Automated Blending of Landsat and Modis Surface Reflectances at Global Scales

The 30 meter spatial resolution of Landsat sensors makes it one of the most suitable datasets for bridging the gap between local measurements and large-scale studies of biophysical processes, yet its usage is limited by the availability of cloud-free surface observations. Methods have been proposed to combine Landsat data with measurements from sensors of coarse spatial resolution but high temporal frequency to enhance their applications [1][2]. In particular, the spatial and temporal adaptive reflectance fusion model (STARFM) [1] utilizes associations between concurrent Landsat and MODIS observations, and temporal changes in MODIS observations to predict corresponding changes in Landsat data. However, because STARFM uses an unsupervised classification algorithm to cluster spectrally similar pixels in a Landsat scene, it is computationally timeconsuming and thus not suitable for applications at continental and global scales. As an on-going community project, we started to process a global Landsat dataset (the Global Land Survey 2005) to create a global LAI dataset that is compatible with MODIS. A key component of this project is to adjust the reflectance of GLS2005 data, which have different dates of acquisition (DOA), to approximately its peakgrowing-season (PGS) level when vegetation reaches its maximum greenness. To do this, therefore, we developed an alternative approach to expedite the classification algorithm in STARFM and apply it to process the global data. The method and the preliminary results are briefly described below.