Home > General > The browning of Australia

The browning of Australia

September 30th, 2006

Reader Proust points me to this helpful BOM site showing rainfall trends in Australia. You can choose your own region, season and time period.

Here’s the most relevant to consideration of the effects of global warming, the trend since 1970, which demonstrates how much drier the climate has become over the period in which warming has been observed. As various people have pointed out, is was even drier during the famous Federation drought at the beginning of C20, so the role of global warming isn’t conclusively established, but it would certainly seem unwise to bet on a rapid return to the average observed in the historical record

Rainfall trend

Categories: General Tags:
  1. proust
    September 30th, 2006 at 19:45 | #1

    Heh. Nice try. But dishonest (or just dumb, either way pretty sad coming from soneone in your position).

    Here’s the 1900-present trend map. Professor Q, how about you post both side-by-side and then explain why your map has nothing to say about AGW whatsoever?

    (hint – it’s a first-year statistics exercise).

    Proust – please note policy on civilised discussion. Your comments so far have been uniformly rude, and (the two usually go together) uninformed. The point about 1900 being drier is noted in the post, and had been discussed at length before you raised it. Also, if you want to make claims implying some degree of expertise in statistics, either demonstrate it, or use a real name and state your qualifications. JQ

  2. scott
    September 30th, 2006 at 22:10 | #2

    Great tool BOM.

    Of course, as they point out “because rainfall is such a variable element, trend values are highly dependent on the start and end dates of the analysis” and the fact they are simply using linear interpolation it is very difficult to derive anything meaningful in terms of climate change from just one map.

    Analysing each plot in sequence from 1910 to 1970, however, and the trend towards drying in the east and wetting in the northwest seems fairly consistent.

  3. scott
    September 30th, 2006 at 22:13 | #3

    Here is an animation illustrating this.

  4. stephenl
    September 30th, 2006 at 22:39 | #4

    Actually, the 1970-2005 and 1900-2005 maps have a lot in common. Drying in coastal WA, Queensland and Tasmania, increases in rainfall in Northern WA (which doesn’t need it) and inland WA (where it is of no economic use). The bid difference is in NSW and Vic. (And in inland SA where rainfall is so low to start off with that it won’t make much difference).

    What has happened is that from 1900-1970 trends were in different directions in different parts of the country and generally not particularly strong. Since 1970 we have seen exactly what global warming models predict – more rainfall in the North-West and some desert areas and less in the major agricultural regions. In some cases this has either combined with, or overcome, pre-1970 trends. In other places it might take a bit longer. Nevertheless the signature of global warming looks pretty clear.

  5. September 30th, 2006 at 22:56 | #5

    Well, my concern is that the scale stops at -50 mm/10 years. It could be a lot higher in parts of the brown zone: -60 or -70 or even -100mm/10 years. (Aka, one less cm of rain for each year.) Seriously, if you have roughly half of Queensland falling into that category, I think we need BOM to provide more colors to show the full picture.

    I’m not so worried about places like Cairns – but global warming could ruin inland farming regions like the Darling Downs.

  6. October 1st, 2006 at 01:22 | #6

    Its an interesting map. Why has rainfall increased in the West? Roughly rainfall seems to get worst as you go from East to west. StephenL suggests this is consistent with global warming models. Why?

    Would not global warming be expected to produce a uniform average decrease?

  7. rog
    October 1st, 2006 at 07:28 | #7
  8. proust
    October 1st, 2006 at 08:18 | #8

    Your comments so far have been uniformly rude, and (the two usually go together) uninformed.

    How so? Put up the 1900-present map alongside the above, and I’ll believe that you’re not trying to dishonestly distort the debate.

    Also, if you want to make claims implying some degree of expertise in statistics, either demonstrate it

    Hmm, ok, start simple.

    - Bayes rule: P(A|B) P(B) = P(B|A) P(A)

    [I've always preferred writing it like that because then it is manifestly symmetric beyween events A and B]
    Bayes rule is the fundamental building block of probabilistic reasoning. However, an infrequently mentioned caveat: it doesn’t work if either A or B are events of zero probability. “Conditioning on events of zero probability” is naughty. Remember those sneaky proofs of “1=0″ you discovered in high-school? They always relied on dividing by zero somewhere. Well, that’s the equivalent of using Bayes rule to reason about events of probability zero – you can establish anything you like.

    Bit more complicated, but no less fundamental is the

    - Strong Law of Large Numbers: The sample mean of any infinite sequence of independent and identically distributed (iid) random variables with mean x converges almost surely to x.

    The “Strong Law” is why we can represent samples by averages.

    What’s this weasely “almost surely” you ask? Well, it really means what it says. It means that the measure (or probability) of the samples whose mean does not converge to x is zero. Or if you’re into tossing coins, it means that the probability of the average number of heads in a long sequence of tosses differing significantly from one half gets smaller and smaller (towards zero) as the length of the sequence of tosses increases [strictly, that last statement was closer to a statement of the weak law of large numbers, but we can leave the distinction between the two laws for another day]

    Since we’re talking trends here, how about a bit of Stochastic Calculus:

    - Ito’s Equation: If x(t) is a generalized Wiener process

    dx = a dt + b dWt

    then f(x(t), t) is also a generalized Wiener process with:

    df = (a f_x + f_t + 1/2 b^2 f_xx) dt + b f_x dWt

    [f_x means the partial derivative of f wrt x. f_xx is the scond partial derivative]

    Suprisingly, if you’re cavalier with interchanging limits and integrals, Ito’s formula is trivial to prove.

    But, what is a Wiener process I hear you ask? No, it is not a sausage machine. It is the equivalent of a normal distribution for processes. Wt – Ws behaves like a normal distribution of (t-s). I won’t go into more details, except to say that the financial markets (particularly hedge funds) run on Ito’s formula and extensions (eg Black-Scholes option-pricing is just Ito’s formula applied to options).

  9. David
    October 1st, 2006 at 08:34 | #9

    Harry – I was thinking the same thing about WA. I wonder could it be the influence of an increasing number of cyclones (caused by global warming???) hitting the north west coast and moving inland in a south easterly direction, providing brief but heavy rain to an otherwise dry area? Just speculation, I have no data.

  10. jquiggin
    October 1st, 2006 at 09:19 | #10

    OK, Proust, you can talk the talk. That makes it easy.

    In the previous thread, I gave an informal Bayesian argument for the proposition that AGW is leading to generally hotter and drier weather in SE Australia. You should have no trouble filling in the details, and pointing out where you think I’ve gone wrong. Alternatively, set out your own priors and show why you think the proposition should be rejected.

  11. chrisl
    October 1st, 2006 at 10:33 | #11

    The map shows that the Australian rainfall is variable and you farm it at your own peril

  12. Ian Castles
    October 1st, 2006 at 11:45 | #12

    John, You say that the BoM map of the trend in rainfall in southeastern Australia since 1970 “demonstrates how much drier the climate has become over the period in which warming has been observed.” Even if that were true, you would only be demonstrating a correlation between rising temperatures and declining rainfall, not that one had caused the other.

    But “the period in which warming has been observed” did not begin in 1970. In fact, over half of the increase in the 11-year moving average of temperature in the BoM’s “Southeastern Australia” between 1950 and 2000 occurred BEFORE 1970. (I haven’t checked my arithmetic, but this looks to be right from the BoM chart). Yet the early 1970s were uncommonly wet, and the BoM’s very helpful sorted data shows that 1973 was the second wettest and 1974 was the third wettest year in the 106-year instrumental record of rainfall in southeastern Australia.

    So an obvious question presents itself: if the warming of about 0.4 degrees C in temperatures in southeastern Australia in the 1970s, 1980s and 1990s caused the lower rainfall of the recent past, why was the warming of 0.4 degrees C in the 1950s and 1960s followed by a period of above-average rainfall?

    But the cause-and-effect argument that you are mounting must confront a more serious problem: according to my calculations, the seasonal pattern of the recent change in rainfall in south-eastern Australia is completely at odds with CSIRO’s projections.

    The paper prepared by CSIRO and published by the Australian Greenhouse Office that I’ve already cited (“Climate change scenarios for initial assessment of risk in accordance with risk management guidance”, May 2006) provides projections FOR EACH SEASON of the change in rainfall between 1990 and 2030 for ten Australian regions, two of which (NSW and Victoria) make up “southeastern Australia.”

