Kip Hansen’s brief note – August 17, 2022
I did research for other people’s book projects (I don’t write books). One of the topics I’ve been looking at recently is USCRN – US Surface Climatic Observation Reference Network (noaa.gov); Self-described as ” United States Climate Reference Network (USCRN) is a systematic and sustainable network of climate monitoring stations with locations worldwide continuous United States, Alaska and Hawaii. These stations use high quality equipment to measure temperature, precipitation, wind speed, soil conditions, etc.”
A major temperature data product manufactured by USCRN is Abnormal average temperature for the entire network for its entire length of about 17 years. It is shown updated here at WUWT in the References section as “Surface temperature, United States. Climate Reference Network, 2005 to present“Where it looks like this:
Now a lot of people want to come in and start finding the trend lines and tell us Average US Temperature Abnormal is “going up” or “going down” and how fast is it.
But let’s start with a more pragmatic approach and ask first: “What do we see here?”
I suggest the following:
1. What is the range for the time period presented (2005-2022)?
From highest to lowest, range is about 11°F or 6°C. This range does not represent a rise or fall of the index, but rather change (natural or forced). Look at the difference between the late 2005 high and the early 2021 low. If this chart were not labeled, I would define it as semi-chaotic.
2. Is the visual irregularity increasing or decreasing?
For me, it is hard to say. Oddly enough, the anomaly seems to run slightly above “0” – which tells us that the origin period for the anomaly must be from another period. And USCRN uses 1981-2010 the base time interval for “0” when determining these outliers, the base time is not within the time range of this particular time series dataset.
However, we may require Excel to tell us mathematically what the trend is over the entire time period.
There, now you know. Or is it you? MS Excel say USCRN Average Temperature Abnormal is trending upwards, quite a bit, by about 1°F (0.6°C) over 17 years.
~ ~ ~
Now to FUN!
I have arbitrarily chosen five-year time increments because they are about 1/3 of the entire time period. Three five-year trends (last trend, slightly longer), all of which are downtrends, add up to one uptrend chart when placed from start to finish in date order.
Lessons we can learn:
one. Do not use short periods when determining trends in a time series. The trend is always sensitive to start and end dates.
b. This phenomenon is similar arrive Simpson’s paradox: “A phenomenon in probability and statistics in which a trend appears in some groups of data but disappears or reverses when the groups are combined.”
“In his 2022 book Shape: Hidden geometry of information, biology, strategy, democracy and everything else, Jordan Ellenberg argues that Simpson’s paradox is misnamed: “
“Paradox” is not the right name for it, really, since there are no constraints involved, just two different ways of thinking about the same data. … The lesson of Simpson’s paradox doesn’t really tell us which view to take, but rather emphasizes that we must keep both parts and the whole in mind at the same time. ” [ source ]
c. It does take note of other datasets that change trends (or even signs of trends) when considered over different time periods – sea level rise in mind, with satellite records short claims to be twice the rate of SLR measuring tides spanning centuries .
d. Why look at trends that are obviously unreliable on different time scales? This is a philosophical question. Can a longer trend be real if all the shorter components of the trend have opposite signs? Can three shorter downtrends form a longer uptrend with real-world applicability? Or is it just an artifact of the chosen time scale? Or is the opposite true? Are the three shorter downtrends real if they add up to an uptrend? (When I say “real” I don’t mean I’m just mathematically correct – it’s physical right.)
e. Are we dealing with Simpson-like aberrations here? Is there anything important to learn from this? Both views are valid but seem unlikely.
f. Or, what we’re seeing here is just trying to force a short data set with a lot of variation to follow real-world trends? Are we fooling ourselves in interpreting the USCRN Mean Temperature Anomaly as trending up – when the physical reality is that this rather short data set is simply described as “highly volatile” “?
# # # # #
Author’s comment:
I hope some readers will find this Brief Note interesting and that it may lead to some deeper thinking than “the average and its trends must be right – they are simply math. “.
Many of the CliSci measures are viewed on a time scale that is artificially assigned to “since the beginning of the Modern Industrial Age” commonly understood to be the end of the 19th century.order century, circa 1860 to 1890. Judith Curry, in her recent interview at Mind and Matter suggests that this is literally “short-sighted” and that for many indicators, longer period of time should be considered.
I hope I have time to update your comments, I try to answer all that is mentioned clearly to me.
Thanks for reading.
# # # # #
Related
