Transcript for Hydroclimate projections informed by new models and science video

[ Speaker: Rachel Brown (DEECA)]

Welcome everyone to this webinar being hosted by the Hydrology and Climate Science team at the Department of Energy, Environment and Climate Action, where we're going to showcase some of the research from the Victorian Water and Climate Initiative.

So for those who have been long before you might recognize like previous webinars, we'll be recording this session. So, you might have seen that pop up the intention of being of this is to make the webinar itself available on our website after the event.

For people who don't know me, my name is Rachel Brown. I work in the hydrology and climate science team at DEECA where we manage the Victorian water and climate initiative. Other team members are working hard behind the scenes today to coordinate the session and make sure all the technology runs smoothly. So thanks to Geoff, Sandra, Jacqui and Gil.

So to get started, I'd like to acknowledge the traditional owners on the lands on which we are meeting today. In my case, that's the land of the Wurundjeri people of the Kulin nation, and I'd like to pay my respects to their elders, past and present, and extend that acknowledgement to those across other parts of Victoria and Australia, wherever you might be joining us from today.

The second phase of the Victorian Water and Climate Initiative, or VicWaCI, for short, has been underway for a few years now, and in fact is rapidly approaching it’s conclusion. So we're going to mark that with the release of a synthesis report later in the year and, in parallel, this will be supported by a series of webinars, including this webinar that you're joining us for today.

So I'm delighted to say that David Robertson from the Commonwealth Science, Scientific and Industrial Research Organization, or CSIRO, will talk to us about hydrology projections. Today, he's going to run through the climate models that were used to inform the IPCC assessment reports, termed the Coupled Model Intercomparison project models, or CMIP for short, and talk to us about these from a water perspective.

CMIP 5 models were used to inform the 5th assessment report back in 2015, whereas CMIP 6 models were used for the more recent assessment report a couple of years ago, and they climate projections from CMIP 6 models are becoming available, and so David will provide an update on CSIRO’s investigations into the similarities and differences between CMIP 5 and 6, as well as insights from the application of the new models in relation to water resources.

Our team in DEECA is currently in the process of considering how we might update the projections that we use for water resource applications in Victoria and the work that David runs through today and described in this presentation will help to provide insights and be relevant for how we go about the development of any update to these projections.

So just a bit of admin, by default, you'll see that the webinar is set up with audience microphones and cameras switched off, but I definitely invite the audience to use the Q&A function to post your questions or comments at any time, and we'll put these to David, once he's run through his formal presentation, so just finally I'd like to thank everybody for joining us and hope that you enjoy this webinar and hopefully you’ll join us again in the future ones as well.

So just to get started, it's time for David to pop online if you can.
So I'll just give you a bit of a brief introduction before we get started.
Doctor David Robertson is a principal research scientist in CSIRO’s environment research unit, and he leads the water forecasting team. He has extensive research experience in water forecasting, hydrology and water resources. David's research interest is in understanding hydrological change and using this knowledge to support the development of methods to generate hydrological projections and ensemble hydrological forecasts with lead times from 12 from hours to 12 months.
His research has supported the development of forecasting services used by the Bureau of Meteorology and other operational services tailored to the needs of water managers in Australia.

And obviously he's involved in the VicWaCI research program and will talk to us today about this work on using the CMIP models. Thank you, David. Take it away.
Hopefully you're able to share some slides on your screen. Great, I can see those so over to you.

[ Speaker: Robertson, David (CSIRO) ]


Thanks for the introduction, Rachel. So, I think Rachel's provided a summary of what I'll talk about, which is Hydroclimate projections, and particularly those informed by the new models, the new CMIP 6 projections.

Before I start I just want to acknowledge the other CSIRO VicWaCI team. Who have contributed in various forms to this presentation.

Today what I broadly want to cover is 4 points, the first one is around runoff projections that we've generated that have been informed by global climate models, and particularly contrasting those from the CMIP 5 and CMIP 6 models.

We'll touch a little bit on the dynamical downscaling results. There are increasingly becoming available across Australia. I'll talk about some of the findings from the runoff projections that we've generated and their potential impacts on water security and just highlight some of the new and emerging methods and research that we're doing we've been doing through the VicWaCI program.

