Automated Groundwater Monitoring in Twin Cities Aquifers Shows Anthropogenic Changes in Water Quality Measures Open Access

In states with extensive groundwater resources, like Minnesota, monitoring and management are critical for maintaining water quality and quantity. Groundwater provides the baseflow of many river systems and supports various ecosystems. It is also a major water source for human consumption, agriculture, and industry, but these uses can strain aquifers through growing populations, over-pumping, and anthropogenic pollution. Existing monitoring efforts are robust, but a more automated monitoring network would provide timely information about parameters like dissolved constituents and temperature, along with hydraulic head, and give resource managers new tools to help ensure that these resources continue to meet ecological and societal needs into the future.

Traditionally, groundwater monitoring in Minnesota has been performed by employing geotechnicians to physically visit some number of wells per year, collect water samples and field measurements, and send those samples to a lab for geochemical analyses. This approach provides comprehensive, high-quality data but only from a small fraction of existing wells. The process is people intensive and does not scale easily to the thousands of wells spread across large geographies and multiple aquifers. As a result, an individual well is likely sampled only once per calendar year at best, obscuring long-term trends and hiding sub-annual variations in water chemistry. The goal of this contribution is to demonstrate that real-time, continuous monitoring of a few simple parameters can shed light on anthropogenic impacts on groundwater and better assist monitoring agencies as they allocate their limited resources to the fullest effect.

This work builds on previous research using historical data to understand the trends in chloride concentration ([Cl⁻]) in Twin Cities Metropolitan Area (TCMA) groundwater (McDaris et al., 2022) and the relationships between [Cl⁻] and specific conductance (SC) in several TCMA aquifers (McDaris et al., 2023). By installing automated sensors in observation wells, we have gathered dense time series data on water depth, temperature, and SC. These high-frequency data allow us to explore (1) the continued infiltration of chloride into the aquifer system of the TCMA, (2) the effects of localized pumping on groundwater chemistry in a heavily modified urban setting, and (3) increasing groundwater temperatures in the broader TCMA. These results argue strongly for the implementation of a sentinel network of simple sensors installed in wells of various depths across the TCMA to generate new data to aid in safeguarding the region’s critical groundwater resources.

The larger TCMA basin is composed of numerous Paleozoic sedimentary bedrock layers overlain by Quaternary deposits of varying thicknesses and compositions. There are several major bedrock aquifers in the region, primarily composed of sandstones and carbonates with extensive secondary porosity (Runkel et al., 2003, 2006).

This work focuses on the area surrounding the East Bank campus of the University of Minnesota–Twin Cities in the heart of the urban TCMA. Geologically, the area is dominated by the Mississippi River, which flows through the middle of campus (Figure 1). The local stratigraphy includes Quaternary deposits ranging in thickness from 8 to 11 m (NRRI, 2024) and the top bedrock unit is the Platteville Formation (OPVL; ∼10 m) with outliers of Decorah Shale. The Platteville is immediately underlain by the Glenwood Shale (1 m), which acts as a leaky aquitard. Underlying these units is the well-sorted and weakly cemented St. Peter Sandstone, which forms the base of the Mississippi River and its shorelines.

The river is also the key hydrological feature in the area (Figure 2). On the east bank of the river, groundwater flows generally NE-SW toward the river, with some azimuthal variation between aquifer units. The Quaternary Water Table Aquifer (QWTA) occurs within the unconsolidated Quaternary sand and gravel above the uppermost bedrock, with less than 3 m of confining material (GWMAP, 1999). This unit is directly connected to lakes and streams at the surface and is heavily impacted by anthropogenic pollution (GWMAP, 1999; Berg, 2021; and McDaris et al., 2022). The OPVL is best characterized as a hybrid hydrogeological unit, having characteristics of both an aquitard and an aquifer, depending on the context (Anderson et al., 2011; Runkel et al., 2018). The limestone and dolostone of the OPVL has low matrix permeability (10⁻⁵ m/d or less), which can greatly impede vertical flow in less eroded areas without throughgoing vertical fractures. However, well-connected horizontal fractures and dissolution features create significant secondary porosity with permeabilities up to 10³ m/d or more. These horizontal features are prominent sources of springs where they intersect the ground surface throughout the region (Brick, 1997).

