Improving groundwater treatment decisions using Giardia and total aerobic spores to assess surface water influence

US EPA

US Environmental Protection Agency

GWUDI

Ground Water Under the Direct Influence (of surface water)(US)

GUDI

Ground Water Under the Direct Influence (of Surface Water) (Canada)

IMS

immunomagnetic separation

IFA

immunofluorescent assay

LT2ESWTR

Long Term 2 Enhanced Surface Water Treatment Rule (US)

MPA

microscopic particulate analysis

PWS (or PWSs)

public water system(s)

TAS

total aerobic spores

SFU

spore forming unit

UCMR3

Third Unregulated Contaminant Monitoring Rule (US)

AGI

acute gastrointestinal illness

Ground water regulators are challenged by the need to determine whether drinking water from public water systems (PWSs), originating from a well, spring, or infiltration gallery water production device, yields either aged ground water or ground water that was recently surface water. The distinction is important because required drinking water treatment is more costly and technologically complex for devices producing recent surface water (Berger et al. 2018).

Colford et al. (2009) evaluated highly credible acute gastrointestinal illness (AGI) in Sonoma County from the drinking water and found a statistically significant 12% attributable AGI risk. Based on an exchange of letters between Sonoma County Water Agency and their state regulators (State of California 1993, unpublished), using horizontal collector well and total coliform data, one well was determined to be under the influence of surface water and shut in during the winter months; however, all wells were measured to have other infrequent indicators that make them possibly influenced by surface water (SCWA 2013). Each person in the United States, on average, is estimated to incur 0.60–0.65 AGI cases per year (Roy et al. 2006; Jones et al. 2007). For a Sonoma County population of 600,000 people served by PWS devices, these data suggest that there occur each year 43,000–48,000 AGI cases attributable to drinking water and, in particular, attributable to the drinking water from the five horizontal collector wells that are regulated (all or in part) as ground water rather than as recent surface water. There is a need for improved regulation and/or guidance on identifying recent surface water to reduce these postulated harmful effects to the public (e.g. National Drinking Water Advisory Council [NDWAC] Letter Report to the US EPA Administrator 2023).

Both Canada and the United States have similar concepts for identifying recent surface water. In Canada and the United States, some devices producing recent surface water are identified as producing drinking water that is regulated as Ground Water Under the Direct Influence (GWUDI or GUDI) of surface water (Health Canada 2019). GWUDI was first formally identified in the 1989 US Environmental Protection Agency (US EPA) Surface Water Treatment Rule (US Federal Register 1989) and subsequently adopted by some Canadian provinces, including British Columbia, Alberta, Saskatchewan, Ontario, and Nova Scotia. GWUDI is defined in the US Federal Code, National Primary Drinking Water Regulations (US Federal Register 1989) (S1). Note that the most current US GWUDI requirements include Cryptosporidium in the GWUDI definition (US Federal Register 2006). The GWUDI regulatory definition allows regulatory agencies (i.e., states, tribes, provinces, and First Nations) to establish a scientific decision tree appropriate to their local conditions.

Most regulatory agencies require PWSs to sample ground water using microscopic particulate analysis (MPA) (US EPA 1991; Vasconcelos & Harris 1992) as a key decision assay. The MPA-suggested method is designed to capture surface water, hyporheic zone, and shallow ground water microorganisms. The method counts classes of organisms and assigns a score based on those counts. The score is weighted based on professional judgement that the organism is a good indicator of surface water. For example, nematodes live primarily in shallow ground water, so they do not get a high score when there is a high count. In contrast, diatoms live only in surface water, so whole green diatom presence and counts are scored high (Vasconcelos & Harris 1992).

