A recent post at WUWT discusses projections for increase in the number of very hot days in northern Australia, based on expected temperature change estimated by computer models. It’s actually a follow-up to a previous post on that topic. The theme seems to be doubt about the projections based on general incredulity, combined with devising some “rules” designed to make people mistrust computer models in general.
I expect that as the world gets hotter (which it will), so will Australia, and as a consequence the number of very hot days will increase. The original projections of how much they will increase are based on taking recent temperature data and increasing it by the forecast warming. This amounts to assuming that as Australia warms, the mean of the distribution will increase but the shape of the distribution will remain the same.
That’s not a bad approach as a first approximation. But I’m interested in whether or not we might also see a change in the shape of the temperature distribution, and how that might affect the number of very hot days.
To investigate the issue, I’ll look at the daily high temperature data from Darwin, Australia according to the Bureau of Meteorology (BOM). The daily high temperature since 1910 looks like this:
The horizontal red line marks 35°C (95°F); the adopted definition of “very hot” is a daily high temperature above that. We can count the number of such days each year, giving this:
Clearly the number of very hot days has been on the rise. But what about the distribution of daily high temperatures?
I divided the time range into 20-year intervals, starting with 1910-1930, ending with 2010-present (which is only a little less than 7 years). Then I estimated the probability distribution of daily high temperature in each time interval. That gives this:
The steady increase in average temperature is clear. But so too is the steady decrease in the width of the distribution. As a result, the probability of “cool” days has decreased faster than the probability of “hot” days has increased.
We can also look at the survival function at each temperature (the probability of being that hot or hotter):
The probability of a “very hot” day is now more than three times what it used to be. It’ll rise even more in the future.
But it may not rise quite as fast as some expect, because the shape of the distribution is narrowing. Still, it is rising at a disturbing rate. Also, the conclusion that the shape of the distribution is narrowing (that the probability of “cool” days has decreased faster than the probability of “hot” days has increased) is based on observed data, which makes it more real than a computer model simulation, but also means that we don’t really have a theoretical basis for expecting that trend to continue, other than to extrapolate based on assuming the trend will continue.
By the way, counting the number of very hot (T > 35°C) days per calendar year has a drawback, in that a very hot summer will be split between two calendar years because in the southern hemisphere, the summer season straddles the new year. A better idea, in my opinion, is to count the number of very hot days reckoning the “year” as July-through-June rather than January-through-December. On that basis, the counts look like this:
Note that when the yearly count includes the entire hot season like this, the hot summers of 2002-2003 and 2012-2013 really stand out, with 29 “very hot” days each. Note also that 2015-2016 already has seen 19 such days, in spite of the fact that the data for calendar year 2016 isn’t yet listed so that counts only half of the July-through-June year.
Deniers like the regular crew at WUWT will continue to try to persuade you not to expect any bother from the increase in hot weather, be it northern Australia or anywhere else. Meanwhile, the people of Darwin will swelter.
This blog is made possible by readers like you; join others by donating at Peaseblossom’s Closet.
Open Mind 
It would be nice to do the same with dewpoint temperature, IMHO, since it is a better indicator of discomfort.
“The steady increase in average temperature is clear. But so too is the steady decrease in the width of the distribution. As a result, the probability of “cool” days has decreased faster than the probability of “hot” days has increased.”
Has similar analysis been done with global temperature data? The decrease in cold (higher minimums) is a major factor in spread of pests to the north, like the various beetle larvae that kill trees.
One pest that may be of particular interest is the aedes aegypti, a species of mosquito. It carries dengue,yellow fever, zika and other diseases. Raise the annual minimum temperature and the frost line moves poleward. What used to be a seasonal pest that had to be “imported” each year becomes endemic, a native that is able to maintain a much higher population and the diseases that it carries may become much more prevalent. Northern Taiwan has been seeing this. In contrast, more equatorial regions may actually see a decrease in transmission due to temperatures in excess of 30°C damaging mosquitoes. Higher relative humidity may also decrease transmission, perhaps as the result of reducing water-stress and consequently the frequency with which mosquitoes bite.
Please see:
Morin, Cory W., Andrew C. Comrie, and Kacey Ernst. “Climate and dengue transmission: evidence and implications.” Environmental Health Perspectives (Online) 121.11-12 (2013): 1264.
http://ehp.niehs.nih.gov/1306556/
Nice analysis as always.
For recent months, you *can* get the daily data by month from the BoM site. (Go to home -> climate and past weather -> weather and climate data -> recent observations.) Assuming you’re using data from station 014015 (Darwin Airport) I can confirm that there have been 16 days with daily max over 35C in the first 6 months of 2016, including a run of 6 consecutive days in Apr. That would make an impressive 35 days for 2015/16 and a new record.
