Increases in Great Lake winds and extreme events facilitate interbasin coupling and reduce water quality in Lake Erie

Climate change has increased water temperature and altered wind-driven water movements in aquatic systems^1,2. This applies not only to the mean conditions^3,4, but also to the frequency of extreme events (i.e., near the upper ends of the range of observed values⁵, > 80th percentile). For example, high air temperature or powerful winds^5,6,7,8,9 has affected the behaviour of surface gravity waves¹⁰. Understanding the changes in wind and wave climate provides insight into the prediction and management of climate change impacts related to coastal dynamics, such as coastal erosion and sediment budgets, water motions, and biological responses^6,11,12. Several studies on the impacts of climate change on oceanic waves^12,13,14,15 have been undertaken, including a recent study¹⁶ that shows a 0.41% annual increase in global wave power (WP; the transport of energy by waves, which represents the temporal variations of energy transferred from the atmosphere to the ocean surface motion over cumulative periods of time^16,17 (Eq. 2) due to stronger winds caused by increases in sea surface temperature. The oceanic wave climate also responds to global atmospheric phenomena (e.g., El Niño Southern Oscillation and the Atlantic Multidecadal Oscillation), in which sea surface temperature modifies wind patterns and storm cyclogenesis^12,18,19,20. A systematic long-term assessment of climate warming impacts on waves in lakes remains to be undertaken, but should include winds, which are one of the principal sources of mechanical energy for lake circulation and interbasin coupling (e.g., exchange)^21,22,23,24.

The Laurentian Great Lakes, which consist of lakes Superior, Michigan, Huron, Erie, and Ontario (Fig. 1a), are the largest group of freshwater lakes on Earth; they contain 21% of the world’s volume of fresh surface water. These lakes have been affected by climate change in several ways including increased surface water temperature, longer summer stratification related hypoxia (i.e., dissolved oxygen [DO] concentrations < 2 mg L⁻¹ observed in Lake Erie, Lake Superior (Green Bay) and Lake Michigan (Saginaw Bay)) as well as increased occurrence of weather extremes and harmful algal blooms (HAB), which degrade water quality^3,7,25,26. Lake Erie is the shallowest (average depth ~ 19 m) of the Great Lakes, has the shortest residence time (~ 3 years)⁷, and yet its watershed is home to one-third of the total human population of the Great Lakes basin²¹. Not surprisingly, it has a long history of eutrophication and hypoxia related to urbanization, industry, and agriculture^25,26. For example, the largest HAB recorded in Lake Erie occurred in 2015 when the algal bloom spread ~ 200 kms across most of the lake²⁷. Moreover, during a previous HAB in 2014, half a million people living around the southwestern basin of the lake suffered a domestic water use ban due to hepatotoxins produced by the cyanobacterium Microcystis²⁸.

Lake Erie is not uniform spatially, rather it consists of three interconnected basins (Fig. 1a) with dominant wind forcing from synoptic high winds (5–10 days period) and diurnal sea breeze in the coastal and shallow regions during summer^29,30,31. Thermal structure in the basins are different in the summertime. The relatively shallow western basin thermally stratifies diurnally or for short periods of calm and warm atmospheric conditions (~ 4.5 days) that can result in hypoxia^30,32 due to oxygen depletion from the mineralization of organic matter in the hypolimnion and within the sediment. The deeper central and eastern basins stratify seasonally^21,33,34, which in the former has increased in spatial extent since the mid-1990s; in recent times, this has been primarily due to increased dissolved reactive phosphorus loads from agriculture³⁵. Hypoxia, especially in the western basin, can have significant negative impacts on fish distributions^36,37 and benthic invertebrates including burrowing mayflies in the genus Hexagenia^38,39. Whereas the eastward hydraulic flow from Detroit River into the central basin is the dominant flow direction in the western basin, the intrusion of hypolimnetic water from the central basin from south of Point Pelee (Fig. 1b) can cause episodic stratification in the western basin during August^22,40,41. This occurs when high winds (> 8 m s⁻¹) from the south and southwest that are the common wind directions over the Great Lakes²³, tilt the thermocline upward in the western and northern part of the central basin due to Ekman transport of surface water southward^{22,38,42,43,44,45}. As this hypolimnetic water upwells into shallower depths it can be transported counter clockwise by the alongshore surface currents moving to the west³². If there is a calm period following the high winds, the upwelled water in the northwestern part of the central basin will flow southward because of the pressure gradient and also in a clockwise direction (to the west) because of the Coriolis effect, and so will intrude into the western basin (i.e., a geostrophic flow) opposite to the hydraulic flow from the Detroit River (Fig. 1c)^22,32,46. This causes the rapid (on the order of hours) formation of a thermocline within the northeastern portion of the western basin (Pelee Passage) due to the intrusion of low temperature bottom water^22,42, which can also be hypoxic²² or anoxic (i.e., DO (approx) 0) at the sediment surface²² and contain high soluble reactive phosphorus concentrations (SRP; 0.02–0.05 mg L⁻¹)^47,48,49. Low values of sediment oxygen uptake are observed during these events in the western basin due to stratification and weak bottom shear and turbulence, which results in thicker diffusive sublayer²².

Interbasin exchange has been observed in lakes with multiple basins elsewhere (e.g., Lake Geneva⁵⁰, Nechako Reservoir⁵¹) as well as in the Great Lakes region (e.g., Muskegon Bay⁵², Green Bay⁵³, Kempenfelt Bay⁵⁴, Pere Marquette River⁵⁵). In Lake Michigan, for example, high winds can lead to coastal upwelling into Muskegon Lake causing episodic hypoxia⁵². In case of Lake Erie, interbasin exchange was identified as the dominant cause (63%) of hypoxia in the northeastern portion of the western basin during biweekly fishing trawls in August over the past 30 years²². However, there are no long-term continuous water quality observations to assess the occurrence and historic trends in these hypoxic events. Extreme winds prevailing from upwelling favourable directions (i.e., from the south and southwest) can generate strong surface waves and water currents through momentum flux at the air–water interface. Therefore, WP can be used as an indicator or proxy (but not the cause) of interbasin exchange. Here, we examine the historical trends in water temperature, winds and resultant waves in the context of climate change in the summer in the Great Lakes (Fig. 1a) with an emphasis on the western basin of Lake Erie (Fig. 1b). We examine data for August, which is the month when hypoxia is most likely to occur in dimictic north temperate lakes before the fall turnover, and when large HAB have been observed in the western basin of Lake Erie. August is also the time when the spatial extent of hypoxia in the central basin is the largest and when the aforementioned upwelling into the western basin is likely to occur^22,40,56. The data examined are from buoys with the longest historical records (Fig. 1a and Table S1). We examine winds from the south and southwest directions, which are the common wind directions over the Great Lakes during August, and which are favourable for upwelling into the western basin of Lake Erie. The results show that the WP in Great Lakes has increased in the past 40 years. A pattern in WP (a proxy for hypoxic upwelling events into the western basin of Lake Erie) has also increased in frequency over this time, which has implications for the water quality (e.g., dissolved oxygen and total phosphorus) of the lake. The increased frequency of interbasin upwelling was confirmed using historical records of lake bottom water temperature (LBT), as well as dissolved oxygen and total phosphorus concentrations. This is the first time that WP has been identified as an indicator of climate change-driven biogeochemical responses in lakes.

Long-term trends in WP and LST in the Great Lakes

First, we investigate the historical trends in average lake surface temperature (LST), wind, and waves in the Great Lakes during August. Results show that LST and LST_w (hereinafter subscript ”w” is used to denote the variables measured during upwelling favourable winds from 180° to 270°, clockwise from north) have both increased significantly (p < 0.05) by ~ 0.5 °C decade⁻¹ (> 0.2% year⁻¹) since 1980, although lower trends were observed in lake Erie and Michigan (Figs. S1–S5 ((a) and (b)) and Table S1). These changes in the LST correspond to a warming trend in air temperature (Tair); the average Tair over the Great Lakes increased significantly by ~ 0.4 (pm) 0.2 ((pm) standard error) ^oC decade⁻¹ since 1980 (Fig. S6a,b). There was an associated significant increase in wind speed (W) over the Great Lakes during August (W_w) of ~ 0.4 (pm) 0.1 m s⁻¹ decade⁻¹ for winds from the south and southwest (Figs. S1–S5 ((c) and (d)) and Table S1). Consequently, the wind stress associated with wind from the south and southwest over the water surface of the Great Lakes (({tau }_{w}=0.0012{rho }_{air}{W}_{w}^{2}), where ({rho }_{air})=1.22 kg m⁻³ is the density of air⁵⁷, and the wind speed is measured 10 m above the water) increased significantly by 0.006 (pm) 0.002 Pa decade⁻¹ during August (3.0 (pm) 0.9% year⁻¹; Figs. S1–S5 ((e) and (f)) and Table S1).

The effects of increased wind stress can also be seen in wave power, which is a function of the square of significant wave height (the mean value of the largest third of the wave heights during typically 1 h, SWH) and the wave period (({T}_{p}); i.e., (WP propto {{T}_{p} times SWH}^{2})); and changes in wind are reflected in wave power ((WP propto {W}^{2.4}) and (propto {W}^{5}) for developing and fully developed waves, respectively; see “Materials and methods”). The average SWH and SWH_w in the Great Lakes during August have increased significantly by 0.03 (pm) 0.02 and 0.04 (pm) 0.03 m decade⁻¹, respectively (i.e., ~ 1.0 (pm) 0.8% and ~ 1.7 (pm) 1.5% year⁻¹, respectively), and this is largely driven by the increase in the frequency of extreme surface winds⁵⁸ (Figs. S1–S5g and h; WP responds to changes in mean values, but it is more sensitive to extreme events because WP (propto { SWH}^{2})¹⁶). Consequently, the average WP and WP_w in the Great Lakes during August have increased by ~ 0.04 (pm) 0.02 and ~ 0.06 (pm) 0.03 kW m⁻¹ decade⁻¹, respectively (i.e., ~ 1.0 (pm) 0.6% and ~ 2.0 (pm) 0.9% year⁻¹, respectively; Fig. 2). In Lake Erie, WP_w during August increased significantly by 0.02 (pm) 0.01 kW m⁻¹ decade⁻¹ (1.4 (pm) 0.2% year⁻¹; Fig. 2 and Table S1; the increasing trend in WP = 0.02 (pm) 0.02 or 0.5 (pm) 0.1% was not statistically significant). It is relevant to note that these results are based on observations from a single buoy per lake; the one with the longest available data records (Fig. 1a and Table S2). However, the wind records and historical wave trends between buoys Sta. NDBC 45005 and Port Stanley in Lake Erie (Fig. 1a), which are ~ 130 km apart, are consistent based on the available records. Specifically, wind speed and direction in 2018 have Pearson correlation coefficients, r > 0.6 (Fig. S7a,b, respectively); W_w and WP_w are also correlated with r = 0.51 and 0.67, respectively, during August of 1990–2018 and the buoys show similar temporal increases in WP_w (~ 0.025 (pm) 0.02 and 0.02 ± 0.01 kW m⁻¹ decade⁻¹ in Port Stanley and Sta. NDBC 45005, respectively). The trends in historical LST_w and WP_w are related statistically (i.e., higher mutual information; Fig. S8) similar to the relationship described for global sea surface temperature and oceanic WP used as an indicator of climate change¹⁶.

