 Well, thanks everyone for joining us today for our latest Hydrotera webinar. Today we're joined by Dr Paul Fakima from the Bureau of Metrology and it's a bit of a reunion for me. I've known Paul for many years, so thanks for coming along today, Paul, really appreciate it. The topic is monitoring and forecasting catchment hydrology and before we charge into that a few housekeeping matters and I'll give you a bit of the background of Paul. As I mentioned, we're the two presenters, so I'm the managing director of Hydrotera and Paul is team leader of Water Forecasting at the Bureau of Metrology. We love your questions and thanks to those people who have sent us some early bird questions already. We will look forward to answering those questions at the end of this presentation. In terms of raising other questions, you just use your Q&A button on the top of your screen, I think it is, and type away in there and I will read those questions out to Paul at the end and he'll do his best to answer them. Why does Hydrotera undertake these webinars? Well, they're proving very popular and it's great to see so many people coming along today, but we really are passionate about sharing knowledge and today's a really good example of not just sharing our own knowledge but sharing the knowledge of others as well. We do believe in facilitating education. We feel that in the industry at the moment there's probably a bit of a lack of really good relevant hands-on training happening, so we're doing our best to provide some of that through this forum and we're also trying to show a bit of an industry leadership position bringing up topics which seem timely and this one around catchment hydrology, I don't think there's been a bigger phase in Australia's history where catchment hydrology hasn't been more important, whether it's related to floods or to resilience in agriculture. OK, so a little bit about Paul. I first met Paul and we were both doing our masters at Melbourne University studying hydrogeology and we both then also got some funding to join what was then the Cooperative Research Centre for Catchment Hydrology. Paul's taken his passion for all things hydrology and forests on a really long and stellar career once he finished his PhD, which was also done with the CRC and was related to forest hydrology. He then went on and worked for DELP in Victoria on plantation research and then went to Melbourne University as a forest hydrology research fellow. Since then he has moved into the Bureau of Meteorology and currently is team leader responsible for all things water forecasting. So I feel very lucky to have Paul here today. In terms of the things that Paul's going to talk about today, he's going to talk about the services around seven day streamflow forecasts, seasonal streamflow forecasts and new landscape water forecasts. The reason we decided to put this webinar on today is we've been involved with monitoring of catchment projects recently and as part of one of those monitoring projects, we really have had the opportunity with some funding from DPI in New South Wales to investigate what all those resources are out there that can be used as monitoring information around those catchments and I was really impressed with what's available in this sort of spatial context around forecasts and it truly blown away with things like forecasts of planned available water and that sort of thing. So I felt enlightened and I thought Paul, be a good guy to inform us all a bit more broadly. So without further ado, Paul, over to you. Wonderful. Thanks, Richard. And look, thanks for having me along. It's good to rekindle all relationships. So and always, you know, enjoy talking about the Bureau's water forecasting capabilities. So when we go, we'll go to the next slide. So yeah, just for people just to let you know, I look after a team that's responsible for water forecasting, delivering the water forecasting services in the Bureau, been there for nine years, undergone some transformations recently, but at the moment we look after seven day streamflow forecasts and the seasonal streamflow forecasts. Also very new landscape forecasts that were released later last year. And so we'll talk about those as well. So next slide. Thanks, Richard. So that's really the topic for today. So those are the the I guess the services I'll talk about. But before we do that, I'll I'll cover just an introduction to place this into some context next thanks, Richard. So the Bureau covers in terms of its water information and water forecasting services, covers across a lot of temporal and spatial scales. Just click again. Thanks, Richard. And we'll be covering sort of in that this that that in that that red sort of circle there in terms of looking at the from seven days out to three month forecasts. And we'll be looking at both catchment and continental scale forecasts as well. Thanks, Richard. So I guess in terms of our forecasts, you know, when we started the water information program started in the Bureau about 10 or 11 years ago now, the Bureau did provide flood forecast, but we didn't provide any other water information or water forecasting information back then. And it was during the drought that the Bureau became, I guess, a central point in in organising the the country's water information and and providing new services and some of those new services included providing water forecast. So one of the first things I guess we did was understand the potential users or applications of these water forecasts. So over that time, we've we've developed relationships with a number of different managers in different areas and users. So our forecast generally used for for managing storages for helping regulate environmental flows for river managers. They certainly find it useful both in seven day and seasonal areas for farmers and to develop their management and cropping strategies as well. There are recreational users as well that use the forecasts, whether it's boating or fishing or camping along rivers. And we and particularly in the seven day area, those forecasts provide important information to support flood forecasting as well. And we're also working on a project to actually integrate our seven day and our flood forecasting services more deeply. Thanks, Richard. So I guess, you know, one of the things just to give you an idea is a bit of context as to what goes into making a good forecast. We need user needs. Like I just said before, one of the, you know, the first thing we do is understand why we're why we're developing these services and who might be using them to data sharing is a really important one to gathering the data, particularly updating our models and being able to update those forecast regularly is really important. So that's where you all, I guess, come in in terms of monitoring in particular, having access to good data and having access to it quickly is really important. And those systems that facilitate that, the science and systems comes into it as well. So we've worked very closely with Siro, Melbourne Uni, for example, develop systems to allow that data sharing and forecast generation to happen quickly. And then we don't assume that that the forecast will be used straight away. So we need to we need to work with with potential users as well to understand for that to help them understand how the forecast may be useful for them and that it's a continuous loop, I guess. Thanks, Richard. So moving now on to the seven day forecast. There's a catchment based forecast. As I mentioned, we operate at different different spatial scales. Next, thanks, Richard. And so the Bureau has one to go on a process where we will we will continue to provide catchment scale forecast for seven day and seasonal forecast, but we also at the continental scale provide a gridded forecast as well. That being that at the catchment scale, we generally have finer resolution and better performance. So particularly in the cases of flood forecasting and seven day forecasting, we really need that catchment based those catchment based forecast for that better performance within those catchments of interest where they're going to have the biggest impact. So the seven day forecast service started about 2015, upgraded in 2017 or 18 to provide ensemble forecast. That's another big area that the the the Bureau has been investing in rather than providing deterministic or a single forecast out for seven days. We provide an ensemble. So to give give users an indication of the uncertainty and the likelihood of certain river levels and river volumes being reached. So we would we can't provide a service for over 200 locations across the country. They are grouped in 100 or so catchments. So we have a number of we generally have a number of forecast sites within a given catchment. The forecast are updated every morning and they go out for seven days. So on the left hand panel provides an indication of what the what the website looks like of for each of their services, I've provided a link. So you're certainly more than welcome to access that link. And there's lots of other background information as well. Next things Richard. So these are the main products, I guess, that that you'll come across on the website, the top two showing daily and hourly forecast. So within each of those panels, you've got on the left side of each of those panels in the blue, there's the observations and then to the right in red of the forecast on the lower parts of those upper charts. You've got the river or the stream flow volume forecast or information on the top you've got rainfall. So you've got, I guess, both observed and forecast rainfall and stream flow in those plots. And again, it gives you an idea of what the plumes and uncertainty what those uncertainties are around those estimates as well. At the bottom, the accumulated rainfall and stream flow forecast give you an idea of accumulations for the next seven days and the uncertainty around those. And then on the right, we've got information that provides you information on how good these forecasts are. One of the things the Bureau is pretty strong on is verification and testing of performance of these forecasts. So it gives you and we we generally benchmark that against climatology. So what would you if by chance you were to estimate, you know, what the flow might be in the next seven days? How will do our forecast to compare to what you could otherwise do by just looking at long term averages? So green green boxes there show where the forecast is better than climatology and purple show whether they're not as good as climatology. So we generally for this particular forecast you'll see that the forecast is better than the climatology for the next four days or so. Thanks, Richard. I should mention also that each of these products also has a description and information on how they derive on the information sits behind them and you can also download the data that sits behind any one of those images as well. So just to give you an idea of how we go about or how we how we develop these forecasts or generate them and the data that goes in. The top left shows the inputs so largely past information is a really important. Hence the monitoring and the data systems that allow us to ingest that data very quickly is important. So rainfall, it's a daily rainfall for the previous few days, as well as past daily stream flow are really important inputs as well as the seven days ahead early or three early rainfall forecast as well from our systems. So they go in to a rainfall runoff model that runs on an hourly time step. There's some post processing that occurs and that's where the daily stream flow comes in and that is then used to generate the seven day ahead forecast. Thanks, Richard. One of, again, talking about the, you know, how we increase adoption, I guess, and how we've worked with users in the past. One of the ways we've done those is develop a number of case studies. This one's available on the website as well where we work with the Goldenbrook and Catchment Management Authority and we helped them with a decision that was made or that would have been made to release water from Lake Yildon. So one of the challenges they have in relation to environmental flows or the gold Murray water has in relation to environmental flows releasing water from Lake Yildon. It's meeting certain targets but then avoiding unwanted flooding further down in the lower part of the broken gold and catchment. So now we don't provide. So one of the things that we provide is seven day forecast below Lake Yildon. So the trips that flow into the golden and so that provided them with guidance as to whether they should or shouldn't release water from Lake Yildon. And this was suggesting that increased flows were likely in the trips below Lake Yildon. So they actually made the decision not to release water which could have potentially led to unwanted flooding further downstream. So that was one, I guess, a case study that was shown. Thanks, Richard. So moving on, next. Also talking about the seasonal streamflow forecast service which is probably one of the oldest or the earliest services that was developed by the Bureau as part of the Water Information Program back in 2010. I think they've initially released them to registered users. And so we've probably had most experience with this particular service in working with customers. So this one is one where we provide cumulative forecast of streamflow volumes out to three months ahead. We do that at over 200 forecast locations or that's actually over 340 for registered users as well. And we tend to put sites into registered users before we then transfer them across to the public. Anyone can access those registered user sites. You just have to let me know. And we can send you the details. So that's the first service as well. Again, the link is at the bottom left there. And the main product, the first product that users or visitors have shown is the box plot to the lower right. Again, to the left showing the observed and to the right in red showing the forecast. Here we've got, so the red forecast are overlaid onto the blue historical reference. So that is the long-term historical reference, what you would expect for this time of year at this location. And then the red forecast are overlaid onto that. So it gives you an idea of what you are likely to expect or what is more likely to happen at this site for the next three months, whether median flows or likely to be higher or lower or near median. And again, you'll see there are box plots. I mean, in this case, we've got 5,000 ensemble members that go into these distributions here. So that gives you an idea of the spread, I guess in the variability that we see. And again, you can navigate through a dynamic map on the top there and move your way around the country. Next, thanks, Richard. So here we've got, the top left was the product that I showed before. We have two other products to, I guess, communicate what the forecast is suggesting the top right is flow categories. So where we split up the historical flow at a particular location, at a particular time period into low, near median or high flows into equal thirds. And then we represent what the forecast, this tells you this particular forecast is suggesting that there's a higher chance of having near median or high flows for the humedown for this particular, for those three periods. And yeah, we can see that you've got a one, one, two and one, two, three month accumulative flow flows for that particular site. The lower one probability of exceedance curves as well, just a different way of representing the same information. Thanks, Richard. As I mentioned, we're pretty fussy in relation to verification. So providing information on how will the forecast perform is really important. So we do hindcasts where we have a cross-validation procedure where we look over the past January, 2030 years and we test to see how the forecast would have gone had we applied them for each of those time periods across those 20 or 30 years. So the top left is a skill score metric that is commonly used. And so this again, it's a box plot. We've got a rather than a single figure, we've developed what's called a bootstrapping method to get a distribution of skill or performance. So it has a median and a distribution around that median again, we provide skill or performance measures relative to the historical reference. So many water managers in the past would have used spreadsheets and looked at the historical average over a certain period and that would have guided them in relation to what they might expect for the coming months. So we provide, we sort of benchmark our forecast in relation to what you could expect by using the long-term average. So these box plots, if they're above the blue line, they are better than using the historical reference. And I guess the amount, the distance above the line gives you a better feel for how well. So what I guess those box plots are telling you is that certainly for those except for May to July, you've got 100% chance that our forecast will be better than using the long-term average as a guide. And in some cases, forecasts aren't as good. You'll see that the forecast, this is a particularly good site inflows into the Hume Dam. Many sites in many parts of the country don't have as good a skill as using historical reference. So that's a guide really that you shouldn't place too much emphasis on those forecasts. And so that's an important guide to users as to know when or what sort of, I guess, assurance or confidence they place in the forecast. And you'll see also that the forecast, and this is something we see quite often perform. It's a particularly those transition in the Southeast mainland anyway, when we transition from dry to wet, it's generally a challenge. And one of the reasons, as we'll talk about in the next slide, I think it is, is that one of the inputs we have is the previous month's stream flow. So persistence in flows is a really important part of our performance. And so that on the right, just a different, I guess, giving users an idea of where our forecast fell in terms of or where the observations fell in terms of our forecast related, being able to relate our forecast to the observed flows that occurred during that time in the past. And in the bottom, again, this is something that users were keen to have more information on and that really was about how well have our forecast gone over the past year, those three months periods. And the red shows the forecast themselves. So you'll see that certainly during July, September for this particular, back in 2019, this was that our forecast were tended to over estimate the flows that were observed during that winter period. Thanks, Richard. So climate influences top left. So the inside ID are main ones, but there are others as well. And then on the right, just give you an idea of the anti-seed and catchment conditions. So really what we're talking about is stream flow from the previous months. They're the two bits of, the main bits of information. And this was a system that was an approach developed known as BJP, a Bayesian joint probability model developed by colleagues at Siro. And that allows us to generate 5,000 ensemble members, the distribution, and then we then provide that distribution after one, two, and three months ahead. Thanks, Richard. Don't tell me it's happened again. No, good. And that allows us to provide a national overview. So one of those forecast products I mentioned, those bar charts in terms of flow categories, if we take the most dominant flow category from each location, and this is the current forecast that we're about to release on Monday, I think. So this is the overview. And this provides, I guess, a national overview of where we expect flows to be either high, near, medium, or low. And we also have a normal flow category. That one is one we developed only a few years ago to deal with the situation. This is one of the challenges we have with the national system, that these models work well in some areas and maybe less well in others. One of the challenges we have, in particular with arid and northern areas, are no flows or zero flow days. How do you forecast that? And so we made a decision that where flow volumes are very low and we had a certain threshold that we specified across the catchment, then it makes no sense to try and forecast a flow when it's likely to be zero. So for those, we provide a category called normal flow where really you just look at the long-term average as your guide to what you would expect at those locations. So typically at this time of year, not surprisingly, a lot of the normal flows that are expected in the northern part of the country. But yeah, it gives you an idea of where flows are high and where they're not. Our flood forecast sometimes uses information as well, particularly at the moment in the Southeast, not surprisingly, lots of high flows. And that is something, I guess, along with wet catchments that our forecasters look at together with the forecast rainfall as to what the next season might hold ahead. Next one, thanks, Richard. And again, so the skill scores we do in the same way we categorize those skills scores according to certain, in certain categories so that we can communicate how well the forecast are expected to perform across the country. And