Managing the Red Lake Nation’s and Minnesota’s largest lake: monitoring and paleolimnology support a site-specific standard for Upper and Lower Red Lakes (Red Lake Nation and Minnesota, United States)

Abstract

Burge DRL, Edlund MB, Bowe S, Bowe K, Anderson JP, Bouchard RW, Heathcote AJ, Leavitt PR, Engstrom DR. 2023. Managing the Red Lake Nation’s and Minnesota’s largest lake: monitoring and paleolimnology support a site-specific standard for Upper and Lower Red Lakes (Red Lake Nation and Minnesota, United States). Lake Reserv Manage. 40:1–17.

Water quality and the ability of lakes to provide benefits such as recreation and aquatic life are negatively impacted by climate change, aquatic invasive species, and nutrient enrichment. To address these lake-management challenges, regionally tailored water quality criteria need to assign achievable and protective goals while minimizing assessment errors. In Minnesota, when lakes exceed state water-quality standards, restoration plans are developed and implemented based on eco-regional standards, or in cases where it is determined that the regional standard is not appropriate, a site-specific standard may be developed. Upper Red Lake and Lower Red Lake, located within the boundaries of the Red Lake Indian Reservation (the Red Lake Nation) in northwestern Minnesota (United States), have been recently experiencing cyanobacteria blooms, as well as routinely exceeding the regional water quality criteria for total phosphorus and chlorophyll a. The last 20 yr of water quality monitoring suggest a stable recent history with no significant trends. To extend this time perspective, biogeochemical evidence from paleolimnology was used to reconstruct a 150 yr history for these 2 large, shallow lakes to determine whether site-specific water quality standards might be recommended. Biogeochemical evidence (e.g., sedimentation rates, phosphorus fractions, biogenic silica, diatom community composition, and fossil algal pigments) from 10 sediment cores revealed complex sedimentary dynamics together with a small gradual increase in limnological productivity over the last 2 centuries. Despite the implied productivity change, diatom-inferred phosphorus showed no significant increase among the sediment cores. Furthermore, the reconstructed total phosphorus values were within the range of modern monitored values. Because the paleolimnological evidence suggests little change to the aquatic ecology of the Red Lakes over the last 150 yr, and monitoring data show regular exceedance of regional TP standards with no trend in the last 20 yr, site-specific water quality standards are deemed appropriate for this large, shallow lake system.

Keywords:

• Diatoms
• lake monitoring
• nutrient criteria
• paleolimnology
• shallow lakes

Lakes provide highly valued ecosystem services such as clean drinking water, recreation, transportation, fisheries, and biotic diversity. In turn, lake water quality can impact recreational revenue (Keeler et al. 2015), commercial revenue, and local property values (Henry et al. 1988, Dziuk and Heiskary 2003, Reynaud and Lanzanova 2017). Water quality management is complicated by watershed land-use alterations, nonnative species invasion, undesirable algal blooms, toxic pollutants, and climate change. For example, both nutrient loading from landscape changes and the warming of lakes can promote cyanobacteria growth and higher concentration of their toxins, which are potentially lethal to humans, livestock, and domestic pets (Paerl and Huisman 2008, Wagner and Adrian 2009, Kosten et al. 2012). Even in remote, pristine watersheds, lakes can be impacted by atmospheric deposition of nutrients and pollutants (Spaulding et al. 2015, Hobbs et al. 2016) or increasing global temperatures (Woolway et al. 2019). Long-term lake monitoring and knowledge of historical influences are crucial to sustainable management of aquatic ecosystems and are fundamental to understanding the biogeochemical trajectory of lakes (European Union 2000, US Environmental Protection Agency 2000).

In Minnesota, where lake ecology varies along major gradients of climate, geography, geology, lake morphometry, and land-use history (Eddy 1938, Moyle 1945, Moyle 1956, Bright 1968, Heiskary et al. 1987), a “weight of evidence” approach has been taken to develop lake eutrophication standards (Heiskary and Wilson 2008). Recognizing the regional diversity of Minnesota’s lake types (Heiskary et al. 1987), criteria were initially established based on the explanatory variable total phosphorus (TP) and 2 response variables, chlorophyll a (Chl-a) and Secchi depth (Heiskary and Walker 1988, Heiskary and Wilson 2008). It was noted that shallow polymictic lakes (maximum depth < 8 m) were distinctly different from deeper dimictic lakes (maximum depth > 10 m) lakes in the Central Hardwood Forests and Western Corn Belt Plains ecoregions. Shallow lakes in these regions tended to have higher TP and Chl-a and reduced Secchi depths, and were often characterized as eutrophic even in reference systems. Furthermore, paleolimnological studies comparing diatom-inferred phosphorus pre- and post-European settlement showed a nutrient increase in shallow lakes with disturbed watersheds (Heiskary and Swain 2002, Heiskary et al. 2003, Ramstack et al. 2003, Heiskary et al. 2004, Ramstack et al. 2004, Heiskary and Lindon 2005). Although Heiskary and Wilson (2008) identified differences in trophic state between polymictic and dimictic lakes in most ecoregions, these differences were determined to be small in the Northern Lakes and Forests (NLF) ecoregion. As a result, a single set of lake eutrophication standards was adopted for both lake types occurring in the NLF (Heiskary and Wilson 2008).

Paleolimnological studies have provided a crucial line of evidence for determining historical lake conditions and establishing nutrient criteria for Minnesota’s lakes (Heiskary and Wilson 2008). Such studies employ a wide variety of geochemical (e.g., sedimentation, sediment phosphorus) and biotic proxies (e.g., diatoms, algal pigments) from a lake sediment record to reconstruct ecological change from perturbations such as hydrological modification (Bradbury et al. 2004, Yang et al. 2018) and nutrient enrichment (Wang et al. 1988, Spaulding et al. 2015). Diatoms have been a particularly useful proxy for reconstructing historical phosphorus concentrations and ecological change, especially in Minnesota lakes (Ramstack et al. 2003, Ramstack et al. 2004).

Combining biogeochemical and biotic approaches, multiproxy paleolimnology was used to reconstruct environmental change in Lake of the Woods, a large, shallow lake basin located just north of the Red Lakes on the US-Canadian border (Anderson et al. 2017, Edlund et al. 2017, Reavie et al. 2017). The sediment records from Lake of the Woods recorded 20th century nutrient enrichment from municipal and industrial sources, which once abated resulted in the ecosystem slowly shifting to a novel ecology rather than recovering to its predisturbance state. Given their proximity and morphometric similarity to Lake of the Woods, the Red Lakes, which have never been impacted by point-source nutrient pollution, provide a potentially important contrast to recent ecological change in large, shallow, sub-boreal lakes.

Upper Red Lake and Lower Red Lake are 2 large, shallow (<10 m) and connected basins located almost entirely within the Red Lake Indian Reservation (Red Lake Nation) in northwestern Minnesota. Like Lake of the Woods, they are remnants of glacial Lake Agassiz (Wright et al. 1992). The watershed of the Red Lakes is at the crossroads of 3 of Minnesota’s ecoregions, with the expansive Red Lake Peatlands draining into Upper Red Lake and a mosaic of hardwood and coniferous forests draining into Lower Red Lake. Upper Red Lake (z-max: 5.4 m, 482.7 km²) has a residence time of 10.8 yr and drains into Lower Red Lake (z-max: 10.6 m, 686.1 km²), which has a 12.7 yr residence time (RESPEC 2016, Anderson 2017, Bowe et al. 2021). Although Secchi depth, Chl-a, and TP in both Upper Red Lake and Lower Red Lake have exceeded Minnesota’s water quality standards for the NLF ecoregion, no significant trends for these parameters have been observed in 20 yr of monitoring (Anderson 2017), despite recent observations of cyanobacterial bloom in both lakes. Dollinger et al. (2017) found that for the relatively small, undisturbed watershed of the Red Lakes, about one-third of the tributaries and lakes were impaired for aquatic life or recreational use due to natural factors such as excessive sediment or high organic matter. Naturally occurring, high concentrations of organic matter can provide challenges to lake management in northern Minnesota, as natural bog staining of waters contributes to reduced Secchi depth measurements (Brezonik et al. 2019).

Historical records of anthropogenic modification of the Red Lakes and their watershed include selective logging at the beginning of the 20th century (Albrecht and Thomas 1977), attempted draining of the Red Lake Peatlands during the 1910s to 1930s (Volstead Act of 1906), damming of the lake in 1931, and collapse and recovery of the walleye (Sander vitreus) fishery at the turn of the 21st century (Pereira et al. 1992, Gangl and Pereira 2003). Regional climate trends indicate increasing winter temperatures and decreasing average wind speeds (Reavie et al. 2017), which could result in more frequent lake stratification and internal phosphorus loading. In the absence of a longer monitoring record, a paleolimnological investigation was undertaken to ascertain changes in the ecology of Upper Red Lake and Lower Red Lake and to place current conditions in context with a longer historical record (Reavie et al. 2017, Yang et al. 2018).

Here, we use modern and paleolimnological evidence to (1) characterize phosphorus levels within Upper Red Lake and Lower Red Lake, (2) establish ecological trajectories of the Red Lakes over the past 150 yr, and (3) determine whether site-specific nutrient criteria are appropriate based upon historical nutrient reconstructions and modern monitoring data.

Materials and methods

Coring methods

Three piston cores were collected from the ice surface of each Red Lake along east–west transects during March 2016. Four additional cores were collected in the western portion of Upper Red Lake during February 2018 (Fig. 1, Table 1). Sediment cores were collected using a piston and drive-rod system equipped with a 6.5 cm diameter polycarbonate core tube (Wright 1991). Core recovery ranged from 69 to 102 cm. The upper watery sediments were stabilized with Zorbitrol and the cores were transported to the lab for sectioning. All sediment cores were subsampled in 0.5 cm increments for the top 10 cm and 1 cm increments below that, except core UC1B, which was sectioned in 1 cm increments in its entirety.

Figure 1. A watershed map showing sediment coring sites for Upper Red Lake (UC) and Lower Red Lake (LC) within the Red Lake Nation and Minnesota, USA. Black filled circles indicate cores with conformable sedimentation that were used for subsequent analyses, whereas white filled circles indicate rejected cores with problematic sedimentation identified with radiometric dating.

Table 1. Sediment core recovery information and geochemical proxies analyzed in addition to loss on ignition and ²¹⁰Pb dating (Cs = ¹³⁷Cs dating; phosphorus fractions, P-frac; diatoms, D; pigments, Pig).

Core ID	Recovery year	Lat. N	Long. E	Depth (m)	Length recovered (m)	Analyses
LC1	2016	47.99023	−95.15025	7.5	0.81	P-frac, D
LC2	2016	47.96323	−94.87279	8.0	0.81	P-frac , D
LC3B	2016	48.00795	−95.03203	7.7	0.69	P-frac, D, Pig
UC1	2016	48.10329	−94.94996	4.7	0.84	P-frac, D
UC1B	2018	48.10331	−94.94995	4.4	1.01	Pig
UC2A	2016	48.12668	−94.75829	4.6	0.81	^137Cs
UC3	2016	48.14822	−94.61871	4.2	0.80	^137Cs
UC4	2018	48.15018	−94.85711	4.2	1.02	P-frac, D
UC5	2018	48.17389	−94.75387	4.1	1.03	–

Water quality

Spatial trends in TP (μg/L), Chl-a (μg/L), and Secchi depth (m) were explored using monitoring data provided by the Red Lake Department of Natural Resources from 1999 to 2019. Samples were collected monthly at 5 stations on east–west transects within each basin. Growing season data, June to September (Heiskary and Wilson 2008), were explored using one-way analysis of variance (ANOVA; R Core Team 2014) to detect within-basin differences, and values from each basin were compiled in a second one-way ANOVA to detect between-basin differences.

Dating and geochemistry

Sediment chronology was determined using both ²¹⁰Pb and ¹³⁷Cs dating. Lead-210 was analyzed by isotope-dilution alpha spectrometry with dates and sedimentation rates calculated using the constant rate of supply (c.r.s.) model (Appleby and Oldfield 1978, Binford 1990, Edlund et al. 2017). Cores were measured for ¹³⁷Cs by gamma spectrometry to identify the 1963 peak in atmospheric nuclear testing and deposition (Robbins and Edgington 1975). Cores with clear evidence of depositional hiatuses (obvious discontinuities in the ²¹⁰Pb activity profiles and misplaced ¹³⁷Cs dates) were excluded from further analyses and included the paired cores UC2A and UC2B, and UC3 (Supplemental Fig. 1). Bulk density (dry mass per volume of fresh sediment), inorganic content, organic content, and carbonate content of sediments were determined using loss-on-ignition techniques (Dean 1974). The ²¹⁰Pb dating models were used to estimate sedimentation rates as dry mass accumulation rates (DMAR; g/cm²/yr) for each core. Bulk sediment DMAR was combined with concentrations of sediment constituents, biogenic silica (BSi), and sediment P concentrations to determine down-core rates of accumulation.

BSi, a proxy for historical siliceous algal productivity, was measured using weighed subsamples (∼30 mg) extracted with dilute NaCO₃ (DeMaster 1979, Conley and Schelske 2001). Dissolved silica was measured colorimetrically on a Unity Scientific SmartChem 170 discrete analyzer as molybdate reactive silica (SmartChem 2012a).

Sediment phosphorus (P) fractions were analyzed following the sequential extraction procedures in Engstrom (2005), Engstrom and Wright (1984), Psenner and Puckso (1988), and Kopácek et al. (2005). Extracts were analyzed colorimetrically on a Unity Scientific SmartChem 170 discrete analyzer using methods described by SmartChem (2012b). In addition to TP in cores, sediment fractions include the refractory forms of mineral-bound P, recalcitrant organic P, and Al-bound P, and the labile or readily exchangeable forms of Fe-bound, labile organic P, and loosely bound P.

Pigment and isotope analyses

Algal pigment analyses were performed on 15 sections each from cores LC3B and UC1B. Carotenoids, chlorophylls, and derivatives were extracted (4 C, dark, N₂) from freeze-dried sediments according to Leavitt and Hodgson (2001), measured on a Hewlett-Packard model 1050 high-performance liquid chromatography system, and are reported relative to total organic carbon (TOC; Hall et al. 1999). Stable C and N isotope ratios and elemental composition were determined on unacidified whole sediment samples from core LC3B using a ThermoQuest (F-MAT) Delta^PLUS XL isotope ratio mass spectrometer equipped with a continuous flow (Con Flo II) unit, an automated Carlo Erba elemental analyzer as an inlet device, and following standard procedures of Savage et al. (2004). Stable N (δ¹⁵N) and C (δ¹³C) isotopic compositions were expressed in the conventional δ-notation in units of per million (‰) deviation from atmospheric N₂ and an organic C standard that had been calibrated previously against authentic Vienna Pee Dee Belemnite. Sample reproducibility was <0.25 ‰ and <0.10 ‰ for δ¹⁵N and δ¹³C determinations, respectively.

Diatom preparation and enumeration

Diatoms were prepared by digesting 30 mg freeze dried sediment in 10% v/v HCl to dissolve carbonates followed by 30% H₂O₂ to oxidize organic matter. The samples were then centrifuged, rinsed, and transferred to Battarbee chambers, where the cleaned material was dried onto coverslips (Battarbee 1973). Coverslips were permanently attached to microscope slides using Zrax mounting medium (Ramstack et al. 2008). Diatoms were identified along measured random transects to the lowest taxonomic level under 1000–1250× magnification (full immersion optics of numerical aperture [NA] > 1.3). Six hundred valves were enumerated in each sample and encountered taxa were documented using the voucher flora method (Bishop et al. 2017). Valve density, a measure of overall diatom abundance, was calculated using a modification to the equation provided by Scherer (1994) where the area of the bottom of the beaker was replaced by the area of the Battarbee settling chamber. Identification of diatoms relied on floras and monographs such as Hustedt (1927–1966, 1930), Patrick and Reimer (1966, 1975), Krammer and Lange-Bertalot (1986–1991), Reavie and Smol (1998), Camburn and Charles (2000), and Fallu et al. (2000). Diatom counts were converted to percentage by species or taxon; abundances are reported relative to total diatom counts in each sample.

Diatom community analysis

To determine community similarity and changes within and among cores, nonmetric multidimensional scaling (NMDS) was implemented in the vegan package in R (Oksanen et al. 2013, R Core Team 2014). Square root transformation was applied to the relative proportion diatom data using the Hellinger method (Legendre and Gallagher 2001). Transformed data were then converted into a distance matrix using Euclidian measures and the NMDS was performed using Bray-Curtis dissimilarity. Stratigraphy of predominant diatoms (species with greater than or equal to 3% relative abundance in one or more core depths) was plotted against core date using the rioja R package (Juggins and Juggins 2019). Temporal relationships among the dominant diatom communities within each sediment core were explored using the Euclidian distance matrices in a constrained cluster analysis (CONISS). In the rioja R package, significant cluster groups were determined using the broken stick model (MacArthur 1957).

Diatom-inferred phosphorus

Down-core diatom communities were also used to reconstruct historical epilimnetic TP levels. A transfer function for reconstructing historical Briggsian log-transformed TP (logTP) was developed previously on the relationship between modern diatom communities and modern environmental variables in 89 Minnesota lakes (Ramstack et al. 2003, Edlund and Ramstack 2006) using weighted averaging (WA) regression with inverse deshrinking and bootstrap error estimation (Juggins 2003, Juggins and Juggins 2019). Data are modeled as logTP values but presented as back-transformed values of TP in μg/L.

To evaluate the strength of the reconstruction, we determined the amount of variance in the diatom data that can be accounted for by the TP reconstruction. This is calculated by the variance explained by the first axis of an ordination of the sediment assemblages constrained to diatom-inferred TP, divided by the variation explained by an unconstrained ordination of the sediment assemblages (λr/λp). A maximum λr/λp value of 1.0 would mean that TP was a perfect explanatory variable of diatom community change (Juggins et al. 2013). Furthermore, correlation between the diatom-inferred TP (DI-TP) and community variation was also examined. Using ANOVA (R Core Team 2014), the measured TP, modern DI-TP, and historical DI-TP data were examined for significant differences to infer any changes between and within Upper and Lower Red Lake sediment cores. Furthermore, if any significant diatom community clusters were identified through constrained cluster analysis, then the modern and historical DI-TP values were subset around these periods of change in the sediment core to evaluate changes in P between periods.

Results

Sediments and dating

Unsupported ²¹⁰Pb activities ranged from 0.08 pCi/g to 21.2 pCi/g and generally followed a monotonic down-core trend consistent with radioactive decay and conformable sediment deposition (Fig. 2). All ²¹⁰Pb profiles exhibited some flattening over their uppermost 10 cm, likely caused by physical sediment mixing from wave and current action in these large, shallow basins. Levels of supported ²¹⁰Pb were reached at depths ranging from 11 cm to 31 cm. Sediment core UC2A recorded an anomalous spike in unsupported ²¹⁰Pb between 20 cm and 30 cm, and core UC3 had low and variable unsupported ²¹⁰Pb that was limited to the top 7 cm (Supplemental Fig. 1). Bimodal ¹³⁷Cs peaks in UC2A and UC3 provide further evidence of nonconformable deposition in these cores. These problematic dating records indicate that large areas of central and eastern Upper Red Lake have nonconformable sedimentation owing to their shallow depths and prevailing westerly winds.

Figure 2. Profiles for unsupported ²¹⁰Pb (pCi/g), dry mass accumulation rate (DMAR; g/cm²/yr), and ²¹⁰Pb age (calendar year) against core depth (cm) for sediment cores recovered from Upper Red Lake (dashed line) and Lower Red Lake (solid line). Lines and points are colored by coring location: LC1 (red), LC2 (olive green), LC3B (green), UC1 (blue), and UC4 (magenta).

Dry mass accumulation rates (DMAR) followed a general increasing trend across all cores, ranging from 0.0025 g/cm²/yr to 0.0968 g/cm²/yr. The sediment cores LC1, LC2, LC3B, UC1B, and UC4 had low but steadily increasing sediment rates (<0.027 g/cm²/yr), whereas UC1, UC2A, and UC3 displayed the highest sedimentation rates with variable trends (Fig. 2). The cores UC1, UC2A, and UC3 were all marked down core by a high deposition event (>0.075 g/cm²/yr) estimated at 25, 33, and 8 cm respectively. Given the highly variable nature of UC1, UC2A, and UC3, unsupported ²¹⁰Pb trends were checked using ¹³⁷Cs. The uniform ¹³⁷Cs peak for UC1 occurred between 13 cm to 14 cm, which corresponded with the ²¹⁰Pb years estimated as 1968 to 1962, respectively. All cores from Lower Red Lake and UC1, UC1B, and UC4 demonstrated conformable sedimentation, and dating was modeled back to at least the late 19th century. The sediment cores that exhibited the greatest variability also had the highest sedimentation rates and proved to be nonconformable sites in Upper Red Lake.

Geochemistry

The Lower Red Lake cores were composed of 35% to 45% organic matter by weight, higher than the Upper Red Lake cores, which were usually between 20% and 30% organic matter (Fig. 3). The opposite was true for inorganic matter, with Upper Red Lake sediments composed of 55% to 65% inorganics and Lower Red Lake sediments ranging between 40% and 55%. Upper Red Lake also had 10% to 20% CaCO₃, about 5% more carbonate content than Lower Red Lake (Supplemental Fig. 2). All cores showed gradual decreases in the percentage of inorganic sediments and, conversely, increases in the percentage of organic sediments up core, marked by a spike in organic matter at the very top of the cores. Dry mass accumulation rates gradually increased over the 20th century in all cores. Core UC1 showed a marked accumulation event of 0.097 g/cm²/yr dated to 1914, composed of a 15% increase of inorganic sediments. Despite the dominance of inorganic materials, the organic flux of the 1914 event reached levels similar to the modern sediments for UC1, LC1, and LC3B.

Figure 3. The percentage (dry weight %) and flux (mg/cm²/yr) of organic matter and biogenic Si, percent Fe-P, diatom-inferred total phosphorus (μg/L), and percentage of planktonic diatom taxa for Upper Red Lake (dashed line) and Lower Red Lake (solid line); sediment cores plotted by year on the y-axis. Lines and points are colored by coring location: LC1 (red), LC2 (olive green), LC3B (green), UC1 (blue), and UC4 (magenta).

All cores showed increases in sediment phosphorus concentration and flux between the oldest and most recent ²¹⁰Pb dated sediments. Upper Red Lake had lower sediment TP concentrations compared to Lower Red Lake (Supplemental Fig. 3). In the Lower Red Lake cores, the recalcitrant organic P fraction was greatest, while in Upper Red Lake, the recalcitrant organic P and mineral-bound P fractions had similar concentrations. The concentration of mineral-bound P decreased over time in all cores, while the other fractions—recalcitrant organic P, labile organic P, Al-bound P, Fe-bound P, and loosely bound P—increased in concentration up core.

Pigments and stable isotopes

Fossil algal pigments were analyzed on cores LC3B and UC1B. Pigment concentrations from both cores showed a sharp increase at the core top (Supplemental Figs. 4 and 5). Fucoxanthin (siliceous algae) was low and remained nominal throughout the cores until the most recent decades, when it began a sharp increase to 50 nmol/g C and 73 nmol/g C for LC3B and UC1B, respectively. Diatoxanthin (diatoms) ranged from 7.09 nmol/g C to 44.73 nmol/g C, with a slight increase in UC1B over the 20th century and no clear trends in LCB3. Diadinoxanthin (dinoflagellate, diatoms, chrysophytes, cryptophytes) and alloxanthin (cryptophytes) showed a gradual increase over the 20th century. Canthaxanthin (colonial cyanobacteria) was higher in LC3B (36 nmol/g C on average), but lower in UC1B (21 nmol/g C on average). Aphanoxanthophyll (N-fixing cyanobacteria) had no early occurrences in LC3B and UC1B but peaked sharply at the beginning of the 21st century to 5 nmol/g C and 31 nmol/g C, respectively. Phaeophytin-b and β-carotene (total productivity) showed small increases at the beginning of the 20th century and variable trends until a spike in the 21st century. Chlorophyll a (total productivity) showed a slow gradual increasing trend in both cores over the 20th century followed by a sharp peak at the top of the core. Chlorophyll b (green algae) followed a trend similar to that of chlorophyll a except UC1B had no detections of chlorophyll b throughout the 20th century. Echinone (cyanobacteria) and lutein (green and red algae) showed no clear trends over the 20th century in the Red Lakes.

Carbon (δ ¹³C) and nitrogen (δ ¹⁵N) isotope geochemistry from core LC3B revealed decreasing trends in the δ ¹³C and C:N (Supplemental Fig. 6), while the percentage of δ ¹⁵N increased toward the top of the core.

Diatoms and inferred historical TP

Historical diatom productivity was explored using BSi concentration and flux and a count-based abundance of diatom microfossils. Percent BSi was lower in Upper Red Lake compared to Lower Red Lake, and both lakes exhibited a gradual trend of increased concentration and BSi flux over the 20th century, suggesting increasing productivity trends (Fig. 3). Diatom valve density for the most part mirrored the productivity increases shown by BSi flux, although valve density in UC1 was variable relative to the other BSi records during the sedimentary period of the early 20th century.