So what I'll firstly just talk about is how we go through the process of understanding what the future climate or the climate change impacts on water are. So, when we try and do that, we have a a comprehensive modelling chain that really, I guess, makes use of a few different modelling tools. Specifically, we're using the output of global climate models that have been forced by some sort of projections of future greenhouse gas emissions.

We can then take those and make them look more like rainfall that falls on the ground by using some sort of downscaling method and then we can generate hydrological projections by running those downscaled data through hydrological models.

So just touching on where our projections that we've been looking at have come from. So, we were very reliant on global climate models to produce projections of future climate. Now these models are really representations of the globe and the biophysical processes that occur.

And they're designed to represent the interactions between a whole lot of those biophysical processes.
So those climate models, though, when we're looking at projections of future climates are then driven, or in the jargon forced, by some sort of emission scenarios that describe future trajectories of greenhouse gas emissions. Now they have taken various names in the past. Various acronyms in the (CMIP6) work. They were Representative Concentration Pathways or RCPS and in the current lot of models they’re Shared Socioeconomic Pathways, which describe emission scenarios and projections of future emissions.

And there's a range of those from low emissions futures to high emissions futures and a number of the points in between. So, Global climate models and emission scenarios have been put together and ran forward through the coupled model intercomparison projects, the so-called CMIP models project that have been used to inform the IPCC, The International Panel on Climate Changes Assessment, reports and I've got some dates there about when the CMIP projects finished, or when the final sets of data became available and how they've been used to inform some of those reports.

So the differences between these different sets of projects are really, I guess there's a there's a number of them, but it's how the models used in each project are different.
As the projects have rolled out, I guess they've been representing more of the global biophyscial physical processes that influence our global climate, and they've also, over time, been increasing the resolution of the models. So, the original CMIP5 models, a lot of them were very coarse at two and two and a half degree spatial resolutions across the globe. The current set of CMIP6 models have much finer resolution of the order of one or degree or less.

So within each of these comparison projects, there's been lots of agencies from around the world contributing modelling results. Which creates what's termed an ensemble of convenience. That is, lots of people are contributing their results and the ensemble that's generated is just, you know what those agencies have been contributing.

So what that means is that all of those different agencies are using different models.
So we have a range of projections for each emission scenario that we need to understand. So given that little bit of background context, what are the projections looking like?

So firstly I just want to touch on briefly the rainfall projections and compare the rainfall projections from the CMIP 5 and CMIP 6 GCMS across all of Australia.
So, here we have on the right hand side we have models from the CMIP 6 projections from the CMIP 6 GCMs and CMIP5 GCMs in the middle, and this is for a future period centered on 2060. Relative to a historical period centered on 1990.
We've got 3 rows of data here, one is for annual rainfall, middle row is for summer rainfall and the bottom rows for winter rainfall, and then for each set of projections we have a median projection. So that's roughly the middle projection of the range that from the set of GCMs and then something towards the tails of the distribution at the dry end and at the wet end.

So if we look at the projections of annual rainfall from the two sets of models, we see that they are broadly similar across large parts of Australia. Particularly if we look at the median projection and we're focusing on Victoria, the CMIP5 models were suggesting a median reduction in rainfall of roughly 6%, whereas the CMIP6 ones are roughly 4% reduction.

So we're broadly saying that there are they are similar in their characteristics of projections of annual rainfall. What we do see is that there are more models projecting drier future than a wetter future in terms of annual rainfall, particularly in terms of winter rainfall, where pretty much across all of Australia, the median projection is for a drier future. Even for the wet end models, we can see that parts of southwest Western Australia and also coastal Southern Australia, we're seeing projections that are for predominantly for a drier future.

So this is our rainfall projections for Australia and of mean annual rainfall and seasonal rainfall.

I models are not just projecting though changes in mean, or you know, long term average rainfall we can also go and do some more analysis to understand how extreme rainfall or the variability of rainfall is likely to change and what we're seeing here is that the top panel set of results here are for extreme daily rainfall. So when we say extreme daily rainfall, we're talking about the rainfall that might be exceeded two to three times a year, so the high rainfall events that occur two to three times a year.

So what we're seeing here is in contrast to the average rainfall, we're seeing that the median projection of these extreme daily rainfalls is increasing over large parts of Australia and that is consistent with our understanding of what will happen in a future warmer atmosphere.

The CMIP6 models tend to show larger increases across all of Australia than the CMIP5 models in that extreme daily rainfall. If we look at the variability of rainfall, so this is the variability of the interannual variability of rainfall, what we're also seeing is that after the median projection we're seeing it increase under both CMIP 5 and CMIP 6 projections, but particularly in Southern Australia, we're seeing that the median increase in variability tends to be somewhat a little bit larger than the CMIP5 models. In a a few slides I'll talk a little bit about the consequences of this for water management.

So if we then take those rainfall projections and also projections of future temperatures leading future evapotranspiration and run those through our rainfall runoff models, we can look at changes in mean annual streamflow and we've zoomed into Victoria for this particular piece of analysis. So, here we have again our median and 10th and 90th percentiles, dry, wet and end of the spectrum informed by those CMIP6 GCMS on the top panel and the CMIP5 GCMS on the bottom panel.

So what we can see here is that broadly, the median projection is, or all, projections are broadly similar, that is that the median projection for all of Victoria is for a future with lower mean annual streamflow. For large parts of the South, and particularly southwest Victoria, even under the wetter end there is drying, runoff will be lower.
The future runoff is highly likely to be lower than current levels.

But there are some differences between the CMIP5 and CMIP6 projections and particularly on the dry end we're tending to see that the range of projection, particularly on the dry end is smaller under the CMIP6 models than the CMIP5 models. So that's, I guess, a story about mean annual rainfall. But we're not just interested in that, you know, mean annual rainfall is an important indicator, but not the only indicator we need to think about when managing water resources. We also need to think about droughts and how the likelihood of particular droughts might change in the future.

So when we look back at the historical annual runoff, we can see that there are long wet periods and long dry periods and when we go look forward, we are expecting to still see those wet and dry periods. But against this background of a declining trend in mean rainfall and hence mean runoff. So, with lower mean annual runoff and as I talked about earlier, higher interannual variability in rainfall, what we're expecting is that the hydrological droughts are likely to become more frequent and more severe.
We highlight this, I guess here, with the frequency of a three-year hydrological drought and expecting to see that to occur more frequently under a future climate, particularly in those areas where the mean runoff is declining.

So all of the projections that we've shared so far have been derived from the global climate models. There is a considerable effort at the moment, or it's drawing to a close at the moment, where dynamical downscaling models are being used to provide insights to generate projections at higher spatial resolutions. So, to inform that, there's been a large amount of work done on understanding the strengths and weaknesses of the range of global climate models, to inform some selections of the models that we are to be used for the dynamical downscaling.

Dynamical downscaling is quite computationally intensive, so it's not possible to do it for all of the global climate models. Therefore, there's this evaluation and selection process that's been undertaken.

Given that process, a number of different organisations in Australia have applied different dynamical downscaling methods to the output of global climate models.
So, there's been work done in Queensland, in NSW, here in CSIRO and also in the Bureau to apply these demos dynamical downscaling methods to produce higher resolution projections of climate for Australia.

Then what we've been doing is exploring what the output of those dynamical downscaling models look like. So, here we have a range of projections that have been dynamically downscaled, we have the raw global climate model on the left-hand panel and then the mean annual rainfall projection from the different dynamical downscaling models as different columns in this table in this set of figures.

So what we can see here, and I'm really conscious that there's a lot of detail in this table, but what we can see here is that the different downscaling models often produce different results to the raw GCM. This is because the different downscaling methods that are used to have a range of different assumptions built in. So, sometimes what we're seeing is that the direction of change of future rainfall is different with some of those downscaling models and sometimes the downscaling models actually amplify the signal.

But these models really, can I guess, provide much higher resolution projections and particularly in very local rainfall they can really represent some of the processes that aren't possible in the global models. For instance, the effects of orography on rainfall.

So there's a lot of different ways that these downscaled projections can be generated and I guess when we've gone and then tried to synthesise the message from these downscaled projections is that we're finding that if you combine all of the projections from these different downscaling models that the range is similar to that of the GCM models, individual models with their own assumptions might change the direction or the range of projections. But when you pull them all together, just like we pull all the global climate models together, what we see is that the range of projections is similar to the base GCM ranges.

I now just want to touch on the point I talked about earlier about we're seeing higher inter annual rainfall variability in the GCM projections, so there's a number of reasons why annual rainfall variability is likely to be higher under climate change. You know, it might be related to El Nino, but there's a number of other factors that it can be related to.

