Increasingly, data from multiple sensors are used to gain a more complete understanding of land surface
processes at a variety of scales. Although higher-level products (e.g., vegetation cover, albedo, surface
temperature) derived from different sensors can be validated independently, the degree to which these
sensors and their products can be compared to one another is vastly improved if their relative spectroradiometric
responses are known. Most often, sensors are directly calibrated to diffuse solar irradiation or
vicariously to ground targets. However, space-based targets are not traceable to metrological standards, and
vicarious calibrations are expensive and provide a poor sampling of a sensor's full dynamic range. Cross-calibration
of two sensors can augment these methods if certain conditions can be met: (1) the spectral
responses are similar, (2) the observations are reasonably concurrent (similar atmospheric & solar
illumination conditions), (3) errors due to misregistrations of inhomogeneous surfaces can be minimized
(including scale differences), and (4) the viewing geometry is similar (or, some reasonable knowledge of
surface bi-directional reflectance distribution functions is available).
This study explores the impacts of cross-calibrating sensors when such conditions are met to some degree
but not perfectly. In order to constrain the range of conditions at some level, the analysis is limited to sensors
where cross-calibration studies have been conducted (Enhanced Thematic Mapper Plus (ETM+) on Landsat-
7 (L7), Advance Land Imager (ALI) and Hyperion on Earth Observer-1 (EO-1)) and including systems having
somewhat dissimilar geometry, spatial resolution & spectral response characteristics but are still part of the
so-called "A.M. constellation" (Moderate Resolution Imaging Spectrometer (MODIS) aboard the Terra
platform). Measures for spectral response differences and methods for cross calibrating such sensors are
provided in this study. These instruments are cross calibrated using the Railroad Valley playa in Nevada.
Best fit linear coefficients (slope and offset) are provided for ALI-to-MODIS and ETM+-to-MODIS cross
calibrations, and root-mean-squared errors (RMSEs) and correlation coefficients are provided to quantify the
uncertainty in these relationships. In theory, the linear fits and uncertainties can be used to compare radiance
and reflectance products derived from each instrument.
|