Transcript of the new Victorian climate change projections webinar
Heather Field:
Thank you everyone for joining today's webinar. Our first webinar today in our Climate Webinar Series is on the New Victorian Climate Projections. In today's webinar, John will explain what the projections are telling us, and the resources and products available, including regional reports and application ready datasets.
Heather Field:
John Clarke is a research team leader at CSIRO in Melbourne. John has managed the provision of tailored regional climate change projections for a wide range of national and international stakeholders since 2009. He has over 25 years’ experience undertaking, interpreting, applying, and communicating science to a range of clients in the fields of climate change impacts, and adaptation and conservation ecology. I will now pass over to research team leader at the CSIRO, John Clarke.
John Clarke:
Hi, everybody, good afternoon. Thanks very much, Heather. We're talking today about the Victorian Climate Projections 2019 Project, and the results which were released towards the end of last year, essentially this was a Victorian government-funded body of work for us to undertake new high-resolution dynamical downscaling. I'm not going to go into the details of that in this presentation, but for now just sufficient to say what we did was we downscaled Victoria at 50 kilometers, and then at 5 kilometers, necessary two-step process to get down to that fine resolution. But very importantly, and somewhat uniquely, we also bind those new high-resolution results with as much of the already available climate projections data as we could, so that includes the international collection of 40 odd global climate models, as well as some earlier modeling that we did using CCAM, as well as some of the information from New South Wales from the Wicky Project, for those of you who are familiar with that which focused on water resources.
John Clarke:
To do this work, we needed to subdivide the state in order to be able to get a clear picture of what the changes mean in different regions of Victoria, and the high-resolution of these projections allows us to divide Victoria into 10 regions, so these regions are based on the pre-existing Victorian regional partnership regions. In addition to that, we developed a package of guidance material which provides advice on how to choose the most appropriate data to use from what is now quite a significant body of data available for you.
John Clarke:
Finally, we're in the process of arranging for some ongoing user support for people out there who want to use these projections. The projections were releases right at the end of the project, and so it's important that users of the work can also get hold of us in order to get some guidance if they need it, or to get some bespoke datasets constructed. Finally, the Department have also started work on development of an online portal, and we make the data available for download online through the Climate Change in Australia website, but this portal that's being developed at the moment will be a much user-friendly version, and will have some really nice visualizations of the data online.
John Clarke:
In terms of the products that we've produced out of this work, there's a technical report, this is as the name suggests, technical in nature, and it's intended for readers with a science degree. It doesn't have to be a climate degree, but that's the level at which it's pitched. It covers things like the methodology, and the results, and there's a chapter on how to use the projections and its material. In addition to that, we had regional reports for each of the 10 regions, they're written in a much more readable format, designed for more general readers. In addition to that, there's a set of fact sheets, there are six of those that cover a range of topics from what do we need to do, and other set of projections of Victoria, right through to a set of frequently asked questions. All of this, as I mentioned is made available online, and you can download from those webpages a set of application-ready datasets at 5 kilometer resolution. These are available gridded at that 5 kilometer scale across the whole state, and also for a selection of towns, two towns from each region.
John Clarke:
In addition to that, you can make the uncorrected continuous output from the high-resolution model, that's very much for specialist users, and I'll come back to that towards the end. In addition to that, there are the regionally averaged datasets, and for a lot of work this will be sufficient, particularly if you're embarking on a first-pass vulnerability assessment for your region. Also, there's a set of data at 5 kilometer grid, where we've computed thresholds, and these are for temperature, where you can look for thresholds for example days per year above 35, days per year above 40. There's a range of different thresholds here, and we also have thresholds for minimum temperature so you can explore projected changes in overnight minimum temperatures as well. As I mentioned earlier, there's a suite of guidance material, I mentioned that that's in the technical report, it's also on the webpages in Climate Change in Australia, which you can download the data, and where you will soon be able to download a set of slide packs, including this one that we're going through today.
John Clarke:
These are the regions that I mentioned, and if you've used or if you're familiar with the Regional Partnerships regions they should look quite familiar to you. We're just going to quickly look at this next shot of some results for the state now. This map is showing what we could describe as the range of plausible change. There are many different ways you can look at projected changes, and this is just one where we focused on range of plausible change, because that is something that's often used in impact assessments. What we've done here is we've combined the modeling results from two emission scenarios that were downscaled for this work. That's a medium emission scenario with RCP4.5, if you're familiar with that terminology, and also a high emission scenario called RCP8.5 The high emission scenario is pretty much what we had been tracking over the last 10 or 20 years, in more recent years there's been a bit of dip below that, but you get the idea, it's a fairly high emission's trajectory. The RCP4.5 emission scenario, the scenario that requires a moderate amount of greenhouse gas mitigation, and declining much lower towards the end of the century.
