UMass Boston

MODIS Spatially and Temporally Complete 30 arcsecond BRDF/Albedo Product

Please cite the paper below or appropriately acknowledge the dataset authors if your use of the data results in a publication. These data are available for download from the LPDAAC: https://lpdaac.usgs.gov/products/mcd43gfv006/.

Sun, Q., Z. Wang, Z. Li, A. Erb, and C. L. B. Schaaf (2017), Evaluation of the global MODIS 30 arc-second spatially and temporally complete snow-free land surface albedo and reflectance anisotropy dataset. Intl. Journal of Applied Earth Observation & Geoinformation, 58, 36-49, doi:10.1016/j.jag.2017.01.011.

Product Summary:

The SZA-extended Gap-filled Snow-free BRDF parameters product is based on the 30

arcsecond gap-filled snow-free BRDF parameters product MCD43D. Thus the gap filled parameters products are provided at 30arc second resolution on a global geographic lat/long grid. In addition, the Quality Assurance (QA) map is generated on the same 30 arcsecond resolution grid. The products are stored in HDF format, with one file for each BRDF parameter, of each wavelength, and for each daily retrieval attempt throughout the year. The product includes three parameter files (i.e. isotopic, volumetric and geometric), as well as white sky albedo, black sky albedo at local solar noon, and Nadir BRDF-Adjusted reflectance (NBAR) at local solar noon.

The MCD43 product is not recommended for solar zenith angles beyond 70 degrees Solar Zenith Angle (SZA). However the MODIS Atmosphere Team has a requirement for albedos at SZAs up to 82 degrees SZA for their cloud properties product. Therefore, the data for those areas that exceed the 70 degrees SZA are filed linearly with data the same geographical area from other daily retrievals at lower SZAs.

Methods Overview:

In order to provide spatially complete global maps, the gap-filled BRDF/Albedo products are

developed through the following major steps: (1) high quality observational data are used whenever possible, based on QA flags from the MCD43D product; (2) for pixels without high quality retrievals, a phenological/temporal curve fitting exercise is used to interpolate values; (3) if the temporal fitting is unsuccessful, a spatial fit is used instead; and as a last resort if neither temporal nor spatial fitting work, (4) spatial smoothing is used. To generate a snow-free product, pixels with ephemeral snow were removed using the MCD43D41 snow flags.

The temporal fitting procedure relies on the high quality BRDF retrievals, based on majority vote of the QA values in the input 500 meter resolution MCD43A data that underlies the 30 arcsecond MCD43D data. Lower quality MCD43D pixels made from a mixture of full and magnitude inversions are only used to condition the temporal fit in areas of persistent missing data, and were weighted lower to minimize their effects.

The underlying MCD43D utilizes a BRDF model derived from all available high quality cloud clear reflectance data over a 16 day moving window, centered on and emphasizing the daily day of interest (the ninth day of each retrieval period as reflected in the Julian date in the filename). This 30 arc second BRDF model is then used to produce the daily Albedo and NBAR products (MCD43D). These BRDF model parameters are computed for MODIS spectral bands 1 through 7 (0.47 um, 0.55 um, 0.67 um, 0.86 um, 1.24 um, 1.64 um, 2.1 um), as well as the shortwave infrared band (0.3-5.0 um), visible band (0.3-0.7 um), and near-infrared (0.7-5.0 um) broad bands.

The phenological/temporal fitting is the primary means of data gap filling in MCD43GF. A spatial fitting may still be necessary on the very small proportion of pixels which can not be temporally fit, because the parameter time series was insufficient to complete the temporal fitting. The spatial fitting uses a value borrowed from pixels within 1 degree of latitude. This second method filled the majority of the remaining missing pixels after temporal fitting, but at the end an additional spatial smoothing exercise was performed over the last few scattered pixels that had not yet been filled. The status of each pixel can be found in the Quality Assurance (QA) map:

0 = Best quality, full inversion (Weights of Determination, RMSE majority good)

1 = Good quality, full inversion

2 = Magnitude inversion (numobs >= 7)

3 = Magnitude inversion (numobs >= 2 & < 7)

4 = Temporal fit

5 = Spatial fit

6 = Spatial smoothing

7 = SZA extension between 70 and 82 degrees

Values between 0 and 3 represent the retrievals that do not require gap filling, whereas values 4-7 indicate the gap-filled pixels.

The MCD43GF product includes 67 layers containing black-sky albedo (BSA) at local solar noon, isotropic model parameter (ISO), volumetric model parameter (VOL), geometric model parameter (GEO), quality (QA), Nadir BRDF-Adjusted Reflectance (NBAR) at local solar noon, and white-sky albedo (WSA) for each of the bands and broadbands. Due to the large file size, each data layer is distributed as a separate HDF file. Users are encouraged to download the quality layers for each of the 10 bands to check quality assessment information before using the BRDF/Albedo data.

Version 6.1 reprocessing will begin in the fall of 2019. In the reprocessing, the QA values for the MCD43GF product will change to reflect the band 5 and 6 dead detector issues.

Known Issues:

This product has some known issues. The underlying V006 MCD43D results are based on a direct BRDF at this results (an improvement over the V005 where the MCD43 results were merely an average and reprojection of the MCD43A results). This has resulted an much higher quality MCD43D and thus MCD43GF results. In the high latitudes, however, there still may be primarily poor quality BRDF retrievals, usually dominated by very low illumination. And as mentioned, the BRDF parameters from areas experiencing local noon solar zenith angles of greater than 70 deg should not be utilized. Therefore in high latitudes the QA flags will indicate poorer quality GapFilled results. Therefore the linear fit approach described above has been implemented.

Description of Data Fields:

YDIM (Row) = 21600

XDIM (Col) = 43200

Data conversions:

file_data = BRDF_Parameter / scale_factor

scale_factor=1000

Projection: Lat/Long (Geographic Coordinate System)

Corner Values:

UpperLeftPointMtrs (-180.0, 90.0)

LowerRightPointMtrs (180.0, -90.0)

Dimensions:

21600*43200