Using Stochastic Inverse Modeling Method to Obtain Probabilistic Capture Zones of a Spring in a Complex Fracture Aquifer

Generating a series of stochastic models (realizations) by applying stochastic inverse modeling method is sometimes an efficient way to improve hydrogeological cognition accuracy of a site, such as obtaining a more clarity aquifer structure or a probabilistic capture zone of a spring. However, the borehole data size often cannot meet the requirements of stochastic modeling in a general project. Considering geological analysis result and borehole data, it may be a rational and effective method to translate geophysical prospecting (TEM) points into virtual boreholes to solve the data shortage problem. Using the PEST program, stochastic models established through practical boreholes and virtual boreholes can be screened with groundwater level data as the reference. The stratigraphic structure of the filtered models is then checked artificially to guarantee model geological rationality. In this paper, a total of 503 realizations are generated by using a transition probability Markov chain (T-PORGS) based on 74 data points (including virtual boreholes). With data from 9 groundwater observation points within the site as a benchmark, 67 models that effectively describe the hydrogeological characteristics of the site are selected through PEST. Finally, the probabilistic capture zones of the target spring in a fracture aquifer are calculated from these selected models. This modeling process enables stochastic modeling at a site scale even in the absence of sufficient borehole data, providing valuable hydrogeological information for the site.