Groundwater table prediction can play an essential role in agricultural production, drainage engineering, and water-resources regulation in arid and semi-arid regions. Particularly, shallow groundwater dynamics can also respond rapidly to climate forcing and anthropogenic disturbances. In this study, a groundwater dynamics model was constructed to predict the regional level using spatial heterogeneity. 23 daily monitoring stations were selected in Linwei District, Weinan City, Shaanxi Province, China, a typical semi-arid region with complex non-linear interactions between meteorological forcing and human activities. A prediction framework was also developed for the regional groundwater level. Site-specific explanatory variables were integrated with Long Short-Term Memory (LSTM) networks and intelligent hyperparameter optimization. An independent modeling strategy was adopted for spatial heterogeneity in the study area. A customized prediction model was constructed for each individual observation well. The localized hydrogeological features were implicitly captured in response to external stressors, rather than the explicit spatial parameterization. A standard score outlier detection was applied to the daily water level series after data preprocessing. The 3 $\sigma$ criterion of standard deviations was utilized to identify and then remove anomalous data points. Missing values were subsequently filled using linear interpolation. Feature screening was conducted for the effectiveness of model inputs using Pearson correlation analysis. Eight candidate variables were evaluated, including meteorological and environmental indicators, such as reference crop evapotranspiration, daily maximum and minimum temperatures, precipitation, relative humidity, sunshine hours, wind speed, and soil heat flux. Input features were selected for the absolute correlation coefficient greater than 0.30 with the localized groundwater level. Site-adaptive input selection effectively reduced data redundancy and overfitting. Weakly correlated noise was removed after selection. A systematic investigation was implemented to compare manual hyperparameter tuning, the Salp Swarm Algorithm (SSA), and Bayesian Optimization (BO). The swarming behavior of Salp chains was then simulated to explore the parameter space using a leader-follower mechanism. In contrast, a Gaussian Process was employed as a surrogate model to estimate the distribution of the objective function. An Upper Confidence Bound acquisition function was utilized to balance exploration and exploitation. These optimizers were used to determine the optimal configuration of the model structures. The results demonstrated that intelligent optimization significantly enhanced the predictive accuracy and robustness of the baseline LSTM models. The benchmark LSTM models achieved average Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Nash-Sutcliffe Efficiency (NSE) of 0.15, 0.24, and 0.71, respectively. The SSA reduced the average MAE and RMSE to 0.14 and 0.22, respectively. The LSTM network with BO (LSTM-BO) exhibited the most superior performance, with the average MAE and RMSE decreasing to 0.13 and 0.20, respectively. Crucially, the average NSE reached 0.85 for the LSTM-BO models. All individual stations maintained an efficiency score above 0.61. A more stable and reliable parameter search was obtained in the BO, compared with the SSA, particularly in complex stations where the baseline models previously failed or significantly degraded. The framework was combined with site-specific feature selection with probabilistic hyperparameter optimization. The findings can also offer a powerful and practical tool for groundwater resources of agricultural drainage engineering in semi-arid environments.

The infiltration of muddy water in soil has a complex change of upper boundary process. Among them, the hydraulic conductivity of soil has been one of the most important indicators for the infiltration characteristics of muddy water in the soil. In this study, a series of infiltration tests (9 orthogonal test treatments and 8 groups for the model validation) were performed on the muddy water in the saturated soil column. The hydraulic conductivity of the formed dense layer was also determined during the infiltration of muddy water. The dynamic model was then constructed for the hydraulic conductivity of the sand under multiple factors （sand content of muddy water, clay content, and infiltration time） using multiple regression. A verification test was made on the dynamic models of the hydraulic conductivity of the saturated soil columns in the muddy sandy and silt loam soil. It was assumed that the muddy water was infiltrated into the saturated soil column. The results showed that there were highly significant effects of the muddy water sand content, clay content, the gradation of sediment particles, and infiltration time on the hydraulic conductivity of the sand （ P<0.01）. The first factor was the infiltration time, followed by the sand content, and the last was the clay content. The coefficient of determination （R²）was 0.853 in the dynamic model for the hydraulic conductivity of sand, and the root mean square error was 0.004 cm/min, indicating the better-represented relationship between various factors and hydraulic conductivity. Meanwhile, the values of root mean square error was less than 0.01 cm/min, and the mean absolute value of relative error was less than 7%, indicating the highly reliable dynamic model of hydraulic conductivity. The R² values of the dynamic model were 0.912 and 0.930 for the sandy and silt loam, respectively, and the values of root mean square error were 2×10⁻³ and 5×10⁻⁵ cm/min, respectively. The values of root mean square error were less than 0.01 cm/min, and the mean absolute value of relative error was less than 17% in the model validation, indicating the excellent agreement between the calculated and measured. The finding can greatly contribute to the theoretical basis of the infiltration process of muddy water in the soil, together with the barrier mechanism of the dense layer.

This study aims to investigate the effects of muddy water irrigation and bio-organic fertilizers on soil water infiltration, evaporation, and leaching in the cultivated layer of Xi'an Province, China. An indoor one-dimensional soil column infiltration device was also utilized in the experiment. Four levels of muddy water sediment concentration (0, 4%, 8%, 12%, quality fraction, sediment from the muddy infiltration water was collected from the bottom of the Jinghui Canal irrigation area in the Yellow River basin, China) and four levels of bio-organic fertilizers (0, 2 250, 4 500, 6 750 kg/hm², bio-organic fertilizer from Yuanlong Biotechnology Co., Ltd., Puyang, China.) were set in a total of 16 treatments, where each repeated three times. The experimental indicators were measured, such as the cumulative infiltration, wetting front transport distance, soil water evaporation, soil moisture content, electrical conductivity, organic matter content, and leaching solution loss. The models were about: 1) the relationship between the infiltration time, sediment concentration in muddy water, fertilization amount, and wetting front distance; 2) the relationship between infiltration time, sediment concentration in muddy water, fertilization amount and infiltration rate; 3) The cumulative evaporation and evaporation time were fitted using Black and Rose evaporation models. The results showed that the sediment concentration of muddy water and bio-organic fertilizer posed the significant effects on the wetting front distance, cumulative infiltration amount, and infiltration rate. The higher the muddy water sediment concentration and bio-organic fertilizer level were, the smaller the wetting front distance and cumulative infiltration amount within the same infiltration time were. The muddy water infiltration increased the infiltration time by 9.88%-70.37%, whereas, the cumulative infiltration was reduced by 13.17%-34.13%, compared with the clean water infiltration. The bio-organic fertilizer was applied to increased the infiltration time by 8.96%-51.09%, while the cumulative infiltration amount was reduced by 0.91%-20.69%, compared with no fertilization. The wetting front distance and infiltration time were conformed to the power function relationship, where the coefficient of determination (R²) was more than 0.9. The Kostiakov infiltration model was used to fit the soil water infiltration characteristics and parameters under muddy water irrigation and fertilization conditions (R²>0.9). The infiltration coefficient K gradually decreased, as the muddy water sediment concentration and fertilization level gradually increased, while the infiltration index ɑ enlarged gradually. Muddy water irrigation and bio-organic fertilizer were effectively reduced the soil water evaporation, where the muddy water infiltration was reduced cumulative evaporation by 10.44%-26.24%, compared with clean water infiltration. The bio-organic fertilizer was reduced the accumulated evaporation by 4.76%-15.83%, compared with no fertilization. The cumulative evaporation amount and evaporation time under muddy water irrigation and bio-organic fertilizers conditions were consistent with the Black and Rose evaporation models (R²>0.9), and the SRMSE value of the Black evaporation model was lower than that of the Rose model, indicating that the accuracy of the Black evaporation model was higher than that of the Rose evaporation model. After evaporation, the soil moisture content, electrical conductivity, and organic matter content increased with the increase of soil depth. The pH value in the leaching solution increased with the increase of leaching time, while the electrical conductivity and total dissolved solids decreased with the increase of in leaching time. The muddy water irrigation increased the conductivity, organic matter content in the soil and total dissolved solids in the leaching solution, while reducing soil moisture content, pH, and leaching solution volume, compared with the clean water infiltration. Furthermore, the bio-organic fertilizer was applied to reduce the leaching solution volume, and conductivity, and total dissolved solids in the leaching solution, whereas there was an increase in the organic matter content and conductivity in the soil, compared with no fertilization. Therefore, the better performance was achieved in the muddy water irrigation, bio-organic fertilizer and soil infiltration capacity, and water and fertilizer conservation capacity. The finding can provide the scientific reference for the efficient utilization of muddy water and rational application of bio-organic fertilizer in the arid and semi-arid areas of the Yellow River Basin.

