Your question is not quantitative

What is poor?  in what region is the PM; in what circumstance is there a poor correlation. For what do you need this information.

For specific regions I am sure you can find your answer when you read the papers

I recommend dr Schaap's publications for Europe.

Where are you doing these measurements your PM2.5 data is measured using a cascade impactor or by an IR based instrument or Laser?

Harry is right what is poor, please state the percentage and correlation factor and is this consistently off across seasons? 

Dear @Harry Sir, the poor correlation was obtained between MODIS AOD and surface PM2.5. The value of correlation coefficient (r) was 0.17. PM2.5 data was collected through Fine dust sampler on 24-h basis at Varanasi, India a location at Indo gangetic plains. the period of monitoring was winter (Jan, Feb, March). The purpose of such study is to know the relationship between satellite retrieved AOD and surface particulates in winter as the evidences of biomass burning increases during winter. I am going through the paper of Dr. Schaap recommended by you.

Dear @Saif Uddin sir, we did not study for different seasons. It was only for winters.

I think there is quite often fog in that region in winter: how well are the images screened for that. Also: AOD is only available for daytime and you have 24-hr PM data. What is the diurnal variation in PM or rather what difference is there between day and night time PM?

Dear sir, you are right. there were a number of days with dense fog during the study period.  however, we did not study the diurnal variation because we have monitored PM mass concentrations through fine particular sampler continuously operated for 24 hrs. I think this could be one of the reason behind this discrepancy. But sir as we had studied under those given conditions and resources and found such results, should we report it for publication by mentioning all the limitations and circumstances? 

Thanks Dr Backer

May be you can discuss the 24 hr PM2.5 variation with local meteorology and discuss the possible sources of the same.

And AOD from MODIS is for the whole column also you are taking only PM2.5 which is not the total aerosols.

And I completely agree with Prof. Brink and Prof Backer.

Parth,

Yes you are right. We have done the things in similar ways. Thank you for your suggestions.

a. AOD is an optical property while PM is a surface one

b. PM 2.5 represents a portion (depending on the time/area) of total aerosols

But if we are talking for a well mixed and not highly variable (with height)  layer, with no transported (high elevated) and no coarse aerosols then

-the spatial resolution used by modis can be tricky as a 5x5Km modis/aod can be highly uncertain due to clouds and maybe ground reflectance if we are talking about a non homogenous terrain, while the 5x5Km spatial variability of PM 2.5 over a city can be also high. Finally, time coincidence (and not single satellite overpass time vs 24h mean) could also help.

There can be a lot of discussion on the above, but i hope they are helpful.

The simplest (but least helpful) answer is we don't always know what causes discrepancies between AOD and PM2.5. 

There are at least three influential factors to the PM2.5/AOD relationship, namely the relationship to ground-scatter and total column extinction, midday vs daily ground scatter measurements, and scattering efficiency (as in aerosol composition/size distribution).

From what I can tell, the relationship between ground and column is often most important; vertical distribution and composition plays a big role, but midday biases less so. 

There are also issues with how representative the air sampling stations(s) is/are compared with the area measured by satellite. If the surface PM distribution is concentrated in urban areas but low in surrounding rural, the satellite-measured AOD may smear out localized peaks. Also particular mountainous areas and small islands are challenging.

On top of that, sometimes it's possible the instruments are wrong! say if filters are improperly handled (e.g. not postweighed at low enough RH, size cut/flow problems, or filters sufficiently heated to lose semi-volatiles).