    Take NSW first. The CSIRO projections are for an INCREASE of 1.5% (low global warming scenario) and 3.5% (high warming scenario) in rainfall in summer and autumn; and for a DECREASE of 3% (low global warming scenario) and 7.5% (high global warming scenario) in winter and spring. I’ve checked the BoM data for rainfall for the most recent 11-year period for which the information is available (Sept. 1995-August 2006) with the immediately preceding 11-year period (Sept. 1984-Aug. 1995), and I find exactly the opposite. In summer and autumn, when rainfall should have increased with global warming according to the CSIRO’s projections, there was a decrease of 16% in rainfall in the more recent period. But in winter and spring, when CSIRO projected decreases in rainfall, my calculations show an increase in rainfall in 1995 to 2006 compared with 1984 to 1995.

    Similarly for Victoria. The CSIRO projections are for no change in rainfall in the summer, both in the low and the high global warming scenarios. My calculations for Victoria from the BoM data show a larger reduction in rainfall in summer (14%) than in average annual rainfall (13%) in the more recent period. For spring, it is the opposite. The CSIRO projections show larger reductions in spring than in any of the other seasons (5% for the low global warming scenario, 11% for the high global warming scenario), But the BoM data of actual rainfall show that the smallest reduction in rainfall in the more recent period was in the spring (down 4% compared with reductions of of 14% in summer, 16% in winter and 21% in autumn).

    These figures are E&OE – I’d be happy to have them checked, perhaps by one of the water supply planners who are using climate model results that predict permanently lower levels of rainfall in southeastern Australia.

    It’s not my argument that such an assumption should not be made: that raises other issues. My point is simply that observations of the actual rainfall trend in recent times are not consistent with CSIRO model projections (I will of course withdraw this comment if my figuring is shown to be wrong). PLease note that the CSIRO projections I am citing were published only four months ago.

  13. jquiggin
    October 1st, 2006 at 12:17 | #13

    Ian, I am well aware of the distinction between correlation and causation, and I didn’t claim that causality could be inferred from the trend alone.

    As regards your first point, I don’t think it’s appropriate to try and correlate rainfall in SE Australia with temperature in the same area. It’s my understanding (not expert) that sea-surface temperatures in the Pacific are more important for rainfall, and the Indian Ocean may also be relevant. The sustained rise in global temperatures dates from the 1970s, so I don’t see a big problem here.

    Your second point is, as you say, more serious. On the face of it, the seasonal projections aren’t supported by the trends observed so far, although the aggregate trends seem to be about right. I’ll look into this.

    On the final point, if we return to your original point, my article made the claim that it was appropriate to assume permanently lower levels, and proceeded to consider the implications. It appears you don’t dispute this.

  14. Smiley
    October 1st, 2006 at 13:07 | #14

    This doesn’t correlate with the map of predicted rainfall that Al Gore showed in his movie. He verbally noted the drying around the Sahara, but the map showed an increase in rainfall on the east coast of Australia and a reduction on the west coast.

    This weather predicting thing looks like a dangerous game to be in. A bit like playing the money markets. But I guess that you’re not going to be sent to jail for getting your weather predictions wrong (though the farmers might not be happy).

  15. Ian Castles
    October 1st, 2006 at 13:40 | #15

    John, If it’s the increase in global temperatures that’s relevant to southeastern Australian rainfall (which I’m happy to accept), the same problem arises. The increase in global temperatures between 1910 and 1945 was comparable with that since 1970: see Table 2.1 of the IPCC Third Assessment Report at http://www.grida.no/climate/ipcc_tar/wg1/053.htm .

    So I’ll reformulate my question: “if the increase of about 0.5 degrees C in global temperatures since 1970 caused the recent downward trend in rainfall in southeastern Australia, why was the increase of this order in global temperatures between 1910 and 1945 followed by the strong increase in rainfall in southeastern Australia which occurred in the 1950s?

    On the final point, I’m merely trying to distinguish the issues. Your article made the claim that it was appropriate to assume permanently lower levels of rainfall because of global warming. If the modelled effects of global warming are not being borne out by the observations, as it appears may be the case, I wouldn’t see any reason arising from the grounds that you cited to assume permanently lower levels of rainfall.

  16. Ian Castles
    October 1st, 2006 at 14:22 | #16

    I hadn’t noticed Roger Jones’ informative posting on the “Water again” thread until now. I probably would have expressed some of my points differently if I’d read Roger’s comments, but I think most of my arguments stand. I’d be interested in Roger’s reaction to them.

  17. jquiggin
    October 1st, 2006 at 15:19 | #17

    Ian,

    I have a number of problems with your revised question, and before I can respond, I need a better idea of where we agree and disagree. As I’ve made pretty clear, I think the “consensus” view on climate change is broadly correct. In particular,

    1. There is a clear global warming trend, both at the surface and in the troposphere
    2. This warming is due, to a substantial extent to human activity, particularly emissions of greenhouse gases
    3. The cooling observed in the mid-20th century was in part due to emissions of particulates, and reductions in these emissions will tend to exacerbate warming
    4. At the global level, climate models represent these processes reasonably accurately. Analyses at the regional and seasonal level are less reliable, but still represent the best available basis for future planning (in particular, better than planning on the basis of distributions derived from C20 experience)
    5. Under any plausible interpretation of “business as usual”, there is a high probability of warming of 2 degrees or more, relative to baseline, by 2100
    6. Such warming would have severe environmental consequences

    For the purposes of our current discussion concerning experience from 1900 to the present, only 1-4 are really important. If you broadly agree with these, I think I can respond usefully to your question.

  18. Hermit
    October 1st, 2006 at 16:36 | #18

    The map may need revising at end 2006 because the greenish WA wheatbelt is much drier this year. Whatever statistical hypothesis testing reveals, I believe most Australians over 30 now believe rainfall patterns are different. This can’t be all due to biased memory. On the other hand they don’t want to pay for it in terms of food prices, water restrictions, carbon taxes and so on. I’d call it begrudging acceptance.

  19. Ian Castles
    October 1st, 2006 at 16:52 | #19

    Thanks John. I didn’t put my “final point” correctly – what I should have said was that, if the current understanding of the causes of changes in regional rainfall is as low as it appears to be, it would not be appropriate to count on a recovery to levels that have prevailed in the past – and that would be true irrespective of the view taken on the implications of global warming.

    My question as reformulated was the one that I would have asked in the first place if I had realised that your reference to “warming [which] has been observed” related to changes in mean temperature at the global level. As the title of the thread is “The Browning of Australia” and you’d said in your introductory post that “the role of global warming isn’t conclusively established”, I didn’t think that my revised question would give you problems. But it is not a matter of great importance.

  20. Graeme Bird says:
    October 1st, 2006 at 19:15 | #20

    We went over this at Prodeo.

    And I discovered (as it seems did proust) that the only significant thing it showed was that the 1970′s were likely an excellent year for rainfall in Australia.

    The other thing is they cut the time-series off earlier then they needed to. Which was tendentious since it is implied that having made a diagram with the later information would have shown almost nothing untoward.

  21. chrisl
    October 1st, 2006 at 21:38 | #21

    I think we need a new word here..
    Castled(verb)
    To be castled.. To be shot down in flames statistically

  22. jquiggin
    October 1st, 2006 at 21:55 | #22

    Or, alternatively

    Christled(verb)

    To be christled… to receive support of negative value, or criticism that increases your credibility.

  23. chrisl
    October 1st, 2006 at 22:17 | #23

    Touche
    But you’ve got to admit he is very good

  24. proust
    October 2nd, 2006 at 08:37 | #24

    In the previous thread, I gave an informal Bayesian argument for the proposition that AGW is leading to generally hotter and drier weather in SE Australia

    I searched the thread for the string “Bayes” – I could not find it. Could you please reproduce the argument here? Feel free to put it in mathematical terms – the more precise the better – that way there can be no confusion.

    Sure. My propositions are:

    A: Global warming implies drier weather in SE Australia

    (Here ~A encompasses everything else “Global warming is not taking place or does not imply drier weather in SE Australia”)

    B: Hot, dry weather as observed in the period 1995-2006 occurs

    My prior probabilities are Pr(A)=0.5, Pr(B|~A)=0.25, Pr(B|A)=0.75. Applying Bayes rule, the posterior probability for A|B is 0.75, as claimed.

