As a critical ecological barrier and carbon reservoir on the eastern Qinghai-Xizang Plateau, the Qilian Mountains (QLMs) play a vital role in maintaining regional ecological security. This study employed an enhanced Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model with dynamic interannual carbon density parameterization to evaluate carbon storage (CS) variations in the QLMs during the historical period of 1990-2020 and future period of 2030-2100 under climate scenarios (SSP126, SSP245, and SSP585, where SSP is the Shared Socio-economic Pathway). By integrating GeoDetector analysis, random forest modeling, and scenario simulations, we further assessed climate-human interactions and their impacts on CS. The QLMs had a multi-year average CS of 1.49×10⁹ t during 1990-2020, with higher values in the southeast and lower values in the northwest. The spatial heterogeneity of CS was primarily driven by warming-wetting climatic gradients. Multi-scenario projections revealed divergent trajectories: under the SSP126 scenario, ecological restoration measures (e.g., mountain closure and afforestation) increased CS, whereas CS under SSP245 and SSP585 scenarios exhibited inverted U-shaped patterns. Notably, SSP585 resulted in 8.00×10⁷ t losses due to extensive conversions from grassland to urban area during 2050-2100. Grassland and cropland CS, together with precipitation, emerged as the key determinants of regional CS dynamics, with a synergistic interaction between precipitation and temperature. This study introduced a century-scale, multi-scenario framework that captures temporal interactions and dynamically parameterizes carbon density coefficients for era-specific accuracy. The findings suggest optimizing vegetation coverage, managing urban expansion, and implementing climate adaptation strategies to enhance the total CS of this ecologically sensitive region.

The Yellow River Basin (YRB), located in the mid-latitude region of China, encompasses diverse ecosystem types and is highly sensitive to climate change. However, the temporal patterns and environmental drivers of net ecosystem CO₂ exchange (NEE) across multiple time scales remain poorly understood. Using eddy covariance observations from the ChinaFLUX network collected between 2003 and 2020, this study investigated the temporal dynamics of NEE and its primary environmental controls in five representative ecosystem types within the YRB and its adjacent 100-km buffer zone: cropland, forest, grassland, shrubland, and wetland ecosystems. The results showed that all five ecosystems exhibited a generally U-shaped diurnal pattern from May to September, characterized by net CO₂ uptake during the daytime and net CO₂ release at night. At the daily scale, cropland displayed a typical bimodal carbon uptake pattern, whereas forest ecosystem exhibited the greatest day-to-day variability in NEE. In contrast, grassland, shrubland, and wetland ecosystems showed relatively smooth daily fluctuations. The net CO₂ source-sink functions derived from NEE differed substantially among ecosystem types. Forest ecosystems acted as the most stable and persistent carbon sinks, whereas croplands exhibited short-term but high-intensity carbon uptake. Wetlands showed pronounced interannual variability, including an extreme net CO₂ release event at the Haibei wetland site in 2007. Grassland and shrubland ecosystems were more susceptible to environmental stress and could shift from net CO₂ sinks to net CO₂ sources during drought years. The environmental controls on NEE exhibited clear time-scale dependence. At the half-hourly scale, photosynthetically active radiation (PAR) was the dominant driver of NEE variability, the influence of temperature increased progressively from the daily to monthly scales. These findings improve the understanding of regional carbon dynamics in the YRB, provide insights into the net CO₂ source-sink status of different ecosystem types, and elucidate the mechanisms regulating ecosystem CO₂ exchange across multiple temporal scales.

High false alarm rates (FARs) in snowmelt flood forecasting persist, largely due to an insufficient understanding of the coupled effects of multi-source hydro-meteorological drivers and their inherent time lags. This study addressed this gap by developing a hybrid snowmelt-flood forecasting framework that combined multiple linear regression (MLR) and backpropagation neural network (BPNN) models, with a simulated annealing (SA) algorithm employed to optimize the ensemble weighting. The developed hybrid model that incorporated daily hydro-meteorological inputs and temporal factor was validated using data from 1978 to 2011 for the Hutubi River Basin, an arid inland basin on the northern slope of the Tianshan Mountains, China. The findings demonstrated that the hybrid model achieved a specificity of 0.8549, significantly outperforming standalone MLR (0.6024) and BPNN (0.6436) models. Correspondingly, the FAR was reduced to 0.1451, which was substantially lower than that of MLR (0.3976) and BPNN (0.3564). Upon integrating temporal factor, the FAR was further reduced to 0.0959, markedly enhancing overall predictive robustness. Collectively, this study offers a robust methodological framework for optimizing snowmelt flood forecasting by effectively integrating multi-source data and temporal dependencies.

