Heatwaves have become a publicity tool. Far from there being a clear trend in Australian heatwaves, Geoff Sherrington shows that it’s also legitimate to claim heatwaves were worse 80 – 100 years ago in Adelaide and Melbourne and things are getting better. Those officials who cherrypick their claims might be technically correct, but it’s outrageously deceitful and unscientific at the same time.
Just how hard is it to get a record heatwave? It’s so easy that if it’s summer in Australia, it’s hard not to set a record. That’s because heatwaves come in so many flavors – there are seven capital cities which can all have 3 day, 4 day, 5 day or 6 day heatwaves. Then there are the heatwaves over 40C, or over 38 C, or over 35C… already that makes 84 flavours of wave. If a hot spell doesn’t break one type of wave, it could easily break another. Then there is the pre-heatwave, and there would be another 84 types of heatwaves that we haven’t had, but might get, you never know. You might think I’m kidding, but pre-heatwaves get headlines already:
“More Canberra heatwaves forecast”
“A heatwave could return to Canberra next month, according to the Bureau of Meteorology.”
Did we need to ask the Bureau of Meteorology if there might be a heatwave in February in Australia?
Guest Post by Geoff Sherrington
Preparing for heat waves is important, for fire, health, and electricity supply. There is a message being spread that recent heat waves are getting worse, that global warming will make them more common. A BoM paper (N&F 2013, see below) tells us that
“The number of heat-related deaths in temperate Australian cities is expected to rise considerably by 2050 as the frequency and intensity of heatwaves is projected to increase under climate change from global warming.”
N&F 2013 also assert that –
“While heatwaves are not unusual for Australians, the trend towards more frequent and intense heatwaves (Alexander et al. 2007) is of significant concern.”
Some regular temperature records start as long ago as 1850, yet this Alexander et. al. prediction concentrates on the years 1951-2003. I wondered what happened to the early years of data? So I looked at heatwave temperatures in Melbourne and Adelaide. These graphs summarize the hottest maximum temperature heat waves in each city.
Adelaide has a population about 1.3 million today. The main weather station (23090) is now at Kent Town, 34.9211S, 138.6216E, records from there or nearby West Terrace (23000) start from 1887.
Melbourne has a population about 4.3 million. The main weather station (86071) is at 37.8075S, 144.9700E, with records from there or nearby from 1856 onwards.
This graphic evidence suggests that if anything, we are having fewer heatwaves than we had before (say) 1975. And some less hot. I qualify this: “Australia’s past climate records are not good enough to support sophisticated analysis and accurate forecasting.” There is only so much that one can do with one Tmax temperature and one Tmin each day.
Conclusions about heat waves depend on many factors:
- How you select places as examples. (E.g. Hobart is not so interesting lately).
- Definition of ‘heat wave’. (E.g. 3, 4, 5 or 6 day, even longer, or more complex?)
- Choice of weather station. (E.g. Kent Town versus Adelaide Airport 2009 heat wave).
- Choice of weather record. (Do we use Acorn or “raw” data?)
- Quality of data keeping. (In some datasets the max temperature is sometimes lower than the minimum on the same day.)
The official approach appears clumsy and error-prone:
- Short past data series are used when long ones are available.
- There is subjective estimation of missing values, instead of deletion.
- Adjusted data is used when raw data is available.
- Complicated indices are invented when simple ones tell an adequate story.
- There is an uncritical reliance on the manipulated records of others.
Let us step through these five points.
Point 1. How you select places as examples.
Selection of publicized locations is a process of cherry picking. In order to generate headlines about global warming, cities with large populations and a history of fires or energy shortages are the best candidates, particularly if emergencies are recent, overhyped and in the minds of many people. By these criteria, Adelaide and Melbourne emerge as prime targets. Darwin has an insignificant bush fire history, so apparently doesn’t get a mention, and Hobart has not seen much recent heat wave activity as shown here.
Concentrated publicity of problems or theoretical trends in places like Adelaide and Melbourne can cause people to think that problems exist for people living in other places where there might be far less of a problem. Heatwave awareness should be encouraged, but not inapplicable awareness.
Point 2. Definition of ‘heat wave’.
The definition of a heat wave is surprisingly vague. An internal report of 84 pages dated March 2013 by Nairn & Fawcett of Australia’s Bureau of Meteorology (BoM) is a state of the art example of heat wave wisdom (hereafter N&F 2013). It uses a concept of EHF (Excess Heat Factor) which I cannot find defined in a way that allows replication. There is an extended definition in a report by Price Waterhouse 2011 that has in its Appendix B -
The Excess Heat Factor (EHF) value is constructed by multiplying a factor that represents a short-term temperature anomaly factor (heat stress) by a long-term temperature anomaly (excess heat).
These factors are defined as follows:
Excess heat: Unusually high heat that is not sufficiently discharged overnight due to unusually high overnight temperature. Maximum and subsequent minimum temperatures averaged over a three day period are compared against a climate reference value. This is expressed as a long term (climate scale) temperature anomaly.” Etc etc.
