Friday, March 20, 2015

Projected Changes in Snowfall over the Western U.S.

An paper published last month in Water Resources Research, Lute et al. (2015) provide an interesting and insightful look at projected changes in snowfall and winter storms over the western U.S. through the mid 21st century (2040–2069) based on climate models run for the most recent IPCC Fifth Assessment Report.  They use statistical techniques trained with historical data to downscale these course-resolution climate model projections and obtain snowfall and winter storm trends at SNOTEL station locations across the west.  Not surprisingly, the results are fairly sobering for the skiers and lovers of snow.

To begin, we need to review the climatology of SNOTEL stations across the west.  SNOTELs are remote weather and snowpack observing stations used for a variety of purposes, but primarily water resource management and forecasting.  They tend to be located in sheltered locations and, within a given mountain range, typically at mid elevations (e.g., mid mountain Snowbird, but not at the summit).  The wintertime (Nov-Mar) average temperature at these SNOTEL stations tends to be higher in the Cascades, Sierras, and Arizona Mountains, and lowest over the high terrain of Wyoming, Utah, and Colorado.

Average wintertime (Nov-Mar) temperatures at SNOTEL stations.  Source: Lute et al. (2015).
The average western U.S. temperature and precipitation change from 1950–2005 to 2040–2069 for the climate models used by Lute et al. (2015) is summarized in the graph below below.  Most of the models call for 2–4ºC of warming between these periods, but there are a couple colder and warmer outliers.  Precipitation increases range from 3% to 18%.  The key point here is that there are a range of outcomes, which we will ultimately consider.

Source: Lute et al. (2015).
The average percent change in annual snowfall water equivalent (SFE, the liquid equivalent of snowfall) produced by the Lute et al. (2015) downscaled climate model projections for 1950–2005 to 2040–2069 is shown below.  The losses are greatest at relatively warm SNOTEL sites in the Cascades, Sierra Nevada, and Arizona mountains, in some instances > 55%.  As one moves to the interior where most SNOTELs are higher and the climate is more continental (i.e., colder in winter), the losses are smaller, although there are a few stations with a warm storm climatology like Ben Lomond Peak in northern Utah that experience larger loses.  At the coldest, high-elevation SNOTELs in Wyoming and Colorado, the decline is so small that it is not statistically significant (small circles) and it is unclear if such changes would be detectable given the large ups-and-downs in snowfall from year to year.  Keep in mind that the highest SNOTELs are at about 11,500 feet.  Perhaps above that elevation in Colorado they might have obtained a slight increase in snowfall.  
Source: Lute et al. 2015
Below is the trend for the number of snow days.  The pattern is fairly similar with declines everywhere, but largest at warmer sites and smaller at colder sites.

Source: Lute et al. (2015).
Although snowfall and the number of snow days declines, a more complicated picture emerges if you look at the average size of the largest 3-day snowfall events (the top 10% based on snowfall liquid equivalent).  Although there are declines in the Cascades, Sierra, and Arizona mountains, especially at warmer sites, the trends are either small or weakly positive at cooler, interior sites over Wyoming, Utah, and Colorado.

Source: Lute et al. (2015).
These results are broadly consistent with the expectation that storms become less frequent but of greater intensity in a warming world due to the higher atmospheric water-vapor content at higher temperatures.  At warmer sites, however, these bigger storms end up producing rain or mixed-precipitation, rather than snow, leading to declines.  At colder sites, these more intense storms will still be producing snow, hence a slight upward trend.

Those are the averages, but an important factor to consider is the range of model projections.  Lute et al. (2015) have a nice figure for showing this range by presenting the mean change in annual snowfall water equivalent (SFE), snow days, and the largest 3-day snowfall events within bins based on current average wintertime temperatures.  For SFE, declines increase with increasing temperature.  The most "optimistic" model is MRI-CGCM3, which calls for only about 1.5ºC of warming and about a 15% increase in precipitation, yielding a slight increase in wintertime SFE at the coldest sites and the smallest declines at warmer sites.  All the models, however, produce declines in the number of snow days, including MRI-CGCM3, consistent with a shift toward fewer, more intense storms.

Source: Lute et al. (2015).
Finally, the bottom figure above shows the change in the average size of the largest 3-day snow storms.  At the coldest stations they get bigger, whereas at the warmest stations they decline.  Perhaps Colorado will see an increase in deep powder days between now and mid century?

