Composted sewage sludge (CSS) can rehabilitate degraded land by enhancing soil fertility and organic carbon (SOC). However, the temperature sensitivity (Q10) of SOC mineralization in sludge-amended soils remains poorly understood. Degraded grassland soil samples amended with 0-20% composted sewage sludge (CK to M20) at 25-35 degrees C for 35-days. The 'equal time', 'equal-C', 'one C pool model' and 'two C pool model' approaches evaluated the Q10 of SOC mineralization. Chemical properties and enzyme activities were analyzed to identify key factors influencing Q10 values of SOC mineralization. SOC, total nitrogen and phosphorus increased with sludge dose, while the C:N ratio decreased. Soil enzyme activities increased then decreased with increasing sludge rates, peaking at M5 to M15 treatment. Cumulative mineralized carbon rose with both sludge dose and temperature. Different models yielded contrasting Q10 patterns. The 'equal time' and 'equal-C' methods showed decreased Q10 with sludge addition, whereas the 'one C pool' and 'two C pool' models showed an initial increase followed by a decrease. Random forest analysis indicated that SOC, TN and TP contents were primary contributors to Q10 variations. These findings indicate that CSS alters the Q10 of SOC mineralization, with nutrient status being a more significant factor than enzyme activity.

Study regionThe Wangmaogou watershed (5.97 km2), a representative catchment in the hilly-gully region of China's Loess Plateau, middle Yellow River Basin—an area characterized by severe historical erosion and decades of intensive soil and water conservation.Study focusThis study quantifies the evolving sediment reduction benefits of check dams under progressive land use change (LUC) using a decoupling-based quantification framework. The WaTEM/SEDEM model and multi-scenario simulations were employed to isolate check dam benefits conditional on the LUC-determined sediment supply regime.New hydrological insights for the regionBetween 1970 and 2017, LUC reduced total watershed erosion by >60%, fundamentally reconfiguring the sediment supply regime. Under this evolving background, decoupling analysis reveals a supply‑driven decline in check dam benefits—from 90.9% under the 1970 high‑supply regime to 26% after 2012—while the sediment delivery ratio remained relatively stable, confirming that this decline reflects a regime shift rather than engineering failure. A double‑logarithmic nested model was developed to characterize the response of check dam benefits to declining erosion modulus. Extreme event analysis (2017 "7·26" rainstorm) further demonstrates that check dams retain irreplaceable "safety net" value, intercepting 1.78 times mean annual retention under transient high‑supply conditions. These findings indicate that on a greening Loess Plateau, check dams have evolved from primary sediment control structures to regime‑dependent safety nets, supporting their strategic functional transformation toward multi‑purpose water resource hubs.

The longitudinal dispersion coefficient Dx is a key parameter in the advection-dispersion equation and directly influences the accuracy of pollutant-transport simulations in rivers. Empirical formulas currently used to predict Dx often yield widely varying results and lack general applicability, particularly under complex hydraulic conditions. In vegetated channels, aquatic vegetation produces pronounced lateral shear and heterogeneous velocity fields, which makes the prediction of Dx a challenging task. A vegetation-constrained physics-informed neural network (V-PINN) is developed to predict Dx by embedding a zonal vegetation model into the physical constraints of the learning process. Training on datasets covering multiple vegetation types and flow regimes enables V-PINN to capture vegetation-induced mixing processes and to predict Dx with Relative Error (RE) below 0.233 under sparse or missing data. The proposed V-PINN exhibits superior performance over both the traditional Nzone formula and the standard PINN in estimating Dx. The standard PINN achieves reliable prediction of Dx and prediction of pollutant transport in natural rivers. The predicted Dx is integrated into a two-dimensional PINN in natural rivers to simulate pollutant transport, with RE below 0.064, demonstrating improved alignment with measured data and enhanced applicability in real rivers. The standard PINN also achieves reliable prediction of Dx and pollutant transport in natural rivers. The proposed V-PINN framework provides physically consistent predictions across diverse vegetation conditions and demonstrates the potential of physics-informed deep learning as an accurate and interpretable extension of classical hydraulic theory for modeling vegetation-flow interactions.

