TCS Daily

Snowpack in a Greenhouse?

By George Taylor - May 26, 2005 12:00 AM

Several scientific articles have been published in recent years suggesting that western U.S. temperatures are rising, snowpack is declining, and that greenhouse gas emissions are largely responsible. Here are some examples:

From Mote, 2003*:

"Trends during the 20th century [in the Pacific Northwest] in annually averaged temperatures (0.7-0.9°C) and precipitation (13%-38%) exceed the global averages."

"...trends in urban areas are no larger than those in rural areas."

"Long-term trends in temperature are fairly well represented by linear trends."

From Mote, et al, 2004*:

"A study of springtime mountain snowpack in the Pacific Northwest showed widespread declines in snowpack since 1950 at most locations with largest declines at lower elevations indicating temperature effects."

"Substantial declines (some in excess of 50%) were common in the Cascades, especially in Oregon."

"Trends in November-March temperature ... were positive (generally >1.0°C) at nearly every location,"

From Mote, et al, 2005*:

"Widespread declines in springtime SWE have occurred in much of the North American West over the period 1925-2000, especially since mid-century."

"The increases in temperature over the West are consistent with rising greenhouse gases, and will almost certainly continue."

"It is ... likely that the losses in snowpack observed to date will continue and even accelerate, with faster losses in milder climates like the Cascades."

An examination of temperature and snowpack data for Oregon (and for limited areas in Washington) suggests that the declines described above may be largely a function of the period of record studied. What is true for a 30- or 50-year period may be very different if a longer period of data is examined.

I. Temperatures of the last century in Oregon

Oregon climate history data go back to the 19th century, but early records are sparse and discontinuous. Over the last 100 years or so, we have data from several dozen weather stations which are more or less complete. The highest quality stations have become part of the NOAA's Historical Climate Network (HCN).

Figure 1 shows charts of annual average temperature at twelve long-term Oregon stations. Notice the similarities: at nearly every station, the warmest decade of the last century was the 1930s. The single warmest year overall was 1934.



Figure 1. Annual average temperatures at Oregon climate stations

Most of the charts, in fact, resemble the U.S. temperature history (left side of Figure 2). However, the last two charts in Figure 1, for Forest Grove and Portland, more closely resemble the global temperature chart (right side of Figure 2). Are there reasons for the similarities?

Yes. HCN stations are nearly always in rural areas to avoid "contamination" problems from nearby human activities. This is true of the U.S. chart as well as most of the Oregon station charts. But Forest Grove and Portland have seen significant nearby urban growth, so their recent upward trends may reflect and urban influence rather than large-scale climate change. The same is true of the global data set, which is often criticized for "urban contamination."


 Figure 2. U.S. annual temperatures (left) and global temperatures (right) from Goddard Institute of Space Science

Scientists have long known about these urban effects, and coined the term "urban heat island effect" to describe the influence of urbanization on temperature measurements. Many of our long-term monitoring stations are heavily influenced by local land use change, and their data should be viewed with skepticism.

In Figures 3 through 5 are three pairs of stations, an urban one and a nearby rural one, illustrating this difference. If Portland, Seattle or Bellingham data were used for the purposes of determining climate trends, the warming in recent decades would be overestimated.

Figure 3. Annual temperatures since 1949 at Portland(urban) and Corvallis (rural)


 Figure 4. Annual temperatures since 1909 at Seattle (urban) and Long Beach (small town)


 Figure 5. Annual temperatures since 1909 at Bellingham (urban) and Cedar Lake (rural)


The warmest decade in the last century in Oregon, according to rural station records, was the 1930s. The last several decades have indeed seen a warming, but current temperatures remain below those reported 70 years ago. What's more, one must use temperature data with great caution in order to avoid contamination caused by land use changes, including the urban heat island effect.

II. Snow trends

The Mote, et al papers referenced earlier included the statement:

        "A study of springtime mountain snowpack in the Pacific Northwest showed widespread 
        declines in snowpack since 1950 at most locations with largest declines at lower 
        elevations indicating temperature effects."

Note the starting point for this analysis; the late 1940s-early 1950s were an exceptionally snowy period in Oregon and the Pacific Northwest. The Mote, et al papers used 1950 as a starting point because snowpack measurements were "widespread by the late 1940s" (Mote, et al, 2005) and much less extensive earlier. However, in view of the fact that climate conditions prior to the late 1940s were very different, one might wonder if inclusion of longer period data sets would change the result. We explore that here.

Snow course measurements are made throughout the winter by USDA and other agencies. These are currently collected and archived by the USDA Natural Resources Conservation Service (NRCS). Early measurements were made manually, usually once per month.

