Understanding Climate Risk

Science, policy and decision-making

Northern Victorian Flood Review Part III

with 4 comments

This is the third post covering the January 2011 floods in northern Victoria. Part I described how the floods unfolded and Part II described the hydrology of the Loddon catchment and how its history has affected flood behaviour. This final post covers the climatic influences on the floods. The floods themselves are discussed here, and affects on the family farm in Kerang are shown here.

Northern Victoria’s rainfall is influenced by the El Niño–Southern Oscillation (ENSO; spring–summer), Indian Ocean Dipole (IOD; winter–spring) and Southern Annular Mode (SAM; winter). Correlations of northern Victorian rainfall with simple indices are -0.51 for SOI May–Jan, ‑0.48 and -0.45 for two IOD indices May–Jan, and -0.25 for SAM for winter rainfall.Recent research has linked exceptionally wet conditions in south-eastern Australia with the combination of La Niña and a negative Indian Ocean Dipole (-IOD) (Ummenhofer et al., 2009; Ummenhofer et al., 2011).

The -IOD is associated with exceptionally warm waters off north-western Australia resulting in ‘north-west’ cloud bands that bring flooding rains into northern Victoria. The recent extended drought in Victoria has been characterised by an absence of these weather patterns. While central and north-western Australia have become wetter, this rainfall source has been conspicuously absent from south-eastern Australia.

Ummenhofer et al. (2009; UEA) nominate the following years as combined La Niña-negative IOD years: 1906, 1909, 1916, 1917, 1933, 1942 and 1975. These dates are based on Meyers et al.’s (2007; MEA) classification of -IOD that includes upwelling and/or strong winds near Java (MEA). MEA also included 1910, 1928, 1950 and 1981. Based on a more straightforward use of the Indian Ocean Dipole Mode Index, the Bureau of Meteorology nominates the years 1964, 1971, 1974 and 1975 as La Niña – negative IOD years since 1958 (BoM). However, if the simple Dipole Mode Index (DMI) is used, there are a good many false positives, so the accompanying climatological analysis by UEA and MEA nominating positive and negative IOD years is vital for assessing the  statistics.

When flood years are ranked according to the highest 40 monthly flows at Laanecoorie Reservoir in Victoria since 1900, of the 31 years represented, 11 (15) are combined La Niña -IOD years, 5 (8) are La Niña years, 11 (13) are neutral years, one is an El Niño year, one a -IOD year and one a +IOD year (numbers in brackets are occurrences). Laanecoorie Floods Table. All 7 of UEA’s dates for La Niña –IOD register, one of four extra MEA dates, two of BoM’s three dates  and 2010 as yet not analysed. These add up to the at least eleven La Niña-IOD flood years estimated to have occurred in the past 110 years. Over the same period 29 La Niña years with neutral IOD occurred and one La Niña coinciding with a +IOD (UEA 2009, 2011).

Most La Niña -IOD years have produced a major flow at Laanecoorie and about half have been major floods in northern Victoria: 1909, 1933, 1974, 1975 and 2010–11. La Niña years with major floods are 1956 and 1973, but the statistics for major monthly flows and La Niña years are not as high – they occur at a frequency slightly higher than La Niña years coinciding with neutral IODs. Neutral conditions (No ENSO or IOD event) occur about 50% of the time, but are associated with about one-third of major flow years. Therefore there is about a 20% chance of getting a high monthly flow in a neutral year. El Niño and +IOD events occurring with high flow events are rare but not unprecedented.

The lesson from this  is clear. La Niña-IOD years are highly likely to have moderate or major flooding. La Niña years have a slightly higher than chance probability of  having high flow events. However, because La Niña years are on the whole wetter  than normal, regional catchments are likely to be wetter – they are  preconditioned to flood. Negative IOD conditions usually only occur after a La Niña is in place – in 2010, negative IOD conditions arose halfway through August (see Figure). Flooding followed immediately. La Niña conditions and/or  catchment wetness should be used to establish flood readiness – these are conditions when heavy rain, if it occurs, is likely to cause flooding.