Forest stand composition can regulate soil nutrient cycling by altering aggregate formation, nutrient partitioning and microbial extracellular enzyme activity. Here, we examined pure Pinus tabuliformis forest (YS), pure Quercus acutissima forest (ML) and mixed coniferous-broadleaved forest (HJ) in the Ziwuling forest region of the Chinese Loess Plateau. Soil aggregate composition, soil organic carbon (SOC), total nitrogen (TN), total phosphorus (TP) and extracellular enzyme activities were quantified across different aggregate-size fractions in the 0-100 cm soil profile. Except for the >5 mm aggregate fraction, the proportions of all other aggregate-size classes followed the order ML > HJ > YS, with ML and HJ showing increases of 19.94%-66.98% and 8.76%-35.01%, respectively, relative to YS. Mixed forest significantly promoted SOC content, with SOC contents 46.9% and 76.1% higher than those in YS and ML, respectively. In contrast, TN content was highest in YS and was 20.3% higher than that in HJ, whereas TP showed only small differences among forest types. SOC and TN were mainly enriched in smaller aggregate fractions, accounting for 49.7%-79.1% and 44.7%-81.3% of their total stocks, respectively, while TP was preferentially associated with larger aggregates, accounting for 54.9%-82.1%. Compared with YS, HJ increased EG, LAP, NAG and ACP activities by 42.2%, 14.9%, 18.0% and 42.5%, respectively. Compared with ML, HJ also showed generally higher extracellular enzyme activities, indicating that mixed forest favored the enhancement of most enzyme-mediated nutrient acquisition processes. Overall, forest stand type regulated extracellular enzyme activity by reshaping soil aggregate composition and aggregate-associated nutrient distribution. These findings help improve our understanding of aggregate-associated nutrient cycling processes in restored forest soils on the Loess Plateau and may provide a reference for future comparative studies on restoration effects among different forest types.

Erosion and deposition critically influence soil organic carbon mobilization and stabilization, yet how microbial necromass contributes to soil organic carbon under these processes remains unclear. This study investigated the contribution of microbial necromass to soil organic carbon accumulation and stability across land-use types in a dam-controlled watershed on the Chinese Loess Plateau. Soil samples (0–20 cm) were collected from erosional (sloping farmland, woodland, grassland, and shrubland) and depositional (dammed land) zones. Soil properties and soil organic carbon mineralization were analyzed, and the contributions of bacterial and fungal necromass carbon to soil organic carbon were quantified. Key drivers were identified via redundancy analysis. Soil organic carbon content and stability—characterized by a higher proportion of stabilized carbon pools and longer mean residence time—were higher in erosional soils. The proportion of microbial necromass carbon in soil organic carbon ranged from 20.97

Hillslope management practices significantly alter the erosion process and its accompanying spatiotemporal dynamics of soil carbon. However, the hydrodynamic mechanisms regulating multiphase carbon loss under different management practices remain unclear. This study investigates how hydrodynamic processes drive the transport of dissolved organic carbon (DOC), dissolved inorganic carbon (DIC), soil organic carbon (SOC), and soil inorganic carbon (SIC) during erosion events. Field rainfall simulation experiments were conducted to assess four typical slope management treatments on the Loess Plateau: terrace (TP), grassland (GP), root system (RP), and bare land (CK, control). Key hydraulic parameters (flow velocity, Reynolds number, Froude number, resistance coefficient, shear stress, and stream power) were measured or derived to characterize the hydraulic conditions influencing carbon loss. Results showed that compared with CK, the treatments reduced DOC flux by 38.4-68.7 %, DIC flux by 1.2-47.9 %, SOC flux by 85.8-98.5 %, and SIC flux by 89.3-98.3 %. CK and RP treatments primarily lost carbon through sediment migration (89.7 % and 57.3 %, respectively), while TP and GP treatments primarily lost carbon through runoff migration (50.1 % and 76.2 %, respectively). Moreover, eroded carbon was predominantly lost in the form of inorganic carbon. Hydraulic parameters indirectly influenced carbon loss by modifying runoff rate and sediment discharge rates. Under bare slope conditions, both runoff-and sediment-associated carbon fluxes were primarily controlled by sediment transport dynamics. Following surface interventions, runoff-associated carbon flux became increasingly governed by runoff redistribution, while sediment-associated carbon flux remained closely coupled with sediment discharge. These findings offer important insights into the multiphase carbon transport mechanisms associated with soil erosion under different slope management practices.

