Lake Superior Research Articles

Lake Superior evaporation: A long-term eddy covariance dataset at Stannard Rock Lighthouse (2008–2022)

Robust, accurate, and direct measurements of evaporation and related energy fluxes on the Laurentian Great Lakes are necessary to understand the large historical range in water levels, regional climatology, lake hydrodynamics, and lake-effect snowfall, all of which inform water management. Despite the societal and scientific importance of this information, few long-term, full-year, in situ measurements exist due to logistical, financial, and safety-related challenges. We present 15 years (2008–2022) of eddy covariance data from Stannard Rock, a historic lighthouse on Lake Superior located 38 km southeast of Manitou Island and 72 km north of Marquette, Michigan. We provide information about the site and instrumentation, as well as data availability and processing. Analysis of this unique long-term dataset, available through the AmeriFlux network (US-GL1), will improve our ability to understand the drivers and patterns of large-lake surface energy fluxes and will advance predictions of evaporative regimes over Lake Superior.

What controls the onset of winter stratification in a deep, dimictic lake?

AbstractThe transition from summer stratification to winter stratification is considered for deep, dimictic Lake Superior. The fall transition is dynamically distinct from the better‐studied spring transition; it is characterized by a wind‐driven collapse of weakening summer stratification, a surface‐cooling–driven period during which the water column is essentially isothermal, and eventual onset of inverse (cold upper layer) stratification, after which sub‐thermocline water temperatures are largely fixed for the rest of the season. Due to the small value of thermal expansivity near the temperature of maximum density, temperature gradients do not impart much stability to the water column, and it is difficult for winter stratification to form immediately upon dropping below the temperature of maximum density. Instead, the water column continues to cool until the thermal expansivity is sufficiently large in magnitude to allow temperature gradients to impart stability. This observation implies that the temperature locked into the deep water for the winter season will be a consequence of the specific meteorological forcing experienced in a given year; there may be long‐term climate change driven trends in the timing of the onset of stratification but not trends in winter deep‐water temperature. To address this hypothesis, 15 yr of moored temperature data collected in Lake Superior are examined, and a scaling argument and climatological data are applied to better understand interannual variability in the onset of winter stratification.

Stable isotope analysis of western Lake Superior predatory fishes, part one: Trophic niche overlap

Characterizing food web interactions between introduced and native fish predators is important to quantify niche overlap, assess predator–prey balance, and support fisheries management. Lake Superior provides an ideal setting to investigate potential trophic overlap between native top predators, lake trout (Salvelinus namaycush) and walleye (Sander vitreus), and a diverse assemblage of introduced salmonids. Angler-caught fish were used to measure trophic position and niche overlap between predatory fishes in the western arm of Lake Superior based on carbon and nitrogen stable isotope ratios (δ13C, δ15N). Within a species, we did not find consistent differences in either δ13C or δ15N values by sex, collection year, capture location, or length. Among species, the range in δ15N values was >5 ‰, indicating that these species occupy ∼2 trophic positions, with dietary reliance ranging from primarily invertebrates to primarily fish, whereas the range in δ13C values indicated dietary reliance on a mix of pelagic and benthic prey. Combined, the δ13C and δ15N values revealed siscowet lake trout as the apex predator having a distinct trophic niche, while walleye and lean lake trout were intermediate trophic position predators with overlapping isotopic niches, and introduced salmonids were the lowest trophic position predators with overlapping trophic niches. Overlapping trophic niches indicated similar resource reliance and habitat occupancy among native lake trout and walleye, as well as among introduced salmonids, but less so between native and introduced predators.

Development of a total variation diminishing (TVD) sea ice transport scheme and its application in an ocean (SCHISM v5.11) and sea ice (Icepack v1.3.4) coupled model on unstructured grids

Abstract. As the demand for increased resolution and complexity in unstructured sea ice models is growing, higher demands are also placed on the sea ice transport scheme. In this study, we couple the Semi-implicit Cross-scale Hydro-science Integrated System Model (SCHISM, v5.11) with Icepack (v1.3.4), the column physics package of the Los Alamos sea ice model (CICE); a key step is to implement a total variation diminishing (TVD) transport scheme for the multi-class sea ice module in the coupled model. Compared with the second-order upwind scheme and the finite-element flux-corrected transport (FEM-FCT) scheme, the TVD transport scheme is overall superior when evaluated based on conservation, accuracy, efficiency (even with very high resolution), and strict monotonicity. Although it is slightly weaker than FEM-FCT in terms of accuracy alone, the TVD scheme still outperforms the other two schemes in comprehensive performance. The new coupled model outperforms the existing single-class ice model of SCHISM in the case of Lake Superior. For the Arctic Ocean case, it successfully reproduces the long-term changes in the sea ice extent, sea ice boundary, concentration observations from satellites, and thickness from in situ measurement.

Per- and Polyfluoroalkyl Substances in the Duluth, Minnesota Area: Exposure to and Biomarker Responses in Tree Swallows Relative to Known Fire-Fighting Foam Sources.

Tree swallow nest boxes were deployed at sites proximal to two putative aqueous film forming foam (AFFF) sources in the Duluth, MN area, as well as along the St. Louis River and a reference lake for comparative purposes in 2019, 2020 and 2021. The two AFFF sites were the current Duluth Air National Guard Base (ANG) and the Lake Superior College Emergency Response Training Center. Between 13 and 40 per- and polyfluoroalkyl substances (PFAS), depending on year, were detected and quantified in tree swallow egg, nestling carcasses, and stomach contents. Assessments were made of oxidative stress and ethoxyresorufin-O-dealkylase activity in liver tissue, thyroid hormone levels in plasma and thyroid glands, DNA damage in red blood cells, and two measures of immune response (haptoglobin-like activity and immunoglobulin) in plasma of the nestlings. Additionally, other contaminants, such as polychlorinated biphenyls, legacy organochlorine pesticides, and trace elements, were assessed at sites with no previous data. Total egg PFAS concentrations at the ANG site and north of that site were 30-40 times higher than at the reference lake, while nestling PFAS concentrations were 10-15 times higher. In contrast, the St. Louis River sites had slightly, but non-statistically significant, elevated egg and nestling PFAS concentrations relative to the reference lake (2-5 times higher). One PFAS, perfluorohexane sulfonate (PFHxS), was higher, as a proportion of total PFAS, at sites with a known AFFF source compared to the reference lake, as well as compared to sites along the St. Louis River with mainly urban and industrial sources of PFAS. The ratio of total carboxylates to total sulfonates also distinguished between PFAS sources. There were few to no differences in biomarker responses among sites, and no association with PFAS exposure.

Watershed inputs of suspended sediment drive patterns of total phosphorus in Chequamegon Bay, Lake Superior

Lake Superior is considered the least anthropogenically stressed and has the lowest offshore phosphorus (P) concentrations of the five Laurentian Great Lakes. However, nearshore habitats in Lake Superior are showing evidence of nutrient related stress. We examined drivers of total P dynamics in Chequamegon Bay, a shallow embayment in southwestern Lake Superior located in a region of the Laurentian Great Lakes that is primarily forested with low human development. Over a nine-year period (2014–2022) we measured total and soluble reactive phosphorus (TP and SRP, respectively), total suspended solids (TSS), and chlorophyll-a (Chl-a) at 12 locations distributed across Chequamegon Bay. Path analysis revealed that TP in this region of Lake Superior is largely sediment bound and driven by watershed inputs of suspended sediment. SRP and Chl-a make up only a small portion of TP. TP and TSS were highly correlated, with a stronger correlation at the most nearshore locations and following extreme precipitation events in 2016 and 2018. TP and Chl-a had a weak positive correlation at low TP concentrations, and lack of correlation at high TP concentrations. This suggests that despite high TP inputs from runoff events, Chl-a response was minimal, likely due to low light availability and limited bioavailability of sediment-bound P. Understanding conditions where episodic inputs of TP could contribute to the reactive P pool and how hydrodynamics affect biogeochemical processes and algal response to nutrient inputs are critical to understanding how an expected increase in extreme events will influence nearshore water quality in large lakes.

