Multiple scattering in the remote sensing of natural surfaces
Radiosity models, working at scales much greater than the wavelength of light, predict the amount of light scattered many times (multiple scattering) among scene elements in addition to light interacting with a surface only once (single scattering). Such models are little used in remote-sensing studies because they require accurate digital terrain models and, typically, large amounts of computer time. We have developed a practical radiosity model that runs relatively quickly within suitable accuracy limits, and have used it to explore problems caused by multiple scattering in image calibration, terrain correction, and surface roughness estimation for optical images. We applied the model to real surfaces spatial scales of 30 m and 1 cm, separating multiple-scattering effects into those resolved by the Landsat TM and unresolved subpixel effects. Calculated radiosities were used to estimate quantitatively the magnitude of the multiple-scattering effects for different solar illumination geometries, surface reflectivities, sky illuminations and surface roughnesses. At the 30-m scale, multiple scattering can account for as much as 10 per cent of the radiance from sunlit slopes, and much more for shadowed slopes; at the 1-cm scale, the multiple scattering can locally account for as much as 70 per cent. Because the amount of multiple scattering increases with reflectivity as well as roughness, multiple scattering effects will distort the shape of reflectance spectra as well as changing their overall amplitude. Our results have significant implications for determining reflectivity and surface roughness in remote sensing and for energy-balance calculations.