John Clarke:
What we're showing here are the projected changes by mid-century in summertime maximum temperature. These are changes relative to the 1986 to 2005 baseline. We used that baseline because that's the baseline that was used in the last IPCC Assessment Report, the fifth assessment report, it was also the baseline that was used for the most recent Australian projections, the Climate Change in Australia work. You can see that there's a range across the state, but it's not an enormous range, if you look at Mallee region for example, the high emissions upper limit of projected change is around 2.9°C. Now that's an average value for that whole region. The same region, the lowest value, the lowest plausible value for medium emissions is 1.2°C. You can get an idea that what we're showing you here is the least amount of change you might expect by mid-century, right up to the highest amount of change you might expect in mid-century, irrespective of which of those two emission scenarios we're following.
John Clarke:
The small labels here just show us what the average summer maximum temperature for a location in each region was over that period, 1986 to 2005. You can see from Mildura that average value is 31.5, so you can add that on to this upper limit, and work out by mid-century if we do follow the high emissions pathway, the upper limit of that will be averaging out around 34.4°C. You can see that there are a range of values for those other regions as well. Hopefully, by now you've zoomed in on the ... focused in on the area of your interest, and I'll move on to the next slide.
John Clarke:
This gives us a representation of the six simulations that we did at high-resolution. The reason for showing you this is because it's important that you understand that there is a range of projected change. Every model gives a slightly different result, and you can see that captured here. We only had time to do six simulations, so the way this works is you take a global climate model, and you take the results from that global model and feed them into the CCAM high-resolution model, and the CCAM model then does its own simulation of the region that you're interested in, but at the higher resolution. That's what these labels reflect, that's the name of the global host model, so global climate model, and CCAM is indicating that it was the CCAM downscaling method.
John Clarke:
You can see we've downscaled from ACCESS 1.0, which is one of the CSIRO Bureau of Meteorology models, downscaled CNRM-CM-5, GFDL-ESM2M, HadGEM2-C, MIROC5, and NorESM1-M. Those six models were selected because as global climate models they give a good representation of the range of projected change across Australia, across all seasons for temperature, rainfall, and wind. They are all models that perform well in the comprehensive analysis that was done for the Climate Change in Australia work. One of the important things to note here, is apart from the fact that every simulation gives a different result, you can see there's also quite a lot of spatial patterning in this, hitting greater amounts of change in some areas, and lesser amounts of change in others. Importantly, this work has given us what we describe as a plausible hotter case. Later on you'll see that this hot case here simulated by the HadGEM2-CC version of CCAM is giving greater increases in maximum temperature than we've seen before from any of the modeling work that's been done previously.
John Clarke:
We switch now to rainfall, focused in here are in winter rainfall because that's a relevant season to look at given that we've observed a decline in cool season rainfall in Southeastern Australia generally, and particularly in Victoria. We're showing here winter rainfall change by mid-century, in this case I'm only showing the high emissions results. It doesn't make a lot of sense to combine the two RCPs with the rainfall in this way, so what we're showing here now is ... if you look at the case for Gippsland showing the 10th percentile change. So 10% of the results fall below that value, and the 90th percentile result, where 10% of the results fall above that value, and the small number in the middle is median or the 50th percentile. So half of the model results fall above that, half of the model results fall below it. Remembering that these values represent the average across this whole region, and see that for Gippsland all of the simulations show a decline in rainfall, ranging from a decline of 3% to a decline of 4%.
John Clarke:
For Bairnsdale, you can see that over the same season the average rainfall is 146.2 millimeters in winter during the period of '81 to 2005. You can see that there's a bit of a different story across different parts of the state, and in the Mallee for example there is one simulation showing an increase in rainfall, but the remainder showing a decrease, and that's quite a strong decrease, 23% decline in rainfall. Whilst there's a range, there's a pretty consistent story that for winter rainfall we expect to see a decline. If we look at other seasons, the story is sometimes less clear, but certainly it's a strong signal in winter.
John Clarke:
Now, this is similar to the slide I showed before for maximum temperature, except in this case we're looking at rainfall, and it's winter rainfall. You can see again, that there's a range of different results from each of the six simulations into the range, and the averages that we just saw on the map. It's the same models obviously, and we're also seeing the same model that was a hot model temperature is also showing us the driest future, that's generally the case. These models are usually also the driest. Importantly though, if we zoom into the Ovens Murray region, this is an important new finding from this downscaling work, is that we see here a range of change on the windward slopes of the ranges. These are the Northwest-facing slopes of the ranges, we're seeing a much more enhanced drying signal than we've seen before. We looked very closely at this result, as we did the result for the large increase in maximum temperature.