NINA3 and Indian Ocean Dipole Index Winter-Spring 2010

There are several Dipole Mode Indices measuring the IOD. The straight index does not serve as a diagnostic for flood conditions. An attempt was made to see if the relationship between northern Victorian rainfall ENSO index and IOD index has changed over the past century. This created a regression relationship between Northern Victorian rainfall and the ENSO and IOD indices. No evidence of change was found, whereas eastern Australian rainfall has increased relative to ENSO. However, the model does seriously underpredict the last two -IOD related flood events on the Loddon in 1992 and 2010. The figure below shows that the model estimates are out by a couple of hundred mm. Adding ENSO and IOD give better estimates than using only one predictor. Are extreme rainfall events getting wetter? This provides some evidence that it may be the case, but it’s not very compelling evidnce. Further research is required to build a better model for diagnosis and prediction that develops MEA and UEA’s work further.

Simple regression model to predict northern Victorian rainfall from ENSO and IOD indices and the IOD index alone.

Lastly, a great deal of research has been carried out to investigate the cause of the recent extended drought (Timbal et al., 2010). An anthropogenic component is likely, and has been linked to the strengthening of the sub-tropical ridge, linked to less storm generation over the mid latitudes and strengthening of the circum-polar vortex, pushing the rain-bearing frontal systems south. Therefore, dry conditions are likely to continue but do not rule out large floods occurring.

What are my recommendations for future future management?

  1. Build flood modelling capacity to estimate flood volume, speed, location and peaks. This can also be used for mitigation before events by using past floods with current storage and soil moisture and also to run scenarios
  2. When negative IOD conditions, accompanied by phenomena such as a shallow thermocline and upwelling in the Java-Sumatra region, are sustained for more than two weeks during a La Niña event, the northern catchments of Victoria should be placed on full flood alert.
  3. A wet catchment with full storages, La Nina conditions or a negative IOD should place the catchments on flood watch. This is time to look at potential mitigation, review flood plans and ensure that flood mitigation capacity is in good shape.
  4. During full flood alert any temporary works limiting flood mitigation capacity should either be completed urgently, or mothballed if that is not possible, to make that capacity available. Storages such as the Kerang Lakes that can be partially emptied should be partially emptied. Modelling should be used to set the limits for reducing storage levels that best allow trade-offs between meeting environmental flow targets and flood mitigation. Agreements between the states at the scale of the Murray Darling Basin are also required. Major anabranches should not be closed when such conditions occur, and seldom filled wetlands can be earmarked and flooded opportunistically.
  5. During floods the local situation should be managed by local flood wardens with outside assistance operating under their guidance. It is not possible to import instant experts who know enough about what they are doing. This will require some training. See Stuart Sims’ comment about how not to manage a severe flood situation.
  6. Further research to develop a better climate based prediction system that can reliably pinpoint likely conditions for extreme rainfall needs to be undertaken. Built into a warning and flood mitigation system, such research has the potential to save millions of dollars damage into the future.
  7. Northern Victoria is a water-sensitive landscape. Floods larger than the January 2011 are possible, if the indications provided by the simple model are correct (although this is low confidence). A drying climate with wetter extremes seems the most plausible climate outcome for coming decades given the current scientific state of knowledge. Long-term planning is required to adapt to these and other plausible changes tha also take into accounting changing environmental, social and economic conditions.


Meyers, G., McIntosh, P., Pigot, L. and Pook, M. (2007) The years of El Niño, La Niña and interactions with the
tropical Indian Ocean. Journal of Climate, 20, 2872-2880.

Timbal, B., Arblaster, J., Braganza, K., Fernandez, E., Hendon, H., Murphy, B., Raupach, M., Rakich, C., Smith, I., Whan, K. and Wheeler, M. (2010) Understanding the anthropogenic nature of the observed rainfall decline across south-eastern Australia, The Centre for Australian Weather and Climate Research, Melbourne.

Ummenhofer, C.C., England, M.H., McIntosh, P.C., Meyers, G.A., Pook, M.J., Risbey, J.S., Gupta, A.S. and Taschetto, A.S. (2009) What causes southeast Australia’s worst droughts? Geophysical Research Letters, 36, L04706.

