Spatiotemporal Snowfall Variability in the Lake Michigan Region: How is Warming Affecting Wintertime Snowfall?

Abstract

This study has investigated the spatiotemporal structure and changes in Lake Michigan snowfall for the period 1950–2013. With data quality caveats acknowledged, a larger envelope of stations was included than in previous studies to explore the data using time series analysis, principal component analysis, and geographic information systems. Results indicate warming in recent decades, a near-dearth of serial correlation, midwinter dependence on teleconnection patterns, strong sensitivity of snowfall to temperature, peak snowfall variability and dependence on temperature within the lake-effect belt, an increasing fraction of seasonal snowfall occurring from December to February, and temporal behavior consistent with the previously reported trend reversal in snowfall.

Spatio‐temporal November and March snowfall trends in the Lake Michigan region

“…Consistent with the increasing fraction of seasonal snowfall which occurs from December–February (Clark et al, ), the combined fall and spring transition season snowfall has decreased significantly, while we cannot reject the null hypothesis of stationary December–February snowfall (Figure and Table ). Snowfall trends are negative throughout the fall and spring, with larger and significant decreases in both November and (for two of the three test measurements) March (Table ).…”

“…A map of the region, stations and subregion indicators is shown in Figure , while a list of station information, period of availability, and fraction of missing days is provided in Clark et al . (, Supporting information).…”

“…The snowfall and fraction of snow days to precipitation days are highly dependent on temperature in both months (Figure ), suggesting warming as a driving force for the snowfall decrease in recent decades. This is in contrast with December–February snowfall, which has not decreased despite a similarly strong dependence of snowfall on temperature (Clark et al ., ). Because the temperature trends in November and March (Table and Figure S2) have not been larger than during the core winter months (Clark et al ., ), the contrast in snowfall trends may simply reflect that the warmest Novembers (especially) and Marches have few days cold enough for snowfall.…”

“…This is in contrast with December–February snowfall, which has not decreased despite a similarly strong dependence of snowfall on temperature (Clark et al ., ). Because the temperature trends in November and March (Table and Figure S2) have not been larger than during the core winter months (Clark et al ., ), the contrast in snowfall trends may simply reflect that the warmest Novembers (especially) and Marches have few days cold enough for snowfall. The snowfall response has been especially large during November for stations in the southern subregions; for a dramatic example, South Bend, IN (located in the southeast subregion) experienced no snow accumulation‐free Novembers from 1950 to 1989, yet seven occurred from 1990 to 2012 (not shown).…”

“…Clark et al . () examined the variability of snowfall characteristics from winters 1950–1951 through 2012–2013 in the Lake Michigan basin using 47 snowfall locations in order to allow a detailed spatio‐temporal analysis. Results included a near‐absence of autocorrelation, dependence of temperature and snowfall on mid‐winter teleconnection patterns, strong sensitivity of snowfall to temperature, and peak inter‐annual variability within the lake‐effect belt.…”