So what we know from a water resources perspective is that the reliability of water resources systems is related to the coefficient of variation of rainfall. So, what we've done is some analysis that investigates the relative contributions of changes in mean annual rainfall and changes in the coefficient of variation of rainfall on the reliability of a characteristic storage system.

So we see that the higher coefficient of variation in rainfall leads to a higher coefficient of variation in runoff, and that the higher coefficient of variation in runoff reduces the reliability of water resource systems.

But what we've been trying to show here in this figure on the right and some of the other analysis that we've done is that the increases in the variability of runoff on system reliability is a lot smaller than the impact of the projected changes in
the mean rainfall or mean runoff, so in a way, I guess the conclusion that we're coming to from this piece of analysis is, probably from a water resources reliability perspective we should be a lot more concerned about the magnitude of the projected changes in mean rainfall than the projected changes in rainfall variability.

I'll now just move briefly on to some of the new insights and methods that we've been applying to our analysis of the CMIP6 GCMS. So, what we know from the GCM projections is that it really is only a limited set of projections that we have, we only have a small number of models and a small number of scenarios and that those projections are really trying to identify or characterise multiple sources of uncertainty.

So there's uncertainties in variability in rainfall that's due to, when we look at the projections, that's due to the emission scenarios, there’s variability and uncertainty that's due to the modelled responses to the emission scenarios and there's also a source of variability, the internal variability, the natural variability of rainfall, and depending on how we analyse the projections, we can be placing an emphasis on different sources of uncertainty.

Using different methods for analysis, we can see that we can obtain different ranges for projected future changes. So, the results that I showed so far have been made looking at one future time slice relative to a historical time slice and what that does is it mixes all sources of uncertainty. It mixes the internal variability, the climate response to emission scenarios and the uncertainty due to modelling.

But there are other ways where we can can separate the internal variability signal from the underlying change in climate change signal and the method we’ve been using, we're calling it pattern scaling, where it lets us identify what that climate change signal is relative to and remove the effect on internal climate variability.

So, that's useful if we're trying to understand changes in mean annual rainfall or runoff or projected future changes and reduce the uncertainties and narrow the range of the projections that we're trying to understand. But if we do that, we also need to consider when we're applying those methods.

We need to be able to characterise the internal climate variability, and we've been working on stochastic modelling methods illustrated by the panel on the right, which seek to model both the internal climate variability and climate change, and model those things independently so that we can get better insights into the relative impacts of variability and change on water resources availability.

I'm going to provide a a teaser to a future seminar where we, as a part of the VicWaCI work, we've also been doing a fair bit of analysis on characterising and modeling hydrological nonstationarity, and that will be the topic of our future seminar.

So at this point I just want to wrap up and share I guess some of the key messages that we‘ve been finding. So what we're finding is that the streamflow projections for Victoria, informed by climate change signals by other CMIP5 or CMIP6 GCMs, do tend to be relatively similar. There are small differences, but they are relatively similar.

We're seeing a large number of dynamical dance scaling products becoming available and these do seem to add value over local areas and over orography where rainfall is influenced by topography and so forth. However, what we do know is that the different downscaling methods can be producing different projections and ideally we would try, for practical applications, use all of the projections available from the different downscaling methods.

What we're seeing in the projections is that we're seeing higher interannual rainfall variability, and that can be modelled as leading to higher variability and runoff as well. The high variability will reduce the reliability of water resources systems, but the impact is relatively small compared to the significant reductions in mean stream flow from rainfall reductions and increases in evapotranspiration.

There is research going on, looking at understanding the relative magnitudes of climate change and internal variability on projections. Also we're looking at how these different components and this understanding can be incorporated into water planning.

And finally, I think the hydroclimate projections that are described in the DEECA guidelines to be relevant for assessing climate change impacts on water availability.
But I think some of the insights that we're gaining may provide some updates in the future on those guidelines.

Thank you.

[ Speaker: Rachel Brown (DEECA)]

Thank you, David Umm and it's great to hear about these latest findings about the climate projections and the models.
So, very insightful and I think probably highlights to people on the line that there's a large body of work that goes into, I guess, investigating these models and how they might be useful for providing contemporary projections for the water sector and other applications elsewhere for those people that are interested in their own work.