John Clarke:
We looked into the workings of the model, and make sure that we trusted the result, that it was stimulating the relevant climate processes accurately or well plausibly, and we reached the conclusion that both of these new results are plausible. We considered this as an important new finding, and we've published this work in international literature. Basically the take-home message is from this is that for the alpine regions of Victoria we're seeing enhanced drying on these Western slopes, which you can see just across the border. This is the Gippsland region down here, you can see that there's much less drying, or in some cases little change in this other side of the range. This makes sense when we look at the convection systems that are at work in these topographic areas.
John Clarke:
This is one of the main reasons for doing high-resolution modeling, it gives us the potential to produce new and interesting information, Because it operates at a finer scale, and is able to better simulate some of those important processes that happen at small scales like convection. But with all downscaling work, you can't guarantee that it will give you new and important information, you need to do a systematic evaluation of those results to make sure that you have confidence in what they say. That's what we did, and that gives us significant confidence that this drying result is a real thing.
John Clarke:
Now, I'm going to just quickly flick through some temperature and rainfall results for the regions. First up, I'm going to look at this plot just to give you an explanation of how this works. What we're showing here is ... sorry I've got a thing in my way that I can't get rid of. What we're showing here are the results from the high-resolution modeling alongside some results from global climate models, and this is for temperature change, and in this case we're looking at, I think this is actually Loddon Campaspe, but I've got the Melbourne map showing. I'll check that. Anyway, what we have ... if we look at this top plot, these pinkish bars show us the range of projected change from the six high-resolution modeling simulations for four future time periods into the 21st century, 2030, 2050, 2070, and 2090.
John Clarke:
This top plot is showing us the results for the medium emissions scenario, that's RCP4.5, and at the bottom there's another bar, and it's burgundy color, burgundy on my screen, that shows the range of projected change from all of the global climate models for 2090. That's there so that we can compare results from the global models with at least the 2090 results from CCAM simulations. The things to look for here are whether there is anything important to note from these two sets of data. You can see I've highlighted that there is a hotter case, 2090 under medium emissions for this region from the CCAM results. Substantially hotter than the global climate model results. Also, notable is the fact that these global climate models simulate a plausible cooler case than any of the CCAM simulations for 2090, likewise down below, this bottom chart shows the same, but for the high emissions scenario, RCP8.5, you see that it's showing that there are both hotter, and cooler cases simulated by CCAM, but the range from the GTMs is also shown for reference.
John Clarke:
These plots are used in the regional reports, so if you haven't yet got hold of the copy of the regional reports I encourage you to do so, this is the kind of thing that you can ... Sorry, these plots are taken from those regional reports. Over here on the left though, I'm showing you an example of a spreadsheet that you can download from Climate Change in Australia VCP 19 webpages, and these contain the underlying results for all of this shown in the plots. I'll explain that in a bit more detail later.
John Clarke:
Okay, so I've just double-checked, these are definitely the results for Greater Melbourne, and these are the same, this is the same information that's showing for Loddon Campaspe region. I did intend to delete that Greater Melbourne slide, but I forgot. The reason I wanted to show you this one is because this is the region in which I live, but anyway I've explained all of that. There is a difference here though, you can see that whilst there's a hotter case ... sorry, I've forgotten my line of thought. You can ignore what I was about to say.
John Clarke:
All right, so I'll move on to the next one, which is showing us the rainfall projected changes. This one is for Central Highlands, sorry, for Loddon Campaspe, and again you can see that the pattern is similar, blue bars represent the projected change from the six CCAM simulations for 2030, 2050, 2070, 2090, just as with the temperature plots, only in this case we don't show global climate model results for medium emissions, instead because rainfall is so variable we have combined results from not just the global climate models, but also every other piece of data we could get hold of, including the Victorian Climate Initiative Projections, but also the New South Wales NARCliM results in some earlier CCAM modeling. All of those data sources are represented by this green bar, which we only show for the high emissions scenario, probably because of data availability.
John Clarke:
Again, what I'm pointing out here are the important things to look for, particularly important if you wanted to evaluate the wettest plausible future climate, for the end of the century it shows us that it's the CCAM simulations that give us that plausible wetter case is significantly wetter, than all of the other results combined. But in terms of drying, well, this region, there's a plausible case that it's slightly drier on the other data sources than it's simulated by CCAM, but it's a fairly small difference. Again, you can obtain the detailed results from the spreadsheets that you can download.
