Impact of Satellite Geometric Distortions on Landscape Analysis: Effects on Albedo

Abstract

Data from wide field-of-view sensors have been providing information about the Earth's surface since the early 1980's. This manuscript is the result of investigations designed to determine the effective resolution and geometric variability of the NASA Earth Observing System MODerate Resolution Imaging Spectroradiometer (MODIS) and Visible Infrared Imager Radiometer Suite (VIIRS) gridded data. Although the wide field-of-view and high temporal frequency of MODIS provide near-daily global coverage, inconsistent observation assignment in geolocated MODIS pixels measurably demonstrates how spatial accuracy is affected by pixel-size growth (up to 4.8x) along-scan. For studying the effective resolution, the point spread function of nominal 250m MODIS gridded surface reflectance products (L2G) was estimated from [man-made] large size targets. The findings indicate that in near-optimal locations the resolution of (sinusoidal grid) gridded products varies between 344m-835m along-scan for a range of viewing angles, but also indicate location-dependent variability with along-scan and along-track ranges of 314m-1363m and 284m-501m respectively. Albedo was identified as a well-known physical metric to study the effects of geometric variability, thus a broadband albedo using MODIS-like geometry was simulated for five EOS validation sites. Results of each site simulation exhibit compounded uncertainty attributable to the geometric distortion in ranges sufficient to influence climate models (i.e. ranges from 0.01-0.045 albedo). A second series of broadband albedo simulations was developed for the same five EOS validation sites using VIIRS-like geometries and aggregation zones. Spatially heterogeneous land cover demonstrated a marginally significant difference in the mean albedo between aggregation zones (< 0.015). Results from data simulating temporal compositing, demonstrate the influence of geometric artifacts through differing levels of uncertainty between periods (i.e. ranges from 0.01-0.05 albedo). The variability in both MODIS and VIIRS L2G questions the standard application of a global fixed grid, and indicates that regional projections combined with a representative grid cell 4x the nominal detector size (i.e. 1000m and 1500m for MODIS and VIIRS, respectively) are potentially useful for products using off-nadir views. This work ultimately resolves the surface-feature representation of temporo-spatial wide field-of-view instrument observations and quantifies the results of associating inherently-variable observations into an artificially-fixed and geometrically-regular space.