Climate change significantly affects the arid/humid processes and patterns in China, directly impacting management decisions related to adaptive agriculture and water resources management, desertification control, and spatial ecological restoration. However, current studies primarily focus on changes in arid/humid climate variables, lacking quantitative characterization of the dynamic evolution of areal systems and their nonlinear responses. Based on the data of national meteorological stations from 1961 to 2020, we systematically quantified the nonlinear response of arid/humid patterns to climate change. The results revealed that 6.98% of eco-geographical arid/humid regions underwent type shifts over the past six decades, with 4.95% transitioning toward wetter conditions. Humid and semi-arid regions expanded significantly while sub-humid and arid regions contracted significantly. In the late 1990s, trends of the humid and sub-humid region shifted. Humid region contraction in northern China was driven primarily by precipitation decline, whereas the Tibetan Plateau responded to increasing potential evapotranspiration. During the same period, the retreat rate of the arid region slowed, linked to intensified aridification in the west part of northern China and a decelerating wetting trend in northwest China, both primarily driven by precipitation trends. Our study reveals the nonlinear response of the arid/humid patterns under climate change, providing a scientific basis for the improvement of regional climate resilience.

Drought cascade propagation critically constrains sustainable water resource utilization, yet quantitative characterization of dynamic drought propagation thresholds and their multi-process drivers remains limited. This study integrates meteorological (SPI), hydrological (SRI), and soil moisture drought indices (SSI), run theory, Copula models, and machine learning to analyze spatiotemporal patterns and drivers of drought propagation thresholds in the Mongolian Plateau (1982-2021). The results show that over 30% of the study region exhibited drying trends for SPI (-0.25/10a), SRI (-0.19/10a), and SSI (-0.34/10a), with SPI/SRI showing "high frequency, short duration, low intensity" features versus SSI's "low frequency, long duration, high intensity" pattern. Droughts propagate sequentially as "meteorological drought→ hydrological drought→soil moisture drought", averaging 2 and 9.01 months between stages. The probability of drought cascade propagation showed the triple characteristics of "positive correlation of rank, negative correlation of intensity and attenuation of path". Propagation thresholds are lower in mountainous forests (drought-resistant) but higher in central desert-grassland transition zones (prone to cascading effects), forming a pattern of core vulnerability and edge insensitivity. Notably, 59.58% and 49.20% of areas show rising SPI→SRI and SRI→SSI propagation thresholds, respectively. In addition, precipitation dominated the SPI→SRI propagation (contribution rate 43.56%), soil moisture content (contribution rate 60.14%) became the key bottleneck of SRI→SSI propagation, and NDVI, potential evapotranspiration and temperature synergistically regulated the cascade risk of the marginal transition zone. The results of the study can provide a basis for precise early warning of cascading drought in arid and semi-arid zones, as well as theoretical references to reduce the negative impacts of drought propagation on ecosystems and realize the sustainable development of regional ecology and agriculture and animal husbandry.

With the rapid global warming and increasing atmospheric pollutant emissions, frequent urban ozone pollution events pose a serious threat to human health. Ozone formation is complexly influenced by multiple factors, including pollutant emissions and meteorological conditions, leading to generally low accuracy in current ozone concentration forecasts. This study constructs an undirected graph based on the spatial relationships among monitoring sites, using observational data from environmental quality monitoring stations, meteorological reanalysis data, and emission data. A spatiotemporal graph convolution model based on GCN-GRU (Graph Convolutional Networks-Gated Recurrent Unit) was developed to simulate the spatiotemporal patterns of ozone concentrations in 24 typical cities across the Yangtze River Delta region. The results demonstrate that: (1) The GCN-GRU model reduces MAE and RMSE by 17.7% and 16.3%, respectively, compared to conventional statistical and machine learning models; (2) In case simulations of high ozone pollution events, the model effectively captures the periodic variations and critical turning points of observed ozone concentrations (r = 0.735, p < 0.05), and accurately reflects spatial distribution characteristics; (3) Simulated ozone concentrations from 2015 to 2023 show a fluctuating upward trend, with an average annual increase of +0.56 μg m^-³ a⁻¹ (p < 0.05), which is largely consistent with the observed trends. Both simulated and observed data reveal increasing trends in inland areas and decreasing trends in coastal regions. This research provides a robust scientific basis for precise ozone concentrations prediction and pollution control policy formulation in the Yangtze River Delta.

