Archive for October, 2017

Summer Temperatures in South-Central Queensland Part 2: Weather Events and Spikes

October 30, 2017

In my last post I showed how on average temperature changed diurnally across a number of Queensland BOM stations.  In this post I will show examples of temperature change at some of these stations.  I am using “one minute data’, which despite its name, is really the value at the last second of every minute, in other words, sampling at 60 second intervals.

We know that temperatures spike up and down every few seconds, but these spikes are not captured by the Bureau unless they are the highest and lowest for each minute, and only noticed by a keen observer if the highest or lowest temperature spike so far that day occurs in the same minute (usually on the hour or half hour) as final second temperature reported at the Station Observations page.

Let us begin with this plot of a temperature spike at Maryborough Airport on 15 February, kindly reposted by Anthony Watts.  This was one of many examples from different locations around Australia of times when the maximum temperature of the day occurred in the same minute as a half-hourly recording, but exceeded it by a large amount (1.5 degrees in this example).

Figure 1:

Mboro 15 Feb

Please note that for this plot I only had access to the half hourly data from the Bureau, supplemented with some time offset data from the UK Met Office, usually 10 minutes before the BOM values.  With the higher resolution given by one minute data, we can gain a better appreciation of what was happening on this day.

Figure 2:

1 min T Mboro 15 feb

Note the spike at 13:00.  It is just part of the constant fluctuation during daylight hours which is not apparent from the data available for Figure 1.

Let’s have a closer look at the period from 12:00 to 14:00.

With the caveat that we can only guess at the 59 one second values in between the final second samples, we can use the latter values to investigate temperature response by day and night to various influences.  Assuming that the intervening one second fluctuations are approximately equally above and below a 60 second de facto mean represented by the value at the final second (as the Bureau’s Fast Facts would have us believe), a centred 5 minute mean of one minute (final second) data would approximate a mean of the complete 300 seconds.  I use a centred 5 minute mean to compare with the one minute data, but please understand this is an approximation, a best guess, when applied to short time lengths.  Its real value will be with all 115,200 data points- more later.

Figure 3:

1 min T Mboro 15 feb 12 to 2

Firstly, note how well the five minute centred mean represents most of the larger fluctuations, while considerably smoothing the final second data.

Secondly, note that the day’s maximum, 33.7C, was reached in the final second of 12:59, and was still at 33.7C at some second of the next minute, before falling 1.5 degrees to 32.2C in the final second of 13:00.

Thirdly, note that if this was a station in the USA, where 5 minute means are used, the maximum for the day would have been approximately 32.5C, still 1.2C less than the official value.

The temperature also fell 1.6C in the 60 seconds to 12:53.   And here are all the minute to minute temperature changes at Maryborough on 15 February (large outliers circled).

Figure 4:

1 min T change by hr of day Mboro 15 Feb

As shown in the previous post, this is the typical diurnal pattern.  Figure 5 shows one minute temperature fluctuation for the whole period, 1 January to 21 March 2017.

Figure 5:

1 min T change by hr of day Mboro 1 Jan 21 Mar 2017

Note the swelling of fluctuation in daylight hours, the constriction at sunset and sunrise as heating/ cooling regimes change, and the outliers: values can change by up to +2.3C or -2.1C in 60 seconds.

And here is an example of how a day’s temperature can change quite naturally, but we have to ask: would a mercury thermometer be able to match this?

Figure 6:

1 min T Mboro 6 mar

I now turn to other stations.  Hervey Bay Airport is about 30km from Maryborough Airport, only a couple of kilometres from the sea.  Firstly, how temperature changes from one minute to the next for the whole period.

Figure 7:

1 min T change by hr of day Hervey Bay 1 Jan 21 Mar 2017

Note that the daily increase in fluctuation is much less than at Maryborough.  Hervey Bay Airport is only a couple of kilometres from Sandy Strait, and proximity to a water body may be a tempering influence.