The destructive impact of soil wind erosion and water erosion on surface erosion is particularly prominent. The sandy coarse sand area in the middle reaches of the Yellow River, as a typical global wind-water composite erosion zone, directly threatens watershed soil and water security and the stability of the river's course and flood control in the downstream Yellow River. This study addresses the issue of erosion mechanism differentiation caused by differences in soil surface properties in this region. By focusing on erosion energy as the central link and combining wind tunnel experiments and hydrometeorological data, an energy-based wind-water composite watershed erosion and sediment transport model is constructed. The results show that the wind erosion energy-based sediment transport prediction models constructed for the sand-covered bedrock, sand-covered loess and loess erosion areas exhibit high fitting degree and reliability, with R-2 values reaching 0.944, 0.979 and 0.978, and RSR values being 0.305, 0.188 and 0.194, respectively. Wind erosion and water erosion energy show significant spatiotemporal differentiation, with wind erosion concentrated in the spring and water erosion concentrated in the summer. The average annual energy values are sand-covered loess area > sand-covered bedrock area > loess area. The compound erosion sediment yield model shows good adaptability in all three types of soil areas, with R-2 values no less than 0.75 and RSR values all below 0.5 during both the calibration and validation periods. Among them, the model achieves the highest accuracy in the sand-covered bedrock area, with a high R-2 of 0.989 and a low RSR of 0.104. The study confirms that the energy-based model can accurately simulate composite erosion sediment yield and provide scientific support for precise soil and water conservation management in the region.

Flood event sediment load plays a critical role in evaluating the effectiveness of soil and water conservation measures on the Loess Plateau and advancing understanding of water and sediment dynamics in the middle reaches of the Yellow River Basin. Thus, 524 flood events from 1960 to 2020 were analysed to identify the changes in flood event sediment load and their driving mechanisms in the Dali River Basin on the Loess Plateau, China. Results indicate that flood events consistently dominate sediment load, accounting for more than 90% of annual sediment load in most years. Two abrupt change points in the annual flood sediment load are identified in 1971 and 2002, accompanied by substantial declines in both annual maximum peak sediment concentration and flood sediment load across three periods. Runoff-sediment relationships become increasingly clustered with reduced slopes, indicating weakened sediment load sensitivity to runoff. Check dams, terraces, and vegetation exhibit significant negative relationships with flood sediment load, whereas five extreme precipitation indices are positively correlated (p < 0.05). These patterns suggest that although extreme precipitation remains a key hydrological driver of sediment mobilisation, its sediment-producing effect is increasingly constrained by soil and water conservation measures. Overall, coordinated implementation of check dams, terraces, and vegetation reduces sediment availability and connectivity within the basin, making sediment load during extreme flood events unlikely to exceed historical upper limits under current conservation conditions. The findings establish a scientific basis for sediment management and the optimization of soil and water conservation strategies.