  25. melanie
    October 2nd, 2006 at 16:34 | #25

    Suppose we take 1900 as the base, admittedly an exceptionally dry period. The Queensland coastal areas still present a problem, as does southwest WA.

    What if you statistiians were to plot the rate of rise/decline in rainfall against the rate of increase in water consumption in the east?

  26. Ian Castles
    October 2nd, 2006 at 18:59 | #26

    John, You said on the “Water, againâ€? thread that you thought the CSIRO projections were for lower rainfall for SE Australia generally, but that you’d look at the sources I’d cited. Have you had a chance to look at the AGO publication “Climate change scenarios for initial assessment of risk …â€?? For NSW the CSIRO central projections are for NO CHANGE both for the low and high global warming scenarios, and for Victoria their projections are for fractional decreases of 1.5% and 3.5% respectively over the 40-year period 1990 to 2030. During this period CSIRO’s projections of global temperature (based on the flawed SRES scenarios) are for increases of 0.54 deg. C and 1.24 deg. C.

    In short, CSIRO’s expectation of change in annual rainfall in SE Australia, on a decadal timescale, ranges from nil to imperceptible. Moreover, as I’ve shown, seasonal variations have generally been in the opposite direction to those projected on the assumption of global warming. These differences between projected and observed trends in rainfall seem to raise serious questions about the ability of the models to predict changes in rainfall – though I’d be interested in CSIRO views, especially on whether it is appropriate to use successive 11-year averages as measures of outcome and, if it is not, how the relationship between projections and outcome should be monitored.

    You also mentioned that there was an expectation of drier weather in SW Australia, and that this had influenced your thinking in terms of rejecting the hypothesis of no change. I’ve now looked at the observed rainfall in SW Australia, using the same successive 11-year periods as I used above for NSW and Vic. Again I find a uniformly negative correlation between the CSIRO projections of rainfall and the observed outcomes. In SW Australia there was virtually no change in the observed annual rainfall between the running average for 1984-95 and that for 1995-2006 (a decrease of 1%). But here too there was a marked change in seasonal patterns, and here too it was in precisely the opposite direction to the CSIRO projections.

    These assumed that the largest decreases in rainfall would be in winter and spring (decreases of 5% and 11% between 1990 and 2030 on the low and high global warming scenarios respectively), but rainfall was fractionally HIGHER in winter/spring in the more recent period (1995-2006) than in the previous 11-year period. In summer and autumn the CSIRO projections were for smaller decreases in rainfall than in winter and spring, but the observed change was a substantial decrease: in fact, as large a decrease between the successive 11-year periods as CSIRO projected on the high global warming scenario over the 40-year period from 1990 to 2030. This too raises seems to raise questions about the CSIRO’s projections.

    I’d be interested in your views.

  27. rog
    October 2nd, 2006 at 20:01 | #27

    From the BOM the average rainfall in Australia, over all the sites, was 400mm and in 2000 it was 500mm (and in the ’70s it shot up to 800mm.)

    In 25 words or less, is this a drier or wetter climate?

  28. jquiggin
    October 2nd, 2006 at 20:24 | #28

    Rog, you’re making less sense than usual (which is not much). At a minimum, there’s a date missing somewhere, and you appear not to have followed the discussion of regional variations. And please avoid peremptory demands for responses – if there’s one thing I can’t stand it’s a bossy troll.

  29. melanie
    October 2nd, 2006 at 20:28 | #29

    The sea surface maps are interesting. If you look at the 1900-2005 map, there is a slight warming all around. Now look at 1970-present. Who wants to tell me this doesn’t have anything to do with the rainfall pattern over the same period?

  30. melanie
    October 2nd, 2006 at 20:29 | #30

    Sorry. Can’t post images in comments. Here’s the link.
    http://www.bom.gov.au/web01/ncc/www/cli_chg/trendmap/sst/0112/aus/1970/latest.gif

  31. jquiggin
    October 2nd, 2006 at 20:30 | #31

    Ian, looking at the projections it certainly appears that increased evaporation plays a bigger role than decreases in rainfall, but of course the two interact, and the reduction in streamflow is larger again (commonly by a factor of three) than the reduction in rainfall net of evaporation. Since reductions in streamflow are the variable of interest for most of my work, I’ve tended to focus on this, and I’ll be more careful to say so in future.

    As regards the issue of seasonality, while one would expect less reliability at the seasonal level than the annual level, there certainly appears to be a problem here. I’ll look into this.

    As I mentioned above, in responding to other points you’ve raised it would be very helpful if you could clarify your own position, so that I can see how much common ground we share.

  32. Chris O’Neill
    October 2nd, 2006 at 21:41 | #32

    It’s a pity the Bureau doesn’t have an evaporation graph along with the rainfall graph because the expectation with rising temperatures is that evaporation rises. So if rainfall doesn’t also rise at the same time then you can look forward to a drier future. Regional decreases in rainfall just make the problem worse.

    In general, nearly every continental area is warming faster than the oceans and since continental evaporation is tied to continental temperatures but continental rainfall also depends on surrounding oceanic temperatures, most of the world’s landmasses can look forward to a drier future (drier meaning evaporation going up faster than rainfall).

  33. derrida derider
    October 2nd, 2006 at 21:52 | #33

    IMO it’s rash to be dogmatic on such a small sample – the current long dry may or may not be a product of ACC, and thus may or may not be a harbingerof things to come.

    But in any case our water planning has got to be better anyway, because what the current long dry tends to confirm is what people have said in the past, using the Federation drought as an example – that drought in Australia can manifest as dry *decades*, not just dry years.

  34. Ian Castles
    October 2nd, 2006 at 22:48 | #34

    Chris O’Neill, I’ve recently posted on another thread on this site the conclusions of a Workshop supported by the Australian Academy of Science and the Australian Greenhouse Office on “Pan Evaporation: an example of the detection and attribution of trends in climate variables” (22-23 November 2004). Among these was the following:

    “It is affirmed that global atmospheric warming does not necessarily mean a more drying atmosphere or a drier land surface.”

    This suggests that this issue may not be as clear cut as your posting implies. Are you able to provide some references to support your view that most of the world’s landmasses can look forward to a drier future?

    John, On the “Presentation: Precautionary Principle …” thread you told me that you think it’s “unhelpful to conflate discussion of climate-science issues like the modelling of SO2, about which none of us here know very much, with discussion of economic projections, where we can have a useful discussion.”

    Against that background, it’s not clear to me why, in responding to other points I’ve raised, you’d find it very helpful if I were to clarify my position on issues such as points 1-4 of your posting on this thread at 3.19 pm on 1 October. I’m not a climate scientist, and I don’t understand why you’re not able to answer my questions unless I state my position on climate-science issues about which I don’t know very much. To put it bluntly, why do my views matter?

  35. Roger Jones
    October 3rd, 2006 at 00:54 | #35

    Ian,

    from a climatological view, there’s no point in trying to contrast 11-year averages with climate model projections. The projections are calculated as a function of global warming over the entire period of a model run e.g. from nominal year 1881-2100, or 1901-2100 depending on the model. Therefore, the projections contain little of decadal or shorter signals of variability – we’re quite confident they have the global warming signal as represented by the model (except in desert regions, where one event can define the signal). Whether we have confidence in that signal depends on convergence in the agreement of seasonal signals from the models, their representation of current climate etc.

    This suggests that yes, as both you and John have concluded, its best not to read the attribution of a rainfall signal from acceleration of warming from the 1960s or 70s on the basis of correlation. There are too many variables affecting rainfall. One would want some supporting physics in terms of circulation changes that are reflected in models and obs or similar.

    As to your comments on the CSIRO projections of change. Yes, the rainfall projections are not large. This is a good thing. We should be grateful the numbers are not larger. However, what we found when we looked at central NSW, was that climate variability could produce swings in decadal mean rainfall of +/- 20% over the long-term mean. This means that we have benign combinations of variability and change, and really nasty combinations. Which will eventuate? (and over the next century, decadal variability will give us wetter and drier periods imposed on, and maybe interacting with, a greenhouse signal) We don’t know for sure, and so we must try and assess the risks with the info we have at hand.

    Let’s say for example, that we have a scenario in SE Aust where summer and autumn rainfall increase by +5% and winter-spring decreases by -5%. The summer rainfall will have little effect on streamflow and the winter spring-rainfall will lead (with increases in evaporative demand) to a decrease in streamflow of >10%. This is what John was referring to as the amplification effect. And why in the report CSIRO did for the Vic govt, we found few signs of streamflow increase anywhere in Victoria (link given in earlier comment above). Record extended periods of low flow in a number of rivers (including the Murray) must give us pause.