Note also the large outlier of -2C in one minute- still less than the 2.2C downwards spike on 22 February in less than a minute, which prompted my first query to the Bureau!  What could have caused such an outlier?  Here’s the one minute temperature plot for 16 March:

Figure 8:

1 min T Hervey Bay 16 March 2017

This outlier was the result of an entirely natural weather event, a sudden cool change, possibly a storm front: 4.4mm of rain was recorded at 09:00 on the 17th.  Would a mercury thermometer be sensitive enough to capture that?

And here’s 22 February:

Figure 9:

1 min T Hervey Bay 22 Feb 2017

Note the unusual spiking between about 04:30 and 06:30.  Something was going on.  Note also that the minimum temperature at 06:00 was far below at 23.2C, 1.6 degrees below any other temperature that day- for one second.

I now turn to Thangool Airport, a few kilometres from Biloela in the Callide Valley, 150km from the coast.

Figure 10:

1 min T change by hr of day Thangool 1 Jan 21 Mar 2017

Note the same shape, and though much further inland, not apparently different range from Maryborough.  Most of the change between 09:00 and 15:00 is within the bounds of +/- 1 degree each minute, but there are many outliers.

I shall now look at how temperature changed on a sample of days.  Firstly, 31 January shows a typical temperature curve for a clear sunny day.

Figure 11:

1 min T Thangool 31 jan

Figure 12 shows 7 January, a day with a mid-morning drop.  0.2mm of rain was recorded on the 8th.

Figure 12:

1 min T Thangool 7 jan

Note how after the sudden plunge the temperature quickly returns to “normal” as if nothing has happened.

28 January shows a late afternoon drop with a smaller recovery until sundown.

Figure 13:

1 min T Thangool 28 jan

Figure 14:

1 min T Thangool 24 jan

Note the typical warming curve which lasts until 16:47 when there is a sudden drop of 2.3 degrees in 3 minutes, with continued cooling.  I suspect a wind change was the cause.

Figure 15:

1 min T Thangool 20 mar

This shows a midday weather event, with the rapid return to the “normal” curve.  6mm was measured next morning.

Figure 16:

1 min T Thangool 17 feb

Note the sudden spike mid-morning.  The temperature spikes nearly 4 degrees in a few minutes to a value not expected for another hour or two.  This is odd and I cannot think of a natural weather event that could be the cause.  Whatever the cause, I doubt a mercury thermometer would track this change.

The final station for this post is at Lady Elliott Island, about 80km off the coast in the Coral Sea.  The screen is on white coral sand, about 100 metres from the water to the east.  First, one minute change over the whole period.

Figure 17:

1 min T change by hr of day L Elliott Is 1 Jan 21 Mar 2017

Note again the typical shape, but with much smaller daytime range of changes than inland sites.  Upward outliers are muted (there is only one instance of a temperature change in one minute of more than one degree).  However, downwards outliers are large and occur throughout the 24 hour period.

Here are some plots of several days on a tropic island.

Figure 18:

1 min T L Elliott Is 7 jan

Note the early morning downward spikes: rain showers.

Figure 19:

1 min T L Elliott Is 16 jan

Note the sudden drop just before midday: another rain shower.  But note how the temperature quickly returns to nearly what it was before.

Figure 20:

1 min T L Elliott Is 28 jan

Again, morning showers (quite normal near the sea in the wet season).

Now for the largest one minute temperature drop of -2.3 degrees just before midnight on 14 March.

Figure 21:

1 min T L Elliott Is 14 mar

Now watch the temperature recovery next day.

Figure 22:

1 min T L Elliott Is 14 15 mar

So, with a drop of nearly 6 degrees in a few minutes, this was a perfectly natural weather event.  Apart from sudden weather generated decreases like those shown above, it seems that there is a floor to minima of about 26C to 27C, due of course to the sea temperature.

While these examples are interesting, what about a day with sunny, fine weather?  Here’s the plot for 16 February.

