[Objective]To address the challenges in short-term rainstorm disaster risk assessment, including strong suddenness of rainstorms, limited warning time, and insufficient assessment accuracy, a refined risk assessment and early-warning method is proposed, which integrates real-time meteorological data with dynamic risk factors. [Methods] Based on hourly observed and forecast precipitation data on a 3km×3km grid in Nanjing, a risk quantification model of disaster-causing factors was developed (including five indicators such as rainstorm warning level, affected area, and process precipitation). Combined with vulnerability indicators of exposed elements (including river water level, dynamic population density, and disaster data) and the precipitation probability index, the comprehensive risk index (R) was calculated by GIS spatial overlay and a multiplicative integration approach to classify four risk levels: extremely severe (R≥9), severe (7<R<9), moderate (5<R<7), and general (3<R<5). The weights of the factors in the model were preliminarily validated through historical event inversion and sensitivity analysis. [Results] The model validation during the September 2024 Nanjing rainstorm event demonstrated that high-risk areas could be accurately identified (at 0.03°×0.03° spatial resolution). The risk indices calculated reached 9. 2 in southern Baima Town of Lishui District and 9. 1 in northern Yaxi Town of Gaochun District, both classified as extremely severe level, which was consistent with the field disaster investigation result (1 652 cases of fallen trunks/ branches and 79 cases of fallen billboards). With realtime data updates, the dynamic assessment response time was maintained within one hour, with an early warning accuracy rate of 87.3%. [Conclusion] Through multi-source data integration and grid-based dynamic calculation, the proposed method significantly improves the timeliness and accuracy of short-term rainstorm risk assessment and provides street-level refined early warning support for urban disaster prevention and mitigation.

[Objective] To clarify the quantitative impact mechanism of landscape pattern reconstruction on river water quality under the compound influence of geological disasters and intensive human activities in dry-hot valley basins, thereby providing a scientific basis for the coordinated management of soil and water resources in ecologically fragile areas of Southwest China. [Methods] Taking the Xiaojiang River Basin in the lower reaches of the Jinsha River as the study area, a synchronous monitoring network for water, sediment, and water quality was integrated. Landscape patterns were quantitatively analyzed using Fragstats 4.2 software. Redundancy analysis was applied to systematically investigate the impact of landscape reconstruction on river water quality under the combined action of human activities and geological disasters. [Results] The results showed that:(1) the average values of key water quality indicators such as total phosphorus, total nitrogen, ammonia nitrogen, and COD during the flood season were 2.00, 1.88, 3.21 and 4.43 times higher than those of the non-flood season, respectively.(2) The average concentrations of total nitrogen and COD were 2.28 mg/L and 27.36 mg/L in the disaster-affected areas, 2.03 mg/L and 9.23 mg/L in the residential areas, and 2.00 mg/L and 6.60 mg/L in the cultivated land.(3) Under rainstorm events, the pre-rain fluxes of total phosphorus, total nitrogen, ammonia nitrogen, and COD were 1.95, 2.41, 2.34, and 4.74 times higher than the post-rain fluxes, with the most pronounced increase observed in the disaster-affected areas.(4) At different buffer zone scales, the selected landscape indices explained over 94.00% of the water quality variation during the flood season. The indices with the highest explanation rate were the aggregation index of disaster-affected areas(43.10%), the dispersion and adjacency index of residential areas(40.70%), and the division index of cultivated land(16.80%). [Conclusion] Water quality indicators in the Xiaojiang River Basin exhibit significant seasonal variations. Spatially, pollutant concentration gradient follows the pattern: disaster-affected areas > residential areas > cultivated land. These spatial disparities are further amplified under the influence of rainstorm. Landscape pattern reconstruction caused by landslides and debris flows—similar to that driven by urbanization and agricultural expansion—may cause further deterioration of river water quality. The findings provide scientific guidance for soil and water conservation, aquatic environment protection, and disaster mitigation in the dry-hot valleys of Southwest China.

