Project Funding: 2016 - 2018

NRA: 2014 NASA: Land Cover / Land Use Change   

Funded by NASA

Abstract:
This modestly budgeted proposal, with detailed proof of concept, provides an unprecedented opportunity to support the needs of several NASA programs as well as the needs of the fire science and applications communities. It is directly responsive to the primary focus of the call - developing algorithms and prototyping products for combined use of data from Landsat-8 and Sentinel-2 toward global land monitoring, and advances the virtual constellation paradigm for mid-resolution land imaging. This proposal will prototype a global burned area product by combination of Landsat-8, Sentinel-2A, and - 2B data. The algorithm will be applied on a multi-temporal per-pixel basis with temporal consistency checks to reduce commission errors. Seamless integration of the different sensor data will be enabled by a random forest change regression that is parameterized with synthetic training data and that models reflective wavelength change considering the different sensor spectral response functions. In this way the algorithm can be implemented considering each sensor alone and in conjunction. Temporal consistency checks and active fire detections derived from Landsat-8 and Sentinel-2 and contemporaneous MODIS and VIIRS active fire detections will be used to differentiate between burned areas and spectrally similar non-fire surface changes. The product will be prototyped for one fire season for all of South Africa (Year 1) and Southern Africa (Year 2) to reflect the early continental availability of Sentinel-2 data and our long standing regional experience and collaborations. Global product implementation will be demonstrated by generation of one year of global 30m Landsat-8 WELD weekly products on the NASA Earth Exchange (NEX) and using these and contemporaneous Sentinel-2 time series to produce the product at 28 globally distributed WELD tiles (Year 3). Product validation will be conducted by comparison with visually interpreted commercial high resolution data and will be endorsed by a regional network of African fire scientists and practitioners. The product will be compared with coarser resolution MODIS and VIIRS burned area products, and ESA funded, burned area products, to assess product differences and at the validation sites to quantify their relative accuracies.

Publications:

Roy, D. P., Huang, H., Boschetti, L., Giglio, L., Yan, L., Zhang, H. H., Li, Z. 2019. Landsat-8 and Sentinel-2 burned area mapping - A combined sensor multi-temporal change detection approach. Remote Sensing of Environment. 231, 111254. DOI: 10.1016/j.rse.2019.111254