Spatiotemporally monitoring forest landscape for giant panda habitat through a high learning-sensitive neural network in Guanyinshan Nature Reserve in the Qinling Mountains, China
During the 1970s and 1990s, Guanyinshan Nature Reserve (GNR), a giant panda (Ailuropoda melanoleuca) distribution area historically, had experienced periodic commercial logging. After officially logging stopping in 1998 and converting to a giant panda nature reserve in 2002, GNR got the chance on forest restoration. It is very necessary to monitor the spatiotemporal change of its forest habitat. It is also widely known that it is difficult to make accurate mapping the mountainous area based on images through traditional classification algorithm. So, this study aims to monitor the spatiotemporal change of mountainous habitat in GNR in order to provide proper suggestions for giant panda conservation. The research applied a multilayer perceptron model, a high learning-sensitive algorithm, to classify the land cover types and monitor habitat change in GNR by using Landsat images acquired in 1978, 1988, 1997 and 2007, respectively. Our results showed that: (1) three types of forests composed the main landscape of the GNR, and an increase of 7.7% forest coverage occurred within 30 years. (2) Due to logging, there were many forest clearing-cutting areas in 1997 and swaths of shrub–grass in 1978 and 1988. However, these two types of landscape were strongly reduced by 2007 due to more attention and protection. (3) A decrease in the number of patches, an increase in the mean patch size, and an over-time decreasing in the mean nearest neighbor distance all revealed a decreasing on habitat fragmentation. Therefore, reduction in detrimental human activities has helped enhance and expand giant panda habitat toward a healthier and more stable ecosystem.
