Extraction Of Rocky Desertification Information In Karst Area Of Southwest China Based On Time-series Remote Sensing Data

Rocky desertification has become a major long-term problem for the ecological environment in karst areas of southwest China.Remote sensing data plays an irreplaceable role in regional ecological environment monitoring because it can provide spatiotemporal continuous changes of earth surface elements.To establish an efficient and accurate information extraction model of rocky desertification based on remote sensing time series data and quantitatively evaluate its evolution process is great significant for the prediction,ecological assessment and restoration of rocky desertification in karst areas.In this paper,peak cluster depressions located in Guangxi,Yunnan and Guizhou provinces were selected as the study area,and all available Landsat remote sensing images and field measurement data of six years including 1994,1999,2004,2009,2014 and 2018 were collected to construct a time-series data set.Firstly,the lithology vector and land use type data were used to mask the non-karst area.Secondly,the single index of greenness,humidity and brightness was extracted by using the hat transform to construct the Karst Rocky Desertification Composite Index(KRDI).Then,the time spectrum feature space model(TSFS)was used to determine the degree of rocky desertification.Finally,the spatial and temporal evolution characteristics of rocky desertification were described through the indicators of evolution direction,evolution rate and evolution frequency,and the driving factors of rocky desertification were further explored.The influences of lithology,slope,elevation,precipitation and human activities on rocky desertification were analyzed.The main results are as follows:(1)KRDI is a sensitive index of karst rocky desertification.With the increase of the degree of rocky desertification,the KRDI value shows an increasing trend,that is,the KRDI value of heavy rocky desertification is the highest,while that of light rocky desertification is low.In addition,there are certain differences in KRDI values in different seasons.The value of KRDI is the lowest in summer and the highest in winter.(2)TSFS model can accurately monitor and identify different degrees of rocky desertification.The selected key indicators performed well,with Kappa coefficient of 0.90 and overall accuracy of 0.92.Therefore,the index combining spectral information with spatial and temporal information is a promising method to extract karst rocky desertification information,which has important application potential under similar geological conditions.(3)The spatial distribution pattern of rocky desertification is heavier in northwest and lighter in southeast,according to the evolution pattern,rate and frequency of rocky desertification.The long-term change trend shows that most of the study areas have been improved and effectively contained rocky desertification,but the evolution of local rocky desertification has a fluctuation.To sum up,rocky desertification area,the vegetation and green degree of rock,the temporal characteristic of humidity and brightness changes with the seasons as discriminant rocky desertification provides a solid theoretical basis.The method of spectral analysis in this paper,based on the building-extraction model of karst rocky desertification,provides a new method to accurately extract desertification information,have important potential applications in the similar geological conditions.Based on this model,the spatial and temporal distribution of rocky desertification in peak-cluster depression during 1994-2018 was extracted,and then the temporal and spatial evolution characteristics of rocky desertification were analyzed.It provides a scientific basis for the analysis of the driving force of rocky desertification,the understanding of the development mechanism of rocky desertification,and the prevention and treatment of rocky desertification.