Transcript for VicWaCI Webinar Drought and Deluge streamflow regime shifts after multi-year drought in south-eastern Australia

[Speaker: Sandra Dharmadi (DEECA)]

Hi everyone.
We are just going to start the webinar in a few minutes.
We just wait for a few more people to join us.
Thanks for joining us today, everyone.
We will start in a minute or two.
Okay. Welcome.
Welcome everyone to this webinar that's being hosted by the hydrology and Climate Science team at the Department of Energy Environment and Climate Action.

To showcase the research from the Victorian Water and Climate Initiative, like previous webinars, we'll be recording this presentation with the intention of making it available after the event on our website.

To start with, I like to acknowledge the Traditional Owners of the lands on which we are meeting today.
In my case, this is the land of the Wurundjeri of the Kulin nations.
And I like to pay my respects to their elders past and present and extend that acknowledgement to those across other parts of Victoria and Australia where people may be joining us today.
My name is Sandra Dharmadi and I work in the hydrology and climate science team where we manage the Victorian Water and Climate Initiative.

Other team members have been working hard to plan and coordinate this session and to make sure that the technology runs smoothly.
So thank you to Geoff Steendham, Rachel Brown, Jacqui Lloyd, and Gil Aitken.
The second phase of the Victorian Water and Climate Initiative or VicWaCI for short has been underway for a few years now.
We are rapidly approaching the end of the second phase, which will be marked by the release of our VicWaCI synthesis report later in this year, and supported by a webinar series, including this webinar today.

So today we have Dr. Ulrike Bende-Michl from the Bureau of Meteorology to talk about how stream flow patterns have changed during and after multi-year drought in Southeastern Australia.

This study is part of a broader research effort by our research partners to understand how and why we've observed a shift in for run of relationship in some catchments during the millennium drought and in some cases continuing after the drought.
We are planning to have further webinars on the broader topic of shifts in rainfall runoff relationship in this webinar series. So stay tuned.

By default, the webinar will run with audience microphones and camera switched off. However, I invite you, the audience, to use the Q&A function in Teams to post questions or comments at any time. We will present this question to Ulrike following her presentation.
So before I start formally introducing our speaker, I'd like to thank everybody for joining us today and hope you enjoy the event.

So Dr. Ulrike Bende-Michl is a senior research scientist at the Australian Bureau of Meteorology, where she focuses on seamless hydro climatic impact studies.
Her work explores processes driving changes in hydrological systems, analyzing past impacts of hydro climatic extremes like droughts and projects, how they might change in the future.
Ulrike has led the first national hydrological projection and is currently working on the National Climate Risk Assessment.
She has also collaborated widely with CSIRO and various international institutions.
So welcome Ulrike and take it away.

[Speaker: Dr. Ulrike Bende-Michl (Australian Bureau of Meteorology)]

Thanks so much, Sandra, for your lovely introduction and also thanks everyone for joining this webinar today.
As Sandra explained, in this webinar, we will explore further how Streamflow has changed after multi-year droughts in southeastern Victoria or respectively Victoria or Southeastern Australia.
So what do we mean by multi-year drought? Is Victoria also frequently exposed to droughts? The most pronounced one was the millennium drought, happening between 1997 and 2009.
So just a quick recap on the left-hand side. What some of the characteristics of the Millenium drought were?
So, you might remember that, the drought was characterized by strong below-average cool season rainfall that led to large declines in water availability across Victoria.
Agriculture was particularly impacted, so there were large losses.
It also had impacts on the environment including water quality with frequent fish kills and a decline in aquatic diversity.
Water supply was also largely impacted and that led to changes in how water was managed across Victoria and elsewhere, but also changes to water policies.
Some of the mitigation strategies included, the introduction of permanent water-saving measures, but also the exploration of alternative water sources.
And in terms of water policies, the most impactful introduction was the Water Act in 2007.
So post-Millenium Drought, even though that was in 2010/11 and 11/12, we had two strong La Nina years, which really broke the drought with record rainfall.
We observed that there was a non-recovery in what's called the rainfall runoff relationships.
So for one unit of rainfall, there was a two to three or even more louder amplification of runoff reduction and therefore leading to strong reductions of yield.
And that was also a key feature of the drought.
Of course, on the right-hand side, you see this sort of visualization behavior in the diagram as produced by Saft, but there's also, of course, other researchers including Francis Chiew, Albert Vandyke, who had really investigated this topic quite thoroughly.
New research from Tim Peterson also sort of investigated and concluded that some of the catchments may never recover from drought.
And there's a paper back here, Fowler and others who try to explain multiple explanations for this behavior.
So our research doesn't focus so much on the rainfall runoff relationship itself, but we were interested in how stream flow patterns change itself.
So comparing it during and post-drought conditions, we were specifically interested in where catchments changed and how, and we were specifically interested in stream flow regime changes.

