Understanding Climate Risk

Science, policy and decision-making

The big wet Part 1

with 6 comments

In this long post I describe how climate change and variability have contributed to the big wet of the past year and outline a role for climate change that challenges the orthodox model of understanding how climate changes through trend analysis.

Last year was the second wettest year recorded in Australia since 1900. Deadly floods occurred in Queensland and a series of increasingly severe floods affected Victoria and New South Wales during spring and summer. There is no doubt the origin of the wet year was due to climate variability. The strongest La Niña on record with an SOI of 21.5 from Sep–Dec was the origin of the wet conditions.

  • Total Sep–Jan rainfall in eastern Australia was 602 mm second only to 1972–3, just beating 1974–5. Sep–Dec rainfall was 425 mm, the wettest on record.
  • The El Niño-Southern Oscillation (ENSO) Index averaged 21.5 over the Sep–Feb period; the strongest on record, producing very strong La Niña conditions.
  • The Indian Ocean Dipole (IOD) was negative between Aug–Oct 2010. A negative IOD is characterised by warmer than usual waters near Indonesia. The combination of a negative IOD and La Niña is rare and results in wetter than usual conditions across much of Australia (see http://www.bom.gov.au/climate/IOD/negative/).
  • Victoria recorded its wettest January at 119 mm and Sep–Jan at 532 mm.
  • Pan evaporation and diurnal temperature range over eastern Australia in 2010 were the lowest on record.

Neville Nicholls has an excellent summary as part of his March president’s column for the Australian Meteorological and Oceanographic Society. He says:

But was the impact of the 2010/11 La Niña on Australian rainfall stronger because of the record warm sea surface temperatures around northern Australia in 2010? These waters … are now about a degree warmer than early in the 20th century. If these warmer waters were enhancing the impact of La Niña on Australian rainfall we might expect to be seeing heavier rains in recent decades, relative to the rains that accompanied earlier strong La Niña events. There is some evidence of this (e.g., Nicholls et al 1996), and there has been a weak tendency towards increased rainfall since 1900, independent of the influence from the El Niño – Southern Oscillation.  Perhaps this trend towards increased rainfall might be related to the warmer sea surface temperatures – but much more work is needed to test this. The effect, if there is one, does not look very strong.

I’ll have a go at doing that here. I’m working on a hypothesis that human-induced climate change proceeds in a series of steps, rather than as a trend. This work has been bubbling along for two decades, but has accelerated as I’ve investigated the step change in rainfall and temperature that occurred in SE Australia in late 1996. The basics come from hydrometeorology, where heat fluxes between the surface and the atmosphere are dominated by latent and sensible heat. Latent heat is the energy wrapped up in evaporation, so follows the water cycle. Let’s call it coolth. Sensible heat is warmth. The two are partitioned in the environment, where latent heat is governed by energy and moisture and sensible heat is measured by the temperature of the environment.

So when it is wet, the environment is cooler, evapotranspiration is higher and potential evap is lower. When it is dry, the environment is dominated by warmth rather then coolth, actual evap is low and potential evap is high. The robustness of this relationship was shown by Budyko, the famous Russian scientist in the mid 20th century. The ratio between warmth and coolth varies linearly in the complementary relationship, coined by Bouchet in the early 1960s. This relationship is a gift to the hydrometeorologist and means that most of the action describing the behaviour of energy and moisture in the environment can be described with a couple of equations. There is a proviso – it only works averaged over time (at least a day, preferably several) and is large scale – more than patch size. The complexity of the hydrological cycle balloons at small space and time scales, making precise predictions of hydrological behaviour extremely difficult (and not all process are that well understood), while generalised estimates are comparatively easy.

So, it’s a great tool for climate studies and for making pretty good guesstimates about environmental change. Here’s a picture of the complementary principle published in a 2005 paper. If you follow  the solid lines, as rainfall goes up so does actual evap (Ea) and potential evap (Ep) comes down. Where those two lines meet is the point at which the environment is not moisture limited, and where Ep=Ea is called wet area evap.

Schematic of the complementary relationship

Setting aside the point that water has a different albedo to vegetation, if you put an evaporation pan in a dry paddock, it will evaporate like crazy, if you flood the paddock, pan evaporation goes right down. Because pan evaporation is a proxy for Ep, Ep will follow the same relationship.