The MPA method is based on the use of a polypropylene yarn-wound cartridge filter, which does not have an absolute rated pore size and is not designed to capture Cryptosporidium oocysts. The filter is a 1-μm nominal porosity yarn-wound cartridge in a filter housing. US EPA Method 1623 (includes Method 1623.1) (US EPA 2005, 2012) for Giardia and Cryptosporidium requires filters such as EnviroChek HV and Filta-Max, that have an absolute (EnviroChek HV) or nominal (Filta-Max) rated filter. Because the MPA method lacks features essential to US EPA Method 1623, particularly the use of immunomagnetic beads for the capture of cysts and oocysts, the MPA method is not favorable for Giardia or Cryptosporidium recovery. The MPA-suggested method lacks the innovations to separate, tag, and confirm Giardia or Cryptosporidium from all the other debris and bioparticles; Giardia or Cryptosporidium detection in an MPA sample is unlikely. Giardia or Cryptosporidium detection in an MPA sample results in a high MPA score independent of other bioparticle occurrences (Vasconcelos & Harris 1992). Given the rare Giardia cyst and Cryptosporidium oocyst detections in MPA samples and the uncertainty associated with those detections without the use of US EPA Method 1623, there are few available data that confirm in a single sample that MPA indicator organisms (e.g. diatoms), correlate with Giardia and/or Cryptosporidium occurrence in ground water.

The use of immunofluorescent assay (IFA) to aid in the location and identification of cysts and oocysts on microscope slides was included in the Information Collection Rule (ICR) method (US EPA 1995) and subsequently in US EPA Method 1623. The staining procedure originally required considerable time and effort but has been significantly reduced with modifications incorporating pre-conjugated monoclonal antibodies with fluorophores. IFA provides substantial assistance with identifying cysts and oocytes (Abbaszadegan et al. 2011).

Although there is a focus on the MPA assay in most States and Provinces, in Canada and the US, regulators emphasize using all available site and local hydrogeological information as part of the GWUDI determination (Vasconcelos & Harris 1992). Chaudhary et al. (2009) provided a summary of State and Provincial GWUDI determination protocols and similarly recommended more detailed hydrogeologic assessment and reliance on bacteriological sampling with analyses of turbidity, conductivity, and temperature. Abbaszadegan et al. (2011) summarized MPA use in some States and Provinces (e.g. Quebec does not use MPA) and recommended the use of microbial indicators (e.g. diatoms, green algae, fecal coliforms, and total aerobic spore (TAS)) of Cryptosporidium and Giardia for GWUDI determination. TAS is suggested as a Cryptosporidium surrogate microorganism since both species are spherical and have environmentally resistant coatings that protect against inactivation (Bradford et al. 2015; Headd & Bradford 2016). TAS is slightly smaller in diameter than the Cryptosporidium oocyst (i.e. 1 μm vs. 4–6 μm) but their soil abundance, longevity, and low-cost assay make them a useful Cryptosporidium surrogate.

Berger et al. (2018) published approximately 15 years of TAS data from vertical and horizontal collector wells at Casper, WY. Using paired samples from the South Platte River and the adjacent well, Berger et al. calculated log removals by subsurface passage. TAS log removal values inform bioparticle (e.g. Cryptosporidium) removal efficiency during subsurface passage. Subsurface log removal efficiency mimics the performance of a surface water treatment plant. Thus, TAS subsurface log removal studies assess any site-specific need for additional surface water treatment plant particle removal. With significantly fewer samples, TAS log removal can also inform site-specific Cryptosporidium risk by providing snapshot values for the local subsurface removal efficiency.

Although most States and Provinces use MPA as part of their GWUDI determinations (Chaudhary et al. 2009), there are few peer-reviewed scientific papers reporting MPA results at a site. Like most waterborne pathogens, unambiguous Giardia or Cryptosporidium detection in a routine ground water sample using US EPA methods appears infrequent (Chique et al. 2020) (see Stokdyk et al. (2019) for an alternative view). Absent parasitic protozoa occurrence, there are few objective measures of human health risk sufficient to warrant making a positive GWUDI determination, which could result in adding engineered treatment.