[Objective] The floods caused by dam failures exhibit strong impact forces, extensive inundation areas, fast flow velocities, and great destructive power. They pose a significant threat to public life, property safety, and social and environmental stability in downstream areas. To improve evacuation efficiency in dam failure accidents, it is crucial to conduct research on flood evacuation route planning. [Methods] Minimum evacuation time and lowest public safety risks during evacuation were set as the goals, and the dynamic flooding propagation caused by dam failures was used as a constraint. By using unmanned aerial vehicle(UAV) tilt photogrammetry technology to map the road network in the downstream of the reservoir and integrating it with flooding propagation data, a coupling model of flood-road network information was developed. Then, using this model and Dijkstra's algorithm, a flood evacuation route planning method was proposed, which was further validated through a case study of the Shanmei Reservoir in Fujian, China. [Results] The result shows that based on the dynamic process of road network failure in the downstream of Shanmei Reservoir, four evacuation plans were designed for 8 disaster-affected sites, with 13 evacuation routes in total. All evacuees from disaster-affected sites successfully avoided floods during the evacuation. [Conclusion] The findings demonstrate that the flood evacuation route planning method for dam failure accidents based on coupling flood and road network information effectively ensures evacuation route safety and meets both safety and timeliness requirements of emergency management for flood evacuation routes in the event of dam failure accidents.

[Objective] In the context of global climate change, extreme hot and humid events have become more frequent and widespread, posing significant threats to human health, socioeconomic development, and ecosystems. Investigating the spatiotemporal evolution of extreme hot and humid events and their response patterns to preceding extreme precipitation provides a scientific basis for improving regional climate adaptation strategies and early warning systems for extreme weather events. [Methods] Based on ERA5 reanalysis data from 17 cities in Hubei Province from 1985 to 2024, extreme hot and humid events were identified using the simplified Wet-Bulb Globe Temperature(sWBGT) index. Trend tests, multiple regression, and other method were applied to systematically analyze the spatiotemporal evolution characteristics of extreme hot and humid events in the study area and their response mechanisms to extreme precipitation. [Results] (1) Extreme hot and humid events in Hubei Province exhibited a significant increasing trend, and their duration showed an upward trend, particularly for HS-EP events(hot and humid events with preceding extreme precipitation).(2) Extreme hot and humid events showed different spatial distribution patterns. The western region experienced high frequency(>80 times), long duration(>5.6 days), and low intensity(<35), while the central region exhibited low frequency(<70 times), short duration(<4.8 days), and high intensity(>36).(3) Preceding extreme precipitation played a regulatory role in extreme hot and humid events. The logarithmic ratios of three characteristics(frequency, duration, and intensity) between HS-EP events and HS-NEP events(hot and humid events without preceding extreme precipitation) were-1.200, 0.148, and-0.005, respectively.(4) The impact of preceding extreme precipitation showed notable lagged effects and regional differences, exhibiting an evolution pattern of “short-term suppression(1~2 days)-long-term enhancement(7~10 days)”[Conclusion] The findings reveal the spatiotemporal distribution patterns of extreme hot and humid events in Hubei Province and their response patterns to preceding precipitation, providing new insights into the interactive response mechanisms between regional extreme hot and humid events and extreme precipitation under global warming.

[Objective] This study aims to investigate streamflow variations and hydrological drought characteristics in the lower Songhua River Basin under future climate scenarios, providing a scientific basis for regional water resource management and guidance for formulating comprehensive drought and flood disaster prevention and control strategies. [Methods] Based on hydrometeorological data from 1970—2014, a SWAT model was constructed and driven by 14 CMIP6 climate models. Five optimal climate models were selected through Taylor diagram analysis and combined with SSP1-2. 6, SSP2-4. 5, and SSP5-8. 5 scenarios to simulate streamflow variations from 2015 to 2100. [Results] The SWAT model demonstrated excellent performance at Tongjiang hydrological station, with R² and NSE values exceeding 0. 7 during both calibration and validation periods. The absolute values of PBIAS were below 10. 7%, confirming model reliability. Future climate predictions indicated increasing trends in precipitation, temperature, and evapotranspiration across the river basin. The magnitude of temperature and evapotranspiration increases followed the order of SSP5-8. 5 > SSP2-4. 5 > SSP1-2. 6, while precipitation increases ranked as SSP5-4. 5 > SSP1-2. 6 > SSP2-8. 5. Seasonally, SSP1-2. 6 and SSP5-8. 5 scenarios showed significant precipitation increases in autumn and winter, while SSP2-4. 5 scenario exhibited notable increases in summer and winter. Streamflow variations strongly depended on scenarios: SSP1-2. 6 showed significant increases, while variations under SSP2-4. 5 and SSP5-8. 5 were not significant. Standardized Streamflow Index(SSI) analysis showed significant increases under SSP1-2. 6, but declining trends under other scenarios. Drought characteristic analysis indicated that under SSP1-2. 6, all drought indices except average drought duration decreased, while under SSP2-4. 5 and SSP5-8. 5, all indices except drought frequency increased. Seasonally, SSP1-2. 6 increased spring-summer proportions with elevated winter drought frequency, while SSP2-4. 5 and SSP5-8. 5 increased autumn-winter proportions with corresponding frequency changes. [Conclusion] Future climate change will significantly impact hydrological characteristics in the lower Songhua River Basin. The SSP1-2. 6 scenario promotes streamflow increase and drought mitigation, while the SSP2-4. 5 and SSP5-8. 5 scenarios may exacerbate drought risks. The findings provide decision-making support for regional water resource management and disaster prevention, recommending differentiated adaptation strategies tailored to specific climate scenarios.

