Long-term hourly data records of the Automated Meteorological Data Acquisition System (AMeDAS) with snow depth measurements at three high elevation sites were analyzed in central Japan to assess the decadal scale winter climate variabilities with the effects of observation biases. Year-to-year variations of persistent snow cover periods (PSCPs) did not show a significant shortening trend with the delays of starting the PSCP associated with El Niño in December and forward to the ending of the PSCP with nationwide warming in March. Significant warming trends in the surface air temperature averaged in PSCPs were not identified, and length of the PSCP strongly contributed to the year-to-year variations of monthly averaged winter air temperatures. Long-term variabilities of precipitation records were primarily affected by changing the gauge type and data resolutions. After a simple correction of gauge undercatch by air temperature and wind speed, the amount of precipitation increased by around 20 % at the northern station but only by 3 % in the south due to frequent rain-on-snow events. In the two northern stations, snowfall amounts on the order of around 10 % were estimated to be missing in the precipitation records, where a long-term decreasing trend in occurrences of freezing precipitation was also found. The statistics calculated by AMeDAS data were verified using the data of the Vaisala Present Weather Detector in the Sugadaira Highland.