As a key soil and water conservation measure, check dams play an important role in erosion control, flood mitigation, and ecological restoration. Their scientific siting is the core prerequisite for realizing these composite benefits. To address the limitations of traditional siting methods, which are characterized by strong subjectivity, low efficiency, and incomplete multi-objective collaborative optimization, a multi-objective optimization framework integrating GIS, hydrological-hydrodynamic coupled models, and a genetic algorithm is proposed in this study. By parameterizing the distance from the watershed outlet (S) and the dam height (H), a continuous decision space is constructed, establishing quantitative mapping relationships with the dam crest length, sediment storage capacity, and silted land area. By coupling the HEC-HMS and HEC-RAS models, a flood surrogate model is developed to dynamically predict the peak flood reduction benefits under a 100-year flood scenario. Based on the Nondominated Sorting Genetic Algorithm II, a multi-objective optimization model incorporating the construction cost (C), peak flow attenuation (P), and sediment retention and farmland creation (SRFC) benefits (E) is constructed, revealing the nonlinear regulatory mechanisms of the decision variables on the objectives. A case study demonstrates that the optimal solution set significantly clusters in the middle and lower reaches of the Yangjiagou watershed (S < 4.2 km). High dams located near the outlet (S < 2.2 km and H > 20 m) correspond to schemes with strong flood-control performance (P > 60 %). Schemes with advantageous benefit-cost ratios (E/C > 2.5) are distributed in the middle reaches (S > 3.2 km) and are characterized by low-dam systems (H < 17 m). This framework overcomes the spatial discretization and empirical dependence limitations of traditional check dam siting methods. It achieves prediction errors below 20 %, providing a scientific tool that combines mechanistic interpretability and decision-making efficiency for check dam planning. The proposed framework further demonstrates engineering feasibility and transferability to other watersheds, offering practical value for soil and water conservation planning.

Lateral carbon transport induced by soil erosion is a key component of the global carbon cycle. Over the past 50 years, ecological restoration projects in the Loess Plateau of China have increased vegetation coverage. However, the effects of land use, precipitation variability, and landscape factors on lateral carbon transport remain unclear. This study integrates field sampling data with the distributed erosion model to quantify lateral carbon transport in typical watersheds of the Loess Plateau under erosion-induced conditions, using geographic detectors to identify key influencing factors. The results show that land use changes, particularly the conversion from bare land to forest and grassland, significantly reduced both soil erosion and lateral carbon transport by 74.0% and 74.5%, respectively, from 1970 to 2017. Extreme rainfall events exacerbated both soil erosion and carbon transport, with lateral organic carbon transport increasing from 89 t under the 50% design frequency of annual rainfall erosivity to 649 t under the 1% design frequency. Vegetation restoration has mitigated the negative impacts of extreme rainfall events on both soil erosion and carbon transport. Under 1970 land use conditions, combined with average annual rainfall erosivity (50%), soil erosion and lateral carbon transport were significantly higher than under 2017 conditions with a 10-year recurrence interval rainfall erosivity (10%). Among landscape control factors, land use and terrain ruggedness exhibited the highest explanatory power for lateral carbon transport distribution. In conclusion, this study provides a novel framework for quantifying lateral carbon transport and offers insights for policymakers to develop carbon management strategies during watershed restoration.

The frequency and intensity of flood events driven by extreme rainfall have increased significantly due to climate change. However, the causes of urban flood disasters and the measures used to govern them are more complex, especially given limited experience with floods that exceed historical climate patterns. Using global flood loss data, this paper highlights the drivers of extreme flood disasters in cities and presents a series of mitigation strategies. The results indicate that the magnitude of damage caused by rainfall events and the severe flooding that follows has shown an alarming upward trend in recent years. The total economic losses caused by global flooding in 2023 reached as high as 87.7 billion USD, mainly due to prominent climate change issues, changes in the urban environment, and social factors. Based on this comprehensive analysis, this study focuses on adaptation strategies from a resilience management perspective, including an improved top-level design for urban water systems, enhanced emergency response capabilities, and institutionalized learning and reflection mechanisms. In conclusion, relatively comprehensive development strategies, such as regional multilateral agreements, may be integrated into flood management to protect inland cities from increasing flood-induced threats.

