What Phnom Penh Air Stations Say About Housing

Phnom Penh does not have one permanent station count that can be quoted without a date and a definition. In an April 5, 2026 report citing the Ministry of Environment's first-quarter statement, the Phnom Penh Post recorded 59 monitoring locations nationwide and 12 fixed stations in the capital. That is a useful Q1 2026 snapshot, not a timeless inventory.

GGGI's earlier assessment described a different configuration: 50 monitoring stations across Cambodia, including 10 reference stations and a wider network of lower-cost sensors. The same assessment treated reference equipment and sensors as different parts of the monitoring system, with different maintenance and quality-control needs. A count of 50 in an earlier technical report, 59 locations in a 2026 ministry statement and 12 fixed Phnom Penh stations can therefore all be internally consistent once the date, geography and station class are kept separate.

As of September 29, 2026, the Ministry of Environment's public air-quality service is online, but the web interface loads its data through JavaScript. The accessible static page does not expose a complete live census with coordinates, instrument models and last timestamps for every site. A current Cambodia Ministry of Environment contributor profile and September IQAir city maps do, however, confirm several MoE points in Phnom Penh, including the Institute of Technology of Cambodia, Wat Langka Front Garden, the University of Health Sciences, Chroy Changvar Cannon Roundabout Garden, Ounalom Pagoda Front Garden and Makro Superstore Compound.

The table below should therefore be read as a dated, confirmed subset rather than a replacement for the ministry's live registry. It also keeps an independent AirGradient sensor visibly separate from MoE sites. This distinction matters because an aggregator can place government, school, private and community sensors on the same map, while the number of visible pins says nothing by itself about even spatial coverage across the city.

A map pin answers a limited question: a data source exists there, or existed there recently enough to remain visible. For housing comparisons, that is not enough. You also need the provider or owner, the instrument class when published, when reporting began, the latest timestamp and whether there were long interruptions. Otherwise two neat-looking averages may represent different operating periods rather than different neighbourhood conditions.

For the MoE sites visible through current IQAir contributor and city pages, there is clear evidence that the network is reporting in 2026; many of those pages mark the stations as established in April 2026. What is not exposed through the ministry's static web view is a complete raw history for each site. This page therefore does not invent a 90-day completeness percentage, and it does not assign precise first or last observations where the source does not publish them. A current value on an aggregator helps establish that a feed is alive, but it is not a substitute for a gap history.

The AUPP Liger Leadership Academy sensor in OpenAQ shows why that distinction matters. OpenAQ identifies it as a stationary AirGradient ONE Generation 9 measuring PM2.5 among other parameters and reports data beginning on August 15, 2025. At the September 29, 2026 check, however, the page described its latest update only as about seven months ago. A sensor can therefore remain discoverable on a map long after it has stopped being useful for today's comparison.

A missing period is not evidence of clean or dirty air; it is simply missing measurement. The same logic applies to averages: sites with different operating windows should not be ranked mechanically. Cambodia's GGGI-supported QA/QC procedure explicitly treats maintenance, calibration, data review and siting documentation as part of producing reliable monitoring data, which is why freshness and provenance matter as much as the number shown on the map.

Snapshot: September 29, 2026. This is a confirmed portion of the public network, not a claim to reproduce the full official registry: the MoE service does not expose the complete live census in its accessible static interface. Government sites are attributed to the operator, but no instrument class is assigned unless published.

PM2.5 concentration and AQI are not interchangeable. PM2.5 concentration is a physical measurement, normally expressed in micrograms per cubic metre. AQI is an index calculated from pollutant concentrations using a particular set of breakpoints and rules, then often translated into a colour or category. A phrase such as “AQI 25 µg/m³” therefore mixes an index with the unit that belongs to concentration.

The distinction becomes important as soon as two apps are open. IQAir can display US AQI⁺ alongside PM2.5 concentration, while the U.S. EPA calculates its AQI by mapping pollutant concentrations to defined index breakpoints over specified averaging periods. That does not establish the formula used by Cambodia's official service. In the primary material accessible for this review, we did not find a document robust enough to reconstruct an official Cambodian AQI formula from colours or from a third-party platform.

