A comparison of consecutive 30-year periods shows that summers in the U.S. and Canada are getting longer, and winters are getting shorter. The length of summer is about 6% to 10% longer in most regions. Winter is about 7% to 11% shorter in the Lower 48, and a staggering 30% shorter in Alaska. As temperatures continue to increase, expect these trends to continue.
Introduction
Is summer longer than it used to be? Is Winter shorter than it used to be? To answer these questions, we first need to define what winter and summer are. Should we think of them as December-February and June-August? If so, then winter and summer are exactly the same length every year. Of course this isn't what you were thinking.
What you really want to know is whether summer heat lasts for a longer period of time and if winter cold is shorter. The answer to these questions are an unambiguous yes for most places. The reason for this is simple, the climate is warming in most places. If the average temperature is warmer, then the comparison to what temperatures used to be like will change accordingly. Think of it this way, if you live in Omaha, Nebraska, and you imagine what summer is like (length and intensity) and then you move to Houston, Texas, you would experience a (much) longer period of summer conditions in Houston that what you are accustomed to. Similarly, if you moved from Chicago to Oklahoma City, you would say that you effectively shortened your winter. With a changing climate, we are effectively relocating our cities farther south.
Australia Study
In March 2020, the Australia Institute published a study titled Out of Season, which looked at this exact question. They asked if summer was getting longer and and winter was getting shorter by comparing similar period from two 20-year time spans. They noted a dramatic increase in the length of summer and a dramatic decrease in the length of winter. There are very important philosophical and methodological differences between the Australian study and the current study that will be noted at the end of this post.
Other Season Length Studies
There are many, many papers that discuss changes in the wet season, the dry season, the snow season, the frost-free season, and so on. Relatively little has been done on changing boreal seasons. The most interesting paper is a European study that looked at the changes in the length of summer over a 63-year period (Peña-Ortiz et al. 2015). The authors did a grid-based study and computed a detrended average of all the June daily temps for each grid cell. The 63-year average June temperature value was declared the unbiased start of summer. The same procedure was performed on September temps to determine an unbiased end of summer date. Next, the raw (untrended) grid cells were evaluated for each year and when a 30-day average daily temperature reached the reference start and end dates, they were flagged as the begin and end of summer for each year. With 63 start/stop summer values for each grid cell, a trend could then be computed.
A new study by Hekmatzadeh et al. (2020) looked for novel ways to assess changes in the summer and winter periods for two cities in Iran. They start by identifying the warmest and coldest 90-day periods of the year (using a 90-day smoothed average of daily temperatures) and define those as summer and winter respectively. For deciding when each summer/winter began/ended each year, I'll be honest, I cannot figure out for the life of me what the heck they did. What ever it was, it was very poorly explained. The result was a time series of begin and end dates for summer and winter in each year. They were then able to conduct linear regression trend analysis on the results. Also, they were able to assess the difference in the timing of the seasons for each year.
An earlier study by Yan et al. (2011) also looked for changes in the warmest and coldest 90-day periods of the year to assess changes in the warm and cold season lengths and intensities. Unlike the previous study, this one assess the changes in annual season begin and end dates for over 500 cities (stations) across China. An elaborate methodology is used remove the "noise" from daily observations. Both this study and the previous study systematically look at each year's temperatures and extract begin and end points of for the seasons from the data. You'll see below that our new study uses constructed climate normals – thus alleviating the problem of noisy daily data.
Methodology for This Study
To identify whether season lengths are changing, we need to define a modern period and a reference period. I decided to use consecutive 30-year periods (1990-2019 as the modern period and 1960-1989 as the reference period). Using the GHCN-M climate database, I identified all stations with at least 20 years of complete data in both time periods. There were 6,625 qualifying stations globally. In the U.S. and Canada, there were 3,250 qualifying stations. For each station, daily climate normals were computed using a 3rd-order (cubic) spline. This very closely approximates the official NCEI climate normals procedure.
For the reference period, the warmest 90-day period of the year and the coldest 90-day period of the year were identified and declared to be the baseline summer/winter periods respectively. We can then compare the climate normals of the new, modern time period and identify the dates where the old seasonal temperatures begin and end.
Examples of changing summer and winter temperatures:
Imagine the warmest 90-day period of the year used to be June 3rd to August 31st and on both June 3rd and August 31st, the daily average temperature was 70°F (it's warmer during the intervening days). Now imagine it's 30 years later and the date where the summer temperature first reaches 70°F is May 30th and the date where the temperature drops back down to 70°F is September 5th. Compared to the earlier period, we now achieve summer heat 4 days earlier and it lingers 5 days longer. This gives us a 99-day period in today's climate that experiences the same length of warmth as the 90-day period of yesteryear – a 9 day increase. Two animation example of 5 stations are shown below. A summer animation and a winter animation. Remember that we are not comparing the intensity of the warmth/cold, just the length of time we experience comparable temperatures.
Figure 1. Changes in summer length for 5 U.S. stations.
Figure 2. Changes in winter length for 5 U.S. stations.
The vast majority of stations globally now experience a longer summer compared to the previous 30-year reference period. The exception to this rule in North America is parts of the Midwest and Great Plains. Areas in blue on the first map show where summer conditions are slightly shorter. For every place else, the closer you get to an ocean, the longer summer gets. This makes sense as the oceans have warmed dramatically and they impart a tremendous influence on the climate.
There is also a notable urban heat island (UHI) effect. Cities are experiencing longer summers than nearby rural areas. Remember that 80% of Americans live in urban areas. The added increase in urban heat is real and affects the lives of those 80% of Americans.
