Lehnert, L.; Meyer, H.; Wang, Y.; Miehe, G.; Thies, B.; Reudenbach, C. & Bendix, J. (2015): <b>Retrieval of grassland plant coverage on the Tibetan Plateau based on a multi-scale, multi-sensor and multi-method approach</b>. <i>Remote Sensing of Environment</i> <b>164</b>, 197-207.
Resource Description
Title:
Retrieval of grassland plant coverage on the Tibetan Plateau based on a multi-scale, multi-sensor and multi-method approach
FOR816dw ID:
3
Publication Date:
2015-05-16
License and Usage Rights:
PAK 823-825 data user agreement. (www.tropicalmountainforest.org/dataagreementp3.do)
email:
bendix <at> staff.uni-marburg.de
Deutschhausstraße 12
Room No. 02 A 48
35032 Marburg
Faculty of Geography
Germany
Abstract:
Plant coverage is a basic indicator of the biomass production in ecosystems. On the Tibetan Plateau, the biomass<br/>
of grasslands provides major ecosystem services with regard to the predominant transhumance economy. The<br/>
pastures, however, are threatened by progressive degradation, resulting in a substantial reduction in plant<br/>
coverage with currently unknown consequences for the hydrological/climate regulation function of the plateau<br/>
and the major river systems of SE Asia that depend on it and provide water for the adjacent lowlands. Thus,<br/>
monitoring of changes in plant coverage is of utmost importance, but no reliable tools have been available to<br/>
date to monitor the changes on the entire plateau. Due to the wide extent and remoteness of the Tibetan Plateau,<br/>
remote sensing is the only tool that can recurrently provide area-wide data for monitoring purposes. In this<br/>
study, we develop and present a grassland-cover product based on multi-sensor satellite data that is applicable<br/>
for monitoring at three spatial resolutions (WorldView type at 2–5 m, Landsat type at 30 m, MODIS at 500 m),<br/>
where the data of the latter resolution cover the entire plateau. Four different retrieval techniques to derive<br/>
plant coverage from satellite data in boreal summer (JJA) were tested. The underlying statistical models are<br/>
derived with the help of field observations of the cover at 640 plots and 14 locations, considering the main<br/>
grassland vegetation types of the Tibetan Plateau. To provide a product for the entire Tibetan Plateau, plant<br/>
coverage estimates derived by means of the higher-resolution data were upscaled to MODIS composites acquired<br/>
between 2011 and 2013. An accuracy assessment of the retrieval methods revealed best results for the retrieval<br/>
using support vector machine regressions (RMSE: 9.97%, 7.13% and 5.51% from the WorldView to the MODIS<br/>
scale). The retrieved values coincide well with published coverage data on the different grassland vegetation<br/>
types.<br/>
Keywords:
| MODIS | Tibetan Plateau | Partial least square regression | Plant coverage | Degradation monitoring | SVM regression | Linear spectral unmixing | Spectral angle mapper | Vegetation indices |