The global temperature data on land that is used to measure climate change is based on daily measurements across thousands of weather stations, some some of which date back to the 1700’s. These stations record the average temperature for each day based on the minimum (night-time) and maximum (day-time) temperatures. Originally these were measured by physical Min-Max thermometers, whereas today they are mostly automated digital recordings. The daily average temperature at each station therefore is just simply
Tav = (Tmin + Tmax) / 2
The monthly average <Tav> is calculated for the 12 months of the year for each station. These temperatures depend on the altitude and location of the station. Therefore in order to compare and combine them together us Temperature “anomalies” instead. For a given station the “anomaly” is the deviation from a 30 year climatology derived for that station by averaging <Tav> over a baseline 30 year period. These values are called “normals” and the 30 year period is a “baseline”. CRU uses a 1961-1990 baseline and NASA uses a 1951 – 1980 baseline. Finally all the monthly temperatures values recorded at each station are subtracted from their monthly “normals” to yield so-called monthly temperature “anomalies”. Once combined with sea surface temperature anomaly data and averaged over the earth’s surface we then measure a Global temperature anomaly. This has “temperature” has risen by ~1.2C since pre-industrial times. The rest of the story we know well.
However we can also define other “global” parameters such as “temperature range” which can be defined as follows:
Trange (Daily Diurnal Range) = (Tmax – Tmin)
We can then follow exactly the same procedure used for temperature anomalies to define Trange “anomalies”, and then compare these with more traditional temperature anomalies. For this comparison I use the GHCN daily temperature of Berkeley Earth because it is based on daily measurements.
Tav and Trange anomalies (Berkeley Earth)
We clearly see that as global land temperature anomalies “rise” so the global diurnal temperature “range” on land “falls”.
This effect can also seen in the Australia as calculated using the daily raw temperature measurements (ACORN), (curtesy Ron Graf).
Average Temperature Range compared to temperature anomaly (ACORN)
So the absolute temperature range is also reducing as the temperature “anomalies” rise.
Could the timing of seasons be changing due to human development? I chose to look at Alice Springs Airport because it is based in central Australia, far from coastal effects yet susceptible to urban development and of course rising CO2 levels.
Comparison of recorded temperatures in Alice Springs from 1950s with those in 2010s
There is no statistically significant effect whatsoever! The climate in Alice Springs seems to be remarkably stable.
Does the drop off in range imply that one value, say night time temperatures are increasing more than day time? Or are both being affected? Most average country temperature records I have looked at show zero trend
The coldest nights in winter have warmed rather than the warmest days.
It would be awesome if you could plot exactly this phenomenon, because it is not deductible from this one. Amazing work by the way.
Global Warming or whatever name you want to give to CO2 put in the atmosphere by humans is primarily a reduction in night time cooling.
I agree.
Why?
I think it is due to a small increase in humidity.
In Seville (Spain) something similar has been happening for years. The minimums go up, but the maximums remain similar. The mean T increases, but the range between maximum and minimum decreases. The climate is more pleasant every year…
Long time ago I looked at Australian data for the 1900s and found that the prevalence of extreme heatwaves was much the same. I looked (AFAICR) for runs of days with max temperature above 40 degrees. Someone remarked that thermometers were less accurate then, but I doubt that makes any difference. I don’t have the data anymore otherwise I would look at it again!
Please correct me if i am wrong ? Tav = (Tmin + Tmax) / 2 that means if during 1 minute second in a day a peak peak is measured of 50 C and for the rest of the day its max 30C Then 50C is taken as Tmax. right ?
How long need a max to be measured to be a a max. and the same for a min.
Yes Tav is always (Tmin+Tmax)/2 . There used to be thermometers that recorded min and max temperatures.
Hugo, it’s a long time since I studied statistics, but I seem to remember that before you go looking for a mean, you have to test for normal distribution. Daily temperatures do not follow a normal distribution pattern since the warming curve is a different shape to the cooling curve. I am not sure what is being measure by tmax-tmin/2. With the limited data sets that we have we need to think on how to use them more meaningfully. I wonder if comparing minimums with minimums and maximums with maximums will give more meaningful information. If we compare the temperature against the same date of the year over a time series (many years) it will be apparent what is warming, or cooling and when it is happening. By doing this, it will be clear whether minimums and maximums are rising and exactly when in the year this is happening. Although this data could be amalgamated and manipulated, I doubt that we have the right tools to be able to draw any meaningful conclusions about global temperature from it, but someone may have a brainwave.
Clive, you are a bulldog when it comes to data analysis. An excellent idea to choose Alice Springs, since the air is dry and it is deep within a continent. I had thought of having a bash at analyzing the data from the Isle of Tiree which would give a good idea about what has been happening in the North Atlantic. Oceans are what drives the weather and I presume are highly influential on climate. Azores might be a good choice of data set to analyze.
Clive – thanks as always – look forward to your weekly updates and don’t say thank you often enough. Slightly off (an excellent) topic “I read somewhere” i.e. I can’t find the reference now – that these 30 year time periods were chosen because 30 years of measurement would give you a statisically meaningful data set. This has probably been covered here & I’ve missed it – if so apologies. It would seem to be an embedded, but flawed, period to use yet it never seems to be challenged. The met office currently offers 3×30 year anomaly periods and the actual measurement on their monthly anomaly charts. Thanks again Kris
The basic problem they had was how to measure the expected CO2 warming signal. I suspect that originally travel brochures used a 30 year average of monthly temperatures to encourage travel to exotic resorts. Their advertised climate in winter was the average temperature for December at that location. So you could expect warm weather in the Cap Verdi islands in November based on the average “normal” climate. They used to also show of course the maximum daytime temperature.
This idea was taken on by climate scientists but then extended globally. The new feature was to measure deltas against this “climatology”. The further assumption was that you can simply average these deltas across the earth to derive The “global” temperature “anomaly. However it can’t really be as simple as that because the global ocean/atmosphere is far more complicated ! 😉
Using T_max and T_min to calculate the average temperature is very crude. I wouldn’t ask a student of mine to calculate the average height of our basketball team by only using the tallest an shortest player in a similar fashion. I understand that back in the day that was east to do with inexpensive thermometers designed to do just that. But modern thermometers must be much better. And there must of have been stations in the past where the temperatures were recorded on an hourly basis. Would be interested to see what averages based on hourly readings would show.