the property and on whether the causes have been corrected. Conversely, an absence of documented incidents means only that none were established in the evidence reviewed. It is not a certificate of future performance.

For an off-plan building, design information should be described in the same disciplined way. A drainage plan, a protected basement ramp and stated site elevations may show that the project has addressed certain pathways. They do not prove how the systems will be installed, commissioned, maintained or operated years later. The gap between design intent and operating evidence should stay visible.

Some uncertainty will remain even after a careful review. A future storm may differ from the events represented in a historical layer. External drains can be blocked or upgraded. Surrounding sites can change runoff. Maintenance can improve or deteriorate. The purpose of the exercise is not to turn those uncertainties into a single probability without a defensible model; it is to identify which uncertainties matter for this property and which can still be reduced.

A useful buyer-facing conclusion therefore sounds specific: this product provides this level of flood or water context for this period; these building-level facts are documented; these items could not be established. That is enough to challenge an overconfident sales claim, request better evidence or decide that a particular unresolved issue deserves more attention. It is not a NovAsia “flood score,” and it should never be presented as a guarantee that the address will or will not flood.