Winter Analysis
The change in the length of winter is even more dramatic than the summer changes. While average summer conditions last 7 days longer (97 days vs 90 days) than they used to, comparable winter conditions last 15 days less (75 days vs 90 days) than they used to. The winter patterns in the Lower 48 are broadly similar to the summer patterns, except the winter warming is slightly less stratified from north to south. In northern Canada and Alaska, the shortening of winter is dramatic. In fact, along the North Slope of Alaska, even the low point is winter during the current period is warmer than any of the days in the 90-day winter from 30 years prior! Urban heat island effects are equally as pronounced in the winter map as they are in the summer map.
Selected Stations
The graphic below shows how long the previous 90-day summer and winter are in today's warmer climate. This is only a partial list. A complete list can be made available upon request.
Changes in Season Timing
Instead of looking at the length of the seasons, what if we just look to see the difference in the timing of the warmest and coldest 90-day periods from the consecutive 30-year periods (1960-1989 vs 1990-2019)? The maps below show the difference in the timings.
In summer, there is a pronounced shift toward earlier peak heating in Alaska, and a pronounced shift later in the northern Rockies. Outside those areas, the changes are for earlier summer heat in southerly latitudes and later summer heat in northerly latitudes.
For winter, the patterns are much less distinct. Most places are seeing peak cold occur sooner. More appropriately, late winter is warming faster. This effectively pushes the coldest part of the year to an earlier point on the calendar.
Figure 5. Change in timing for warmest 90-day period of the year.
Figure 6. Change in timing for coldest 90-day period of the year.
The Australia study noted earlier uses the following methodology.
1) Define the seasons strictly based on calendar months (Jun-Aug and Dec-Feb).
2) Pick two 20-year periods (1950-1969 and 1999-2018).
3) Use a 21-day average (smoothing) filter of daily temperature and construct a 20-year daily climatology.
4) Identify the 1950-1969 begin and end of season temperatures in the reference period and looks for those temperature start/stop points in the 1999-2018 modern period.
5) Compute season length changes for 70 stations south of 25°S latitude.
Comparison of Australian and Current Studies
1) The current study uses constructed daily normal temperatures from monthly data – the same technique used by the NCEI and the WMO. The Australia study uses smoothed daily data. The advantage of using constructed climate normals is that the smoothing is a feature of the spline, whereas the 21-day average of daily data tries to approximate a smoothed line. An argument can be made for either approach.
2) The Australia study sets summer and winter in the reference period to be matched with the calendar months (Jun 1/Aug 31 and Dec 1/Feb 28). The current study identifies the warmest/coldest 90-day periods regardless of how they match with the calendar months. This allows us to track the changes in timing of the seasons.
3) The Australia study uses two 20-year periods that have a 29-year intervening gap. The current study uses two consecutive 30-year periods. This is a non-trivial methodological difference. The annual temperature change in Australia between the 1950-1969 and 1999-2018 time periods (their methodology) was 0.755°C. The annual temperature change in Australia between the 1960-1989 and 1990-2019 time periods (our methodology) was 0.534°C. [Computed from Berkeley Earth global temperature data.]
Results Analysis
Using the Australia study's 20-year periods versus our 30-year periods results in a magnification of the warming by 41% (0.755°C vs 0.534°C). This large difference has a substantial impact on the changes in the length of summer and winter and is largely why they show a much more substantial lengthening of summer and shortening of winter.
The GHCN-M data set contain 215 stations in Australia that met out criteria. We performed the same analysis for Australia that we performed for the U.S. and Canada. In Australia, we determined that summer is now (1990-2019) 9 days longer than it used to be compared to the 1960-1989 reference period (99 days vs. 90 days). This is dramatically different that the 4-week lengthening of summer noted in the Australian study (118 days vs. 90 days). For the U.S. and Canada, the change in summer length was 7 days (97 days vs. 90 days).
For winter in Australia, we computed a decrease in winter length of 9 days in our two reference periods (81 days vs. 90 days). The Australia study indicated a decrease in winter length of 23 days (69 days vs. 92 days). For the U.S. and Canada, the change in winter length was 15 days (75 days vs. 90 days). Much of the U.S. and Canada decrease was a result of dramatic changes in Arctic latitudes.
The method of constructing daily averages vs daily normals, and the selection of reference and modern periods explains the majority of the differences between the two methods for Australia. One method is not necessarily better than the other, but it's important to understand why the results are so different.
Alaska-Centric Maps
These are the same maps show in in figures 3, 4, 5, and 6, but with a focus on Alaska.
Europe Maps
Here are the Europe versions of the same analysis. The modern summer length is now 9 days longer (99 days vs. 90 days). Winter is now 65.6 days compared to 90 days in the 1960-1989 reference period (24.4 days shorter).
Hekmatzadeh, A.A., Kaboli, S. & Torabi Haghighi, A. New indices for assessing changes in seasons and in timing characteristics of air temperature. Theor Appl Climatol (2020). https://doi.org/10.1007/s00704-020-03156-w
Peña-Ortiz, C., D. Barriopedro, and R. García-Herrera, 2015: Multidecadal Variability of the Summer Length in Europe. J. Climate, 28, 5375–5388, https://doi.org/10.1175/JCLI-D-14-00429.1
Swann, Tom and Mark Ogge. 2020: Out of Season: Expanding summers and shrinking winters in subtropical and temperate Australia. The Australia Institute. https://www.tai.org.au/sites/default/files/P834%20Out%20of%20Season%20%5BWEB%5D.pdf
Yan, Z., Xia, J., Qian, C. et al. Changes in seasonal cycle and extremes in China during the period 1960–2008. Adv. Atmos. Sci. 28, 269–283 (2011). https://doi.org/10.1007/s00376-010-0006-3