Muddy water irrigation has been limited to a wide application during agricultural production in recent years. Current challenges also remain on the water shortage and soil salinization in the arid and semi-arid regions. In this study, a series of experiments was conducted on a one-dimensional soil column infiltration in the laboratory. A systematic investigation was also made to explore the effects of different salt contents (S0: 0.80 g/kg, S1: 4.81 g/kg, and S2: 9.45 g/kg) and biochar addition amounts (B0: 0, B1: 0.5%, B2: 1%, and B3: 2%) on the soil water transport, salt ion distribution and leaching under clear water (ρ0: 0) and muddy water (ρ1: 4%) infiltration. The results showed that the cumulative infiltration amount significantly decreased with the increase of salt content (S) and biochar (B). Compared with the clear water (ρ0), the turbid water (ρ1) was reduced by 18.05% to 31.57%. The medium (S1) and high salt (S2) treatments were reduced by 9.40% to 18.58% and 15.38% to 23.86%, respectively, compared with the low salt (S0) treatment. The addition of biochar (B1/B2/B3) was reduced by 0.75% to 21.16%, compared with no addition (B0). The Kostiakov model (coefficient of determination (R²)>0.88, root mean square error <0.09 cm/min) was more suitable to simulate the infiltration rate than the Philip model. Its parameter K decreased with the increase of S and B, while α was the opposite. The distance of the wetting front migration (F) and time (t) were followed by the power function. The turbid water prolonged the time to reach the bottom of the soil column increased by 10.50% to 27.93%. The S and B further delayed the migration. The fitting parameters C and D were negatively exponentially correlated with S and B (R² > 0.92). The soil profile analysis showed that the moisture content decreased with depth, while the electrical conductivity increased. The turbid water was reduced the average moisture content by 0.52% to 4.37%, whereas, the electrical conductivity increased by 1.78% to 16.74%; The moisture content was reduced by 9.09% to 11.59% in the high salt (S2), leading to a sharp increase in the electrical conductivity by 215.21% to 314.95%; While the biochar (B3) increased the moisture content by 2.61% to 5.53%, whereas, the electrical conductivity was reduced by 10.23% to 23.80%. Except for the soil potassium ions (K⁺), the contents of soil calcium ions (Ca²⁺), magnesium ions (Mg²⁺), sodium ions (Na⁺), sulfate ions (SO₄²⁻), chloride ions (Cl⁻), and bicarbonate ions (HCO₃⁻) all increased with the increase in the salt content. There was a more significant increase in the muddy water than that in the clear water treatment; While the biochar addition significantly reduced the contents of Na⁺ and Cl⁻, thus promoting the retention of K⁺ and Ca²⁺. The high salt content (S2) exacerbated the salt accumulation to reduce the volume of the leaching solution. The treatment with ρ1S2B3 showed the highest leaching solution conductivity, with an average conductivity of 88.32 mS/cm. The sediment concentration and salt content of turbid water inhibited the water transport due to the pore blockage and osmotic pressure. The biochar optimized the water salt balance to regulate the ion adsorption and pore structure. The finding can also provide a theoretical basis for the synergistic improvement of the turbid water irrigation and biochar in saline alkali land.

Muddy water irrigation, an effective water-saving irrigation method, has been widely used in the Yellow River Basin in China. To investigate the effect of sand content on water infiltration and dense layer formation under one-dimensional vertical infiltration of muddy water, muddy water infiltration experiments were performed in the laboratory, and five sand contents of muddy water (S=0%, 3%, 6%, 9%, and 12%) were used. Models were established to describe the relationship between the cumulative infiltration amount [I(t)] and the infiltration duration (t); the relationship among the migration distance of the wetting front (Z), S, and t; the thickness of the sedimentary layer [H(t)]; and the relationship between S and t. The results revealed that I(t) and Z decreased significantly with the increase of sand contents, while H(t) increased significantly with the increase of sand contents. I(t) and Z were in the range of 7 cm and 20 cm for each treatment, respectively. The variation in I(t) with t fitted Kostiakov and Philip models, and the coefficients of determination were all greater than 0.99. With the increase in S, the infiltration coefficient gradually decreased, the infiltration index gradually increased, and the sorptivity gradually decreased. The particle composition of the sedimentary layer was similar to that of the argillaceous sediment, and the content of particles with a size of less than 2 mm in the sedimentary layer was lower than that of the argillaceous sediment. Compared with the original soil, the content of particles with a size of less than 0.05 mm and physical clay particles (diameter less than 0.01 mm) in the soil with an infiltration depth of 0-2 cm increased. The retention layer was from the topsoil to the infiltration depth of approximately 2 cm. This study can provide a scientific basis for further research on soil infiltration mechanisms under muddy water.