Actual evapotranspiration (ETa) is a key regulator of land-atmosphere water and energy exchanges in arid regions. However, the coupled effects of its controlling factors along mountain-basin gradients remain poorly quantified, limiting precise water resource management. Based on multi-source data (2001-2022) from Xinjiang Uygur Autonomous Region, China, this study analyzed the spatiotemporal patterns of ETa across five eco-zones: the Altay Mountains, Junggar Basin, Tianshan Mountains, Tarim Basin, and Kunlun Mountains. Using Mann-Kendall and Theil-Sen trend analyses, GeoDetector, and Random Forest model, we quantified the contributions and interactions of seven environmental drivers: air temperature, precipitation, soil moisture, wind speed, net radiation, vapour pressure deficit, and leaf area index. The multi-year mean ETa over Xinjiang was 193.15(±21.16) mm/a, showing a distinct ''high in mountains and low in basins'' spatial pattern. A significant increasing trend of 1.60 mm/a was observed regionally, yet with clear spatial divergence, i.e., ETa increased in the major mountain ranges but decreases or stabilized in the hyper-arid Tarim Basin and transitional Junggar Basin. Water-supply factors (precipitation and soil moisture) explained about 40% of the spatial variance individually, while factor interactions enhanced the explanatory power substantially (maximum interaction gain (Δq)=0.18), indicating that bivariate and nonlinear combinations captured ETa patterns more effectively than single driver. Interaction regimes were strongly ecozone-dependent and varied between wet and dry years. Through comparing the wet year (2003) with the dry year (2022), we found a reorganization of dominant interaction pairs: the hyper-arid Tarim Basin shifted from hydrothermal control to vegetation-water-stress coupling; the Junggar Basin maintained persistent vegetation-energy coupling; the Altay and Tianshan Mountains strengthened vegetation-mediated interactions during drought; and the Kunlun Mountains transition from hydrothermal to moisture-aerodynamic control. These results clarify how complex topography and vegetation physiology modulate macroclimatic forcing in this arid mountain-basin system. These findings offer a process-based foundation for ecohydrological zoning, and provide critical insights for adaptive water ecosystem management and sustainable development in response to the "warming and wetting" trend in global drylands.

The middle and lower reaches of the Irtysh River form a transboundary ecological corridor across the Republic of Kazakhstan and Russia and represent an environmentally sensitive region in arid Central Asia. Understanding long-term vegetation dynamics in this region is essential for evaluating ecological stability and supporting cross-border ecosystem management. However, existing studies are mostly confined to individual administrative units, breaking the eco-hydrological integrity of transboundary basins; meanwhile, most analyses rely on single annual-mean vegetation metrics, failing to reveal the differentiated variation patterns and long-term persistence of LAI across different vegetation growth states. This study used the Global Inventory Modeling and Mapping Studies Leaf Area Index 4g (GIMMS LAI4g) dataset and Climate Research Unit (CRU) precipitation data from 1982 to 2020, to investigate the spatiotemporal dynamics of leaf area index (LAI) across 14 transboundary subregions in the middle and lower reaches of the Irtysh River Basin. Specifically, this study applied Mann-Kendall trend analysis and Theil-Sen median slope estimator to identify long-term trends, employed rescaled range analysis evaluate long-term persistence, conducted Pearson correlation analysis to quantify the relationship between LAI and precipitation, and used spatial pattern analysis to characterize regional heterogeneity. The results showed significant increases in annual mean and maximum LAI, with Sen's slopes of 0.0025/a and 0.0077/a, respectively, whereas annual minimum LAI exhibited a significant decreasing trend (-0.0011/a). Pronounced spatial heterogeneity was observed among mountainous and piedmont regions, steppe-riparian transition zones, and downstream forest-wetland landscapes. Rolling-window analysis revealed relatively stable vegetation dynamics during the early period, followed by enhanced spatial divergence and increased interannual variability after the early 2000s. All LAI indicators exhibited strong persistence, with Hurst exponents exceeding 0.7000 across the study area, indicating that the observed vegetation trajectories are likely to persist in the future. Precipitation showed significant positive correlations with annual maximum and mean LAI in 64.29% and 57.14% of the study area, respectively, whereas annual minimum LAI showed generally weak responses to precipitation variability. These findings improve understanding of vegetation dynamics and future persistence in transboundary arid river basins and provide scientific support for ecological conservation and sustainable watershed management in Central Asia.