Under global climate change, sand and dust storms (SDS) in Northern China have exhibited new spatiotemporal evolution characteristics. To investigate SDS driving mechanisms, this study integrated multi-source remote sensing and reanalysis data (2005-2024), employing SHapley Additive exPlanations (SHAP), receiver operating characteristic curve analysis method, and the extreme gradient boosting machine learning model. The results indicated that SDS frequency increased significantly after 2015, with a 47.6% rise in annual events (from 4 to 6 events/year), while high-intensity events declined from 23.8% to 14.5%, revealing increasing frequency but decreasing intensity. SDS occurrence is primarily triggered by the synergy between strong wind hours and the concurrent VHI. The lag period of eco-hydrological factors such as vegetation, temperature, and moisture on SDS was approximate 5 months. SHAP analysis revealed VHI is dominant driver in hyper-arid and arid transition zones, while soil moisture and evapotranspiration became the primary controls in semi-arid regions. Critical thresholds with narrow 95% confidence intervals for dust occurrence were identified, including VHI below 0.35, strong wind hours exceeding 12.7 h, NDVI below 0.10, and evapotranspiration below 10.8 mm. When VHI exceeds 0.35, the relative SDS risk is reduced by 88.4%, underscoring its potent protective effect. The dynamic thresholds and high-precision prediction model achieved robust performance with an out-of-sample PR-AUC of 0.843, together with strong predictive accuracy as evidenced by a test AUC of 0.935 and annual-scale R2 greater than 0.7, providing a scientific basis for accurate dust forecasting and risk management.

Effective riverine nitrogen pollution control necessitates the accurate quantification of nitrate (NO3-) sources. In this study, water quality monitoring, ion analysis, and environmental isotope tracers (delta H-2-H2O, delta O-18-H2O, delta N-15-NO3-, delta O-18-NO3-) were coupled with the Bayesian mixing model (MixSIAR) to apportion riverine NO3- contributions from four potential sources in the Dan River Basin, which is the headwater region of China's South-to-North Water Diversion Project. The results revealed that nitrate concentrations fluctuated between 1.20 and 5.51 mg/L, with significant seasonal variation (P < 0.05). Isotopic values ranged from 0.70 parts per thousand to 13.53 parts per thousand for delta N-15-NO3- (mean = 6.70 parts per thousand) and from -5.02 to 3.25 parts per thousand for delta O-18-NO3- (mean = -0.40 parts per thousand). Isotopic data demonstrated that during the wet and dry seasons, NO3- was primarily derived from the conservative mixture of multiple sources, with limited influence from biological removal processes. MixSIAR analysis revealed manure and sewage (M&S) as the primary nitrate source, accounting for 67.6% of the nitrate in the dry season and 46.1% in the wet season. Meanwhile, chemical fertilizer (CF) accounted for 16.1% and 34.8% of the nitrate in the dry and wet season, respectively. Atmospheric precipitation (AP) contributed the least (0.8% in dry season, 1.1% in wet season). Uncertainty and sensitivity analyses revealed relatively high uncertainties for M&S, CF, and soil organic nitrogen (SON) (uncertainty index (UI90) = 0.299-0.534), with the delta N-15 isotopic signature of M&S exerting the strongest influence on source apportionment results. Overall, the study provides critical insights to help local authorities develop evidence-based pollution control and sustainable water resource strategies for the region.

Accurate prediction of soil organic carbon (SOC) content is essential for soil management, ecosystem sustainability, and climate change mitigation, particularly in ecologically fragile arid and semi-arid regions. Traditional statistical approaches often struggle to capture the complex and nonlinear relationships between SOC and environmental drivers. In this study, multiple machine learning models—including Random Forest (RF), Support Vector Regression (SVR), Partial Least Squares Regression (PLSR), Artificial Neural Network (ANN), Convolutional Neural Network (CNN), and Extreme Gradient Boosting (XGBoost)—were evaluated for SOC prediction using a large dataset of 8,621 soil samples collected from north-central and north-western China. A comprehensive set of environmental covariates was incorporated, including terrain attributes, climate variables, soil properties, vegetation indices, soil moisture, and erosion indicators. Model performance was assessed using five-fold spatial cross-validation to mitigate potential spatial dependence between training and validation samples. Prediction accuracy was evaluated using multiple metrics, including adjusted R2, MAE, MAPE, MSE, RMSE, and RPIQ. Among the evaluated models, XGBoost achieved the highest predictive performance across land-use types, with an adjusted R2 of 0.775 and an RMSE of 0.610 in the spatial validation set. Rather than emphasizing algorithmic superiority alone, this study further explored model interpretability by integrating Shapley Additive Explanations (SHAP) with generalized additive models (GAMs). This combined interpretive framework was used to characterize land-use–specific nonlinear response patterns and model-derived breakpoints of key environmental predictors. The results indicate that temperature, soil pH, vegetation activity (NDVI), and elevation consistently emerged as important predictors associated with SOC variation, although their relative importance and response patterns differed among forestland, grassland, farmland, and unutilized land. These findings highlight the context dependency of SOC–environment relationships in arid and semi-arid landscapes. Overall, this study provides a spatially robust modeling and interpretation framework for SOC prediction at sampling locations, offering transferable insights into how environmental gradients shape SOC variability across land-use types. The approach establishes a methodological basis for future large-scale SOC mapping and uncertainty assessment.

