Evolution And Prediction Of Valley Type Cities/towns In The Huangshui River Basin Based On Long Time-series Landsat Images In Recent 36 Years

The Huangshui river basin is located in the transition zone between Qinghai-Tibet Plateau and Loess Plateau.Its valley type city/town cluster is an important component of the largest trans-provincial city/town cluster in west China.Its unique geographical location makes it easy to be affected and disturbance by human activities.Plot and obtain long-term and continuous Land Use and Land Cover(LULC)classification spatio-temporal data of the study area,and explore the spatio-temporal conversion characteristics and rules of urban/urban land use/land cover.It is of great significance to analyze and study the expansion and evolution of cities/towns brought by human activities,effectively simulate and predict the development of cities/towns in the future,and scientifically plan and manage cities/towns.Based on the multi-seasonal Landsat long time series remote sensing images provided by Google Earth Engine(GEE)cloud computing platform,this study combined the random forest algorithm and numerous representative characteristic variables extracted from Landsat data,A new method of extracting continuous long time series land cover classification in Huangshui river valley was explored.Then,based on this data set,this paper uses the methods of land cover structure change analysis,transfer matrix,landscape index,etc.,to reveal the spatial and temporal evolution pattern of urban/urban in the Huangshui Valley of Qinghai Province in the past 36 years from 1987 to 2022.Based on PLUS-CA and FLUS-CA models,firstly,the city/town development conditions during2000-2014 and 2014-2022 were simulated and compared,and the optimal model was selected.On this basis,three scenarios of future land use/land cover change were constructed,and the spatio-temporal pattern of urban/urban land cover change in 2030in Huangshui Valley was further predicted.The main research conclusions are as follows:(1)Based on GEE cloud computing platform,the LULC data of nearly 36 years from 1987 to 2022 were obtained by random forest algorithm based on multi-seasonal Landsat remote sensing images combined with auxiliary classification data such as spectral features,terrain features and texture features.The accuracy evaluation results showed that the average overall accuracy was 0.83±0.03,and the average kappa coefficient was 0.78±0.03,indicating that the method was feasible to apply Landsat data with 30 m resolution to land cover mapping in the valley at the basin scale.(2)The land cover classification and post-processing methods based on continuous time series such as null fill algorithm,image denoising algorithm and spatio-temporal consistency detection algorithm can effectively solve the problems of null value,salt and pepper speckle,misclassification and missing grade of classified data caused by image quality problems in the process of drawing continuous long time series images.The change of long time series data is logical and regular.(3)The spatio-temporal evolution analysis showed that the cultivated land and grassland in the study area continued to decrease from 1987 to 2022,and the urban/urban construction land area increased from 274.80 km~2in 1987 to 573.96 km~2in 2022,indicating that the urban/urban area showed a significant expansion trend in the past 36 years.The spatio-temporal changes of land cover in the upper,middle and lower reaches of the valley region and the resulting city/town development patterns are significantly different.The results of this study will help government decision makers and city/town managers to further develop sustainable city/town development policies based on the urban/town development patterns of the previous study areas,and to build more sustainable,green and resilient cities/towns.(4)The prediction of future urban/urban change expansion in the valley shows that in 2030,the study area will expand under the three scenarios.Under the scenario of rapid urban/urban development,the urban/urban construction land change is the most drastic,and the area will increase by 92.01 km~2compared with that in 2022.The area of construction land increased by 80.98 km~2under the urban/urban natural development scenario,and only increased by 36.59 km~2under the urban/urban ecological protection development scenario.The development hotspot of cities/towns in the Huangshui River Valley in the future was similar under the three scenarios.It mainly concentrates in Chengbei District of Xining,south Huangzhong River valley area,Datong County River Valley area and west Lanzhou River Valley area.The forecast results will provide reference for the government’s future development orientation and model of the study area,and help the government to make better future development planning.