    As to the comments about temperatures in 2030 being based on the “flawed” SRES scenarios. Projections of climate change are dependent on changes in radiative forcing. For regional climate projections it does not matter overly much what assumptions underlie that forcing as long as the magnitudes of increase in radiative forcing themselves are plausible. Projected changes for rainfall and temperature in 2030 are mainly due to climate sensitivity, a scientific uncertainty. The highest of that range you quoted, the 1.24°C in 2030 is attached to the A1T scenario that has more agressive sulphate emission reduction than the other SRES scenarios. Interestingly, the recent ABARE AP6 reference emission scenario gives an upper temperature almost as high in 2030 (0.05°C lower and it is largely based on IEA projections). If more aggressive sulphate reductions were to occur, warming would be as high by that time. To suggest (by implication or otherwise) that projected emissions will somehow deliver a lower range, means showing that the entire data-set of available emission scenarios (not just SRES) are over-estimated.

    It would be interesting to try a Bayesian assessment of the different cases of rainfall reductions in the obs and how the current run of dry years affect the probabilities. Given a reduction of 10 years in rainfall what is the posterior probability of it being a short-term departure, long-term but temporary departure and/or change. The longer that period of lower rainfall extends, the higher the posterior probabilities for the latter combinations become. Given the number of possible alternatives, this may not be simple. Posteriors of 11, 12, 13 years etc will change the numbers quite dramatically, I would expect.

  36. October 3rd, 2006 at 05:40 | #36

    I was steered this way by a posting in another blog, and I am afraid that I may not understand Australian geography well enough to pass judgment. I would make one comment on the relationship between SSTs and precipitation, however. I would concur with Ian Castles discussion of the pan evap study that conditions will not be uniformly drier in a warmer world, nor would they be uniformly wetter. Physical theory would lead one to predict a faster evaporation rate, depending upon atmospheric pressure and moisture content, wind speed, etc…

    While I agree with all of you that statistical criteria are important, it is also important to identify physically meaninful relationships in deterministic equations.

  37. rog
    October 3rd, 2006 at 08:12 | #37

    I am not so sure if anything of value can be extracted from the BOM images, as they do warn “.. analysis periods starting after 1970 are considered too short to calculate meaningful trend values” and “..trend values calculated here using past observations should not be used to imply future rates of change”

    An opportune moment to apply the Precautionary Principle?

  38. Ian Castles
    October 3rd, 2006 at 08:27 | #38

    Thanks Roger, especially for the advice on the rainfall projections. I was interested to know whether it was appropriate to use successive 11-year averages to test observations against projections, and I accept your view that it’s not. I also take the point in your last paragraph that it is not simple to test how the current run of dry years affects the probability of a permanent change. Nonetheless it would seem important that this be done.

    I know that temperature projections in the period to 2030 are not highly dependent on projections of GHG emissions, and it is certainly significant that the highest projection for that year is attached to the scenario with lowest sulphate emissions. Do you have any comments on my series of postings on this subject on the “presentation: precautionary principle …” thread of this blog, beginning at 11.08 am on 21 September? The best estimates I have been able to find of sulphur emissions – those by Stern (2003), cited in van Vuuren and O’Neill, “The Consistency of IPCC’s SRES Scenarios to Recent Literature and Recent Projections”, “Climatic Change”, March 2006: Table VI on p. 38 – show a steep drop in the first half of the 1990s to a level comparable to that projected in the A1T scenario for 2030.

    I’d be particularly interested in your view on the question whether, if sulphate concentrations have been as low as A1T projects for 2030 during the past decade, you would have expected global mean temperatures to have risen more than they have.

    You say that the ABARE AP6 reference emission scenario is largely based on IEA projections.The only IEA emissions scenarios to which I give credence are the Reference and Alternative Scenarios published in the Agency’s excellent World Energy Outlook. These are fully articulated scenarios for CO2 that can be tested against realistic projections of power generating capacity, vehicle usage etc.

    The latest IEA Reference Scenario (that in WEO 2005) shows the same increase in CO2 emissions between 2000 and 2030 – the end-point of the projection – as the IPCC B1 scenario. This scenario has the LOWEST emissions in 2100 of the six IPCC illustrative scenarios, whereas the ABARE AP6 scenario projected HIGHER emissions in 2100 than any of the IPCC scenarios. So I do have difficulty understanding how it can be claimed that ABARE’s scenario is based on IEA projections. Are there IEA projections that I don’t know about?

    Of course I realise that projections of climate change are dependent on changes in radiative forcing, but projections of radiative forcing are in turn depend on projections of emissions. The latest relevant ABARE publication (“Economic impact of climate change policy”, ABARE Research Report 06-7) says that global CO2 emissions in its reference case closely follow those under the IPCC’s A2 scenario to 2030 and that the latter scenario assumes a decline in economic growth after that year (pps. 22-23). In fact, projected decadal economic growth under A2 between 2030 and 2040 exceeds 40%, a higher rate than in any of the four preceding decades (SRES, p. 471).

    Your statement that “it does not matter overly much what assumptions underlie that forcing as long as the magnitudes of increase in radiative forcing themselves are plausible” seems to me to be entirely circular. If the underlying assumptions about population, energy use etc. are implausible, the radiative forcing assumptions that depend on them can’t be plausible. As Erwin Diewert says in his comment on John Quiggin’s critique of Castles and Henderson, the SRES projections should be reestimated. I have no doubt that he would take the same view of the ABARE reference scenario and most (if not all) of “the entire data-set of available emission scenarios” to which you refer.

  39. Ian Castles
    October 3rd, 2006 at 08:51 | #39

    I meant to say that the ABARE reference scenario projects higher fossil CO2 emissions in 2100 than any of the six IPCC illustrative scenarios – some of the other 29 IPCC scenarios used in the 1.4-5.8°C are higher than ABARE’s.

  40. jquiggin
    October 3rd, 2006 at 13:06 | #40

    Ian, I’d appreciate a statement of your position for several reasons

    (1) My responses to several points you’ve raised depend on my understanding of the climate science. If you have a different view, I’d prefer to know about it first.

    (2) In any scientific debate, it’s always possible to make lots of (not necessarily consistent) criticisms of a model, but this isn’t of much value unless there is an alternative, better model.

    (3) On this specific issue, I’ve found little to be gained in arguing with those who are sceptics/denialists regarding the basic science. Even though the main arguments that gave their position some credibility (urban heat islands, discrepancies with the satellite data, the Medieval Warm period and so on) have collapsed, they don’t change their views. Rather they just keep on finding new arguments to support a predetermined position. I’ve only got a finite amount of time to allocate, and I’d rather use it productively.

  41. Ian Castles
    October 3rd, 2006 at 14:33 | #41

    John, I also have only a finite amount of time to allocate (in all probability a good deal less than you have, given our respective ages), and I’m not interested in devoting any more time to explaining my views to someone whose predetermined position is that “sceptics/denialists” keep on finding new arguments to support a predetermined position.

  42. rog
    October 3rd, 2006 at 19:29 | #42

    That was a overt display of petulance JQ, using BOM images to prove a point was always a risky enterprise.

  43. proust
    October 3rd, 2006 at 21:55 | #43

    jquiggin:

    Sure. My propositions are:

    A: Global warming implies drier weather in SE Australia

    (Here ~A encompasses everything else “Global warming is not taking place or does not imply drier weather in SE Australia�)

    B: Hot, dry weather as observed in the period 1995-2006 occurs

    My prior probabilities are Pr(A)=0.5, Pr(B|~A)=0.25, Pr(B|A)=0.75. Applying Bayes rule, the posterior probability for A|B is 0.75, as claimed.

    Hmm – and I thought you were being serious for a moment there. I thought we were going to get something sophisticated like an ARM analysis of the rainfall timeseries for SE Australia, and from that a demonstration that the current drying has a residual anomaly that cannot easily be explained by the historical variation in the data.

    Instead, we get the Milly-Molly-Mandy analysis.

    Of course you can get any number you like for Pr(A|B) if you’re allowed to plug in arbitrary numbers for Pr(A), Pr(B) and Pr(B|A). But that just begs the question: are your assumptions justified by the data?