The Horqin Sandy Land is one of the four major sandy lands in China, and is located in the western part of the Northeast China Plain and the southeastern part of Inner Mongolia. The research on the spatio-temporal variation patterns and driving factors of land desertification is an important foundation for scientificly carrying out the sand prevention and control work. In this study, soil desertification indicators were retrieved from Landsat remote sensing imagery, and the soil wind erosion intensity was modeled using data including meteorological conditions, NDVI, DEM and soil texture. The spatio-temporal distribution pattern of soil desertification and its response to soil wind erosion in the Horqin Sandy Land from 1991 to 2020 were revealed by using methods such as slope trend analysis and Pearson correlation analysis. The results showed that: (1) From 1991 to 2020, the land desertification in the study area exhibited a reversal trend, and the area of the reversal region accounted for 71.27% of the total, whereas the area experiencing intensified desertification accounted for 17.54% and was sporadically distributed. (2) From 1991 to 2020, the changes in land desertification were basically consistent with the changes in soil wind erosion intensity, showing a trend of first increasing and then decreasing. (3) In the regions where wind erosion played a dominant role in the desertification, the area affected by human activities accounted for 66.36%. In the regions where wind erosion weakened and desertification reversed, the positive impact of human activities was significant, accounting for 78.34%; however, in the regions where wind erosion increased and desertification intensified, the contribution of climate change (57.69%) was greater than that of human activities (42.31%), and these two regions were adjacent in distribution, mainly located in Horqin Right Wing Middle Banner in the north, the western part of Tongyu county and Horqin district. This research investigated the spatio-temporal evolution of desertification in the study area over a 30-year period and examined its response to wind erosion intensity. These results provided theoretical basis and scientific guidance for the prevention and control of desertification, and contributed to the continuous restoration and improvement of the ecological environment in the Horqin Sandy Land.

The Baihetan Reservoir area is the region with the highest sediment yield and transportation intensity in the lower Jinsha River. Multiple tributaries flowing into the reservoir carry high sediment concentrations, and the sediment deposition in the estuary and main stream reservoir areas caused by the inflow of sediment from tributaries has significant impacts on the operation of the reservoir and local aquatic ecosystems. This study focuses on the Heishui River, currently the tributary with the highest sediment concentration in the Jinsha River Basin. Based on complete time-series observations of rainfall, runoff, and sediment across the watershed, the characteristics and main influencing factors of water and sediment transport and their changes in the basin are analyzed in detail. A high-precision distributed erosion and sediment yield model is established, and the simulation of the process of runoff and sediment yield are carried out. The results show that the trend of increasing rainfall runoff in the Heishui River Basin (HRB) is not significant. Due to soil and water conservation projects, the average annual sediment transport during 2001-2020 decreased by approximately 2.14 million tons compared to 1981-2000, representing a 34.5% reduction; Rainfall is the primary driver of sediment yield in the basin. Rainfall amount, spatial distribution, intensity, and duration all affect sediment yield and transport. Heavy rainfall in the middle-lower reaches (Puge to Ningnan section) is more likely to cause high sediment transport at the Ningnan hydrological station. Calculations show that during 2021-2023, the HRB experienced relatively low runoff, with average annual runoff of 1.68 billion m³ and average annual sediment transport of 1.64 million tons. The reduced runoff caused a 65.3% decrease in sediment transport compared to the multi-year average.

Gridded meteorological data obtained by spatial interpolation play a crucial role in ecological, hydrological, and land surface process models. The key challenge in current spatial interpolation is to effectively match the physical behavior of meteorological variables with the mechanistic properties of interpolation methods to maximize accuracy. This study constructed 12 interpolation models using thin plate spline (TPS) and random forest (RF) with six combinations of covariates including elevation, slope, aspect, and reanalysis data. These models were applied to daily interpolation and comparative analysis of eight meteorological variables (maximum temperature, minimum temperature, precipitation, skin temperature, wind speed, relative humidity, surface pressure, and sunshine duration) across China from 2000 to 2020. The results indicated that: (1) TPS with elevation as a covariate was optimal for maximum temperature, minimum temperature, skin temperature, relative humidity, and sunshine duration; RF with elevation as a covariate performed best for precipitation and wind speed; RF with elevation, slope, and aspect as covariates was the best for surface pressure. (2) Minimum temperature and skin temperature were better interpolated using TPS with elevation, slope, and aspect in winter, while surface pressure was better interpolated using TPS with elevation in spring. The optimal interpolation methods for the other variables remained consistent across seasons. (3) The interpolation accuracy of maximum temperature showed an interval-dependent preference, with RF performing better for daily maximum temperature below 15 °C, whereas TPS was more accurate above this threshold.