The point to all of this is that it explains the relationship between temperature and rainfall in a particular place. High rainfall, in addition to the added cloud cover blocking radiation and limiting inward energy during the day, will lead to higher evaporation, therefore more coolth and less warmth (sensible heat). Annual rainfall (P) and maximum temperature (Tmax) are highly anti-correlated: -0.65 in south-eastern Australia 1910–1967. What happens if the climate warms?

The next figure is what we theorised would happen back in 2005. The extra energy raises the whole relationship. The example shows what can happen even if rainfall increases from P to P´, Ep can still go up, and so will Ea. If rainfall goes down, Ep will rise even more.

Complementary relationship under climate change

To get back to temperature, it is possible using the high correlation between the two to estimate Tmax from P:

P = α + βTmax,

The estimate of Tmax produced from this equation is referred to as TmaxP. It is estimate of temperature you would get if it was totally depedent on rainfall. An additional strong relationship is between Tmin and Tmax (correlation 0.47 1910–1967).

A number of Australians have worked on these relationships: Mike Coughlan , Scott Power et al., Neville Nicholls et al. and David Karoly & Karl Braganza who looked at observations and models. They all investigated these relationships as trends. Interestingly there has been little interest about this approach in other countries. Is it too simple?

Using the implication from the above figure – if you inject more sensible heat into the climate from climate change what happens to the relationship between Tmax and P, and Tmin and Tmax? I hypothesised that the β variable would remain constant, while the α variable increased. That suggests for a region that if it warms, the relationship between the partitioning of sensible and latent heat remains the same as it did before. That is, the slope remains the same while the constant increases. I assumed the same thing for TminTmax. This is not as safe an assumption because of potential land surface feedbacks that I won’t go into.

I use a test called the bivariate test to estimate shifts in time series: it gives dates, amounts and significance. The relationship with Tmax and P and Tmax and Tmin in south-eastern Australia is stationary until 1972, then Tmin increases relative to Tmax. Tmax increases relative to P in 1996, with a non-significant shift in 1971.

The relationship between Tmax and P, and Tmin and Tmax for 1910–1972 was then used to estimate TmaxP and TmaxTmin to 2009. What did it show? It showed that the residual of the relationship did not emerge until the late 1960s for Tmax and maybe a little earlier for Tmin. I call the residual for Tmax, TmaxARW and for Tmin, TminARW. ARW stands for anthropogenic regional warming. The statistics show that the shift for TminARW occurred in 1968 and for TmaxARW, 1973. A second shift for TmaxARW occurred in 1997.

11-year running means of Tmax, TmaxP and TmaxARW (top) and Tmin, TminTmax and TminARW (bottom) for south-eastern Australia

1968 is an interesting date. It is the date from which rainfall in SW WA started to decrease. Fredricksen et al. identify a reduction in storm generation in southern Australia from that date. A statistically significant change in low pressure system central pressure and density occurs from that time. I then had a look at the GISS zonal, hemispheric and global average temps to look for step changes. The zone 24–44°S and the southern hemisphere average shifts upward by 0.4°C in 1969 and 44–64°S by 0.3°C in 1972. The next shift for all three zones is in 1997. All statistically significant. There is no trend between these shifts.

I think the variation of shift dates for Tmax in 1973 and Tmin in 1968 is due to interannual variability; the series of wet years in the early 1970s could be affecting when the step change registers.

My contention is that TmaxP and TmaxTmin is a measure of climate variability, and TmaxARW and TminARW are a measure of climate change. And that the latter have been stepping, rather than trending upwards. And furthermore, if the other areas of the globe are looked at, they all do similar things but with different timing. Except for 1997, when nearly all regions shift upwards, contributing to a statistically significant upward step of 0.3°C at the global scale.

What this method doesn’t account for is any human influence on rainfall. To do that, you need an independent variable – not Tmax. Higher temperatures may be contributing to increased rainfall as suggested by the theory, models and now some observational studies. Neville Nicholls showed  a graph of southern Oscillation Index and Sep 10–Feb 11 rainfall, to illustrate there is no evidence of climate change in the SOI record and a weak trend in rainfall. Hence his conclusion that The effect, if there is one, does not look very strong.

