Change Detection Of Semi-arid Forest Ecosystem Based On Remote Sensing Time Series Analysis

Due to the influence of natural factors and human activities,the Earth's surface is constantly undergoing complex changes at different time and space scales.Based on remote sensing time series data,effective remote sensing indicators are developed to detect,describe and interpret the changes of ecosystems and the environment.The interaction is important.This study analyzes the changes of forest ecosystems by using remote sensing change detection methods for constructing time-space information.The change of forest ecosystem of Saihanba Forest Farm in Hebei Province from 1989 to 2017 is the research object,which is constructed for different detection targets.The dynamic analysis method of vegetation change at different time scales was used to extract long-term trend changes and short-term disturbance events in the dynamic process of vegetation,and further evaluate the elasticity of forest ecosystems,and finally realize the forest ecosystem of Saihanba forest farm.Comprehensive evaluation.The main work and conclusions of the study are as follows:(1)Using the double logistic regression model to simulate the characteristics of seasonal phase fluctuations in NDVI time series signals,construct NDVI residual time series to eliminate the phenological differences of vegetation pixels in the regional scale,so as to enhance the short-term mutation and long-term gradual characteristics in time series signals..The results show that the residual time series method can better eliminate seasonal items,which provides a possibility for subsequent trend discrimination and mutation extraction.(2)Selecting the indicators that characterize the change of vegetation greenness and its local spatial correlation to describe the vegetation dynamics,and develop the spatiotemporal feature extraction method for the long-term changes of vegetation development based on non-parametric trend test method.The results show that the greenness of the area with low vegetation coverage at the initial moment shows an increase in volatility,while the spatial autocorrelation shows a steady increase.The greenness of the region with a certain vegetation coverage at the initial moment shows a steady increase,but spatial autocorrelation exhibits multiple forms of decline.(3)For the current fusion time-space information,the perturbation detection algorithm uses small scale and the threshold setting is affected by subjective factors.The spatiotemporal correlation continuous spurious feature is used to extract the disturbance events in the study area and analyze the time and space of the disturbance event in the study area.law.The results show that the disturbance frequency of the most concentrated and most active man-made forest farms is the most frequent of all the subfields.(4)Exploring the remote sensing analysis method of forest ecosystem based on time series data analysis.Mathematical simulation of the NDVI residual time series of the disturbance process is carried out to construct a remote sensing index that measures the elasticity of the forest ecosystem and apply it to the disturbance events in the study area.The results show that the duration of the disturbance and the duration of the disturbance phase are significantly reduced,reflecting the elastic changes of the forest ecosystem in Saihanba.