Emerging negative trends in snow depth and cover days highlight the challenges posed by changing snow patterns around the world. They suggest that snow-dependent regions in southern Europe could be affected by these changes because the number of days with snow on the ground (DSG) determines soil processes and water-flow in rivers, streams, lakes and reservoirs. We present here the first homogeneous, annually-resolved (from October to April), multi-centennial (1681-2018 CE) DSG time-series for the Parma meteorological observatory (OBS), in northern Italy, which to date is also the longest DSG series reconstructed in the world. DSG data are in fact still poorly documented and misunderstood due to the limited and fragmentary data measurements of the past. DSG recording only began in 1938 at Parma OBS. To generate the long-term annual DSG time-series at the study site, we develop a model consistent with calibration (1938-1990 CE) and validation (1991-2018 CE) samples of observed data. We show that the variability of DSG depends on winter precipitation and air temperature, as well as on winter-spring temperature variability, suggesting that long sequences of DSG are dominated by cold air masses in years with cold weather and high variability. Modeled DSG data show a downward trend from the 19th century, in the transition period from the cold of the Little Ice Age to the warmth of modern times, followed by a more rapid decline in the five most recent decades. The DSG at Parma OBS appear to have followed over the last century trends similar to those observed throughout Eurasia and across the Northern Hemisphere, where a marked decline of snow-cover duration has been reported in the transition seasons (spring and autumn).