Impact of long-term mining activity on groundwater dynamics in a mining district in Xinjiang coal Mine Base, Northwest China: insight from geochemical fingerprint and machine learning.

Abstract

Long-term coal mining could lead to a serious of geo-environmental problems. However, less comprehensive identification of factors controlling the groundwater dynamics were involved in previous studies. This study focused on 68 groundwater samples collected before and after mining activities, Self-Organizing Maps (SOM) combining with Principal Component Analysis (PCA) derived that the groundwater samples were classified into five clusters. Clusters 1-5 (C1-C5) represented the groundwater quality affected by different hydrochemical processes, mainly including mineral (carbonate and evaporite) dissolution and cation exchange, which were controlled by the hydrochemical environment at different stages of mining activities. Combining with the time-series data, the Extreme Gradient Boosting Decision Trees (XGBoost) derived that the mine water inflow (feature relative importance of 40.0%) and unit goaf area (feature relative importance of 29.2%) were dominant factors affecting the confined groundwater level, but had less or lagged impact on phreatic groundwater level. This was closely related to the height of the water flow fractured zone and hydraulic connection between aquifers. The results of this study on the coupled evolution of groundwater dynamics could enhance our understanding of the effects of mining on aquifer systems and contribute to the prevention of water hazards in the coalfields.