关键词: eCognition, 多尺度分割, 面向对象分类, 碳转换系数, 碳收支能力

Abstract:

Global warming is one of the most daunting challenges that our humanity is facing. It has urged the action to study carbon emissions and carbon sequestration ability to control the regional climate change. Some studies show that human activities, especially the combustion of fossil fuel, cause the carbon emissions and the global climate warming. Therefore, it is important to work on energy saving and emission reduction. At the same time, the carbon sink capacity of forest, grassland, crops and other vegetations become one of the most powerful approaches to ease the global climate change. Thus, we conduct a research in this area, in order to improve the carbon source utilization efficiency, reduce the carbon intensity, prefect the energy-saving and emission-reduction works, and well manage the carbon budget capacity and some other problems. Taking Youjiang district of Baise city in Guangxi Province as the experimental area, this article uses the object-oriented classification technology and extracts the area geographic information from the Landsat 8 OLI and Google Earth images. Due to the complex terrain of the study area, different parameter settings of the multi-scale segmentation are used and the optimal scaling for the image segmentation is selected. We also use the membership function method, the closest classification method and the CART decision tree classifier method to complete the object-oriented classification layer by layer and evaluate the accuracy of the classification results, based on the spectral difference, geometric shapes, objects, texture and other characteristics. Through summarizing the conversion relationship between the land and carbon coefficient, combining with the high-precision object-oriented classification results, the estimation model of carbon budget capacity is built based on land-cover types. Finally, according to the CASA model of carbon budget capacity, we check the accuracy of the estimation method. The carbon budget capacity of Youjiang district is estimated to be -3996.4 kt, according to the coefficient that corresponds to the feature carbon conversion relationship. Integrated with the administrative planning, population distribution, DEM, and other relevant data, the carbon budget capacity of Youjiang district is analyzed thematically. The results showed that: (1) the use of RS and GIS technologies in the studies of regional carbon budget capacity reveals a distinct advantage. The multiscale segmentation and object classification method can efficiently eliminate the extraction error caused by spectral confusion, and solve problems such as the large quantity of spatial data faced by the traditional classification, the classification of "salt and pepper", the exact utilization of different classification methods, and the improvement of the classification accuracy of carbon. (2) We summarized the findings of the coefficients for carbon balance capacity from domestic and international researches, and applied it in the construction of carbon balance estimation model. The object-oriented method and carbon balance estimation model were used to interpret the land cover data to estimate the Youjiang area carbon balance capability. Results show that the high degree of forest land, grassland, cultivated land and other types of vegetation areas is responsible for the performance of carbon sink in the study area. At the same time the construction land consumes a lot of fossil fuel to play as a large source of carbon. But the overall carbon sink ability of Youjiang area is stronger than its carbon emissions. It is conducive to the stable development of the regional ecological system, and may ease the regional threat of climate change. (3) Combined with the administrative divisions of Youjiang district and DEM data, the spatial analysis was carried out. We summed up the overall characteristics of the carbon budget capacity in the Youjiang district. The carbon budget is large in quantity in the central districts. Moreover, studies have shown that the elevation, slope aspect and human activities could affect the carbon budget capacity. It specifically demonstrated that the areas of high-altitude, steep slopes and less human activity reveal a large amount of carbon sink and less carbon footprint. On the contrary, the areas of low-altitude, lower slopes and more human activity reveal a large carbon footprint and lower amount of carbon.

Key words: eCognition, multi-scale segmentation, object-oriented classification, carbon conversion index, carbon budget