[Objective] The extent and contribution of meteorological drought's influence on agricultural drought across different land use types in the three provinces of Northeast China remain unclear. Given that meteorological drought serves as a key driver of agricultural drought, it is necessary to quantify its influence and contribution. [Methods] The Standardized Precipitation Evapotranspiration Index(SPEI) and the Vegetation Condition Index(VCI) were used to represent meteorological drought and agricultural drought, respectively. Using trend analysis, correlation coefficients, and residual trend analysis, the spatiotemporal variation characteristics and response relationships between meteorological and agricultural drought across different land use types in the three provinces of Northeast China from 2000 to 2023 were investigated. Additionally, the contributions of climatic factors(SPEI and precipitation) and human activities to agricultural drought were evaluated. [Results] The result showed that:(1) across different regions and different land use types, both SPEI and VCI in the three provinces of Northeast China exhibited upward trends, indicating a gradual alleviation of meteorological and agricultural drought.(2) SPEI and VCI were positively correlated across different regions and different land use types, with forest land being the least sensitive.(3) The combined influence of climatic factors(SPEI and precipitation) and human activities drove changes in agricultural drought in the three provinces of Northeast China. Human activities were identified as the primary driver of agricultural drought changes in cropland, accounting for 70.5% of the total contribution. Moreover, the contribution of human activities significantly exceeded that of climatic factors, especially across the entire Liaoning Province, southeastern Jilin Province, and western Heilongjiang Province, where positive contributions from human activities exceeded 80%. [Conclusion] These findings reveal the variation patterns of meteorological and agricultural droughts in the three provinces of Northeast China and provide a theoretical basis for evaluating drought disaster relationships.

Accurate precipitation prediction plays a crucial role in regional flood prevention and mitigation,water resources management,and socioeconomic development.However,the precipitation process is influenced by the interaction of multi-scale meteorological factors and shows significant nonlinearity and spatiotemporal heterogeneity.Traditional numerical models fail to effectively capture its complex evolution patterns. [Methods]Based on random forest stacking techniques,six hybrid prediction models were constructed:KNN-LSTM,SARIMA-KNN,SARIMA-Prophet,SARIMA-LSTM,Prophet-LSTM,and Prophet-KNN.Monthly precipitation data from 1990 to 2023 at station 58238 in Nanjing were used for modeling,with data from 1990 to 2020 used as the training set and data from 2021 to 2023 used as the testing set.The regional generalization ability was validated using contemporaneous data from 12 independent meteorological stations in Jiangsu Province. [Results]The result showed that the SARIMA-LSTM hybrid model,which integrated the seasonal decomposition advantage of SARIMA with the long-term dependency capturing ability of LSTM,achieved the highest prediction accuracy on the testing set,with R²=0.904,MAE=16.16 mm,and MSE=477.87 mm².The regional generalization validation demonstrated that the model achieved $ \overline{R^{2}}$=0.919,$ \overline{M A E}$ =15.33 mm,and $ \overline{M S E}$=537.52 mm² across 13 meteorological stations in Jiangsu Province,indicating good spatial generalization capability. [Conclusion] The constructed hybrid models exhibit excellent predictive performance,providing reliable technical support for precipitation prediction in the lower Yangtze River region.This holds significant application value for the optimization of regional water resource allocation and disaster early warning.

