The cloud screening procedures for the two products use similar principles but different tests due in part to different bands used in the retrieval algorithms and different pixel resolutions used in algorithms. Really 'cloud mask' is not 100% correct as a term, they are more like a 'suitable pixel mask', and both algorithms have different requirements for pixel type (e.g. Dark Target has  a surface brightness criterion).

One reason is that AOD is normally low in these areas so AOD variability is similar in size to the retrieval error.  Another reason is that these areas may have bright surfaces and/or atypically colored soils which the two products will retrieve differently.

The cloud screening in the MODIS dark target aerosol algorithm is a series of internal and external tests, and is different over land and ocean. The details of the cloud masking scheme is described in section A2.2 of the ATBD. The Collection 6 aerosol products contain a new SDS called “Aerosol_Cldmask_Land_Ocean” which characterizes each 500 m input pixels as “clear” or “cloudy”

The MODIS Dark Target algorithm uses a set of ratios and relationships between the 0.47, 0.67 and 2.1 µm channels to account for the surface signal.

These relationships begin to break down over the brighter surfaces of urban areas.   Improving the product over cities is an active area of research.  Please contact us if you would like more information.

There are several possible reasons why the AOD product is less accurate in the Western U.S.  Some of these reasons include: bright land surfaces, high altitudes and or rapidly changing topography, and potential inaccuracies in the MODIS aerosol models for the Western U.S.

Currently there is no definitive test or product parameter available to conclude that the MODIS aerosol product is measuring dust or smoke.  Over ocean a low fine mode fraction is a strong indicator of dust but is not definitive.  Often ancillary information must be included to try to make this determination.  This ancillary information can include: AERONET data, AIRS dust product (over ocean), MISR data, OMI aerosol index, VIIRS suspended matter product, back trajectories which link to a dust or smoke origin.

The dark target retrieval algorithm assumes that aerosols over a surface will brighten the scene.  Where the surface is too bright this assumption breaks down. 

The merged product is a single SDS (AOD_550_Dark_Target_Deep_Blue_Combined) comprised of only high quality dark target (QA=3 over land, QA > 0 over ocean) and deep blue (QA 2 & 3) data to produce a global 10 Km product. Over ocean this product uses only dark target retrievals. Over land, monthly NDVI maps are used to assign which retrieval will fill the merged SDS. Over bright surfaces (NDVI < 0.2) the deep blue value is selected. Over dark vegetated surfaces (NDVI > 0.3) the dark target value is selected.

This is a big question and we can only provide very general answers in this FAQ format.

Dark target has separate algorithms for land and ocean.  Deep blue in the MODIS aerosol products is a land retrieval only. All total column aerosol retrieval algorithms must account for and remove the surface reflectance signal to accurately determine the aerosol signal. Dark target and deep blue retrievals have different ways of accounting for the reflectance signal coming from the land surface.