The diatom communities of Upper Red Lake and Lower Red Lake clustered together in ordination space by lake, indicating each lake’s distinct community composition (Fig. 4). Sediment core trajectories over time were centered around the oldest sample except at the tops of the cores. Upper Red Lake had a greater abundance of ­benthic diatoms, and the historical communities showed a significant shift in the diatom assemblage coinciding with decreases in Staurosira construens Ehrenberg and S. venter (Ehrenberg) Cleve & J. D. Möller, replaced by a modern community of increasing Achnanthidium minutissimum (Kütz.) Czarn. and Navicula cryptotenella Lange-Bertalot over the late 20th century. Diatom communities of Lower Red Lake contrasted with Upper Red Lake and had about 15% greater abundance of planktonic species such as Asterionella formosa Hassall, Aulacoseira ambigua (Grunow) Simonsen, A. granulata (Ehrenberg) Simonsen, Lindavia bodanica (Eulenst. ex Grunow) Nakov et al., and Stephanodiscus niagarae Ehrenberg (Figs. 5 and 6, Supplemental Figs. 8–10).

Figure 4. Two-dimensional solution for a nonmetric multi-dimensional scaling ordination of common diatoms observed in sediment cores from Upper Red Lake (dashed lines) and Lower Red Lake (solid lines); (stress = 0.14, non-metric stress r² = 0.98).

Figure 5. Stratigraphic plot of the dominant diatoms, as percent relative abundance, from the Upper Red Lake core UC1 plotted against 210Pb age (year) on the y-axis. The solid blue line represents a significant break in the constrained cluster analysis of diatom assemblages as shown on the right. 

Figure 6. Stratigraphic plot of the dominant diatoms, as percent relative abundance, from the Lower Red Lake core LC1 plotted against 210Pb age on the y-axis. A constrained cluster analysis of diatom assemblages as shown on the right.

Constrained cluster analysis of Lower Red Lake cores did not identify any significant shifts in diatom communities when compared against a broken stick model (Fig. 6, Supplemental Figs. 8–10). Diatom communities from 2 Upper Red Lake cores, UC1 and UC4, showed a historical shift in the late 1960s (Fig. 5, Supplemental Fig. 8). Changes in the dominant diatoms for these UC cores included increased Achnanthidium minutissimum, Aulacoseira ambigua, Navicula cryptotenella, and other benthic species that replaced the previously more dominant Staurosira construens and to a lesser degree S. venter and Staurosirella pinnata (Ehrenberg) D.M. Williams & Round.

The range of historical diatom-inferred TP (DI-TP) across the Red Lakes over the last 150 yr was 30.1 μg/L to 52.0 μg/L; the average DI-TP for UC1 and UC4 was 38.6 μg/L, while that for LC1, LC2, and LC3B was 43.7 μg/L (Fig. 3). The proportion of variation in the diatom data that could be explained by TP (λr/λp) for LC1, LC2, LC3B, UC1, and UC4 was 0.25, 0.41, 0.58, 0.21, and 0.47, respectively. These proportions were significantly correlated with diatom community variability (Supplemental Table 2). Analysis of variance revealed a significant difference between Upper Red Lake and Lower Red Lake; however, Tukey pairwise comparisons of modern DI-TP, historical DI-TP, and monitored TP values revealed the only significant difference between basins was for the monitored TP values (Figs. 7 and 8, Supplemental Table 1).

Figure 7. Box plots of monitored total phosphorus (μg/L), chlorophyll a (μg/L) and Secchi depth (m) for 5 sampling stations in Lower Red Lake (left) and Upper Red Lake (right). The red lines on each plot represent the Northern Lakes and Forests (NLF) ecoregion criteria for TP, Chl-a, and Secchi depth (Heiskary and Wilson 2008).

Figure 8. Box plots of monitored total phosphorus (μg/L), chlorophyll a (μg/L), and Secchi depth (m) averaged by lake for Lower Red Lake (left) and Upper Red Lake (right). The solid black lines on each plot represent the NLF ecoregion nutrient criteria for TP, Chl-a, and Secchi depth. The dashed blue lines represent suggested site-specific standards for Lower Red Lake (LRL) and the dashed red line represents suggested standards for Upper Red Lake (URL) sufficient to protect aquatic life and recreational uses. Boxes are bisected by the median and represent the interquartile range (IR), dots represent outliers, and whiskers represent the maximum and minimum as defined by the IR limits plus and minus 1.5 × IR, respectively.

Water quality monitoring

Monitored TP showed no significant temporal or spatial differences within each basin (Fig. 7, Supplemental Table 1, Supplemental Fig. 11), although Upper Red Lake (44.8 μg/L) had significantly higher water column TP values compared to Lower Red Lake (36.1 μg/L; P < 0.001). Monitored chlorophyll a was also significantly higher (P < 0.001) in Upper Red Lake compared to Lower Red Lake; 16.1 μg/L and 10.2 μg/L, respectively. Chlorophyll a corrected for phaeophytin was higher in Upper Red Lake compared to Lower Red Lake, at 12.3 μg/L and 11.2 μg/L, respectively, although the difference was not significant (P = 0.06). Average Secchi depth was significantly lower (P < 0.001) in Upper Red Lake compared to Lower Red Lake, at 0.68 and 1.13 m, respectively.