Ummenhofer, C.C., Sen Gupta, A., Briggs, P.R., England, M.H., McIntosh, P.C., Meyers, G.A., Pook, M.J., Raupach, M.R. and Risbey, J.S. (2011) Indian and Pacific Ocean influences on southeast Australian drought and soil moisture. Journal of Climate, 24, 1313-1336.


Written by Roger Jones

June 3, 2011 at 3:37 pm

4 Responses

Subscribe to comments with RSS.

  1. What happens if SAM, or the AAO, is added to your mix? It may be that the Southern Annular Mode only has an influence on rainfall south of the Divide, ie the western districts, the south east of SA etc. Has anyone done any modelling using all three modes, ENSO, IOD and SAM? The BoM seems to be preoccupied with ENSO, although of late they seem to have discovered the need to add the IOD. I have just started massaging NOAA data to see if I can do a multivariate analysis. Do you think this would work? Cheers.

    Kevin Brewer

    September 2, 2015 at 8:04 pm

    • Kevin, I’m a bit out of touch with the latest on this, but…

      The reason that the IOD and ENSO matter is because they are interannual phenomena. SAM is a decadal mode that conditions some of these other phenomena, so works on a different timescale. The Interdecadal Pacific Oscillation is similar in the way it works. With regard to 2011, warming off NW Australia is critical in the amount of water that came across (therefore SAM is not really influential). Because we are in a highly non-stationary environment, if we can get a general idea of what influences events, that’s great, but there is unlikely to be too much gained from looking into the statistics of the past to get much of an edge beyond that.

      Part of the problem with multivariate analysis is having data that goes far back enough and the fact that we are now highly non-stationary. Climate change and variability are combining to create different phenomena. There is no doubt many of the existing processes are playing a role, but warming is messing with that. Model experiments have become a critical part of this process, because they allow some of these influences to be assessed one at a time.

      Curiosity may be rewarded though, so why not have a look?

      Roger Jones

      September 3, 2015 at 10:01 pm

      • Roger, my reading of SAM is that it fluctuates on a very short time scale. BoM reckons they can predict it 3 weeks out, but looking at the data, eyeing it in, it seems to often have a very short period, a matter of days. That is why they are not as interested in it as ENSO and IOD, or the MJO. Reading a lot of papers lately very few seem to take non-stationarity as a problem for their analysis, which seems odd, as there are now methods of dealing with it like Empirical Mode Decomposition. Using R and massaging one’s time series makes it possible for even the likes of me to do it. It is the interpretation of the IMFs that is the hard bit. And I have just read about kriging as a means of doing geospatial analysis, which seems a useful technique for plotting the influence of the acronyms all over the place. James S. Risbey, Michael J. Pook, Peter C. McIntosh, Matthew C. Wheeler, and Harry H. Hendon, 2009: On the remote drivers of rainfall variability in Australia. Mon. Wea. Rev., 137, 3233–3253 published by the American meteorological society online has a fairly detailed analysis of SAM, IOD, SOI and a category called blocking, which is a high pressure cell over the Tasman that has considerable influence on rainfall on the east coast (and on summer temps in se Australia through the hot northerlies it produces). They use the standard methods of correlation, not spatial correlation, nor non-stationarity. Then there was a small body of work from the 20s to the 50s in Australia that is now ignored, which picked up signals of solar affects. Modern meteorology hates the word sunspot, but I have read a couple papers that use them as a way of dealing with changes in the influence of the sun on the troposphere, and how those changes move down into the atmosphere.

        Kevin Brewer

        September 9, 2015 at 5:38 pm

  2. Kevin, the SAM index is multi-decadal.
    See here http://www.nature.com/articles/srep08909
    The current national climate science plan also provides a bit of insight cawcr.gov.au/projects/vicci/documents/Science_plan_final.pdf

    The Risbey et al paper is relevant, but the big question is about things like blocking and cut-off lows and what they may do in future

    Roger Jones

    September 9, 2015 at 8:46 pm

Leave a Reply

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

WordPress.com Logo

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

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s

%d bloggers like this: