The groundwater monitoring network (GMN) design balancing prediction accuracy with cost minimization has paramount importance in depth to groundwater (DTW) monitoring, modeling, and management. This study demonstrates a robust and transferable GMN design framework for the underexplored statistically homogenous DTW to facilitate appropriate, long-term, cost-effective, and regional-scale groundwater monitoring. The inherent homogeneity in regional scale DTW data was investigated to cluster the observation wells into 18 strongly structured clusters/strata using optimal cluster number (k = 18), corresponding to the highest mean silhouette score of 0.72 and the K-means clustering algorithm. The clustered observation wells were used to evaluate the cluster random sampling (CRS) and stratified random sampling (SRS) for GMN design. The sampling techniques were evaluated to capture the spatial variability of DTW over the whole study region through spatial modeling at an adequate accuracy by the inverse distance weighting interpolation tools. The performance metrics for SRS recommended it as a more robust and appropriate technique for regional-scale GMN design than CRS. Then, the GMN was designed with 540 wells corresponding to 70% sampling under SRS for accurate, cost-effective, and long-term groundwater monitoring. By reducing 30% of observation wells, the cost-saving can contribute to cutting-edge monitoring infrastructure development. Additionally, the need-based balancing of the accuracy in spatial modeling and monitoring cost may further moderate the number of observation wells using the performance metrics of SRS corresponding to sampling percentages. Conclusively, this novel and transferable framework can be adopted globally with internally homogenous DTW data for effective GMN design.
Read full abstract