Lake Superior fish community and fisheries, 2001–2022: An era of stability

Lake Superior is the least anthropogenically impacted of the Laurentian Great Lakes ecosystems, yet dramatic changes to the fish community are evident. Previous published works chronicled those changes and the efforts to rehabilitate the fish community through the year 2000. Here, we review through the year 2022, where post-rehabilitation stability was driven by lean lake trout (Salvelinus namaycush namaycush) as the most abundant piscivore in nearshore waters, siscowet lake trout (Salvelinus namaycush siscowet) as the most abundant piscivore in offshore waters, and a healthy, intact assemblage of native prey species, which created ecological redundancies and helped stabilize the food web. Stocking of non-native salmonines was reduced 74%, and populations of Chinook salmon (Oncorhynchus tshawytscha) and coho salmon (Oncorhynchus kisutch) were maintained through natural reproduction. Despite reduced stocking, yield from recreational fisheries was stable. Likewise, developments in population modeling led to evaluations and refinement of management strategies that helped create stability for lake trout, lake whitefish (Coregonus clupeaformis), and cisco (Coregonus artedi) fisheries. With lake trout rehabilitation achieved, focus shifted toward rehabilitation of native brook trout (Salvelinus fontinalis), lake sturgeon (Acipenser fulvescens), and walleye (Sander vitreus). Despite continued control efforts, sea lamprey (Petromyzon marinus) abundance increased considerably, and estimates of fish killed by lampreys averaged 2.65 million kg annually. Environmental changes have benefited sea lampreys and fostered thermal habitats more suitable to non-native organisms, posing new challenges for managers and researchers. Nevertheless, the post-rehabilitation stability in the contemporary fish community will help provide resilience to future perturbations in the ecosystem.

Metals in sediment of the upper Great Lakes: Spatial distribution, temporal trends, anthropogenic enrichment, and risk assessments

Thirteen elements including Al, Ag, As, Co, Cu, Cd, Cr, Fe, Mn, Ni, Se, Zn, and Pb were measured in 107 surface grab sediment samples and 175 segments of eight cores from Lakes Superior, Michigan, and Huron, using inductively coupled plasma mass spectrometry (ICP-MS). Concentrations in Ponar grabs vary considerably among metals and among locations, ranging from the highest median for Fe in Lake Superior (42,000 mg/kg) to the lowest median for Ag in the main Lake Huron (0.05 mg/kg). The inventory at coring sites ranged from 7 × 106 mg/m2 of Fe to 3 mg/m2 of Ag. The background concentrations were estimated from deeper core segments, and enrichment factors (EFs) were calculated with Fe or Al as the reference element. The results show that Al, Fe, Co, Cr, and Mn did not enrich, Ag, Cu, and Ni were present higher than expected from natural sources alone, while Pb, Cd, Se, Zn, and As have been enriched at most sites after European settlement in the region. EFs of most metals are higher for Lake Michigan than the other lakes. However, EF comparison among sampling sites revealed intrinsic problems of this approach for the assessment of human interference. Preliminary risk assessment, conducted by calculating risk quotients, revealed environmental risks of some metals in each lake; however, the results should be interpreted with caution because the approach used is considered to be conservative.

