WETLAND DISTURBANCE PATTERNS BASED ON THE MULTI-SCALE REMOTE SENSING IMAGE SEGMENTATION METHOD
KONG Bo 1,2, TAO He-ping 1,2, LIU Bin-tao 1,2, CHEN Wei-li 1, ZHANG Ji-fei 1
1. Institute of Mountain Hazards and Environment, CAS, Chengdu 610041, China;
2.National Remote Sensing Application Engineering Center, Southwest Subcenter, Chengdu 610041, China
This paper has used the differences of space-time Landsat TM images of NDVI standardized change intensity indexes to monitor land use changes. Eight clusters of disturbance patterns were identified in Jiansanjiang plain wetland after experimenting with a multi-scale remote sensing image segmentation method, which measured the proportion of disturbance (PD) and its adjacency (PDD) of the land cover and made a correlative analysis between them. The results show that the change of land use has exerted an obvious influence on the distribution of patterns of disturbance at the multi-scale, and the total disturbance rates of cropland, grassland and swamp are high and characterized by frequency variation. The proportion of the disturbance clusters C1 and C2 is distributed mainly in water, forest and marsh areas, and the value and the affected area are small. However, the rice land and crop land account respectively for about 74.38% and 61.76% of the proportion of the disturbance clusters C7, C8, and the value is high and the affected area is big, which indicates that cultivation is the main affecting factor that disturbs the wetland ecological system in this area. Disturbance adjacency at multi-scales has proved that human disturbance has appeared to a certain extent: when PD<0.4 or PD>0.7 and PDD>PD, human disturbance has an effect on land cover. This study provides ecological indexes for the assessment of frangibility and restoration capability of the wetland ecological system in Jiansanjiang area.