The study of surface waters has benefited from technological advances in real-time, continuous monitoring that enable the detection of various high-frequency behaviors, including spatial variability in the biogeochemistry of streams and lakes (e.g., Crawford and Lee, 2015), episodic weather-related events in lakes (e.g., Jennings et al., 2012), hydrochemical monitoring of urbanized river systems (e.g., Halliday et al., 2015; Yang and Moyer, 2020), periodic fluxes of road de-icing salts into urban streams (e.g., Corsi et al., 2010; Moore et al., 2020), and the relationship between land use and surface water quality (e.g., Zampella et al., 2007). By contrast, groundwater studies have benefited comparatively less from advances in continuously automated measurements. Some highlights include the development of full water quality assessments from field-based sensors (e.g., Granato and Smith, 1999; MacDonald et al., 2017; and Saraceno et al., 2018), the use of real-time SC data to study seawater intrusion into coastal aquifers (e.g., Lee et al., 2007; Roehl et al., 2013), the exploration of increasing groundwater salinity in agricultural areas (e.g., Deverel and Fujii, 1988; Hamilton and Helsel, 1995; and Abanyeh et al., 2005), and the investigation of groundwater–surface water interactions (e.g., Chen et al., 2006; Fernald and Guldan, 2006; and Cox et al., 2007). In Minnesota, the Minnesota Pollution Control Agency (MPCA) is experimenting with continuous (although not real-time) monitoring using SC in a small number of wells (Kroening, 2022).

This study builds on historical data collected by state agencies that capture long-term chloride trends in TCMA aquifers over the last 50 years (McDaris et al., 2022). This historical data also demonstrate a relationship between [Cl⁻] and SC in various aquifers, as well as the impact of local hydrogeological conditions on those relationships (McDaris et al., 2023). Nearly all the region’s aquifers show some amount of anthropogenic chloride infiltration, although most measurements remain below the state’s chronic environmental water quality threshold of 230 mg/L. [Cl⁻] is highest in Quaternary deposits and tends to be higher in areas with more impervious surfaces, which tend to focus runoff during rainfall and snowmelt events. Bedrock aquifers overlain by other bedrock units or significant amounts of Quaternary clays tend to have lower [Cl⁻]. We also observed an empirical relationship between the average concentration of chloride in an aquifer and the rate at which [Cl⁻] is increasing. This relation predicts that nearly 45 percent of recently monitored wells will exceed 230 mg/L of chloride by 2050. Furthermore, some aquifers yield a robust correlation between [Cl⁻] and SC using all available data such that SC can be used as a direct chloride proxy. In units where this relation is sufficiently strong (particularly QWTA wells) measurements of SC can be used to accurately (91.5 percent) characterize water samples as below, between, or above the Minnesota state environmental water quality thresholds of 230 and 860 mg/L of chloride.

For this study, sensors were installed in six observation wells on the University of Minnesota–Twin Cities campus (Figure 1). Most data originate from sensors installed in two wells adjacent to Williamson Hall (denoted with the yellow star in Figure 1) in May 2021. These two wells are laterally separated by only 2 m, but they are screened in QWTA and OPVL, respectively (Figure 2, inset). These wells were selected because of their proximity to each other, ease of site access on campus, and opportunity to study groundwater variability in successive aquifers separated by an aquitard (the Decorah Shale). They have also been used for hydrogeology teaching and research purposes in the past, which provided some historical data for comparison. Finally, there is a weather station less than 2 km away maintained by Hennepin County, for which hourly rainfall data are available for comparison.

The Williamson Hall installation consists of a METER Hydros 21 (MH21) sensor (METER Environment, Pullman, WA, 2025) in each well (measuring SC, temperature, and water depth), and a waterproof case with a pressure vent at the surface containing a Bosch BME 280 sensor (Adafruit, Brooklyn NY) for atmospheric pressure correction at the surface), battery backup power (Adafruit, Brooklyn NY), a cellular modem (EnviroDIY LTE Bee, Avondale, PA) and antenna (Taoglas Maximus FXUB66, Wexford, Ireland), and an EnviroDIY Mayfly data logger (Avondale, PA) that serves both wells (EnviroDIY, 2024a). Data from the sensors are stored on a local SD card and are telemetered to the Monitor My Watershed website (MMW, 2024), where they are publicly available in real time for download and analysis (McDaris et al., 2024). Data were initially collected and transmitted every 5 minutes, but this was revised to bihourly after a year.

