| The Huangshui River Basin,located in the northeastern part of the Qinghai-Tibet Plateau,is characterized by diverse landforms,fragmented land surfaces,and complex land use.It serves as an important economic hub and ecological function area for the Qinghai provincial capital and surrounding areas,bearing the significant responsibility of driving economic development in Qinghai Province and ensuring ecological security for the eastern region.However,the socio-economic development in the basin has led to an imbalance between land use and ecological protection,posing a significant scientific challenge for local governments and researchers to strike a sustainable balance between the two.Previous similar studies have predominantly focused on developed provinces and cities in the eastern region of China,with limited attention given to underdeveloped western areas,including Qinghai-Tibet Plateau’s ecologically fragile regions and key ecological functional areas.Therefore,the unique geographical environment,diverse ecosystems and complex landforms of the Qinghai-Tibet Plateau make the classification of land use / land cover(LULC)extremely complicated,and how to acquire accurate long time sequence LULC classification data of this region has become an important challenge for the field of plateau earth science.Research on the land use/land cover(LULC)classification,driving forces,and simulation in the Huangshui River Basin holds regional significance.The findings from this study can provide valuable insights for the development of plateau regions and other areas.This study focuses on the Huangshui River Basin and employs the Google Earth Engine(GEE)cloud computing platform,using long time series imagery from Landsat TM/ETM+/OLI as the data source.By optimizing key steps and factors such as samples,temporal composition,and features of classification and classifiers,yearly LULC classification datasets for the period 1987-2023 are obtained.The study explores the spatiotemporal characteristics of LULC changes over the past 37 years,considering aspects such as LULC quantity,structure,transition patterns,land use intensity,spatial autocorrelation,and landscape patterns.Furthermore,the study employs a geographic detector to investigate the natural,socio-economic,and policy-driven factors influencing LULC changes in the basin,aiming to uncover the driving mechanisms behind LULC.Additionally,mainstream models including ANN-CA,FLUS,and PLUS are applied to simulate and compare historical LULC patterns in the basin,with the accuracy of these models compared.Development scenarios such as inertial development,cultivated land protection,and ecological protection are constructed.The LULC simulation model with the highest accuracy is selected,and the spatial distribution patterns of LULC in the basin under different development scenarios are predicted for the year 2040.These predictions aim to provide scientific support for the future planning and decision-making regarding land resources in the basin.Major conclusions of this study are presented as follows:(1)On the GEE platform,semi-supervised classification for sample selection,multi-seasonal images,integrated learning classifiers and classification features optimization were adopted to produce the watershed time sequence LULC classification data set.The overall accuracy(OA)of each yearly LULC dataset ranges from 83.6% to 89.8%,with Kappa coefficients ranging from 0.818 to 0.885,indicating a good level of accuracy.To address issues such as missing LULC information,“salt and pepper” noise,and classification errors in the initial LULC datasets,temporal LULC interpolation and spatiotemporal filtering methods are employed for optimization.The optimized temporal LULC dataset shows improved OA ranging from87.6% to 91.5%,with Kappa coefficients ranging from 0.86 to 0.90.,which can provide an accurate LULC classification data set for LUCC analysis in later phases.(2)Grassland is the most predominant LULC type within the basin,with an area percentage ranging from 35.26% to 39.17%,and it’s collectively distributed in some high mountainous areas and valley-shallow mountain transition belts.Over the past 37 years,there has been an overall decreasing trend in grassland area.Cultivated land ranks as the second most prevalent LULC type in the basin,collectively distributed in shallow mountains and middle-high mountains,with an area percentage ranging from 25.99%to 29.77%.Around the year 2000,there was a transition from an increasing to a decreasing trend in cultivated land area.The area percentage of forest land(27.20% to29.63%)is comparable to that of cultivated land,and the forest land is collectively distributed in middle-high mountains.Over the past 37 years,forest land area has shown a steady growth.Unused land occupies a percentage ranging from 3.05% to 3.71%,exhibiting a slow decreasing trend since 1987.Following the implementation of the Western Development Strategy after 2000,the area of land designated for construction within the basin increased rapidly,with an area percentage ranging from 1.38% to3.89%,and spatially,it is largely distributed along the X-shaped valley centered around Xining urban areas.While water bodies within the basin account for a small proportion of area(0.91% to 1.60%),their percentage has experienced substantial fluctuations,notably displaying a consistent decrease after 2005.