Limnol. Oceanogr., 44(6), 1999, 1486–1497

The calendar date of ice break-up on southern Lake Baikal has been recorded uninterruptedly since 1869. A strong trend to earlier thawing up to around 1920 (1 d per 3.3 yr) is followed by the lack of any significant trend thereafter. For the period 1931–1994, the timing of break-up is related to local surface air temperatures integrated over periods of 1–3 months. Although highest unimodal correlations are with the 3-month mean air temperature, a bimodal relationship between break-up and air temperature exists at shorter integration times, with break-up date being related not only to the air temperature prevailing during thawing (April) but also to that prevailing during the time of ice formation, when air temperatures are lowest (February). High-frequency (interannual) fluctuations in the timing of break-up appear to be influenced mainly by the air temperatures prevailing during thawing, and low-frequency (interdecadal) fluctuations by those prevailing during ice formation. Whereas correlations with April air temperatures are always significant, those with February air temperatures are only significant during the latter part of this century, i.e., after cessation of the tendency toward earlier thawing. The high correlation between break-up date and integrated air temperature is not merely local but extends over most of Siberia and parts of northern China. Because air temperatures in Siberia contain a strong winter NAO (North Atlantic Oscillation) signal, so does the Lake Baikal break-up date, with up to 14% of the variance in the observed date of break-up being explained by the seasonal NAO index from January to March. As in the case of the air temperature data, a significant NAO signal in the break-up date can be detected only during the latter part of this century, implying that the influence of the NAO on the thawing of Lake Baikal during the early part of this century was probably negligible. Lake Baikal (Fig. 1), situated at 456 m above sea level in the Baikal Rift Zone of eastern Siberia, extends about 636 km from southwest to northeast, with an average breadth of 49 km (Shimaraev et al. 1994). The lake is estimated to be over 25 million yr old and is the deepest freshwater lake on earth (;1,650 m maximum depth) and the largest by volume (;23,000 km3), containing no less than 20% of all liquid fresh water on the surface of the earth. It is so large that it exerts a significant tempering influence on local climate (e.g., Woeikof 1900; Hutchinson 1957). Lake Baikal, which was declared a United Nations World Heritage Site in 1996, has the highest biodiversity of any lake now existing and Acknowledgments Thanks are due to M. N. Shimaraev and R. Kipfer for making available the Listvyanka ice data and the Babushkin air temperature data, to I. Hajdas for her help in translating some of the relevant Russian literature, to F. Peeters for programming the sensible heat transfer model, and to F. Peeters and R. Kipfer for their comments on an original draft of this paper. This research was made possible by funding from the Swiss Federal Office of Education and Science (BBW; grants 95.0518–1 and 97.0344) within the framework of the European Union Environment and Climate projects REFLECT (Response of European Freshwater Lakes to Environmental and Climatic Change; contract no. ENV4-CT97-0453) and MOLAR (Measuring and Modelling the Dynamic Response of Remote Mountain Lake Ecosystems to Environmental Change: A Programme of Mountain Lake Research; contract no. ENV4-CT95-007); and by financial support to the Baikal International Centre of Ecological Research (BICER) from the Swiss Federal Office of Education and Science (BBW), the Swiss Federal Institute of Technology (ETH), and the Swiss Federal Institute of Environmental Science and Technology (EAWAG). supports a unique ecological system with about 1,500 floral and more than 3,500 faunal species and subspecies (Timoshkin 1997). About 35% of the floral and 54% of the faunal species are endemic (Martin 1994). This makes it imperative that efforts be undertaken to understand this system and build up a knowledge base so that the local and international communities can react responsibly to any potential threat to it which may emerge. Ecologically, one of the most important aspects of the physical environment of Lake Baikal is the fact that it is frozen over during 4–5 months of the year. A description of the freezing and thawing of the lake has been given by Verbolov et al. (1965) and Shimaraev et al. (1994). Because of the difference in climate along the more than 48 of latitude spanned by the lake, there is a north–south gradient in both the time of freeze-up and the time of break-up. Freezing begins in late October, and most of the northern basin is usually frozen over by early December; the southern basin, however, does not freeze over until about a month later. Maximum ice thickness varies from about 1 m in the northern basin to ,80 cm in the southern basin. Ice decay in the southern basin begins in late March or early April and by the middle of May the southern basin