John Clarke:
Now I'm just going to run through fairly quickly the temperature results on the left, rainfall results on the right for each of the regions. I won't dwell on this one because we just looked at it, these are the changes for Barwon, and as I flicked through these you can see that there is a variation in these ranges across state. Another important thing to note with these plots, is here you can see how at 2070 the medium emission's rainfall results have a much smaller range than for 2050, and also for 2030, and that's a bit counterintuitive. Normally we expect the range of change to expand as you go further into the future, and if we had the global climate model results alongside these results that's what you would see. This occurs because there's only six simulations in these plots, that means that these range of change, which is from the 10th to the 90th percentile is very sensitive to just the results from a single model out of that six model ensemble.
John Clarke:
If you look down here for example you can see that there's one model, this dots represent the results from one simulation. You see, there are six dots, but one of them sits right out here, so if by chance one of those six models in 2070 had fallen out here the range would be much more like what we'd expect to see. The reason it's not is because the models are doing a good job of simulating natural variability, and by virtue of that natural variability, which actually occurs at random over time in future climate simulations. Just by chance it happened that all of the results are clustered together around 2070. If we ran all of those simulations again with different parameters we would find that those results would look different, that doesn't mean they're not plausible. They're still all plausible results. To finish off, just flicking through the rest of these results, this is for Great South Coast, this is for the Mallee Region, this is for Ovens Murray, which we zoomed in on before with the gridded results on the map, and Wimmera Southern Mallee.
John Clarke:
Now, hopefully you got a sense then that's consistent with what I showed on the map earlier, that essentially the rainfall change for all of Victoria, it's for a reduction in a rainfall. These results are an annual value though, an annual average so when we pull this apart by season we would see that there is some complexity there that is not reflected in these results. Those seasonal results are available through those spreadsheets.
John Clarke:
One more thing that I'll mention before we move on to how to get hold of data is that ... not as part of the work that we did with this high-resolution modeling, but as part of a sister project of the National Environmental Science Program, some more detailed analysis of trends and projections for fire danger have been undertaken. We've assessed those results for Victoria, and we present a brief summary of those results in the regional reports. Essentially we know that fire weather has become more dangerous since the mid-1900s, this is very well-documented, equally well-documented is the fact that fire seasons have become longer, and starting earlier, and it's somewhat ironic really that these results came out just before the most severe fire season we've ever seen. It's particularly frustrating that we've been, the climate science community has been projecting these changes in fire danger for a very long time, somewhat disheartening to see it actually come to pass, and unfortunately these trends are expected to continue.
John Clarke:
Here, just as an example of the type of results that we provided in the regional reports gives an example for Bairnsdale, and these words are lifted straight out of the regional report says, "The number of days when forest fire danger index is greater than the 95th percentile for the period 1986 to 2005, that number of days is predicted to increase by a median value, 5.8 days per year by mid-century under high emissions." So that's a 32% increase, some model results give a higher increase than that, some model results give a lower increase, the median increase is 32%. What that means is when we look at the forest fire danger index data from the period 1986 to 2005 we took the worst 5%, and then we looked at how that is expected to change in the future. The numbers differ across the different regions of Victoria, but the fundamental story is the same, so we're seeing increases in those, in the most dangerous fire weather across all of the 10 regions.
John Clarke:
So moving on to how to get a hold of the datasets. There are a range of different datasets available according to the level of detail you need, and to an extent the level of expertise you have in working with these datasets. Apart from the plots that we provided in the regional reports, the easiest results to use are these regional average change summaries. These are available for each of the 10 regions, and they provide data on 11 different climate variables. They show the results from the high-resolution modeling, as well as from the global climate model results. The results are presented as a median, 10th to 90th percentile, that's exactly the same as we've been looking at in the bar plots, only in this case, in addition to the annual results, which is what is shown in the bar plots, you can also access the seasonal changes for standard temperate seasons. These data are available for four time periods, and with two emissions scenarios, medium and high, I mentioned earlier.
John Clarke:
Again, this is what these spreadsheet data looked like. You can see that it's a bunch of information about maximum daily temperature. It tells you that the codename for maximum daily temperature is tasmax, it identifies which season the results relate to, and it gives the median, lower, and the upper value, so that's the ... lower is the 10th percentile, the upper is the 90th percentile for this time period in emissions scenario combination. This one's early in the century under medium emissions, this one's early in the century under high emissions, mid-century, later in the century, and scroll across to the right, and see that there's 2090 as well. Important to note also that down the bottom you can see that there are a number of tabs, so these are different sheets in the spreadsheet, the one that we've been looking at so far is the summary sheet, which brings together information from the CCAM changes, along with the regionally averaged changes. If you wanted to look at the regionally averaged global climate model results you just go to that sheet. If you want to look at just the CCAM results. You go to that sheet.
John Clarke:
The summary sheet shows you the results from the CCAM downscaling to the high-resolution work, but it indicates if there is something you need to look at from the global climate model results. In this case we've got lower results for annual daily maximum temperature, and it's got these carets symbol, that tells us that there is something, some information in the global climate model results that we need to consider if we're interested in this lower value. Essentially that will mean that the global climate model give a lower result than CCAM results, and similarly with the upper values, if the global climate model give a value that's higher than the CCAM simulations, they'll be a symbol to indicate that. You can go to the GCM sheet and look up what that value is. In addition to the summary, and the two individual sheets on CCAM and GCM Changes, there's also a Read Me sheet, which provides you some explanation.
John Clarke:
These are the variables for which data are available through these spreadsheets. These are all available for download now, except for potential evapotranspiration results, which we're still working on, available soon. So that's how to get hold of those regionally averaged results. If however you want gridded datasets then there's another set of results that you can download from the Climate Change in Australia VCP 19 webpages, and these represent the gridded change since the 1990s on this 5 kilometer grid. They present the results as what we described as climatological averages, in other words they might be a value for the average January maximum temperature, or the annual winter rainfall, things like that. The results are only available as individual model results, so there are six different sets of results for each variable. There are data on the annual changes, seasonal changes, and the monthly changes, available for the same four time periods as we've been talking about previously, in the same two emission scenarios. There's a sample of one of the datasets.
John Clarke:
On the bottom left, these are the data that I used to produce those changed plots that I showed you earlier the maps with the gridded changes. The important thing about this data is that they are bias-free, important to know that any climate modeling result will contain a bias of some description. That means that for example a given model might be slightly too hot compared to the real world, or it might be slightly too wet compared to the real world, but those biases are generally consistent throughout the entire run of the climate model.
John Clarke:
A simple way to remove that bias is to compute what's called the change factor or a delta value, you might have heard those terms. Essentially what this means is that for each model you compute an average over 20 years from that model's version of the 20th century, and compare that to a 20-year average value from the future period that you're interested in, in the future projection simulations. Then you just subtract those two values, and what you're left with is the change value which is free of any bias. You can then apply that change to actual observations that you have, or that you can get from the Bureau of Meteorology, to produce a plausible future set of data in absolute terms.
John Clarke:
Second, and the next most complicated set of data that we have available, and what we've called application-ready gridded datasets. Now these are datasets where we have done exactly what I just described, we've taken the change values from CCAM simulations, and applied them to real world observations from the 20th century. In this case we've done it in a fairly sophisticated manner, where we've done what's called a quantile-quantile matching. That means that for example if you've got a day of rainfall in the historic dataset that represents, say the 99th percentile rainfall, so one of the wettest days in the historic datasets, we would then scale that according to the 99th percentile change value from the projections. That means in practice that the extreme occurrences, extremely wet days, extremely hot days is scaled by the corresponding change in that extreme event in the future dataset. The result of that is that we've produced a 5 kilometer gridded set of data, and at each of those grid points you can obtain what look like, data that looked like real weather for a 30-year time series.
John Clarke:
These are pretty useful datasets. They have the actual future values that you could plug into a model, for example, that requires a daily input of temperature or rainfall, or whatever. Again, this data is available for individual models, so you have to have a method for deciding which models you're going to use if you're going to use them all. These time series datasets are available at daily, seasonal, and monthly, as well as annual, and these are available for four overlapping time periods. The 30-year time series for 2030 overlaps with the 30-year time series for 2050, but it's important to understand that these data are not a continuous time series for the 21st century, to obtain that you have to use a much more sophisticated analysis, which we haven't done.
John Clarke:
These files are quite large, they're about 1.2 gigabytes each, and this data, as with the gridded change data are only available as netCDF files, but there are some conversion tools around that you can also bring netCDF files into ArcGIS, and tools like that, and ... data format that's extremely efficient for climate data. So if you're going to work a lot with climate data, it's a format that's recommended you use. An equivalent dataset is also available for a selection of towns, two towns from each region, and these are essentially the data extracted from a single grid point of that 30-year ... of that gridded dataset.
John Clarke:
Finally, the most complicated datasets that you can obtain are what could be described as raw CCAM output, and you can download these. You have to be aware that these contain the inherent model bias that I mentioned earlier, but if you have the ability, and the technology to handle that, and to apply some form of bias correction, these are an extremely valuable dataset, largely because of the number of variables to which data are available. There's more than a hundred variables of monthly data. There are 13 variables daily time step. There's also the historical run from the CCAM as well. These are continuous, unlike those time series I mentioned before, which are not continuous. Throughout the 21st century this dataset is a continuous run from 2006 to 2099, and that's really important for some applications.
John Clarke:
Finally, another way of looking at these data is to use the Climate Futures Toolkit, which is a tool that's on climate change in Australia. If you look at one of the VCP 19 Regions, and this case we've chosen Goulburn as the region, generate a matrix that shows you where the results from the CCAM data, as well as all the other available data that we have, all in terms of projected change in rainfall and temperature, or any other two variables that you are interested in. Here we're showing an example of what would be described as the maximum consensus future, that's where there's the greatest number of models in agreement, in this case it's a warmer, drier, future, and it contains both GCM and the new CCAM results. You can see in the far lane here it says, CCAM, and that number R3355, that is VCP 19 data. These others are global climate models because they don't have the name of the downscaling model.
John Clarke:
There's also an example here of the hottest-driest future that's simulated by these collection of models, and in this case there's only a global climate model in that future. So if it was really important for you to understand the plausible hottest-driest future, this is an important one, and essentially unless you've got a really good reason not to, you should use these global climate model results to represent that. That's an example of one of the ways in which you can use tools to determine which of the available datasets is most appropriate. Obviously if you absolutely have to have data at 5 kilometer resolution, and the global climate model won't meet that requirement, so you would have to find the hottest-driest of the new CCAM simulations that are available.
John Clarke:
Finally, if you are going to download some of these data you need to go through this thing that's called a THREDDS Server, and it's actually ... if you've not used these before it's a bit daunting, if you have, they always work like this so it's a standard way of obtaining datasets. Essentially from the landing page on the CCIA VCP 19 page, you click on the datasets, that takes you to a page like this, which list the variables that are available for a particular set of data that you've chosen. Then chose the time period that you're interested in, and when you want to download a particular model you get presented with these options. If you just want to download the netCDF file you use this HTTP server version, that's a link that will just trigger your browser's download function. If you wanted to though, you can subset the results, and say cut out a smaller region, and download that using this netCDF subset version.
John Clarke:
The other versions are quite specialized, and you're probably not going to use those unless you're planning to use a web service to directly obtain data from this THREDDS Server automatically. When you download these data they have a very long and clunky name, and there's a description of the filenaming protocol to use, so you can see what all the terms mean.
John Clarke:
Finally, there's a range of other places you can get the data that are sometimes referred to in either the technical report, the regional reports, or through the Climate Futures Tool. You can obtain, for example data from the NARCliM work, which is at 10 kilometers, and that covers Victoria. You can obtain data from the VicCI Project, and as I mentioned earlier that has a focus on water availability. Also, coming soon are some more water-focused results from the new VicWACI Project, which is the second iteration of the VicCI Project. Also, coming fairly soon are some National Hydrological projections that are being done by the Bureau of Meteorology, and you can obtain a bunch of data of course from the Climate Change in Australia website, including VCP 19 results.
John Clarke:
For those of you who don't know how to import a netCDF file into ArcGIS, we have an explanation of how to do that. The FAQ is on the Climate Change in Australia website, and that's a link directly to that page. Don't forget that there's a bunch of guidance material also available to help you work your way through these datasets. Here is the URL up here if you want to obtain anything, any of the products out of the VCP 19 work just go to climatechangeinaustralia.gov.au/vcp19, and you'll land on this page. From there you can download publications, the datasets, and read the online guidance material. Thank you very much.
Heather Field:
Thank you, John. That was a fantastic, and very thought-provoking presentation and lots of resources there for everyone to follow up should they wish to. We do have a few questions come in, and just out of interest we did have close to 300 registered for today's webinar, and about 170 participants online live this afternoon, which is great.
John Clarke:
Excellent.
Heather Field:
Lots of interest out there. So we do have a question from Sky, she wants to know why is Gippsland not split in to smaller areas given that the East is quite different to the West?
John Clarke:
That's a really good question. It's an interesting thing. Many of you would have encountered this, it doesn't matter how you decide to divide an area up, there is always going to be some problems, and that includes Sky's point, but it also includes what if your town is right on the boundary in one of these regions. No matter how you slice and dice an area there are going to be issues, and you always have to strike a compromise between the level of detail that's important, and the amount of effort that it would take to get to that level of detail.
John Clarke:
So right at the start of the project, we conducted a stakeholder consultation session, and one of the things we asked people was how they wanted the state to be divided up. This is where we landed as a result of that consultation. We could have, of course done any number of different ways, so it came out of the stakeholder consultation process. Another thing that's important to remember though, is that what's critical for climate projections isn't so much the current climate patterns as a pattern of change in the future. The pattern of change across Gippsland, it has some spatial variability as you can see in some of the plots that I showed earlier, but by and large the change signal is pretty consistent across that whole area. What that means is if you've got East Gippsland which is wetter than West Gippsland, then whatever the projected change in rainfall is for that area it's going to affect the East and the West in the same way. They're both going to get drier, it's just that East Gippsland will stay wetter than West Gippsland, if that makes sense.