[Objective] Soil moisture is a key variable affecting the precipitation-runoff process. However, the lack of soil moisture observations in complex terrain regions makes it difficult to clarify its mutual feedback mechanisms with precipitation and runoff, which has become a major bottleneck for improving the understanding of river basin hydrological processes and flood forecasting accuracy. Therefore, revealing the co-evolution patterns of soil moisture, precipitation, and runoff, as well as their regulatory mechanisms on runoff, is of great significance for water resources management under changing environments. [Methods] Based on daily observational data from the flood season(April to September) from 2010 to 2021 in the Jianxi River Basin in the upper reaches of the Min River, multiple method including trend tests, ordered clustering, lagged regression, hierarchical analysis, and model comparison were employed to explore the coupling mechanisms of precipitation, soil moisture, and runoff. [Results] The result showed that:(1) after a synchronous abrupt change in precipitation and runoff in July 2010, both entered a long-term stable stage, while soil moisture remained stable, exhibiting the characteristics of a “hydrological buffer”.(2) A significant scale effect was observed in the relationships among the three elements. At the monthly scale, precipitation dominated runoff, whereas at the daily scale, soil moisture had stronger regulatory capacity, indicating that the dominant factor varied with scale.(3) The regulation of soil moisture on precipitation and runoff was state-dependent, with runoff efficiency under wet conditions being 5 to 6 times higher than under dry conditions, and it exhibited a differentiated lag response pattern.(4) Model comparison verified the above mechanisms. The hybrid model integrating precipitation and runoff memory(AR+XGBoost) showed optimal performance(NSE = 0.876), with the contribution rate of its antecedent runoff memory characteristics reaching 13.3%. [Conclusion] The precipitation-runoff transformation in the Jianxi River Basin is a nonlinear process driven by precipitation, with soil moisture serving as a state variable. Its transformation efficiency depends on the antecedent soil moisture state and its lagged regulatory mode. The findings provide a scientific basis for water resources management and flood forecasting in complex terrain river basins.

[Objective] To address the limitations of unsystematic integration of natural and social dual characteristics and insufficient modeling of spatiotemporal heterogeneity in water consumption prediction, a spatiotemporal collaborative prediction framework for water consumption in the nine provinces and regions along the Yellow River is constructed based on deep learning method. [Methods] A preliminary dataset was constructed using 29 characteristic factors influencing water consumption. The importance of these factors was ranked using the random forest algorithm, and redundant features were eliminated. Considering the characteristics and applicable scenarios of different deep learning algorithms, a hybrid prediction model based on convolutional neural network(CNN), long short-term memory(LSTM) network, and attention mechanism(AM) was established and compared with other baseline models. To address the problem of extreme errors, a dual-attention collaborative mechanism was designed to optimize the model. [Results] In the study area, the CNN-LSTM-AM model achieved better simulation result than other models, with mean absolute error(MAE), mean absolute percentage error(MAPE), and root mean square error(RMSE) reduced by 7.7%~40.6%, 22.6%~44.1%, and 0.7%~32.1%, respectively, indicating superior overall performance. After introducing the dual-attention collaborative mechanism, extreme errors were reduced while maintaining small fluctuations in overall accuracy. The model demonstrated good generalization ability and was able to predict future water consumption in the study area with high accuracy. [Conclusion] In the study area, the current model shows good applicability and prediction accuracy, providing a new technical approach for spatiotemporal collaborative prediction of water consumption. Future research should consider the balance among model adaptability, complexity, and stability based on task requirements, and construct a comprehensive prediction system through multi-dimensional analysis.

[Objective] Data assimilation method can integrate observed soil moisture into hydrological models to enhance the accuracy of the simulation process. However, in the absence of observed data for validation, assimilating a single source of unobserved soil moisture under certain precipitation conditions may introduce greater simulation deviations. To address this, a multi-source soil moisture assimilation approach based on precipitation scenario selection is proposed. [Methods] Taking the upstream watershed of Hongjiata station as an example, an assimilation scheme was constructed by coupling the WEP-L model with the Ensemble Kalman Filter algorithm. Four precipitation scenarios—heavy rain, moderate rain, light rain, and no rain—were classified using watershed precipitation data. The assimilation performance of three soil moisture datasets was evaluated under different precipitation scenarios using relative error as the evaluation indicator. Based on the strategy of assimilating the better-performing data under the corresponding precipitation scenarios, an optimized scheme(SM4-1) was developed. [Results] The results showed that scheme SM4-1 improved runoff simulation performance across different precipitation scenarios, with the Nash-Sutcliffe efficiency coefficient increasing by 0.032 and the relative error decreasing by 17.4%. The errors in runoff coefficients for heavy rainfall events and annual periods were reduced by 4.06% and 17.38%, respectively, compared to the original scheme. [Conclusion] The results indicate that the proposed multi-source soil moisture assimilation method based on precipitation scenario selection enhances both the magnitude and spatial distribution of soil moisture under corresponding rainfall conditions. This approach demonstrates potential for optimizing runoff simulation performance and enhancing overall model accuracy.

[Objective] In order to investigate the mechanical response law of the overreach small conduit-arch anchor support system in the construction of fractured perimeter rock section, Tongzi Tunnel was taken as the engineering background, and constructs the synergistic mechanical model of “overreach small conduit-steel arch-anchor”[Methods] Based on the theory of structural mechanics, combining Simpson's numerical integration method with the principle of virtual work, the load transfer path is analyzed step by step, and the parameters are dynamically calibrated by using the actual tunnel measurement data. [Results] The result show that the model can accurately predict the transfer loads(shear force, bending moment, axial force), arch foot displacement and arch settlement under different excavation steps, and the error is less than 9% compared with the on-site monitoring data, which verifies the reliability of the model; the overtopping small conduit grouting improves the stability of the surrounding rock significantly, and its anchors share about 6% of the surrounding rock load, and the reduction in the spacing of the steel arch frame can inhibit the settlement effectively, but it needs to take into account the economy. The spacing of steel arch can effectively inhibit the settlement, but need to take into account the economy, it is recommended that the spacing of Ⅳ perimeter rock section take 80 cm. [Conclusion] The three-dimensional cooperative mechanical model proposed in this study realizes for the first time the load analysis of the whole path of the overrun support system from the grouting anchor solid to the locking foot anchor, which provides a reference basis of both theoretical rigor and engineering practicability for the design of the dynamic support of overrun small catheters.

[Objective] Image-based flow measurement is an effective non-contact method for discharge measurement, which has been widely used in open channels and natural rivers. However, its applicability in confined and complex environments such as the drainage ditches of high-arch dam galleries has not been fully validated. The aim is to systematically evaluate the performance of this technology in this specific scenario. [Methods] Validation was conducted using a combination of model tests and field tests. Camera calibration was performed by placing control points. In the model tests, three camera perspectives(upstream, downstream, and lateral) were tested while controlling discharge variations. Water levels were obtained using a monocular vision-based method without a staff gauge, and surface flow fields were calculated using the particle image velocimetry(PIV) technique. Comparisons were made between conditions with and without tracer particles, as well as before and after image preprocessing. Field tests were conducted in a real gallery drainage ditch. Measurements from staff gauges and a measuring weir were used as benchmarks to validate the engineering applicability of this technology. [Results] The result showed that in the model tests, the relative errors of the image-based water level recognition method were all below 7%. Under conditions with tracer particles, the flow measurement errors across different camera perspectives were all less than 4%. The accuracy of the flow measurement result from the lateral perspective was lower than that from the upstream and downstream perspectives. The flow calculation error increased with the increase of discharge, rising from 5.77% at a low discharge(21.40 m³/h) to 35.17% at a high discharge(65.10 m³/h). After image preprocessing, the errors could be reduced to the range of 3.60%~28.74%. In the field tests, the water level measurement error was 0.78%. Under conditions with tracer particles and under conditions without tracer particles but with image preprocessing, the relative errors of discharge calculation reached 1.17% and 1.81%, respectively. [Conclusion] The result show that image-based flow measurement technology is suitable for discharge monitoring in the drainage ditches of high-arch dam galleries, and the upstream and downstream perspectives are the optimal camera perspectives. Under conditions with natural or artificial tracers, this technology can achieve high measurement accuracy. Image preprocessing can enhance the flow measurement performance, while an increase in discharge can reduce the algorithm's ability to identify flow field features, thereby decreasing the accuracy of image-based flow measurement. By analyzing the application of image-based flow measurement technology in both model and field settings, it is confirmed that this technology has good potential for application in the drainage ditches of high-arch dam galleries and is expected to provide a valuable reference for algorithm optimization and broader application.