Another interesting statistic would be the duration of heat waves – consecutive days over some limit.
If you can’t find the data, drop me an email and I’ll give you more detail on how to find it so you don’t have to trust my counting :)
Again you have produced a fascinating and enlightening post. There are a couple of questions that I would like to ask as a result of this.
The first is procedural: Why did you not use 30 year periods as 30 years seems to be the standard period for measuring climate change? I can see that your method gives a very good impression of the likely changes, but would overlapping 30 year periods give more confidence in the results.
The second question is about the narrowing distribution. Is there a theoretical maximum temperature that can be obtained in an area? This would mean that the maximums might only increase occasionally and by small increments, while the mass of days will increase more regularly and by larger amounts– so in a heating world the distribution might narrow because of the laws of physics .
This may be seen as relevant.
https://www.theguardian.com/environment/damian-carrington-blog/2013/jan/08/australia-bush-fires-heatwave-temperature-scale
Good work as usual.
Is anyone trying to track the number of WUWT participants? Are any people there getting the message and quietly retiring from ‘prominent denier’ status? I guess there is always some coming and going in comment forums so it may be hard to get reliable data.
It just seems like some people must notice after a while that they picked the wrong side, or at least entertain the possibility that they might have done, and decide to be a be less ‘forthright’ about it.
The finding of the narrowing of the distribution is really interesting – and intriguing.
My hypothesis is, that high temperatures are cut off by the sea, with its greater thermal inertia. This is supported by the fact, that we have a 70 % – probability there of winds coming from the sea, according to
http://www.bom.gov.au/cgi-bin/climate/cgi_bin_scripts/windrose_selector.cgi?period=Annual&type=3&location=14015&Submit=Get+Rose
.
It would be interesting to link ultra high temps with wind direction on that particular day, or to do a similar statistics for an inland location.
A long way further from the equator down in Perth (32 degrees south) we happily get maxima in the 40’s every summer, ( a lot warmer than Darwin), provided we have dry winds coming off the land.
With Darwin adjacent to a very warm sea, I think on the hottest of days more of the energy is going into increasing the humidity rather than the temperature. Hence the narrowing distribution.
Hi
I suspect you’ve used ACORN-SAT data – which is OK – but just be aware that it is joins the old Darwin post office site with the current airport site (with a bias correction). You may be better off just considering post 1941, when the move to the airport occurred.
The limiting state to the maximum temperature during the build-up season is the arrival of the seabreeze – and whilst ACORN-SAT does a good job in estimating the overall average impacts of site moves it may be a step to far to assess the impact on extremes (cue vigorous debate).
You don;t really need to split over the southern summer as peak temperatures occur in the pre monsoon”build-up”.during September/October/November. The number of days in December/January/February greater than 35 is really quite small due to the monsoon trough being well developed over continental northern Australia (and the prevailing wet season dynamics limits maximum temperature).
GB
Here is an example of hot day analyst using R and the weatherData package.
https://rclimate.wordpress.com/2016/08/29/rclimate-script-to-assess-local-hot-day-trends/
Just a somewhat pedantic, but also in my opinion interesting point on splitting the year July-June: Darwin is close enough to the equator that the max temperature is hotter in spring and autumn than in summer. Summer is cooled by the wet season. Minimum temperatures are close to the standard pattern, but still peaks early with Nov and Dec averages the same. On the other hand the number of hot days may follow ENSO cycle which are going to be better represented by a July-June split. Also at a guess a failed wet season is going to result in a spike in hot temperatures from say Nov-Apr, and an especially good wet season may result in the opposite.
p.s. I think there may be some questions on the data for Darwin, and GISS discards significant portions of the BOM data and assesses the trend as rather flat. I looked into this after Watts and co made a big deal of adjustments at this station being the ‘smoking gun’, but what I discovered is that they started their analysis with the ‘raw’ GHCN data which had already made adjustments to BOM data (and shows no trend), and ended with the final GHCN data (showing a warming trend) which GISS then applies further analysis to and rejects all data prior to a rather recent date and shows no trend (or at least did maybe 5 years ago when this argument was made). There are several moves on the Darwin station, and it is very isolated so no good quality nearby stations to use in homogenization.
GISS current trend maps at 250k resolution, and with ocean excluded for 1950 to 2015 shows a red area just to the NW of where I think Darwing should be, and blank area to the east and south. I’m not sure if GISS is accepting the Darwin data but plotting it a bit further NW than I think it should be, or rejecting it as unusable and plotting data for some offshore islands.