Time Series Analysis Of Forest Ecosystem Changes In Typical Red Soil Region

Due to soil erosion,forest vegetation in red soil area is constantly undergoing complex changes.Remote sensing technology has unique advantages in forest ecosystem change detection because of its macroscopic,dynamic and rapid monitoring characteristics.Based on remote sensing timing analysis methods,tracking forest change and describing its change process,which is of great significance for characterizing long-time-scale environmental changes.The study area of this paper was selected in Hengyang Basin,Hunan Province,China.Landsat images from 1985 to 2019,field surveys and auxiliary data were collected,and the disturbance sensitive vegetation index(DSVI)was established as an indicator of forest change.Long-term(1985–2019)forest disturbance and recovery changes were detected using trajectory spectral-time segmentation algorithm(Land Trendr)on the Google Earth Engine(GEE)platform,while three evaluation aspects of velocity,frequency,and variance were used to describe the processes of forest gowth process characteristics in red soil regions.Then,the stability of forest ecosystem was evaluated.The main work and conclusions are as follows:(1)The forest change response sensitive vegetation index(DSVI)was constructed,and DSVI had strong sensitivity compared to other indicators(NDVI and SWIR / NIR).In addition,DSVI showed a good ability to detect different types of forest changes,such as abrupt changes caused by deforestation and urban expansion,and gradual changes caused by drought and severe soil erosion.(2)Based on the Land Trendr algorithm,the disturbance and recovery changes of forest ecosystem were detected.By extracting the year,duration and magnitude of disturbance and recovery,it was found that about 45% of the forests were disturbed,and the disturbance occurred most frequently during 1985-1990,and the duration of most forest disturbances was short.The forest change caused by moderate disturbance(95.34%)was much more than that caused by intensity disturbance(4.66%),and the intense disturbance mainly distributed in the areas affected by human activities,such as the areas around cities and forest margins.26.11% of the forests were in recovery state,and the recovery lasted for a long time after disturbance.The moderate recovery(90.95%)was about 10 times of the intensity recovery(9.05%).Finally,the accuracy of the disturbance process was evaluated,and the overall accuracy was 82%,indicating that the Landtrendr algorithm could effectively describe the dynamic process of forest disturbance and recovery.(3)Three characteristic indexes of velocity,frequency and variance were constructed to describe the characteristics of long-term forest growth in typical red soil region.The results showed that approximately 2/3 forests showed an increasing trend,1/3 showed a decreasing trend,the forest interference frequency was mainly four disturbances,accounting for 28.33%,and the variance of the forest growth curve was mainly low(48.46%)or medium(28.84%).It can be seen that the forest ecology in the red soil region is fragile and relatively unstable in the long-term growth process.The three evaluation indexes can effectively analyze the forest long-term change process.DSVI is a sensitive spectral index for forest change monitoring,which can be used to monitor forest landscape changes under land degradation.Combined with DSVI index and Landtrendr algorithm,the disturbance and recovery changes of forest were effectively detected.In the analysis of forest characteristics,three evaluation indexes(velocity,frequency,and variance)were used to describe the long-term forest growth process quantitatively and accurately.