An empirical topographic normalization method for forest TM data

A new empirical topographic correction method is proposed and applied to multi-temporal Landsat TM data (bands 1-5 and 7) collected in different periods of the growing season (April, June, July), for a mixed deciduous forest area (343 ha) in the northern Apennines. The method is based on the empirical observation that while for near-Lambertian behavior a linear relationship exists between radiance and the cosine of the incidence angle (i), for more anisotropic reflection a linear regression between the radiance logarithm and cos(i) is more effective at modeling topographic dependence.