For station-to-station comparison, the cleaner starting point is one pollutant, one unit and one time window. An hourly PM2.5 reading at one site should not be compared with a 24-hour average at another as if the difference were purely geographic. A quarterly city average is further removed again: it cannot be placed beside a current reading near a building and treated as the same kind of observation.

Aggregator pages also show how dynamic the visible network can be. September 2026 IQAir snapshots for Phnom Penh displayed different numbers of available stations and contributors at different update times, ranging from the high teens to more than twenty visible points on individual crawls. That does not mean physical monitors are being installed and removed every few hours; feeds, providers and current availability change what the platform can show. Matching colours across two apps therefore do not prove identical source data, and different colours do not automatically prove that one sensor is wrong.

When an index methodology is unclear, the most defensible common denominator is the labelled PM2.5 concentration with its date, time and averaging period. AQI remains useful as a communication layer, but it is not a universal measurement unit that can be moved between platforms without checking the scale behind it.

The nearest station is often a sensible place to start, but distance does not create a universal reliability radius. Two addresses a kilometre apart can sit in different local environments: one beside heavy traffic or active construction, another on a more open street with better airflow. The reverse can also happen, where a slightly more distant station is a better indicator of the broader urban background because the closest monitor is strongly affected by a hyperlocal source.

The U.S. EPA's sensor-siting guidance describes the same technical principle. Monitoring goals matter, and buildings, trees, dusty roads, exhausts and other local sources can change how representative a site is. A sensor intended to describe broader ambient conditions needs free airflow and should not be dominated by one highly local source. This is international technical practice rather than Cambodian law, but it makes clear why a rule such as “within two kilometres is reliable” would be artificial.

For a condo near Wat Langka, the useful question is not simply how many metres separate the building from Wat Langka Front Garden. It is where the monitor sits relative to traffic, surrounding structures and airflow, and whether the building faces similar conditions. The same reasoning applies in Chroy Changvar, Dangkao or Mean Chey: a short line on a map shows geographic proximity, not a pollution-dispersion model.

For a housing decision, it can also help to separate the station question from infrastructure and local effects around a property. Construction dust, machinery, congestion and temporary works can create building-level conditions that a city monitor was never designed to represent.

Distance should therefore remain context, not a precision score. If a network does not publish the purpose of a site, its installation details or enough information about its surroundings, there is no defensible basis for assigning a condo the value of the “nearest station” or ranking neighbourhoods from a handful of heterogeneous pins.

Monitoring is most useful when it separates three levels of a housing decision instead of pretending to rank buildings. At the city-and-day level, several fresh, comparable stations can show whether a PM2.5 episode is broad or confined to one feed. Directional agreement matters: if multiple stations rise together over the same window, that is stronger evidence of a citywide change than a single isolated spike.

At the neighbourhood and street level, the question changes. Is there an active station nearby whose setting resembles the environment you care about? A major road, construction site, generator, open burning source or exhaust can make conditions around one façade different from the broader background. If you are comparing BKK1 with Tonle Bassac, uneven station coverage is not evidence that one area is cleaner; it is a limit on how confidently the network can support that claim.

At the apartment level, the outdoor station becomes only one input. EPA guidance on indoor particulate matter notes that outdoor particles enter through windows, doors, cracks and ventilation, while indoor concentrations also depend on filtration and sources generated inside. Two units in the same building can therefore have different air-exchange patterns even under the same outdoor conditions. This is where how floor level changes the conditions of a specific apartment can be relevant without turning height into an automatic “cleaner air” claim.

The same caution applies to the façade and windows. Window orientation and apartment ventilation influence contact with outdoor air and the way the unit is aired, but they do not produce a PM2.5 number on their own. If indoor air is a decisive criterion, a properly placed indoor sensor observed over a representative period provides evidence that an outdoor station cannot supply for a 25th-floor living room.

The practical value of the map is therefore straightforward. It can answer “what is happening to outdoor air in the city and around this part of town?” reasonably well when the data are fresh and comparable. It is weaker at “what is happening at my façade?” and cannot answer “what is the air inside this apartment?” without indoor evidence. That boundary does not make the monitoring network less useful; it makes the network useful for the questions it was actually built to answer.