Some Quick Thoughts on CSIRO Drought Info

First of all, the most important issue in this study is acknowledging Hennessy et al 2008. I had to agree to acknowledge them about 10 separate times to download data and so I do so. Acknowledging Hennessy et al 2008 seems to be more important to the authors than the results themselves. I hereby acknowledge:

Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp,

Their data archive is a total pig. Their archive is all in little micro series, which are tarred and so getting the data into a form that you can use is highly annoying and time consuming. Some of it are Excel files within tar files, making the extraction even more time consuming as each Excel file has to be saved into a csv file in order to be read into R or Matlab for statistical analysis. I’ve managed to organize the data into a few usable R-objects so that instead of downloading multiple tar files, manually unzipping each one. I hereby acknowledge:

Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp,

If you don’t want to waste endless amounts of time wading through the goofy tar files, I’ll save you the effort of repeating all the aggravating hoops that I had to go through (I hereby acknowledge Hennessy et al 2008) and use the following commands (watch the quotation signs form WordPress):

download.file(“http://data.climateaudit.org/data/csiro/csiro.cy.tab”,”temp.dat”,mode=”wb”); load(“temp.dat”)
download.file(“http://data.climateaudit.org/data/csiro/rainf.tab”,”temp.dat”,mode=”wb”); load(“temp.dat”)
download.file(“http://data.climateaudit.org/data/csiro/tempf.tab”,”temp.dat”,mode=”wb”); load(“temp.dat”)

The first yields an R-object “csiro” which is a list of 8 objects with names

# rain.5pc rain.95pc tmax.5pc tmax.95pc tmean.5pc tmean.95pc tmin.5pc tmin.95pc

each of which collates the regional and total time series (see names in column heads). The second and third collate the rainfall and temperature forecasts into objects “rainf” and “tempf”, also lists of 7 objects this time by region with the columns being the different model forecasts. The collation script is shown in a comment.

I took a look at the results for under 5 percentile area rainfall for two regions 4- Queensland; and 1- Murray-Darling, picked at random.

According to my calculations, the average intermodel correlation of the results for Mur-Darling was 0.009 (Qld: 0.027), while the average correlation of the model results to observations for Mur-Darl was -0.013 (Qld: -0.017). [David Stockwell observes that the timing of droughts would be stochastic for a given model, which is fair enough. However, as I observe below, looking at the Qld vs GISS, aside from distributions, the GISS model is showing a 20th century increase in drought while the data is showing a decrease. So if the finding promoted to the public is trend, there seems to be a mismatch for this region at a pretty basic level for this model.)

Shown below are two plots the first one comparing Murray-Darling historical to CSIRO Mk 3.5 and the second Queensland historical to GISS. Oh, yes, I hereby acknowledge:

Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp,

I’ve attached my scripts both for collation and for the calcs, as I’ve done this quickly; I’m not experienced with these data sets; they are poorly organized for statistical analysis and I might have collated apples and oranges along the way, in which case I’ll amend the calcs.

new_pa77.gif

new_pa74.gif


51 Comments

  1. Steve McIntyre
    Posted Jul 25, 2008 at 9:32 AM | Permalink | Reply

    I hereby acknowledge: Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

    The historical rainfall and temperature data are organised into seven sets of uncompressed TAR archives of ASCII text files, one file for each time series. Study regions are denoted in the file names by “aust” for Australia, “murrdarl” for the Murray-Darling Basin, “nsw” for New South Wales, “nwquad” for the Northwest Region (NT + northern WA), “qld” for Queensland, “swaust” for Southwest Western Australia, “swquad” for the Southwest Region (SA + southern WA) and “victas” for Victoria&Tasmania. The periods are denoted by “cy” for calendar year (January to December) values, “fy” for financial year (July to June) values, and “2y” for biennial (January to December) values.
    Here is the collation script requires manual downloading and unzipping. tmean and tmax 5 pct not provided at webpage for some reason/

    url=”d:/climate/data/csiro”
    id=list.files(url)
    region=c(“murrdarl”,”nsw”,”nqquad”,”qld”,”swaust”,”swquad”,”victas”)
    id0=id[grep("cy",id)]
    #48 series rain,tmax,tmean,tmin by region
    K=length(id0)
    item=c(“rain”,”tmax”,”tmean”,”tmin”)
    quant0=c(“.5pc”,”.95pc”)
    g=function(x,y) paste(x,y,sep=””)
    id3=c(t(outer(item,quant0,g)))

    f=function(url){
    test=read.table(loc,skip=1)
    f=ts(test[,2],start=1900)
    f}

    csiro=rep(list(NA),8)
    for (i in 1:4) {
    temp1=!is.na(match(1:48, grep(item[i],id0) ))
    for(j in 1:2) {
    temp2= !is.na(match(1:48, j+seq(0,46,2) ))

    id1=id0[temp1&temp2]
    K=length(id1)
    test=NULL
    for (k in 1:K) {
    loc=file.path(url,id1[k])
    test =ts.union(test,f(loc))}
    dimnames(test)[[2]]=id1
    csiro[[2*(i-1)+j]]=test
    }
    }
    names(csiro)=id3
    csiro.cy=csiro
    save(csiro.cy,file=”d:/climate/data/csiro/csiro.cy.tab”)
    #8 items: 4 items by 2 percentiles

    ##FORECASES
    #Time series of percentage area below the 5th percentile
    #Time series of percentage area above the 95th percentile

    #Percent_area_below_5thpercentile_Rainfall_Qld Percent_area_above_95thpercentile_Temperature

    region1=c(“MDB”,”NSW”,”NWAust”,”Qld”,”SWAust”,”SW-WA”,”VicTas”)

    rainf=rep(list(NA),length(region1))
    for(i in 1:length(region1)){
    loc=file.path(url,paste(“Percent_area_below_5thpercentile_Rainfall_”,region1[i],”.dat”,sep=””))
    fred=readLines(loc);N=length(fred)
    fred=gsub(“\t\t\t\t”,””,fred)
    temp=c(TRUE,!is.na(as.numeric(substr(fred[2:N],1,4))))
    fred=fred[temp]
    writeLines(fred,”temp.dat”)
    test=read.csv(“temp.dat”,sep=”\t”,header=TRUE,colClasses=”numeric”)
    rainf[[i]]=test
    }
    names(rainf)=region
    save(rainf,file=”d:/climate/data/csiro/rainf.tab”)

    tempf=rep(list(NA),length(region1))
    for(i in 1:length(region1)){
    loc=file.path(url,paste(“Percent_area_above_95thpercentile_Temperature_”,region1[i],”.dat”,sep=””))
    fred=readLines(loc);N=length(fred)
    fred=gsub(“\t\t\t\t”,””,fred)
    temp=c(TRUE,!is.na(as.numeric(substr(fred[2:N],1,4))))
    fred=fred[temp]
    writeLines(fred,”temp.dat”)
    test=read.csv(“temp.dat”,sep=”\t”,header=TRUE,colClasses=”numeric”)
    tempf[[i]]=test
    }
    names(tempf)=region
    save(tempf,file=”d:/climate/data/csiro/tempf.tab”)

  2. Steve McIntyre
    Posted Jul 25, 2008 at 9:33 AM | Permalink | Reply

    Here is script to produce figures.
    use0=”pairwise.complete.obs”
    k=4 #k=1 #4- Qld; 1- MDB
    hist= csiro[[ "rain.5pc"]][,k+1]
    fore=rainf[[k]]
    fore=ts(fore[,2:ncol(fore)],start=fore[1,1]) # 1900 2040 1

    ##MEAN CORRELATION
    corry=cor(fore); corry[corry==1]=NA;
    mean(corry,na.rm=TRUE) # Qld- 0.02695259; MD – 0.009061272

    test=ts.union(hist,fore)
    corry=cor(test,use=use0); corry[corry==1]=NA;
    mean(corry[,1],na.rm=TRUE) # Qld- -0.01665452 MD – -0.01369344

    test=ts.union(hist,fore[,"giss_aom"])
    dimnames(test)[[2]]=c(“Reported”,”GISS”)
    plot.ts(test,main=paste(region1[k],”: Area LT 5 Percentile Rainfall”) )

    test=ts.union(hist,fore[,"csiro_mk3.5"])
    dimnames(test)[[2]]=c(“Reported”,”CSIRO Mk3.5″)
    plot.ts(test,main=paste(region1[k],”: Area LT 5 Percentile Rainfall”) )

  3. Sam Urbinto
    Posted Jul 25, 2008 at 9:42 AM | Permalink | Reply

    Hey, you forgot to acknowledge them in post 2.

  4. Patrick M.
    Posted Jul 25, 2008 at 11:03 AM | Permalink | Reply

    (re 3 Sam Urbinto):

    In acknowledging that Steve forgot to acknowledge:

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

    You yourself, Sam, forgot to acknowledge:

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

    which I shall now acknowledge:

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

  5. M Python
    Posted Jul 25, 2008 at 11:27 AM | Permalink | Reply

    This is all getting very silly. (I shall now acknowledge: Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .)

  6. Jim Arndt
    Posted Jul 25, 2008 at 12:24 PM | Permalink | Reply

    Steve this might be of some use. From Pielke’s site on Colorado droughts.

    http://climatesci.org/2008/07/25/the-value-of-paleoclimate-records-in-assessing-vulnerability-to-drought-a-new-paper-meko-et-al-2008/

  7. dicentra
    Posted Jul 25, 2008 at 12:48 PM | Permalink | Reply

    So… where’s this data come from again? You could at least include an acknowledgment where people can find it. To do otherwise is supremely unprofessional.

  8. Steve McIntyre
    Posted Jul 25, 2008 at 12:52 PM | Permalink | Reply

    #7. Sorry, that was an oversight on my part. For reference, the data comes from:

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

  9. Tolz
    Posted Jul 25, 2008 at 1:21 PM | Permalink | Reply

    #6 (Jim),
    Don’t you think you should have acknowleged that the inspiration for your comment came from:

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/climate/droughtec/ .

    ? I certainly do. I acknowledge that the above acknowledgment applies to my comment, which I hereby duly acklowledge as well.

  10. Jim Arndt
    Posted Jul 25, 2008 at 1:29 PM | Permalink | Reply

    #9,

    Fail to see your point. I was merely adding information to the discussion and not taking it one way or the other, hence the link. Just thought it might add something.

