# MODIS Surface Reflectance Data State QA Aerosol Quantity field meaning is unclear

I am trying to compare NDVI values computed from MODIS Surface Reflectance(MOD09) with the NDVI product (MOD13).

MODIS Surface reflectance Collection 5 products include a Data State QA field that I am using to estimate which pixels are best suited for computing NDVI.

A part of this field is labeled Aerosol quantity. The values in this field seem to significantly affect the result of NDVI computation.

The explanation of the possible values of this field in the MOD09 Users' Guide seems not entirely clear to me.

         | 00 - climatology
aerosol  | 01 - low
quantity | 10 - average
| 11 - high


The last three values are clear, while the meaning of the climatology value is somewhat non-intuitive.

I have selected a crop field and calculated NDVI for all pixels clear of clouds or cloud shadows during one vegetation period and observed the following:

It seems that results with low aerosol quantity (green points) match well with the results obtained with the complex rating algorithm used in the standard NDVI MOD13Q1 product. While the data with climatology state (orange points) seems far off.

What does the climatology attribute mean and how to treat it?

• This probably means that the retrieval had insufficient information and the product is based entirely on the a priori, which is obtained from climatology. See Bayesian retrievals and how they combine a priori information with information added by the measurement. – gerrit Apr 3 '17 at 13:19
• I agree with @gerrit . This assumption is reasonable because the data with "aerosol quantity" climatology performs worst which one can expect when the cases use actual aerosol retrievals. – daniel.heydebreck Apr 4 '17 at 16:15
• @gerrit in PGE11 code that generates MOD09 L2 product "climatology state" is equivalent to "AOT (aerosol optical thickness) has climatological values" in Internal cloudmask. So that basically means MODIS aerosol retrieval was not performed for that spot. Could you please post your comment as an answer that I could accept. – Dmitri Chubarov Apr 5 '17 at 10:44

Most likely, climatology means there was no retrieval at all.

Bayesian retrievals combine information from an a priori with information from measurements. When the retrieval fails for whatever reason, or the measurement contains insufficient information, instead of reporting no measurement at all, they copy over information from the a priori and use the flag to indicate that the reported "measurement" is in fact simply the climatology.

For details, see optimal estimation, in particular the book by Rodgers:

Clive D. Rodgers (2000). Inverse Methods for Atmospheric Sounding: Theory and Practice. World Scientific.

Gerrit is correct, the climatological setting indicates that the value is filled in (not an actual retrieval). If you see the MODIS Vegetation Index User’s Guide, it is discussed a few times:

Cloud-free global coverage is achieved by replacing clouds with the historical MODIS time series climatology record (Fig. 5)

and

In the Climatology Fill case, certain highly dynamic regions may show discrepancies where fill values were used. This is most obvious when missing input tiles are replaced, where edges may be visible. For pixels filled due to high cloud contamination, this fill strategy will perform well. The fill completes the two VI layers with data. Other layers will contain their respective fill values, except data layer 11 (#1km pix used), which is set to 0, i.e., no good input data.

and they give an example too, for African dust: