Introduction
Bean Creek, originating from Devils Lake and flowing toward Hudson, Michigan, passes through an area of intensive agricultural activity (Figures 1, 2, and 3a). The rise of megafarms, particularly large-scale dairy operations, have created concern over water quality due to nutrient enrichment, chemical runoff, and ecological contamination. These concentrated animal feeding operations (CAFOs) often utilize thousands of acres of cropland within a 5 to 10-mile radius of the central facility for feed production and manure application (MSU Extension, 2023; EGLE, 2024). Previous studies have linked large-scale agricultural operations to eutrophication, harmful algal blooms, and ecological stress in freshwater systems (Withers & Jarvie, 2014; Akinnawo et al., 2023). However, there has been little direct monitoring of Bean Creek in this specific reach and limited investigation into whether Devils Lake may play a role in filtering upstream pollutants.
This study aims to determine whether megafarm practices between Devils Lake and Hudson are contributing measurable pollutants to Bean Creek, to test whether Devils Lake acts as a natural bioremediator for pollutants entering from upstream, and to identify potential hotspots of contamination downstream.
This study hypothesizes that Bean Creek water quality is degraded between Devils Lake and Hudson, with higher concentrations of pollutants near megafarm operations (Figure 3b). A secondary hypothesis is that Devils Lake acts as a bioremediator, reducing pollutant loads from upstream agricultural activity before water exits into Bean Creek.
Background
Agricultural impacts on freshwater ecosystems have been studied extensively in Michigan and the Midwest. Nutrient runoff, particularly nitrogen and phosphorus, is a major cause of algal blooms and hypoxia (River Raisin Watershed Council, 2009). Lakes often act as sinks or filters for pollutants, with some capacity to reduce nutrient and sediment loads before outflow. However, when pollutant inputs exceed that capacity, lakes can shift from nutrient sinks to nutrient sources, releasing stored phosphorus and nitrogen downstream. Concentrated animal feeding operations (CAFOs) have been shown to significantly increase pollutant loads in nearby water systems, and large dairies in Michigan typically operate across a mosaic of leased and owned fields. Within the area covered by Figures 3a and 3b, approximately 51,000 acres, the land base of these operations could occupy 30,000–40,000 acres of active cropland, consistent with the density of agricultural fields visible in satellite imagery. This area represents the primary zone of nutrient input and potential runoff into Bean Creek. While state and federal agencies have highlighted agricultural runoff as a key driver of water quality decline, localized data for Bean Creek and Devils Lake remain limited. This project helps fill that gap.
Additionally, widespread manure applications were witnessed by the author in early September just before the study period began, possibly introducing elevated nutrient loads into the watershed prior to the first sampling event.
The study area lies within a complex regional hydrological network. At the upper end, Devils Lake (Lenawee County, Michigan) serves as the headwater source for Bean Creek. Bean Creek flows south through Addison, Hudson and crosses into Ohio where it becomes part of the Tiffin River system. The Tiffin River is a tributary of the Maumee River, which ultimately drains into Lake Erie. Thus, pollutant inputs entering the Bean Creek–Devils Lake reach potentially carry all the way to Lake Erie via the Western Lake Erie Basin.
In Michigan’s lower peninsula, many adjacent watersheds (including the River Raisin watershed) also dominate land‑use patterns and nutrient flow. The Bean Creek/Tiffin/Maumee connection places the study reach in a different drainage corridor, flowing into Lake Erie rather than solely into the River Raisin system. Because Lake Erie has experienced periodic harmful algal blooms and nutrient‑driven ecological stress, understanding pollutant transport from localized agricultural activities through this watershed is essential.
Methods of Study
The study area encompassed the reach of Bean Creek between Devils Lake and Hudson, Michigan, including both the inflow and outflow of Devils Lake (Figure 1). Sampling was conducted weekly from September 22, 2025 through October 27, 2025, with at least one sampling event scheduled within 24 hours following a significant rainfall event. Five sites were selected for monitoring: Site 1 at the inflow to Devils Lake (Figure 4), Site 2 at the Devils Lake outflow (Figure 1), Site 3 midway between Devils Lake and Hudson (Figure 3a), Site 4 adjacent to megafarm operations south of Hudson (Figure 3a), and Site 5 downstream of operations near Hudson (Figure 3a).
At each site, water samples were collected manually using 1-liter Nalgene bottles at a consistent depth within the water column to ensure comparability among sites. Samples were transported to the Adrian College Rock Sample Preparation Laboratory immediately after collection, refrigerated upon arrival, and analyzed within 24 hours to maintain sample integrity.
Water quality was assessed through measurement of total phosphorus (TP), total nitrogen (TN), conductivity, and total dissolved solids (TDS). Analytical measurements were performed using a HACH DR/890 Colorimeter for phosphorus and nitrogen and a Hanna direct-read EC/TDS meter for electrical conductivity and total dissolved solids. Conductivity was measured in microsiemens per centimeter (µS/cm). Phosphate (PO₄³⁻) and nitrate (NO₃⁻) concentrations obtained from the colorimeter were converted to total phosphorus (TP=(PO₄³⁻/3.065)/.001) and total nitrogen (TN=NO₃⁻/4.43).
To interpret pollutant concentrations for TP, TN, Conductivity, and TDS, reference thresholds were compared using the River Raisin Watershed Council (2009) water quality scorecard. Total dissolved solids values reported in the scorecard in grams per liter (g/L) were converted to parts per million (ppm) to ensure consistency with my dataset and improve readability. Results were compared both across sites and over time to evaluate spatial and temporal differences in pollutant levels. Particular focus was placed on changes between Devils Lake inflow and outflow in order to test the hypothesis that the lake functioned as a natural bioremediator for upstream agricultural pollutants, as well as potential sources for pollutants downstream.
Results
Across the five sampling sites along Bean Creek, notable variation in pollutants were observed, as summarized in the quality report card and data sets (Table 1). Sites located near intensive agricultural operations, particularly Site 3, consistently exhibited higher concentrations of total phosphorus (TP) and total nitrogen (TN) relative to upstream locations. TP values commonly exceeded 100 µg/L at all sites and exceed the limit of the HACH DR/890 Colorimeter (897.9 µg/L). These results fall within the High to Very High classification range based on the water quality report card, where the threshold for pollutants beginning to negatively affect ecological activity starts at the High (yellow) category (Table 1). TN levels show a different story, only surpassing the High threshold in two data sets. Overall, 72 of the 120 total samples collected exceeded their respective High thresholds, demonstrating the frequency at which pollutant levels reached ecologically concerning concentrations .
In contrast, Sites 1 and 2 (the inflow and outflow of Devils Lake) generally displayed lower nutrient and dissolved solid levels, with higher concentrations of both TP and TN based on rainfall, pollutant input, and reflecting how pollutants were distributed within the Devils Lake system. Electrical conductivity and total dissolved solids (TDS) increased as we looked miles downstream of Devils Lake (Sites 3-5), suggesting cumulative loading from agricultural drainage between Sites 3–5. The most pronounced increases occurred at site 3, where conductivity values often exceeded 700 µS/cm and TDS surpassed 400 ppm, classified as Very High for conductivity and High for dissolved solids, indicating strong anthropogenic influence.
To better illustrate these findings, all measured data were organized into graphical representations for each data set (Figures 6–12). These visualizations help depict how pollutant levels fluctuate over time and vary between sampling sites, making trends in nutrient concentration and water quality more apparent. Rainfall also plays a significant role in this study and is displayed in Figure 13.
Discussion
The graphical analyses reveal clear and consistent pollutant patterns throughout the study period. Conductivity and TDS trends across all six sampling sets show Devils Lake acting as a partial filter. Sites 1 and 2 consistently exhibit lower concentration levels than their downstream counterparts before a sharp spike appears at Site 3. From there, these elevated levels persist as the water flows downstream through Sites 4–5 (Figure 6). This dataset also showed the lowest variability across all datasets, making it the most consistent indicator of pollutant behavior in the watershed.
Figure 7, which illustrates conductivity over time, and Figure 8, which shows TDS over time, both reinforce this consistency. Although displayed separately, they exhibit nearly identical temporal patterns, with contaminant levels at Sites 3–5 remaining elevated throughout the study period. Notably, these concentrations frequently exceed the River Raisin Report Card (RRC) thresholds, 500 µS/cm for conductivity and 300 ppm for TDS, further emphasizing the excessive pollutant loading occurring downstream of Devils Lake.
TN graphs (Figure 9), although not representing a major pollutant in this study, still help illustrate the system’s limited filtering capacity. Devils Lake can buffer nutrient loads up to a point, but once overwhelmed, it releases these pollutants into Bean Creek. You can see this reflected in the fluctuating levels at Sites 1 and 2, where brief drops or spikes in concentration reveal how Devils Lake temporarily absorbs incoming nutrient pulses before reaching saturation. These shifts underscore that while the lake can dampen short‑term variability, it cannot fully regulate or neutralize sustained pollutant inputs, allowing excess nitrogen to move downstream once its buffering capacity is exceeded. It could also indicate that the lake’s ability to remove or process nitrogen is less efficient than its handling of other pollutants, causing more persistent downstream transport. Each sampling set shows a noticeable surge at Site 3 that consistently carries downstream.
Figure 10, which displays TN over time, further supports this pattern. Here, pollutant levels remain consistently below the High threshold defined by the RRC (1.5 mg/L for TN), reinforcing that TN was not a major pollutant of concern during this study. However, the influence of early September manure applications may be a key influence, with elevated nutrient levels appearing during the first two sampling sets.
TP graphs (Figure 11) present the most concerning trends. Concentrations remain High to Very High at nearly every site across almost all sampling sets, noting that the RRC defines High as 75 µg/L and Very High as 150 µg/L. These patterns also underscore rainfall’s influence on pollutant movement within the watershed. Sample sets 1 and 5 were taken within 24 hours of significant rain events and both exhibited elevated TP levels. A drought period between these events corresponded with reduced concentrations, while the heavy rainfall preceding Set 1 likely contributed to the extended spikes seen in Set 2.
Figure 12, which depicts TP over time, highlights an especially concerning trend. All but one data point rises above the High threshold set by the RRC, underscoring the persistent phosphorus burden within the watershed. An interesting pattern emerges in this figure: Sites 2–5 track together, showing similar increases and decreases in pollutant levels throughout the sampling period. In contrast, Site 1, the inflow to Devils Lake, displays an almost inverted pattern. This behavior reflects the timing of pollutant entry into Devils Lake and the delay associated with the lake’s bioremediation processes. As pollutants enter the system, Devils Lake temporarily absorbs and processes part of the load before releasing excess nutrients downstream, resulting in the lagged and mirrored behavior observed between the inflow and outflow sites.
Additionally, rainfall played a key role in shaping these pollutant patterns. Sample Sets 1 and 5 were collected within 24 hours of rainfall events (Figure 13.), totaling 0.52 inches on 09/22 and 0.51 inches on 10/20 according to Weather Underground (2025). These sample sets also show elevated concentrations of pollutants, most notably total phosphorus (TP), highlighting how rainfall can rapidly mobilize stored nutrients into the system. Even with these nearly identical rainfall amounts, the data shows differing lag times for pollutant decline. This discrepancy is most likely due to the large manure applications that occurred in early September, which introduced a substantial nutrient load into the watershed at the beginning of the study. As a result, more pollutants were present in the system early on, requiring longer periods to filter through and degrade.
Conclusion
The observed pollutant gradients along Bean Creek strongly support the hypothesis that megafarm operations contribute measurable nutrient and solute loads to the watershed. Elevated phosphorus and nitrogen concentrations downstream of Devils Lake parallel patterns reported by Withers & Jarvie (2014) and Akinnawo et al. (2023), who found that large-scale agricultural runoff plays a dominant role in driving eutrophication and ecological stress across Midwestern river systems.
Fluctuating nutrient levels at the Devils Lake inflow and outflow indicate that the lake functions as a partial bioremediator, temporarily reducing upstream nutrient loads through sedimentation and biological uptake. However, the downstream rise in pollutant concentrations shows that this buffering capacity is limited. Surface runoff, tile-drain discharge, and field-applied manure and fertilizers from the surrounding megafarm operations reintroduce substantial nutrient and solute loads into Bean Creek. This pattern aligns with statewide assessments noting that lakes can provide short-term nutrient retention but cannot compensate for sustained high-volume inputs from CAFO-associated cropland (EGLE, 2024).
Conductivity and TDS trends further reinforce this interpretation. Elevated ionic loads, primarily nitrates, sulfates, chlorides, and potassium, are characteristic of watersheds influenced by fertilizer application, manure leaching, irrigation return flows, and wastewater inputs, as described by Hem (1985) and Boyd (2015). Additional research by Sharpley et al. (2013) and Carpenter et al. (1998) demonstrates that fertilizers, manure, and pesticides collectively introduce dissolved solids and organic contaminants that directly increase conductivity and nutrient concentrations, matching the trends documented in Bean Creek. The persistence of these elevated values beyond Site 3 suggests limited dilution capacity and continued downstream transport.
Taken together, the spatial, graphical, and quantitative data converge on a clear conclusion: agricultural megafarms between Devils Lake and Hudson exert a defining influence on Bean Creek water quality. Devils Lake functions as a natural but limited bioremediation system, capable of reducing nutrient loads under moderate conditions but unable to withstand the pulses of pollutants introduced during major runoff events, heavy rainfall, or post-manure application periods. During these episodes, nutrient and solute concentrations exceed the lake’s filtering capacity and pass directly into Bean Creek, reinforcing the pollutant surges shown in Figures 6–12.
This integrated evidence supports the study’s central hypothesis that Bean Creek experiences degradation downstream of Devils Lake due to proximity to megafarm operations. It also partially supports the secondary hypothesis: while Devils Lake does provide measurable bioremediation, its buffering capacity is inconsistent and easily exceeded under high-input scenarios. These findings underscore the urgent need for improved manure and fertilizer management practices throughout the watershed to reduce nutrient and solute loading into Bean Creek and downstream ecosystems.
References
Akinnawo, E., Adeyemi, R. A., & Olowofela, J. A. (2023). Agricultural runoff and eutrophication in freshwater systems: A case study of Midwestern watersheds. Journal of Environmental Management, 334, 117283.
Boyd, C. E. (2015). Water Quality: An Introduction (2nd ed.). Springer.
Carpenter, S. R., Caraco, N. F., Correll, D. L., Howarth, R. W., Sharpley, A. N., & Smith, V. H. (1998). Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological Applications, 8(3), 559–568.
EGLE (2024). Michigan Concentrated Animal Feeding Operations (CAFO) Program Report. Michigan Department of Environment, Great Lakes, and Energy.
Hem, J. D. (1985). Study and Interpretation of the Chemical Characteristics of Natural Water (3rd ed.). U.S. Geological Survey Water-Supply Paper 2254.
MSU Extension (2023). Manure Management Guidelines for Michigan Dairies. Michigan State University Extension Publications.
River Raisin Watershed Council (2009). River Raisin Watershed Management Plan. Adrian, MI.
Sharpley, A. N., Kleinman, P. J. A., Jarvie, H. P., & Flaten, D. (2013). Distant connections: Effects of agricultural practices on stream water quality in the United States. Journal of Soil and Water Conservation, 68(2), 36A–42A.
Withers, P. J. A., & Jarvie, H. P. (2014). Delivery and cycling of phosphorus in rivers: A review. Science of the Total Environment, 468–469, 326–344.
Apple Maps (2025). Satellite Imagery and Topographic Context for Bean Creek Watershed. Data accessed via Apple Maps aerial imagery, October 2025.
Maps of World. (2025, April 3). Lenawee County Map [Map]. https://www.mapsofworld.com/usa/states/michigan/counties/lenawee-county-map.html
Weather Underground. (2025). Daily historical weather data for Adrian, MI (KADG) for 9/22, and 10/20. https://www.wunderground.com/history/daily/us/mi/adrian/KADG/