So for example, when we had perennial flow systems or permanent flowing systems that changed to intermittent flows and investigating why those catchments didn't really change back after the drought.
And we wanted to understand factors that were leading to more resilience. So in recovery of those shifts, changes to intermittent flows, or also wanted to investigate why are some of the catchments more vulnerable than others?
And this has really impacts on some of the sort of understanding whether we want to see whether this has impacts on season water resource availability, but also particular with a focus on changes from perennial to intermittent flow, to see the impact on whether it's disruptive to biodiversity and ecosystem functions, but also could potentially have impacts on recreation and cultural water use.
So this brings me straight to our method.
So we had a multiple-step approach. First of all, we wanted to understand these changes in stream flow patterns. So really the where and how.
And for this, we used 160 hydrological reference sites. These are special sorts of data sets which have long and high-quality data records and also are unimpacted by any sort of foreign implications such as water storage releases.

So for these, we really wanted to detect what are the stream flow regime changes and particularly whether, as I mentioned, we see catchments that change from perennial to intermittent flow or also those who have more of an increased intermittence.
The second step followed in terms of the characterization of those changes in stream flow patterns. And here we distinguish between the flow quantities and the relative reduction in flow versus the flow variability where we look into seasonal peaks changes, flow duration and season changes as well as cease to flow characteristics.

We do this all on three different timestamps. So we consider the time period between before the Millenium drought from 1976 to 1997, during the drought is the timestamp from 1998 to 2009, and after the drought.
We consider the time periods between 2010 to 2022. These time periods include some very dry periods during the drought, but then, after the drought-breaking two La Nina years and also in 2021/22, some more wet periods.
The second part of our exploration includes understanding the why, and here we look into various drivers.

of stream flow pattern change.
So first of all, it is sort of the long-term catchment characteristics itself.
And the second one is the rainfall characteristics.
And here we're putting a focus on rainfall intensities.
We also took a look into the water balance itself on annual and seasonal timescales, but also how the drought propagated itself and what are some of the events that could have led to the non-recovery of these stream flow changes.

So I'll go straight to the results, starting with the question, where did we have stream flow changes discovered across Victoria?
And on the left hand side you can see a map which shows different stream flow regime types that we detected and which show different patterns before, during, and after the Millennium droughts.
So in sort of lighter blue, we see those catchments that were perennial before the drought and did not change or shift.
And these are about, more than 50% of the catchments that did not change at all.
We also had perennial catchments that shifted or changed to intermittent flow and that recovered.
And these were about 5% of the 160 HRS catchments.
In about 11% of the cases, we had perennial flow catchments that were perennial before the drought changed to intermittent flow regimes and then did not recover after the drought.
The droughts were persistent in these intermittent flow regimes.
And then secondly, we had intermittent catchments that changed to a more intensification of the cease to flow days.
These were about 4%, but they sort of changed back into the cease to flow days before the drought.
And those catchments who are sort of marked in this dark purple, which have about 26% of the HRS catchments of 160 HRS catchments, were intermittent and they became more intensified and stayed in this sort of intensified flow regimes.

And the characteristics that we associated with some of these unrecovered catchments is quite astonishing, that they're more situated on lower elevations.
And so they are situated on catchments that are on less than 350 meters above sea levels.
They have generally a flatter topography, which means that runoff is possibly dominated by groundwater and surface runoff.
We also see that they have less forested cover, so they have either 50% forest cover or less, which means they have less interception and higher evaporation.
These non-covered catchments are all situated in sub-humid climates, whereas non-covered ones are more humid climates.
They have, compared to the recovered catchments, less rainfall, but similar PET, which means the rainfall runoff coefficient is smaller.
So in general, for these catchments, they have less water availability for runoff generation.

