MISR abstract
K.Khlopenkov and A.P. Trishchenko (2004). Analysis of BRDF and Albedo Properties of Pure and Mixed Surface Types From Terra MISR Using Landsat High-Resolution Land Cover and Angular Unmixing Technique. Fourteenth ARM Science Team Meeting Proceedings, Albuquerque, New Mexico, March 22-26, 2004.
Introduction: The bi-directional reflectance distribution function (BRDF) determines the degree of anisotropy of the
surface’s reflective properties (Nicodemus et al. 1977). This is a basic parameter employed in the
vegetation structure retrievals and characterization of surface albedo. An important feature of the BRDF
shape is the amplitude and width of the hot spot and dark spot areas in the principal plane direction. The
general BRDF shape in the direction of the perpendicular plane is also important. Retrieving the BRDF
shape from a satellite observation is a difficult task (Luo et al. 2004). In general, satellite sensors do not
provide simultaneous measurements of reflectance in all viewing directions. For example, in BRDF
retrievals from moderate-resolution imaging spectroradiometer (MODIS) observations (Schaaf et al.
2002), angular coverage is achieved by assembling clear-sky pixels from different orbits within a certain
interval of time: 8, 10, or 16 days. Multiangular observations from multi-angle imaging spectrometer
(MISR) contain simultaneous measurements at nine viewing directions, thus providing much better
capabilities for surface BRDF retrievals and an atmospheric correction for the presence of the
atmospheric aerosols. However, since MISR has a narrow swath and operates on sun-synchronous orbit,
it cannot provide complete coverage of the entire angular domain (Luo et al. 2004). Therefore, it is
important to have the BRDF models for generic landcover types, which will be adjusted for each
particular case, based on a limited sampling of the angular measurements.
To construct the generic BRDF for a specific landcover type, one needs to ensure an adequate sampling
of observations in the viewing zenith angle vs. relative azimuth angle (VZA-RAA) plane for various
ranges of the solar zenith angle (SZA). The best strategy to achieve this task is to collect all of the
reflectance data for all of the pixels with the same landcover type, and then use these data to derive the
generic BRDF shape (Luo et al. 2004). In reality, however, the definition of landcover type may not be
a perfect identification of the type of surface element, from the BRDF point of view. For example, “cropland” is a frequently used landcover type to denote a large variety of vegetation types, such as
wheat, corn, soybeans, and many others. Depending on the season, the area identified as cropland may
correspond also to baresoil or the mixed conditions – baresoil and vegetation. Another problem is
related to the mixing of various landcover types within one pixel. If the pixel size is larger than the
typical size of agricultural fields, then the pixel contains a mixture of various fields. Also, there is a
problem associated with similar landcover types in different conditions and/or stages of growth.
Although identical in the landcover type, their BRDF properties may look very different.
Below, we present some of the results of our analysis aimed at studying the problem of the mixing of
various landcover types and its effect on the BRDF at a coarse spatial resolution. In particular, we
analyzed MISR observations and deduced the BRDFs for pure landcover types using the angular
unmixing method. We also studied the relationship between the BDRF shape of the mixed landcover
classes and the level of the Normalized Difference Vegetation Index (NDVI).