The vegetation ecosystem in the Yellow River Basin is highly sensitive to climate change.Analyzing the driving mechanism of climate on vegetation dynamics plays an important role in the ecological management of the Yellow River Basin.Based on the data of the normalized difference vegetation index,total primary productivity,leaf area index,enhanced vegetation index,chlorophyll fluorescence,and vegetation coverage from 2001 to 2020,combined with the IGBP vegetation classification system,the differences and trends of vegetation characteristics of different vegetation types were quantified.By constructing a multi-scale geographically weighted regression model and Geodetector,the spatial heterogeneity and interaction mechanism of climatic factors and soil moisture on vegetation coverage were revealed.The results showed that:① From 2001 to 2020,the vegetation in the central and eastern parts of the Yellow River Basin was significantly improved,whereas that in the northwest showed a trend of degradation,while the interannual variability of the middle reaches of the Loess Plateau was the largest.② The vegetation characteristic index value of forest type was significantly higher than that of other types,and the vegetation coverage was the most sensitive to the differentiation of vegetation types.③ In most areas of the Yellow River Basin,precipitation promoted vegetation growth,temperature inhibited vegetation growth in the north and middle,and saturated vapor pressure difference significantly promoted vegetation coverage in the southeast and northwest.In the humid areas of the south and southeast,the saturated vapor pressure difference showed a promoting effect,and the soil moisture showed an inhibitory effect.In the whole basin,precipitation and temperature,saturated vapor pressure difference,and soil moisture had a strong interaction,which promoted vegetation coverage.The research results provide theoretical support for the ecological zoning management of the Yellow River Basin.

The water-sediment interface acts as a core reactor for the riverine carbon cycle, releasing substantial CO2 into the atmosphere through the microbial mineralization of dissolved organic carbon (DOC). However, the role of riverbed sediment in regulating organic carbon mineralization and its quantitative mechanisms remain unclear, particularly in ecosystems undergoing large-scale vegetation restoration. This study investigated rivers within a revegetated small watershed on the Loess Plateau. Using laboratory dark incubation experiments at constant temperature, combined with analyses of microbial communities and functional genes, we elucidated the seasonal mechanisms governing the sediment's contribution to organic carbon mineralization. Results indicated that riverbed sediment served as a continuous carbon source for the overlying water; after a 42-day incubation, particulate organic carbon (POC) content was at 0.598-0.824 times its initial value. The sediment significantly enhanced total CO2 emission primarily by alleviating nitrogen limitation in the water column and altering DOC composition. The contribution of sediment to organic carbon mineralization potential (CS) peaked during the wet season (52.020 +/- 8.054%), coinciding with the lowest cumulative mineralization, and reached a minimum during the dry season (43.417 +/- 6.688%), when cumulative mineralization was highest. A high CS state was characterized by a low abundance of carbon degradation (e.g., xylA, sga) and carbon fixation (e.g., rbcL) genes in the initial water column, and was significantly positively correlated with sediment fungi capable of decomposing recalcitrant organic matter (e.g., Amniculicola, Castanediella). These findings provide a theoretical reference for understanding the drivers of microbial spatiotemporal patterns in riverine carbon cycling and offer a scientific basis for comprehensively assessing the ecological benefits of vegetation restoration on the Loess Plateau.