The "rule of thumb" for high elevations is that the deepest snow pack occurs on or about April 1. For the purposes of this analysis, historical values of April 1 snow water equivalent (SWE) -- the water content of the snow pack -- were examined.

Initially, the linear trend in SWE from 1950-current was calculated. It is recognized that linear trends are inappropriate for many time series, but they were used in the Mote, et al analyses and we wished to be consistent. For stations whose period of record extended back well before 1950, linear trends for the entire period of record were calculated.

The figures below (Figure 6 through Figure 9) show the two sets of trends (1950-current of period of record) for four long-term stations: Annie Springs and Summit Lake in the Cascades, and Strawberry and Ochoco Meadows in central Oregon. The equations for the best-fit linear trends are shown as well. For example, Annie Springs shows linear trends of


Y = -.28X + 56.8            1950-2001

            Y = -.09X + 47.1            1930-2001


Where Y = predicted SWE value, X = year


In this case, the slope of the 1950-2001 line drops at a rate of 28 inches per century, while the 1930-2001 drop is only 9 inches per century -- less than one-third as much. Similar differences can be seen in the other plots.



Figure 6. Trends in April 1 Snow Water Equivalent, Annie Spring, 1950-2001 and 1930-2001



Figure 7. Trends in April 1 Snow Water Equivalent, Summit Lake, 1950-2002 and 1929-2002



Figure 8. Trends in April 1 Snow Water Equivalent, Strawberry, 1950-2002 and 1938-2002



Figure 9. Trends in April 1 Snow Water Equivalent, Ochoco Meadows, 1950-2001 and 1928-2001



The longest available SWE data set in the region is from Bumping Lake, Washington, for which data are available back to 1915. As above, trends in the data were computed for 1950-current and for period of record. They are shown in Figure 10, with the shorter data set exhibiting a downward trend of 16" per century, but the period of record data actually showing a slight increase.



 Figure 10. Trends in April 1 Snow Water Equivalent, Bumping Lake, 1950-2001 and 1915-2001




Figure 11. April 1 Snow Water Equivalent, Bumping Lake, 1915-2003, fitted to third-order polynomial



Figure 12 shows Bumping Lake's record fitted with a piecewise linear trend, with each period corresponding approximately to a Pacific Decadal Oscillation regime (PDO shown in Figure 13).



Figure 12. April 1 Snow Water Equivalent, Bumping Lake, 1915-2003, fitted to piecewise linear function corresponding to Pacific Decadal Oscillation (PDO)



Figure 13. Monthly vales for the Pacific Decadal Oscillation (PDO) index


Finally, in an attempt to answer the question what is the primary variable causing variations in snow pack? we present Figure 14, a double scatterplot of total monthly snowfall versus average monthly temperature and monthly snowfall versus monthly precipitation for January (1953-2003). The site is Government Camp, at about 4000 feet in elevation on the south side of Mt. Hood in the northern Oregon Cascades; January is the snowiest month at that location. The chart reveals an expected positive correlation between precipitation and snowfall and a negative temperature-snowfall relationship. Note, however, the r-squared values: .08 for temperature and .55 for precipitation. In other words, temperature variability explains only 8% of the variance in the snowfall values, while precipitation trends explain 55%. Note, however, that no systematic study of the relationship shown in Figure 14 has yet been conducted.



Figure 14. Double scatterplot of total monthly snowfall versus average monthly temperature (blue symbols) and monthly snowfall versus monthly precipitation (green) for January (1953-2003), Government Camp, Oregon. Best fit linear trend equations and R-squared values shown.


The use of snowpack trends from 1950 through current suggests a much different (steeper) trend than if period of record measurements are used. Granted, there exist relatively few stations that extend back prior to 1940, but those stations whose records are available make it clear than monotonic decreases in snow pack do not occur through the entire period of record.


Based on a limited analysis, there are indications that precipitation is a much more significant influence on snow pack than is temperature.




Mote, P. W., 2003. Trends in Temperature and Precipitation in the Pacific Northwest During the Twentieth Century. Northwest Science, 77, 271-282.


Mote, P. W., M. Clark, and A. F. Hamlet, 2004. Variability and Trends in Mountain Snowpack in Western North America. 15th Symposium on Global Change and Climate Variations, Seattle, Washington.


Mote, P. W., A.F. Hamlet, M.P. Clark and D.P. Lettenemier, 2005. Declining Mountain Snowpack in Western North America. Bull. Amer. Meteo. Soc., 86, 39-49.




TCS Daily Archives