Discussion

Over the last 20 yr, Upper Red Lake and Lower Red Lake have exceeded TP, Secchi depth, and Chl-a standards for the NLFs ecoregion (30 μg/L TP, 9 μg/L Chl-a, 2 m Secchi depth; Heiskary and Wilson 2008). In addition to exceeding lake eutrophication standards, cyanobacteria blooms have been frequently reported, especially in Upper Red Lake. Paleolimnological evidence covering the past 150 yr showed only minor biogeochemical changes in Upper Red Lake and Lower Red Lake. Dry mass accumulation rates showed slight increases in most cores, and sediment phosphorus, in particular the percentages of Fe-P and loosely bound P, also showed gradual increases over the 20th century. However, each of these variables is subject to core-top enrichment by diagenetic processes (dissolution and up-core porewater diffusion). Historical events may have contributed to the long-term increase in sediment P and sedimentation rates during early logging, damming of the lake, and draining of the peatlands, though all these perturbations were relatively local and modest in observable impact. Moreover, models currently estimate that only 32% of P input to the lakes is from watershed tributaries, while 47% is from atmospheric deposition (RESPEC 2016, RMB 2017). With most watershed changes occurring in the late 19th and early 20th centuries, historical land-use effects should have quickly subsided with reforestation. The increase in sediment Fe-P could be indicative of increased summertime anoxia events and internal loading driven by warmer temperatures and decreased regional wind speeds (Reavie et al. 2017). Another indicator of a gradual increase in productivity is the δ ¹³C (%) and δ ¹⁵N (%) isotopes increasing above historical conditions over the 20th century (Leavitt and Hodgson 2001). The slight decrease in C:N isotopic fractionation may indicate increased algal productivity. Overall, phosphorus and sediment accumulation and sediment proxies of in-lake productivity appear to be gradually increasing in the lakes over the 20th century.

A notable change in lake ecology is revealed in the paleodiatom community record from Upper Red Lake, where a significant change—increases of Achnanthidium minutissimum and Navicula cryptotenella—occurred around the late 1960s. These 2 species differ from the other dominant diatoms owing to their benthic life strategies; N. cryptotenella often lives in the benthos and A. minutissimum is often found epiphytic on another alga or macrophytes (Wehr et al. 2015). An increased signal of benthic diatoms could be indicative of increased sediment mixing or increased water clarity, whereas the increase in A. minutissimum may also coincide with increases in filamentous algae or aquatic macrophytes.

These minor paleoecological changes in Upper and Lower Red Lake, however, contrast with other biogeochemical evidence indicating ecological stability for the previous 200 yr. The Lower Red Lake diatom sediment record showed no significant community shifts, and there were no significant differences or trends in the diatom-inferred TP or biogenic silica content for either lake. The average modern and historical diatom-inferred TP for each basin was within the monitored averages for each basin (Anderson 2017), with no significant differences between the DI-TP and monitored TP values. This suggests that the nutrient condition represented by diatom ecology is functionally similar between the 2 basins. Based on diatom ecology and the shallow morphology of the Red Lake basins, Upper Red Lake and Lower Red Lake were more like moderately productive western Minnesota lakes from the training set, rather than northern Minnesota lakes (Ramstack et al. 2003). The low variance estimates for diatom-inferred phosphorus draw caution (Juggins et al. 2013); in lieu of improving the taxonomic coverage between the model and training set (Bennion et al. 2001) or training set sample size (Reavie and Juggins 2011), we can look for supporting evidence based on similarities between the diatoms and historical proxies.

The fossil pigment data similarly show no major community shifts within the cyanobacteria or the broader algal community. Landscape factors that might explain why the modern ecology of Upper Red Lake and Lower Red Lake is similar to the historical ecology include only minor and transient watershed disturbances, low portions of watershed land development, and a relatively small watershed to lake surface area ratio. The paleo and modern biogeochemistry for Upper Red Lake and Lower Red Lake was temporally consistent within each respective basin; however, a few characteristics differentiate the 2 basins. Upper Red Lake is half the depth of Lower Red Lake, on average 3.6 m versus 7 m, respectively. This shallower nature contributes to potentially greater wind-driven mixing in Upper Red Lake. The effects of this are pronounced in the higher inorganic content of sediments and large portions of the basin with nonconformable sediment profiles. Diatom ecology also separates the 2 lakes, with higher proportions of planktonic diatoms in Lower Red Lake, whereas Upper Red Lake had higher proportions of benthic diatoms. These are subtle differences compared to the basin differences observed in Lake of the Woods. Diatoms in the large shallow southern basin of Lake of the Woods have changed from presettlement community compositions notably in response to P enrichment (Reavie et al. 2017), whereas diatom communities in the northern and deeper Ontario portions of the lake have shifted contemporaneously with a reduction in ice cover (Rühland et al. 2010). Mineral-bound phosphorus and inorganic sediments were also higher in Upper Red Lake compared to Lower Red Lake, indicating that depth and mixing may be a primary determining factor for the spatial distribution of sedimentation within the 2 basins. Within Upper Red Lake, nonconformable sediment accumulation in the easternmost cores provides evidence of intensive sediment resuspension. More frequent mixing and lower overall capability to bury P are likely reflected in higher monitored TP, Chl-a, phaeophytin-corrected Chl-a, and Secchi depth, which together showed somewhat more naturally eutrophic conditions in Upper Red Lake. Sediment-based evidence suggests that wind-driven mixing plays a larger role in the ecology of Upper Red Lake compared to Lower Red Lake.

Understanding depth, wind, and sediment dynamics is not only important for informing the ecological management of a lake, but also necessary for site selection in paleolimnological studies. Lakes near one another may reflect environmental perturbations differently (Forbes and Hickman 1981), while separated basins in the same lake can yield different sedimentary records and experience varying water quality (Rühland et al. 2010, Reavie et al. 2017). Given the complex nature of working in large and productive shallow lakes (Heathcote et al. 2015), the evidence presented here and in other studies highlights the importance of a multicore, multibasin approach to establish reliable historical reconstructions.

While Upper Red Lake and Lower Red Lake may have some geochemical sedimentation differences, they both recorded subtle, uniform increased organic sediments, increased phosphorus accumulation, and increased diatom and algal productivity beginning in the early 20th century and continuing to modern times. Mirroring a decline in mineral-bound phosphorus, the proportion of phosphorus bound by iron increases over this time. This shift begins around the time of the initial land-use changes within the watershed: selective logging, draining of wetlands, and damming of the lake. We know these watershed activities can be associated with phosphorus contributions to aquatic environments (Nieminen 2004), and that shallow lakes in particular take longer to recover (Jeppesen et al. 2005, Jeppesen et al. 2007). However, these activities were only modest in scope and, since the early 20th century, have been largely curtailed, yet the lakes continue to increase (albeit very modestly) in available nutrients and productivity. As Edlund et al. (2017) suggested for nearby Lake of the Woods, regional climate forcing may be driving changes in the Upper Red Lake and Lower Red Lake. One factor might be recent increases in precipitation and runoff (Kharel and Kirilenko 2015). Another climate effect may be the decreasing trend in regional wind (Reavie et al. 2017), which could increase the number of anoxic stratification events contributing to internal loading (Nürnberg 1995, James 2017a, 2017b). Increased winter temperatures can lead to longer open-water seasons and longer algae growing seasons (O’Beirne et al. 2017). In addition to potential climate-driven increases in internal loading, the Red Lakes may be subject to enrichment through the atmospheric deposition of dust (Todhunter and Cihacek 1999, Zhu et al. 2019). Future studies in support of determining water quality standards should examine the role of climate change and atmospheric dust deposition in lake-nutrient cycling.

Given the paleolimnological evidence and 20 yr of modern water quality monitoring that show only subtle historical trends and no recent changes, we conclude that Upper Red Lake and Lower Red Lake have likely exceeded the current regional total phosphorus standard (30 µg/L; Heiskary and Wilson 2008) for at least the last 200 yr. With this understanding, we conclude that the adoption of a US Environmental Protection Agency approved site-specific standard for the Red Lakes is warranted (Heiskary and Wilson 2008, Heiskary and Wasley 2011, MPCA 2011, Bouchard et al. 2017). For Upper Red Lake and Lower Red Lake we suggest adopting the 75th percentile based on the 20 yr of growing season (Jun–Sep) monitored water quality as site-specific standards for TP and Chl-a, and the 25th percentile of the monitored Secchi depth. Using this framework, Lower Red Lake should be assessed for impairment using thresholds at 45 µg/L TP, 14 µg/L Chl-a, and 0.91 m Secchi depth, and Upper Red Lake should be assessed using 54 µg/L, 20 µg/L Chl-a, and 0.61 m Secchi depth. An alternative solution would be to regulate Upper Red Lake and Lower Red Lake as part of the Central Hardwood Forests ecoregion (60 µg/L TP, 20 µg/L Chl-a, 1 m Secchi depth). The recommended site-specific standards should be sufficient to protect aquatic life and recreational uses for these lakes. Although the walleye fishery collapsed in these lakes in the 1990s, this collapse was attributed to overexploitation (Gangl and Pereira 2003) and not to a decline in water quality. The diatom-inferred phosphorus and monitoring data reported here indicate a history of walleye-favorable conditions (Oglesby et al. 1987, Ward et al. 2007). Through the collaborative Red Lake Department of Natural Resources (DNR) and Minnesota DNR special fisheries management regulations in the Red Lakes, the walleye population has rebounded, demonstrating that current water quality conditions can support a high-quality and healthy fishery. In light of fisheries limits relaxing over the past decade, the current and apparently stable water quality of the Red Lakes supports recreational fishing as a primary beneficial use. Here we propose site-specific standards above the average for the diatom-inferred TP and the monitored TP, but within the higher trophic limits of the observed variability. These empirically derived standards are intended to be incorporated by lake managers when considering site-specific standards that incorporate and support the designated uses of the Red Lakes.

Conclusion

We found the combination of multiple sediment cores and multiple paleoproxies invaluable for reconstructing the history of Upper Red Lake and Lower Red Lake, in particular sediment dating and the use of diatom microfossils paired with modern monitored data. The geochemical and biotic sediment record of Upper Red Lake and Lower Red Lake do not record any major changes to the lakes’ ecology over the last 150 yr. The sediment record does capture slow but minor increases in overall productivity that could be linked to climate forcing and activities within or outside the watershed. Combined with 20 yr of monitoring data that also show no significant trends, the productive nature of the lake, and strong potential for large-scale sediment resuspension, we recommend adopting alternative nutrient standards based on recent monitoring data for the site-specific regulation of Upper Red Lake and Lower Red Lake. The ongoing monitoring of stable water quality and recovery of the Red Lakes’ fisheries highlights the successful cooperation and collaborative management of the lakes by the Red Lake Nation and the state of Minnesota.

Disclosure statement

No potential conflict of interest was reported by the authors.