So I come now to question number two: What are some of the characteristics of these stream flow changes?
And here I just want to present a subset of the characteristics that we had a look into.
We want to look into how monthly stream flow peak changed in terms of the peaks itself, but also the seasonality.
We will look into stream flow exceedance and the cease to flow days.

So you can see a representation of the stream flow characteristics and the monthly changes.
And in the following, I mostly focused on those perennial catchments, which switched to the intermittent flow regime and then did not respectively recover.
So you see a representation of monthly flows averaged across these three different stream flow types.
On the left hand side, this is a flow regime that's not changed.
This one here is changed and recovered.
And the one on the bottom is changed and not recovered.
And what we see here is the monthly flows across all these stream flow regime type catchments.
And the thick line here in blue shows the pre-drought condition.

The green one shows the post-drought condition, and then the orange line is the monthly flow during the drought.
The bars across here show the tenth and ninetieth percentile, but for the time being, it's probably just good to focus on the fiftieth percentile.
So what we can see during the Millennium drought is that there is a strong decline in all flows and monthly flows during the cool season flow.
So we can see still that the peak of the Millennium drought is still in September across all these catchment types.
But then this sort of peak flow shifted post-drought to one month earlier.
And we can see here that for the remainder of the cool season flow post-drought, this stream flow did not recover.
And this is really irrespective of all the stream flow types.

What we can also see here is for those catchments who either did not change or those recovered, the early part of the cool season showed a full recovery in stream flow, but this was not the case for those non-recovered catchments, as you can see here.
And the second point to make is also that the cool season rainfall or the stream flow in the later part of the cool season rainfall shows different amplitudes of change.
So, which means that in response to the lower sort of recovery in the second part of the cool season, there's a different amplitude in the non-recovery as well.

So secondly, we had a look into the impact on the flow exceedance.
And again, on the top, we compare the not changed catchments versus the changed ones that recovered.
And then on the bottom, we look into the changed and not recovered catchments.
And what we see here again in blue is the pre-drought condition.

The one in orange is the drought condition, and the green one is sort of the post-drought condition.
And what we can see across all the different stream flow types is that there's a strong shift downwards in all flow components during the drought.
That's impactful on both the low flows and the high flows across all the catchments.
But we can see that the ones who have shifted during the drought to an interrupted or intermittent flow regime, we can see that the impact on the flow is particularly strong.
And in the non-recovered catchment, this impact on the low flow exceedance is much stronger compared to all the other stream flow regime shifts.
So we see about a 20% reduction in the low flows during the drought.
But what we also can see, as marked here in green, these sort of low flows do not recover post-drought.

So really, these processes affecting base flow generation are a key characteristic of these shifts or changes.
We can also now look and further characterize how impactful these low flow conditions were.
And this is by looking into a comparison plot of the different flow regime types.
And this also in this case includes our intermittent catchments.
So on the bottom, you see those perennial catchments that did not have any changes.
And we sort of stratified this sort of heat map in terms of before, during, and post-drought condition.
So again, on the bottom here, we see our perennial catchments that did not show any change in cease to flow conditions.
So they remained as zero.
Then we see our catchments that had intermittent flow during the Millennium drought of about an average of 20 days.
And then these recovered to zero flow days.

And we can see our perennial catchments that did not recover.
So they switched from zero cease to flow days to about 34 days of no flow days, which is even higher than those which were just temporarily intermittent.
And interestingly, post-drought, we see indeed also an increase in those cease to flow days.
If we look into the non-perennial catchments again, we see their pre-drought conditions, they had on average about 30-plus days of non-flows that about doubled and then sort of did not switch back.
And in fact, it also increased in terms of the cease to flow days post-drought.

A similar pattern for those catchments who were non-perennial before the drought but intensified strongly in terms of the intermittence during the drought, but then were able to recover to some degree, but still had, you know, very high days of no flows.

So if we now move into looking into what could potentially cause the recoveries of this behavior of perennial catchments, the behavior of perennial catchments in terms of switching to intermittence and non-recovery, there are three things we looked into more closely.
And this was the influence from the rainfall intensity itself, but also we looked into how the drought propagated in terms of the relationships between rainfall, stream flow, evapotranspiration, and soil moisture.
And we also wanted to go, we have a better understanding of what really happens within the catchments and under the surface.
So we try to model hydrological pathways by using the bureau's operational model.