    Specifically, do the climate models and existing timeseries data support your assumption that the current dry period is three times more likely to be due to AGW (Pr(B|A) = 0.75), than otherwise (Pr(B|~A) = 0.25)?

    I think not.

  44. jquiggin
    October 3rd, 2006 at 22:41 | #44

    Well, Proust, I understood you had a “first-year stats” demonstration of error. Instead, all we get is blather. If you disagree with my priors, state and justify yours.

    Just as a hint, I’ll start with (Pr(B|~A) = 0.25). You can get this one straight from the historical data. Look at the most recent data (including 2006 so far), take whatever smoothing period you like to get the shortfall from the average, and then look at the frequency distribution. Or, you could try an ARMA (I assume that’s what you meant) analysis to get the same thing if you’re feeling energetic.

  45. econwit
    October 3rd, 2006 at 23:02 | #45

    “on my understanding of the climate science”

    No doubt the weather keeps changing and a lot of hot air is expired air discussing that fact.

  46. October 4th, 2006 at 00:52 | #46

    Tell that to the former citizens of Tuvalu and New Orleans, econowit.

  47. econwit
    October 4th, 2006 at 01:19 | #47

    Bob has the answer to predicting the weather.

    How many roads must a man walk down
    Before you call him a man?
    Yes, ‘n’ how many seas must a white dove soil
    Before she sleeps in the sand?
    Yes, ‘n’ how many times must the carbon dioxide fly
    Before they’re forever banned?
    The answer, my friend, is blowin’ in the wind,
    The answer is blowin’ in the wind.

    How many times must a man look up
    Before he can’t see the sky?
    Yes, ‘n’ how many ears must one man have
    Before he can hear people fry?
    Yes, ‘n’ how many droughts will it take till he knows
    That too many people have dried?
    The answer, my friend, is blowin’ in the wind,
    The answer is blowin’ in the wind.

    How many years can a mountain exist
    Before it’s rained to the sea?
    Yes, ‘n’ how many years can methane exist
    Before they’re allowed it to be green?
    Yes, ‘n’ how many times can a man turn his head,
    Pretending he just doesn’t see?
    The answer, my friend, is blowin’ in the wind,
    The answer is blowin’ in the wind.

  48. econwit
    October 4th, 2006 at 01:36 | #48

    OK wbb,

    I’ll tell them not to blame “the weather”

  49. taust
    October 4th, 2006 at 08:57 | #49

    re the observation that skeptics denialists keep on finding new explanations.

    My reading of the data linked to below is that the rate of increase in glogal temperature has markedly decreased for the last three years.

    /hadleycentre/obsdata/globaltemperature.html

    Am I interpreting the data correctly. If my intepretation is corrrect do any of the computer simulations give such a signal.

    What is the explanation for the signal?

    Before I get monstered (or put on JQ’s list) I am not denier nor a mitigator but for Australia an adaptationer.

  50. taust
    October 4th, 2006 at 09:23 | #50

    My apolgies let me have a second go of posting a link correctly

    http://www.met-office.gov.uk/research/hadleycentre/obsdata/globaltemperature.html

  51. Ian Castles
    October 4th, 2006 at 09:37 | #51

    Roger, I’m hoping that you will be able to answer my question about the implications for modelled temperature of the steep decline in global sulphur dioxide emissions in the 1990s which has been identified in several studies published since the last IPCC Assessment Report.

    The Summary for Policymakers of that report said that “The higher projected temperatures [than in the previous Assessment Report] are due primarily to the lower projected sulphur dioxide emissions in the SRES scenarios relative to the IS92 scenarios.”

    In your posting abve, you said that the highest temperature of the range I’d quoted for quoted for 2030 “is attached to the A1T scenario that has more aggressive sulphate emission reduction than the other SRES scenarios.” You also said that “the recent ABARE AP6 reference emission scenario gives an upper temperature almost as high in 2030 (0.05°C lower” and that “If more aggressive sulphate reductions were to occur, warming would be as high [as in A1T] by that time.”

    If you assume, in accordance with the results of the studies cited above, that “more aggressive sulphate reductions” in line with the A1T projections for 2030 had already occurred by the mid-1990s, what does that imply for the increase in temperature that would occur from now on?

    Specifically, if sulphur emissions as estimated in Stern D. I. (2005) “Global sulfur emissions from 1850 to 2000″, Chemosphere 58, 163-175 and the database supporting that paper are substituted for those that were used to produce the SRES and/or ABARE projections, what is the effect on the global mean temperature up to now, and the projected increase between now and 2030?

    Is the ABARE temperature projection that you cited publicly available? And where do I find the projections of sulphur emissions that are used in the ABARE temperature projection that you cited?

  52. Roger Jones
    October 4th, 2006 at 11:31 | #52

    Ian,

    re your last post. I can try this (substituting alternative SO2 pathways) using the MAGICC model, and it will be an interesting little test but it will have to wait until some of my contractual obligations are exhausted (that would be a fair while, if I was strict about putting such matters first all the time).

    Note also that the raditiave uncertainties surrounding sulphates are large, so any analysis (like all of those discussed on this thread) would be informative rather than conclusive.

    The ABARE temperatures in 2100 were quoted in their report (Matysek et al., 2006). I ran them through MAGICC (on the understanding that their origin would be acknowledged), so no, the full numbers are not yet publicly available (see the comment on contractual obligations above, it’s a time issue). The sulphur emissions in the scenario were assumed because the numbers were run really quickly for a short deadline. A sensitivity analysis using recent emissions estimates and some more aggressive reductions would also be interesting. Using “best bet” forcing estimates, early reductions in sulphates can accelerate warming by a few points of a degree.

    Re the plausibility of radiative forcing. I stand by my earlier comment. If high emission rates are plausible, then we should be factoring the impacts due to these into risk assessments. The methods we have developed and are using draw on the breadth of emission pathways from low to high. If work comes out that suggests that part of that range is less likely to occur, we can assess that, but any underlying priors in the risk assessment are subjective. This is why we apply uncertainty analysis and Bayesian inference to the outcomes to see how much different types of information affect the results.

  53. Ian Castles
    October 4th, 2006 at 16:56 | #53

    Thanks Roger. I realise that radiative uncertainties surrounding sulphates are large, and that’s why I found it surprising that the IPCC Summary for Policymakers in 2001 stated without qualification that the increase in the range of temperature increase to 2100 compared with that in the previous assessment in 1995 was “due PRIMARILY to the lower projected sulphur dioxide emissions in the SRES scenarios relative to the IS92 scenariosâ€?. (CAPITALS added). I questioned this statement in my initial correspondence on the IPCC emissions scenarios – it seems that analyses of the radiative effect of sulphate emissions are conclusive if they have been included in an IPCC SPM but are at best informative if carried out by other researchers.

    Taken literally, the IPCC statement meant that the projected reduction in SOx emissions accounted for the greater part of the increase of 2.3°C in the upper end of the IPCC range. If one-third of this projected reduction had already occurred before the IPCC Report was published, as the evidence now available suggests, that seems to me to be one of the many issues that the IPCC should have investigated before deciding that ‘the SRES scenarios provide a credible and sound set of projections, appropriate for use in the AR4.’

    I appreciate that you have contractual obligations that preclude you investigating this important matter for some time. However the IPCC’s decision to reuse the SRES projections in AR4 was made three years ago. It does seem rather extraordinary that the Panel decided that the scenarios were acceptable for this purpose, without reviewing the underlying assumptions in the light of all of the new evidence that had become available since the projections were prepared in the late-1990s.

    Of course I agree with you that “If high emission rates are plausible, then we should be factoring the impacts due to these into risk assessments.” But how does one determine whether the high emission rates are plausible? In your paper (co-authored with Wenju Cai) presented to the Pan Evaporation Workshop at the Academy of Science in Canberra in late 2004, which I attended, you used the SRES A2 scenario projections to reach the conclusion that “By 2100, the equivalent CO2 reaches a level that is more than three times the level of 1870 (concentration ppm).”

    But surely the projection of a global population of 15 billion in 2001 upon which the A2 scenario (and therefore your projection that the CO2 concentration may treble from 1870 levels) is predicated is no longer plausible, if it ever was? As long ago as 2001, the organisation that produced this projection for the IPCC (the International Institute of Applied Systems Analysis) published probabilistic estimates that put the 95% confidence limits for global population in 2100 at 4.3–14.3 billion. Now that another five years have passed without any upsurge in global fertility levels, the prospect of a 15 billion population by end-century must be vanishingly small.