I took the same view that SOI was not being influenced by global warming, regressed Sep–Feb rainfall against it and tested the residual for a step change. Hey presto, there is a statistically significant shift in 1973, the same year as Tmax shifts. Rainfall is higher for a given value of the SOI after 1973 than it was before. The pre and post 1973 residual is also different according to the t-test (used logarithms to reduce skewness in the sample). Sep–Feb rainfall in eastern Australia shifted upwards relative to SOI by 14% in 1973, the same date that earlier work by Nicholls suggests there was a change. Curious, I thought, and in the process of downloading the rainfall data spotted Australian average rainfall record. Sure enough, there is a statistically significant shift of 14% to wetter conditions in 1973. Much of this increase has been in the north and north-west while the southwest is labouring under record dry conditions and the SE was in drought from late 1996 until this year.

Residuals of the Sep-Feb rainfall regressed against SOI for eastern Australia

This is strong evidence that the warmer conditions in the southern hemisphere since the late 1960s, measured as an upward step change in regional and local temperatures, have enhanced Australian rainfall by around 14%. Climate variability will be playing a part also, so the actual amount is approximate. And because this is a climatological analysis, it is still hard to attribute individual events. The wet conditions last year and early this year were due to natural climate variability, but it was probably turbo-charged by warmer conditions as have the years from 1973. Because flood damages are highly non-linear, this means there is a premium being paid for the floods we are experiencing that is due to human-induced warming.

On the other hand, it also means the dry conditions in southern Australia are wetter than they otherwise would have been if solely due to climate variability. If they are due to climate change though, that is little consolation. The date of 1968 for drying in the west and 1996–7 for the onset of dry conditions in the east and drier conditions in the west matches regional step changes in temperature too closely to be attributable to climate variability. There looks to be a strong role for climate change, a conclusion that has been supported by other analyses.

More information on individual situations requires sufficient modelling to simulate a climatology of specific conditions, not a trivial task. Therefore this is a general conclusion that cannot be used to attribute proportional attribution to specific events.

I’m aware that the line of reasoning in this post is complex. If people get this far and have trouble following it, or wish to question the assumptions and analysis, feel free to respond. One aim in posting this is to improve my communication of this rather complex material.

References and further reading
Bureau of Meteorology (2006) An exceptionally dry decade in parts of southern and eastern Australia: October 1996-september 2006. Special Climate Statement, vol 9. Bureau of Meteorology, Melbourne
Coughlan MJ (1979) Recent variations in annual-mean maximum temperatures over australia. Quarterly Journal of the Royal Meteorological Society 105:707-719
CSIRO, BoM (eds) (2007) Climate change in Australia. CSIRO, Melbourne
Fredriksen, J., Fredriksen, C., Osbrough, S. and Sisson, J. (2010) Causes of changing Southern Hemisphere weather systems. In Managing Climate Change: Papers from the Greenhouse 2009 Conference (eds Jubb, I., Holper, P. and Cai, W.J.). CSIRO Publishing, Melbourne, pp. 85-98.
Karoly DJ, Braganza K (2005a) Attribution of recent temperature changes in the Australian region. Journal of Climate 18 (3):457-464. doi:doi:10.1175/JCLI-3265.1
Karoly DJ, Braganza K (2005b) A new approach to detection of anthropogenic temperature changes in the Australian region. Meteorology and Atmospheric Physics 89 (1):57-67. doi:10.1007/s00703-005-0121-3
Maronna R, Yohai VJ (1978) A bivariate test for the detection of a systematic change in mean. Journal of the American Statistical Association 73 (363):640-645
Murphy BF, Timbal B (2008) A review of recent climate variability and climate change in southeastern Australia. International Journal of Climatology 28 (7):859-879
Nicholls N (2003) Continued anomalous warming in Australia. Geophysical Research Letters 30 (7):1370. doi:10.1029/2003gl017037
Nicholls N (2010) Local and remote causes of the southern Australian autumn-winter rainfall decline, 1958–2007. Climate Dynamics 34 (6):835-845
Nicholls N, Dellamarta P, Collins D (2004) 20th century changes in temperature and rainfall in New South Wales. Australian Meteorological Magazine 53:263-268
Power S, Tseitkin F, Torok S, Lavery B, McAvaney B (1998) Australian temperature, Australian rainfall, and the southern oscillation, 1910-1996: Coherent variability and recent changes. Australian Meteorological Magazine 47:85-101
Timbal B (2009) The continuing decline in south-east Australian rainfall – update to may 2009. CAWCR Research Letters 2:4-11
Timbal B, Arblaster J, Braganza K, Fernandez E, Hendon H, Murphy B, Raupach M, Rakich C, Smith I, Whan K, Wheeler M (2010) Understanding the anthropogenic nature of the observed rainfall decline across south-eastern Australia. CAWCR Technical Report. The Centre for Australian Weather and Climate Research, Melbourne
Ummenhofer CC, England MH, McIntosh PC, Meyers GA, Pook MJ, Risbey JS, Gupta AS, Taschetto AS (2009) What causes southeast Australia’s worst droughts? Geophysical Research Letters 36 (4):L04706. doi:10.1029/2008gl036801
Vivès B, Jones RN (2005) Detection of abrupt changes in Australian decadal rainfall (1890-1989). CSIRO Atmospheric Research Technical Paper, vol No. 73. CSIRO Atmospheric Research, Melbourne