[Objective] Pressure waves have been widely applied in the field of pipeline leakage detection due to their ability to carry energy and information. The aims are to address the limitations of existing pressure wave generators and to further investigate the excitation and propagation characteristics of pressure waves in leaking water pipelines. [Methods] A nozzle-type pressure wave generator was designed, which actively excited pressure waves by using compressed gas to impact the water pipeline through a nozzle, and was applied to an experimental platform for water pipeline leakage detection. In a pipeline about 120 meters long, the effects of intake pressure, intake time, leakage aperture, and leakage location on the upstream and downstream pressure wave characteristics of the leakage hole were tested under both leakage and non-leakage conditions. Leakage localization was performed using two methods: the difference method based on the pressure curve difference between leakage and non-leakage conditions, and the slope method based on the slope of the pressure curve under leakage conditions. [Results] Experimental results showed that the presence of pipeline leakage led to distortion of the pressure waveform and the occurrence of waveform separation at the upstream monitoring point of the leakage hole, and also caused overall pressure attenuation at the downstream monitoring point. As the intake pressure increased, intake time extended, and leakage aperture enlarged, these leakage characteristics became more pronounced. The distortion position of the pressure waveform at the upstream monitoring point varied with the leakage location. The distortion time point could be effectively identified, and the leakage could be localized using the difference method and the slope method. The maximum absolute localization errors using the difference method and the slope method were-2.31 m and-2.78 m, respectively, with maximum relative errors of-5.64% and-8.24%, respectively. [Conclusion] The research findings verify the feasibility of using a nozzle-type pressure wave generator for water pipeline leakage detection and can provide references for leakage identification and localization technologies in water pipelines.

[Objective] Flood discharge atomization has become a critical issue in the safety protection of high dam flood discharge. The aims are to address the issues of low computational accuracy and high computational load in traditional random splashing models of water droplets, as well as the limitations of the graded random splashing model of water droplets and the number theory graded random splashing model of water droplets in engineering practices. [Methods] A novel random splashing model of water droplets was established based on the theory of high-dimensional numerical integration using good point sets, combined with GPU parallel computing technology. The model was verified through numerical experiments. [Results] The result showed that under the same conditions and with the same number of water droplets, the numerical solution of the proposed model exhibited smaller total absolute errors than those of traditional models. As the number of water droplets increased, the reduction in total absolute errors became generally more pronounced. When the number of water droplets exceeded 2 million, the reduction stabilized at approximately 50%. With the aid of GPU parallel computing technology, the computational efficiency of the model was significantly enhanced, achieving an acceleration ratio of about 100 times. Especially when processing larger-scale data, the computational time of the model was significantly reduced, further highlighting its advantage in computational performance. [Conclusion] The findings indicate that the proposed model can effectively handle computational cases with four-dimensional random variables as initial conditions. It significantly improves computational accuracy and efficiency compared to traditional random splashing models of water droplets, making it suitable for practical applications in large-scale simulations of random splashing of water droplets.

[Objective] Ecological flow serves as a fundamental constraint in determining the transferable water volume for the Western Route of the South-North Water Transfer Project, requiring comprehensive quantification that balances multiple requirements such as the maintenance of downstream ecological functions, water resource utilization, and scheduling management. However, the optimal selection of ecological flow method under different requirements in the Western Route Project requires further research. [Methods]64 years of observed hydrological data from two key control stations—Yajiang station and Dajin station—in the water transfer watershed were utilized, and ten typical hydrological method were selected to calculate ecological flow, including the Tennant method, Texas method, and intra-annual distribution method. Additionally, a fuzzy optimization method was applied to evaluate the suitability of these hydrological method for ecological flow under different requirement weights. [Results] The result showed that ecological flow result calculated by different hydrological method exhibited different suitability under four critical requirements: hydrological regime consistency, downstream ecological requirement assurance, upstream water transfer capacity, and long-term inflow satisfaction. When these four requirements were comprehensively evaluated using the fuzzy optimization method, the Texas method demonstrated the highest suitability. Intra-annual ecological flow ranges were recommended at 64.6~608 m³/s for the Yajiang station and 48.4~491 m³/s for the Dajin station. [Conclusion] The suitability assessment of hydrological method under multi-objective requirements has enhanced the technical framework for research on ecological flow in the Western Route Project. It provides important scientific references for ensuring ecological safety, optimizing water resource allocation, and informing subsequent engineering design and scheduling management.

