The sawfly,Cephalcia arvensisPanzer, is a pest of the Norway spruce (Picea abiesKarst.) and has caused large outbreaks in Europe. The variability in the diapause length and the adaptability of the immature stages to a wide range of temperatures makes it difficult to predict the attacks and the phenology. In this paper, the effects of variable temperature regimes on both embryonic and larval development are studied in a natural environment. The experimental conditions covered the whole temperature range of the areas colonized by the pest. Several models were considered and their estimation were performed using the DEVAR software program applied to natural fluctuating temperatures. Before the application of DEVAR, its predictive power was tested using hypothetical data sets. The program was able to find parameter values leading to an excellent model fit. Four models were validated using data from three generations of natural populations of the pest. All models gave a good prediction for both embryos and larvae in all the years considered. The model can be usefully applied in population dynamics and in timing the application of both insecticides and entomopathogenic nematodes against the pest.