Alex, the wind rose is a tool that can be used to partition dispersion patterns from a source and as related to any downwind receptors. While it allows for some diagnosis of the probability of impacts downwind it is not the same as applying wind information in a numeric AQ model (e.g., EPA dispersion models). The use of nearby locations or sites (data at "points where air quality data was collected") is often used in AQ models but again it depends on what you are hoping to accomplish.

Sir, I think the application of the wind rose diagram is know the prevailing direction which air pollutants plume follows ie the direction or trend where most or the least of the  pollutant plume takes.

Some models may use this information to analyze the dispersion of pollutants of a local source at the study área.

You will be better off constructing a pollution rose to assess the directional impact of wind transport from the source. To do so, you may plot the concentration levels against the wind direction and obtain the frequency distribution of concentrations versus wind direction. This exercise can be done for various wind classes to assess the impact of wind speed on concentrations and transport.

You may use, I think  Hysplit NOAA model. It's a good tool which  shows the trajektories of particles. It will be an important support for your predictive model.

Alex:

Conventionally, a wind rose shows the direction from which the wind blows. However, some software, allow you to show it in the direction to which the wind blows. I usually leave my windrose to show the direction to which the wind blows (for easy interpretation). I obtain the map of the study area and then plot the coordinates of the air quality sampling locations on the map. Sometimes too, I do indicate the concentrations of the air quality parameter(s) on the map. I then superimpose the windrose plot on the map to give a pictorial view of the directions to which the wind will carry the air contaminant. Most often than not, people concentrate their analysis on the predominant wind direction. However, the other directions  the windrose may show can also be of significant concern. You may also consider using pollution rose.

On your investigation of the influence of wind speed on the air contaminant, the higher the wind speed, the more likely your monitors will pick lower concentrations. This is because, the wind will cause dilution of the contaminant. This case is also true when you are doing air particulates study. But there can also be the issue of higher particle (especially dust) concentration with increasing wind speed, as increasing wind speed can cause an already settled dust particle to become airborne.

I hope these highlights help. Let me know if you require further information. I can also send you some materials if you provide your email address.

Side note: In some long-term cases/studies any 'calm' wind reports (either light and variable or zero wind speed) are distributed among the wind directions in order account for all observations (and relative frequencies).

Alex, I suggest using wind rose with 16 wind sectors (direction)  to have more resolution. I agree in any cases with R. Bellasio that this indicates more local situation.

Now, you should be aware that wind direction change with altitude (especially in rough terrain) so surface wind direction might not be representative of the above surface layer flow direction and therefore the direction from where the pollution comes from (i.e., the source regions). So you should complement your wind rose analysis with HYSPLIT NOAA back-trajectory  tool: http://ready.arl.noaa.gov/HYSPLIT.php

Do you have rough topography or flat terrain ?

In case of rough terrain around your measurement site, both back-trajectories or wind rose are questionable.

Another method is analyzing your air quality data by synoptic situation or by weather parameters (windspeed, temperature, pressure, etc.)

Good luck