Regional vegetation mapping and direct land surface parameterization from remotely sensed and site data

Abstract

Given the ability to define vegetation and land cover at the site level based on attributes such as physiognomy, horizontal and vertical structure, vegetation phenology and leaf morphology, direct parameterization and mapping using remotely sensed data can enhance the ability to characterize and monitor those important biogeophysical parameters. This research reports on the integrated analysis of site and remote sensing data to directly predict and map land surface biophysical parameters relating to structure, morphology, phenology and physiognomy, as an alternative to using intermediate classification categorizations to relate these site variables/parameters to remote sensing features. Correlation analyses were performed with the objective of finding appropriate remote sensing features to discriminate site variables/parameters, given site-level biogeophysical and ecological information. The direct parameterization technique that best predicted regional vegetation parameters was the Gaussian ARTMAP neural network supervised classification algorithm.