Abstract:
This paper analyses monthly data on sea ice cover extent from January 1979 to April 2024 and on snow cover extent from January 1967 to April 2024 for various regions of the world using fractional integration methods. A statistically significant time trend is found when the errors in the regression model are incorrectly assumed to exhibit short memory. However, this evidence vanishes when the errors are allowed instead to follow an I(d) process with d different from zero and thus to be characterised by long memory (a well-known property of most climatological series) – in this case, the time trend becomes insignificant for all series. This implies that previous findings suggesting an irreversible decline in both sea ice and snow cover extent might be misleading because of an incorrect model specification.