Saturday, 15 October 2016

Heat wave

The official explanation is sort-of defined here.  That is a bit too complicated for my programming skills - without worrying about the somewhat normative issues  that arise - but I have tried to come up with some definitions that follow the same principles.

To begin with I decided that a maximum of 30oC was above average temperature.  I also adopted the concept of three days or more at that level.  My initial idea was to count the number of periods that fitted that definition in a year but:
  • that got difficult/impossible to do automatically; 
  • it is probably better, as measure of weather-nastiness, to count the number of days in the periods (ie a period of 6 day each is much less pleasant than two periods of 3 days); and
  • it overcomes the problem of allocating to a month when a wave starts (eg) on the last day of month "a"and runs for (eg) 5 days in the following month.
In what follows I have referred to that as heat wave days.

A first modification was to exclude those days in which the overnight low was below 10, giving some relief.  If that meant the period of consecutive days became less than three, the entire episode was no longer counted.  I refer to this as adjusted heatwave days.

I did a bit of manual coding and then used ACCESS to count and cross-tabulate the outcome.  With a little help from EXCEL I have created a few charts!

The first chart shows that under either the basic or adjusted measures heatwaves occur between October and March.  Probably not a surprise!  By definition the adjustment reduces the number of heatwave days, but the pattern is quite similar..
 As the Southern hemisphere finds Summer is split by years end I have based the annual series on a year ended in April.  Thus the information for October 2015 to April 2016 is shown under the label of "2016".  There is not a significant trend with the data appearing to wobble around a pretty static value.

So, thus far, I am unconvinced of the need to fine tune my indicator by excluding the warm days following cool nights.

It has been suggested that to give a better analogy with my measures of frost (which work on a simple day by day basis) I should simply use the number of days with a maximum temperature of 35+oC as being uncomfortably hot.   On compiling such a series it shows a fair degree of difference to the heat wave day series in that there are several periods in which they move in different directions.
At present I am inclined to use the heat wave day series as it is close(ish) to the official BoM) definition and is relatively easy to monitor

No comments: