We analyse air temperature data from 14 sites in Switzerland, each with two weather stations in close proximity, one under a forest canopy and the other in the open. We use the statistics of extremes to investigate how extremely high maximum and extremely low minimum temperatures depend on the effect of forest cover. Our analysis shows that temperature maxima at two nearby stations are less dependent than are temperature minima. Maxima under the canopy are influenced by altitude: for higher sites, the maxima are less variable and depend less on the open-field data. Southerly orientations increase the dependence of minimum temperatures and so reduce the sheltering effect during cold periods. Extreme maximum and minimum temperatures occur less within conifer forests, indicating that the insulation provided by conifers all over the year is more efficient than that provided by deciduous species. Steepness of slopes has a complex impact on distributions of extremes and on their dependence. (C) 2011 Elsevier B.V. All rights reserved.