Search

Drylands account for approximately 27 % of global soil organic carbon (SOC) reserves and play a vital role in regulating the global terrestrial ecosystem carbon cycle. Patchy patterns are common in drylands due to severe water shortages, making it difficult to predict sediment and SOC losses accurately. Therefore, a detailed field experiment was conducted to investigate the effects of revegetation on sediment and SOC losses by runoff on typical dryland hillslopes on the Loess Plateau. Three machine learning models, namely, extreme gradient boosting (XGBoost), support vector machine (SVM), and random forest (RF) models, were used to simulate sediment and SOC losses, and the accuracies of these models were evaluated. The results revealed that the minimum value of SOC loss was 1.84 g kg−1 in the grass-covered plot, which was significantly lower than the loss of 17.73 g kg−1 observed in the bare plot. Compared with those in the bare plots, the sediment yields in the revegetation plots with shrubs, shrub–grass, trees, and grass were reduced by 46.9 %, 59.3 %, 70.7 %, and 94.4 %, respectively. The sediment and SOC losses were affected primarily by hydraulic characteristics (flow velocity and unit stream power) and soil properties (percentages of sediment fractions and median particle size). In addition, when the performances of the machine learning models were compared, the XGBoost and RF models yielded the best predictions of the SOC concentration, sediment loss, and SOC loss (NSE > 0.80, RRMSE < 0.14). In general, revegetation can significantly reduce SOC loss on hillslopes. Machine learning-based models provide a new method for sediment and SOC loss prediction in situations where severe erosion is occurring on dryland hillslopes. • Re-vegetation significantly reduces sediment and SOC loss on dryland hillslopes. • The SOC loss was primarily affected by the hydraulic and soil characteristics. • XGBoost and RF models outperform traditional regression in predicting SOC loss. [ABSTRACT FROM AUTHOR]

Purpose: Soil erosion drastically perturbs the global carbon cycle, and splash erosion is the first critical stage of soil erosion by water. Thus, studying the soil organic carbon (SOC) dynamic caused by splash erosion is essential. Methods: Simulated rainfall splash experiments were conducted on three cultivated soils with different SOC contents (black soil, purple soil, and red soil) to investigate the dynamics of SOC related to splash erosion, and the splash distance was 0–5, 5–10, 10–17, 17–28, 28–40 cm, respectively. Results: SOC was unevenly distributed and enriched in different-sized aggregates. The enrichment ratio of SOC (EROC) was basically greater than 1 and increased as the splash distance increased. Moreover, the variation of EROC differed from that in the aggregate enrichment ratio (ERA), and the EROC in large macroaggregates (> 0.5 mm) decreased with increasing ERA. The weight proportion of SOC mobilization with different-sized aggregates was consistent with the mass percentage of aggregates. Specifically, there was a significant positive exponential relationship between SOC mobilization and the aggregate mass of the greatest proportion, indicating that the mobilization of SOC was mainly affected by the aggregate mass of the greatest proportion. Conclusions: Our findings indicated that SOC migration can be realized by aggregates transport under splash erosion, indicating that improving soil structure will help to reduce SOC migration and even loss. [ABSTRACT FROM AUTHOR]

Highlights • Applying the electrolyte tracer pulse method to measure shallow water flow velocity on hill-slope. • An equation for estimating flow velocity on the basis of the measurement distance was established. • The new equation could be used to simulate results from other experimental data sets. • The steady flow velocities and the distances when water flow reached stability were determined by the new equation. Abstract Understanding the velocity distribution of shallow water flow along hill-slope is of great significance in soil erosion studies. This study proposes a method to estimate the steady flow velocity and the distance when water flow reaches stability, using hill-slope flow velocity distribution data along a simulated rill measured by the electrolyte tracer pulse method. Laboratory experiments were performed using a flume of 12 m long, 0.1 m wide and 0.3 m height under five slope gradients (5°, 10°, 15°, 20° and 25°) and four flow discharges (2, 4, 8 and 16 L min−1). The electrolyte tracer pulse method was employed to measure the flow velocities at locations of 1, 2, 3, 4, 6, 8, 10 and 12 m from the inlet of the flowing water. The flow velocities measured by the dye tracer method served as a control. The results showed that the flow velocities measured by the electrolyte tracer pulse method initially accelerated and then reached a steady value. An equation for estimating flow velocity on the basis of the measurement distance was established. The flow velocities calculated by the equation agreed well with those measured by the electrolyte tracer pulse method. The determination (R2) and the Nash-Sutcliffe model efficiency (NSE) were greater than 0.800, except for a few cases. In addition, the equation established in this study was shown to successfully predict flow velocities measured in other studies. The steady flow velocities and the distances when water flow reached stability under different slope gradients and flow discharges were determined by this equation. When the differences between the measured flow velocity and the steady flow velocity were very small, that is, 5% and 10%, the steady distances ranged from 2.239 m to 4.772 m and from 2.843 m to 6.059 m, respectively. By comparing the steady flow velocities with the flow velocities measured by the dye tracer method, the results indicated that the steady flow velocities were 0.702–0.735 times that of the flow velocities measured by the dye tracer method at various slope gradients and flow discharges. A linear function existed between the two, and the regressed parameter k (i.e., 0.718) could be used as the correction factor between the two. In general, the equation established in this study can facilitate the prediction of the steady flow velocity. [ABSTRACT FROM AUTHOR]

