Validation of the Extend Suite of MOD09 and SMAC Processed Reflectance Products for Australian Terrestrial Supersites: A Case Study

Abstract:

The Australian Terrestrial Ecosystem Research Network (TERN) brings together ecosystem scientists allowing them to collect, contribute, store, integrate data and collaborate across numerous disciplines. The TERN AusCover Facility comprises a national expert network that provides remote sensing data such as satellite-derive biophysical products, advanced remote sensing products and ground-validation information free and online to the research community.

TERN and AusCover have collected in situ data for a number of 5 km x 5 km supersites from nearly every state and territory in Australia. These data include spectrophotometer data, sun photometer and ozonometer data, airborne and terrestrial LIDAR data and airborne hyperspectral data.

As part of the AusCover facility and in conjunction with Landgate, Western Australia, Curtin University has modified the atmospheric correction and surface reflectance processing scheme from Landgate to process 12 extra MODIS bands to surface reflectance, thus providing 19 bands. This processing scheme uses the Simple Method for Atmospheric Correction (SMAC) to produce reflectance data.

Until recently, only the first 7 MODIS bands were available with the MODIS institutional algorithm for surface reflectance, MOD09, but this has altered to now also provide 9 extra bands. MOD09 is based around 6S to produce reflectance data.

This case study makes use of hyperspectral airborne data captured over the Credo TERN supersite to compare the surface reflectance products from MOD09 and the SMAC-based 19-band reflectance process.

This work required validating the airborne data against a set in situ reflectance measurements of large calibration targets. The validated airborne data were resampled spatially and spectrally to MODIS bands and both the airborne and MODIS data were mapped to the same spatial grid.

Direct pixel comparisons have been made between the airborne data and the two algorithms, and between the algorithms themselves.

The algorithms show good agreement to the airborne data and to each other with RMS errors less than 0.025 and r² values greater than 0.95. This would indicate both algorithms are viable methods to produce surface reflectance data over the Credo supersite and provides a roadmap to extend the comparison to the other AusCover sites.