Validation Method Overview
MODIS and VIIRS dark target aerosol products for land and ocean are validated by comparing Aerosol Optical Depth (AOD) values against AOD version 3 data from AERONET ground based instruments. You can learn more about AERONET here. The AERONET AOD values are considered to be "true" and the satellite values are compared against the AERONET values on a site by site basis. Statistics for the global comparisons are found on the Validation Results page. Results for all of the individual sites can be found on the Validation Maps page. Currently validation results are shown only for satellite AOD at 0.55 µm. AERONET values at 0.50 µm are interpolated to 0.55 µm for these comparisons.
Temporal and spatial matching of the satellite and ground data as well as data quality screening (see below) is employed for all points used for the validation analysis. MODIS and VIIRS DT aerosol products are considered "validated" when 67% of the points fall within a bounding envelop of "expected error" (EE) as compared to AERONET. The comparison data set of satellite vs AERONET points is what is used to determine EE. That is, we adjust the bounds of the EE until at least 67% of the points fall within the bounding envelope. Thus EE can and does change for different data collections and sensors..
The expected error envelopes for C61 MODIS and the VIIRS DT products have been calculated as:
Expected Error (EE) = +/- (0.05 + Tau * 15%).
The collection 5 validation included data from the AERONET Maritime Aerosol Network (MAN) for global or regional comparisons. This data set was not included for the MODIS C61 and VIIRS validations.
Validation Method: Data Selection
AERONET instruments measure AOD using two different protocols. One of these derives AOD by measuring incoming radiation from the direct beam of the sun while the second protocol uses diffuse radiation measurements from various directions in the sky to derive the AOD. These measurements are made several times per hour the exact number depending on the angle of the sun. Additionally AERONET AOD data is computed for three different quality levels: Level 1 (unscreened), Level 1.5 (cloud-screened) and Level 2 (cloud-screened and quality assured). The AERONET data used for MODIS AOD validation is only from direct sun radiation measurements of Level 2 quality.
AERONET instruments collect data in multiple wavelengths many of which are slightly different from the MODIS channels. AERONET data using their 0.44, 0.50, 0.67, and 0.87 µm channels is converted to MODIS equivalent channels using a quadratic interpolation. Comparisons are made at the MODIS 0.47, 0.55, 0.67, and 0.87 µm wavelengths. In the same way that the satellite data processing is updated and organized into collections the AERONET data processing has different versions. Version 3 is the most recently updated data set for AERONET (as of 2019) and has been used for all comparisons.
MODIS data comes with a quality assurance (QA) designation with values ranging from 0 to 3 where 3 is the highest quality. For Land validation only QA 3 data is used. For Ocean validation all QA above 0 is used.
Validation Method: Spatial and Temporal Data Matching
For a point to be including in the validation analyis the AERONET and MODIS measurements must be matched in space and time. For the spatial matching we draw a circle of fixed radius around the aeronet site of 27.5 km for the 10 km product and 7.5 km for the 3 km product. We require a minimum of 20% of all possible satellite retrievals of the required QA within this area to use this point in the validation data set. We average all qualifying MODIS AOD values to compare against the AERONET AOD. For the temporal matching we require at least two AERONET direct sun level 2 measurements within 30 minutes of the MODIS overpass ( a temporal window of 60 minutes). We average all qualifying AERONET measurements to compare against the MODIS AOD.
The drawing above shows the selection criteria for the spatial and temporal matching of MODIS satellite and AERONET ground measurements used in our validation studies for the 10 km product. For the 3 km product the subset surface circle diamter would be approximately 7.5 km at nadir. Image courtesey of Maksym Petrenko.