[Objective] To address the complex hydrodynamic conditions in mountainous river reservoir systems, a three-dimensional flow-temperature-concentration coupled numerical model based on the lattice Boltzmann method is developed to describe water flow, thermal stratification, and pollutant transport processes. [Methods] A multi-distribution-function scheme was employed to solve the flow field, temperature field, and concentration field in parallel. The model was validated using three-dimensional double-diffusive natural convection cases, and the relative errors of the averaged Nusselt and Sherwood numbers under multiple Rayleigh number conditions were all less than 0.8%. On this basis, the model was applied to a typical river reservoir in the southwestern plateau to simulate water temperature evolution and short-term pollutant transport under cold-water inflow conditions. [Results] The result showed that after the cold-water inflow, a subsurface undercurrent formed at the reservoir bottom, inducing thermal stratification. When the main cold-water flow reached the dam site, downstream water temperature decreased from approximately 11.5 ℃ to approximately 6.5 ℃ within about 60 minutes. Pollutants accumulated locally in low-velocity zones, with high-concentration regions extending up to approximately 150 m, and were transported longitudinally in the mainstream predominantly by advection. [Conclusion] The coupled model can stably characterize the multi-field interactions of flow, temperature, and concentration under complex boundary conditions. It is applicable to high-resolution simulations of stratification induced by cold-water undercurrents and short-term pollutant transport in mountainous river reservoir systems, providing a reference for stratified water intake scheduling and pollution risk analysis.

[Objective] High temperatures generated by coal seam spontaneous combustion cause thermal damage to surrounding rock, leading to significant changes in its mechanical properties and structure and potentially inducing geological disasters such as water inrushes and roof collapses. The evolution patterns of mechanical properties, permeability, and fracture characteristics of rock under different burning temperatures and burial depths are investigated in this study. [Methods] Sandstone specimens subjected to different high-temperature pretreatments were selected as the research objects. Triaxial compression seepage tests were conducted under constant seepage pressure and varying confining pressures, with synchronous acoustic emission monitoring. [Results] The result showed that the mechanical properties of sandstone progressively strengthened with increasing confining pressure, while they initially exhibited slight enhancement and then gradual deterioration as the burning temperature rose. Under varying confining pressures and burning temperatures, the permeability followed a trend of initial reduction followed by a subsequent increase. [Conclusion] The result indicate that:(1) during triaxial compression, the peak strength and elastic modulus of sandstone initially increase and then decrease with rising burning temperature, while both parameters increase with higher confining pressure. Axial peak strain increases continuously with elevated burning temperature and also rises gradually with increasing confining pressure. Closure stress, crack initiation stress, and yield stress exhibit a slight initial increase followed by a progressive decline as burning temperature increases, with a temperature threshold observed near 200 ℃.(2) The permeability evolution patterns of burnt sandstone are classified into three types: “√”-shaped, “U”-shaped, and “”-shaped. Both confining pressure and burning temperature promote a transition from the “√”-shaped pattern to the “U”-shaped and “”-shaped patterns.(3) Under seepage pressure conditions, acoustic emission ring-down counts remain low before rock failure but exhibit a sharp surge immediately prior to peak failure.(4) Burning temperature has no significant effect on the variation characteristics of acoustic emission ring-down counts, although acoustic emission signals are notably enhanced during the compaction stage from 400 ℃ to 600 ℃, and become intense during the yielding stage after 800 ℃.(5) The triaxial failure modes of sandstone are consistent with shear failure. Both confining pressure and burning temperature exert an inhibitory effect on the permeability of dominant fractures.

[Objective] The precise acquisition of rock mass structural plane parameters is crucial for the stability evaluation of deep underground engineering. Existing manual recognition method for digital borehole images suffer from subjectivity and low efficiency, while image processing-based automatic recognition method still face challenges in accuracy due to the diversity of structural types and insufficient recognition robustness. Therefore, this paper proposes a deep learning-based algorithm for the identification of fractures, veins, and karst, aiming to achieve efficient recognition of structural planes and high-precision parameter extraction. [Methods] In response to the fine segmentation requirements for fractures, veins, and karst, a multi-target semantic segmentation label system is constructed. The SpectDA-ResU-Net segmentation model is designed, integrating spectral gating modules, dynamic channel attention mechanisms, attention gating units, and deep supervision mechanisms, significantly enhancing the accuracy of complex structure recognition. Additionally, an automatic 3D geometric parameter extraction method based on segmentation masks is proposed. [Results] Ablation experiments on an enhanced dataset of 2 148 borehole images show that after integrating all modules, the proposed model achieves an F1-score of 94.46%(an improvement of 4.65%) and an mIoU of 89.59%(an improvement of 8.69%). In comparison with the U-Net model, the mIoU improves by 15.77%, and other evaluation metrics are significantly better than those of existing mainstream segmentation networks. Case studies in the Guizhou karst region show that the automatic extraction error of structural attitudes is controlled within 4%. [Conclusion] The research demonstrates that the proposed algorithm has significant advantages in improving the accuracy of digital borehole image structural plane recognition and the efficiency of parameter extraction.

