Large-Scale BRDF Retrieval over New Mexico with a Multiangular NOAA AVHRR Dataset

In this study a number of linear semiempirical kernelrado. Subsets of these data have been used in other studies as a means of investigating variations in the bidirectional driven (LiSK) bidirectional reflectance distribution function (BRDF) models are adjusted against an extensive Adreflectance distribution function (BRDF) and albedo at the grassland and grassland-mesquite transition sites within vanced Very High Resolution Radiometer (AVHRR) dataset collected over a variety of semiarid cover types in the souththe United States Department of Agriculture Agricultural Research Service (USDA-ARS) Jornada Experimental ern part of New Mexico and parts of Chihuahua, Mexico as part of the May 1997 Prototype Validation Exercise Range (Hyman et al., 1998; Barnsley et al., 2000) and of validating geometric-optical and radiative transfer (GORT) (PROVE) campaign, an activity of the NASA Earth Observing System Terra validation program. The aim is to investiBRDF models (Ni and Li, 2000). In this study a number of linear semiempirical kernelgate model behavior under conditions of sparse angular sampling such as that provided by the AVHRRs and MODIS over driven (LiSK) BRDF models are adjusted against these AVHRR data, with the focus on kernel combinations using a wide variety of southwestern desert surface types. Linear semiempirical models of the type to be used in the MODIS/ the LiSparse surface geometric-optical kernels with either the RossThin or the RossThick volume scattering kernel. MISR BRDF/albedo product (MOD43) are inverted, since these are appropriate for use over large areas. The results The models are derived in Wanner et al. (1995) and are based on the formulation of Roujean et al. (1992b); detailed of the inversions show that these models are able to describe BRDF for a wide variety of surfaces and provide both a descriptions will not be given here. These models have been suggested as part of a multimodel algorithm for MOmeans for correcting for directional phenomena in satellite data and for extracting structural information from multiDIS BRDF processing (Strahler et al., 1996). The aim is to investigate model behavior under conditions of sparse angular reflectance datasets. Published by Elsevier Science Inc. angular sampling (as provided by the AVHRRs) over the southern part of New Mexico and parts of Chihuahua, Mexico. This area encompasses both the USDA-ARS Jornada Experimental Range (JER) and the Sevilleta National INTRODUCTION Wildlife Refuge (NWR) and a wide range of cover types, As part of the Prototype Validation Exercise (PROVE) including desert grassland, semiarid grassland, grasslandcampaign, an activity of the NASA Earth Observing System shrub transition, desert shrubland, forest, dry lake bed, Terra-1 validation program, 22 High Resolution Picture alkali flats, and the gypsum dunes of White Sands National Transmission (HRPT) datasets from the NOAA Advanced Monument, among others. LiSK models are appropriate Very High Resolution Radiometers (AVHRR) on the NOAAfor use over large areas (since they are rapidly inverted) 12 and NOAA-14 satellites corresponding to the campaign and can be used with sparse angular datasets (since the period were acquired, georeferenced, and calibrated to small number of parameters tends to lead to robust retrievpercent albedo by researchers at the University of Coloals). They are particularly appropriate for use with AVHRR data over arid and semiarid regions where cloud cover is not prohibitive. The ability to account for BRDF effects * USDA-ARS Hydrology Laboratory, Beltsville Agricultural Research Center-West, Beltsville, Maryland in AVHRR Channel 1 (visible) and 2 (near-infrared) data Address correspondence to Mark J. Chopping, USDA-ARS Hydrolis important since many researchers have recognized limiogy Lab., Beltsville Agricultural Research Ctr.-West, Beltsville, MD tations with these multiangular data when uncorrected 20705-2350. E-mail: [email protected] Received 15 July 1999; revised 20 August 1999. (Roujean et al., 1992a; Goward et al., 1993; Kremer and