Obtaining surface reflectance factors from atmospheric and view angle corrected SPOT-1 HRV data

SPOT-1 High-Resolution Visible (HRV) multispectral (XS) and panchromatic data were acquired over an agricultural area on two consecutive days in June 1987, June 1988, and April 1989, at view zenith angles of approximately 23° and 10°. Digital data were converted to surface reflectance factors (ρ_s) by use of the sensor calibration coefficients, measurements of atmospheric optical depth, and a radiative transfer model. View-angle corrections (C_v) were derived from ground-based measurements of bidirectional radiance of bare soil, and used to convert nadir ground- and aircraft-based measurements to off-nadir values (ρ_g and ρ_a, respectively) for comparison with SPOT HRV data. The absolute error of ρ_s values, relative to ρ_g and ρ_a, was less than 10% for most XS bands on all six days over the reflectance range 0.1-0.4. However, there was a systematic trend for ρ_s estimates to be slightly higher than ρ_g and ρ_a measurements, particularly at low surface reflectances. The C_v coefficients were then applied to SPOT HRV data for a variety of cover types to assess the effectiveness of a simple, view-angle correction over a complex landscape. For rough, unvegetated surfaces, ρ_s values that had originally differed by more than 0.09 in reflectance on the two days were brought to within 0.01 in all three XS bands. For vegetated surfaces, C_v appeared to be wavelength dependent; the soil-based C_v worked well for data in the red and green wavebands but overcorrected the near-IR data. The C_v correction overcompensated for view angle effects over planar surfaces (i.e., water and roads) in all wavebands.