1a.Objectives (from AD-416):
The overall objective for this project is to better quantify albedo changes associated with land cover change, vegetation disturbance, and recovery from disturbance by fusing MODIS BRDF/Albedo and Landsat directional reflectance observations. There are four separate objectives for this project. We will work on objective 1 and create a global lookup table of albedo values as well as BRDF parameters for typical land cover types, as a function of global ecoregion. This objective supports global, historical analysis of the consequences of land cover conversion, and builds on earlier studies published by co-investigators using MODIS data. The high quality MODIS BRDF parameters LUT will be applied for different crop types using USDA NASS Crop Data Layer to estimate albedo values for different crops. The accumulated BRDF parameters for different crop types will be used to correct directional effects from wide-swath satellite data such as MODIS, AVHRR and AWiFS. The accumulated high quality BRDF look-up-table will be used to compute albedo for typical crops and correct angular effects from wide-swath polar-orbiting satellite such as MODIS, AVHRR and AWiFS. Consistent high quality remote sensing data are required for data fusion application in crop condition monitoring which will improve evapotranspiration estimation, crop yield forecasting and drought detection.
1b.Approach (from AD-416):
In previous studies, we have constructed an BRDF/albedo LUT based on global MODIS BRDF/albedos and MODIS IGBP land cover classes. The inter- and intra-annual variability of albedo for different IGBP classes have been examined under snow free and snow covered conditions. This work was based on the MODIS Climate Modeling Grid (CMG) albedo product (0.05 degree) and MODIS IGBP land cover map. The IGBP class at MODIS scale represents the majority class, and statistics derived from this product may be affected by mixed pixels. Here we propose to extend our previous work by using Landsat to quantify class homogeneity at the MODIS scale, and to retrieve albedo from “pure” examples of IGBP classes. We will use the 2000 Global Land Survey (GLS) Landsat dataset to select “pure”, homogeneous MODIS pixels globally. Each Landsat scene will be reprojected and aggregated from 30m Landsat resolution to 500m MODIS resolution. There are about 240 Landsat pixels included in each MODIS cell. We will check the homogeneity of a MODIS pixel based on these ~240 Landsat pixels using either unsupervised classification or statistics such as the mean and standard deviation from all bands. Only “pure” MODIS samples that fall in a homogeneous area (ie. small spectral standard deviation) will be used to compute mean per-class albedos and their variance. The Albedo results from these “pure” MODIS pixels will be assembled into a new per-class LUT, and compared and analyzed to one derived from all MODIS pixels.
In Year 1 of this project, a beta version of global snow-free albedo look-up table (LUT) has been built. The multi-dimensional albedo LUT includes statistics (mean and standard deviation) of albedo for different land cover types at various spatial and temporal resolutions. Ten years MODIS global data products were used to build a global climatology albedo data set. Landsat global survey (GLS) 2000 and 2005 data were used to extract the homogeneous pixels. The albedo results from these “pure” MODIS pixels were assembled into a new per-class LUT in a hierarchical data structure. Modules to use the hierarchical albedo LUT were built and tested. The beta version of global albedo LUT has been delivered to NASA research group for studying the global climate change due to the changes of land covers. A separate albedo LUT under snow-covered condition is under construction by scientists from University of Maryland.