Les Henry’s Article on Temperature Records from the
Swift Current Research Station, which is posted on my blog February 16th drew a
comment from Ravin Onthill regarding the use of Simple Moving Averages (SMA).
My curiosity got the best of me, so I did some reading on SMAs and then on how
temperature norms are calculated and applied to determine Global Climate
Change, which I posted on March 6th. I found the historic Swift
Current data and followed through on my determination to play my own games with
it. (Raven, this is why I haven't had time to follow up on the links you sent me).
The Swift Current Historic Temperatures were found at https://climate.weather.gc.ca/historical_data/search_historic_data_stations_e.html?searchType=stnName&timeframe=1&txtStationName=Swift+Current+&searchMethod=contains&optLimit=yearRange&StartYear=1840&EndYear=2020&Year=2020&Month=3&Day=6&selRowPerPage=25.
The raw monthly mean temperatures were compiled from between
1886 and 2019 (with Jan and Feb 2020 included). No corrections were made to
account for any changes in time of reading, changes of location, changes of
elevation, or changes of equipment. That was left for the experts.
When more than one site is compared, a normal period of e.g.
30 years is used for each one and the differences from normal calculated. As
only one site is looked at here, it is not necessary (and I spend two days
doing it to learn that).
Annual and monthly mean temperatures are charted as
scattergrams with the y-axis coordinates set to include even the widest
outlying observation and the x-axis coordinates set between 1880 and 2020. Mean
and Standard Deviation were calculated for each set of data. About two-thirds
of observations fit within the mean plus or minus one standard deviation, an
estimate of how widely (wildly) the observations vary from the mean.
A 6th order polynomial trend line (I have no idea what
that means) was calculated for each chart as it produced the largest value for
R2 i.e. the best fit. The R2 value was posted to the
chart. R2 is a statistical measure of how close
the data are to the fitted regression line. It is also known as the
coefficient of determination, or the coefficient of multiple determination for
multiple regression. The smaller the R2 the less the trend line
explains variability in the observations. 0% indicates that the model explains
none of the variability of the response data around its mean. The polynomial
trend line is in blue on the charts. A five year SMA, in red on
the charts, was calculated just to bring the scattergram into a more
comprehendible format.
April temperatures varied from -2ºC to +11ºC with a mean of 4.7ºC and standard deviation of ±2.6ºC. The trendline explains about 6% of the monthly mean temperature variability. May temperatures varied from 6ºC to 16ºC with a mean of 10.9ºC and standard deviation of ±1.9ºC. The trendline explains only 3% of the variability of the monthly mean temperatures from the mean. June temperatures varied from 11.5ºC to 21ºC with a mean of 15.5ºC and standard deviation of ±1.7ºC. The trendline explains about 11% of the variability of the monthly mean temperatures from the mean.
July temperatures varied between 15ºC and 24ºC
with a mean of 18.8ºC and
standard deviation of ±1.5ºC.
The trendline explained about 11% of the variability. August temperatures varied between 14.5ºC and 21.5ºC with a mean of 17.8ºC and standard deviation of ±1.7ºC The August trendline is about
as flat as pee on a plate, explaining less than 1% of the variability of
observations around the mean. September temperatures varied between just over 6ºC to just under 18ºC with a mean of 12.1ºC and standard deviation of ±2.0ºC. The trendline explains less
than 5% of the variability.
October temperatures varied from a low of -1ºC to a high of +11ºC with a mean of 5.7ºC and standard deviation of ±2.3ºC. The trendline only explains
4% of the variability from normal. November monthly mean temperatures varied from a low of just
under -15ºC to a high of +5ºC with a mean of -3.4ºC and standard deviation of ±3.6ºC. The trendline explains only
3% of the variability. December monthly mean temperatures varied from a low of -21ºC to a high of -2ºC with a mean of -9.4ºC and standard deviation of ±3.8ºC. However, the trendline was
essentially flat, explaining only 2% of the variability.
Making the best of a bad situation, the annual and five of
the months had R2 of over 10%, so the trend lines might mean
something. The Annual trend seems to be cooling from about 2010. The January
trend is warming from 1980, while the February trend is cooling from sometime
after 2000. March seems to have been warming up since about 1970. June monthly
mean temps took a sharp rise in the 2010s while July has been pretty much
normal since 1960, slightly cooler around 1980, slightly warmer around 2010 and
showing a small amount of cooling in the last few years.
 |
| Table of Means, Standard Deviations and R2 |
Seven months had R2 of less than 6%, meaning
essentially the trendlines accounted for almost none of the variability of the
monthly mean temps from the average. April and May and August through December,
with R2 less than 6%, could be said to show no significant variance from the
mean. In other words, no trend to warmer or colder than the mean other
than normal year to year variations. April appears to be cooling slightly from
about 1990 and May appears to be warming slightly from 2010. August appears to
be cooling slightly from 2000; September warmed from 1980 to 2010 then cooled;
October has been cooler than normal since 1970 and really cooled off after
2010. November shows no change since about 1950, while December shows no change
since 1930.
One thing did stand out for me and that is the cyclical
nature of the trendlines, especially on the Annual chart where they appear to
be about 60 to 70 years in length. I have read research articles, I think from
an Alaskan University professor, which of course I cannot find when I need
them, that talked about 30-year and 60-year cycles superimposed on the longer
term.