Cyanobacteria in cold waters: A study of nearshore cyanobacteria assemblages in Lake Superior

Cyanobacterial blooms, often associated with warm, eutrophic lakes, are of widespread concern owing to their potential disruption of ecological and public health. Recently, cyanobacterial blooms have been observed in oligotrophic lakes, including Lake Superior, a large cold-water system. Despite recent developments, limited research has targeted phytoplankton assemblages in the northern Lake Superior nearshore, where isolated cyanobacteria blooms have emerged since 2019. In 2019, the cyanobacteria assemblage at 10 nearshore monitoring stations was examined to understand potential bloom-forming taxa and their association with water quality. We observed a diverse cyanobacteria assemblage, with highest densities in spring and biomass in fall. Pseudanabaena limnetica, Dolichospermum flos-aquae and Limnothrix sp. were the most prevalent species across sites and seasons. Variation partitioning analysis revealed that water quality conditions were more influential drivers of cyanobacteria density and biomass than meteorological factors, particularly in regions with elevated nutrient inputs and following stormy conditions. However, at present, cyanobacteria compose <15 % of the total phytoplankton biomass in the study areas. Our monitoring data and recent reports of cyanobacteria blooms on the north shore, suggests that current blooms are not widespread. Nonetheless, we found a diverse array of cyanobacteria taxa, with most capable of producing cyanotoxins. Considering the global uncertainty in the factors contributing to cyanobacteria blooms, in association with the large-scale climatic changes affecting Lake Superior, a proactive approach to assessing risks of blooms is suggested. This should include data-generating efforts (e.g., frequent phytoplankton monitoring and reported blooms investigations) to support future collaborative initiatives focused on managing cyanobacterial blooms.

Vertical distribution of Lake Superior cisco (Coregonus artedi) spawning aggregations and implications for population monitoring

Cisco (Coregonus artedi) support an evolving commercial roe fishery in Wisconsin waters of Lake Superior. To monitor trends in spawning cisco abundance, fishery managers recently began estimating adult biomass and exploitation using fall hydroacoustic surveys, which were combined with gill net surveys to inform apportionments of acoustic data. The gill net survey design consisted of paired top-suspended and bottom-set gill nets, but only the sex ratios from top nets are currently used with the hydroacoustic surveys due to an assumption that cisco in Lake Superior are pelagic spawners. However, the vertical sex distribution of cisco during spawning aggregations has been described as dynamic, with males becoming more bottom-oriented throughout the spawning season. We used multilevel aggregated binomial regressions to: 1) determine if there is bias between top and bottom gill net catches of cisco for either sex and if it changes throughout the spawning season, 2) evaluate how the vertical distribution of males and females may create bias in sex ratios used to estimate exploitation, and 3) explore the effect that maturity (i.e., gonadal development) has on vertical distribution during spawning aggregations. We identified sex-specific bias in vertical catch location that has the potential to bias estimates of sex ratio, and the source of this bias may be attributable to maturity driven changes in behavior. These findings highlight a need for caution when relying on gill nets to apportion cisco sex ratios during spawning aggregations and provide support for a non-pelagic alternative hypothesis of spawning behavior.

Aquatic invertebrate diversity in Apostle Islands and Isle Royale waters: Comparison among habitats and sampling gears and to open Lake Superior

Aquatic invertebrate composition in Great Lakes nearshore regions is known to differ from offshore, but studies representing the closest-to-land end of this gradient are primarily from estuaries and rivermouths having substantial watershed connectivity and anthropogenic influence. Here, we present data from aquatic invertebrate surveys conducted in two Lake Superior parks that are distanced from such watersheds and pressures, namely Apostle Islands National Lakeshore (APIS) and Isle Royale National Park (ISRO). Our intensive, multi-gear surveys from 99 APIS stations in 2017 and 165 ISRO stations in 2012 reveal a broad diversity of zooplankton and benthic/littoral macroinvertebrate taxa. Park samples yielded 29 zooplankton and >300 benthic/littoral taxa, with richness exceeding that of comparable nearshore datasets, especially for aquatic insects, leeches, and mites. Station depth was a major factor structuring invertebrates, and benthic/littoral densities were highest and composition most diverse at shallow stations having aquatic vegetation. Species composition and taxa accumulation patterns differed considerably among sampling gears, highlighting the value of multi-gear surveys. Several park mollusk and insect species matched ‘special concern’ listings, and two non-native cladocerans were very abundant. These two surveys added 11 new species to the aquatic macroinvertebrates known from Lake Superior, highlighting the importance of these parks in harboring biodiversity and the importance of individualized assessments of places not well represented in more routine lakewide biological monitoring. Our data are available as baselines for future biological surveys and trend assessments.

The benthic nepheloid layer in the offshore waters of the Great Lakes and its post-dreissenid disappearance

Prior to the appearance of Dreissena, pronounced benthic nepheloid layers (BNL) near the bottom characterized by elevated levels of both turbidity and total phosphorus (TP), were a consistent and extensive feature of the offshore, stratified waters of all the Laurentian Great Lakes, except Lake Superior. In recent (2010–2019) years, the BNL has disappeared from all areas except for central Lake Erie, where only a small decrease in bottom turbidity has occurred. All stratified regions which exhibited a pre-Dreissena BNL, including central Lake Erie, experienced substantial post-Dreissena reductions in near-bottom TP, although the forms of phosphorus (particulate or soluble) responsible for these reductions have varied from lake to lake. Notably, the arrival of Dreissena at offshore sites was not accompanied by an increase in soluble phosphorus. Initiation of changes in the BNL almost invariably preceded appearance of Dreissena in the offshore, suggesting both that dreissenid impacts on the reductions in the BNL were largely remote, and by extension that the source of the BNL was also at least in part remote. Previous researchers’ estimates of the importance of the benthic pool of phosphorus to offshore water column concentration suggest that the post-invasion reductions in bottom phosphorus during the stratified season could be contributing to the offshore oligotrophication of the lakes.

Genetic differentiation across a steep and narrow environmental gradient: Quantitative genetic and genomic insights into Lake Superior populations of Quercus rubra.

Adaptive differentiation of traits and underlying loci can occur at a small geographical scale if natural selection is stronger than countervailing gene flow and drift. We investigated this hypothesis using coupled quantitative genetic and genomic approaches for a wind-pollinated tree species, Quercus rubra, along the steep, narrow gradient of the Lake Superior coast that encompasses four USDA Hardiness Zones within 100 km. For the quantitative genetic component of this study, we examined phenotypic differentiation among eight populations in a common garden, measuring seed mass, germination, height, stem diameter, leaf number, specific leaf area and survival. For the genomic component, we quantified genetic differentiation for 26 populations from the same region using RAD-seq. Because hybridisation with Quercus ellipsoidalis occurs in other parts of the species' range, we included two populations of this congener for comparison. In the common garden study, we found a strong signal of population differentiation that was significantly associated with at least one climate factor for nine of 10 measured traits. In contrast, we found no evidence of genomic differentiation among populations based on FST or any other measures. However, both distance-based and genotype-environment association analyses identified loci showing the signature of selection, with one locus in common across five analyses. This locus was associated with the minimum temperature of the coldest month, a factor that defines the climate zones and was also significant in the common garden analyses. In addition, we documented introgression from Q. ellipsoidalis into Q. rubra, with rates of introgression correlated with the climate gradient. In sum, this study reveals signatures of selection at the quantitative trait and genomic level consistent with climate adaptation, a pattern that is more often documented at a much broader geographical scale, especially in long-lived wind-pollinated species.

Earthworms as a prey source for the insular snake Thamnophis sirtalis (Linnaeus, 1758)

Insular organisms typically evolve in closed, isolated systems; thus, changes caused by introduced species can drastically affect their ecology. Isle Royale National Park (Michigan, USA) is a remote island in Lake Superior that, until relatively recently, was void of earthworms. Its two native natricine snakes are known to feed on earthworms on the mainland; here, we confirm that eastern gartersnakes (Thamnophis sirtalis) are also feeding on earthworms on Isle Royale. It is unknown how this new important dietary source will impact the snake ecology on Isle Royale.

Pinus banksiana Lamb. Regeneration Patterns in a Lacustrine Dune System

Successional patterns in lacustrine sand dunes along Lake Superior begin with grass-dominated plant communities leading to the establishment of Pinus banksiana Lamb. as initial forests. Using maximum entropy models, we predicted P. banksiana seedling and sapling patterns within the Grand Sable Dunes, Pictured Rocks National Lakeshore, USA, based on slope, aspect, forest basal area, and vegetation types. Across the different vegetation types, there were variable probabilities of seedling and sapling occurrence. For both seedlings and saplings, the higher likelihoods of occurrence were observed in coastal pine barrens vegetation types. P. banksiana regeneration is occurring in the Grand Sable Dunes in the absence of fire, with seedlings establishing and saplings being recruited in a variety of vegetation types. With the greatest probabilities in barrens, there is likely a relationship with seed source and canopy density. Understanding regeneration patterns in dune ecosystems is necessary to predict the future forest arrangement and colonization of P. banksiana into the dunes. These results contribute insights into the dynamics of plant communities in lacustrine dune systems, specifically the establishment of P. banksiana seedlings in various vegetation types. Continued forest establishment and increasing P. banksiana density will influence endangered species and non-native species management strategies for Pictured Rocks National Lakeshore.

A hierarchical model of persistent and transient growth variation applied to Lake Superior lake trout

Variability in individual fish growth both within and among populations can interact with mortality to affect variation in size-at-age and other critical features of populations. Herein, we developed and applied mixed-effects, hierarchical growth models to back-calculated length-at-age data from six lake trout (Salvelinus namaycush) populations in Lake Superior to quantify how growth variation was attributable to persistent sources within and among populations and transient (short-term and not consistent over time for an individual) sources. Persistent variation in growth among individuals explained more variability in length-at-age than transient variation, and most of this variation was within populations rather than among populations. Simulations showed that the modeling approach could robustly estimate most growth function parameters, even with mismatches between true and assumed among-population covariation, although a higher number of populations enabled better estimation of certain population-level parameters. An implicit assumption to our interpretation was that lake trout populations in Lake Superior had not experienced substantial size-selective mortality, so their length-at-age patterns largely reflected growth variation rather than size-selective mortality. This assumption should be tested when interpreting future applications of this growth modeling approach.

Synthesis of recent research and attributes of recovered lean Lake Trout populations in Lake Superior, 1993–2022

AbstractObjectiveThis purpose of this paper is to describe the characteristics of recovered Lake Trout Salvelinus namaycush populations in Lake Superior by describing its population dynamics, ecology, and recent research and management activities since 1993, when Lake Trout were declared rehabilitated.MethodsData from commercial fisheries, recreational fisheries, agency stocking reports, and natural resource agency fishery‐independent survey data along with published research findings on lean Lake Trout in Lake Superior between 1993 and 2022 were synthesized and reported.ResultCurrently, Lake Trout populations are self‐sustaining and lightly exploited with only a few areas with elevated total mortality rates. The total annual mortality has been far below the target maximum range of 42–45%. Furthermore, stocking of hatchery Lake Trout is no longer necessary.ConclusionWe have learned from research and management experience that the regulatory role of Lake Trout in the Great Lakes is critical to proper ecosystem function. Thus, continued commitment from natural resource agencies to cooperate and implement effective management actions is required to preserve the accomplishments of lakewide recovery of Lake Trout populations.

Evaluation of eDNA qPCR monitoring as an early detection tool for a non-native mysid in Great Lakes Waters

Early detection of aquatic invasive species (AIS) is vital to cost-effective prevention of their spread in the Great Lakes. Unfortunately, AIS surveillance has been generally too slow and geographically limited to support this purpose. Environmental DNA (eDNA) detection using quantitative polymerase chain reaction (qPCR) offers more rapid and affordable detection of likely AIS presence, but it does not directly discern live/dead status. Vital status verification using conventional surveys following positive eDNA qPCR detections could resolve this barrier, but only if the latter are adequately reliable and sensitive. Here we explore the reliability and sensitivity of eDNA qPCR monitoring for the bloody red shrimp (Hemimysis anomala), an AIS established in the southern Great Lakes but not yet widely distributed in Lake Superior, against conventional microscopy-based methods. We conducted this comparison using 1) harbor water from Muskegon Lake, MI where H. anomala is established, and 2) raw ballast water from ships transporting ballast from lower Lake Michigan to western Lake Superior. Our studies showed positive eDNA qPCR detections of H. anomala in all harbor and ballast samples for which conventional detection results were positive, and in some samples for which conventional results were negative. These results suggest that qPCR assays with adequate specificity could be an important tool in support of more effective and affordable early detection of target species in Great Lakes water, especially when combined with confirmatory conventional monitoring.

Climate refugia along Lake Superior’s shores: disjunct arctic–alpine plants rely on cool shoreline temperatures but are restricted to highly exposed habitat under climate warming

Abstract Climate refugia can serve as a remnant habitat or stepping stones for species dispersal under climate warming. The largest freshwater lake by surface area, Lake Superior, USA and Canada, serves as a model system for understanding cooling-mediated local refugia, as its cool water temperatures and wave action have maintained shoreline habitats suitable for southern disjunct populations of arctic–alpine plants since deglaciation. Here, we seek to explain spatial patterns and environmental drivers of arctic–alpine plant refugia along Lake Superior’s shores, and assess future risk to refugia under moderate (+3.5 °C) and warmest (+5.7 °C) climate warming scenarios. First, we examined how the interactive effects of summer surface water temperatures and wind affected onshore temperatures, resulting in areas of cooler refugia. Second, we developed an ecological niche model for the presence of disjunct arctic–alpine refugia (pooling 1253 occurrences from 58 species) along the lake’s shoreline. Third, we fit species distribution models for 20 of the most common arctic–alpine disjunct species and predicted presence to identify refugia hotspots. Finally, we used the two climate warming scenarios to predict changes in the presence of refugia and disjunct hotspots. Bedrock type, elevation above water, inland distance, July land surface temperature from MODIS/Terra satellite and near-shore depth of water were the best predictors of disjunct occurrences. Overall, we predicted 2236 km of the shoreline (51%) as disjunct refugia habitat for at least one species under current conditions, but this was reduced to 20% and 7% with moderate (894 km) and warmest (313 km) climate change projections.

Model-based prediction of water levels for the Great Lakes: a comparative analysis

This comprehensive study addresses the correlation between water levels and meteorological features, including air temperature, evaporation, and precipitation, to accurately predict water levels in lakes within the Great Lakes basin. Various models, namely multiple linear regression (MLR), nonlinear autoregressive network with exogenous inputs (NARX), Facebook Prophet (FB-Prophet), and long short-term memory (LSTM), are employed to enhance predictions of lake water levels. Results indicate that all models, except for FB-Prophet, perform well, particularly for Lakes Erie, Huron-Michigan, and Superior. However, MLR and LSTM show reduced performance for Lakes Ontario and St. Clair. NARX emerges as the top performer across all lakes, with Lakes Erie and Superior exhibiting the lowest error metrics—root mean square error (RMSE: 0.048 and 0.034), mean absolute error (MAE: 0.036 and 0.026), mean absolute percent error (MAPE: 0.021% and 0.014%), and alongside the highest R-squared value (R2: 0.977 and 0.968), respectively. Similarly, for Lake Huron-Michigan, NARX demonstrates exceptional predictive precision with an RMSE (0.029), MAE (0.022), MAPE (0.013%), and an outstanding R2 value of 0.995. Despite slightly higher error metrics, NARX consistently performs well for Lake Ontario. However, Lake St. Clair presents challenges for predictive performance across all models, with NARX maintaining relatively strong metrics with an RMSE (0.076), MAE (0.050), MAPE (0.029%), and R2 (0.953), reaffirming its position as the leading model for water level prediction in the Great Lakes basin. The findings of this study suggest that the NARX model accurately predicts water levels, providing insights for managing water resources in the Great Lakes region.