    The base year of the population projections produced for the IPCC by IIASA in 1996 was 1995, and the global population aged 0-4 was known to have decreased between 1990 and 1995 at the time that the projections were prepared. The A2 scenario projected an increase of more than 20% in the world’s population aged 0-4 between 1995 and 2005 – in the event, according to UN and US Bureau of the Census estimates, there was a further DECREASE in the global pre-school age population during this period.

    Did you and your co-author realise that the scenario that you used to project a trebling of CO2 concentrations in the 1870 to 2100 period rested on assumptions about one of the key driving forces that were entirely unrealistic?

    You seem to be saying that it doesn’t matter if the assumptions about driving forces are unsound as long as the emission rates themselves are plausible. But how do you determine that the emission rates are plausible, other than by considering the plausibility of the assumptions upon which they are based?

    You say that “If work comes out that suggests that part of that range is less likely to occur, we can assess that.” But less likely to occur than what?

    The IEA has estimated that cumulative energy-sector investment of $17 trillion (in 2004 dollars) will be required by 2030 in order to finance its Reference Scenario, and has said that “Financing the required investments in non-OECD countries is one of the biggest sources of uncertainty surrounding our energy-supply projections� (World Energy Outlook, 2005, p. 79).

    The estimated increase in the global use of electricity between 2000 and 2030 under the IPCC’s B1 scenario is more than twice as great as under the IEA Reference Scenario, and the increase under the IPCC’s A1FI scenario is nearly three times as great.

    Do you feed in this type of information into your uncertainty analysis? If the IEA is uncertain about the prospects of the investments required by its Reference Scenario being financed in developing countries, is there any real likelihood that the funds and infrastructure will be forthcoming to support two or three times the investment in power supply and distribution that the Agency is predicting on the basis of present policies?

    I think that it is entirely fanciful to suppose that such a massive expansion in electricity generation and distribution capacity could take place, bearing in mind that the IEA’s much more modest estimates are supported by a mass of information supplied by governments and power supply authorities.

  54. jquiggin
    October 4th, 2006 at 17:20 | #54

    Ian, the object of modelling is to give reasonable projections of the variable of interest, so what matters is getting the important variables right. Climate change is a lagged result of cumulative emissions, so errors in forecasts of what will happen after 2050 are of much less importance than getting projections right for the next few decades – this is fortunate because of course we can’t know what will happen many decades into the future. For that reason, I think it’s a mistake to worry too much about the projected population in 2100 – what matters is the medium-term growth path. Here, the recent news is good, but not so good as to make our problems go away.

    On any plausible business as usual scenario, emissions will grow substantially, while for any plausible climate science model, we need to reduce emissions substantially if we are to avoid highly damaging climate change.

    Of course, the lower is the BAU projection, the lower the cost of stabilising concentrations of greenhouse gases and the better off we all are, so, like you, I’d welcome the use of more up-to-date population projections.

  55. Ian Castles
    October 4th, 2006 at 18:58 | #55

    John, Of course the object of modelling is to give reasonable projections of the variables of interest, and of course the most important forecasts are those for the decades immediately ahead. As I’ve pointed out, the IEA’s current Reference Scenario projects the same growth in global CO2 emissions to 2030 as the IPCC’s B1. In practice the growth in emissions is likely to be considerably less than this, because the IEA Reference Scenario does not take account of new policies that are under consideration in many countries. Some of these are climate policy-related, but most are linked to other objectives – energy security, urban air quality etc.

    Like all the IPCC scenarios, B1 does not assume that any measures are taken for climate policy reasons (e.g., Kyoto, or carbon sequestration). Yet, again as I’ve pointed out already on this blog, James Hansen and 45 other scientists from 12 research institutions have estimated that the warming to 2100 under the B1 scenario is 1.1°C. Cumulative emissions for the century under the IPCC’s B1T MESSAGE scenario are 20% lower than under B1. Do you think B1T is a plausible scenario? Because if it is, the cost of stabilising concentrations of greenhouse gases is nil. Concentrations under this scenario are stabilised at 540 ppm CO2 equivalent or thereabouts, and by assumption the implementation of expensive technologies for climate policy reasons is excluded.

    On the ABC Four Corners documentary What Price Climate Change? on 28 August, Dr. John Wright, Director of CSIRO’s Energy Transformed Flagship said that the task of avoiding highly damaging climate change is “absolute immense.” He asked rhetorically “Can the world do it?” and answered “Mot without the sort of technologies we are trying to develop here [at CSIRO].”

    Some of those technologies would not be contemplated other than for climate change reasons. Wouldn’t it be a good idea to try and get the IPCC “no policy” scenarios right first.

    I agree with you that 2100 doesn’t matter much, but these poor population projections lead us astray well before then. For population under 20 in the year 2050, the UN’s population projections are: Low, 1.6 billion; medium 2.4 billion; and High, 3.4 billion. The IPCC’s A2 estimates 4.1 billion – not a good guide for policy.

  56. Ian Castles
    October 5th, 2006 at 07:37 | #56

    I said above that the cost of stabilising GHGs under a particular “no climate policy” scenario IS nil. That was carelessly expressed. Of course I should have said that the cost WOULD be nil if the storyline, scenario and model quantification I was discussing were all to be realised.

    Although the B1 scenarios seem to me to be far more plausible than the A2 or A1FI scenarios, the pervasive influence of uncertainty in all aspects of climate change debate must be recognised.

    This was the theme of the policy paper “Uncertainty and Climate Change: the Challenge for Policy� which was published by the Academy of the Social Sciences in Australia in February 2005 (available at http://www.assa.edu.au/publications/op.asp?id=75 ). This paper incorporating articles by three leading Australian experts has been virtually ignored by the Australian Government and the media, including in the voluminous reading lists on climate change issues published on the websites of the Australian Greenhouse Office, CSIRO and the ABC.

    I should also mention the very sensible article by Professor David Pannell of the University of Western Australia on this subject, which is available on his website at http://cyllene.uwa.edu.au/~dpannell/pd/pd0069.htm .

  57. proust
    October 6th, 2006 at 04:31 | #57

    quiggin:

    …all we get is blather.

    Indeed, with yourself the greatest proponent of said art.

    Just as a hint, I’ll start with (Pr(B|~A) = 0.25)…

    And that proves what exactly? If you read my post, you’ll see I objected to your assumption for the ratio Pr(B|A) / Pr(B|~A), not your assumption for Pr(B|~A). Obviously it is the ratio that matters (although I disagree that the data gives 0.25 for Pr(B|~A) – a more justifiable upper bound would be 0.35-0.4, given the limited amount of data available)

    Once again, instead of answering my objections, you make a clumsy attempt to obfuscate.

  58. jquiggin
    October 6th, 2006 at 06:06 | #58

    OK, Proust, let’s go on a couple more steps. Climate projections indicate that the likelihood of more severe droughts, such as the one we’re observing, increases under global warming. I estimate the ratio at 3, based on my reading of the literature on climate change. You claim to be able to prove me wrong. Go ahead – state and justify a lower number.

    BTW, as noted above, I should have paid more attention to the role of increased evaporation in increasing the severity of drought, and reducing streamflow, the relevant variable for my analysis. So, please take this into account in formulating your response

    I also note an error in your analysis so far. You state “a more justifiable upper bound would be 0.35-0.4, given the limited amount of data available)” but there’s no justification for using an upper bound here. Upper bounds don’t play the kind of role in Bayesian analysis you seem to want.

  59. proust
    October 6th, 2006 at 09:39 | #59

    That link is broken.

    I meant lower bound for Pr(B|~A), hence upper bounding the ratio (my mistake). Unless you are planning an infinite hierarchical Bayesian regress (prior on prior on prior…), bounds on prior probabilities are very important in Bayesian analysis.

    Or, if you don’t like that approach, tell me your hyper-prior for Pr(B|~A) and justify it.

  60. jquiggin
    October 6th, 2006 at 16:03 | #60

    My objection is to your apparent desire to use a one-sided bound so as to get a probability substantially higher than the observed relative frequency. If you have a justification for this, state it.

    I’d suggest starting with a diffuse prior over the observed range for annual rainfall, and updating on the observed data. With 100+ observations you should get a posterior distribution pretty close to the historical one, and there’s no reason to suppose a bias of the kind you want to include is going to emerge.