This study reports on the analysis of 1,221 samples collected from 2006 to 2020 in 590 ground water collection devices from several Canadian provinces for GWUDI determination and includes assays for parasitic protozoa using US EPA methods (n = 494 samples). In addition to Giardia and Cryptosporidium detections, the dataset also includes MPA scores, MPA individual bioparticle counts (e.g. diatoms, nematodes, and insects), and TAS counts. Each of these parameters could be used in a GWUDI determination; however, diatoms and TAS were selected for GWUDI determination assessment as the best surrogates (e.g. Abbaszadegan et al. 2011). This dataset of 21 Giardia detections in 494 ground water samples is the most robust Giardia dataset using US EPA methods from ground water PWSs available worldwide. These data also have three Cryptosporidium detections, with each co-occurring with Giardia. A Giardia or Cryptosporidium detection using US EPA methods is assumed to be an unambiguous measure of GWUDI given the large score resulting from that detection. This study demonstrates, through analysis of this large dataset, recommendations for improving GWUDI determination and drinking water treatment decisions by exploring (a) TAS as a single predictor of Giardia to augment an MPA score that is largely based on diatoms; (b) integrating IFA with MPA analysis and utilizing Filta-Max or EnviroChek HV filters in MPA analysis for improved detection of Giardia and Cryptosporidium; and (c) paired sampling of surface and ground water samples for TAS log removal estimates (e.g. Berger et al. 2018).

An expanded, in-depth description of the sample collection and methods used to collect the samples used for the dataset is provided in the Supplemental Information (S1, S2, and S3). In brief, samples were collected from 2006 to 2020 from provinces, territories, and First Nations in Canada where the use of ground water as a potable source is common and regulation requires GWUDI determination, namely Alberta, British Columbia, New Brunswick, Nova Scotia, Ontario, Saskatchewan, and Yukon. Given the long sampling timeframe, a variety of methods were used to analyze these samples (a significant research limitation). MPA analysis was performed according to Vasconcelos & Harris (1992) from 2012 to 2015 using yarn-wound filters. From 2016 to 2020, MPA analysis was conducted similarly, but with Filta-Max or EnviroChek HV filters. TAS was analyzed according to Section 9218B of Standard Methods (APHA 2005). US EPA Method 1623 with modified IFA was also conducted for the detection of Giardia and Cryptosporidium and is briefly described in S2. Giardia or Cryptosporidium samples that did not use US EPA Method 1623 were excluded from the statistical analyses. Giardia and Cryptosporidium results filtered with the polypropylene yarn-wound filter were also excluded due to the issues with pore size discussed earlier. Nonetheless, diatom data collected with the yarn-wound filter (Vasconcelos & Harris 1992) were evaluated (S3) for comparison of these data with other data from the scientific literature. Samples were collected using either the Filta-Max (McCuin & Clancy 2003) or EnviroChek HV filters (approved in US EPA Method 1623, 2005). Due to cost constraints, ease of field usage, and well-pressure effects, Filta-Max filters were favored. However, under low-pressure conditions, EnviroChek HV filters were used. Samples per device were variable, ranging from 1 to approximately 50 samples, depending on the requirements of the funding entity and the need for repeat samples over time.

A brief description of the database is provided in the following:

• • Number of MPA method (Vasconcelos & Harris 1992) yarn-wound filter samples from ground water.

• • o 403 devices

  • o 727 ground water samples (on average, about 1.8 samples per device)

• • Number of MPA (or Giardia/Crypto) samples from ground water using US EPA Method 1623 Filta-Max or EnviroChek HV filters:

• • o 205 devices

  • o 494 ground water samples (on average, about 2.4 samples per device)

A summary of the key findings of the dataset is provided in Table 1. Contingency tables of diatoms and Giardia (Table S1) and TAS and Giardia (Table S2) provide a tabular alternative information summary of the results. A brief description of the dataset used for the statistical analysis is:

• 205 devices

• 494 ground water MPA diatom, Giardia, and Cryptosporidium samples; 251/494 with TAS assay

• 21 Giardia detects from 16 devices

• 3 Cryptosporidium detects from three devices

• Co-occurring Giardia and Cryptosporidium in three samples

TAS, diatoms, and Giardia detections in the same sample were compared among all samples (Figure 1) to investigate the utility of TAS for supplementing MPA analysis. Neither TAS (p-value = 0.38) nor diatoms (p-value = 0.27) were significant single predictors of Giardia by sample. Sensitivity for TAS and diatoms was 78 and 24%, respectively, by sample (Figure 1). These results indicate that the presence of TAS was more helpful than diatoms for finding true positives of Giardia in samples. Specificity for TAS and diatoms was 46 and 88%, respectively, by sample (Figure 1), which indicates that the absence of diatoms is a better indicator that a sample will not have Giardia. The ROC curves for Giardia presence/absence using diatoms (Figure S1; AUC = 0.58) and spores (Figure S3; AUC = 0.83) support these findings, and reinforce spores as a helpful surrogate for Giardia detection.

