An examination of newly digitized U.S. snowfall records extending back to the early 20th Century revealed numerous inconsistencies. In the lake-effect snowbelts of the upper Midwest, it was discovered that long-term trends of neighboring stations differed greatly. Such differences seem unlikely to be real (although this cannot be discounted with complete certainty), but instead probably reflect inhomogeneities in the snowfall records. Some of the station records do not appear to be internally consistent; some stations with increasing snowfall do not show upward trends in maximum seasonal snow depth. One can imagine physical reasons for such inconsistencies that reflect changes in other aspects of the climate, but is there any evidence that this could be the case. There are other features that suggest inhomogeneities. A time series of the annual median ratio of snowfall to reported liquid equivalent precipitation exhibits an increase from about 10 in the early part of the record to 13 in the 1990s. This probably reflects changing observation practices. However, there are spatial differences. In the west, there is very little trend. Thus, there appear to be periods of time in the past when operational approaches to snowfall and winter precipitation measurement differed by region.

The above examples complicate the identification of trends in snow variables. The issues are not easy to address and it may not be possible to resolve all potential inconsistencies. At the very least, great care should be taken in interpretation of time series of snow-related variables.