John Clarke:
The underlying patterns in the climate are maintained, it's just that these changes can be reasonably uniform, but of course if you dig in to more detail then you start to see that there are variations. If you need to work at that level of detail then you probably need to use the gridded datasets rather than the regionally averaged ones.
Heather Field:
Thanks, John. Just a reminder to everyone, if you do have a question, and you want to ask that of John, you can write that in the chat box or raise your hand, and I can unmute you, and you can ask your question yourself. We do have another question from Bronwyn, the data points in rainfall are useful for understanding the impacts of the models on the range, was this presentation considered for temperature?
John Clarke:
We sort of decided ... Yes, we did think about it, short answer. We didn't ... Just in the interest of reducing the clutter on the plots, there's no other reason really. You might remember from those plots, I'll bring one up, the range of change in temperature is generally much less than in rainfall. So if you compare, say this is for Loddon Campaspe range of change in temperature, compared to the range of change in rainfall, of course rainfall is more difficult to simulate, and because it's subject to a lot more random processes than temperature, you always get a bigger range in rainfall.
John Clarke:
In addition to that, you really see that in the high emissions plot down the bottom here, you get some projections for increased rainfall, and some projections for decreased rainfall because of that. It creates a lot of ... It makes it a lot harder for people to deal with. When presented with this type of information people are often just bamboozled, and think, "Well, we can't really make any decisions using that information." But by putting in the results with dots showing where each of the six simulations falls within that range, you can see that there's only one model that gives that increased rainfall result for this particular region. It allows you to make an informed judgment that, well, if you take that on balance then you can see that five of the six models are showing a drying signal. So you probably give more emphasis to the drying signal than to the wetting signal if you like. Even though, we would always say that you should acknowledge the fact that there is this wet future that is plausible. As far as we can tell there's no reason to discount that particular wet result.
John Clarke:
It's just more difficult working with the rainfall results so we felt that adding the dots, even though it clutters up the plot a bit, it provide some really valuable information. Whereas with the temperature, all of the changes are in the same direction, and the range is really just due to the variation, and the internal variability that's going on within each model, so we think it's less important. Others have done it differently, if you go online and look at the Queensland high-resolution projections that have been released in the last year or so, they show you where each model sits for both temperature and rainfall variables. It does clutter up the plot, but it's also interesting information, but that's the reason why we chose to do it that way, this work.
Heather Field:
Thanks, John. We are close to time, but I have got a couple more questions if you've got the time, John.
John Clarke:
Sure.
Heather Field:
From Ross, which organization is working this modeling through to the implications for water availability, by when, and will conclusions be ready, or when will conclusions be ready?
John Clarke:
I can't answer the last question very well, or the last part of the question very well, I'm not quite sure when those results will be ready, but there's a number of things going on. There's the National Hydrological Projections, which are being done by the Bureau of Meteorology, and they are using CCAM results in their work, but not only CCAM results. The VicWACI Project, which is underway are using a statistical method to work out the projections at high-resolution, but they are also taking these VCP 19 results, and working those through their models as well. So that will provide a direct comparison between the statistical methods, and our CCAM dynamical downscaling methods. They are particularly focused on hydrology, so they will be doing things like streamflow, soil moisture, those kinds of things, some of which if you're a sophisticated user of climate model data you can obtain data on that directly from the CCAM output.
John Clarke:
But it's not a dataset we've made available in any of those value-added ways, mainly because we were aware that this new hydrological work is being done by DELWP, and in conjunction with CSIRO's Hydrologist Team. They are the people that I am aware of doing that work in Victoria. I think the VicWACI data will be coming out this year, but don't hold me to that, I'm not really sure.
Heather Field:
Thanks, John. We've got Roger with his hand up to ask a question, so I'll try and unmute you, Roger, and if you can just say who you are, and where you're from, and ask your question. Now, go ahead, Roger. I don't think we can hear Roger. You may want to ask your question in the chat box, we can't hear you. I'll just move on to another question from Sharoma, "A very informative presentation, do you recommend using one model output when using application-ready data?"
