In this study we analyze and simulate (with statistical downscaling techniques) the snow trends observed in the Northern Iberian Peninsula using daily snow occurrence data from a network of 33 stations ranging from 60 to 1350 metres. We first analyze the annual snow frequency measured as the annual Number of Snow Days (NSD), obtaining a significant decreasing trend since the mid seventies with a NSD reduction of about 50%; moreover, this magnitude is similar for low and high stations and for winter and spring separately. Then, we analyze the existing correlations with mean temperature and precipitation occurrence obtaining different relationships depending on the season and elevation. Finally, we simulate the observed trends using the connection of Daily Snow Occurrence (DSO) with large-scale fields simulated by a General Circulation Model (GCM); for this purpose we apply an analog-based statistical downscaling method to obtain a deterministic prediction of DSO, working in perfect prognosis conditions using reanalysis data. On the one hand, the downscaling method is able to reproduce the DSO with typical values of Hit and False Alarm Rates around 60% and 2%, respectively. On the other hand, the annual frequency obtained by averaging the DSO predictions reproduces very well both the observed trends and the high interannual variability. These promising results open the possibility to future research in seasonal or climate change predictions of snow frequency.