|HOME SCHEDULE AUTHOR INDEX SUBJECT INDEX|
Understory vegetation mapping from remote sensing data and its implications for giant panda habitat conservation.
Linderman, Marc 1, Liu, Jianguo1, Qi, Jiaguo1, Li, An1, Ouyang, Zhiyun2, Yang, Jian3, 1 2 3
ABSTRACT- Remote sensing has been very successful at recognizing general distributions, disturbances and boundaries of landcover classes. Traditional landcover mapping, based on dominant overstory features, has been successfully applied to landscape, regional and global analyses. Many ecological processes and patterns, however, are the result of variations in the structure and floristic composition of understory vegetation features as well as those of the overstory features. Advances in active sensors such as lidar and radar offer potential sources of sub-canopy structure information. Large-extent studies, however, have either neglected or generalized the composition of understory vegetation, as mapping of understory features has remained unavailable from remote sensing data. An excellent example is the lack of information on the spatial distribution of understory bamboo for giant panda conservation. Wild pandas are dependent on bamboo for over 99% of their diet. In addition, the distribution of bamboo can vary significantly, spatially and temporally, independent of the dominant vegetation. Methods to map regional distributions of understory of bamboo are critical for the conservation and restoration of habitat. Here, we demonstrate a new approach to classifying bamboo directly from remote sensing data. Through the use of an artificial neural network, limited ground truth, and leaf-on Landsat TM data, a classification accuracy of approximately 80% was achieved in predicting the presence or absence of bamboo. Significant differences in habitat availability and fragmentation were found by including these data in landscape analyses.
KEY WORDS: remote sensing, understory vegetation, landscape pattern