  11. Jim Arndt
    Posted Jul 25, 2008 at 2:08 PM | Permalink | Reply

    #9

    Ok get it. Sarcasm.

  12. conard
    Posted Jul 25, 2008 at 2:15 PM | Permalink | Reply

    I just downloaded all of the files without any problem or agreements. The page lists the citation information and contains a link to a site where the underlying data can be purchased.

    Steve: If you’ve downloaded the files, then you’ll surely agree that they they are not organized for modern statistical users. Otherwise I’d be able to specify the urls, and read the data directly into R. I had to waste a lot of time on mechanics. YEah, you can wade through it, but anyone who’s interested in using the data would be better off to use my collation. As to the purchasing of the underlying, that’s another story in itself. Why should David Stockwell have to purchase the data used in this study?

  13. Tolz
    Posted Jul 25, 2008 at 2:41 PM | Permalink | Reply

    #11. Acknowledged.

  14. TonyA
    Posted Jul 25, 2008 at 3:00 PM | Permalink | Reply

    Hennessy et al. (2008) Peace Be Upon Them

  15. Posted Jul 25, 2008 at 3:27 PM | Permalink | Reply

    Funny. Thanks Steve. I’ve just been doing a bit of reading. I think this is an job for an extreme value statistic, excess over value Pareto distribution (aka power law). Our friend Koutsoyannis has been there, done all that. See http://sama.ipsl.jussieu.fr/Documents/articles/2006_11_17_Rust.pdf. Definitely not a difference of means test.

  16. Steve McIntyre
    Posted Jul 25, 2008 at 3:40 PM | Permalink | Reply

    #15. Bur David, before we even talk about statistics, it’s interesting to look at intermodel comparisons and how the models compare to observations. Even for Michael Mann, a correlation of -0.013 between model and observation wouldn’t be enough. For verification, he’d probably require at least 0.0005.

  17. Posted Jul 25, 2008 at 3:45 PM | Permalink | Reply

    Steve, is that necessary? I mean, it only needs to match the statistical properties, identical distributions, not be predictive.

  18. Steve McIntyre
    Posted Jul 25, 2008 at 4:02 PM | Permalink | Reply

    #17. The GISS model and the Queensland actual look different to me on any level. Whether the properties of the model change is an interesting analysis in its own right. But let’s also look at the properties of the observed and not just the before and after model versions.

  19. Posted Jul 25, 2008 at 4:32 PM | Permalink | Reply

    Steve, There has to be some connection of the model past with observations – granted. It just wouldn’t have to match drought-for-drought, like hurricane-for-hurricane, and so high r2 values are not a necessary condition. But then what else is there? You could just generate numbers with the same distribution otherwise. I like to work out exactly what I am going to test for.

  20. Louis Hissink
    Posted Jul 25, 2008 at 5:19 PM | Permalink | Reply

    Shouldn’t the Oz droughts show some negative correlation with the position of Aquarius as observed from the Hubble telescope? Surely CISRO must have some data sharing arrangements. In any case 1/r^2 would be a better looking number – you could form the conclusion that drought frequency correlates according to the inverse square law of the position of Aquarius wrt the horizon.

    Ahem.

  21. Posted Jul 25, 2008 at 5:42 PM | Permalink | Reply

    Louis, I guess what you both are saying is that because the models have no skill at predicting regional climate the whole DECR thing was a waste of time.

  22. ianl
    Posted Jul 25, 2008 at 6:25 PM | Permalink | Reply

    Steve McIntyre/David Stockwell:

    that you got this far is amazing. I’m genuinely impressed.

    David Stockwell:

    “because the models have no skill at predicting regional climate the whole DECR thing was a waste of time”

    from whose viewpoint ?

  23. Posted Jul 25, 2008 at 6:41 PM | Permalink | Reply

    Well from the viewpoint of the results meaning anything. What I hear the claim is, that if the models do not correlate with regional observations then doing any regional work with them is blocked.

  24. Geoff Sherrington
    Posted Jul 25, 2008 at 6:55 PM | Permalink | Reply

    The generalist again, with more Q than A.

    The graphs from now to 2040 have a different “texture” to earlier years. Now, how does that arise? Is an explanation given in

    Hennessy et al. (2008): “An assessment of the impact of climate change on the nature and frequency of exceptional climatic events”. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp

    apart from noting that future models were used? I would imagine a good starting point to help gain confidence in a model for the future would take the patterns of Mother Nature into account. She tends not to recognise models that change abruptly at the date of a paper.

    But politicians have been known to.

  25. Posted Jul 25, 2008 at 7:07 PM | Permalink | Reply

    Geoff. The process is define exactly what needs to be tested, then test it, then report it. Otherwise, they come back with all sorts of rebuttals. If absence is r2 correlation is enough, then good, but is it? I can understand r2 correlation being needed to test medieval warm period height for example, but is r2 correlation necessary for testing drought frequency, or what is more standard return frequency?

  26. Posted Jul 25, 2008 at 7:11 PM | Permalink | Reply

    return period I mean.

