The frequency dependence of statistics of radar signals scattered from forest are investigated through a simulation study based on a radiative-transfer theory. A forest canopy is modelled as a volume of needle-shaped or disc-shaped leaves and finite-length, cylindrical branches. It is expected that at radar frequencies above the X-band, scattering away from normal incidence is largely dominated by leaves, while a mixture of leaves and branches is important to model a forest canopy at the C-band. Thus, simulations of both a volume of leaves and a mixture of leaves and branches are of interest. It is assumed that all leaves and branches are randomly oriented. It is found that for leaves that are small compared with the incident wavelength, signal amplitude is Rayleigh distributed. Generally, this means the length of a needle-shaped leaf can be as long as kh (wavenumber x length) equal to 2, because its volume is small. However, for a disc-shaped leaf ka (wavenumber x radius) should be no more than 0.5 because its volume is much larger. When the leaves are comparable in size to the incident wavelength, signal amplitude assumes a Gamma distribution. The signal amplitude distribution for a mixture of branches and leaves is also well represented by a Gamma distribution because, here, the scatterer volume is also large in dimension relative to the incident wavelength.
