2:00 PM - 2:15 PM
[HTT17-02] State and fate of taiga-steppe ecotone in Mongolia
Keywords:Transition zone, Remote Sensing, Machine Learning, Taiga, Steppe, Ecotone
Within the ecological transition zone, a biome gradually shifts from one to the others, and ecological succession due to climate changes is more obviously seen than in the regions characterized by the typical biome. Mongolia involves the transition zone with the minimum anthropogenic land modification such as cultivation. Deciduous needle-leaf forests, corresponding to southern edges of Siberian taiga, extend over the northern territories, while grasslands (steppe) and bare area dominate mostly middle to southern territories. The boundaries separating taiga and steppe biomes, however, are not definitely obvious. This study aims to quantify present distribution of the taiga–steppe transition zone in Mongolia, providing some insights into prediction of shifting this zone possibly caused by ongoing climate changes.
To end this, we firstly referred to tree cover multi-year dataset (GFCC30TCv003/2015), which is derived from Landsat 5 Thematic Mapper (TM) and Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images. This global dataset shows areal ratio of tree cover with 30-meter resolution, although it remains some ambiguities especially for low tree coverage as for taiga-steppe transition zones. The pixels in the dataset were categorized into 4 classes in accordance with tree cover ratio. The pixels with >20 %, 5–20 %, 1–4 %, and 0 % of tree cover were designated as taiga (TA), transition (TR), steppe (ST), and desert class (DE), respectively. The ST classes were further classified into two (TR2 and ST1). TR2 classes are adjacent to the pixels with tree cover rates of >4%, and the latter as ST1. Then TR and TR2 classes correspond to the taiga–steppe transition zone.
Require for improving this dataset to more realistic representation of taiga-steppe transition zones in Mongolia, therefore further corrected by support vector regression (SVR) analysis using the quadrat-based visual observations on the Google Earth images. We found that the GFCC30TCv003/2015 tends to underestimate tree coverage for the areas of less than 40 % tree coverages, while overestimates above 40 %.
As for the next step, using machine learning algorithms we will correlate the improved tree coverage with conditioning factors as topographical and meteorological parameters to assess the sensitivity of taiga-steppe transition zones under ongoing climate changes.
To end this, we firstly referred to tree cover multi-year dataset (GFCC30TCv003/2015), which is derived from Landsat 5 Thematic Mapper (TM) and Landsat 7 Enhanced Thematic Mapper Plus (ETM+) images. This global dataset shows areal ratio of tree cover with 30-meter resolution, although it remains some ambiguities especially for low tree coverage as for taiga-steppe transition zones. The pixels in the dataset were categorized into 4 classes in accordance with tree cover ratio. The pixels with >20 %, 5–20 %, 1–4 %, and 0 % of tree cover were designated as taiga (TA), transition (TR), steppe (ST), and desert class (DE), respectively. The ST classes were further classified into two (TR2 and ST1). TR2 classes are adjacent to the pixels with tree cover rates of >4%, and the latter as ST1. Then TR and TR2 classes correspond to the taiga–steppe transition zone.
Require for improving this dataset to more realistic representation of taiga-steppe transition zones in Mongolia, therefore further corrected by support vector regression (SVR) analysis using the quadrat-based visual observations on the Google Earth images. We found that the GFCC30TCv003/2015 tends to underestimate tree coverage for the areas of less than 40 % tree coverages, while overestimates above 40 %.
As for the next step, using machine learning algorithms we will correlate the improved tree coverage with conditioning factors as topographical and meteorological parameters to assess the sensitivity of taiga-steppe transition zones under ongoing climate changes.