(3)Based on the evolution law of Land Use Intensity Index(LUII)and Land Use Overall Change Index(LUOI)in the basin,the period from 1987 to 2000 was classified as a phase of natural development in LULC within the basin,during which there was a rapid conversion of a significant amount of grassland into cultivated land,while changes in construction land and unused land were relatively slow.From 2000 to 2010,the basin entered a phase of rapid LULC development.Influenced by policies such as the Western Development Strategy,returning farmland to forest(grassland),and the construction of water conservation forests,cultivated land rapidly shifted to forest and grassland,and the area of construction land doubled.The period from 2010 to 2023 represents a phase of stable LULC development within the basin,when the area of construction land continued to increase,but the speed of conversion between the main LULC types(cultivated land,grassland,and forest land)in the basin slowed down.Over the past 37 years,most of the LULC transition patterns within the basin were aligned with the general patterns of LUCC in China,but unique transition models also emerged,such as the transformation of cultivated land to forest land and grassland types,a result of policies issued to return farmland to forest(grassland)for ecological protection.(4)Over the long term,there has been significant positive spatial autocorrelation among various LULC types within the basin.Cultivated land and forest land exhibit the strongest spatial autocorrelation,while grassland,construction land,and unused land show relatively strong spatial autocorrelation.Water bodies demonstrate weaker spatial autocorrelation.During historical periods,the spatial clustering of cultivated land went from strong to weak starting from 2000,while the spatial clustering of grassland types turned from weak to strong in the same period.The spatial clustering of forest land and construction land continued to strengthen,while water bodies and unused land types gradually reduced their spatial clustering.Analysis of landscape index showed that2,000 years later,the landscape heterogeneity will be enhanced,and patches develop towards the trend of dispersion and fragmentation,leading to a more intricate distribution of various LULC types,indicative of heightened human interference.(5)In the phase of natural LULC development within the basin,the pattern of LULC and Land Use Intensity Index(LUII)changes within the basin was mainly influenced by natural and socio-economic driving factors such as altitude,temperature,anthropogenic disturbance index,and population distribution.Entering the phase of rapid LULC development within the basin,these changes were primarily influenced by strengthened socio-economic and policy-driven factors,while the influence of natural driving factors decreased.In the phase of stable LULC development within the basin,the influence of socio-economic and policy-driven factors continued to strengthen.The patterns of LULC and LUII within the basin during each period were influenced by a combination of various driving factors.The interaction types between natural,socioeconomic,and policy driving factors mainly involved dual-factor enhancement and non-linear enhancement relationships.The combination of most driving factors during each period significantly impacts the spatial distribution of LULC and LUII.(6)Comparing the performance of the PLUS,FLUS,and ANN-CA models in simulating LULC patterns across multiple development stages within the basin,the PLUS model demonstrates the highest accuracy and finer simulation results.The OA ranges from 84.16% to 88.82%,with Kappa coefficients between 0.774 and 0.838.Using the PLUS model,the LULC distribution patterns in the Huangshui River Basin for the year 2040 are predicted under three development scenarios.In the inertial development scenario,the construction land will spatially expand quickly,with a significant conversion of cultivated land to construction land and forest,posing a serious threat to food security in the basin.Under the cultivated land protection development scenario,the ongoing trend of continuous decrease in cultivated land within the basin will be effectively curbed,leading to a short-term recovery in cultivated land area and substantial restrictions on the expansion rate of construction land.In the ecological protection development scenario,the areas of forest land,grassland,and water bodies,among other ecological land types,will see significant increase.The expansion rate of construction land will be restricted to half of that in the inertial development scenario,contributing to the maintenance of ecological security within the basin. |
