International audienceThe creation of temperature fields, in contrasted terrain, must be founded on a model taking into account characteristics and spatial evolutions of the ground. For this reason, multivariate regression often gets the best results, compared to geostatistical spatial interpolation. The parameters used as independent variables in this multivariate regression must be correlated to air temperature. These factors' influences vary according to spatial and temporal studied scales: thus, if at annual scale, elevation is often the only element correlated with air temperature, at daily scale, the model is more complex. We propose here to highlight some topographic parameters derived from geographical information systems, as well a...