So now we are going to have a look into the role of rainfall intensity.
And you see two series of plots here.

So on the top we are looking into the annual rainfall intensity again, in our stream flow regime catchment types, those who did not change and remained perennial, those who went from perennial to intermittent flow and recovered post drought and those who changed and did not recover.
And on the bottom, we had a closer look into the cool season rainfall recovery again amongst our three catchment types.
So what we can see here on the top is that all three flow regime types catchments have a similar distribution in rainfall intensities.
We see a little bit in terms of the intensities, we see that the zero to 10 millimeter day in rainfall intensities occur more frequently again, as marked in orange at the orange line here, and here as well, we see a shift during the

drought to less intensive rainfall, but then post drought we see a recovery of all these rainfall intensities as well.
It's across all the catchments at an annual scale, not as intensive as to pre-drought condition, but it becomes relatively close.
So now we can look into the more cool season rainfall, which we know had major shifts during the Millennium drought.
We can investigate whether rainfall intensity during this cool season rainfall had an influence in the recovery of some of the catchments' non-recovery.
So again, here in orange, we see the depletion in, the decline in rainfall intensities and sort of the lower to moderate rainfall intensities.
This is across all the catchments again, in the non-recovered catchments as well.

And in green, we see the post-rainfall intensity conditions, that it's closed back to the baseline.
And this is for all catchments, even though the change in non-recovered ones had sort of stayed below lower rainfall intensities post-drought.
But what we can see in this diagram is also that intensive rainfall across the change and recovered plus non-changed catchments has recovered.
So those really intensive rainfall events, which have about 50 millimeter daily rainfall intensities or more, have recovered in post-drought, but it has not recovered in those catchments who have changed and not recovered.

So they're really missing out on these intensive rainfall events, which, according to papers from

Jen Tolhurst and Peter Van Rensch can also be attributed to less rainfall from fronts and cyclones.
So we now move into what's really causing more event-driven landscape conditions that lead to recovery and non-recovery of those catchment types.
And this looks a bit like a complicated diagram, but what we want to compare is, according to standardized anomalies, express a standard deviation across four major hydrological variables.
So on the top we have the rainfall itself, then we compare that to stream flow, evapotranspiration, and soil moisture.
And we compare this across our three different catchment types.

So in orange we see these catchments that have not changed to intermittent flow regimes and have not switched back to perennial flows.
And in blue, we see contrary those perennial catchments that have not changed at all.
And in green, these are the ones which went from perennial to intermittent flow and then switched back to perennial flows.
So we see in terms of rainfall anomalies, catchments sort of behave relatively similar.
And this sort of beige box here in the middle is a representation of the minimum drought, and the blue box shows the two very wet La Nina years post-drought.
And this box here is a zoomed-in version of these two very wet years.
And we can see, as I said, the fluctuation between the different flow in terms of their anomalies are not that particularly high.
Sometimes they show a little bit more anomalous variation as expected.
But when we come to the post-drought recovery and look into these very two wet La Nina years, there are some subtle but very meaningful differences.

So what we can see here is during the first year of this very wet La Nina year in 2010/11, we see that our non-covered catchments have slightly more rainfall compared to the other two catchment types.
But this switches when we look into the second La Nina year, when we have suddenly less rainfall in those catchments.
But associated with these very wetter conditions in the first La Nina year, we also see that there was higher evapotranspiration, and this higher transpiration led to less soil moisture stored in these changed catchments.
And also, following from this, we see less stream flow generating from these catchments.
We wanted to have a deeper look into what could cause some of these catchment processes that could be impacted by these dynamics.
And for this, I wanted to understand whether there's any change to flow contributions from different flow generation mechanisms.
And we used the AWRA, the Bureau of Meteorology's AWRA-L model, which is an operational model that operates on a five-kilometre landscape at a continental scale.

So the AWRA models hydrological processes at these five-kilometre scales.
It partitions rainfall between interception losses and net rainfall. It can deviate rainfall further into saturation excess flow or infiltration excess flow—so surface runoff—but also generates flow from subsurface flow.
And this could be either via the interflow dynamics or base source drainage.
But there's also evapotranspiration from all different levels of soil moistures.
So the model has been intensively evaluated, and there are a lot of resources available if you're interested to hear more about the AWRA-L model.