So what does all this mean.  To me this study reinforces the view that there is a temperature dependency of snowfall and that warmer locations in the Cascades, Sierras, and Arizona Mountains, are most vulnerable to the first few degrees of global warming.  At colder, upper-elevation sites, the changes in snowfall for the first few degrees of global warming are smaller and year-to-year variations in snowfall produced by climate variability may dominate over any long-term trend for some time (remember, however, that these results are strictly for snowfall and snowstorms, not snowpack).  However, we're looking here at changes through the mid 21st century so this isn't an end point.  In the long run, much will ultimately depend on future CO2 emissions and just how sensitive the climate is to greenhouse gas concentrations.  If you want to run this experiment out to a CO2 concentration of 1200 ppm, 4 times pre-industrial levels, good luck with that.

8 comments:

  1. Hey Jim, long time reader first time poster.

    A few years back, I found temperature data for SLC dating back to the 50's or so, I played around with the data in Excel, and I found that the max temperature in SLC has been trending upwards, but not as dramatically as the minimum temperature, I also found that we have been receiving more precipitation. I was always skeptical of my findings despite the fact that they agree with what my experience in valley. I have been wanting to do this again, but with R, but I have been unable to find the site with the data. I was wondering if you might be able to direct me to a source where I could get a csv file of this data. I am a graduate student at the U so I could access academic sources as well.

    Additionally since it seems like this year is very similar to 1934. I am really curious to know what the years after 1934 (1935-1945) looked like in regards to trough behavior or just weather in general. I know that meteorology is chaotic, but I can't help wondering if we are in a similar pattern, I just gotta know!

    Thanks ahead of time.

    ReplyDelete
    Replies
    1. Sorry, I forgot to add thanks for all you do on this website, it is great! Keep up the good work.

      Delete
    2. It's been a while since I accessed the data you are looking for directly, but I think you are looking for this: http://www.ncdc.noaa.gov/oa/climate/research/ushcn/. If I remember right, you should be able to download raw and adjusted monthly means.

      With regards to what 1935-1945 looked like, that data can be plotted at http://www.esrl.noaa.gov/psd/data/20thC_Rean/.

      Delete
  2. Hi Jim,
    Do you know what CO2 scenario they used for the climate model runs in the paper you site? Is it the worst case scenario from the IPCC model runs?
    Thanks,
    Lisa

    ReplyDelete
    Replies
    1. They used RCP8.5, which is the high-range scenario. However, there's not much difference between the scenarios in the mid-21st century (perhaps ~0.5ºC in the global mean temperature) since most reduction efforts won't start to have an impact for at least another decade or two. The differences become larger in the late 21st century.

      I was going to discuss this, but the post was already way to long! Thanks for asking.

      Delete
  3. I do not want the Wasatch to become the new Sierra Nevada - big cement snow dumps. Today I wrote a letter to Mitch McConnel, even though I do not live in Kentucky. I asked im to stop his assault on the EPA's plans to phase out coal fired power plants in many states. NY times articloe tells the story - http://www.nytimes.com/2015/03/20/us/politics/mitch-mcconnell-urges-states-to-help-thwart-obamas-war-on-coal.html?_r=0

    ReplyDelete
  4. The IPCC forecast is grim for us snow lovers for sure but the good news is the IPCC models have yet to produce a good long range forecast. The IPCC models still can't even replicate the climate change that has happened since 1850 yet alone the future! The only climate models that have successfully replicated the climate change in the past are astronomically based decadal-scale empirical harmonic models. These models look at the solar system balance which induce solar oscillations, these solar oscillations induce oscillations in the electromagnetic properties of the earth's upper atmosphere. The electromagnetic properties of the upper atmosphere in turn induces similar cycles in cloud cover and terrestrial albedo which in turn drive climate. As long as the IPCC and their models continue to believe that CO2 is the driver of climate change their models will fail to forecast the future climate. It has been shown by many climate scientists that CO2 increase FOLLOWS climate change and is not the driver which is why many climate scientists are now jumping off of the sinking CO2 climate change ship.

    ReplyDelete
    Replies
    1. I agree that CMIP model projections for regional climate need to be taken with a grain of salt, however your other statements are baseless. In response to your assertion of the superiority of astronomically-based decadal-scale empirical harmonic models: http://www.realclimate.org/index.php/archives/2011/08/an-exercise-about-meaningful-numbers-examples-from-celestial-attribution-studies/. In response to the lag between CO2 and rising temperature: http://www.realclimate.org/index.php?p=13.

      Delete