Monday, April 23, 2012

A Look at Snowpack Trends

We've seen bad and good snow years over the past decade in the
Wasatch Mountains.  Is there a trend?
Detecting and determining the cause of snowpack trends over the western United States is a difficult task.  There are a number of reasons for this including:
  • The lack of reliable, continuous snowpack and snowfall records in the early 20th century.  As a result, most analyses begin in about 1930 or 1950, which is a relatively short period.  Further, throughout the historical record, snowfall and snowpack observations have their warts, which complicates quality control and trend attribution.  
  • There are large, natural year-to-year and slowly varying decade-to-decade variations in climate and snowpack characteristics, which complicate trend analysis.  
  • Non-climatic factors can contribute to the observed trends and variations, including changes in observational techniques or locations, land cover (e.g., increased tree cover), and land use (e.g., increased snowmobile traffic).  
  • Long-term trends and variations can also be related to anthropogenic factors besides greenhouse gas emissions, including land-surface change and dust loading.  The latter increases the absorption of solar radiation, leading to an earlier and more rapid snowmelt.    
Recently, the Salt Lake Tribune reported on a soon-to-be published study by Gillies et al. (2012, paywalled, so only the abstract may be available for free to some users) examining snowpack trends in Utah, which was followed by an hour long discussion on KUER (podcast available here).  Given the recent interest in this topic, I thought I would review current understanding in this area.  I'll be wading into shark infested waters, but I will try and summarize the science as best as I can.  

The Trends

For simplicity and because it should be available for free public access, I'm going to concentrate on Pierce et al. (2008), which is probably the most comprehensive paper on western US snowpack trends and their causes.  They examine trends in the ratio of the 1 April snowpack snow water equivalent (SWE, a measure of how much water is in the snowpack) to the water-year-to-date (October to March) precipitation (P).

Using this SWE/P ratio has several advantages, but perhaps the most important is that it reduces the effect of variations in precipitation on the results.  Essentially, the SWE/P ratio tells you how much of the precipitation that has fallen since October is still present in the snowpack on 1 April.  

Results are presented for various regions and elevations for the contiguous western United States and concentrate on sites where "an appreciable fraction of the winter precipitation is retained in the snowpack on 1 April."  

Source: Pierce et al. (2008)
For comparison purposes, Pierce et al. (2008) divide each year's SWE/P by the long-term average for the study period (called fractional SWE/P).  A fractional SWE/P of one indicates that an average mount of water year precipitation is retained in the snowpack on 1 April.  A fractional SWE/P greater than one indicates that an above average fraction is retained, whereas a fractional SWE/P less than one indicates than a below average fraction is retained.

One of the more important figures in their paper presents the fractional SWE/P trends by elevation band, which shows the largest declines in the lowest elevations.  The downward trend, however, decreases with elevation and, above 1910 m, the trends are small and no longer statistically significant.   Keep in mind that these are cumulative statistics for all the stations indicated above, so there may be some variations by region.   
Time series of fractional SWE/P as a function of elevation with least-squares fit
trend line shown as solid if the trend is significant at the 95% confidence
interval and dashed if it is not (Source: Pierce et al. 2008, with figure
caption paraphrased from their text and Fig. 5).
My interpretation of Gillies et al. (2012) is that they found a similar influence of elevation in Utah.  In particular, they conclude that, "it was estimated that the proportion of winter (January-March) precipitation falling as snow has decreased by 9% statewide over a half century, with greater reductions occurring at lower elevations (< 2000 m)" (my emphasis).  


What has caused these declines in low-elevation snowpack?  This is a challenge as the climate of the western United States is influenced not only by long-term climate trends, but also variations in Pacific sea surface temperatures and related large-scale atmospheric circulations (examples include El Nino and the Pacific Decadal Oscillation).  

Pierce et al. (2008) use climate model simulations to evaluate the relative roles of natural climate variability and climate change related to anthropogenic greenhouse gas and aerosol emissions.  Their results suggest that roughly half of the trend is due to anthropogenic warming.  

One study that suggests the anthropogenic influence may be smaller, at least in one region of the contiguous western United States, is Stoelinga et al. (2010).  They found that the spring snowpack in the Cascade Mountains declined 23% from 1930–2007 (a longer period than examined by Pierce et al. 2008, although the trend is not quite significant at the 95% level).  For the 1950–1997 period, they find a 48% decline, but argue that about 80% of this is due to natural variability.  


The detection and attribution of snowpack trends in the western United States is challenging.  It is my conclusion based on recent studies (including many not discussed above) that there have been long-term declines in snowpack over the lower elevations of the western United States, but that these declines decrease with elevation and trends at upper elevations appear to be negligible.  It is likely that anthropogenic warming contributes to the low-elevation snowpack decline, but that natural climate variability may also be playing a role, at least in some regions.  

Evaluating the contribution of anthropogenic climate change, natural climate variability, dust loading, and land-surface change is an area of ongoing research.  In addition to improving our understanding and prediction of past and future long-term snowpack trends, this is an important interdisciplinary research problem that is very relevant for improving seasonal, year-to-year, and decadal scale climate forecasts for water resource management.


  1. Do you think solar minima and maxima influence P (the total quantity of precipitation) and SWE/P, the fraction that is snow? My prior is minima increase both P and SWE, converse for maxima. True?

  2. To my knowledge, a clear linkage between solar output and precipitation (or SWE/P) has not been established.