But for our purposes, we wanted to look into catchments and investigate whether we have a change in flow contributions, which helps us explain some of these recovery/non-recovery behaviors and stream flow.
So we generated, for each of the catchments that we looked into, time series dynamics and looked into anomalies on the contributions of surface runoff, subsurface flow (in green), and whether it's coming from shallow soils marked in the green time series and in orange from deep soil layer contributions, more reflecting base flow contributions.
So here are two catchments in comparison.
On the top, there is one catchment that was changed and did recover.
And the second one is one catchment that changed and did not recover.
So we can see, when we just look into the deep soil and base flow contribution, that we have different reaction times towards the groundwater depletion.
But in terms of our non-recovered catchment example, we can see that the groundwater depletion is beyond the Millennium drought, and even these two very wet years did not recover the groundwater store enough, whereas that was the case for those in this catchment, which recovered.
But what we also can interestingly see is that very intensive rainfall initially in the first phase of the La Nina years sort of led to surface runoff in this catchment, where it wasn't as much the case in this catchment.
So more water in this catchment was really used to replenish the soil moisture store, whereas that wasn't really the case in this example.
As I mentioned, the first initial phase is sort of generating a lot of surface runoff, which is then bypassing the soil profile and not leading to the replenishment of the catchment's water store itself.
So, this is one of the mechanisms which we thought—or the model helped us hypothesize—what is going on.
But to come to my summary and conclusion now:
What we saw in response to the Millennium drought, and even though we were having these two very wet years in 2010/11 and 2022, we see a persistence of those flow regime changes.
The persistence means we have about 11% of flows that have changed from perennial to intermittent flow, and we saw an intensification of those catchments which were intermittent.
So we saw more days of no flows, and the duration of all these no flow days and events was also longer.
And we could also relate this back to non-events in rainfall.

So this is about 26% of the catchments. We saw a disproportionately large reduction in all streamflow percentiles, but particularly low flows were affected, which really supported our thinking that their soil generation processes are affecting and leading to those disproportionately large changes in low flows.

Concerningly, we saw for all catchment an emerging streamflow pattern post-drought that was really irrespective of the streamflow change regime itself. We saw an earlier peak flow, about one month earlier, so a peak shift from September to August. This is really due to an incomplete recovery of rainfall.

Catchments that were vulnerable to those streamflow regime shifts have particular underlying catchment characteristics. They are more arid or sub-humid, so generally have less water available for runoff generation. If there's less rainfall, there’s also less runoff generation. These areas also show more variable flows, with a much higher sensitivity to rainfall and rainfall intensity changes. In fact, we saw what's leading to the non-recovery of those streamflow changes.

Particularly, in areas that shifted from perennial to intermittent, there was a lack of recovery of intense rainfall events, especially those larger than 50 millimeters per day. We saw a lesser ability for soil moisture replenishment, which plays a critical role in runoff generation. In catchments that did shift, we also saw higher spring evapotranspiration that contributed to the early decline in soil moisture stores. Additionally, there were event-driven dynamics. For instance, initially higher rainfall in the first La Niña year was disproportionately reduced by evapotranspiration, which was also supported by higher vapor pressure deficits. The change to less rainfall in the second La Niña also contributed to the non-recovery.

Our modeling of the hydrological flow path gave us a hypothesis of what could have contributed to the non-recovery, particularly the initial high surface runoff dominance bypassing the soil profiles, which can be explored further.

In summary, these low flows are critical in terms of shifting to a different flow regime. We really need to better understand what's generating those base flow processes, including surface-groundwater interactions. This is something that we couldn’t address fully as part of this research. Overall, the findings suggest we need to be better prepared for streamflow pattern changes, especially with respect to the projected future drying in Victoria.

I would like to acknowledge the funding for this research and also thank the other contributors to this project. Thank you for listening to my talk.

Thank you, Ika, for sharing the latest findings around streamflow pattern changes during and after the millennium drought. Your presentation has given us much to think about, particularly how these changes may affect water resource and waterway management.

For the audience, just a quick reminder to pose your questions using the Q&A function. I can see there are already some questions, so I’ll dive right into it.