All Collections
Air quality information
How accurate is AirVisual’s data in Vietnam?
How accurate is AirVisual’s data in Vietnam?
Written by AirVisual
Updated over a week ago

In Vietnam, AirVisual reports data from both governmental and non-governmental monitors. AirVisual clearly displays all the data source contributors for each location on our air quality app. Simply search for a location and click on it to find the monitoring stations for that area. Explore each one to find data source information.

All data are constantly monitored and validated by AirVisual’s cloud-based data validation system before being published on the AirVisual platform. The validation system has been built using Artificial Intelligence (AI) and machine learning, and processes billions of air quality data points in order to effectively identify and stop any unusual or anomalous data from being published. 

Government monitoring stations
Government monitoring stations, such as those run by the Hanoi Environmental Monitoring Portal and the U.S. Embassy in Hanoi, are typically high-cost “reference monitors.” These are generally considered the most accurate and reliable source of measured air quality data. However, government sensors can also experience problems and report inaccurate data, such as a sudden high pollution reading. Reasons for this may include temporary periods of maintenance or defects, or temporary hyperlocal emission sources near the sensor.

The AirVisual system therefore puts all government sensor data through a data validation system before publishing. One example of this validation process: 

  1. From 1 p.m. to 3 p.m., a monitoring station has been reporting hourly measurements of PM2.5 of 10 micrograms per cubic meter (ug/m3), 12ug/m3, 10ug/m3. Then at 4 p.m., it publishes the following data: a PM2.5 reading of 100 ug/m3

  2. AirVisual’s cloud-based system notices that this is a very big difference from the pattern of readings recently reported by that station. It flags the data as having a potential anomaly and doesn’t immediately publish it. 

  3. The AirVisual system cross-checks the 100 ug/m3 reading with the measurements from nearby stations: are they also reporting similarly high readings? 

  4. It cross-checks the reading against historical patterns: is it usual at 4 p.m. for the air at this location to become more polluted? 

  5. It considers other parameters such as weather conditions. Is there a weather-related reason for there to be a sudden spike in pollution? 

  6. Based on the results of these cross-checks, the AirVisual system will decide whether the data reading is correct or not, and publish or discount it accordingly.

Non-governmental monitoring stations
Other sensors are also subject to a data calibration and correction process, in addition to the validation process described above.

How does the calibration system work?

It takes into account:

  1. Environmental conditions such as temperature, humidity, pressure and wind speed/direction. For example, high humidity levels may under some circumstances lead to low-cost sensors over-reporting levels of PM2.5.

  2. Regional historical patterns

The AirVisual system then applies a data calibration algorithm based on the above, and adjusts the PM2.5 measurements if necessary. 

The adjustment level is determined by the cloud-based system that is built on artificial intelligence and machine-learning. The AirVisual system learns from its years of aggregating billions of global air quality data points from reference sensors, AirVisual sensors, meteorology data, and pollution composition from satellite imagery.

If you think a station has a problem
All data that is published on the AirVisual platform has gone through a rigorous checking process. Through years of experience processing billions of air quality data points, AirVisual has built the most advanced air quality data validation system to ensure the best data accuracy and reliability possible. However, if you find something looks unusual, AirVisual offers users the possibility to report a potential problem with a station. Our dedicated data quality team will then look into the report, to check if there are any issues with the data source. This ensures double data validation: both from the system, and from the community. 

Did this answer your question?