Understanding the spatiotemporal dynamics and climatic consequences of long-term urbanization in arid and semi-arid regions is essential for sustainable development and improving human living environments. This study integrates historical documents and satellite imagery to reconstruct the urban expansion of the Hohhot-Baotou-Ordos (HBO) region from the 1900s to 2024. It also quantitatively evaluates the relative contribution of urbanization to regional warming since 1951, using meteorological observations and reanalysis data. Results show that the urban area in the HBO region expanded by approximately 93.73 times from the 1900s to 2024, with accelerated growth and prominent expansion in Hohhot. Concurrently, urban morphology showed reduced compactness and increased fractal dimension, indicating more complex urban boundaries alongside substantial regional socioeconomic growth. Between 1951 and 2023, urbanization predominantly influenced the regional climate by increasing minimum temperatures (urbanization-induced trend [OMR]: +0.44 ℃/decade; contribution rate: 63.17%) and narrowing the diurnal temperature range (observed trend: -0.36 ℃/decade; OMR: -0.41 ℃/decade). In contrast, the impact on maximum temperatures was minimal (OMR: +0.03 ℃/decade), with urban maximum temperatures consistently lower than those in adjacent suburban areas. This study highlights the climatic consequences arising from century-long urbanization in a representative dryland urban agglomeration, which provides a robust foundation for future investigations into urban-climate interactions in arid and semi-arid regions.

Based on continuous monitoring of precipitation amount and meteorological factors and periodic collection of plant samples in the Poyang Lake plain from June 2017 to September 2020, this study investigated the spatiotemporal characteristics and formation mechanisms of Lake-effect precipitation events using a three-component isotopic mixing model and ERA5 reanalysis data. The contribution of Poyang Lake to regional precipitation was quantitatively evaluated. Results showed that: (1) The region exhibited a humid climate with weak below-cloud evaporation. Stable water isotopic compositions in precipitation were enriched in spring and winter but depleted in summer and autumn. The total amount of precipitation and frequency of extreme precipitation events decreased from northwest to southeast. (2) A total of 148 lake-effect precipitation events were observed, characterized by the higher frequency in winter than in summer. These events were classified into two spatial patterns of multi-core rainband and single-core vortex. (3) Lake-atmosphere interactions showed seasonal variations that intensive lake evaporation increased boundary-layer humidity saturation and thus intensified convective activity during spring and summer. By contrast, the elevated thermal gradients enhanced atmospheric instability and thus jointly promoted precipitation during autumn and winter. (4) The evaporative contribution rate of Poyang Lake to lakeside precipitation ranged from 1.65% to 37.93%. In winter, the coupling effect of cold-dry air masses and topographic uplift resulted in the peak contribution rate of 37.93% observed at Guling town. However, the evaporative contribution rate in the plain area remained below 15% characterized by low values in the west and high values in the east during summer, which was associated with the dilution effect of deep lake water levels and marine moisture. These findings provide a theoretical basis for optimizing precipitation forecasting models and flood disaster warnings in the Poyang lake plain region.

Runoff is a critical component of the water cycle. Traditional methods for measuring runoff are heavily influenced by the placement of monitoring stations, while physical runoff models require long-term, high-quality data series encompassing multiple elements. The use of remote images to invert the flow can get rid of the influence of natural, social and other factors, providing a new means for runoff simulation. At present, the relationship between remote sensing indicators and discharge is yet to be further explored. This study constructs seven optical remote sensing indicators using Sentinel-2 images (2017-2023), focusing on the river reach near the Qilijie hydrological station in the Jianxi River Basin. Through comparative analysis of runoff estimation results from three methodologies, the Calibration/Measurement (C/M) method, multiple linear regression, and random forest, across 180 pixel sets, the following findings were revealed: The accuracy of the C/M method is highly sensitive to the selection of M pixels, highlighting the necessity of choosing M pixels responsive to various flow magnitudes for reliable estimations. Key indicators, namely the CM signal, Normalized Difference Water Index (NDWI), Normalized Difference Vegetation Index (NDVI), Bare Soil Index (BSI), and Turbid Water Index (TWI), effectively capture key river characteristics such as water extent, vegetation cover, soil conditions, and water turbidity. Compared with the traditional C/M method, the multiple linear regression and random forest models that incorporate multi-source remote sensing indicators effectively mitigate the influence of M pixel scale and spatial distribution on runoff estimation. Among the three methods, the random forest algorithm demonstrates superior accuracy in runoff estimation, particularly for high-flow conditions. The findings of this study provide a scientific reference for river runoff simulation and offer valuable insights for the intelligent monitoring of river systems, helping to improve the efficiency and accuracy of water resource management.