A century of temperature variability in Lake Superior

A 100-yr-long time series of water temperature measured just downstream of Lake Superior is used to produce proxy time series of open-lake temperature. This analysis suggests that open-water Lake Superior summer temperatures have increased by roughly 3.5uC over the last century, most of that warming occurring in the last three decades. Correspondingly, the length of the positively stratified season has increased from 145 d to 170 d. The observed amount of warming is greater than the observed change in regional temperature over the same time period by roughly a factor of two. The discrepancy can be understood in the context of reduced winter ice cover, and implies that spatially and temporally averaged ice cover in Lake Superior has decreased from 23% to 12% over the last century. Global average air temperatures have recently warmed beyond their natural limits in historic atmospheric records (IPCC 2007). However, the expected response of temperature of oceans and other water bodies is less clear. Water temperatures of large natural systems may respond to the atmospheric warming trend in unexpected ways, due to nonlinearities, geographic variability, and feedback mechanisms. Unlike air temperature records, such as the Goddard Institute for Space Science (GISS) database (Hansen et al. 1999), reliable century-scale records of measured water temperature are exceedingly rare (Nixon et al. 2004); long, continuous records in lakes more so. Some analyses of lake-water temperature trends over nearly a century have been performed in the hypolimnetic waters of tropical lakes such as Lake Tanganyika (O’Reilly et al. 2003; Verburg et al. 2003) and Lake Malawi (Vollmer et al. 2005). Due to the absence of strong seasonal variation in surface heat flux and a lack of seasonal overturn, the deep water of these lakes respond gradually, and roughly proportionally, to changes in climate, and can be reliably analyzed using relatively temporally sparse data. In contrast, mid-latitude lakes with large inter-annual and annual variability compared to the magnitude of a longterm trend (for instance, the Laurentian Great Lakes), require dense temporal coverage in order to extract a statistically significant trend. These sorts of long time series are especially important in large lakes, since it has been shown (Austin and Colman 2007) that the thermal response of a complex system like a large, seasonally ice-covered lake can significantly exceed the rate of temperature change experienced by the regional atmospheric climate. This can be explained in terms of a coincident reduction of winter ice cover, which in turns leads to earlier spring overturn and a longer warming season. One example of such a time series of daily water temperature has been collected in the St. Mary’s River, just downstream of Lake Superior, at a pair of locations near Sault Ste. Marie (McCormick 1996) from 1906 to the present. These data (through 1992) were previously discussed (McCormick and Fahnenstiel 1999) along with several other long time series collected at power plants and municipal water supplies from around the Great Lakes. Their analysis revealed an increase in annual mean temperature on the order of 0.006uC yr21 over the time period 1906–1992, and a lengthening of the positively stratified season from roughly 192 d to 206 d over the same time period. While this time series is, to our knowledge, as long as any measured (as opposed to a paleolimnological time series of temperature proxies) in any lake, it is not clear how representative it is of open-lake conditions due to its location. The work presented here is both an update and an extension of the previous work on the Sault Ste. Marie (SSM) time series. McCormick and Fahnenstiel (1999) considered this time series among others but, at the time, only had access to data through 1992. As we show, noteworthy change has occurred in the subsequent 15 yr. Extending their analysis, we use this downstream, coastal temperature data to infer changes in open waters of Lake Superior. We show that while temperature at SSM does not accurately reflect the absolute surface temperature in Lake Superior (it is significantly warmer at SSM than the open lake), it is highly correlated with open-lake temperatures and can serve as an effective proxy for interannual variability in the lake. Using this proxy instead of the raw coastal temperature results in significantly different results when studying the timings of summer and autumn overturns in the open lake. While such a proxy is not perfect, it does still provide insight into long-term trends in open-lake conditions. The estimated open-lake stratified season is significantly shorter than previously reported, and the rate of increase in the length of the stratified season is greater than 1 Corresponding author ( [email protected]). Acknowledgments The buoy data used here was provided by the National Oceanic and Atmospheric Administration’s National Data Buoy Center. Alvin Klein of the Detroit District of the Army Corps of Engineers generously provided the 1993–2006 Sault Ste. Marie data. We also gratefully acknowledge two anonymous reviewers for providing constructive criticism. Limnol. Oceanogr., 53(6), 2008, 2724–2730 E 2008, by the American Society of Limnology and Oceanography, Inc.