To further reveal the critical depth of phreatic evaporation and its influencing factors for different soil types in the Huaibei Plain, observational data from 62 sets of undisturbed soil lysimeters (0~5 m) collected at the Wudaogou Experimental Station during 2006-2024 were used in this study. Four nonlinear functions were constructed to determine the optimal fitting relationships between monthly phreatic evaporation and phreatic depth from January to December, and the critical depth of phreatic evaporation for each month was identified. The geographical detector method was then applied to analyze the dominant driving factors and their interactions affecting the critical depth of phreatic evaporation. The results showed that: ① the annual critical depth of phreatic evaporation in Shajiang black soil ranged from 1.51 to 2.69 m, with an average value of 2.13 m, while that in Huangchao soil ranged from 2.95 to 4.15 m, with an average value of 3.74 m. The nonlinear fitting results showed good agreement with the observed data. ② For both soil types, meteorological factors exerted a greater overall influence on the critical depth of phreatic evaporation in Shajiang black soil than in Huangchao soil. Surface temperature, air temperature, absolute humidity, and vapor pressure deficit exhibited strong explanatory power (q > 0.3) and were identified as the core driving factors. ③ In the Shajiang black soil area, interactions among the driving factors showed a two-factor enhancement effect, with combinations involving wind speed being the key driving interactions. In contrast, nonlinear enhancement dominated in the Huangchao soil area, and interaction combinations involving relative humidity played a dominant role.

To effectively improve the estimation accuracy of reference crop evapotranspiration (ET ₀) under limited meteorological data conditions in the Ganfu Plain Irrigation District, this study utilized daily meteorological data from the Jiangxi Provincial Irrigation Test Center Station spanning from 1987 to 2024. Using the FAO-56 Penman-Monteith (PM) model calculations as the benchmark, six machine learning models-Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Transformer, Support Vector Regression (SVR), Random Forest (RF), and eXtreme Gradient Boosting (XGB)-were developed and compared against the locally adopted Hargreaves-Samani (HS) model to investigate their applicability under various meteorological input combinations. The results indicated that under the temperature-only scenario (T _max, T _min), deep learning models including Transformer, LSTM, and GRU exhibited the best predictive robustness. Meteorological variables significantly influenced model accuracy: incorporating sunshine hours (n) alongside T _max and T _min substantially improved daily-scale ET ₀ estimation, reducing the average MAE by 39.0% and RMSE by 37.3%. The relative importance of variables to ET ₀ estimation accuracy generally followed the order: sunshine hours (n) > relative humidity (RH) > wind speed (u ₂). Additionally, extraterrestrial radiation (R _a), as a calculable parameter, provided stable accuracy gains under data-limited conditions. Regarding model applicability, for daily-scale ET ₀ estimation, the SVR model achieved the highest accuracy when meteorological inputs included n or the full-variable combinations, effectively avoiding parameter redundancy and capturing short-term fluctuations; under other input combinations (e.g., temperature-only, or with RH or u ₂), deep learning models Transformer, LSTM and GRU demonstrated superior predictive performance. For monthly-scale ET ₀ estimation, deep learning models outperformed ensemble algorithms when inputs comprised temperature (T _max, T _min) combined with R _a, RH or u?, while model differences were minimal when n was included; accordingly, the Transformer model with T _max, T _min and R _a inputs is recommended for monthly-scale ET ₀ estimation. Overall, Machine learning models demonstrated robust applicability in the Ganfu Plain Irrigation District, and optimal model selection should be guided by the specific time scale and data availability. These findings provide a scientific basis for accurate ET ₀ estimation in this region and similar data-scarce areas, and future work will integrate multi-site data to explore model transferability and spatial variability.

The increasing frequency of extreme drought events in southern China has severely impacted regional water security and socio-economic development. Focusing on the Lijiang River Basin, this study investigates the spatiotemporal evolution and propagation laws of meteorological and hydrological droughts at various time scales based on the Standardized Precipitation Index (SPI) and Standardized Runoff Index (SRI), utilizing the Mann-Kendall trend test, Pettitt abrupt change test, run theory, and Pearson correlation analysis. The results indicate a high frequency of extreme meteorological drought events within the basin, whereas hydrological droughts are primarily categorized as mild to moderate. Both SPI and SRI indices exhibit an increasing trend at monthly and annual scales. Significant abrupt changes in hydrological drought were detected at the Lingqu, Chaotian, and Pingle stations. Under mild drought conditions, meteorological droughts are more frequent but shorter and less intense than hydrological droughts. During moderate droughts, the upper basin demonstrates a drought amplification effect, while the middle and lower reaches exhibit stronger drought resilience. The propagation time from meteorological to hydrological drought ranges from 1 to 4 months, and this propagation time increases progressively from upstream to downstream. In 2022, a typical year with an abrupt wet-to-dry transition, the basin exhibited a wet state on an annual scale, yet SPI and SRI shifted by 2.79 and 3.08 on average from spring-summer to autumn-winter. At both monthly and seasonal scales, this year was characterized by frequent but mild meteorological droughts alongside rare but severe hydrological droughts.

