Remotely Sensed Method for Detection of Spatial Distribution Pattern of Dryland Plants in Water Limited Ecosystem
Autor: | Su Riga, Buho Hoshino, Myagmartseren Purevtseren, Christopher McCarthy, Keita Shima, Zoljarga Enkhtuvshin, Ying Tian |
---|---|
Rok vydání: | 2020 |
Předmět: |
Endmember
010504 meteorology & atmospheric sciences Pixel 0211 other engineering and technologies 021107 urban & regional planning 02 engineering and technology Vegetation 01 natural sciences Normalized Difference Vegetation Index Field (geography) Environmental science Ecosystem Satellite Satellite imagery 0105 earth and related environmental sciences Remote sensing |
Zdroj: | IGARSS |
DOI: | 10.1109/igarss39084.2020.9324058 |
Popis: | The Gobi Desert in Mongolia is characterized by sparse and patchy vegetation, interspersed with essentially bare areas. The vegetation pattern is typically formed by perennial shrubs, grasses or annually-herbaceous plant overlying a matrix composed of bare soil. Vegetation patterns, most broadly, refer to the spatial organization of vegetation in a landscape. However, since the plants in the Gobi Desert are sparsely distributed over a vast bare field, it is extremely difficult to accurately observe from satellite imagery. This is because reflectance of dry soil is very high and the reflectance of slightly distributed plants is eliminated by soil reflection. This study solves this problem by using field surveys and methods for combining different satellite sensor data and spectral un-mixing analysis. As a result, the pixel NDVI value of desert plants shows a smaller value than the ground measurement. It is shown that the fraction of the vegetation endmember after pixel un-mixing has a remarkably high correlation with the field measured values (where, R2=0.51 between NDVI of Landsat 8 imagery original pixels and un-mixed pixels and R2=0.79 between plants coverage of field measurement and un-mixed pixels percentage of vegetation endmembers). |
Databáze: | OpenAIRE |
Externí odkaz: |