The Mayfly is an Arduino IDE–compatible data logger programmed using the open-source Modular Sensors library (EnviroDIY, 2024b). The code base is well documented by the user community and has been customized for use with various environmental sensors from different manufacturers. The Mayfly is also an inexpensive data logger option (roughly US$300 for all parts needed to assemble the logger). At the time of writing, a complete installation of the sensor and data logger costs approximately US$1,000. Programming the Mayfly requires coding in Arduino, but the EnviroDIY community has extensive training resources available.

The power needs of this setup are modest, and similar monitoring efforts elsewhere on campus have been fully powered with no issues by small (∼200 cm²) solar panels. Under the typical Modular Sensors program framework, the sensors and data logger are inactive most of the time. A measurement interval is specified in the code (with a reliable maximum measurement rate of every 5 minutes, constrained by sensor and modem wake-up times), at which time the instrumentation turns on, takes a set number of measurements from each sensor and averages them (six measurements for this work), transmits that averaged value for each sensor to MMW, and then goes to sleep. For this study, we made use of hardwired building power in Williamson Hall via a USB adapter for the Mayfly. During occasional power failures to that outlet, the small onboard battery backup supported continued data collection and transmission for 1–2 days before power failed completely. Additional battery capacity could extend this period significantly.

During this study, geochemistry samples were collected at the time of installation and roughly quarterly thereafter and were characterized by the Research Analytical Laboratory at the University of Minnesota. Analyses included ion chromatography and inductively coupled plasma mass spectrometry to ascertain the concentrations of 34 common dissolved constituents (7 anions and 27 cations). Water chemistry data for all samples used in this work can be found in McDaris et al. (2024).

The Williamson Hall sensors were placed in each well at specific depths: 13.1 and 18.0 m below grade in the QWTA and OPVL wells, respectively. This positioning placed the QWTA sensor below the lowest previously measured water level and put the OPVL sensor in a zone of ambient downflow that emanates from a fracture higher in the open hole of the well. The MH21 sensors are held in place by integrated hooks attached to the sensor cables. We found that these hangers can lose compression over long deployments, allowing the sensors to slip out of position by as much as 2 cm, particularly during the collection of water samples for lab analysis. Affixing a zip tie around the cable by the hanger eliminated the slipping, and errant measurements were removed from the dataset.

Some simple data quality measures were used on the dataset prior to analysis. These included removing discontinuous data points associated with the extraction of sensors from wells for water sampling and then running R’s base routine for removing outliers from the SC, temperature, and water depth measurements to avoid statistically errant readings.

Although MMW provides some native time series analysis tools, computation and analysis for this work was conducted in RStudio (R Core Team, 2023). Data from all sensors was downloaded to perform a standard set of analyses on all data. These analyses included linear regressions of the relationships between [Cl⁻] and SC, calculations of simple moving averages plotted with the data cloud, and SC-derived and lab-measured [Cl⁻] comparisons.

In addition, temperature data from the sensors and from a nearby weather station were compared to assess the degree of lag between the surface and each aquifer unit. For each set, the temperature data were first de-trended and centered about zero by finding the linear trend and subtracting it at each point. The resulting data were then normalized by dividing by the maximum value to yield temperature data running between −1 and +1. Then, the time axis was rescaled such that 1 year was equal to 2π. Finally, cosine curves were fit to the data to ascertain the delay in the expression of surface temperatures at depth.

To demonstrate the ability of in situ sensors to resolve real-time aquifer conditions, Figure 3 illustrates measurements taken in the Quaternary well during October 2023. Just before and during that month, there were multiple periods of substantial rainfall (September 29–30, October 12–13, and October 23–27) recorded near the well head. These events show up clearly in the water elevation plot (Figure 3b) in the form of significant increases in water levels in the well within half a day of the rainfall. Groundwater temperature (Figure 3a) does not seem to be immediately impacted by the rainfall events, and we will show that the temperature of this aquifer lags surface air temperature by 6 months. SC (Figure 3c) appears to also be uncorrelated with rainfall. However, the calculated standard deviation of the SC data (Figure 3d) shows a positive response to each rain event even faster than water elevation. The cause of this quick response is uncertain. Regardless, it is clear that the sensors are capable of resolving short-term groundwater changes with these few measurements. The following subsections illustrate cases in which the use of such automated measurements is shown to provide groundwater professionals the information they need to understand, diagnose, and address water quality issues.