Rapid urbanization reshapes landscape structure and threatens ecosystem services, yet operational ecological benchmarks for environmental impact assessment remain limited. Focusing on China’s Guanzhong region, this study examines how landscape fragmentation influences the ecosystem service value (ESV) and identifies critical threshold points for planning and management, using land-use data from 1990 to 2020 combined with spatial and statistical analyses. The results show that urban expansion substantially increased landscape fragmentation, with patch density rising by 109%, and contagion declining by 83.7%. Although the total ESV showed partial recovery over the study period, the landscape pattern mediated approximately 62% of the total effect of land-use change on ESV, constituting the dominant transmission pathway. Three actionable thresholds were identified: patch density of 1.2 patches·km−2, the largest patch index of 18%, and contagion of 15%. Beyond these breakpoints, ESV loss accelerates sharply, with the most pronounced rate increase reaching 355.6%. These findings provide region-specific reference values to support environmental impact assessment, ecological zoning, and land-use regulation in rapidly urbanizing areas.

The sediment transport capacity of overland flow is a key parameter in soil erosion models, and accurately predicting this capacity is a critical challenge. Existing studies often overlook the effects of geomorphological changes and the interactions between runoff and sediment transport in actual soil erosion processes. This study conducted indoor experiments on movable beds under various inflow conditions and slopes. It evaluated the applicability of existing sediment transport capacity equations and established a new equation suitable for the Loess Plateau using dimensional analysis. The results show that among the selected classic sediment transport capacity equations, the Liu equation had the best performance (R-2 = 0.806, NSE = 0.781, RMSE = 0.018 kg & centerdot;m(-1) s(-1)), followed by the Luan equation (R-2 = 0.772, NSE = 0.693, RMSE = 0.021 kg m(-1) s(-1)), the Abrahams equation (R-2 = 0.715, NSE = 0.390, RMSE = 0.029 kg m(-1) s(-1)), the Govers equation (R-2 = 0.709, NSE = 0.320, RMSE = 0.031 kg m(-1) s(-1)), and the Yalin equation (R-2 = 0.618, NSE = -0.3, RMSE = 0.043 kg m(-1) s(-1)). Correlation analysis revealed that stream power and friction velocity significantly influence sediment transport capacity (p < 0.05). Based on these findings, a new equation for sediment transport capacity on steep loess slopes under movable bed conditions was developed and validated (R-2 = 0.925, NSE = 0.908, RMSE = 0.011 kg m(-1) s(-1)), significantly improving prediction accuracy compared to the five selected empirical equations. This study presents a new equation for sediment transport capacity of overland flow applicable to steep loess slopes, derived from existing equations for sediment transport capacity of overland flow, offering more reliable support for the development of soil erosion models in the Loess Plateau.

Wind erosion is a primary driver of soil degradation on the Loess Plateau. Soil physical crust plays a critical role in stabilizing aeolian sandy soil and preventing wind erosion. This study investigated the evolution of physical crust characteristics and their response to soil wind erosion under varying rainfall intensities (40, 80, 120 mm & centerdot;h(-1)) and wind speeds (9, 11, 13 m & centerdot;s(-1)) through indoor simulated rainfall and wind tunnel experiments. Research indicated that increased rainfall intensity significantly facilitated crust development, notably enhancing thickness, hardness, and shear strength. Compared to bare soil, physical crusts reduced wind erosion intensity and sediment transport rate by over 95%, and decreased saltation height by over 70%. As crust strength increased, soil wind erosion intensity, sediment transport rate, and saltation height progressively declined. Conversely, both aerodynamic roughness and friction velocity exhibited nonlinear responses to increasing crust strength, highlighting complex surface-airflow interactions. Redundancy Analysis (RDA) and Structural Equation Modeling (SEM) revealed that rainfall intensity indirectly mitigated wind erosion by regulating crust mechanical properties; specifically, crust hardness and shear strength were identified as the primary controlling factors. Furthermore, a multivariate regression model incorporating crust hardness, shear strength, bulk density, >0.25 mm aggregate content, and erodible particles was established, demonstrating high predictive accuracy (R-2 = 0.85) for wind erosion intensity. These findings provide scientific insights into the anti-erosion mechanisms of physical crusts.