  27. Steve McIntyre
    Posted Jul 25, 2008 at 7:16 PM | Permalink | Reply

    Dave, I see your point. But looking at the Qld vs GISS, aside from distributions, the GISS model is showing a 20th century increase in drought while the data is showing a decrease. So if the finding promoted to the public is trend, there seems to be a mismatch for this region at a pretty basic level for this model.

  28. Vincent Guerrini Jr
    Posted Jul 25, 2008 at 7:46 PM | Permalink | Reply

    Bottom line rainfall data = no change whatsoever?

  29. Posted Jul 26, 2008 at 12:30 AM | Permalink | Reply

    Steve, I think I know what to do as a first step — compare past models to observation across a range of statistics to see which if any the models have skill with.

  30. Geoff Sherrington
    Posted Jul 26, 2008 at 12:46 AM | Permalink | Reply

    Re # 25 David Stockewell

    In brief, I am suggesting that the information is more amenable to philosophic analysis than mathematical analysis. People might provide exceptions to this, but in general Nature does not suddenly change beat in the graphical way depicted. The change in “texture” that I reference has meaning only if you invoke a new percussion in the orchestra and that would seem to be AGW. So, unless you accept AGW starting next year or so, there’s not much point in paralysis by analysis.

  31. MarkR
    Posted Jul 26, 2008 at 1:36 AM | Permalink | Reply

    David and Geoff. I’m sure you already know this:

    At first, Lorenz thought the computer, a Royal McBee, was malfunctioning. Then he realised that he had not entered the initial conditions exactly. The computer stored numbers to an accuracy of six decimal places while, to save space, the initial conditions for the second run were shortened to three decimal places. Even this small discrepancy, of less than 0.1%, changed the end result completely.

    Lorenz

    Also, http://ams.allenpress.com/archive/1520-0469/20/2/pdf/i1520-0469-20-2-130.pdf Page 141.

    Accurate climate models are impossible.

    Steve: Different issue and arguable on other grounds. Nothing to do with this thread.

  32. Louis Hissink
    Posted Jul 26, 2008 at 4:11 AM | Permalink | Reply

    #21 David,

    Yes, precisely – especially when the underlying physical processes producing weather are either poorly understood, ignored, or misunderstood.

    #31

    MarkR – good point and this fact points to AGW satisfying one of Langmuires criteria for pathological science – that changes in a small, almost undetectable, factor generate signficant outcomes – sort of the behaviour of a mathematical singularity, I suppose.

    One reason why I support your conclusion is from experience in the diamond mining business, in particular alluvial diamond mining from an ore-reserve estimation perspective. The rule of thumb in this industry is that you can’t estimate what the in-situ diamond ore-grades are.

    The diamonds, as heavy minerals, are deposited into trapsites in either rivers or beach wave cut platforms by the action of highly turbulent water flow. Once a flowing body of water passes the laminar flow conditions, theory goes very pear-shaped. (Same in exploration geophysics – once the system becomes non-linear, all bets are off).

    I used to work for De Beers and they commissioned the French Geostats gurus to try and see if geostats (Matheron’s theory, after Krige of South Africa) could improve ore grade estimates of the Namib beach mines – they could not.

    Now it is quite possible that we have omitted some factor in the framing of the diamond deposition model, but right now, using existing physical theories, no one bothers trying to estimate the grade of an alluvial diamond mine, because you can’t, at least not accurately to within, say +/- 10%. Well they do in a rudimentary arithmetical way but it’s essentially meaningless. (In Geostats the term is extreme nugget effect).

    Weather is basically the behaviour of a turbulent three phase system – gas, liquid and aerosol – and if we can’t model using a two phase system in diamond exploration, then a three phase system would be even more intractable. (and I am not even including the input from the geophysical electrical system into weather!)

    All these conclusions are from practical experience of course, but if your reasoning is essentially rhetorical, restricted to computer modelling, then anything becomes possible.

    So your conclusion is spot on.

  33. DaleC
    Posted Jul 26, 2008 at 5:10 AM | Permalink | Reply

    #28, Vincent,

    The actual charts of rainfall are interesting. I downloaded the annual averages from the Australian BOM website, and repackaged as a two-panel display, 11 year trailing moving average so that the most recent data can be included. At 50 year resolution, there is a distinct change between 1990-1949 and post 1950. All regions except SouthWest, and all states, have had better rainfall since 1950 than before (see the Averages tables in bottom right corners). Make of that what you will.

    Observe that Tasmania is twice the average. However, much of that falls on the sparsely populated and rugged west coast. Most farming is on the eastern side, so the persistent drop since 1990 is hurting.

    The SouthWest region is in slow decline. Naively, much of the weather for the SouthWest region, Victoria and Tasmania comes sweeping up in arcs from the oceans south of Perth, with the south eastern and south western coasts and immediate inland catching the top fringe of the fronts as they head east. If the arcs move a bit further south, rainfall can decline. That looks like the pattern before 1950. If southern Aus is returning to the pattern of 1900-1950, the matter could get very serious, since infrastructure is mostly predicated on the good times of the 1960s and 1980s. Nearly all of Tasmania’s electricity is hyro, for example.

    The black series is the BOM average. It does not match my average (in red), so I do not understand how it was calculated.