The long-term variations in WP and LST may be related to the global atmospheric phenomena. The LST_w anomaly in all the lakes show an increasing trend beginning in 1995 (Fig. S9a), which corresponds to the switch from the negative mode of the Atlantic Multidecadal Oscillation (AMO) to the positive mode (associated with increased tropical cyclone activity and stronger westerly winds) between the 1980s and the early 2000s (Fig. S9b)¹⁶. Both the WP_w and LST_w anomaly are positively correlated with the AMO (r ~ 0.50 and ~ 0.55, respectively, since 1990). Similar to global oceanic wave power¹⁶, peaks in WP_w in the Great Lakes are associated with strong El Niño years (i.e., Multivariate El Niño/Southern Oscillation (MEI) greater than 1.5; Fig. S9c,d), which can contribute to the enhanced wind energy due to increased cyclonic events¹⁶. MEI and WP_w in Great Lakes are generally correlated by r > 0.45 since 1990, however, the impacts of global atmospheric events on temperature and water dynamics of Great Lakes requires further study.

Episodic hypoxic upwelling events in the western basin of Lake Erie

We used historical records (Table S2) of long-term near-bottom water temperature (1998–2018) and dissolved oxygen (2007–2018) in the northeastern portion of the western basin of Lake Erie as well as wave observations in the western portion of the central basin (1980–2018 in Sta. NDBC 45005, Fig. 1) in August to determine the frequency of hypoxic upwelling events and the impacts of these events on the total phosphorus concentration in the northeast portion of the western basin. These analyses do not include the local hypoxia due to periods of calm and warm atmospheric conditions that may occur annually³¹ and, which are different than episodic upwelling events. Intrusion of cold hypoxic hypolimnetic water from the central basin into the western basin, following high winds from upwelling favourable directions, can cause a sudden drop (on the order of hours) in LBT and dissolved oxygen (DO) when the hypolimnetic water in the central basin is hypoxic²². The LBT time series in the western basin from 2017 to 2018 show that LBT decreased more than 3 °C in less than 12 h during upwelling events; e.g., 9–16, 18–22 and 26–31 August 2018 at Sta E (Fig. 3b) and 24–29 August 2017 at Leamington and Sta E (Fig. S10b). The records of LBT measured by the Ontario Ministry of Natural Resources and Forestry (MNRF) in August in Leamington Ontario between 1998 and 2018 detected 23 events of intrusion of cold water, which are consistent with upwelling (the blue symbols in Fig. 4a).

Using a simple assumption of direct-uniform flow of hypolimnetic water from the central basin, we estimate that there is high probability that hypoxic water can intrude into the western basin when the hypoxic zone of the central basin is less than ~ 12 km east of the Pelee Passage (i.e., the distance the hypolimnetic water travels with an average bottom current speed of ~ 0.10 m s⁻¹³² over ~ 1.5 d²²). We used Zhou et al.’s⁵⁶ estimated spatial extent of deep-water hypoxia in the central basin from 1987 to 2007 to predict the hypoxia of the upwelled hypolimnetic water. Their results indicated a spatially limited hypoxic zone in 1989 and 2005, when upwelled waters into the western basin would not be hypoxic as denoted in Fig. 4a.

Observations from the western portion of the central basin (Sta. NDBC 45005, Fig. 1) show that hypoxic upwelling events in the western basin follow waves from the south and southwest direction with increased WP; e.g., 8–9, 16–17 and 25–26 August 2018 (Fig. 3a) and 22–24 August 2017 (Fig. S10a). The significant wave height and mean wave period during these periods was typically (ge) 0.70 m and (ge) 3 s, respectively (Figs. 3a and S10a). These strong waves were driven by extreme winds > 8 m s⁻¹ from similar directions, which corresponds to the ~ 80th percentile of wind speeds and is greater than the sum of the average and standard deviation of the wind speed (~ 6 and 2 m s⁻¹, respectively). This wind threshold is consistent with Rao et al.’s⁴⁴ wind speed that led to upwelling, which resulted in a fish kill along the north shore of the central basin in 2012.

We used a least-square method to find a wave pattern (i.e., wave direction, duration, and power) that could be applied to predict the number of upwelling events that could be hypoxic between 1998 and 2018 based on LBT observations. A rapid decrease in the LBT at both Sta E and Leamington (12 km vs. 20 km from the Pelee Passage, respectively) occurred during events in which the average WP was > 0.44 kW m⁻¹ (i.e., 22–24 August 2017; Fig. S10a,b). The model predicted 25 upwelling events at Leamington (dark bars in Fig. 4a) of which 23 were observed (as stated above; no data were available for 2012; blue circles in Fig. 4a) for waves from south and southwest that lasted for at least 15 h with an average wave power greater than 0.37 kW m⁻¹. Of the 23 observed events, the model predicted 21 events providing a root mean square error [RMSE] of 0.20 events. We validated the model predictions using the biweekly DO measurements from MNRF cruises between 2007 and 2018, which happened to sample 17 of the 23 observed events of low LBT. We note, however, that two hypoxic upwelling events were also recorded outside the study period, i.e., early September; this supports the study’s focus on August. Hypoxic conditions (DO < 2 mg L⁻¹) were observed for 14 of these 17 upwelling events (82%). Importantly, there was no evidence of upwelling hypoxia in 2005, which is consistent with the limited spatial extent of hypoxia in the central basin in that year⁵⁶; unfortunately, there are no observations in the western basin for 1989.

We extended the wave power method to the wave records since 1980 to estimate the frequency of the hypoxic upwelling events into the western basin. Figure 4a shows an increasing trend in the number of predicted upwelling events; i.e., from 0.4 events year⁻¹ in 1980 to > 1.6 events year⁻¹ in 2018 based on a 10-year moving average. Specifically, 21 of 49 (~ 43%) upwelling events in the last four decades have occurred in the past 10 years. Thirty-two of these were strong events with WP > 0.44 kW m⁻¹, 15 of which (~ 47%) occurred after 2009. Interestingly, this pattern in wave power (i.e., waves from south and southwest that last for > 15 h with an average WP > 0.37 kW m⁻¹ from the historical data) was also observed in August 1980 (Fig. 4a), when the LBT dropped following rapid formation of a thermocline, which at the time was attributed to the upwelling of hypolimnetic water from the central basin^40,42. These results indicate that an increase in extreme winds from south and southwest during August, over the last four decades, has resulted in more frequent upwelling from the central basin into the western basin and consequently a greater number of episodic hypoxic events in that part of Lake Erie.

The effect of upwelling on phosphorus concentrations was examined through an analysis of the water column-average total phosphorus (TP) observations from biweekly cruises conducted by the MNRF at station W5 (Fig. 1b). We examined the available data recorded between 15 July and 15 September from 2000 to 2018 (3–5 records year⁻¹; 66 observations in total), which is a period in which linear patterns in TP vs. sampling date were not evident (p >  > 0.05). The z-score (standard deviate) was determined for the data within a given year (({mathrm{Z}}_{mathrm{TP}}=left(mathrm{TP}-{mathrm{TP}}_{mathrm{mean}}right)/mathrm{SD}), where ({mathrm{TP}}_{mathrm{mean}}) is the annual average of TP and SD is the standard deviation). Positive ({mathrm{Z}}_{mathrm{TP}}) values (i.e., (mathrm{TP}>{mathrm{TP}}_{mathrm{mean}})) were observed in 11 cases in which the sampling occurred < 3 days after upwelling events, 30 of which were predicted in Fig. 4a. Six of these cases were among the seven occurrences of elevated TP concentrations (({mathrm{Z}}_{mathrm{TP}}) > 1) observed during 5 August–8 September sampling (black solid circles in Fig. 4b). Statistical comparison revealed that the average ({mathrm{Z}}_{mathrm{TP}}) was significantly higher during upwelling vs. non-upwelling samples (i.e., 0.95 ± 0.18, n = 11 vs. − 0.26 ± 0.12, n = 25; ANOVA F_1,34 = 29.64, p < 0.001; Fig. 4c). These results indicate elevated TP in the northeastern portion of the western basin of Lake Erie following the upwelling events of high phosphorus containing hypoxic water from the central basin.

Source: Ecology - nature.com