In the arid regions of northern Xinjiang Uygur Autonomous Region (hereafter referred to as northern Xinjiang), China developing intensive and sustainable cropping patterns is crucial for ensuring food security and promoting green agricultural development. As a typical cereal and legume intercropping system, maize and peanut intercropping demonstrates significant potential to enhance yield and resource use efficiency. However, the effects of nitrogen management and row configuration on crop productivity and inter-specific interactions under the unique ecological conditions of in northern Xinjiang remain unclear. This study employed a field experiment with four planting patterns (maize monoculture (MM), peanut monoculture (MP), four rows of maize intercropped with four rows of peanuts (4M4P), and six rows of maize intercropped with six rows of peanuts (6M6P)) and five nitrogen application levels (0 (N₀), 100 (N₁₀₀), 200 (N₂₀₀), 300 (N₃₀₀), and 400 (N₄₀₀) kg/hm²). The results showed that under the 6M6P intercropping pattern combined with the N₂₀₀ application level, maize morphological parameters were significantly improved, with plant height increasing by 10.48%, stem base circumference by 13.64%, and root length density (RLD) by 47.63%. Additionally, adaptive root morphological responses of maize were promoted in deeper soil depths, with RLD increasing by 15.48%-54.76% in the 0-60 cm soil depth. Intercropping significantly inhibited peanut growth and root development, as reflected by reductions of 12.37% in main stem height, 27.01% in dry matter weight, and 15.73%-25.23% in RLD in the 0-20 cm soil dept. However, nitrogen application at the N₂₀₀ level effectively alleviated this inhibition, increasing peanut aboveground nitrogen accumulation by 80.12%-92.93% compared with the N₀ level. Although the peanut nitrogen harvest index (NHI) decreased by 6.21%-7.06% relative to MM, it remained at a relatively high level. This planting configuration also significantly enhanced maize dry matter accumulation (increased by 28.50%) and nitrogen uptake (increased by 38.65%-55.21%) compared with MM. The nitrogen use efficiency (NUE) of both crops was the highest at the N₁₀₀ and N₂₀₀ levels, whereas N₃₀₀ and N₄₀₀ levels suppressed the aforementioned indicators. A positive correlation was observed between dry matter accumulation and nitrogen accumulation. Nitrogen translocation was enhanced in maize and allocation to grains was improved through improved morphological plasticity and root system adaptation, as reflected by its NHI, remaining at a high level of 79.54%-83.00%. This configuration achieved the highest land equivalent ratio (1.30) and maize grain yield under the experimental conditions. Together, these findings elucidate how nitrogen management and row configuration regulate crop growth and competitive dynamics in intercropping systems, providing a scientific basis for optimizing cereal and legume intercropping management in arid regions.

Dryland ecosystems, encompassing arid to semi-arid regions, impose strong climatic and edaphic constraints that profoundly shape plant functional traits and secondary metabolism. Understanding how environmental factors regulate phytochemical biosynthesis is essential for biodiversity conservation and sustainable resource management under increasing aridity. Oliveria decumbens Vent., an endemic medicinal species of the drylands of Fars Province, Iran, provides an excellent model for exploring the ecological determinants of metabolite variability in water-limited habitats. We integrated ecological predictors with machine learning to model the spatial variation of thymol and carvacrol concentrations across 59 georeferenced populations of O. decumbens. Three predictive models—Random Forest (RF), Support Vector Regression (SVR) with a Radial Basis Function (RBF) kernel (SVR-RBF), and a hybrid ensemble (RF+SVR-RBF)—were developed and evaluated. Model performance was quantified using root mean squared error (RMSE), mean absolute error (MAE), coefficient of determination (R²), and the concordance correlation coefficient (CCC). Generalized Linear Model (GLM) was applied to identify key environmental variables regulating metabolite biosynthesis. The hybrid ensemble consistently outperformed individual model, achieving the highest predictive accuracy (R²=0.82 for thymol and R²=0.80 for carvacrol). Spatial mapping revealed pronounced heterogeneity in metabolite distribution, with distinct functional hotspots in the northern and western semi-arid regions of Fars Province. GLM analysis indicated that mean annual temperature, slope aspect, slope degree, and sand content were strong positive predictors of thymol and carvacrol accumulation, whereas high soil potassium, clay percentage, and alkaline pH constrained metabolite production. This study shows that hybrid ensemble modeling effectively captures how environmental gradients regulate secondary metabolism in dryland plants. The proposed framework, combining RF with SVR-RBF, is transferable across arid environments. These findings support trait-based ecological predictions and offer practical insights for conservation and sustainable cultivation of high-value medicinal plants in water-limited ecosystems.