    Be aware that the regions overlap, so my regional average and state average are not the same. The piecewise trend line on the red average for the states (bottom panel) shows the regime shift at 1950.

  34. Geoff Sherrington
    Posted Jul 26, 2008 at 5:15 AM | Permalink | Reply

    Heck, even Lorenz said it at p 141 lower left in the paper cited by MarkR at #31

    “If it is true that two analogues have occurred since atmospheric observations first began, it follows, since the atmosphere has not been observed to be periodic, that the succession of states following these analogues must eventually have differed, and no forecasting scheme could have given correct results both times”. (There is more – see his graphs on p 137 to see how simulations of textural changes fared).

    This was written before 1963, so the word “forecasting” would need to be replaced by “projecting” or something similar.

    Would it not be the case that the 2008 onwards data would need two components, being a continuation of the old texture and the overlay of a further texture due to a new cause. To do maths, would one not have to separate the two components? How? Impossible?

    Typo in name above # 30 – apologies David Stockwell.

  35. The Librarian
    Posted Jul 26, 2008 at 5:20 AM | Permalink | Reply

    #21 David

    You have introduced the term DECR not previously used on this blog.
    I am struggling with this one. Looks like you are referring to separate stages of a modeling workflow?
    viz: Define Execute Correlate Reassess or altenatively Define Evaluate Correct Reappraise etc.

    Please remember the poor librarian. :wink:

  36. Ian Castles
    Posted Jul 26, 2008 at 5:59 AM | Permalink | Reply

    Re #35. I think ‘DECR’ stands for ‘Drought Exceptional Circumstances Report’, which is the cover title of the document that is the subject of this thread.

  37. gb
    Posted Jul 26, 2008 at 6:34 AM | Permalink | Reply

    Steve M: I do not know what you are trying to do but are you comparing directly one GCM simulation with the observation? But that is not correct …

    Steve: David and I have been discussing methods of comparison. The point is that the observed Qld graphic doesn’t look like the GISS model in the observation period. The trends go in opposite directions. So what weight can we place on the fact that the GISS model shows increasing drought in the future? Should we conclude that there will be less drought in the future since GISS got the trends reversed in the past? Or that the model has nothing to say on the matter? I reported this quickly and am not advocating any particular method of evaluating things – the information provided is incomplete and, as someone observed elsewhere, they have reported an oddball statistic, complicating analysis. But the very fact that they picked an oddball statistic means that’s it’s harder for them to assert statistical significance. I’m sure that David will be looking further at exactly how they claimed to have obtained statistical significance, but my first impression is that the calculation is pretty ropy.

  38. Posted Jul 26, 2008 at 1:28 PM | Permalink | Reply

    I think they did no more than average the data before and after present. You need to use extreme value statistics at least, and looking at the return period (expected time to next drought) would be of interest but not in the report. But the return period would be way off for the graphs you show too. Another one to look at would be the Murray Darling Basin. In the grip of drought now, is not predicted for significant drought.

    Geoff, The way I see it there is a chain of implication that needs to be established but in the report is merely assumed. I goes something like.
    Observations 1-> Model past 2-> Model future
    We have been talking about implication 1 and you are talking about implication 2. Both are necessary assumptions though. There are plenty of papers around that show GCMs are just not good enough yet for regional forecasting.

  39. Philip_B
    Posted Jul 27, 2008 at 2:50 PM | Permalink | Reply

    The SouthWest region is in slow decline. Naively, much of the weather for the SouthWest region, Victoria and Tasmania comes sweeping up in arcs from the oceans south of Perth, with the south eastern and south western coasts and immediate inland catching the top fringe of the fronts as they head east. If the arcs move a bit further south, rainfall can decline.

    I realize this is the conventional wisdom and doubtless embedded in the models, but it only partly fits the facts, and as an analysis it suffers from drawing conclusions based on averaging what has happened in a relatively small area over a (much) larger area.

    Substantial rainfall deficits are concentrated along the coast south of Perth (and to a lesser extent further north). Incidentally the wettest part of southern WA. Go north and east and rainfall deficits decrease and only a 100 or 200 Ks inland you start to see above average rainfall since 1970. So the weakening fronts argument fits the data along the coast but not further inland. See here for maps.

    Figure 10 at the link is interesting. It shows rainfall trends for 3 overlapping periods. It shows the coastal area has had a decreasing rainfall trend, and away from the coast increasing rainfall trend, since 1900.

    In summary, the data doesn’t fit moreorless uniform decreases in rainfall across SW Australia from weaker Southern Ocean cold fronts.

    This year is probably illustrative of why. We had a couple of mid-lattitude disturbances over Perth that produced heavy rain – about 100mm each, which is about 20% of our annual rainfall and the percentage effect further inland would be larger. These mid-lattitude disturbances don’t seem to fit any annual pattern and vary a great deal in size and where they produce rainfall and how much. It’s my observation that they seem to account for the difference between a wet and a dry year, and any effect from weaker cold fronts is probably secondary.

    The models and AGW theory predict less rain from weaker cold fronts. Average the rainfall data over a large area and bingo the data supports the theory. Shades of the global mean surface temperature showing global warming.

