Dynamic Purging for Groundwater Sampling Informed by a Storage–Mixing Model

Groundwater sampling is a critical component of hydrogeological investigations and is essential for accurate hydrogeochemical analyses. Fully representative samples can be obtained after effective purging of non-representative stagnant water from monitoring wells. However, the procedure and threshold for sufficient well purge remain unresolved. In practice, wells are often purged from 5 to 60 minutes according to the stabilization of chemical-physical parameters or 3–5 well volumes without a rigorous scientific basis, after which samples are collected under the assumption that they represent formation conditions. This introduces substantial uncertainty and potential errors into sampling data.

To address this issue, we develop a well storage-mixing model to characterize the combined effects of two key processes during purging: well storage depletion and wellbore mixing. By modeling the dimensionless completion variables of these two processes, η_q [-] and η_c [-], we demonstrate that sufficient well purge is controlled by the process with the longer characteristic timescale. In high-yield aquifers or large, deep wells, wellbore mixing limits the time required to achieve sufficient purging; conversely, in low-yield aquifers or shallow, small wells, storage depletion is the limiting process. Field data from recent sampling campaigns in Rome, Italy, and Inner Mongolia, China, exhibit mixing-limited and storage-limited purge modes, respectively, indicating that different well geometries and hydrogeological settings lead to distinct purge times and volumes. Accordingly, purge criteria should be dynamic to avoid over-purging (unnecessary capital costs) or insufficient purging (non-representative samples). Dynamic purging informed by the storage-mixing model significantly improves data accuracy while reducing capital costs and should be widely adopted, particularly for monitoring well networks with highly variable well geometries and aquifer conditions.