We see that SC captures the broad contours of the lab-measured [Cl⁻] in the QWTA well at Williamson Hall even though individual lab measurements differ (Figure 4a), especially in more recent months. In Figure 4b, we see that Eq. 1 is also able to relate SC and [Cl⁻] with similar success in the OPVL aquifer. The [Cl⁻] in both wells falls between chronic and acute thresholds by both SC and chloride measurements. Given the stratigraphic position of OPVL and its hydrological properties, it is logical that groundwater in this unit is strongly influenced by contaminants from the overlying Quaternary aquifer. These results show explicitly a surface water–groundwater connection.

In both figures, the differences between calculated and measured [Cl⁻] are a reminder that although chloride plays a dominant role in SC, the concentrations of other dissolved constituents (e.g., calcium, magnesium, and sulfate) contribute strongly and vary over time. Pooling the lab-measured chemistry data from both wells, we can approximate the total dissolved solids (TDS) at each sampling step by summing the concentrations of all species in the chemical analysis. This allows a comparison of the robustness of the linear relations for TDS and [Cl⁻] (Figure 5). Since total SC is the sum of contributions from all dissolved ions, it makes sense that the correlation with TDS would be stronger (R² = 0.74). However, the linear fit between SC and [Cl⁻] still explains more than half the variance in the dataset, and the relationship for this small group of samples aligns well with the equation derived from all historical QWTA data available (shown in red in Figure 5).

Chloride build-up in groundwater is of major concern for environmental and societal reasons. The data collected from the Williamson Hall wells shows that [Cl⁻] can vary by more than 50 percent within a single calendar year. Such variations would be invisible in an annual groundwater sampling program. A more granular, high-temporal-resolution understanding of how concentrations change across the multi-level TCMA aquifer system can improve our understanding of the degree to which surface water and groundwater are interconnected, with practical benefits for groundwater managers:

• Monitoring can identify locations where groundwater [Cl⁻] most rapidly responds to changing conditions at the surface. Continuous data from wells in those locations could then be used to determine the efficacy of improved de-icing practices and other strategies intended to reduce groundwater [Cl⁻].

• Elevated surface water [Cl⁻] in the TCMA relates to shallow groundwater chloride problems, with increasing [Cl⁻] in groundwater contributing variably to the increasing concentrations in lakes and streams. Continuous SC monitoring of groundwater will help parse out the relative contributions of groundwater versus surface water runoff, enabling more effective chloride mitigation strategies in surface waters. For instance, in areas where groundwater contributions dominate, improved practices to reduce chloride-laden surface water runoff may not reduce chloride in lakes and streams as quickly and to the degree intended if shallow groundwater chloride levels continue to increase (Cox et al., 2007).

• Efforts to catch road salt chloride “at the curb” and prevent its infiltration will require significant changes to urban infrastructure such as highways and storm sewers (Dietz et al., 2016; Barbier et al., 2018; and Szklarek et al., 2022). Understanding the sub-annual variations in [Cl⁻] will help agencies design solutions that are most effective during times when the contaminant influx is most significant.

• Because only negligible amounts of chloride are taken up by organisms and geological materials, the chloride already present in the groundwater system can be used as a conservative tracer to illuminate groundwater flow pathways that are not yet well understood. Continuous monitoring of chloride using SC is easy and inexpensive and could be used to shed light on the potential dispersal of other conservative contaminants that are not as easily detected in real time (e.g., per- and poly-fluoroalkyl substances).

In addition to temporal variations, there are spatial heterogeneities in groundwater composition even within the same aquifer unit. We have instrumented other nearby wells that are also screened in the Platteville Limestone to begin to understand these differences. These wells have been instrumented for shorter periods than the wells at Williamson Hall, and more time will be necessary to build up the necessary data for rigorous analysis. However, even a small array of sensors allows snapshots in time of the variability of dissolved content across the OPVL in this area. (See Supplement 1.)