Figure 23:

1 min T L Elliott Is 16 feb

Note a much more regular daytime curve (with rapid large spikes between 09:00 and 15:00), peaking only just after midday- except for a spike at about 14:30.  Here’s a closer look at the time from 12:00 to 15:00.

Figure 24:

1 min T L Elliott Is 16 feb 12 to 3

The second largest downwards spike (-1.3C) of the whole record occurred at 14:32.  This was purely a spike, not due to any weather event.  Could a mercury thermometer possibly match this?  If not, it would not reach the same maximum (30.8C).  On a hot sunny day on a coral island 100 metres from the sea, daytime temperature spikes up and down rapidly by up to a degree (or more) at a very high frequency.  Compare this with Maryborough in Figure 3.

This confirms generalisations I made in my last post:

“Temperatures in daylight hours are very volatile, while at night temperatures change very little except in unusual weather events.  Fastest and most sustained warming is in the hour after sunrise.  Fastest and most sustained cooling is also in daylight hours.  Night time cooling is much more gradual.  Cooling is on average more rapid than warming.  Rapid warming occurs when the sun suddenly appears.  Rapid cooling is associated with weather events such as rain storms.”

The Bureau of Meteorology have claimed that their AWS sensors are so designed that they mimic the mercury in glass thermometers they have replaced.   They claim a mercury in glass thermometer would track the above fluctuations closely.  However they have as yet provided no papers or comparative data to back this up.  From analysis of these stations’ data, I find that hard to believe.

Again we say, show us the data.


Summer Temperatures in South-Central Queensland Part 1: Diurnal Patterns of Temperature Change

October 15, 2017

In March of this year I purchased from the Bureau of Meteorology one-minute temperature data for the period 1 January to 21 March 2017, for a number of Queensland stations within 250km of Bundaberg.  “One-minute temperature data” is not the temperature of the whole minute, but means temperatures at  of the final second of each minute, so are spot samples taken at regular intervals.  Temperatures can be higher and lower in the intervening seconds, and so for example daily maxima can be several tenths of a degree or more above the final second values, as I demonstrated in earlier posts.

I have analysed data from these stations:  Maryborough, Hervey Bay Airport, Gayndah Airport, Thangool Airport, Bundaberg Airport, Rosslyn Bay, Gladstone Radar, Gladstone Airport, Rundle Island, Nambour, Kingaroy, Tewantin, Maroochydore, Gympie, Double Island Point Lighthouse, and Lady Elliott Island.  Most of these have few missing observations, but all still needed tedious checking.  Kingaroy’s record is atrocious, with days and weeks of intermittent data drop out.

I looked at: one minute temperature change, that is, from one data point to the next; temperature change after 10 minutes; the number of minutes of uninterrupted rise; the number of minutes of uninterrupted fall; and the number of minutes the temperature remained at the same value.

In this post I firstly plot averages of the above metrics across all 16 stations by time of day, to show the range of temperature variation from one minute to the next throughout the day and night, in distinctive diurnal patterns.

Figure 1:  One minute temperature change:-

Mean 1 minute dT

All stations show this distinctive shape, with some variance in range from island to inland stations.

Remember, this plot shows the average of 16 stations every minute of every day for 80 days.

Note the narrow range (averaging less than +/-0.1C) between sunset and sunrise, and the much larger swings from one minute to the next in daylight hours, especially between 09:00 and 15:00.  Outlier points are from weather events at individual stations.

The next plot shows the range of temperature change over 10 minute periods:

Figure 2:  10 minute temperature change:-

Mean 10 minute dT

Note the sharp increase from shortly after sunrise to an early morning peak, then a gradual decrease in the mean to a small dip at around 6 p.m..  Note again the small variation in the absence of the sun, and the many individual weather events shown by outliers.

The next plot counts the number of minutes when the temperature increases each minute at least +0.1C.

Figure 3:  Uninterrupted temperature increase:-

Mean Duration Rising

As you might expect, temperatures rise predominantly during daylight hours, with a sudden jump up just after sunrise, and a dip at sunset.