In areas with complex planting structures and fragmented plots, conducting fine-scale monitoring of crop planting structures using medium and high-resolution remote sensing images still faces challenges such as mixed pixel interference and spectral similarity among crops. To systematically evaluate the applicability of different models in such scenarios, Xinjin District of Chengdu City was selected as the study area. Based on multi-source time-series images from Sentinel-2 and Landsat 8/9, the spectral-temporal characteristics of overwintering (small spring) and spring-sown (large spring) crops were analyzed across the blue, green, red, and near-infrared bands. Subsequently, four models - TempCNN, LSTM, Transformer, and Random Forest (RF) - were constructed to evaluate and compare their performance in planting structure extraction. The results show that TempCNN not only has high patch integrity and good spatial continuity in the identification of major crops such as wheat, rapeseed, rice, and corn, but also has significantly higher overall classification accuracy than other models. Although LSTM and Transformer perform well in some crop categories and have relatively complete patch recognition, their overall accuracy is still lower than that of TempCNN. In contrast, the extraction performance of RF is significantly inferior to the other three deep learning models. In conclusion, among the evaluated models, TempCNN performs best in extracting planting structures using medium-and high-resolution remote sensing images in areas with fragmented plots and complex planting structures.

To explore the effects of biogas slurry, biochar, and dicyandiamide application on the growth, yield, and greenhouse gas emissions of greenhouse tomatoes, seven fertilization treatments were applied under equal nitrogen input conditions from August to December 2023. These treatments included conventional fertilization (CK1), single application of biogas slurry (CK2), 0.5% biochar + biogas slurry (T1), 2% biochar + biogas slurry (T2), dicyandiamide + biogas slurry (T3), 0.5% biochar + biogas slurry + dicyandiamide (T4), and 2% biochar + biogas slurry + dicyandiamide (T5). The study analyzed the effects of these treatments on tomato root traits, yield, irrigation water use efficiency, nitrogen fertilizer partial productivity, and soil greenhouse gas emissions, and performed a comprehensive evaluation using the grey relational analysis method. The results indicated that different fertilization treatments significantly affected tomato root traits (P < 0.05), with the T5 treatment showing the best performance in root length, average diameter, total surface area, total volume, and root vitality, achieving the highest yield (139 260.0 kg/hm²). Additionally, the T5 treatment demonstrated the highest irrigation water use efficiency (47.538 kg/m³) and nitrogen fertilizer partial productivity (357.08 kg/kg). Greenhouse gas emission analysis indicated that CO₂, N₂O, and CH₄ fluxes during the entire tomato growth cycle exhibited a "rise-then-fall" trend. Among the treatments, T5 demonstrated the most outstanding performance in reducing N₂O and CH₄ emissions, showing decreases of 30.07% and 16.99%, respectively, compared with CK1. In contrast, T2 had the highest CO₂ emissions. Through AHP-entropy-weighted combination and grey relational analysis, the T5 treatment showed the highest grey relational degree, highlighting its comprehensive advantages in improving root traits, reducing greenhouse gas emissions, and enhancing tomato yield. In conclusion, the T5 treatment (2% biochar + biogas slurry + dicyandiamide) can be recommended as an effective mode for achieving high yield, high efficiency, and emission reduction in greenhouse tomato cultivation.