The number of hot days in a heat wave is obviously an important parameter in a definition. N&F 2013 is based around 3 days. Periods of 1, 2, 3, 4, 5 and 6 days are mentioned in the literature. Because drying of vegetation and other materials is important for fire propagation and because duration is important for sustained electrical consumption, I favour the longer, 6-day term, but that is merely a personal preference. The matter does not lend itself to sophistication. A period of 3 days at 45 deg, then one day at 30, then 3 more at 45, is not a defined 6 day heatwave (but it could be the “hottest week on Earth” instead thinks Jo).
Here is a table that shows how different numbers of days influence the historical ranking. As N&F 2013 note,
“At this level of interplay between multiple variables, units and outcomes, it is difficult to visualise or compare heatwaves across time or compare the severity of local, national or international events.”
Do you see how easy it is to create or rank records by choice of definition?
Point 3. Choice of weather station.
Some cities have several weather stations. The choice of one over another can include or exclude a heat wave according to its definition, as some run hotter than others. Here is an extract from the 2009 Adelaide heat wave, showing a Tmax difference of up to 6.5 degrees on a given day between Adelaide’s main station at Kent Town and Adelaide airport which is 11 km to its NNW. The frequency and warmth of Adelaide’s heat waves form a different pattern depending on the choice of Kent Town or Airport station 23034 as shown here. Of course, this type of observation applies anywhere and places limits on the sophistication of data manipulation.
I’m uneasy using the official temperature record from Melbourne Regional station. Here is a set of daily Tmax graphs from the start of 2013 comparing that station to 3 suburban ones nearby. There are systematic differences of at least half a degree up and down that might be caused by Urban Heat Island, UHI, effects. Until the problem is sorted, I’d be wary – and this has been the poster child of how good stations should operate.
UHI would be expected to artificially raise the intensity of heat waves, but by how much exactly? This comment is from Wales at. al., 2012 –
Morris has been conducting an ongoing study on UHI effects in the wider urban area of Melbourne. The study uncovered that UHI is present in the central business district (CBD) and industrial suburbs of the city. The average temperature from 1985 to 1994 in these areas was 4°C higher than the average temperature in rural suburbs during summer and 3.2°C higher during winter. The temperature of the City of Melbourne’s CBD can be up to 7°C higher than other suburbs in hot weather.
Point 4. Choice of weather record.
One could write a book about the ups and downs of Australia’s weather record. It is shameful that none has been written by an expert. I have never been privileged to see any original raw data recorded by pen and paper or automatically or whatever. I do not know if the BoM release it to the public. The Climate Data Online at the BOM web site could be raw data – I cannot confirm if it is free of all past adjustment. A form of this was sold by the BoM as late as 2007 on compact disc. Here, I term this ‘CD2007’ for shorthand.
The next closest approach to ‘raw’ is the homogenised data set treated by Simon Torok in his doctoral dissertation and related work by Torok & Nicholls of the BoM.
There was a later set designated High Quality or HQ, in the 2000s. In 2012 the BoM replaced the HQ set with ACORN-SAT, here Acorn, a more homogenized set. There is another data set, Australian Water Availability Project (AWAP). The N&F 2013 paper references diverse international sources of Australian time series temperature records, but there comes a point where one needs to be a connoisseur of temperature, adjustments and detective work to understand the shambles.
Research and derivatives such as constants in calibration equations can be rendered useless if one data set version is replaced by another. Versions are not numbered and their variations are poorly described. Acorn, for example, starts in year 1910, long after many valuable longer records started. It is not simply that Stevenson screens mostly replaced other screens after 1910, because we are seeing temperature differences rather greater than that effect.
Here is a graph of the annual Tmax from 86071, Melbourne Regional, Tmax, where annual temperatures from Acorn have been subtracted from the ‘raw’ CD2007. I do not think that Nature acts like this, in hindsight.
Point 5. Quality of data keeping.
There has been commendable record keeping at times, though long periods when advanced uses of data were probably not contemplated. Unfortunately, we seem to be getting worse.
For example, Melbourne regional has these temperatures at the given dates. You have to spot the Acorn problem to see if you are paying attention. There are many more examples that question quality control. The Acorn adjustments inexplicably result in some days when the maximums have been adjusted to be below the minima, or vice versa. Did nobody even did a basic check to notice?
It is time for the BoM to publish the practical magnitude of errors of individual temperature readings. Can it be less than the magnitude of these adjustments?
I’ve tried various ways to present heat wave data, all simple. The following way minimises past step adjustment effects and introduces another complication, long term cycles. There is no space to develop this, but cycles are another factor hampering too sophisticated an analysis. In the end result, the value of the study of extremes is useful into the future only so far as a regular weather forecast. The date that they will happen cannot be predicted in the longer term.
A final thought on raw ingredients
During several trips to remote parts of China decades ago, I found the custom to include a convivial trip to the kitchen before dining, to meet the chef and to view the raw ingredients for quality. Later, one would eat the raw ingredients that could be cooked in a variety of ways, by various secret recipes, some to taste, some not.
So it could be with climate data, but for now, few ‘outsiders’ get to view the raw ingredients. We have to rely upon the cooked product, recipe often unknown.
I’m saying it is time to return the meal to the kitchen and to have a refund of the credit card charge. It tastes like the raw ingredients have been overcooked.