In June 2023, there was an abrupt increase in SC in the OPVL well at Williamson Hall that did not appear in the data from the QWTA well (Figure 6a). To rule out sensor failure, we performed manual testing and confirmed that the sensor was returning accurate results. Water samples also confirmed that the chemistry of the water had changed (Figure 4b) with [Cl⁻] increasing substantially on short timescales (122 mg/L [32.6 percent] between April 10 and August 4). There is also an approximately 25-cm drop in water elevation in the OPVL well during this period (Figure 6b), although the correlation with the water chemistry changes is unclear and there was no equivalent dip in water elevation in the QWTA well.

Consultation with personnel from the Minnesota Department of Health made us aware of de-watering occurring at a construction site less than 500 m north of this well. When contacted, the geotechnical contractor provided the pumping start and stop dates, well construction information, and weekly volume of water pumped. Of the six de-watering wells, three were entirely in the Quaternary sediments and three extended into the OPVL by as much as 1.5 m. Altogether, the wells pumped nearly 138,000 m³ in the 20 weeks they ran.

The pumping start and stop dates correlate strongly with the onset and reversal of the SC excursion in the observation well (Figure 6a). Pumping at the site began on June 2, and the precipitous upswing in SC began June 5. Similarly, pumping ceased on October 16, and the collapse of SC began on October 20. There was a 3- to 4-day lag between changes to the local hydrogeology and the signal appearing in the data.

This excursion in SC represented a substantial change in the overall water chemistry at this well. Between April and August 2023, increases were observed in concentrations of chloride (33 percent), sulfate (20 percent), boron (13 percent), calcium (14 percent), iron (30 percent), lithium (41 percent), magnesium (9 percent), manganese (30 percent), sodium (13 percent), sulfur (6 percent), silicon (8 percent), strontium (23 percent), and zinc (507 percent). In the January 2024 sampling, all of these except zinc had returned to their April 2023 concentrations or less. Zinc remained elevated eight times above its pre-pumping concentration. (See McDaris et al., 2024, for the full set of chemistry data.) The quick return to pre-pumping water chemistry seems to indicate that no significant and long-lasting alterations were made to the flow system as a result of the pumping. However, the modeled [Cl⁻] is persistently higher than the measured level since the cessation of pumping, which suggests that at least one other dissolved constituent remained at a higher-than-normal concentration for a longer period (Figure 4a).

The physical mechanism for the observed water chemistry changes is unclear but could be related to water from Quaternary sediments being drawn down into the OPVL by pumping, shifting flow pathways due to changes in the hydraulic head drawing water from a new area within the OPVL, or some combination of these and other factors. High-frequency data from other wells near the site would have allowed investigation of those hypotheses in real time. A network of such real-time, continuous groundwater sensors would enable high-resolution study of flow pathways and contaminant movement in the TCMA, much like EarthScope and USArray enabled high-definition visualization of the mantle underneath the North American plate with seismic waves (Long et al., 2014).

Regardless of the mechanism responsible for these changes, it is clear that pumping related to construction temporarily changed the local flow regime supplying water to the Williamson Hall OPVL well. Given that de-watering permits are common in association with construction projects across the TCMA, these short-term anthropogenic activities could be modifying the direction and extent of groundwater–surface water interactions in a regionally extensive manner. Given that the uppermost aquifers in the TCMA are extensively impacted by contaminants, including chloride, in variable concentrations, the effects of that modification will vary across the landscape. A de-watering effort could be in an area of relatively low contamination, resulting in the intrusion of more highly contaminated water. Conversely, if the de-watering happens in an area of relatively high contamination, the pumping could slow or divert the natural spread of that contamination. It could even remove those contaminants from the regional groundwater entirely as the pumped water is generally discharged into the storm sewers, which empty directly into rivers in the TCMA. Understanding the context in which these extractions are taking place could help groundwater regulators mitigate or enhance the impacts of construction de-watering.