[Objective] The anti-penetration performance of coral reef limestone is of great significance for underground engineering construction and defensive capacity enhancement on islands and reefs. [Methods] To rapidly evaluate the anti-penetration performance of coral reef limestone and establish a penetration depth calculation formula, coral reef limestone with a sandstone-gravel structure from a specific island reef in the South China Sea was used as the research object. The penetration tests with small-mass arc-shaped pointed kinetic energy projectiles within the velocity range of 500~1 100 m/s were conducted using a self-developed high-velocity impact penetration test system. The failure characteristics of the projectiles and the coral reef limestone targets at different penetration velocities were investigated, and the influence of pore structure characteristics on penetration depth and the tunnel zone was analyzed. A penetration depth calculation model was established based on the test result, and the influence of the projectile size effect on penetration depth was discussed. [Results] The results showed that with increasing impact velocity, the degree of projectile abrasion and nose bluntness significantly increased, and a mass loss rate of 7.2% was observed at an impact velocity of 1 038.96 m/s. Within the tested penetration velocity range, the projectile and target were approximately in a rigid-body penetration state, and higher impact velocities resulted in larger crater volumes. The irregular pore distribution within the coral reef limestone targets caused projectile deflection and tunnel zone bending, Resulting in a reduction in penetration depth. Furthermore, the trajectory deflection in the tunnel zone weakened with increasing penetration velocity. Based on the penetration test data of coral reef limestone, an empirical formula for calculating penetration depth was obtained using dimensional analysis and the Levenberg-Marquardt algorithm fitting. [Conclusion] The penetration depth calculation formula for coral reef limestone can guide the design of protective layer thickness and chamber burial depth, providing theoretical references for ensuring the safety of island and reef engineering.

[Objective] To establish a model for the rapid prediction of the compression coefficient, multiple easily obtainable physical property indicators are used as input features, and an improved stacking model based on principal component analysis(PCA) for dimensionality reduction is proposed. [Methods] Based on 360 sets of actual silty soil data collected from the Binhaiwan New District in Dongguan City, the model was trained and validated. To address the issues of high dimensionality of soil data and potential multicollinearity among features, PCA was first applied to extract the top five principal components with a cumulative variance contribution rate exceeding 95% from the feature data. These top five principal components mainly reflected the comprehensive variation characteristics of physical properties such as water content, natural void ratio, liquid limit, plasticity index, and specific gravity. The constructed stacking model consisted of two learning layers: base learners and a meta-learner. Random forest and XGBoost models were adopted as base learners, while a support vector regression(SVR) model was employed as the meta-learner. Out-of-fold predictions generated via cross-validation were used as input features for the meta-learner layer. [Results] The validation result indicated that the constructed stacking model achieved a coefficient of determination(R²) of 0.790 and a mean squared error(MSE) of 0.037 MPa^-2 on the training set, and an R² of 0.781 and an MSE of 0.033 MPa^-2 on the test set. Compared to traditional machine learning models, the proposed improved stacking model achieved higher prediction accuracy. Specifically, compared to the relatively better-performing random forest model, the R² increased by approximately 11.89%, while the MSE decreased by about 25%. Furthermore, a comparison was made with the model without PCA dimensionality reduction, which achieved an R² of 0.747 and an MSE of 0.048 MPa^-2 on the test set, both worse than the model using PCA for dimensionality reduction. [Conclusion] To enhance the model's generalization under varying data distributions across regions, a calibration method based on median mapping for prediction result is introduced. By comparing the median and variance relationship of the compression coefficient distributions between new and source regions, linear correction is applied to the prediction result, mitigating the impact of distribution drift on the model's prediction accuracy. The result demonstrate that the proposed model exhibits high reliability, generalizability, and application value in the rapid and accurate prediction of the compression coefficient in soft soil.