Spatiotemporal Evolution And Future Trend Prediction Of Vegetation NPP In Huangshui River Basin Of Qilian Mountains

Net Primary Production（NPP）is a measure of the total organic matter produced by green plants through photosynthesis during a certain period of time,after subtracting the portion consumed by plant respiration and growth.It reflects the productivity of the vegetation community and can be used to estimate the carbon balance of terrestrial ecosystems,as well as to describe the processes of carbon cycling and energy flow.The Qilian Mountains,which span the Gansu and Qinghai provinces in western China,are an important ecological conservation area.The Huangshui River Basin,located in the southeastern part of the Qilian Mountains,is at the junction of the Qinghai-Tibet Plateau and the Loess Plateau,and is a demonstration area for the Qilian Mountains ecological conservation efforts.However,the effects of the conservation policies implemented by China in the Huangshui ecological conservation area on future vegetation growth in the region remain unclear.Moreover,most research on vegetation NPP in the Huangshui River Basin is based on existing data and focuses on studying the spatial and temporal characteristics of past vegetation changes.The spatial and temporal evolution of vegetation NPP is usually the result of multi-factor interactive responses and has uncertain non-linear characteristics.As a result,the guidance significance of research results is limited,and exploration of future spatial and temporal trends of vegetation has been rarely mentioned.Therefore,analyzing the spatiotemporal evolution of vegetation in the past,and even predicting future spatiotemporal trends based on existing vegetation index data,is of great significance for maintaining regional ecological quality and monitoring vegetation changes.In this study,we used multiple data sources,including temperature,precipitation,solar radiation,and vegetation type,to estimate the vegetation net primary productivity（NPP）data in the Huangshui River Basin from 2000 to 2019 using the CASA model.We investigated the spatiotemporal variation patterns of NPP,including spatiotemporal distribution features,NPP centroid migration,relationships with various factors,and specific performances under different vegetation types.Subsequently,we constructed a Conv GRU prediction model that considers spatial neighborhood information and can adapt to nonlinear and complex relationships.We compared and evaluated this model with four other prediction models,including MLP,SVR,LSTM,and GRU,on seven indicators.We employed the Conv GRU neural network prediction model to predict the future spatiotemporal trends of vegetation NPP（2020-2029）based on the existing vegetation NPP data in the Huangshui River Basin from 2000 to 2019.Furthermore,we analyzed the spatiotemporal distribution features and spatial agglomeration of future vegetation NPP,and traced the reasons for the spatiotemporal variation of future vegetation NPP from a seasonal perspective.This study intends to provide theoretical support for understanding the regional ecological environment of the Qilian Mountains and for the formulation and implementation of ecological measures.In summary,the main conclusions of this study are as follows:（1）From 2000 to 2019,the net primary productivity（NPP）of the Huangshui River basin in the Qilian Mountains exhibited significant spatial and temporal heterogeneity,ranging from400 to 600 g C·m^-2 and showing a clear increasing trend.Most regions showed a significant increase in NPP,with only a few areas displaying a decreasing trend.Among the four different vegetation types,grassland and alpine meadow had growth rates of 7.27 g C·m^-2·a^-1 and 7.72g C·m^-2·a^-1,respectively,while cropland and forest had growth rates of 8.72 g C·m^-2·a^-1 and 8.44g C·m^-2·a^-1,respectively.The changes in NPP over time were highly similar in grassland and alpine meadow,as well as in forest and cropland types.The NPP centroid of grassland and alpine meadow types showed a trend of"north to south,west to east"displacement,while the NPP centroid of cropland type shifted more towards the southeast.In the forest type,the NPP centroid shifted towards the south in the north-south direction,and had a turning point in the east-west direction.NPP was mainly negatively correlated with temperature（77.54%of the area）,but the correlation was weak.NPP was mainly positively correlated with precipitation,with only a few areas showing a weak negative correlation.NPP was significantly positively correlated with solar radiation,which may be influenced by other factors.About 67.22%of the regions showed a positive correlation between NPP and NDVI,while only 32.78%of the regions showed a negative correlation.（2）A spatiotemporal prediction model for vegetation NPP in the Huangshui River basin from 2020 to 2030 was developed using a Conv GRU neural network that considers spatial neighborhood information and adapts to nonlinear and complex relationships.The original data used for the model was the 2000-2019 vegetation NPP data obtained from the inversion of the CASA model.The Conv GRU model demonstrated superior performance compared to MLP,SVR,LSTM,and GRU in seven performance metrics,preserving the spatial continuity of the data while ensuring small errors and high accuracy,and capturing the spatiotemporal characteristics of the raster data.（3）From 2020 to 2030,the overall vegetation NPP in the Qilian Mountains basin remained at a level of 640-690g C·m^-2,exhibiting a clear increasing trend with an annual growth rate of3.19g C·m^-2·a^-1.Most areas within the region showed an increasing and expanding trend in vegetation NPP,while a few areas exhibited a decreasing trend.The global Moran’s I index for NPP in the Huangshui River basin was 0.84 and 0.85 for the periods of 2000-2019 and 2020-2030,respectively,demonstrating significant spatial distribution characteristics.A significant hot spot contraction or cold spot expansion trend was observed in the region.In the upstream of the Huangshui River,the northwest and southern corners of the basin exhibited a hot spot expansion trend,while the north and upstream areas of the Shatang River and the northwest section of the Datong River exhibited a hot spot contraction trend,indicating that these areas may face the risk of vegetation degradation in the future.