In coastal area of China， there are many aquifers widely distributed in the soil layer， which need dewatering in the process of excavation construction. If it is easy for water to leak between different aquifers， and difficult to fully block the hydraulic connection between inside and outside the excavation， it will cause a large change of water level outside the excavation in the process of dewatering， leading to a series of environmental problems. Therefore， the deformation pattern and control method caused by the dewatering in multi-aquifer leaking excavation are worthy of further study. Based on the dewatering test results of the excavation of a subway station in Tianjin， this paper adopts the finite element method to study the influence of silt lens and leaking aquifers on the excavation deformation in the toe soil layer of the diaphragm wall induced by dewatering. Moreover， the mechanism of influence of silt lens on dewatering induced deformation and the comparative study of deformation induced by different dewatering schemes are also conducted. The results show that hydraulic connection between the upper and lower aquifers at the depth of the toe of the diaphragm wall can reduce the deformation of the diaphragm wall while increasing the settlement outside the excavation. The mechanism behind this is that the hydraulic connection reduces the pore pressure difference between the sides of the diaphragm wall， but increases the drawdown outside the excavation. In this case， it is recommended to adopt the “deep and shallow well” dewatering system with independent filters for the phreatic and confined aquifers， along with timely installation of horizontal supports. Dewatering should be implemented as needed to control the deformation induced by the dewatering process.

In response to the demand for termite prevention in hydraulic engineering， this study constructed a predictive model for termite habitat suitability in Zhejiang Province by integrating the MaxEnt ecological niche model with GIS spatial analysis technology， based on termite census data from Zhejiang Province’s hydraulic engineering projects （2023-2024）. The research systematically revealed the spatial distribution patterns and environmental driving mechanisms of termite habitats in Zhejiang Province. Through a dynamic screening process of “preliminary selection-contribution analysis-collinearity test-iterative optimization”， 22 initial environmental factors were refined to 14， achieving a model AUC value exceeding 0.963. The Jackknife test identified soil pH （6.0-6.4）， annual precipitation （1280-1450 mm） and elevation （80-200 m） as the core environmental drivers most suitable for termite survival in hydraulic engineering. Spatial simulations indicated that the total area of termite-suitable habitats in Zhejiang Province is 40，300 km²， with high-suitability zones （6 400 km²） concentrated in the hilly valleys of central and southern Zhejiang， medium-suitability areas （13，700 km²） distributed in patches in central and northern Zhejiang， and low-suitability zones （20，200 km²） covering all river basins in the province with potential diffusion risks. The research not only provides a reusable technical paradigm for biological disaster risk assessment in hydraulic engineering but also offers important references for formulating precise "one-reservoir-one-policy" prevention systems， guiding the layout of monitoring stations， prioritizing prevention efforts， and developing ecological barrier technologies.

Using portable CO₂ detectors to locate tunnel entrances and nests of the subterranean termite Odontotermes formosanus （Shiraki） in agricultural， forestry， and water conservancy facilities facilitates rapid nest localization and enables scientific control measures.Using a special portable carbon dioxide detection device， the detection area was set up in the serious occurrence plot of O. formosanus. The device was used to detect the ground carbon dioxide concentration， and the concentration anomaly point was excavated to verify the nests and nests entrance of O. formosanus. In addition， the carbon dioxide gas source point at the entrance of the simulated termite nest was set up to determine the concentration diffusion range of carbon dioxide under different wind speeds and wind directions. The device used in the environment with atmospheric wind speed of 0~0.12 m/s can quickly detect abnormal points of carbon dioxide concentration （ termite nest entrance， decayed dead wood or leaves ）. After excavation， it is confirmed that the carbon dioxide concentration at the entrance of the ant road is above 0.10%. In simulations of the carbon dioxide gas source， at wind speeds of 0， 0.1， 0.2， 0.4， 0.8， 1.6， and 3.2 m/s， the downwind distance from the source point to where the CO₂ concentration reaches 0.05% is 0.4， 1.0， 1.0， 0.6， 0.4， 0.2， and 0 m， respectively. The distance from the site with a concentration of more than 0.05% in the crosswind direction to the gas source point is 0.4， 0.4， 0.2， 0.2， 0 and 0 m， respectively. The distance between the abnormal site of carbon dioxide concentration and the gas source point is significantly affected by wind direction and wind speed. The portable carbon dioxide detection device can easily detect the abnormal value of carbon dioxide concentration at the entrance of the termite nest in the termite occurrence site. Combined with wind speed， wind direction， measuring point distance， etc.， the position of the O. formosanus nests entrance can be quickly detected after eliminating interference factors.

Soil-dwelling termites that excavate tunnels and construct nests within embankments represent one of the most critical latent threats to hydraulic engineering. Consequently，accurate and efficient localization of termite nests is essential for effective prevention and control. This study，grounded in the biological behavioral patterns of termites，presents a novel localization approach that integrates deep learning-based object detection with probabilistic field modeling. First，an enhanced YOLOv5 architecture is developed to automatically identify termite castes，including soldiers，workers，and nymphs. Second，an improved multi-object tracking framework based on DeepSORT is employed to generate counting zones for each caste，quantify their populations，and determine aggregated dominant movement directions. Finally，leveraging the characteristic foraging distances and activity distributions of the three castes，a behavioral-habit-based probabilistic model is constructed. This model incorporates dynamic distance and direction weighting functions to generate a two-dimensional probabilistic field centered on each bait-recognition point. The feasibility of the proposed method is validated through experimental comparisons between predicted and actual nest locations. Furthermore，by applying global probability extension and likelihood fusion，the framework supports multi-point matrix-based cooperative localization. Integrating computer vision，ecological behavior analysis，and probabilistic modeling，the method provides a low-cost，high-efficiency alternative to traditional techniques，with scalability for large-scale data analysis and compatibility with IoT-based remote monitoring systems，thereby substantially enhancing the efficiency of termite nest localization.