I'm working on a project aiming at finding some potential point as a new air pollutant monitor site for a city to measure $SO_2$, $NO_2$, $PM_{2.5}$, etc.
After modeling the surface concentration of the area, here is my select standard:

  1. The violation situation

    Finding the area which exceed the local air quality standard most in the modeling area.

  2. The representation area

    Each modeling grid concentration is a time-series dataframe. Compute the
    spatial correlation coefficient (r) between one grid with surroundings area,
    we can get each grid's represent area assuming there is a cut-off coefficient $r_c$.

    For example, 3 grid i, j, k in the city.
    $r_(ij) > r_c$ → $grid_i$ can represent the temporal variation of $grid_j$
    $r_(ik) < r_c$ → $grid_i$ can't represent the temporal variation of $grid_k$

  3. Population density

    Considering the environmental exposure healthy risks, I add the pop-ind data for each concentration grid

Here is my question

  • Are these principle enough for a monitoring site?
  • How to combine the results from the analysis based on multi-air pollutant data?
    ( In practice, one monitoring site often measure various air pollutants simultaneously, but my method is based on a specific type of air pollutant owing to the different emission source & chemical reactivity)

  • Combining each site, We can get a monitoring network, how to estimate the network's strengths and weaknesses?


1 Answer 1


Each one of the monitor siting examples you list has a different purpose. As you have listed, a monitor will be sited for air regulatory compliance purposes (e.g. determining the highest concentrations downwind of a major source), understanding the region (e.g. representative site in well-mixed location), or health exposure risks (e.g. densely populated areas). From a health and regulatory perspective, those types of site locations are sufficient. However, you may want to consider other monitor site locations in your network that will help to evaluate the performance of the air quality model and assess the importance of various sources:

  1. Monitors at a variety of elevations, so that vertical mixing can be evaluated (or use sondes)
  2. Monitors both upwind and downwind of major sources, so that their contributions can be determined to a greater accuracy than represented in the original emissions inventory.
  3. Monitors behind complex terrain, which affects transport, and can be difficult to model correctly.

Air quality modeling helps to fill in the gaps left by monitoring networks, since they are expensive to site and maintain. Monitor networks are relatively sparse compared to the number of major source and receptor locations. So, it is important to carefully consider each site location and pick ones that satisfy multiple criteria. Satellite retrievals and local testimonials can be used to evaluate the initial value of your network. However, you will benefit from quality controlling your model in the areas where your model performs poorly.

For reference, here is an example of how modeling and air monitors can be used together to make quality forecasts. Environment and Climate Change Canada’s FireWork air quality (AQ) forecast system for North America with wildfire emissions has a post process called UMOS-AQ (see below). The Canadian researchers built a system that modifies their air quality predictions based on recent surface monitor model/observations comparisons. The Regional Air Quality Deterministic Prediction System (RAQDPS)uses GEM-MACH (Global Environmental Multi-scale - Modelling Air quality and CHemistry) and BlueSky Canada. Here is a paper that describes the process.

The RAQDPS includes other components besides GEM-MACH. One postprocessing product is the Updateable Model Output Statistics for Air Quality (UMOS-AQ) package, which applies statistics for bias correction to compensate for systematic AQ model errors and account for unresolved subgrid-scale phenomena at particular locations, namely the locations of AQ measurement stations (Wilson and Vallée, 2002; Antonopoulos et al., 2010; Moran et al., 2012). Hourly model forecasts of both pollutant concentrations and meteorological quantities at specific locations are combined with available hourly surface AQ measurements to generate location-specific hourly statistical forecasts for areas where AQ is measured. The location-specific O3, PM2.5, and NO2 concentrations calculated by UMOS-AQ are also the ones used in the calculation of the AQHI values that are disseminated to the public.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.