Methodology Of Landsat Time Series Modeling For Analyzing Complex Spatio-temporal Evolution Characteristics Of Ecosystem

Terrestrial ecosystems changes are continuous and complex at various spatial and temporal scales.The remote sensing time series analysis has a unique advantage in the research of change detection.Based on the LandTrendr algorithm,this study constructed a detection model for monitoring forest ecosystem disturbance-recovery dynamic process incorporating the spatial structure characteristics using a dense Landsat time series.And we obtained the continuous and complex change characteristics of the forest ecosystem and revealed the corresponding essential features of the environment.The main contributions and conclusions are as follows:(1)Aiming at the problems of limited amount of Landsat remote sensing images in the subtropical red soil erosion area and the complicated evolution process of forest ecosystem,this study constructed a detection model for monitoring forest ecosystem disturbance-recovery dynamic process which incorporated spatial structural features based on dynamic time warping and Dense-LandTrendr algorithm using dense Landsat time series.The disturbance-recovery dynamic process including suddenly change(abrupt changes)and continuous and slowly changing(gradual changes)in forest ecosystems was effectively described,and the accuracy of overall time of identifying the recovery process was improved over three times than the previous algorithm.(2)Using dense Landsat images as model input,calculating the contrast of the gray level co-occurrence matrix and the landscape fractal dimension measure in the spatial sliding window,and constructing the dense spatial structure features time series of forest ecosystem can accurately identify the ecology change processes and characteristics and maximum eliminate the influence of vegetation phenology and spectral noise.It is more accurate to identify the changes of the two patterns of forest composition and structure(forest-forest),forest-farmland conversion and the changes of the pixels in the ecozone of the landscape patch edge.(3)Based on the detection maps of forest ecosystem disturbance-recovery processes using dense Landsat time series,we developed a linkage analysis of the spatio-temporal changes of forest ecosystems,exploring the spatio-temporal correlation of forest ecosystem dynamics,and combining ecological change processes with the characteristics of landscape spatial patterns,we analyzed four changes: the spatiotemporal trend change pattern of forest ecosystems in the study area,the change pattern of landscape spatial pattern and change process,and the change cahracteristics of the spatio-temporal index and the change process of different landscape types.The landscape heterogeneity in the study area shows a decreasing fragmentation pattern as a whole;the structural heterogeneity of the homogeneous landscape is generally increasing,while the decrease may be caused by changes in the internal structure of the landscape.(4)Based on the quantitative parameters obtained from the forest ecosystem disturbance-recovery process detection,we developed the forest ecosystem stability indexes,namely resistance,resilience,and variability.This study analyzed the spatiotemporal change characteristics of the resistance,resilience,and variability of forest ecosystems in red soil regions.The overall accuracy of identifying different classes of ecosystem variability is 87.14%,which indicates that the method of evaluating the stability of forest ecosystems based on the dense Landsat time series analysis is effective.Experimental results show that,the forest ecosystems have strong selfsustainability,disturbances from climate and human activities to soil and vegetation reduced gradually,and ecosystem stability is generally on the rise.