Retrieving albedo in small sample size

摘要

A new criteria for kernels' selection in Ambrals model is provided. Instead of using "least square" methods, "least variance" of white albedo which derived from Tarantola's information theory is used. Several tests showed "least variance" had many advantages. First, it is less sensitive to noise. Second, it operated well with a small sample size. Third, it is less sensitive to the sampling position.