Written by Roger Jones

March 20, 2011 at 4:32 pm

6 Responses

Subscribe to comments with RSS.

  1. Thanks Roger. Yes it was a bit too technical for this layman, but I believe I followed the gist of your reasoning.

    I now await George Pell’s refutation (after all, he spends a lot of time studying this stuff).


    March 20, 2011 at 5:39 pm

  2. Roger: quick question – what is the meaning of Tmin_{Tmax} (in the fourth paragraph down from your second figure)? Is it just a typo? Or do you just mean Tmin(Tmax) (i.e. Tmin as a function of Tmax)?


    March 20, 2011 at 7:08 pm

    • Jess: the formatting in your question hasn’t survived so I’m not sure where exactly you are referring to.

      Roger Jones

      March 20, 2011 at 11:36 pm

      • Sorry about how late this is – I’m in NZ atm and have been away from the computer.

        My question was really a minor quibble about your notation – using subscripts like that to denote a function dependence between your temperature estimated from P is a bit confusing to read. Might it be better to do something like T_{est}(P), which makes it clear which are estimated temperatures, and what they’ve been estimated from?


        March 25, 2011 at 8:20 am

  3. Roger, not having any maths to speak of it becomes hard going when you introduce that formula after the second figure. I gather the take-out message is this:

    The wet conditions last year and early this year was due to natural climate variability, but it was probably turbo-charged by warmer conditions as have the years from 1973. Because flood damages are highly non-linear, this means there is a premium being paid for the floods we are experiencing that is due to human-induced warming.

    On the other hand, it also means the dry conditions in southern Australia are wetter than they otherwise would have been if solely due to climate variability. If they are due to climate change though, that is little consolation. The date of 1968 for drying in the west and 1996–7 for the onset of dry conditions in the east and drier conditions in the west matches regional step changes in temperature too closely to be attributable to climate variability. There looks to be a strong role for climate change, a conclusion that has been supported by other analyses.

    Which is in plain English. I’d be looking for a similar plain English statement as to why the step changes happen. I think I understand it but wouldn’t attempt a summary myself, which means that I’m stll unsure.

    This is beyond the scope of the post, but I’m still wondering what causes changes in ENSO other than it happens because it happens.

    I think one minor glitsch. I think the Indian Ocean dipole is negative rather than positive. The only way I can remember is that it is the opposite to what you would think.


    March 21, 2011 at 11:15 am

    • Brian,

      thanks – corrected (I tend to get the positive temperatures/indices mixed up).

      The step changes in temp are really clear once they’re analysed and shown. I will post on this soon. As to why, I have a hypothesis but it requires looking at sea level and ocean heat flux, which is where I think the answer is. I’ll get the aid of those far more capable than I to address that.

      Roger Jones

      March 21, 2011 at 11:32 am

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: