Date Published: June 29, 2017
Publisher: Public Library of Science
Author(s): Pertti Hari, Tuomas Aakala, Emmi Hilasvuori, Risto Häkkinen, Atte Korhola, Mikko Korpela, Tapio Linkosalo, Harri Mäkinen, Eero Nikinmaa, Pekka Nöjd, Heikki Seppä, Mika Sulkava, Juhani Terhivuo, Heikki Tuomenvirta, Jan Weckström, Jaakko Hollmén, Juan A. Añel.
We collected relevant observational and measured annual-resolution time series dealing with climate in northern Europe, focusing in Finland. We analysed these series for the reliability of their temperature signal at annual and seasonal resolutions. Importantly, we analysed all of the indicators within the same statistical framework, which allows for their meaningful comparison. In this framework, we employed a cross-validation procedure designed to reduce the adverse effects of estimation bias that may inflate the reliability of various temperature indicators, especially when several indicators are used in a multiple regression model. In our data sets, timing of phenological observations and ice break-up were connected with spring, tree ring characteristics (width, density, carbon isotopic composition) with summer and ice formation with autumn temperatures. Baltic Sea ice extent and the duration of ice cover in different watercourses were good indicators of winter temperatures. Using combinations of various temperature indicator series resulted in reliable temperature signals for each of the four seasons, as well as a reliable annual temperature signal. The results hence demonstrated that we can obtain reliable temperature information over different seasons, using a careful selection of indicators, combining the results with regression analysis, and by determining the reliability of the obtained indicator.
The present global warming highlights the importance of detailed and reliable information on past climate to place the current change in a historical context, to predict the response of biological and physical phenomena to global warming , and to develop and assess the performance of climate models. Instrumental temperature measurements cover a relatively short period, usually less than 150 years. In addition, the observation network becomes sparser the further back in time we go, and the more remote the area of interest is. Thus, indirect temperature series are needed to expand the temporal and spatial coverage of temperature information.
The reliability of temperature signal obtained from individual indicator time series varied within the year (Fig 2). Analyzed seasonally, spring (March, April, May) temperatures were strongly correlated with the plant phenological observations, the reliability being over 0.60 for birch bud burst, bird cherry flowering and rowan flowering (Fig 3). The melting dates of the lakes provided slightly less reliable information on spring temperatures (0.52–0.56), followed by the 0.44 of the River Tornionjoki melting. Ice cover duration of the lakes provided some information on spring, autumn (September, October, November) and winter (December, January, February) temperatures, with the reliability ranging between 0.20 and 0.30. The freezing dates of the lakes carried the strongest autumn temperature signal, especially for Lakes Oulujärvi and Kallavesi. However, the ice area of the Baltic Sea had the most reliable temperature signal among all indicators for predicting winter temperatures, with a reliability of 0.72.
In this study, we analyzed the reliability of temperature signal in various commonly used temperature indicators. These indicators differ in their characteristics, including their sensitivity to climate, measurability and feasibility, and spatiotemporal scope (cf. ). Hence, their applicability to obtain information on changing temperatures in the past and to reconstruct past temperatures varies.
We studied the reliability of temperature signals in a collection of biological and physical indicators within the same statistical framework, using cross-validation methodology. This approach diminishes the adverse effect that estimation bias can have on the generalizability of our findings, as the cross-validation is not prone to overfitting and results in estimates of the true reliability of the indicators.