If you are looking for 95% confidence levels, don’t go into
the weather/climate business. It may surprise many people that science --
the de facto source of dependable knowledge about the natural world
-- cannot deliver an unqualified, unanimous answer about something as important
as climate change. NASA.
Annual and monthly mean temperatures
over the past 135 years went up and down like a toilet seat at a keg party. A
normal year is how it usetawas or next year.
This was fun but please don’t bet the farm on it one way or
another. Anyone who thinks climatology is simple, is just kidding themselves.
I’m smarter than I was – I won’t do this again. This was a week's worth of
number crunching.
Thanks. Unfortunately your figures 2-4 are not showing up on my system - are the links broken?
ReplyDeleteThey don't show up on my phone either but if I whack the blank with my finger they show up. I will go check my computer
DeleteThe images are showing up on my computer. Don't know what is wrong.
ReplyDeleteI first went to Alaska in 1985. I returned in 2004 and found it hard to believe how much the glaciers had retreated. Speaking with natives on the North Shore, no one can remember when there wasn't shore ice. People of the arctic are the best judges of climate change. Without catastrophic incident, climate changes occur over thousands of years - the current change has visibly taken place over the past fifty years. Like denying pandemic virus, we fail to acknowledge climate change at our own peril.
ReplyDeletethe Ol'Buzzard
Ol'Buzzard, The Arctic is one of the fastest warming regions. When you look at the maps of temperature changes, along with areas that are red, there are areas that are white and blue. Just because some areas are not red does not mean the earth is not warming i.e. retaining more heat than it is giving off. That is why I want to know what is happening in small areas. People are much more likely to take things seriously when it is in their own back yard. I did see some work done on intensity and frequency of storms in several states that showed a definite increase, for example. I do have the monthly and annual precipitation records for Swift Current as well, which might be interesting. But when in the month and how much at once makes a big difference. An inch of rain in early June does far more than a tenth of an inch every day for ten days at the end of June.
ReplyDeleteMy brain is currently drained by writing, but I'll come back and revisit this post when Book 15 is wrapped up and I have some spare processing power again. I'm always interested in data analysis!
ReplyDeleteSounds good, Diane. Congratulations on Book 15. Still no offers from Netflix to make them into a miniseries?
DeleteOur host was nice enough to send along the graphics I couldn't see. Thank you.
ReplyDeleteIt does look like the January data shows a clear warming trend. The rest, like you say, is ambiguous. So far we have had perhaps 1°C warming globally (see, for instance, this figure: https://www.ipcc.ch/site/assets/uploads/2018/02/FigTS-01_errata-2.jpg) and the annual variation is of course much greater. It's hard to see anything in weather data without careful analysis; from the viewpoint of climate analysis annual weather is noise and there is a lot of noise.
The Fifth IPCC assessment devotes an entire chapter to this subject and I think this FAQ paragraph is relevant: "Warming trends associated with global change are generally more evident in averages of global temperature than in time series of local temperature (‘local’ here refers generally to individual locations, or small regional averages). This is because most of the local variability of local climate is averaged away in the global mean. Multi-decadal warming trends detected in many regions are considered to be outside the range of trends one might expect from natural internal variability of the climate system, but such trends will only become obvious when the local mean climate emerges from the ‘noise’ of year-to-year variability. How quickly this happens depends on both the rate of the warming trend and the amount of local variability. Future warming trends cannot be predicted precisely, especially at local scales, so estimates of the future time of emergence of a warming trend cannot be made with precision. - IPCC Fifth Assessment, "Detection and Attribution of Climate Change: from Global to Regional," https://www.ipcc.ch/site/assets/uploads/2018/02/WG1AR5_Chapter10_FINAL.pdf, p. 928
Raven, I tried to fix the images. Thanks for your input. January and March appear to be warming and I expect local anecdotes will bear that out. Mostly it was just interesting to see what happened over 135 years and see just how variable the weather is from year to year.
ReplyDeleteMeasuring an infinitesimal rise of 0.15 degrees every 10 years with any degree of accuracy has got to be tricky.
My education has been expanded at any rate. I finally looked up radiative forcing (In the context of climate change, the term "forcing" is restricted to changes in the radiation balance of the surface-troposphere system imposed by external factors, with no changes in stratospheric dynamics, no surface, and tropospheric feedbacks in operation (i.e., no secondary effects induced because of changes in tropospheric motions or its thermodynamic state), and no dynamically induced changes in the amount and distribution of atmospheric water (vapour, liquid, and solid forms. I'm not sure I am any smarter than before) and decadal cycles which means cycles longer than 10 years.
(Sorry this has taken so long. The past two days contained such wonders as a detour into accounting and some epidemiological statistics. I am currently looking at a frequency chart which I suspect god would like to file and forget.)
ReplyDeleteI can see the graphics now, so I think you fixed it -- thanks.
I think you have done a good deed. I figure the more people who engage the data with an open mind, the better. As you say, "Measuring an infinitesimal rise of 0.15 degrees every 10 years with any degree of accuracy has got to be tricky." Climate researchers deal with this by using lots and lots of data, huge global data sets and lots and lots of computing power. When I was at one of the US national labs doing building science, I would snarf up a small bit of one of their big clusters, and the climate scientists, and the geneticists, and the physicists would run these immense jobs using hundreds of nodes.
I'm glad the fix worked, Raven. If I can get my hands on other long term temp data (dating back to 1880s) I may do this again. Just out of curiosity.
ReplyDelete