John Clarke:
A really good question, thanks, Sharoma. Generally, but it depends what you're doing. I kind of hinted earlier that if you're just doing, I shouldn't say just, but very often when people are trying to work out their exposure to climate change-induced risks you start by looking at all of your areas of interests across an area. It might be the Barwon Region for example, and you can just use those regionally averaged values that are in the plots, even just rough values will be enough to help you do a first pass assessment of your vulnerability. You can work out things like, "Okay, well, summertime's temperature is projected to increase by up to this much, what does that mean for our hospital systems? What does it mean for our energy bills?" Those sorts of things.
John Clarke:
But if you need that high-level of detail, we recommend people use some method to subset the results into particular cases important to the decisions you're making. We've developed the Climate Futures Toolkit to do that, where you can look at projected changes as described by the changes in two variables, and see where all of the different model results fall in terms of those two variables. You can end up with things like, say a much hotter, a much drier future which you might describe as a worst case, say if you've got an agriculture application that would be a worst case situation.
John Clarke:
You should also look at the best case, again an agricultural example it might be wetter, and least hot future. Then, with those two futures you could use the average values from all of the models that fall within each of those two values. A hot, dry future might have four models in it, or your least hot, wettest model might have five models in it. You could use the average results from those, or if you need to plug multiple variables into, let's say continuing the agriculture example, let's say you're running app sim, and you need to simulate the growth of wheat, an area for your worst case and your best case, then you'll need results from single climate model in order that the variables retain their internal consistency. There's a relationship among the variables in the real world, and the simplest example is you don't generally get rain on a cloudless day, that's a really simple example of what I mean by internal consistency.
John Clarke:
If you're going to do that we have another tool in the Climate Futures Toolkit that allows you to identify one model that is the most representative of a given climate future, and if you needed the data of 5 kilometer resolution then you just go through that process until you come to one of the CCAM 5 kilometer models, and identify that as your hottest and driest VCP 19 model, or you could do the same for the GCM model if you didn't absolutely need that 5 K dynamically downscaled output.
John Clarke:
So usually we encourage people to consider at least two models, three if you can manage it, or if you could, use all six of these CCAMs that would be even better, using information from those bar plots that I showed earlier to gain an understanding of how well the CCAM results span the range of plausible change. If they don't quite cover the range of change from all of the available results then you might simulate, use another simulation from the global models, or at least just make a note about that in your report. Unless you've got a really good reason, and you can describe that reasoning in terms that will pass peer review, you should use more than one model.
Heather Field:
Thanks, John. Roger has written in his question, and wants to know was the starting point of 1986 due to available datasets or some other constraint?
John Clarke:
Good question. It's a combination of things. I mentioned before that it's a ... we've gone with that time period because it's consistent with the IPCC in their work. When we did the Australian projections, which were released back in 2015, we used the IPCC's baseline period as it's called, and we've carried that through to this VCP 19 work. It's largely defined by ... when the IPCC think about this they're looking for a baseline period that's as recent as possible, but for which they can run simulations sensibly. When ... It's a bit hard to explain this without a diagram. When you look at the future climate runs, for example, they don't have any data before 2010 for the future projections.
John Clarke:
The IPCC takes the approach that they want, use the most recent time period they can, and that's defined by the starting point of future projections in the modeling. Part of this is because it takes around about seven years to do the climate modeling that the international community does. There's always that sort of seven year lag, so that usually translates into a 10-year lag pretty much by the time you round it off. The next set of IPCC results will be for a new time period, they're moving the time period, if I remember correctly forward five years, so that will be '91 to 2010, I think, something like that. It's possible though to compute the projected changes or any time period we want for which there are historic runs available. I sort of implied before that those historic run ended in 2010 currently, that means, sorry ... yeah, 2010, but we can go back to earlier baseline periods if we want to.
John Clarke:
So as part of the VCP 19 work, for example we computed projected changes compared to the pre-industrial baseline, which is the baseline reference period that is relevant to the Paris Agreement targets of a 1.5 and 2 degrees of change. So it's important to understand that the results will be different depending which baseline you use. I probably haven't answered that very well, but hopefully that gives you an idea.
Heather Field:
Right, thanks, John. It's 10 past one, so we might pull it up there. I don't have any other burning questions coming through, so I would love to thank John for his valuable time in his presentation today, and answering a those questions. So if there is any other questions that do go come through, I can collate those if you email those to me, and pass them on to John.
Heather Field:
I just want to remind you all, that there is a survey at the end of the webinar when you close out at the webinar, and we do appreciate you completing that if you can. Also, I just want to let everyone know that we will have a second webinar in our series, and that will be on Monday, the 24th of February. We'll be hearing Dr. Luke Shelley from the Bureau of Meteorology, and he'll be talking about the bureau's new climate guides, so stay tuned for some information that will come out from that. Thanks again, John, and thank you, everyone, for your participation today.
John Clarke:
Thanks, Heather. Thanks, everyone.