Accurate estimation of sediment concentration is essential to establish physically-based erosion models. The objectives of this study were to evaluate the effects of flow discharge (Q), slope gradient (S), flow velocity (V), shear stress (τ), stream power (ω) and unit stream power (U) on sediment concentration. Laboratory experiments were conducted using a 10 × 0.1 m rill flume under four flow discharges (2, 4, 8 and 16 L min −1 ), and five slope gradients (5°, 10°, 15°, 20° and 25°). The results showed that the measured sediment concentration varied from 87.08 to 620.80 kg m −3 with a mean value of 343.13 kg m −3 . Sediment concentration increased as a power function with flow discharge and slope gradient, with R 2  = 0.975 and NSE = 0.945. The sediment concentration was more sensitive to slope gradient than to flow discharge. The sediment concentration was well predicted by unit stream power (R 2  = 0.937, NSE = 0.865), whereas less satisfactorily by flow velocity (R 2  = 0.470, NSE = 0.539) and stream power (R 2  = 0.773, NSE = 0.732). In addition, using the equations to simulate the measured sediment concentration of other studies, the result further indicated that slope gradient, flow discharge and unit stream power were good predictors of sediment concentration. In general, slope gradient, flow discharge and unit stream power seem to be the preferred predictors for estimating sediment concentration. [ABSTRACT FROM AUTHOR]

• The soil erosion performances of frozen, nonfrozen and thawed soil were evaluated. • Four models were implemented to simulate the key parameters for thawed soil erosion. • The performance of SR and machine learning models were compared. • Influencing factors of straw returning on thawed soil were analyzed. Snow and ice‐melted water flow over thawed soil cause severe soil erosion. The application of straw mulch has long been considered as an effective measure to reduce soil erosion, but the effects of straw mulch and straw incorporation on thawed soil erosion remain uncertain. This study conducted detailed laboratory experiments to investigate the effects of straw mulch and straw incorporation on soil erosion and sediment transport of thawed soil. The median particle size (D 50) of transported sediment, flow velocity and sediment concentration for thawed soil were compared with that of unfrozen and frozen soil. Models based on Random Forest (RF), Support Vector Machine (SVM), Back Propagation Neural Network (BP) and stepwise regression (SR) were implemented to simulate the key parameters of thawed soil erosion, and the accuracy of the developed models was evaluated. The results showed that the sediment yield increased significantly in thawed soil compared with that in unfrozen soil. The maximum flow velocity over thawed soil decreased by 20.7 % compared with that over frozen soil. The D 50 of transported sediment decreased by 29.4 % after the soil suffered the effects of freeze–thaw, and the finer sizes were more likely to be entrained. The straw mulch effectively reduced the D 50 of transported sediment, flow velocity and sediment concentration. In addition, the RF model showed a better result to predict the sediment concentration, flow velocity and D 50 of transported sediment (NSE > 0.849, RRMSE < 0.493) than the BP, SVM and SR models. Comparing the performance of the RF model with the models reported in previous studies, the RF model showed the best performance. In general, the results of this study provide important information for soil erosion in thawed soil. [ABSTRACT FROM AUTHOR]

• Splash erosion affected the redistribution of soil particles and SIC. • The redistribution of splashed material and SIC varied with soil type. • SIC had sorting and enrichment characteristics. Splash erosion caused by raindrops is the first stage of water erosion that detaches and transports soil particles. As an important component of soil particles, soil inorganic carbon (SIC) contains little information on its redistribution during splash erosion. In this study, two soils, purple soil and loess, were selected and placed in a soil tray with a diameter of 20 cm at the center of a splash tray, and a rainfall simulator was used to simulate three raindrop energies in experiments to investigate the migration of soil particles and SIC. The results showed differences in the amount and selective migration of splashed material between the soils, which was evident in the obvious sorting of the particle size of purple soil with distance and the insignificant change of loess soil. The SIC sorting characteristic performed differently in the two soils, that is, the SIC content of purple soil increased with particle size, while that of loess was the opposite. The enrichment and depletion of SIC during splash erosion varied with soil type, and differed from the general enrichment of soil organic carbon (SOC). The SIC enrichment ratio (ER IC) of purple soil was generally > 1, whereas the ER IC values of loess were < 1. In addition, multiple linear regression (MLR) models were established to predict the SIC content of splashed material. Rainfall splash plays an important role in SIC migration and may further affect the carbon cycle. These preliminary results provide a basis for further study of SIC response to rainfall splash. [ABSTRACT FROM AUTHOR]