  40. Vincent Guerrini Jr
    Posted Jul 27, 2008 at 3:33 PM | Permalink | Reply

    #33 dale C thanks for your expose: so rainfall has been INCREASING since 1950? This is completely contrary to AGW (more prolongued frequent droughts). BTW seems that this winter so far rainfall above average for qld anyway.

  41. Jonathan Schafer
    Posted Jul 27, 2008 at 3:46 PM | Permalink | Reply

    #40

    Not really. So far as I can tell, NOTHING is contrary to AGW climate effects. Increased heat/cold, increased precip/drought, increased/decreased TC’s, etc. etc. It appears no matter what happens, the AGW crowd has their bases covered.

  42. DaleC
    Posted Jul 27, 2008 at 9:12 PM | Permalink | Reply

    re Vincent #40,

    Expose is a bit strong – it could be that less rain is falling more often, so soil moisture could decline even as total annualised rainfall goes up. I was just attempting to give others a context to help follow the arcaneness of the model debate. The prima facie contradiction of increased rainfall nonetheless implying via modelling more drought does demand some explanation, however.

    re Philip_B, #39

    Thanks for the greater detail. I tried to cover myself with the qualifier “Naively”. The basis for my statement on fronts is nothing other than long term observation of TV weather reports. It is remarkable how often the fronts come from the ocean below the continental south west. Even the hot north winds which plague Victoria in summer often seem to be just a wider arc, which then curves down southwards after being heated over the central deserts. When I watch the fronts arcing in, I get the impression that the inhabitants of the southern coastal fringes are living on a climatic knife-edge. The difference in rainfall volume from copping the body of a front, as opposed to its edges, is perhaps apparent in Tasmania’s doubling of the Victorian average, even though it is only a few hundred kilometres further south. But maybe topography explains it.

    Any further illumination on the drivers of Australia’s southern climate are most welcome. What I would really like to know is what happened in the mid 1940s? Why are the per state time series so much flatter before 1950? Why the greater averages since then? Why the much greater swings in amplitude after 1950? And this seems to be continent-wide except for the south west.

    Finally, in my original post at #33,

    “…there is a distinct change between 1990-1949 and post 1950″

    should of course read

    “…there is a distinct change between 1900-1949 and post 1950″.

    PostScript:

    It’s always worth remembering Dorothea Mackellar describing the climate of the late 19th century: Australia is “a land of drought and flooding rains”, of “flood and fire and famine” where “sick at heart, we watch the cattle die”. So not much has changed, it seems.

  43. tty
    Posted Jul 28, 2008 at 12:46 AM | Permalink | Reply

    Re 40

    I’ve never been able to understand where they got that idea that AGW will cause increased drought. The paleoclimatic data unambiguously shows that warm = wet and cold = dry almost everywhere. Deserts expand during glaciations and shrink during interglacials, monsoons strengthen and spread further north and south during interglacials. Lake-levels rise during interglacials and drop during glacials. Rainforests expand during interglacials and are replaced by grasslands during glaciations etc etc.

    This is not strictly true everywhere of course. Arid areas caused by rain shadow are not much affected by climate swings (they may even become slightly wetter during cold glaciations due to decreased evaporation). However about the only areas I know of where there is actual evidence that interglacials can be drier than glacials are the intermontane basins of the western US and – possibly – southeast Australia.

  44. Ian Castles
    Posted Jul 28, 2008 at 1:45 AM | Permalink | Reply

    Re the discussion of observed and projected trends in rainfall and droughts in south-west Western Australia (SWWA) at #39 and #42. I’ve noted on the ‘CSIRO: A Limited Hang out??’ thread (post #35) that the recently published and much-cited paper

    Hennessy et al. (2008): ‘An assessment of the impact of climate change on the nature and frequency of exceptional climatic events’. A consultancy report by CSIRO and the Australian Bureau of Meteorology for the Australian Bureau of Rural Sciences, 33pp, http://www.bom.gov.au/

    mistakenly cites the bibliographical details of a paper entitled ‘Effect of GCM bias on downscaled precipitation and runoff projections for the Serpentine catchment, Western Australia’ (which is in SWWA) in lieu of the paper on Australian droughts (Mpelasoka et al, 2008, Comparison of suitable drought indices for climate change impacts assessment over Australia towards resource management’, Int. J. Climatol., 28: 1283-1292) to which Kevin Hennessy and his CSIRO/Bureau of Meteorology (BoM) co-authors presumably intended to direct their readers.

    Hennessy et al (2008) has it that there was a decrease of 62 mm. in the regionally averaged total rainfall in SWWA between 1950 and 2007 (Table 1, p. 7). However, according to Gallant et al (2007) ‘Trends in rainfall indices for six Australian regions: 1910-2005′, Australian Meteorological Magazine [AMM], December 2007 – which is also co-authored by Kevin Hennessy – the trend rate of decline in total annual rainfall for the Southwest region (which appears to correspond to SWWA: see map, p. 226) was 23.54 mm per decade between 1950 and 2005. This implies a decrease of around 130 mm over the five-and-a-half decades, which is more than twice as great as the Hennessy et al (2008) figure. And Gallant et al (2007) also reported a trend rate of decline for the region of 20.66 mm per decade for almost a century (1910 to 2005), implying a total decrease of as much as 200 mm in average annual rainfall during this period. It’s difficult to reconcile this with the data plotted in the relevant time series graph (p. 32) in Hennessy et al (2008), which reveals that the 11-year moving average of rainfall in SWWA showed relatively modest movements over the century (and no net change).

