There has been substantial discussion on the blog over the last few months as to the severity of the current drought afflicting much of Australia and its cause. A fair emphasis has been as to whether the current drought is the “absolute worst” on record or whether they have been “historical anecdotes and personal rainfall data sets before sufficient formal Bureau records, or paleological droughts recorded in coral cores” that were worse. Via the Australian newspaper, Barrie Hunt has contributed the results of a 10,000 year Global Climate Model run to further complicate this argument suggesting we only have a narrow view of what our real climate variability involves.
To some extent does it really matter? Is it just a public bar sports debating point. We know that the drought is widespread, severe, entrenched and multi-year in nature. It’s not much fun if you’re in the middle of it. Surely it is “among” the worst droughts on record in many important areas. All droughts have their own unique signatures and different areas show different impacts. It is simply not possible to say whether this drought is other than natural variability – although one can argue there may be climate change aspects or influences exacerbating the situation. We’ll know for sure in 30-40 years time.
What does matter is to whether we are adapting and responding to drought risk, water security, land use and population changes in a more mature manner. This involves seeing drought as part of normal and planning for it appropriately at farm, regional, industry, state and national levels.
Nevertheless I believe we do need a better spatial understanding of this multi-year drought which some have said has been building for 5, 6 or 7 years. A reasonable way to do this is to undertake a decile analysis of rainfall for the Australian rainfall network. We need to examine how bad the sequence of years are for rainfall deficit. The Bureau of Meteorology has come under quite a bit of criticism by blog commentators, but notwithstanding, I asked Dr David Jones, Head of Climate Analysis, from the National Climate Centre to supply a 5 year and a 7 year decile sequence analysis for their reference network which he has kindly provided. Why debate the issue in a data free vacuum when we can make a phone call or email !
The analyses are included as maps below.
The maps represent a 5 and 7 year decile analyses based on the sequence of 1900 till now. For example – where does the recent 5 or 7 annual rainfall sequences fit into any of the sequences since 1900.
Anyone can easily do these analyses for a single station or stations – you just need to do the mathematics of adding up the year-totals in a running mean fashion and grouping your data into decile bins. So for the five year decile – arrange your data from earliest to latest records, add up the first 5 years in the record and record the total, move the band along one on the rainfall record sequence and do it again, and so on and so on. When all done, rank the new totals highest to lowest in decile bins (0-10%, 20-30%, 30-40% and so on). You can then see where any individual sequence, or the last 5 years in this case, fits in the distribution of all 5 year sequences. Or in fine detail the lowest and highest on record.
Similar logic for the 7 year analysis.
But as you may appreciate a lot of work to do spatially and interpolate the results.
And it is important that it is done as a sequence of years – one cannot add the individual year deciles/percentiles together.
The maps unlike the time series show we live in two Australias – one wetter and one drier. In the majority of areas where we have major agriculture or urban populations we have drought issues – some moderate – some severe in parts. Meanwhile life in the Gibson desert is pretty good (relatively speaking). This why the times series data are not helpful – the addition of real good and real bad = average which doesn’t represent reality
Of course this is only rainfall – there are obvious issues of pattern and intensity of rainfall, how plants and soil respond, and antecedent conditions (i.e catchments currently dry as a chip) – if we’re debating streams and rivers I really think one needs the same analysis but on streamflow.
Similarly one can use a pasture or crop model to get decile crop yields or pasture biomass (decile of wheat and grass!) However experience suggests that often these further analyses amplify effects and it will look worse for drought. Usually same with streamflow.
I understand the Bureau will be undertaking some more work on this issue in coming months.
I encourage any readers to do their own numbers with their own data – but please explain what you have done so we’re all clear.
PS obviously where station density is low – i.e. central Australia and central WA the spatial interpolation is only as good as the data point grid. But I believe we have a sufficient grid for some intelligent analysis in areas that people live or major agriculture occurs. One could obviously argue the spatial interpolation mathematics and accuracy in fine small areas (i.e. SE Qld). The data do not start in the 1890s and the Bureau do have issues with station quality before 1900. The data are of mixed quality.