Bidirectional reflectance factors of agricultural targets: A comparison of ground-, aircraft-, and satellite-based observations

Abstract

The bidirectional reflectance factor (BRF) of most natural surfaces exhibits substantial departure from lambertian behavior, yet the magnitude of its effect in off-nadir imagery from the SPOT satellite remains mostly undocumented. Field investigations were conducted at the Maricopa Agricultural Center in Arizona to explore this phenomenon using three wavelength intervals (green, 0.50-0.59 μm; red, 0.61-0.68 μm; and NIR, 0.79-0.89 μm). During 1988 and 1989 observations spanned two consecutive days during which the SPOT radiometer viewing angles differed by 34°. Ground-based BRFs were obtained for nadir and off-nadir viewing angles over replicated transects in cultivated soil, cotton, and wheat fields using radiometers mounted on a backpack. Aircraft data were measured from flight lines about 150 m above ground level with a nadir-oriented radiometer. SPOT digital imagery was converted to ground equivalent BRFs using specific calibration factors for the sensor, estimates of exoterrestrial radiation, solar zenith angles, and measurements of atmospheric optical depths. Reflectance factors varied with sensor, target, wavelength interval, and viewing and illumination geometry. Ground data were a strong function of solar zenith angle, revealing systematic differences in BRF behavior of each cover type. Good agreement was found between ground- and aircraft-based observations collected with nadir-pointed radiometers. For each wavelength interval, linear regression yielded R 2 > 0.99 and regression slopes averaging 0.96. Ground and aircraft data also revealed negligible changes in day-to-day reflectance from cultivated soil, cotton, and wheat targets. However, large differences were observed in SPOT imagery acquired on consecutive days. As expected, correlations between nadir aircraft observations and off-nadir satellite data were relatively poor ( 0.66 < R 2 < 0.80; slope ≈ 0.76 ) partially due to differences in sensor viewing angles. Off-nadir satellite and off-nadir ground data were better correlated ( 0.91 < R 2 < 0.96; slope ≈ 0.83 ). However, satellite-based BRFs were nearly double the ground-based values at low levels of reflectance (visible bands over dense vegetation). Possible reasons for discrepancies between satellite and ground off-nadir measurements include systematic bias in surface reflectance measurements, inadequate atmospheric correction of satellite data, adjacency effect from nearby surfaces, and inaccurate satellite sensor calibration.