Simulating Topographic Effects On Landsat Time-Series Change Detection

Landsat time series change detection and topographic correction are hotspots in the field of remote sensing.However,the existing research on change detection based on Landsat time series data rarely including topographic correction preprocessing.The topographic effects in Landsat time series change detection are still unknown and rare quantified.This study chose the China-Laos border as the site area,and used Land Trendr time series change detection algorithm based on the 1988-2016 annual Landsat time-series data to identify the threshold of forest disturbance magnitude along the China-Laos border.The SCS topographic correction model which performed well in the densely vegetated area was used to model variations of solar incident angle and forest disturbance index(Atmospherically Resistant Vegetation Index,ARVI)with slope,aspect and DOY at the study area center(22 ° N,101.5 ° E).The simulation diagrams of atmospherically resistant vegetation index in each month of one year in the study area were acquired and then detected ‘pseudo change’ based on difference change detection between each other.The ‘pseudo change’ in the time series data was determined according to the regional forest disturbance magnitude threshold.So as to determine the terrain effect in Landsat time series data and the time series change detection strategy across the China-Laos border.Finally,temporal-spatial distributions of the forest disturbance along the border was determined according to the strategy.The study shows that,first,the threshold of forest disturbance magnitude in this study area is ΔARVI=0.32.Second,when the slope is less than 10°,the terrain and time have little effect on the sun incidence and forest disturbance index.When the slope is less than 40 ° and the Landsat image acquired within 2 months,the terrain effect is little.The time series change detection should consider the images quality of the study area,and the acquisition time should be in the same season and the time span should be less than 2 months.The time series Landsat data were acquired from February to March,and the pre-processing of topographic correction was not necessary.Third,the overall accuracy of forest disturbance is 89.72% ± 0.67% and the estimated forest disturbance area between 1991 and 2016 reaches 4366.14 km2 ± 887.17 km2(at approximately the 95% confidence interval),the disturbance area accounts for 16.73% of the total area of forest land in 1991.The trend in forest disturbance area increased from1991 to 1995 and then continued downward.The forest disturbance area across the China-Laos border is closely related to global rubber prices as well as the polices and economies of the two countries.Compared with Laos,forest disturbance within China is stronger except for a few years.