Groundwater has been getting warmer over the last century, with elevated temperatures measured tens of meters below grade (Menberg et al., 2013; Zhu et al., 2014; and Farr et al., 2017). In rural areas, this warming is due primarily to global climate change, but in urbanized areas, the increased heat is coming from both above and below ground level (Menberg et al., 2014; Tissen et al., 2019). In addition to climate change, these built environments absorb and reradiate heat (i.e., the urban heat island; Oke, 1982; Smoliak et al., 2015), and buried infrastructure, including building foundations, sewers, steam tunnels, and stormwater systems, conduct heat directly into the surrounding subsurface. Increasing groundwater temperatures contribute to increasing temperatures in lakes and streams fed by groundwater, which can be detrimental to native animal and plant species. Elevated groundwater temperatures also increase the rate of infrastructure corrosion and have implications for ground-source geothermal heating and cooling. Increasing groundwater temperatures are also linked with changes in water quality, both positive and negative (Bonte et al., 2014; Saito et al., 2016; and Riedel, 2019).

Measurements of urban groundwater temperatures are not abundant, and it is often necessary to estimate temperatures underground using an offset from temperatures at the surface (Benz et al., 2017). These models are becoming more sophisticated but necessarily rely on spatial averages, which may not capture all heat inputs in a particular place. Increasing the amount of groundwater temperature measurements will lead to a better understanding of real temperature distributions, as well as better models.

The earliest available data on groundwater temperature for the TCMA is from the 1830s. During several of his expeditions to map the upper Mississippi River basin, Joseph Nicollet inserted temperature probes into the ground at places where springs emerged. His measurements at a spring near Bdote (the Dakota name for the confluence of the Minnesota and Mississippi Rivers) yielded shallow groundwater temperatures that varied between 46°F (7.8°C) in July and 45.5°F (7.3°C) in January (Nicollet et al., 1843). This spring, now named Coldwater Spring and located in the heart of the TCMA, occurs within the uppermost portion of the OPVL, which is sourced primarily by water from the QWTA. Its modern temperature ranges between 10.7°C and 13.1°C, and it lags air temperatures by 4–5 months (Kasahara, 2016). Nicollet’s measurements in January and July likely did not capture the full range of spring temperatures in the 1830s. Kasahara’s data show that water temperatures in those months lie around the middle of the sinusoidal distribution each year. Assuming the seasonal sinusoid has remained about the same, measurements in May and November would have yielded more representative minimum and maximum values, respectively.

The work presented here constitutes fundamental science required to bring a new paradigm to groundwater monitoring. Continuous monitoring has been the norm in surface water studies for many years, and groundwater management can also benefit. We advocate for the creation of a sentinel network of real-time, continuous sensors to monitor groundwater quality at the regional scale in the TCMA. Because of the distributed statutory responsibility for groundwater monitoring, it would make sense for management agencies and researchers to collaborate on the installation of such a network to share the burden of its creation. This sensor network would benefit from the extensive subsurface knowledge developed by state agencies such as the Minnesota Geological Survey, Department of Natural Resources, Pollution Control Agency, Department of Agriculture, and Department of Health, as well as regional bodies like the Metropolitan Council and the vast expertise available in academia and the private sector. If archived in a single database, the data gathered in this effort would better illuminate poorly understood flow pathways and processes that can help resource managers ensure the continued availability of high-quality groundwater for the environment and society. For example, continuous groundwater temperature monitoring near streams and lakes that receive a significant component of groundwater discharge could help predict increased warming of these surface waters due to increased groundwater temperatures.

There are barriers to implementing such a plan, some of which are more challenging than others. We believe these barriers to be surmountable with concerted effort.

• Funding and purchasing: The sensor and data logger stations described here are relatively inexpensive, around US$1,000. With a modest allocation of funding each year, a gradual network roll-out could be achieved by having monitoring staff install sensor stations at wells following regular testing. Each year, the number of instrumented wells would grow in the normal course of monitoring work. State agencies do face contractual limitations on vendors from which equipment can be purchased, but these could be renegotiated with appropriate lead time.

• Workforce: A transition toward broad sensor-based monitoring would necessitate a shift of some positions from solely water sampling toward station installation and maintenance in support of the network. The sensor network will not replace human geotechnical staff, and there would always be the need for in-person water testing. The network would simply provide a tool to prioritize the wells most in need of that intervention.

• Data management: With the advent of real-time, continuous data, data management will become paramount. In just 3 years of the pilot study of two wells, we have generated over 120,000 measurements of seven discrete data points. Such large datasets bring issues of storage, quality control, accessibility, and analysis to the fore. Real-time access to this real-time data will also be an important part of maintaining its usefulness for water managers. With multiple bodies potentially contributing to the full network, detailed plans for data sharing and collaboration should be established at the start. Ideally, a single, dedicated office or agency at the state level would take in, store, analyze, and disseminate the increased volume of groundwater quality data being generated throughout the network.