    Maybe the numbers given in Gallant et al (2007) have now been found to be incorrect. Ailie Gallant advised me on 8 May that I had been ‘correct in pointing out the mistakes in rain day values’ in most of the tables in that paper, and said that she’d sent revised tables (also incorporating other corrections) to the editor of AMM. The journal’s website was amended on 4 June to say that a full update of the Gallant et al (2007) paper would take place ‘shortly’, but the corrections that Ms Gallant notified on 4 May have yet to be effected.

    So far as projections are concerned, the Second Order Draft (SOD) of the ‘Australia and New Zealand’ chapter of the IPCC’s 2007 Assessment Report used ‘Mpelasoka et al. 2005′ to support the claim that ‘Two climate models simulate … up to 80% more droughts in south-western Australia.’ Unfortunately, Coordinating Lead Author Kevin Hennessy seems to have misinterpreted the Australian Government’s comment ‘’Two climate models …’ More detail is needed here. Which two and why only these two?’ (Comment G-11-89). Instead of answering these pertinent questions, he responded ‘Pilot study with limited resources. Reference to two models deleted.’ As a result, the final IPCC report did not mention that the statement ‘..[U]p to 80% more droughts [are simulated] by 2070 in south-western Australia (Mpelasoka et al., 2007)’ (p. 515) was based on results from only two models: CSIRO MkII and the Canadian CCCmaI. It is however clear from Mpelasoka et al (2008), Table II, p. 1290, that this is the case.

    The SOD also said that ‘A tendency for decreased rainfall is likely over most of Australia, except Tasmania and NSW’ (p. 11, line 40). This was questioned by expert reviewer Janice Lough, senior research scientist at the Australian Institute of Marine Science, who commented that she was ‘still a bit uncomfortable’ with statements such as this (Comment E-11-189). Kevin Hennessy replied ‘Noted. A surprising comment given the detail in Table 11.3 … Suppiah et al (2006) also reported results for the 15-model average changes in …. rainfall over Australia, which clearly showed decreased rainfall over the whole continent. This is now mentioned.’

    Rainfall projections for five regions of Australia were given in Table 11.3 (p. 12) of the SOD, and for five somewhat different regions in Table 11.4 (p. 515) of the final report. The ranges of uncertainty in the SOD Table are said to be ‘based on results from … 18 climate models (Suppiah et al, 2006)’, and the ranges in the table in the final report are said to be ‘based on results from … fifteen climate models for various locations in each region (Suppiah et al, 2007).’ The final report also cites Suppiah et al, 2007 in support of the claim that ‘The 15-model average shows decreasing rainfall over the whole continent.’

    This is the very kind of generalisation with which Dr. Lough said that she was uncomfortable, and is completely at odds with the results from the same source in the accompanying table on the same page, which show the projected change in rainfall in ‘Northern NSW, Tasmania and central Northern Territory’ for 2080 relative to 1990 as between minus 27% and plus 54%; in ‘Central South Australia, southern NSW and north of latitude 20 S, except central NT’ over the same period as between minus 27% and plus 27%; and in ‘Inland Queensland’ over the same period as between minus 54% and plus 54%.

    The paper cited as ‘Suppiah et al, 2006′ in the SOD was included in the list of References as ‘Suppiah, R., K.J. Hennessy, P.H. Whetton, K. McInnes, I. Macadam, J. Bathols, J. Ricketts, 2006, and, 2006 [sic] Australian Climate Change Scenarios Derived from AR4 GCM Experiments. Australian Meteorological Magazine (submitted)'; and the paper cited as ‘Suppiah et al, 2007′ in the final report was listed in the References with the same seven authors (all from CSIRO) but with a somewhat changed title (‘Australian climate change projections derived from simulations performed for the IPCC 4th Assessment Report’) and with the status of the paper changed from ‘submitted’ to AMM to ‘accepted’ by AMM.

    In fact, it appears that the paper cited as ‘submitted’ to AMM in the SOD had not been submitted to that journal; and the paper cited as ‘accepted’ by AMM in the final report had not been accepted by the journal. The evidence for this is that the manuscript of the only paper by Suppiah et al that has been published in AMM is stated on the cover page to have been received in March 2007 (i.e., after the text of the IPCC’s Working Group II report had been finalised) and to have been revised in August 2007 (after the relevant IPCC report had been published). And the paper that was published in AMM had an additional co-author who had not been included in the IPCC References lists (C.M. Page).

    As finally published, Suppiah et al (2007) ‘present[s] climate change projections based on the results from 23 climate model simulations performed for the IPCC 4th Assessment Report. Statistical methods are used to test how well each model simulated observed average (1961-1990) patterns of mean sea level pressure, temperature and rainfall over the Australian region. The 15 models with the highest pattern correlations and rms errors are identified … Using the best 15 climate models, annual and seasonal average projections of Australian rainfall and temperature change are derived …’

    An ironic result of this procedure is that neither of the models developed by the Canadian Climate Centre (CCC) that were available from the 23 simulations performed for the IPCC was included in the suite of 15 models that best reproduced Australian average patterns of climate in the 1961-90 period. Yet a model developed by the CCC is one of the two used in the Mpelasoka et al (2008) paper in the August 2008 issue of the International Journal of Climatology (IJC). And, just to make confusion worse confounded, Hennessy et al (2008), which has three co-authors in common with Mpelasoka et al (2008), lists the latter paper – but with an incorrect citation to a 2007 issue of IJC: see above – as one of six ‘previous studies [which] have not adequately done this analysis’ (p. 13). The only apparent reason proffered for this appears to be that ‘SOME of these studies used the Palmer Drought Severity Index’ (EMPHASIS added), but on my reading of the paper Mpelasoka et al (2008) did not use that Index.

    My apologies in advance for any inadvertent errors or misrepresentation in the above analysis, and my thanks to David Holland and Steve McIntyre for securing the release of successive drafts of IPCC reports, reviewer comments thereon and author responses thereto. Without this invaluable database, much of the above analysis would not have been possible. I believe that the above is enough to show that Australia deserves better. I await with interest the further examination of Hennessy et al (2008) which David Stockwell has under way.

  45. BradH
    Posted Jul 28, 2008 at 6:15 PM | Permalink | Reply

    So, tell me: how much grant money will the CSIRO give you to come up with a model with negative historical correlation? Has the world gone completely insane?

    #33 DaleC – nice job. There might be something there after 1950, but honestly, it all looks pretty random to me.

  46. Murray Harris
    Posted Jul 28, 2008 at 8:33 PM | Permalink | Reply

    We need an open and honest debate in Australia.
    It almost needs exceptionally simple language presented
    in a way that every person can understand and
    is challenged to respond to, much like modern advertising methods.

  47. Doug
    Posted Jul 28, 2008 at 8:36 PM | Permalink | Reply

    well theyve been pushing for a royal commission into climate change but the pollies dont seem to think thats important

  48. Doug
    Posted Jul 28, 2008 at 8:37 PM | Permalink | Reply

    dont forget sydney DIDNT get snow yesterday it was merely soft hail or as someone else put it – hard rain ROFL

  49. John McLean
    Posted Jul 29, 2008 at 5:34 AM | Permalink | Reply

    You may be interested in my two reviews (one p-r’ed, one not) of CSIRO climate change reports. The peer reviewed paper is this PDF and the other, which the CSIRO created with the Bureau of Meteorology, is here.

    Also relevant to this subject is …
    Wang, G. and H. Hendon (2007) “Sensitivity of Australian Rainfall to Inter-El Nino Variations”, Journal of Climate, v20 pp 4211-4226 (15 Aug 2007)

    The paper’s conclusion begins

    Australia tends to experience drought during El Niño, but not all El Niños are equally influential. Drought usually develops during El Niño, especially in the eastern two-thirds of the continent during austral spring season, because the Walker circulation shifts eastward, resulting in higher surface pressure and anomalous sinking motion at Australian longitudes. El Niño events that develop with the strongest SST anomalies shifted into the far eastern Pacific (e.g., El Niño 1997) tend not to produce as severe a response as those events that are more focused near the date line (e.g., El Niño 2002). This sensitivity stems from the anomalous Walker circulation shifting too far east to effectively suppress Australian rainfall during El Niño events that maximize in the far eastern Pacific.

    Given that for most of the period 1997 to 2007 conditions in the Pacific were close to El Nino or beyond that threshold, Wang and Hendon indicates that sustained drought conditions in Australia across that period are no surprise.

    The question is whether Hennessy et al’s data can be attributed to these natural causes.

  50. Bob Cormack
    Posted Jul 30, 2008 at 5:19 PM | Permalink | Reply

    Steve:

    Some of it are Excel files within tar files, making the extraction even more time consuming as each Excel file has to be saved into a csv file in order to be read into R or Matlab for statistical analysis.

    I don’t know about R, but Matlab can read Excel files using the xlsread function. This would save you having to open the file in Excel, then saving it (assuming that you know which columns are which). It is even possible to have the Matlab script do a little parsing of the Excel file, such as searching for specific column headings, etc.

    Alternatively, Matlab can drive Excel using the DDE functions. A Matlab-controlled Excel can load data, then ship it to Matlab.

    Current versions of Matlab also have tar, untar, zip, unzip, Gzip, Gunzip functions which can deal directly with archives.

    Also, any program that can be run from the command line (including operating system commands) can be run from inside Matlab by prefacing the command with an exclamation mark, (!).

    I’ve had a lot of practice reading weird data into Matlab (although my Matlab license is very out of data: ver 11.1); so if you would like to email me with some specific problems, I’d be glad to work up some scripts/functions for you.

  51. AndyC
    Posted Aug 11, 2008 at 10:00 PM | Permalink | Reply

    So, is it gonna rain or not?

One Trackback

  1. By Niche Modeling » CSIRO and BoM Report on Mar 14, 2010 at 8:42 PM

    [...] Some quick thoughts on CSIRO drought info. [...]

Post a Comment

Required fields are marked *

*
*

Follow

Get every new post delivered to your Inbox.

Join 3,131 other followers

%d bloggers like this: