Bamboo has an extensive habitat range and is found on every continent except Europe. Giant pandas choose their habitat based on the availability of suitable bamboo foraging. Understorey bamboo levels can be important in assessing the quality of giant panda habitat, and remote sensing techniques have been under development to facilitate this mapping process. GIS techniques offer potential for an efficient and cost-effective way of conducting the necessary assessment for large tracts of land. Specific GIS expert systems have been utilized to detect the presence or absence of specific species in the understorey with a combination of remotely sensed data and auxiliary GIS data layers.
This study covered an area in the Qinling Mountains above an elevation of 2000m known to be giant panda habitats. Pandas usually spend 3-4 months of the year in this location. The area known to provide their winter habitat was not included in this study. The two bamboo species of interest are Fargesia qinlingensis and Fargesia nitida. The study treats them as the same species since they are so morphologically similar and inhabit much of the same areas. The researchers collected field data using a Garmin 12XL GPS and collected biotic factors in field plots that corresponded to points of interest. The bamboo coverage was categorized based on percentage. A cloud-free and snow-free ASTER image of the area was obtained from NASA’s EOS Data Gateway. The researcher used the field data along with the ASTER bands to train a neural network to recognize vegetation type along with many other factors. The researchers employed thousands of iterations in order to maximize the accuracy of their hybrid neural network/GIS expert system. After assessment, the hybrid system produced more accurate maps than the other two methods (pure neural network and pure expert system).
(Area of study)
(Area of study)
(Field campaign data acquisition)
(Shows process of making a hybrid system like the one used in this study)
(Portrayal of vegetation type in the study area from a neural network and knowledge-based classifier)
(Side-by-side comparison of the various models with the last being from the most accurate hybrid network)
Wang, T. G. (2009). Improved understorey bamboo cover mapping using a novel hybrid neural network and expert system. International Journal Of Remote Sensing, 30(4), 965-981.