Western Snow Trends and Global Warming: Part I

Source: American Meteorological Society

David Pierce and Dan Cayan of the Scripps Institution of Oceanography recently published an interesting paper in the Journal of Climate entitled The Uneven Response of Different Snow Measures to Human-Induced Climate Warming.  The paper has important implications for the detection of climate change and the future of skiing over the western United States.

Their findings are based on an analysis of climate change as represented by 13 downscaled climate models driven by two greenhouse gas and aerosol emissions scenarios known as RCP4.5 and RCP8.5.  They are considered medium and high "growth" scenarios with CO2-equivalent concentrations reaching about 800 ppm and 1300 ppm, respectively, by 2100.  We currently sit just over 400 ppm (CO2-equivalent combines the global warming potential of CO2, methane, and other human generated greenhouse gas into a single number for convenience).  In this post, we will concentrate on their cumulative results for the western US, which are based on downscaled model data for the colored grid boxes below.  

Source: Pierce and Cayan (2013)

What makes the study unique, as suggested by the title, is their analysis of the "uneven" response of different snow measures to human-induced climate warming.  This aspect of the study is important because it has implications for: (1) the detection of climate change and (2) planning by snow-sensitive groups.  These snow measures include:

• SWE: The amount of water in the snowpack on 1 April
• SWE/P: The fraction of cold-season (1 October - 31 March) precipitation P that remains in the snowpack on 1 April.  
• Snowfall (SFE): The total cold-season snowfall measured as the amount of water in the snow (abbreviated as snowfall water equivalent, SFE).
• SFE/P: Fraction of total water in the cold-season precipitation that falls as snow

Downscaled model trends in these and other climate variables are presented below relative to the 1976–2005 climatology for the high growth (RCP8.5) scenario (results are similar for RCP4.5, but the change is smaller).  

Changes in snow-related and other climate variables over the western U.S. Source: Pierce and Cayan (2013). 

The graphs above show the uneven response noted by the authors.  First, note that the relative size of the average change (black line) projected by the 13 downscaled models differs depending on what variable you examine.  For instance, there is a larger relative decline in the average fraction of precipitation that is retained in the snowpack on 1 April (SWE/P) than in the fraction of precipitation that falls as snow (SFE/P).  This makes some sense since warming would not only result in a decrease in the fraction of precipitation as snow, but more melting (and sublimation) of that snow once it is on the ground, so that the change in SWE/P is greater than SFE/P.  

Second, the size of the average change relative to the range of projections produced by the 13 downscaled models (the spaghetti of lines of various colors) is also uneven.  For instance, there is a wider range of "spaghetti" for the average fraction of precipitation that is retained in the snowpack on 1 April (SWE/P) than for the fraction of precipitation that falls as snow (SFE/P).  This means that there is more year-to-year variability in SWE/P (relative to it's typical size) than there is for SFE/P.  Again, this makes some sense since SWE/P is affected by changes in the fraction of precipitation that falls as snow and changes in the frequency and intensity of melting events.  Thus the swings in SWE/P are larger from warm to cold years than for SFE/P.

So, think of the black line as the signal or long-term trend produced by global warming and the spaghetti as the noise or year-to-year variations produced by climate variability.  For any given variable presented above, the long-term trend due to global warming will take longer to emerge if the variability (i.e. spaghetti) is large compared to the trend (black line).  A good example of this is cold-season precipitation which for the western US as a whole has a large year-to-year variability compared to the small long-term trend.  Identifying a the small trend in precipitation given all the ups-and-downs from year to year is very difficult, although it should be noted that the analysis above is for the western US as a whole and it could be that a regional signal might be more detectable (e.g., a wetter Northwest and a drier Southwest).

The variables for which the long-term trend due to global warming emerges the earliest are temperature and the fraction of precipitation that falls as snow (SFE/P).  The trend in SFE/P is not large (in a relative sense) compared to some variables, but the variability is smaller, and that makes the trend more significant and detectable.  Then comes the fraction of cold-season precipitation retained in the snowpack on 1 April (SWE/P) followed by 1 April SWE.  

Thus, we shouldn't expect all these climate and snow variables to display "smoking gun" characteristics at the same time.  Trends in some variables will emerge later than others.  Temperature and the fraction of cold-season precipitation falling as snow will be the first to exhibit changes large enough to discern from the year-to-year ups and downs in the climate system.  Indeed, there have been several studies showing significant rends in these to variables in recent decades.  Then comes the 1 April snowpack variables, followed lastly by snowfall. 

This analysis is for the western US as a whole and as Pierce and Cayan (2013) discuss, there are important variations with region and elevation (i.e., for a given climate or snow variable, the smoking gun will emerge earlier in some regions and elevations and later in others).  The veracity of the analysis is also dependent on the models providing a reasonable projection of future climate change.  We will have a closer look at the regional variations and their implications in a future post.