A similar analysis of TAS, diatoms, and Giardia by device rather than by sample was performed (Figure 2). Sensitivity for TAS and diatoms was 90 and 44%, respectively, by device (Figure 2). Specificity for TAS and diatoms was 39 and 87%, respectively, by device (Figure 2). Similar to the above, these results suggest that TAS was a strong indicator of GWUDI, but the absence of diatoms was a strong indicator of a device negative for Giardia since the last sample collection. TAS presence was not a significant factor in a linear model for predicting the absence or presence of Giardia in a specific collection device (p-value = 0.723), which follows from the high number of false positives (Figure 2). Nonetheless, the detection of TAS in a device was a useful predictor of Giardia detection occurring at the device (Figure S4; AUC = 0.77).

Diatom presence/absence was a significant factor in a linear model for predicting the presence or absence of Giardia in a specific collection device (p-value = 0.022). The significance of diatoms in the linear model is likely driven by the high proportion of true negatives to false positives. Whole green diatoms have been previously suggested to be the most unambiguous MPA indicator (US EPA 2010; US EPA 2016). Given the low true-positive to false-negative ratio (5:16), the model for predicting a Giardia detection in a collection device, based on at least one sample indicating diatom presence, displayed close to random accuracy (Figure S2; AUC = 0.63). These results reinforce the low sensitivity from using diatoms as an indicator (i.e. a limited predictor of Giardia).

In the drinking water regulations, overly protective actions in support of public health are generally favored (i.e., sensitivity is considered more important than specificity). The presence of TAS at a device displayed high sensitivity but low specificity. Despite the low sample size and low specificity, these findings support those of Berger et al. (2018), which demonstrate a benefit to using TAS log removal in assessing Cryptosporidium risk for alternative treatment by subsurface passage and for site-specific Cryptosporidium risk, even with very limited sample numbers. However, it is emphasized that Berger et al. (2018) only sampled sand aquifers (or sand with gravel in small proportions). GWUDI aquifers are very heterogeneous and include karst limestone, fractured bedrock, and coarse gravel aquifers, as well as any aquifer with limited soil cover. Over 15 years, TAS log removal was demonstrated to be informative in sand aquifers (Berger et al. 2018), whereas only TAS count, but perhaps not log removal, is likely useful in all aquifer types. Sampling variability, differences in environmental conditions, and temporal variability may partially explain the false positives and negatives associated with the utility of diatoms or spores for predicting Giardia. Future research into this topic could be improved by careful selection of similar hydrogeological sites for sampling and the performance of sampling at more regular intervals (e.g. one wet season and one dry season sample per well).

Limited bank filtration TAS data were used in the past to suggest that there may be a background TAS level in shallow ground water (US EPA 2010). Berger et al. (2018) analyzed TAS data without censoring for a putative background limit. These data shows that Giardia detection can occur at TAS counts below 10 SFU/100 mL (the putative background TAS count), which indicates that TAS data should not be censored at 10 SFU/100 mL.

Given the promising results from the TAS co-occurrence analysis, a TAS count threshold (>10 vs. >100 SFU/100 mL) was investigated as a decision metric and is shown in Table 2. For the 116 devices with TAS results and the 10 devices that also had Giardia detections, the detection of higher TAS counts in a device slightly favors Giardia detection (Table 2). The bin with the highest percentage of Giardia detections occurs in those devices with a high maximum TAS count (>100 SFU/100 mL). These limited data suggest that an amendment to MPA would include utilizing an increased MPA score for TAS >100 SFU/mL, although other threshold values may also be suitable. Table 2 shows an apparent increased likelihood of Giardia detection as the maximum TAS count increases, with up to 40% of devices with Giardia detections for TAS count >100 SFU/mL. This relationship may alternatively suggest a non-linear relationship scoring guideline for TAS when amending MPA (e.g. a step function using TAS count and MPA score).