Vegetation restoration is a critical approach for improving soil quality in arid ecosystems, yet its effects on soil extracellular enzyme activities (EEAs), stoichiometric relationships, and nutrient dynamics in the Gobi Desert remain poorly understood. This study aimed to address this knowledge gap by analyzing a chronosequence of three stand ages (3, 11, and 14 a) of artificial sea buckthorn (Hippophae rhamnoides L.) plantations in the 170^th Regiment of the 9^th Agricultural Division of the Xinjiang Production and Construction Corps, China. Three 20 m×20 m quadrats were established for each restoration stage in early to mid-August 2024. Soil samples were collected in 0-10, 10-20, and 20-30 cm soil layers within each quadrat using a five-point method, to explore the ecological changes in Gobi Desert soil throughout the vegetation restoration. The results indicated that soil organic carbon (SOC), particulate organic carbon (POC), dissolved organic carbon (DOC), mineral-associated organic carbon (MAOC), total nitrogen (TN), and microbial biomass carbon (MBC), nitrogen (MBN), and phosphorus (MBP) all significantly increased in the 0-10 cm soil layer during vegetation restoration. In the 10-20 cm soil layer, MAOC and TN significantly increased during vegetation restoration, while in the 20-30 cm soil layer, MAOC, TN, MBN, and MBP significantly increased. Concurrently, the activities of key enzymes (β-1,4-glucosidase and β-1,4-N-acetylglucosaminidase) involved in carbon (C) and nitrogen (N) cycling were also the highest in the 14-year-old plantation in the 0-10 cm soil layer. Further analysis of enzymatic stoichiometry indicated a shift in microbial resource allocation, with significant increases in the C-N and C-phosphorus (P) ratios of EEAs, whereas the N-P ratio remained relatively stable. The drivers of these stoichiometric ratios varied with soil depth: in the 0-20 cm soil layer, the C-N and C-P ratios of EEAs were mainly influenced by soil C content, whereas the N-P ratio was not regulated by any factors. In the 20-30 cm soil layer, the C-N and C-P ratios of EEAs were mainly influenced by soil C, N, and microbial biomass content, whereas the N-P ratio was mainly affected by TP and pH. Overall, the establishment of sea buckthorn plantations, especially over a 14-a period, significantly improved Gobi Desert soil quality by enhancing nutrient content, stimulating microbial activity, and modulating extracellular enzyme strategies. These findings underscore the considerable C sequestration potential and ecological benefits of targeted artificial afforestation in arid land restoration.

Although salinity functions as a key environmental filter for soil microorganisms, its comprehensive effects on soil microbial community structure and assembly processes remain poorly understood, particularly in arid regions prone to salinization. In this study, soil bacterial communities along a natural salinity gradient at the edge of Ebinur Lake in Northwest China were investigated. Using 16S ribosomal RNA (16S rRNA) sequencing, we examined variations in soil bacterial community structure, co-occurrence patterns, and assembly mechanisms across different soil salinity groups, including lightly salinized soils (LSS), moderately salinized soils (MSS), and heavily salinized soils (HSS). The results showed that soil bacterial diversity varied significantly among salinity groups, with the highest value observed in LSS. Community dissimilarity increased notably with greater variations in salinity. Notably, a systematic shift in soil bacterial composition occurred along the salinity gradient, with salt-sensitive bacterial phyla (e.g., Acidobacteria and Gemmatimonadetes) being progressively replaced by salt-tolerant ones (e.g., Firmicutes and Bacteroidetes). Network analysis underscored that increased salinity led to reduced soil bacterial network complexity and stability. More positive correlations among soil bacteria occurred in HSS, suggesting a potential shift toward cooperative microbial strategies under severe salt stress. Moreover, the assembly processes governing soil bacterial communities transitioned from stochastic process, predominantly in LSS (71.43%) and MSS (80.00%), to deterministic process in HSS (66.66%). In summary, the results emphasize the multifaceted role of soil salinity in shaping soil bacterial communities in arid ecosystems, thereby enhancing the understanding of the impacts of soil salinization on soil microbial dynamics.