Abstract
End of summer stratification and onset of ice-on are shifting later across the mid-latitudes, but the lake-specific and climatic drivers of lake thermal structure in autumn remain unclear. We used high-frequency water temperature data (2007–2021) from nine sub-alpine (>600 m ASL) lakes in the Northeastern US to investigate the drivers and coherence of ecologically important autumn processes. The end of summer stratification was highly variable across lakes and mediated by DOC and lake depth. The timing of end of stratification affected the onset of continuous autumn mixing; however, mixing was decoupled from ice-on. Ice-on was highly coherent and driven by air temperature. Extreme weather events with high winds, heavy precipitation, and anomalous air temperatures drove early onset of mixing, re-stratification events, and freeze-melt events. The variability of thermal structure during autumn demonstrates that climate change will not uniformly shift the start and end of autumn season later in the year.
Plain Language Summary
Lakes in the northern hemisphere with seasonal ice-cover can be characterized by three common seasonal thermal structure events: isothermal mixing in the spring, thermally stratified layers in the summer, and isothermal mixing in the fall. Due to warming air temperatures, summer stratification is lasting longer and ice-on is occurring later. In our regional study of small lakes that mix twice a year, we found that the end of summer stratification is highly variable across lakes. This variability is determined by bottom water temperatures, lake depth, and dissolved organic carbon concentrations. In contrast, the timing of ice-on was highly coherent and driven by November air temperature. Extreme weather events caused intermittent ice-on and thaw, re-stratification, and early mixing events. These changes to the timing of autumn thermal structure affect lake organisms and water chemistry.
Key Points
-
End of summer stratification was the most variable autumn thermal structure event and was driven by lake morphometry and DOC
-
The timing of end of stratification influenced the onset of continuous autumn mixing, and autumn mixing was decoupled from ice-on
-
Extreme weather events led to coherent disruptions of autumn thermal structure, such as re-stratification and intermittent ice-on events
1 Introduction
Identifying periods of seasonal mixing and stratification is important to understanding lake ecology and biogeochemistry. In dimictic lakes, autumn mixing redistributes heat, nutrients, dissolved gases, and planktonic biomass for the first time following summer stratification (Dakos et al., 2009; Doda et al., 2022; Sommer et al., 1986), and the timing and duration of this period can impact trophic interactions, food webs, dissolved oxygen dynamics, and lake metabolic balance (Flaim et al., 2020; Hamdan et al., 2021; Hebért et al., 2021). Despite the ecological significance of lake processes during autumn, our understanding of autumn thermal structure lags behind other seasons (Ferrato et al., 2025). As autumn air temperatures and storm frequency increase across the mid-latitudes (Cohen et al., 2014; Wang et al., 2021), assessing the interannual variability and drivers of autumn thermal structure is essential for understanding lake dynamics in a changing climate.
In dimictic lakes, autumn is generally characterized as a mixing period driven by surface heat loss that occurs between the dissolution of the summer epilimnion and the onset of ice-cover (MacIntyre & Melack, 2009). End of summer stratification and onset of autumn mixing turnover are often used interchangeably (Woolway, Denfeld, et al., 2021); however, many smaller, dimictic lakes exhibit a variable transition period of discontinuous mixing between the end of summer stratification and onset of continuous mixing (Alvarez-Cobelas et al., 2005; Smits et al., 2021). The drivers of autumn mixing have been quantified with monthly profile data (Nürnberg, 1988) or observations on one or two lakes (Alvarez-Cobelas et al., 2005; Flaim et al., 2020), but the climatic and lake-specific factors that control the shift from summer stratification to autumn mixing remain unexplored. By integrating a regional spatial scale with high-frequency temporal resolution, we can disentangle the climatic and lake-specific drivers (Magnuson et al., 1990) of autumn thermal structure in small, dimictic lakes.
Thermal structure phenology is altered by the press of long-term climate change (Woolway & Merchant, 2019) and pulse of extreme weather events (Bender et al., 1984; Klug et al., 2012). The autumn mixing window could be shortened as a result of longer summer stratification (Niedrist et al., 2018) or extended due to later onset of ice-on (Flaim et al., 2020), which might imply autumn mixing simply starts and ends later. However, heterogenous rates of seasonal air temperature change could shorten or lengthen the duration of autumn weather (Winslow et al., 2017) and asymmetrically alter the autumn mixing window (Huang et al., 2024). In the short-term, extreme weather events, such as autumn storms, heatwaves, and weather whiplash (rapid air temperature changes) (Casson et al., 2019) events, are increasing in frequency and severity and can abruptly alter thermal structure (Klug et al., 2012; Sadro & Melack, 2012). This interaction of long-term climatic presses and extreme event pulses could shift the autumn season in lakes toward a highly variable combination of mixing, re-stratification, and freeze-melt events.
