Enhancing the understanding of hydrological responses induced by ecological water replenishment using improved machine learning models: A case study in Yongding River.

The ecological water replenishment (EWR) of Yongding River has been an important project implemented in response to the Development of an Ecological Civilization policy in China since 2016. A reasonable amount of EWR requires a systematic understanding of the relationship among the surface water, groundwater, ecology and economy. However, studying surface water-groundwater interactions still remains an important issue. Thus, a coupled model integrating a Muskingum method-based open channel flow model and machine learning-based groundwater model is developed to describe the dynamic changes in streamflow and groundwater level in response to the EWR of Yongding River. The model is calibrated using observed streamflow data as well as groundwater level data on a daily scale for the spring EWR in 2020. The simulated results match well with the observed data and suggest that significant groundwater level increases occur only around the main channel of Yongding River. Fifteen scenarios under different EWR schemes are set to obtain reasonable streamflow during EWR, and then the responses of streamflow and groundwater level changes are simulated. Reasonable streamflow at the Guanting Reservoir need to be above 65 m3/s to ensure the streamflow can pass through Beijing and significant groundwater level recoveries of 170 million m3 through EWR. The developed models can improve the understanding of the interaction between surface water and groundwater and provide a quick assessment of the factors influencing the different EWR schemes and thus aid in effective EWR project management.

Estimating dewatering in an underground mine by using a 3D finite element model.

Groundwater inflow to an underground mine will seriously affect its mining plan and engineering geology safety. Groundwater models are powerful tools commonly used in the mines to develop dewatering strategies. Many mines in the Kolwezi area have been present since the 1950s, and groundwater flow patterns have been significantly influenced by mining activities. A mining plan is developed for an underground mine with overturned syncline strata in Kolwezi, Congo. Previous groundwater models using layered homogeneous media lowered model accuracies. A new three-dimensional groundwater model using FEFLOW, consisting of a combined regionally and locally geology models integrating 16 hydrogeological cross-sections and borehole logging data, are formulated to predict the underground dewatering in the study area. A 31-days pumping tests with 3 pumping wells and 28 observation wells are carried out to estimate the hydrogeological properties. The simulated water level data match the observed data rather well. Under 8 scenarios of possible well designs, the model predicts a possible dewatering capacity greater 23,900 m3/d at the initial stage of mining. The concept of the model and its application can be a reference for other mines with complex geology for mining safety in the region of interest.