The next plot counts the number of minutes when the temperature decreases each minute at least -0.1C.

Figure 4:  Uninterrupted temperature decrease:-

Mean Duration Falling

Temperatures generally don’t fall very much just after sunrise.  However note that between 0900 and 1800 it is very rare for the temperature to be falling for zero minutes.  Most long temperature falls occur in daylight hours.  Surprising? What goes up must come down.

The next plot shows the length of time when the temperature does not change from one minute to the next:

Figure 5:  Unchanged temperature:-

Mean Duration Unchanged

Note that during the night on average temperatures are never the same for zero minutes (i.e. they are frequently the same), while in daylight hours temperatures are much less stationary, with a gradual rise from 1500.

The next graphs show the range of these metrics for individual stations.  This will be explored further in a future post.

Figure 6:  One minute temperature change:-

Max min dT comp

This shows the fastest minute to minute temperature change, both up and down.

Figure 7:  10 minute temperature change:-

Max min dT10 comp

Note that there was much faster cooling than warming over 10 minute periods, mostly associated with rain showers, storms, or cool changes.

Figure 8:  Uninterrupted temperature increase:-

Max Duration Rising comp

Figure 9:  Uninterrupted temperature decrease:-

Max Duration Falling comp

Note that Lady Elliott Island (far out to sea) and Rundle Island (in Gladstone Harbour) both had shorter periods of constantly rising and falling temperature.

Figure 10:  Unchanged temperature:-

Max Duration Unchanged comp

On the night of the 6th March at Maroochydore Airport the temperature was 26.1 degrees for 118 minutes.  As you can see nearly all stations had stable temperatures for nearly an hour on at least one occasion.

These results confirm that temperatures in daylight hours are very volatile, while at night temperatures change very little except in unusual weather events.  Fastest and most sustained warming is in the hour after sunrise.  Fastest and most sustained cooling is also in daylight hours.  Night time cooling is much more gradual.  Cooling is on average more rapid than warming.  Rapid warming occurs when the sun suddenly appears.  Rapid cooling is associated with weather events such as rain storms.

In Part 2 (probably not for a week or two) I will look at daily warming and cooling at individual stations.

Replicating Lewis et. al. (2017): Another Junk Paper

October 9, 2017

The recently released scarey predictions about “50 degree temperatures for Sydney and Melbourne” touted by Sophie Lewis are hardly worth wasting time on.  The paper is

Australia’s unprecedented future temperature extremes under Paris limits to warming, Sophie C. Lewis , Andrew D. King  and Daniel M. Mitchel, (no publication details available).

The paper is junk.  It has some very sciencey sounding words but is at heart pure speculation.  Like most “projections” by Global Warming Enthusiasts, the predictions are untestable.  Scarey temperatures are possible IF (and only if) IPCC scenarios are valid and we get either 1.5C or 2C warming by the last decade of the century.  That’s what the paper rests on.

The paper looks at Australian summer means, Coral Sea autumn means, and New South Wales and Victorian daily January maxima.  AWAP data are used for Australia and NSW and Victoria, and HadCruT4 for the Coral Sea region (which includes most of Queensland).

I have just looked at Australian Summer Means, and that was enough for me.  Lewis say that the decadal mean from 2091-2100 may have Australia wide summer means of 2 to 2.4 degrees above the mean of 2012-13, or 30.1 to 30.5C, with resultant very high daily maxima in southern cities.

I could have saved them the trouble, and at considerably less cost.

All I needed was the AWAP data for summer means (I purchased monthly AWAP data up to 2013 a couple of years ago), and plotted it with a 2nd order polynomial (quadratic) trend line:

lewis predictions summers1

And also showing decadal means (although the first and last decades have several missing summers):

lewis predictions summers2

There: the trend line goes smack through the higher (+2 degrees) projection, so it must be right!

Only trouble is, extrapolating with a quadratic trend is not a good idea. Lots can go wrong in the meantime.

So my plot is about as useful as the Lewis paper, and that’s not much.