To elucidate the distribution patterns and correlations of pollutant concentrations in grassland soil solution during the utilization of rural domestic wastewater, and to provide a scientific basis for risk management in rural wastewater reclamation, an outdoor grassland experiment was conducted. The study investigated the effects of hydraulic loading on grassland soil organic matter,microbial α diversity,and concentrations of chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) in soil solution. The variations of soil organic matter and soil solution COD, TN, and TP in both horizontal and vertical dimensions were analyzed, along with their Pearson correlation coefficients (r). The results revealed that increasing the hydraulic load contributed to a decrease in the concentrations of COD and TP in the lower soil solution and an increase in microbial diversity of the grassland system, but also led to a reduction in the average soil organic matter content and an elevation in TN concentration in the lower soil solution. A strong positive correlation (r > 0.6) was observed between TN and TP concentrations in the soil solution along the vertical profile, and part of the nitrogen and phosphorus nutrients utilized by the grassland originated from the upward migration of inorganic nitrogen and phosphorus from the deeper soil layers. In the process of utilizing rural domestic wastewater for grassland assimilation, a hydraulic load of 0.005 m3/(m2·d) was identified as appropriate; exceeding this load would increase the risk of TN accumulation and leaching. This study provides a scientific basis for the safe resource utilization and environmental risk management of rural domestic wastewater, and holds important practical value for advancing rural domestic wastewater treatment and the recycling of water and nutrient resources.

This study aimed to optimize the irrigation mode of mid-season rice in central Hunan and improve the efficiency of water and fertilizer utilization. The long-term rainfall series in central Hunan was analyzed using the frequency method. Utilizing the Penman-Monteith method, single crop coefficient method and water balance method as the theoretical foundation, the net irrigation water requirement was determined incorporating seasonal rainfall during the growth period of mid-season rice. A questionnaire on water management for fertilizer and pesticide application in the field was collected, and the water requirements for physiological and ecological processes and efficient utilization of fertilizer and pesticide during the cultivation of mid-season rice were statistically summarized. Five irrigation modes, namely flooding, wetting, deep water storage, water control and semi-dry, were optimized, and field experiments on water-fertilizer controlled irrigation were conducted. The field experiment results showed that water control irrigation was the best mode for mid-season rice in central Hunan, with a net irrigation amount of about 284 mm, a fertilizer application level of 142 kg/hm2, a water productivity of 1.50 kg/m3 and a partial factor productivity of fertilizer of 55.6 kg/kg.

To provide a scientific basis for formulating rational water and potassium management strategies in kiwifruit orchards, this study investigated the effects of different drip irrigation water-potassium coupling treatments on the growth, development and fruit volume of ‘Jinyan’ kiwifruit under drip-fertigation. Two water-deficit levels (low water, LW; and high water, HW, applied at 60% and 80% of the control [CK], respectively) and three potassium-deficit levels (low potassium, LK; medium potassium, MK; and high potassium, HK, applied at 40%, 60%, and 80% of CK, respectively) were applied during the flowering-fruit set, fruit swelling, and fruit maturation stages. The results showed that the net photosynthetic rate (P_n ) was highest in the CK treatment at the pre-midday peak during flowering-fruit set [9.95 μmol/(m2·s)], while the high-water-high-potassium (HWHK) treatment increased the post-midday peakby 15.27% compared with CK. The CK treatment performed best during fruit swelling. During fruit maturation, the high-water-medium-potassium (HWMK) treatment significantly raised the afternoon P_n peak by 44.82% over CK. The intrinsic water-use efficiency (WUE?) during flowering-fruit set increased by 4.40%~122.01% in all treatments relative to CK, and the HWHK treatment elevated the morning-peak WUE? by 15.22% in the fruit-swelling stage. Shoot growth rates during flowering-fruit set were 17.98%~25.83% lower in all treatments than in CK. After entering the fruit-swelling stage, compensatory growth occurred in the HWLK and HWMK treatments, with shoot growth rates increasing by 4.24%~7.70% compared with CK. The HWHK treatment enhanced the growth rate by 39.38% during fruit swelling and by 31.46% during fruit maturation relative to CK. The mean leaf area index (LAI) over the whole growth period ranged from 2.30 to 2.69 for all water-potassium deficit treatments, showing a declining trend compared with the CK value of 2.70. The HWHK treatment exhibited the smallest reduction, decreasing by only 0.65%~2.53%. In terms of fruit volume, during the rapid growth period of fruit maturity, the volume growth rate of HWHK treatment and HWMK treatment were significantly higher than that of CK. In terms of yield, HWMK during the flowering and fruit-setting period and the fruit ripening period performed best, while HWHK during the fruit expansion period performed best. Therefore, it is recommended to apply the HWMK treatment during the flowering-fruit set and fruit maturation stages, and the HWHK treatment during the fruit expansion stage. This approach can reduce water and potassium inputs and improve resource-use efficiency without affecting tree growth or fruit formation.