To investigate the progression and variability of lake thermal structure during autumn, we identified five key parameters to explore: end of summer stratification, onset of continuous mixing, number of days between end of summer stratification and onset of continuous mixing, total mixing days, and ice-on. Using a decadal water temperature and geochemical data set from nine mountain lakes in the Northeastern US, our objectives were to assess: (a) coherence of autumn events across a region, (b) the effects of climate and lake-specific characteristics on the progression of thermal structure in autumn, and (c) the effect of extreme weather events on thermal structure in autumn. These lakes provide an ideal location to assess drivers of autumn phenology in a globally abundant lake type: wind-sheltered, dimictic, seasonally ice-covered lakes (Verpoorter et al., 2014). Aquatic systems in relatively low-elevation mountain ranges, such as the Appalachian Mountains, are less impacted by anthropogenic land-use than lowland systems and can serve as climate reference sites for the numerous small lakes in north temperate and boreal ecosystems (Landers et al., 1988). Despite overall warming trends in the Northeastern US, the autumn conditions here are highly variable (Crimmins et al., 2023) and offer an ideal location to assess a range of autumn weather patterns representative of conditions across the mid-latitudes.
2 Methods
We analyzed thermal structure phenology during the autumn season in nine dimictic mountain lakes in Maine across a range of DOC concentrations (2–13 mg/L). All lakes are oligotrophic and fit the functional definition of lakes (Table S1 in Supporting Information S1) (Kahl & Scott, 1994; Richardson et al., 2022). The lakes span 150 km across Maine's Northern Climate Division (Figure 1a) and represent sub-alpine aquatic systems in the Northeastern US (Nelson et al., 2021).
(a) Location of study lakes in Maine and the contiguous US. Dark orange topographical lines show areas greater than 600 m ASL. The red circle shows the location of the lakes within Maine. The photograph of the Horns Pond summit illustrates the subalpine vegetation and high relief that are characteristic of mountain lakes in Maine (PC: Rachel Hovel). (b) Average autumn air temperature for the state of Maine from 1896 to 2022 (NCEI, 2024). Breakpoint analysis shows a break in slope in 1992 (dashed, vertical line), accelerated rate of change (red, solid line), and standard error (light orange ribbon).
Temperature sensors (HOBO Pendant Temperature/Light Data Logger) were deployed from 2007 to 2021 (Figure S1 in Supporting Information S1). Loggers were placed 0.5 m below the water's surface and 0.5 m above the lake bottom and logged measurements every 30 min. End of summer stratification is the first total mixing event, defined as the first time after sustained summer stratification (Wetzel, 2001; Woolway, Sharma, et al., 2021) when the difference between surface and bottom water temperature is less than 0.65°C (Gray et al., 2019; Read et al., 2011). The onset of continuous autumn mixing (hereinafter, onset of continuous mixing) is the first day of sustained mixing, defined as less than 0.65°C difference between surface and bottom loggers, for at least 70% of all 30-min observations within a day. This matches or is more strict than temperature gradients used in similar systems (Gray et al., 2019) and was validated with higher resolution temperature and density gradients (Text S1 and Figures S2–S5 in Supporting Information S1).
The date of persistent ice-on was calculated as the onset of inverse stratification, defined here as the difference between surface and bottom water temperature less than −0.65°C. This method is consistent with those used in wind-sheltered, relatively small lakes (Pierson et al., 2011; Smits et al., 2021; Yang et al., 2021) and was validated with satellite imagery (Text S2 in Supporting Information S1). DOC samples for each lake were collected once per year in either summer or autumn, consistent with methods in Gavin et al. (2018) and Nelson et al. (2021). Daily climate data for each watershed were sourced from the Parameter-elevation Regressions on Independent Slopes Model (PRISM) 4 km resolution gridded temperature and precipitation data set (Daly et al., 2008) and validated with the 800 m PRISM data set and measurements from a weather station deployed in the Mountain Pond watershed (Text S3 and Figures S6–S10 in Supporting Information S1). To examine long-term temperature trends, we used the NOAA U.S. Climate Divisional Database (NCEI, 2024) (Text S3 in Supporting Information S1).
All data analyses were conducted in the R statistical environment (R Core Team, 2024). We assessed temporal trends for autumn events and DOC using Theil-Sen's slope and significance at p-value < 0.05 from the “trend” package (Pohlert, 2023). To assess the breakpoint and change in autumn (September-November) air temperature in Maine, we used the “segmented” package (Muggeo, 2003). To assess the temporal coherence (Gu et al., 2021; Magnuson et al., 1990) of autumn thermal structure events among lakes, we calculated the Spearman's rank correlation coefficients for all lake pairs. Adjusted p-values based on Benjamini and Hochberg (1995) reduced the probability of Type 1 errors for multiple pairwise tests performed on a single set of data and were calculated using the “p.adjust” function in the “stats” package (R Core Team, 2024).
End of summer stratification and onset of continuous mixing can occur simultaneously or as distinct events up to two months apart (Figure S11 in Supporting Information S1). When these events are distinct, a period of discontinuous mixing occurs in between. For lakes and years where these events were distinct, we used piecewise structural equation modeling (SEM) from the “piecewiseSEM” package to evaluate the effect of climatic and lake-specific controls on thermal structure progression during autumn (J. Lefcheck et al., 2015). SEM is suited to address a chain of effects, such as the extent to which the progression of mixing and stratification events are mediated by antecedent events (J. S. Lefcheck, 2016). PiecewiseSEM is also suited to smaller sample sizes (Grace et al., 2015). Component models were built with hierarchical linear mixed models, using lake as a random variable (Pinheiro et al., 2024). We calculated marginal and conditional R2 with the “MuMIn” package to parse out the variability explained by just fixed effects and the combination of fixed and random effects, respectively (Bartoń, 2024). Potential explanatory variables were selected based on the existing literature and availability within our data set (Text S4 and Table S2 in Supporting Information S1). Initial models were refined using Shipley's test of directed separation and Akaike information criterion, and Fisher's C statistic assessed global model fits (Shipley, 2013). For lakes and years where end of summer stratification and onset of continuous mixing were simultaneous events, we used hierarchical linear mixed models to compare this subset of data with the component models from the piecewiseSEM.
To assess whether extreme weather events elicited coherent lake responses in autumn thermal structure, we calculated the number of days until thermal structure response following five-well documented extreme events: a Nor'easter in October 2011 and subsequent heatwave (Casson et al., 2019), Hurricane Sandy in October 2012, and strong October wind storms in 2017 and 2019 (Simonson et al., 2020).
3 Results
3.1 Trends in Climate, DOC, and Autumn Thermal Structure
Autumn air temperature in Maine increased from 1896 to 2022. Breakpoint analysis revealed that the rate of change increased by 0.04°C/year (from 0.007 to 0.05°C/year) (p-value <0.05) in 1992, with a standard error of 8.5 years (Figure 1a).
Over the study period, we largely found no temporal trends in DOC or the timing of autumn events in each lake (Figures S12–S15 in Supporting Information S1). The exception was Cranberry Pond, in which end of summer stratification (slope = −3.8 days/year; p-value < 0.05) and ice-on (slope = −1.67 days/year; p-value < 0.05) trended earlier, and DOC decreased (slope = −2.7 mg/year; p-value < 0.01). More data for Cranberry (12 years of DOC and ice-on observations compared to an average of 7 and 8 years, respectively, for the rest of the lakes) could influence the likelihood of identifying trends. End of summer stratification also trended earlier in Midway (slope = −2 days/year; p-value < 0.05).
3.2 Temporal Coherence
Across all lakes, end of summer stratification ranged from August 8 to October 23, and the onset of continuous mixing ranged from August 29 to October 28 (Figure 2). The duration of discontinuous mixing between these events ranged from zero to 63 days. End of summer stratification and onset of continuous mixing were discrete events in 74% of all observations and often occurred simultaneously in four of the five deepest lakes. In Cloud, end of summer stratification and onset of continuous mixing always occurred simultaneously. In Mountain, Midway, and South, these events occurred simultaneously in 80%, 50%, and 33% of the years sampled, respectively. Total number of mixing days ranged from 18 to 108 days. The onset of persistent ice-on ranged from November 6 to December 11.
Top: range of dates for end of summer stratification, onset of continuous mixing, and ice-on; y-axis is continuous. Bottom left: duration of discontinuous mixing (the number of days between the end of summer stratification and the onset of continuous mixing). Bottom right: the total number of mixing days. Lakes are ordered from shallowest (brown, <10 m deep) to deepest (blue, >10 m deep) in the plots (left to right).
Coherence of autumn events across lakes was lowest for early autumn events and increased as the season progressed (Figure 3). End of summer stratification was least coherent, followed by the onset of continuous mixing. Date of ice-on was the most coherent variable.
Timeseries of (a) end of summer stratification, (b) onset of continuous mixing, and (c) ice-on date from 2007 to 2021 for the nine study lakes to demonstrate coherence rather than trends over time. Spearman's mean rho and number of significant pairs are in the upper right corners.
3.3 Progression of Thermal Structure in Autumn
When end of summer stratification was distinct from the onset of continuous mixing, the variability in autumn events was well described by the piecewiseSEM model (Figure 4 and Table S3 in Supporting Information S1). Deeper lakes with higher DOC had lower August bottom water temperatures and later end of summer stratification. DOC had an indirect effect on end of summer stratification via August bottom water temperature and a smaller direct effect. The inverse relationship between July air temperatures and end of summer stratification was the smallest standard estimate in the model. Together, these two component models described 65% of the variability in end of summer stratification. October air temperature and precipitation and end of summer stratification date drove the onset of continuous mixing. No relationship was identified between the onset of continuous mixing and ice-on, which is predominantly driven by November air temperature. D-separation test revealed a relationship between warmer Octobers and earlier ice-on, and this path was specified as a correlated error and omitted from the model (Shipley, 2000). When end of summer stratification and onset of continuous mixing occurred simultaneously, August bottom water and autumn air temperature were the best predictors (Table S4 in Supporting Information S1). There was no relationship between onset of continuous mixing and ice-on, which was also driven by November air temperature (Table S5 in Supporting Information S1).
Piecewise structural equation model to quantify the progression of autumn season thermal structure as functions of lake-specific characteristics, climatic drivers, and antecedent conditions. Arrow size is relative to the standardized coefficients (solid arrow, significant (p < 0.05)) that can be used to compare relative strength of relationships. mR2 and cR2 refer to the variance explained by the fixed effects alone and full model (fixed and random effects), respectively. No significant relationship between onset of continuous mixing and ice-on was identified.
3.4 Effect of Extreme Events
On 31 October 2011, below average air temperatures associated with a Nor'easter drove intermittent ice-on in Cranberry and Tumbledown that persisted until a heat wave with air temperatures greater than 10°C above the 40-year average (Figures 5a and 5c). This was followed by a second ice-on and melt event occurred in all six lakes (Figure 5c and Figure S16 in Supporting Information S1). Inverse stratification began one day after the mean air temperature fell below freezing, and intermittent ice-on began 5 days later on November 21, as air temperatures rapidly dropped. The November 30th melt was associated with a large rain event and air temperatures 17°C above the 40-year average (Figures 5a and 5b). In late October 2012, all eight lakes re-stratified during the extreme heat, rain, and wind that corresponded with two storms passing through Maine (Figures 5d–5f). The onset of coherent re-stratification correlated with air temperatures rising above 5°C 4 days before Hurricane Sandy and persisted through Sandy's heavy rain and high winds. Strong windstorms on 30 October 2017 and 17 October 2019 correlated with synchronous mixing events (Figures S16 and S17 in Supporting Information S1). Cloud, Eddy, Mountain, and South temporarily re-stratified during another strong windstorm on 31 October 2019 (Figure S17 in Supporting Information S1).
Examples of effects of extreme weather events on thermal structure during the autumn season. Left column: Freeze-melt events during a cold Nor'easter and subsequent weather whiplash event. Right column: Coherent re-stratification event during high air temperature, precipitation, and wind speeds. (a, d) PRISM daily average air temperature from all watersheds in 2011 (2012) (gold) and the entire study period (1981–2022) (turquoise), (b, e) PRISM daily precipitation averaged across all watersheds in 2011 (2012) (dark purple) and the entire study period (1981–2022) (light purple), and (c, f) surface (red) and bottom (orange) water temperature time series. Blue ribbons reflect intermittent ice-on during the Nor'easter and coherent freeze-melt during the weather whiplash event. Yellow ribbons reflect coherent re-stratification events in all lakes before and during Hurricane Sandy.
4 Discussion
Thermal structure events in autumn exhibit varying degrees of coherence and effects from lake-specific characteristics and climatic drivers. We provide evidence that the transition from summer stratification to autumn mixing across a region is highly variable and driven by lake-specific characteristics like DOC and maximum depth, while the synchronous transition from mixing to ice-on is constrained by regionally coherent air temperatures. This supports past findings on the weak temporal coherence of stratification phenology (Read et al., 2014) and the strong synchronicity of ice-cover (Benson et al., 2000).
4.1 Transition From Summer Stratification to Autumn Mixing Is Highly Variable
The end of summer stratification and onset of continuous mixing can occur simultaneously or up to 2 months apart. Only Cloud, the deepest and most northern lake, consistently underwent a rapid transition from summer stratification to mixing, while other deep lakes occasionally experienced discontinuous mixing. Speck was the only lake with maximum depth greater than 10 m where summer stratification and mixing were always distinct events. Speck is also the highest-elevation lake in the study set, so colder air temperature and higher winds could drive discontinuous mixing (Woolway, Sharma, et al., 2021). Shallow lakes (5–10 m maximum depth) always had a discontinuous mixing period.
End of summer stratification is highly variable and driven by August bottom water temperature and July air temperature. The role of late-summer bottom water temperature is in agreement with foundational research on the onset of autumn mixing (Nürnberg, 1988), but here, we have the temporal resolution to distinguish end of summer stratification from onset of continuous mixing and show that August bottom water temperatures also drives end of summer stratification. DOC and lake depth indirectly affect end of summer stratification via clarity-induced changes in mixing depth and stratification strength (Heiskanen et al., 2015; Kling, 1988).
Latent, sensible, and radiative heat loss from lakes is driven by decreasing autumn air temperatures (MacIntyre & Melack, 2009), and as such, air temperature had the strongest effect on mixing. August bottom water temperature affected the onset of continuous mixing in both groups, but only indirectly in lakes with a discontinuous mixing period. Late summer hypolimnion temperatures are correlated with ice-out and duration of spring mixing (Dugan, 2021; Oleksy & Richardson, 2021), so it is possible that spring dynamics have a cross-seasonal impact on early autumn thermal structure and contributed to some of the unaccounted variability in our model. Precipitation was not important for lakes without a discontinuous mixing period. The greater stability (Butcher et al., 2015) and relatively smaller ratio of cold precipitation to total lake volume in relatively deeper lakes (Holgerson et al., 2022) likely increases their resilience to precipitation disturbance.
Wind shear effect on mixing is considered minor in wind-sheltered lakes (Fee et al., 1996), but wind data could have further constrained our models (Doda et al., 2024; Sadro et al., 2019). The 17 October 2019 windstorm instigated a coherent onset of mixing that was 10 days earlier than the study period mean, which suggests that high wind affects the onset of continuous mixing. Coherent re-stratification during two events with high wind, the 31 October 2019 windstorm and Hurricane Sandy in 2012, demonstrates that air temperature can outweigh the impact of strong winds. Although the end of stratification and onset of continuous mixing were relatively asynchronous across the region, synchronous stratification and mixing events were associated with storms. The increasing frequency and severity of autumn storms and heatwaves across the mid- and northern latitudes (Coumou & Rahmstorf, 2012; Francis & Skific, 2015; Walsh et al., 2020) could drive increased coherence in autumn mixing within a region while also increasing the frequency of coherent re-stratification and freeze-melt events.
4.2 Ice-On Is Highly Coherent and Decoupled From Mixing
Delayed ice-on can increase the duration of autumn mixing (Flaim et al., 2020; Smits et al., 2021), but we found that the onset of continuous mixing was not causally linked to the onset of ice-on. The decoupling of autumn mixing and ice-on reflects the indirect lake-specific drivers that contribute to onset of continuous mixing and the temporal heterogeneity of autumn air temperatures. For example, the inverse relationship between October air temperature and ice-on identified by D-separation tests suggested that warmer Octobers led to earlier ice-on. Instead, a warm October and late onset of continuous mixing does not preclude a cold November and early onset of ice-on because relatively small lakes require less time and energy to erode stratification and mix isothermally (Kirillin & Shatwell, 2016). This finding does not undermine correlations between warm autumn air temperatures and delayed ice-on (Smits et al., 2021). Alternatively, it highlights that autumn air temperatures are not temporally consistent within a season and the ecological memory within the autumn season is not persistent enough to couple the onset of continuous mixing and ice-on (Padisák, 1992).
4.3 Implications for Mixing Regime and Ecological Processes
Loss of annual ice-cover is expected to shift dimictic lakes toward monomictic or polymictic regimes (Woolway et al., 2020), but the variability for end of summer stratification suggests that changing water clarity could push shallower lakes toward a different mixing regime. DOC decreased by over 60% in Cranberry Pond, and the end of summer stratification trended earlier. Increasing clarity can reduce the temperature gradient and weaken stratification, thus driving an earlier end of summer stratification (Riis & Sand-Jensen, 1998). The direct and indirect roles of DOC in autumn season events and this anomalous shift toward earlier end of summer stratification in Cranberry support increasing water clarity as a mechanism to shift dimictic lakes to a polymictic regime that occasionally stratifies in the summer (Holgerson et al., 2022). However, given multi-decadal increases in DOC across a larger subset of Maine mountain lakes (Gavin et al., 2018), longer thermal structure timeseries could reveal prolonged summer stratification driven by thermal shielding (Bartosiewicz et al., 2019). As DOC drivers in regions impacted by acidification shift from atmospheric deposition dominance to climate dominance (Herreid et al., 2024), clarity driven shifts in stratification strength will increase variability in end of summer stratification.
Quantifying the drivers of autumn thermal structure across a region is critical for understanding the ecological consequences of shifting stratification and mixing phenology. We demonstrate that lake-specific characteristics mediate the transition from summer stratified conditions to autumn mixing, while regional climatic forcing drives a highly coherent onset of ice-on. As weather whiplash, extreme precipitation, and windstorms become more common in northern latitudes, they will induce freeze-melt, mixing, and re-stratification events, respectively, that will disrupt the expected progression of summer stratification to autumn mixing to ice-on. Given the importance of lake-specific drivers to early autumn events, it is unlikely that stratification and mixing phenology during the autumn season will homogenously shift to later onset. This variability in autumn mixing and stratification phenology can shift phytoplankton community dynamics (Köster & Pienitz, 2006; Saros et al., 2012; Stockwell et al., 2020), decouple nutrient and trophic interactions (Shatwell et al., 2008), and alter biogeochemical processes (Pilla et al., 2023; Sadro & Melack, 2012) with cross-seasonal implications for ecosystem structure and function (Chen & Folt, 1996; Hebért et al., 2021).
Acknowledgments
Funding was provided by NSF award 2021713 (J Saros, PI), Water Resources Sustainability Research Grant USGS 104b (J Saros PI, A Gavin, J Daly, R Hovel, S Birkel co-PIs), and the Maine Economic Improvement Fund Small Campus Initiative (R Hovel, PI, J Daly co-PI). The US Environmental Protection Agency Long-Term Monitoring project was funded by EPA ORD and EPA CAMD awards IAG 06HQGR0143 (2013–2018), DW-014-92478401 (2018–2023), and 84055301 (2023–2024) and processed through the New Hampshire Water Resources Research Center and US Geological Survey Grants/Cooperative Agreements G11AP20128 (2013–2018) and G18AP00030 (2018–2023). This manuscript has not been subjected to agency review and no official endorsement by any agency should be inferred. This work is supported by the Hatch project award ME-022403, from the U.S. Department of Agriculture's National Institute of Food and Agriculture. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and should not be construed to represent any official USDA or U.S. Government determination or policy. We thank the undergraduate students at the University of Maine-Farmington who helped to collect and curate data from 2012 to 2023 and everyone who participated in the High-Elevation Lake Monitoring (HELM) project from 1986 to 2023. This is contribution #2040 of the Institute of Environment at Florida International University.
Open Research
Data Availability Statement
Data and metadata is available in the Environmental Data Initiative data repository: Gavin et al. (2025).