Human health risk from consumption of groundwater is widely documented and particularly challenging to address in private wells, where testing is not required and is infrequent. Furthermore, the common approach of assessing health risk based on whether individual contaminants exceed a health threshold does not account for how close a concentration is to the threshold nor for cumulative effects across contaminants. Assessing cumulative human health risk from drinking water is relatively new and has primarily been conducted on datasets collected from discrete sampling campaigns where all data produced has a common set of analytes and similar detection limits. These sample campaigns are cost prohibitive for many communities and more efficient approaches for conducting tier 1 (screening) level human health risks are needed.

In this work, we leveraged a publicly available database for Montana groundwater and adapted methods developed by USGS to conduct a statewide cumulative human health risk assessment across 19 inorganic contaminants. This type of analysis requires decisions about which thresholds to apply, which data is most relevant to include, and what minimum data availability is considered sufficient. Sensitivity of results to each of these decisions was assessed and results for many alternative analysis scenarios are provided so users can assess what scenarios might be best suited to their assessment needs. Also included is code/output for histograms of contaminant concentrations and detection limit for non-detect concentrations. These histograms were important for identifying outliers from errant data and for informing what detection limits were considered adequately low for non-detect data to be included in the analysis. Histograms revealed that concentration data for some analytes are normally distributed, which could allow for exploration of alternative methods for handling non-detect data, such as the NADA Package in R Statistical Software. The NADA package was not feasible in our analysis due to non-detect concentrations outnumbering detection data for 7 out of 19 analytes. For datasets with a lower frequency of non-detect data, users could re-examine potential for use of NADA to numerically represent non-detect concentrations for this kind of analysis.

For users specifically working with the Montana Bureau of Mines and Geology, Groundwater Information Center database, the code provided here can be used to compile data and create metadata fields (detection limit, qualifiers, non-detect, etc.) from the somewhat cumbersome single field the database uses to store numeric results and metadata.

This data resource includes all data, code, and analysis products for the accompanying manuscript so that users can easily assess, apply, or adapt these methods for other datasets and applications.