Two years ago, an initial study of global snow cover announced how in 78 percent of mountain areas, snowfall had declined over the past 20 years.
To bolster these observations, the same author has now extended the start of data collection to 1982, thus covering a 38-year period up to 2020.
“Unfortunately, there are no denials of these trends, only confirmations,” explains Claudia Notarnicola, physicist, deputy director of Eurac Research’s Institute for Earth Observation and sole author of the study. “With a few exceptions, data on both the extent and duration of snow cover are clearly decreasing.” Specifically, there are on average, 15 fewer days that snow remains on the ground, with peaks of 20 or even 30 fewer days in Canada’s western provinces (dark orange, figure 1). Coverage has decreased by a total of four percent, which when multiplied by the extent of the areas affected by negative trends (yellow, figure 2) presents an alarming dataset.
The few existing counter-trend figures are hardly encouraging. For example, in the early 1980s, the eruption of Mexico’s El Chichon volcano caused a slight cooling that counteracted the decrease in snowfall, but this phenomenon has not lasted long. “In general, over these 38 years of analysis there are also some cases of increases in both coverage and snow days (blue, figure 2). Specifically in parts of Central Asia and some valleys in the United States,” Notarnicola continues. “There are no consensus explanations for these phenomena, but they could be other effects of climate change, for example variations in currents and winds or specific microclimatic conditions. In any case, these are a few exceptions in a very negative global context.”
The study itself took a hybrid approach by combining MODIS satellite data time series from 2000 until 2020 with a resolution of 500 meters, and mathematical models. “I chose a highly attested NASA global model as the basis and then refined it. In fact, for the period when the data from the model and the more precise satellite images overlapped, I was able to better calibrate the model thanks to so-called ‘artificial neural networks’, a computational system that falls under the aegis of artificial intelligence,” Notarnicola explains.