  61. proust
    October 6th, 2006 at 17:41 | #61

    Quiggin:

    With 100+ observations…

    There are not 100+ independent observations of the random variable relevant to your claim that the 1995-2006 drying is caused by AGW (wp 0.75). There are approximately 10 such observations (going back to 1900).

    Another elementary error from the professor. But you’re on the right track now, keep trying.

  62. jquiggin
    October 6th, 2006 at 17:57 | #62

    Proust, I didn’t claim that the observations were independent, and your estimate assumes fixed 11-year cycles. As was stated way back in the thread, these are used just to even out year-to-year noise, the exact opposite of what you seem to be claiming here.

    In any case, it’s time for you to put up or shut up. The data set is there – present your own analysis or admit that you don’t have a case.

  63. proust
    October 6th, 2006 at 18:25 | #63

    Quiggin:

    Proust, I didn’t claim that the observations were independent

    If they’re not independent then your claim

    With 100+ observations you should get a posterior distribution pretty close to the historical one

    is nonsensical. Tossing a coin once and then observing it 1,000,000+ times doesn’t get you any closer to a sharp posterior on the probability of heads.

    and your estimate assumes fixed 11-year cycles.

    Well, if you recall, it was you who picked 11 years:

    B: Hot, dry weather as observed in the period 1995-2006 occurs

    You can put the 11 year cycles where you like. You can even make 100+ overlapping ones if you want. But there’s still only approximately 10 independent observations in there.

    You seem remarkably out of your depth for someone in your position.

    In any case, it’s time for you to put up or shut up. The data set is there – present your own analysis or admit that you don’t have a case.

    So far you have justified your estimate of Pr(B|~A) with an incorrect analysis of the probability of 11-year dry periods, and offered me a broken link as support for your estimate of Pr(B|A).

    Sorry, you haven’t yet got the ball back over the net. At least fix the link.

  64. jquiggin
    October 6th, 2006 at 20:47 | #64

    Proust, this is nonsense. Check the first-order autocorrelation in the series. You need a value above 0.9 for your claims to stand up.

    There is some short-run autocorrelation due to El Nino cycles but nothing like what you need for this claim.

    The 11-year moving average was first mentioned by Ian Castles, not me, and, as I said, it has the exact opposite purpose to the one you claim. It’s meant to smooth out short run noise so long-run movements, if any, can be detected. The claim you’re supposed to be defending is that there are no long-run movements to detect, since all we have is random autocorrelated variation around a stable mean. (At least, I think that’s what you’re supposed to be defending – you haven’t stated a position).

    Here’s the link

  65. proust
    October 9th, 2006 at 13:39 | #65

    Finally got time to load the rainfall data for SE Australia into a spreadsheet.

    Average rainfall 1900-2005: 601mm. Std dev: 103mm
    Average rainfall 1995-2005: 585mm
    6/11 years in the period 1995-2005 had rainfall above the long-term average
    5/11 had rainfall below the long term average.

    On that measure, the 11 years from 1995-2205 were about as unremarkable as it gets.

    So, your claim that in the absence of any significant AGW effects,

    B: Hot, dry weather as observed in the period 1995-2006 occurs

    with a probability of only 0.25 does not seem to be supported by the data at all.

  66. October 9th, 2006 at 15:18 | #66

    Greens environmental consultant Aron Gingis says AGW has nothing to do with reduced rainfall:

    “The major cause of the substantial reduction in rainfall for Sydney is air pollution coming from as far away as Melbourne’s Latrobe Valley power stations,” said Mr Gingis, adding that global warming was not to blame.

    According to Gingis, particles cause clouds to be “constipated.”

  67. tam o’shanter
    October 9th, 2006 at 16:37 | #67

    Various points:

    JQ on Ian Castles said: “I don’t think it’s appropriate to try and correlate rainfall in SE Australia with temperature in the same area.” Why not? global implies warming everywhere, so Australian temperature rises are just as relevant as – and are proxies for – rises in the oceans and elsewhere (which are indeed correlated with each other). Your original claim came close as may be to asserting the correlation you now qualify.

    JQ again: “It’s my understanding (not expert) that sea-surface temperatures in the Pacific are more important for rainfall, and the Indian Ocean may also be relevant.” Yet it is clear from BoM that sea surface temperature trends around Australia broadly match the land trends, and elsewhere you have asserted that evaporation is part of the “browning” problem. I was taught that evaporation is part and parcel of the rainfall system. The “science” on this thread looks more and more dubious and the statistics even more again. Using the BoM 1960-1990 base period and deriving rainfall anomalies to match the mean temperature anomalies, there is always a significant (good ts) positive correlation between variations in both anomalies, even for 1970-2005. The browning map for 1970 to the present that began this thread is entirely an artifact of the very wet years at the beginning of that period, as Ian Castles has noted. Please note that the BoM maps for temperature and rainfall are not strictly comparable, as the former use anomalies and the latter use absolute values.

    “On the final point, if we return to your original point, my article made the claim that it was appropriate to assume permanently lower levels, and proceeded to consider the implications. It appears you don’t dispute this.”

    Ian can and has answered for himself, from which I cannot see that he has agreed it is “appropriate to assume permanently lower levels (of rainfall)”; the BoM shows no such thing.

  68. tam o’shanter
    October 9th, 2006 at 17:06 | #68

    Perhaps I should spell it out more emphatically, this time regressing the 11-year moving averages of rainfall on temperature, which show that more heat implies more rain, with R2 at .24, the X coefficient at 2.85, standard error 0.547, and t at 5.2. The science underlying this is more heat to more evaporation to more rain, pace JQ. BTW, burning hydrocarbons yields both CO2 (more Heat) and H2O or water vapour (i.e more Rain). Curiously, the IPCC stress the physics and ignore such chemistry!

  69. jquiggin
    October 9th, 2006 at 19:40 | #69

    Proust, you can’t exclude 2006 which has been very dry. The rainfall deficiency so far is between 200 and 400 mm for most of SE Australia, and is unlikely to improve. The mean rainfall for the 11 years ending in 2006 is going to be well below the 585 you give for 1995-2005 and will be in the bottom quartile of the moving average values for the entire period, supporting the 25 per cent estimate I gave.

    More importantly, as I’ve said several times now, in assessing the claim that the climate is hotter and drier you also have to take account of evaporation, which is now exacerbating droughts to the point that inflows to the Murray-Darling for 2006 are the lowest on record. (Post on this coming soon).

  70. jquiggin
    October 9th, 2006 at 19:42 | #70

    TOS, you should submit your interesting ideas to a scientific journal. If you want a quicker response, why don’t you give them a run over at RealClimate – for a softer ride, you might try Climate Audit or Jennifer Marohasy’s blog.

  71. tam o’shanter
    October 9th, 2006 at 21:16 | #71

    JQ

    Thanks – your comment applies a fortiori to your new thread, I hope you will take your own advice. What goes up comes down so far as evaporation is concerned. If not where does it go? It’s certainly interesting when a professor considers statistically significant data showing a positive correlation between heating and rainfall is evidence for his faith in the opposite. Using Bayes, I will accept your odds for a bet on higher rainfall in 2007 than this year. As you would say, put up – or…

  72. proust
    October 9th, 2006 at 22:53 | #72

    Proust, you can’t exclude 2006 which has been very dry. The rainfall deficiency so far is between 200 and 400 mm for most of SE Australia, and is unlikely to improve. The mean rainfall for the 11 years ending in 2006 is going to be well below the 585 you give for 1995-2005 and will be in the bottom quartile of the moving average values for the entire period, supporting the 25 per cent estimate I gave.

    Professor Q, I’ll let you work out what is wrong with this reasoning, since you are, after all, a professor. But here is a hint: if the addition of one data point from a year that is not yet complete (2006), changes the conclusion from “unremarkable rainfall” to “remarkably dry”, what does that tell you about your approach?

    [BTW, being in the bottom quartile is not the same as saying probability 0.25 - I'll let you work that one out as well]

    More importantly, as I’ve said several times now, in assessing the claim that the climate is hotter and drier you also have to take account of evaporation

    The map at the top of this post is a rainfall map. It is your blog, you can keep moving the goalposts if you want, but it doesn’t alter the fact that your original post was typical enviro-alarmism.

  73. Roger Jones
    October 10th, 2006 at 08:49 | #73

    TOS,

    your reasoning would hold if what goes up comes down again in the same vicinity. Pity about atmospheric circulation. Global warming is increasing global average rainfall, but recent research by the UK Hadley Centre suggests that rainfall is becoming more spatially variable on a global basis. Harsher droughts, more deluges.

    Furthermore, time series of annual average temperature and rainfall anomalies in temperate Australia are anti-correlated. This is because it rains when cloudy.

  74. tam o’shanter
    October 10th, 2006 at 11:52 | #74

    Roger Jones: of course I know about atmospheric circulation, but perhaps you will have more luck than Ian Castles and myself in getting JQ to agree that as you say “global warming IS (my emphasis) increasing global average rainfall”. So we have a distribution problem, indicating more dams in flood areas for transfer to drought areas (I previously worked on this in various parts of Africa until the World Bank joined the greens in banning dams). I am glad to see JQ’s new thread is a step in that direction.

    However you are wrong about the time series: the actual coefficient on rainfall anomaly from 1960-1990 as function of mean temp anomaly is POSITIVE and statistically significant (t = 2.8) even for 1970-2005.

  75. Roger Jones
    October 10th, 2006 at 12:48 | #75

    Dear TOS,

    I am prepared to be wrong, but taking the 1950-2005 time series of annual temperature and rainfall from the area of SE Australia in the grid box bounded by 140.5°E to 154.5°E and 27.5°S to 39.5°S, rainfall is negatively correlated with temperature to the value of -0.46. This is comparable to the value one gets for individual sites in the region.

  76. tam o’shanter
    October 10th, 2006 at 16:52 | #76

    Dear Roger

    Thanks – but your comment about SE Australia conflicts with your previous comment on atmospheric circulation, as such a small area is unlikely to have a meaningful relationship between heat and rain however statistically significant your negative correlation may be (what is the t?).

    Regressing your rainfall in SE Australia on mean temperature for the whole country from 1950 to 2005 produces once again a POSITIVE correlation with R2 = .27, X = 0.6, t = 4.5, so statistically significant and in this case climatically meaningful.

    As for that part of the Murray catchment in SE Australia, with average annual rainfall of just 328mm in 1940-1947, against 579 in 1998-2005, I doubt the inflows then were better than over the 8 years before the present, pace JQ.

  77. jquiggin
    October 10th, 2006 at 17:09 | #77

    “perhaps you will have more luck than Ian Castles and myself in getting JQ to agree that as you say “global warming IS (my emphasis) increasing global average rainfallâ€?.”

    If you have ever raised this issue before, TOS, I don’t remember it, and a quick scan of your comments didn’t find it. I have no reason to doubt that global warming is increasing mean global rainfall, and I’ve certainly never said anything different.

    Proust, the arithmetic properties of a moving average are what they are. I didn’t pick 11 years, and it’s bit late for you to say now that it’s too sensitive to one very low year. As for your claims about quartiles, the probability that a single observation picked on an objective basis (in this case, the most recent observation on the moving average) will fall into the bottom quartile of a distribution is exactly 0.25. It’s certainly not 0.35, 0.4 or any other number you want to dream up. Of course if you cherry-pick a particular time period like TOS in the comment immediately above, you can get whatever you like – I assume you had something like this in mind, but it’s inapplicable here.

    Coming back to the main point, if you want to continue the dispute over the conclusion that the weather is, as I said at the start of all this, getting hotter and drier [as we have discovered in the discussion, as a result of higher evaporation as well as lower rainfall], I suggest you move the discussion to the thread on “Drying out”.

  78. Roger Jones
    October 10th, 2006 at 17:28 | #78

    Dear TOS,

    I did the correlation but certainly did not assume that temperature drives the relationship. The negative correlation at location is, as I implied, due to local effects only. You left out your domain of analysis in previous posts, hence my quick check.

    Nor am I convinced that your correlation is climatically meaningful. The following paper suggests the correlation of rainfall and max temp over Australia is strongly negative.
    Nicholls, N. ; Lavery, B. ; Frederiksen, C. ; Drosdowsky, W. ; Torok, S. 1996 Recent apparent changes in relationships between the El Niño-Southern Oscillation and Australian rainfall and temperature
    Geophys. Res. Lett. Vol. 23 , No. 23 , p. 3357 (96GL03166)

    I have the Upper Murray inflows to 2000-1. Yes, the forties were low but I am reliably informed by the custodians of the data, that the latest accumulated natural inflows (since 1997) are the lowest on record.

  79. Brian Bahnisch
    October 10th, 2006 at 22:16 | #79

    Roger said:

    Global warming is increasing global average rainfall, but recent research by the UK Hadley Centre suggests that rainfall is becoming more spatially variable on a global basis. Harsher droughts, more deluges.

    That is as I understand it. I heard a climatologist explain on the radio that with a higher temperature the saturation point rises. Hence the atmosphere can hold more water. But when something disturbs it there is more up there to fall down.

    I also recall a climatologist saying that we have been experiencing a predominance of large high pressure systems with the lows tracking further south in recent years. This was said to be related to tighter circulation patterns over Antarctica, I think to do with the ozone hole.

    If this is so, it seems to my simple mind that there may be an upside in that the melting of Antarctica may be delayed, apart from the bit that sticks out. But then the sea is majorly in contact with the ice, especially under the large ice shelves, so I expect it will be chewed out eventually.

    And I wonder what impact the 30% reduction in the thermohaline circulation in the North Atlantic in the last 12 years is having on our weather here.

    But I’m wandering…

  80. proust
    October 10th, 2006 at 23:20 | #80

    Quiggin:

    Proust, the arithmetic properties of a moving average are what they are. I didn’t pick 11 years, and it’s bit late for you to say now that it’s too sensitive to one very low year.

    It is a matter of record that you picked 11 years, this is your statement:

    B: Hot, dry weather as observed in the period 1995-2006 occurs [with probability 0.25]

    On this:

    As for your claims about quartiles, the probability that a single observation picked on an objective basis (in this case, the most recent observation on the moving average) will fall into the bottom quartile of a distribution is exactly 0.25.

    Of course.

    Of course if you cherry-pick a particular time period like TOS in the comment immediately above, you can get whatever you like – I assume you had something like this in mind, but it’s inapplicable here.

    At last!! I could not agree more. Cherry-picking is precisely what you are doing, as my analysis of 1995-2005 shows.

    If, as you claim, 2006 will change the 11-year moving average from “unremarkable” to “astonishingly dry”, that tells you 2006 is an extreme outlier. Therefore, ending the timeseries in 2006 and not 2005 is cherry-picking your period (it just happens to be the most recent period, but 1995-2005 is also recent enough for the purpose of this disscussion, as is 1994-2004, 1993-2003, etc – all “unremarkable”).

    It would be one thing if it had been overly dry for the last several years, for then your argument would be relatively insensitive to the chosen period. But it’s not; your argument depends critically on ending in 2006, not 2005. Which means the support for your argument is essentially one data point: 2006.

  81. jquiggin
    October 12th, 2006 at 06:55 | #81

    Proust, making up quotes, as you’ve done above, is usually an indication of desperation.
    Including 2006 shifts the 11-year moving average for rainfall from below-average to well below average.

    More importantly, the discussion has demonstrated quite clearly that hotter, drier weather, including higher evaporation, is leading to flows into catchments that are well below the historical average and, for the Murray-Darling at all-time (historical) lows. This was the original claim disputed by you and Ian. I’ve restated it in “Drying out”, so if you have anything further to say, please do so there.

  82. proust
    October 12th, 2006 at 14:21 | #82

    Which quotes? Everything in blockquote tags is a direct quote of yours. The other quotes (“unremarkable”, “astonishingly dry”) are what is known as “scare quotes”. Look it up. As a general rule I don’t like them – but they slipped through here.

    Including 2006 shifts the 11-year moving average for rainfall from below-average to well below average.

    Yes, strictly speaking 585mm is below-average. But when the average is 600mm +- 100mm, 585mm is not significantly below average (in the statistical sense). So “unremarkable” is a more appropriate description.

    You cherry-picked. Or just applied sloppy statistics, I don’t know which. Then when caught out, you simply shifted the goalposts. But I wouldn’t worry about it, you’re in good company – your approach seems to be the MO amongst many influential climate scientists.

Comments are closed.