Global warming is increasing lake surface temperatures worldwide, significantly impacting lake functioning, thermal structures, and ecosystem processes. As lakes warm, the available thermal habitat within specific temperature ranges can shrink or expand, affecting organisms based on their thermal tolerances. In some instances, suitable thermal habitats may change to the point that native species struggle while non-native species thrive. Lakes are particularly vulnerable because many species are ectothermic, and their habitats are inherently limited by the lake's boundaries, similar to islands or mountaintops. The resulting shifts in thermal habitats raise concerns about the impact of climate change on lake ecosystems and the biodiversity they support. While it's often assumed that lakes gain warm and lose cold thermal habitats with climate change, the reality is more complex. Temperatures and temperature trends vary vertically, horizontally, and seasonally within lakes. Many lakes show cooling in deeper waters, at least during stratified periods, leading to overall cooler lake volumes in some regions. This deep-water cooling can be a thermal response to surface warming, caused by increased thermal stratification. These contrasting mechanisms make predicting thermal habitat shifts in response to global warming challenging, highlighting the need for a better understanding of lake ecosystem vulnerability to climate change.

Studies have shown that some aquatic species adapt to climate change by adjusting their seasonality (phenology) or depth distributions to track suitable thermal habitats. For example, fish with broad tolerances may shift depth and seasonality to exploit new resources and interactions. However, specialist species with constrained seasonality and depth due to interactions, life history, or resource limitations (light, nutrients, oxygen) are more susceptible to thermal habitat change. *Planktothrix rubescens*, a cyanobacterium, exemplifies this vulnerability, as its adaptation to specific light and stratification conditions limits its ability to track shifting thermal habitats. Similarly, some *Daphnia* species rely on photoperiod for diapause egg development, hindering their ability to track earlier phytoplankton blooms resulting from warming. Anoxic zones can also limit the ability of aerobic species to move to deeper, cooler waters. Therefore, understanding how habitat constraints exacerbate thermal habitat change is crucial for predicting the resilience of lake ecosystems.

This study quantified long-term thermal changes in 139 lakes across six continents, representing a significant portion of the Earth's freshwater habitat. The researchers used decades of lake temperature depth profiles, comprising over 32 million temperature observations, to calculate 'thermal non-overlap' as the core metric of change. This metric quantifies the difference between recent and baseline lake temperatures (the first and second halves of each lake's time series) as the non-overlapping area of their temperature distributions, expressed as a percentage of the combined area. Temperature distributions were volume-weighted to accurately reflect the volumetric habitat available to species. The thermal non-overlap values represent relative thermal change standardized against baseline period variation. To assess the impact of depth and seasonal habitat restrictions, the researchers recalculated thermal non-overlap across various restricted habitat ranges for each lake. These restrictions provided a proxy for species' capacity to cope with temperature change. Boosted regression trees (BRT) were used to model lake-to-lake variability in thermal non-overlap and identify the most strongly related lake characteristics. The interpolation of temperature data across depth and time used linear interpolation for depth and spline interpolation with a Kalman filter for time, aiming for daily temporal and varying depth resolutions. Null thermal non-overlap values were also calculated by randomly dividing the years, providing a baseline for comparison and resulting in standardized thermal non-overlap (TNOs) values. Seasonal and depth habitat restrictions were incorporated factorially, using a continuous scale (0 to 0.95) to represent the severity of the restriction. The BRT analysis included various predictors, such as lake characteristics (mean depth, latitude), time series characteristics (starting year, ending year, seasonal coverage), and habitat restriction values. A 100-fold cross-validation was used to assess model performance and robustness. To facilitate comparisons across lakes, the BRT model was used to remove variation in thermal non-overlap due to differing temporal coverages of the lake time series. This was achieved by setting predictor variables related to the time series to their median values and predicting the thermal non-overlap values using the BRT model. The residuals from this prediction were then added to the predicted values, thus creating adjusted thermal non-overlap values for further analyses.

Over recent decades, volume-weighted whole-lake temperatures increased in 77% of the lakes studied, averaging +0.12°C per decade. However, this overall warming trend masked the complex pattern of thermal habitat change across each lake's temperature spectrum. Most lakes showed both losses and gains in thermal habitats across different temperature ranges. After controlling for differences in time series length and seasonal coverage and standardizing against a null estimate, the mean thermal non-overlap across lakes from 1978 to 2013 was 6.2%. This represents 6.2% of the cumulative temperature distributions across both time periods being composed of thermal habitat losses or gains over specific temperature ranges. When hypothetical habitat restrictions (depth and seasonality) were applied, the average thermal non-overlap increased significantly. Restricting habitats to 5% of available depths increased non-overlap from 6.2% to 9.7%, and restricting to 5% of available days increased it to 11.0%. Combined depth and seasonal restrictions resulted in a dramatic increase to 19.4%, exceeding the additive effect of individual restrictions. The relative importance of these restrictions varied depending on lake characteristics, with seasonal restrictions more impactful in shallower lakes and depth restrictions more influential in deeper lakes. BRT analysis showed that thermal non-overlap was strongly associated with mean lake depth, being higher in deeper lakes, particularly when seasonal restrictions were low. Stronger seasonal restrictions elevated non-overlap in shallow lakes, minimizing the depth-related difference. Tropical lakes exhibited substantially higher thermal non-overlap than lakes at other latitudes, potentially due to the influence of baseline temperature variability on the non-overlap metric and the generally lower temperature variability in tropical lakes. This finding highlights that lake ecosystem sensitivity to climate change, as measured by thermal non-overlap, differs from the global pattern of lake surface warming. Despite slower surface warming rates, tropical lakes, with their often high biodiversity, may face particularly strong impacts from habitat change. The study also found a strong link between lakes with high thermal non-overlap and those identified as having high freshwater biodiversity and endemism, including Lakes Baikal, Biwa, Tanganyika, and Victoria. This suggests an elevated risk of extinction and ecosystem disruption in these biodiverse lakes. The observed changes in thermal habitat may benefit some species, such as *P. rubescens* in Lake Zürich, but overall, increased thermal non-overlap is likely to increase species extinctions and community disruptions.

The findings highlight the discrepancy between whole-lake warming rates and the complexity of thermal habitat change. While some species might benefit from altered thermal habitats, the overall trend suggests increased risk of extinctions and community disruptions due to habitat loss across suitable temperature ranges. The strong association between high thermal non-overlap and lakes with high biodiversity and endemism underscores the need for targeted conservation efforts. Standard conservation measures like in-lake protected areas may be insufficient to mitigate the effects of thermal habitat change. Strategies to enhance the capacity of organisms to shift across depth and season, such as reducing anoxic zones, and improving connectivity among lakes (e.g., dam removal) may be more effective.

This study provides a comprehensive assessment of thermal habitat change in lakes globally, demonstrating that climate change is driving substantial shifts. The metric of thermal non-overlap effectively quantifies the magnitude of these changes and shows the combined impact of temperature increase and habitat constraints. The strong association between thermal habitat change and biodiversity hotspots emphasizes the urgent need for conservation strategies that go beyond traditional in-lake protected areas. Future research should focus on species-specific responses to these changes, incorporating factors like light availability and dissolved oxygen, and refining predictions of biotic change based on these new findings.

The study's reliance on available long-term temperature data introduces limitations. The spatial and temporal resolution of data varies across lakes, potentially influencing the accuracy of thermal non-overlap calculations. The use of hypothetical habitat restrictions assumes certain species’ capabilities to shift in response to temperature change. Furthermore, indirect effects of temperature change on other environmental variables (e.g., light, oxygen) are not fully incorporated, though the study acknowledges their importance and suggests further research in these areas.