• Technology: There is great interest in the development of sensors that can directly measure chemical species of interest, including chloride, in real time rather than relying on proxies like SC (NSF, 2024). These sensors hold enormous promise for revolutionizing many areas of environmental monitoring. However, there are many years of work ahead to create sensors with the precision, accuracy, and robustness for field application necessary to bring them to bear on the challenging issues we face. That ion specificity also means that this generation of sensors will be specialists and less flexible to broader applications beyond their initial design. Sensors measuring basic physical parameters, such as water elevation, temperature, and SC, can be adapted for many different applications, and old data can be analyzed with respect to new problems and contaminants of interest, as has been demonstrated in this work. It makes sense to move forward on the creation of a distributed sentinel network of simple sensors in the TCMA now rather than waiting for more specialized equipment.

• Network density: The resolution with which the sensor network can resolve flow patterns and anthropogenic changes will depend on the number of sensors deployed. The denser the distribution of sensors, the greater ability the network will have to illuminate issues of concern. (e.g., regional water flows in fracture networks and impacts of warming groundwater on lakes and streams). The number needed to provide this level of detail for each of six major aquifers will also depend on the nature of groundwater flow in the unit, which remains poorly resolved at the time of writing.

Minnesota has a wealth of water resources above and below the surface, as well as one of the most extensive water quality monitoring systems in the United States. To continue to protect Minnesota’s groundwater from increasingly numerous and challenging anthropogenic threats, that system needs to evolve to take advantage of new technologies. Incorporating real-time, continuous data into the regulatory monitoring process has the potential to provide a quantum leap in protective capacity for the region’s groundwater. We have demonstrated that data on just three simple parameters (water elevation, temperature, and SC) measured multiple times per hour in two aquifer units can be used to track chloride from the surface into successive aquifers, detect local changes in hydrogeology due to human activities such as transient pumping, and resolve the effects anthropogenic urban heat on groundwater temperatures over short timescales.

This body of work constitutes the basic research necessary for the creation of a sentinel network of distributed real-time sensors to monitor groundwater quality at the regional scale. We advocate for management agencies and researchers in Minnesota to collaborate on the installation of a network of real-time, continuous monitoring sensors in wells around the TCMA. Establishing such a network would illuminate many poorly understood groundwater processes and help groundwater managers ensure the continued availability of high-quality groundwater for the ecosystem and society.

The full dataset for this research is available via this in-text data citation reference: McDaris et al. (2024).

The University of Minnesota is built on the ancestral lands of the Wahpekute band that was ceded to the United States by the Treaty of Traverse des Sioux in July 1851, in an agreement that was not paid in full and whose underlying aim was the dissolution of the Dakota culture. The university has also benefited from Chippewa and Dakota (Mdewakanton, Wahpekuta, Wahpeton, and Sisseton Bands) land ceded by treaty and given to the University of Minnesota via the Morrill Act. Due to its land-grant status, the infrastructure, financial foundations, and faculty, students, and staff at the University of Minnesota all continue to benefit directly from these ceded lands, and we wish to acknowledge this support in our research.

Partial support for this work has been provided to McDaris by the families of Subir Banerjee, Richard Clarence Dennis, Harold M. Mooney, and Samuel Goldich through fellowships awarded by the University of Minnesota Department of Earth and Environmental Sciences (2020, 2021, 2022, and 2023, respectively), as well as the Gill Gabanski Scholarship (2020) from the Minnesota Groundwater Association, the G. A. Harris Fellowship (2023) from METER Group Inc., and the Mistletoe Research Fellowship (2023) from the Momental Foundation.

The authors express gratitude to Nicholas Wiest (Stantec) for his assistance in collecting and analyzing groundwater chemistry data, Stephen Labuz (MDNR) for providing groundwater data from WWMA, and Sharon Kroening (MPCA) for offering her expertise in regulatory agency data management throughout the project.

Supplemental material associated with this article can be found online at

Supplemental 1:https://doi.org/10.2113/EEG-D-24-00088.S1

Supplemental 2:https://doi.org/10.2113/EEG-D-24-00088.S2