Production devices that have high MPA results (MPA score >15–19) in one or more samples are typically adjudicated by the State or Province as GWUDI. There are instances where a decision is challenged, and more data are collected to refine the determination. Although there are reports suggesting high uncertainty associated with an MPA assay (Jacangelo & Seith 2001), the MPA method is a generally accepted tool for use by States or Provinces to make decisions, in conjunction with other easily identifiable information, such as well – surface water setback distance. The inherent logic is that high diatom counts are indicators of recent surface water because only intact diatoms with green chlorophyll are counted. High diatom counts lead to high MPA scores. Thus, a well-producing water with whole, green diatoms as well as other indicator organisms is more likely to be regulated as GWUDI, and that decision is usually accepted. The decision becomes more fraught when organisms that spend part of their life cycle in ground water and part in surface water are counted. Professional judgment circa 1992 was used to establish scores for counts of such dual lifestyle organisms. During the past 30 years, hyporheic zone science was established as a scientific discipline. The hyporheic zone is now a recognized ecotone but there has been little or no overlap between the science of the hyporheic zone and GWUDI determination. No hyporheic zone science has been utilized to update the MPA assay.

Given the extensive MPA use by States and Provinces, it appears that MPA appropriately identifies the most at-risk production devices in relatively simple hydrogeological settings. Typically, the more easily identified GWUDI devices are bank filtration vertical wells in alluvial aquifers (Borchardt et al. 2004). Bank filtration is an established drinking water treatment technique, especially widely used in Europe but also used in Canada and the US, usually as part of a treatment train rather than as a stand-alone process. Except for horizontal collector wells, ground water flow models of alluvial aquifers usually adequately describe the bank filtration flow field, the flow path to the vertical well, and the time of travel. Horizontal collector wells, springs, and infiltration galleries typically have greater flow field complexity. For example, induced infiltration to a vertical well in an alluvial aquifer is typically horizontal flow through the subsurface. In contrast, for a radial collector horizontal well, induced infiltration to a well lateral (always sited in alluvial aquifers) is typically vertical flow through the river alluvium. This contrast in flow field direction (and travel time) could affect MPA organism recovery (Wang et al. 2022).

Finally, we briefly consider the merits of various TAS concentration thresholds, given our data and previous work (US EPA 2010; US EPA 2017; Berger et al. 2018). Table 2 illustrates our putative thresholds (>0, >10, >100 SFU/100 mL). We suggest that the highest threshold (>100 SFU/100 mL) would likely perform poorly where soil cover is patchy and soil bacteria occurrence may be limited. Hydrogeologic settings with such limited soil cover include fractured bedrock in glaciated terrane or in karst topography, dominated by sinkholes that permit direct surface water recharge into the subsurface. Assuming a universally applicable TAS concentration threshold, we suggest that a high threshold would fail to identify some higher-risk devices in these settings. On the other hand, we suggest that a threshold below 10 SFU/100 mL would likely capture too many lower-risk devices. It seems reasonable to conclude that a universal TAS concentration threshold suggesting surface water influence should be somewhere in between those two values.

Investigation into including modified IFA and different filters was performed to assess improved guidance for GWUDI high-risk devices. Table 3 summarizes different methods and techniques for analyzing samples from devices and the proportion classified as high risk. Using the yarn filter for MPA, only 3 of the 403 devices were classified as high risk (Table 3). Analyzing a device sample for Cryptosporidium or Giardia by modified IFA slightly increased the proportion of high-risk devices to 2% (Table 3). These findings suggest a marginal benefit from a sizable effort.

A comprehensive comparison of filter usage for diatom detection from this dataset with that from Abbaszadegan et al. (2011) is described in the Supplemental Text (S3 and S5). In brief, these results supported the findings from Abbaszadegan et al. (2011) that using Filta-Max or EnviroChek HV filters over yarn-wound filters leads to higher diatom recovery. In comparison to only adding modified IFA, using the Filta-Max filter greatly increases the percentage of high-risk devices to 2.4% (Table 3). For a simple change in materials used in the analysis, a significant increase in protective measures was gained. Using both Filta-Max filters and analyzing samples by IFA results in the largest increase in the proportion of high-risk devices (9.3%) (Table 3).

TAS log removal is identified in US EPA guidance (US EPA 2010) as an appropriate method for use in a demonstration of performance for an alternative treatment method under the Long Term 2 Enhanced Surface Water Treatment Rule. With regulatory approval, PWSs regulated as GWUDI may conduct a Demonstration of Performance to show that they are achieving, by bank filtration (in sand or sand and gravel aquifers) as an alternative treatment, the required Cryptosporidium log removal. TAS log removal was evaluated using paired surface water and ground water TAS samples and compared to Giardia and diatom detection (Table 4). Assuming that the Canadian sites are all located in alluvial sand (or sand and gravel) aquifers, the data suggests that lower TAS log removal was associated with higher Giardia detections. For example, as TAS log removal increases from log 2–3 to more than log 3, the Giardia detection percentage decreases from 4.2 to 0%; however, the diatom detection percentage increases from 8.3 to 12.5%. These findings agree with Section 3.1 that diatoms are less sensitive as an indicator of Giardia compared to TAS. TAS log removal of 1–2 results in a comparatively higher proportion of Giardia and diatoms, but this trend is not held when observing TAS log removal of <1 (Table 4). The uneven pattern in the trend suggests (1) that evaluation of TAS log removal may not be a significantly beneficial means for GWUDI determination or (2) that the site hydrogeology was variable, as compared with locations primarily in sand aquifers. These findings may be limited by low sample size and variable site hydrogeology. Nonetheless, these data may also have value to State or Provincial indirect potable reuse regulation (e.g. Eslamian 2016; Oudega et al. 2021; Rauch-Williams et al. 2023) where treated surface water or wastewater is injected or infiltrated into the subsurface. Long-lived, environmentally resistant surrogates such as TAS are more likely to be detected in recovered, reusable water. Log removal calculations can be used to inform the public health risk for re-using the treated injectate or infiltrate.

An expansive, robust dataset of 1,221 samples from 590 Canadian ground water well devices was used to assess improvements to GWUDI determination. Samples were analyzed using a variety of filters for MPA and other methods (i.e. IFA, TAS analysis). There were 21 Giardia detections and 3 Cryptosporidium detections in 16 devices. Most sites that were scored by MPA as high risk occurred because a Giardia and/or Cryptosporidium detection results in a high score. Diatoms still presented as the most practical indicator of Giardia; however, TAS was shown to be more sensitive. This result suggested amending MPA to include TAS.

A non-linear relationship between TAS and Giardia may be beneficial when used in any amendment based on the increasing detection rate of Giardia with high TAS counts (e.g. proportionally more devices with Giardia detections at >100 SFU/100 mL compared to >10 SFU/100 mL).

Using the Filta-Max or EnviroChek HV filters and including IFA results in improved recovery compared to the MPA-suggested method and increases the proportion of high-risk devices. The inclusion of TAS analysis and using a high threshold (>100 SFU/100 mL) further increases the proportion of affected devices, which demonstrates the most conservative approach from this study for GWUDI determination in the protection of public health. TAS log removal analysis using paired surface and ground water samples was minimally beneficial in GWUDI determination; however, a more extensive analysis (including site hydrogeology) may offer improved insights into the utility of this analysis.

Overall, this work indicates the importance of utilizing microbial indicators of GWUDI for determination and supports the recommendations of National Drinking Water Advisory Committee (NDWAC) regarding GWUDI (NDWAC 2023). A thorough investigation into the most beneficial indicators of GWUDI should be conducted in the future. Through these studies, regulators can be empowered with the information needed to develop clear and concise means for GWUDI determination and reduce AGI caused by misclassified water sources.

We thank our reviewers for their helpful comments. Mention of trade names or commercial products does not constitute endorsement or recommendation by the US EPA for use. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.