aa r X i v : . [ phy s i c s . g e o - ph ] A p r El Ni ˜ no signature in Alaskan river breakups G. Boffetta
Dipartimento di Fisica Generale and INFN,Universit`a degli Studi di Torino, v. Pietro Giuria 1, 10125, Torino, Italyand CNR-ISAC, Sezione di Torino, c. Fiume 4, 10133 Torino, Italy
A signature of El Ni˜no-Southern Oscillation is found in the historical dataset of the AlaskanTanana river breakups where the average ice breaking day is found to anticipate of about 3 . The El Ni˜no-Southern Oscillation (ENSO) is the most important recurrent pattern in the inter-annual climate variability. It corresponds to anomalous warm water in the tropical Pacific ocean(El Ni˜no) coupled to an atmospheric pattern over Indian and Pacific oceans which strongly reducestrade winds. Basic physical ingredients for the appearance of El Ni˜no, occurring irregularly every2 to 7 years, are now understood, but the extension of its impact is still partially unknown [1].The effects of El Ni˜no are more consistent in the tropical Pacific areas. At higher latitudes andin remote areas the impact is less deterministic: a single event is mediated by local meteorologicalconditions and ENSO-related variations are better characterized in a statistical sense. Recent sta-tistical studies have reported effects on a wide spectrum of human and natural activities via theso-called teleconnections. To cite some examples, ENSO signature has been found in forest firestatistics, in disease epidemics and in financial market [1]. Here we report the analysis of ENSOteleconnections on the date of ice breakup in Tanana river in Nenana, Alaska. Thanks to the ratherlong time dataset (90 years) it is possible to disentangle the imprint of ENSO signal on the localmeteorological noise with good statistical significance.The Tanana data set is an non-conventional time series obtained from the Nenana Ice Classislottery. Since 1917, participants to the lottery have to guess the day (and minute) of “official”Tanana river breaking, defined as the time at which a wooden tripod frozen into the river movesand stops a clock mechanism at the shore [2]. Nenana is still a small town of about 500 people,and the tripod has been placed in about the same section of the river, therefore we can assumethat local conditions are not much changed in the 90 years of record. Moreover, ice breaking inrivers is a complex phenomenon due to both direct thermal (local temperature) and mechanical(snow melting upstream) effects and therefore it is a good indicator of the seasonal climate of theregion. For this reason, ice breaking in rivers and lakes has been used to infer trends in climatevariation over the past centuries [3]. In particular, the analysis of Tanana dataset revealed anegative (anticipating) trend in the breakup day of about 5 . d = 110 of the year) to May 20 (in 1964) ( d = 140). Withouttaking into account trends (but correcting for leap years), the mean breakup day is < d > = 124 . σ d = 5 . −
03 andin 1997 −
98. In particular, the 1997 −
98 was a strong El Ni˜no winter followed by one of the twoearliest breakup in Tanana (the other, in 1940, also followed an El Ni˜no winter). This coincidencesuggests possible connections between ENSO and ice breaking in Tanana river.We have analyzed the correlation between Tanana dataset and Best index by computing themean breaking day conditioned to El Ni˜no years. The resulting < d ∗ > = 121 . . N = 14 random years uniformlydistributed between 1917 and 2006, we compute the mean breaking day ¯ d conditioned over theseyears. By definition the average value over many realizations is again < ¯ d > = 124 . σ ¯ d = σ d / √ N ≃ .
6. Conditioning over El Ni˜no therefore gives a mean breakup day < d ∗ > which is at 2 . .
1% and we can refute the null hypothesis (no correlation)with a 98 .
9% of confidence (see Fig. 1B).Both temperature and snowfall contribute to the breaking of ice in a complex way. In the caseof lakes, the response to temperature variations is known to be strongly non-linear [6], while thepresence of snow has less known net effect as from one side protects ice from melting but, from theother side, increases the water flow below. In order to get more insight on the physical mechanismsof the breakup, we have analyzed the meteorological data of the Alaska Climate Research Centerof the University of Alaska at Fairbanks [7]. This remarkable dataset gives the minimum andmaximum air temperature, precipitation, snowfall and snow depth with daily resolution for 77years (1930-2006) in the town of Fairbanks, about 90 km upstream Nenana. We have found thatboth maximal temperature and snowfall (and snowdepth) are statistically affected by ENSO. Theaverage maximal temperature of the first 120 days (Jan-Apr) rises from − . o C to − . o C ifconditioned over the 12 El Ni˜no events of the period while the total snowfall on the same perioddecreases during El Ni˜no years from an average 75 cm to 55 cm . The probabilities that thesevariations are due to random fluctuations are comparable and, when evaluated with MonteCarlosurrogates, are around 3%. A direct consequence of reduced snowfall and increased temperatureduring ENSOs is the acceleration of snow melting on ground, which is one of the known mechanismsfor ice breaking. Indeed, we have found that ice breaking days in Nenana are strongly correlatedwith the ”no-snow” days in Fairbanks (defined as the first day without snow on ground). Indeedthe correlation coefficient between the two dates on the whole 90 years dataset is ρ ≃ . [1] M.J. McPhaden, S.E. Zebiak and M.H. Glantz, Science , , 1740 (2006).[2] See [3] J.J Magnuson et al. , Science , , 1743 (2000).[4] R. Sagarin and F. Micheli, Science , , 811 (2001).[5] See http://gcmd.nasa.gov/records/GCMD NOAA NWS CPC BEST.html [6] G.A. Weyhenmeyer, M. Meili and D.N. Livingstone, Geophys. Res. Lett. , , L07203 (2004).[7] Alaska Climate Research Center, see http://climate.gi.alaska.edu/
140 130 120 110 2000 1980 1960 1940 1920 151050-5-10-15 B r ea k up da t e ( da y o f t he y ea r) B r ea k up da t e ( de v i a t i on f r o m m ean ) Year A P r obab ili t y Day of the year B
3. The probability of a random realization of these years is given by the black area left to theline and is equal to 1 ..