is generally free of ice. Break-up in the northern basin occurs 2–3 weeks later. Because the presence or absence of ice cover on a lake critically affects mixing processes, and hence lake chemistry and biology, climatic control of the timing of freeze-up and break-up is of great limnological interest. Shimaraev et al. (1991), for instance, have demonstrated a relationship between freeze-up date and phytoplankton and zooplankton biomass in Lake Baikal. Break-up, with which this paper is 1487 Ice break-up on Lake Baikal Fig. 1. Outline map of Lake Baikal, showing the locations of the ice observation point (Listvyanka limnological station) and the local air temperature measurements (Babushkin meteorological station). concerned, represents a temporally integrated response to the weather conditions to which the lake is subjected during a period of several weeks in spring. Although thawing processes are extremely complex and involve many meteorological and nonmeteorological variables (see e.g., Leppäranta 1983; Gu and Stefan 1990; Vavrus et al. 1996), air temperature appears to be by far the most important of these variables (Ruosteenoja 1986; Vavrus et al. 1996), and air temperature alone is often able statistically to explain 60– 70% of the variance in break-up date (Palecki and Barry 1986; Livingstone 1997). Thus, not only can historical air temperature data be used to estimate the time of break-up of a lake, but also historical observations of the timing of break-up can be employed as proxy data for integrated local and regional air temperatures or temperature changes (e.g., Pfister 1984; Palecki and Barry 1986; Ruosteenoja 1986; Skinner 1986, 1993; Kuusisto 1987, 1993; Robertson et al. 1992; Assel and Robertson 1995; Livingstone 1997), and it has been suggested that satellite remote sensing of ice cover may be of value for estimating air temperatures in sparsely populated areas (Palecki and Barry 1986; Maslanik and Barry 1987; Barry and Maslanik 1993; Hall 1993; Wynne and Lillesand 1993; Wynne et al. 1996). For both limnological and climatological reasons it is therefore important to investigate those few long time-series of historical observations of lake ice break-up that exist with a view to clarifying the relationship between break-up date and air temperature and establishing the significance of ice break-up as a climate indicator. This paper presents the results of a study relating an uninterrupted 128-yr record of observations of the breakup date of Lake Baikal to local and regional air temperatures in Siberia and to the North Atlantic Oscillation (NAO). Data and methods Historical observations of break-up on southern Lake Baikal—The date of break-up on southern Lake Baikal has been registered continuously at the same point of observation, the Listvyanka limnological station (518529N, 1048519E; Fig. 1) since 1869 (Shimaraev 1977; Shimaraev et al. 1994). The term break-up date as used here refers to the first day on which the lake opposite the observation point was observed to be ice-free. It should be borne in mind here that the necessarily restricted field of view is likely to limit the representativeness of the observed break-up date for the southern basin as a whole (Lake Baikal is about 30 km wide at Listvyanka). Depending on wind strength and direction, for instance, ice floes may be carried into or out of the field of view. Despite this reservation, the series of observed breakup dates (Fig. 2) represents a valuable historical record. Break-up has been known to occur any time between 19 April and 20 May, but on average (1869–1996) it occurs on 4 May, with a standard deviation of 68 d. This degree of long-term interannual variability is typical for the break-up of lake ice regardless of lake size or geographical location: some examples of the standard deviation of the calendar date of break-up for other lakes are 8 d for Lej da San Murezzan, Switzerland (Livingstone 1997); 9 d for Lake Kallavesi, Finland (Simojoki 1940); 10 d for lakes in northern Sweden (Williams 1971); and 11 d for Lake Mendota, Wisconsin (office of the state climatologist, Wisconsin). It is apparent from Fig. 2 that long-term changes have been occurring in the timing of break-up on Lake Baikal. Taking the entire record into account, the mean square successive difference test (von Neumann et al. 1941; Moore 1955) confirms the existence of a significant (P , 0.05) overall trend to earlier break-up. The mean rate of change of break-up date from 1869 to 1996 is 0.052 d yr21 or 1 d per 19.3 yr (95% C.I. [confidence interval] 5 60.035 d yr21). However, it is clear that this rate is not characteristic of the entire period. Based on Fig. 2, subjectively the series can be roughly split into two subseries at around 1920. Linear regression of these subseries suggests a shift toward earlier break-up from 1869 to 1920 at the rate of 0.30 d yr21, or 1 d per 3.3 yr (r2 5 0.33, P , 0.001, 95% C.I. 5 60.12 d yr21). In contrast to this, the slight shift toward later breakup from 1920 to 1996 visible in Fig. 2 (0.036 d yr21 or 1 d per 27.7 yr) is statistically indistinguishable from zero (P .