Abstract : A network of 48 stations was setup in the Jura Mountains (France) to interpolate temperatures under forest cover, an environment that differs markedly from open sites. The stations were positioned so as to sample all topographical settings. The series analysed consisted of the daily minima and maxima extracted from observations recorded for a year (356 days). Regressions were used to describe relations between temperatures under forest cover (dependent random variables) and a set of explanatory variables: latitudinal and longitudinal variations, distance to the nearest forest edge, and seven variables derived from a digital terrain model by geomatic computation [elevation, slope gradient, sine and cosine of slope aspect, global radiation (g-rad), valley depth, and hump amplitude]. Elevation is the variable that most strongly affects the spatial variation of temperatures under forest cover followed by valley depth, longitudinal variation, and g-rad (for maxima only). Variables that are significant at the 10% level are combined in multiple regressions with a view to estimating temperatures at 50 m resolution. The MAE values are 0.82 and 0.77 °C for minima and maxima, respectively; adjusted R2 values are 0.47 for minima and 0.64 for maxima. Finally, we present four maps of estimated temperature under forest cover.