As a soil amendment, the effect of biochar application on soil saturated hydraulic conductivity is governed by multiple factors, such as application rate, biochar type, and soil texture, and the underlying mechanism remains poorly understood. This study combined field experiments with three-dimensional modeling to investigate the improvement effects of rice straw biochar and bamboo biochar at different addition rates (3%, 6%) on the saturated hydraulic conductivity of silt loam soil in the Qujialing experimental field. Field results indicate that 3% rice straw charcoal and 6% bamboo charcoal treatments enhanced saturated hydraulic conductivity in the 0~20 cm soil layer, while 6% rice straw charcoal and 3% bamboo charcoal produced inhibitory effects. To elucidate the underlying mechanisms, a 3D pore structure model of the soil-biochar mixture was constructed using Blender software to calculate the mixture porosity. Subsequently, the Terzaghi and Chapuis models were applied to predict the soil hydraulic conductivity. Comparisons revealed that the Terzaghi model more closely approximated measured values, while the Chapuis model systematically underestimated them. Both models reproduced the experimental trends observed with rice straw charcoal treatment, confirming that moderate amounts of rice straw charcoal improve saturated hydraulic conductivity by optimizing pore connectivity. However, both models exhibited certain deviations in predicting bamboo charcoal effects, primarily due to their failure to account for internal bamboo charcoal porosity.

Since 2022, the rapid expansion of automated soil moisture monitoring networks has highlighted an urgent need for high-precision and wide-coverage soil moisture forecasting. To address this operational challenge, this study proposes a hierarchically improved hybrid deep learning framework based on Python (scikit-learn and PyTorch) and applies it to Yunnan Province as a typical study area. The framework follows a hierarchically structured modeling pipeline: baseline model screening, feature enhancement, architecture optimization, optimizer comparison, and physical constraint integration. The main results indicate that: ① The mean squared error (MSE) ranges for traditional time-series, machine learning, and deep learning models are 21.01~256.33, 20.54~29.66 and 15.20~21.04, respectively; deep learning models significantly outperform the others, with the CNN-LSTM model achieving the best baseline performance (MSE = 15.20). ② By integrating a physics-informed multi-level feature engineering system and a bidirectional long short-term memory (BiLSTM) mechanism along with optimizer optimization, the model MSE is further reduced to 7.51. Furthermore, an operationally applicable physically constrained CNN-BiLSTM model is developed (MSE=7.92) representing a 47.9% improvement in prediction accuracy over the baseline while ensuring physical consistency. ③ The addition of an attention mechanism fails to yield performance gains, instead degrading the performance with an MSE increase to 9.80. ④ When applied to the 2026 soil moisture forecast for Yunnan, the model results align well with regional drought evolution, confirming the framework’s strong generalization ability and practical applicability for operational soil moisture prediction.

Affected by concentrated rainfall and the backwater effect induced by the high water level of the Yellow River during the flood season, farmland adjacent to the Yellow River riparian zone is characterized by a shallow groundwater table, strong evaporation, and severe soil salinization, which seriously restrict high-quality agricultural development along the Yellow River. To address the insufficient understanding of the effectiveness of coordinated drip irrigation and subsurface pipe drainage in regulating groundwater and suppressing soil salinization, a monitoring network was established in the Lingsha subsurface drainage experimental area of the Yinbei Irrigation District to periodically monitor the dynamic variations of soil water and salt in drip-irrigated farmland under subsurface drainage regulation. Data from 17 consecutive months of continuous monitoring (May 2024 to September 2025) showed that subsurface drainage could rapidly draw down the elevated groundwater table caused by heavy rainfall during the flood season and the backwater effect of the high water level of the Yellow River to near the design burial depth of the drainpipes. Under the combined action of drip irrigation and subsurface pipe drainage, the soil moisture in the 0~0.6 m soil layer ranged from 20% to 30%, which was close to the field capacity (22%) of the experimental area. Meanwhile, drip irrigation and precipitation drove salts in the crop root zone to migrate downward gradually. Correlation analysis showed that groundwater table depth was significantly correlated with soil electrical conductivity in the 1.0~1.2 m layer (r≤-0.44, p<0.05). These results indicate that drip irrigation mainly maintained soil moisture in the root zone within an appropriate range and promoted downward salt movement, whereas subsurface pipe drainage affected the distribution of water and salt in deeper soil layers by regulating groundwater table. Their combination effectively alleviated soil salt accumulation in the crop root zone.

To explore the sediment deposition characteristics of the barrel section of siphon culverts in plain river network areas and optimize the cross sectional dimensions of the culvert barrel for the purpose of sediment reduction, three-dimensional full-flow field numerical simulations were performed using Computational Fluid Dynamics (CFD) on the inlet, barrel, and outlet sections. Subsequently, the velocity distribution law and low-velocity zone characteristics across key cross-sections were analyzed. Taking the minimum head loss of the culvert as the objective function, the geometric dimensions of the cross-sectional fillets were optimized using the central composite design method, and the optimal fillet dimensions of the culvert cross section were obtained. Based on the non-silting velocity, the sediment deposition patterns of the key cross-sections before and after optimization were compared. The results show that a certain degree of sediment deposition occurs in the barrel section of the siphon culvert under the low-head design flow condition, and the deposition areas are mainly located at the junction between the descending and horizontal sections, as well as the ascending section of the barrel. When the optimal fillet dimensions of the culvert cross section are set as a base length of 0.45 m and a side height of 0.50 m, the flow velocity of the characteristic cross sections of the optimized culvert is increased, and the sediment deposition amount is only 27.28% of that in the initial design scheme. This study confirms that the combined CFD-RSM (Response surface methodology) method can be well applied to optimize the structural dimensions of hydraulic structures. The barrel section of siphon culverts in plain river network areas is prone to sediment deposition zones, and the siphon culverts designed with filleted cross sections can achieve the effect of reducing sediment deposition. This research provides technical support for the structural design of siphon culverts in plain river network areas.

Shanpingtang (mountain) ponds serve as important sites for intercepting sediment deposition during soil erosion in small watersheds, and the grain size distribution (GSD) characteristics of deposited sediments provide a crucial basis for reconstructing the spatiotemporal evolution of regional soil erosion. This study, through field sampling of soil and sediments along with laboratory grain size analysis, comparatively analyzed the particle size distribution characteristics and fractal dimension relationships between deposited sediments in mountain ponds and surface soils from farmland, grassland, and forestland within the watershed. The results indicate: ① The particle size composition of deposited sediments in mountain ponds, as well as surface soils from farmland, grassland, and forestland, is dominated by clay (d<0.002 mm) and medium silt (0.005≤d< 0.02 mm). The average combined proportions of these two fractions are 61.19%, 71.36%, 74.85%, and 82.34%, respectively, while other particle size fractions account for relatively small proportions, particularly coarse sand (0.5≤d<1 mm) and very coarse sand (1≤d<2 mm). Cluster distance matrix analysis further reveals that the GSD of pond sediments is highly similar to that of farmland and grassland soils. ② The sediment profiles are classified into 10 clusters, reflecting a highly similar, multi-cyclic nature of the rainfall erosion-transport-deposition processes within the watershed. ③ The mean grain-size fractal dimension D of particle size for deposited sediments in mountain ponds is 3.62, which is higher than that for farmland (3.58), grassland (3.44), and forestland (3.30). The basic characteristics of particle size in mountain pond sediments show the highest correlation with farmland soil. Overall, the findings indicate that deposited sediments in mountain ponds primarily originate from farmland and grassland, with fine particles being preferentially eroded and deposited in still-water environments. This study provides a scientific basis for soil erosion control and sediment management in small karst watersheds.

The main raw materials of stalk composite pipe are straw and soil, and the water is transported to the crop root zone through the pores between straw and soil. However, the effect of different working pressures, buried depths, and pipe wall thicknesses on soil water movement remain unclear, restricting parameter optimization and performance stability in practical applications. Therefore, based on the theory of soil moisture dynamics, a soil moisture transport model of stalk composite pipe infiltration irrigation was constructed and validated using data from indoor soil-box experiments. The model was then applied to investigate the laws of soil water movement under various working pressures, burial depths, and wall thicknesses. The results indicated that the simulated values of soil moisture content and wetting front migration distance were in good agreement with the measured values, with determination coefficients (R2) greater than 0.85, indicating that the model can accurately simulate the subsurface irrigation process using stalk composite pipes. The discharge per unit length of stalk composite pipe decreased rapidly at the initial stage of infiltration, and then continued to flow with a relatively stable flow rate. The stable flow rate increased with the increase of working pressure and decreased with the increase of pipe wall thickness, but the influence of burial depth was not significant. The wetted soil zone exhibited an approximately circular shape. The wetting front advance distance expanded with higher working pressures, with the horizontal advance being shorter than the vertical advance, while the wall thickness showed a negligible effect on the wetting front. As infiltration time progressed, the soil moisture content rapidly increased initially before stabilizing. The stable soil moisture content was positively correlated with working pressure but negatively correlated with both burial depth and wall thickness.