you do see variations throughout the year. As I said, during the autumn period, particularly, I think April is typically the month where performance of our forecast is the lowest overall and so you can already see a seasonal trend in the skills scores across the country as well. Thanks, Richard. So this is just another example of how these forecasts have been used at case study, which isn't on our website, not one of the published ones. There are a number of other ones as well, but this is a case study where we worked with DPIE in New South Wales. I think they're DPE. Now they've undergone another name change where they were using our forecast inflows into Angola Dam and together with some estimates of their own estimates of evapotranspiration demand in particular. So extractions, they put together different scenarios as to water storage levels for Angola for the next six months and they provide that in their allocation statements that they issue to their customers during the irrigation season. And so we worked together to provide, I guess, a pilot study or just a test to see whether the Bureau's forecast could be used to provide a little bit more certainty around those. So they've got these dry, medium and wet and minimum scenarios that they apply, but they used our inflow forecast along with their demand forecast to develop that green wedge there, which is based on the Bureau's inflow forecast. And you'll see that it's a narrower spread than their own forecast. Next, thanks, Richard. You can probably do a couple of clicks because there's a couple of animations. So thank you. So if you look at the inflow climatology based on their own climatology, so they would have used long-term inflows to guide these scenarios. And here they've used a 20th to 80th percentile inflow climatology, you'd get that spread. Using our inflow climatology, you get the spread shown by the green wedge. And so I guess that's the moral of the story in a lot of our forecasts is that our forecast tends to reduce the uncertainty that you would normally expect at these locations. So to give users a better, yeah, more certainty, I guess, in terms of what flows they may expect. Next, thanks, Richard. I think there's one more popping up. So that was the, I guess, the level that ended up being the actual level. Thanks, next one. So now we'll move on to a relatively new service that the Bureau has provided based on hydrological forecasts on a gridded scale across the country. Thanks, Richard. So the service is known as the Strainwater Outlook and I've shown the link up there. It was released in last year and really we're moving towards a more seamless water balance service. So the Bureau has provided historical gridded information for a number of years now via what was, via the same model called Aura, Strainwater Resource Assessment HIFNL for landscape model, also developed in partnership with Siro over a number of years. And so there's been the number of, so that historical information has been there now for a number of years where we provide output for things like soil moisture and ET and runoff for the, like, I think it's from 1900 or 1911 up until yesterday, so it's updated every day. And you can extract the data that all that history, as you can imagine, there's a lot of technology and systems that underpin it as well because these calculations obviously are not trivial at all. So thank you. Thanks, Richard. Gives you an idea of the type, the information that goes in and comes out of this model. So it's a water balance model as known as Aura. Elle, you may have heard of it. As I mentioned, the Strainwater Resource Assessment Landscape Model across the whole country. It's updated on a daily time step. This is for the historical information, so the monthly, sorry, the seasonal forecast are updated every month. And although that might change in future, it's still, as I said, it's a new service and we're developing as we go still. So there'll be improvements into the future and it operates at a 5K resolution across the country. So inputs, the variable inputs, I guess, are rainfall temperature and solar radiation. There are a number of fixed inputs across the country that relate to the maximum relative available water capacity within the soil, saturated conductivity, leaf area index. These are all fixed parameters across the country. So for the historical service every day, the model is run, it's run on a daily time step. And provides output, as you see on the right, for soil moisture, evapotranspiration and runoff, in particular, they're the ones that are of main interest. And so what I should, yeah, I think the one that's of most interest is the soil moisture. That's certainly from customers that we've heard, particularly landholders calculated as the excess leaving a grid-cell deep drainage. So that's just something to bear in mind, some of the definitions of these parameters or these processes may be different to what people have or perceive them to be. So the cells are not connected, so the water's not routed through the landscapes, that's just something to be aware of. That's something we're likely to address into the future as well. Thanks Richard. So gives you an idea of how the system works to a degree. So we're providing three-month forecasts for individual months. In this case, we have a 99 member ensemble. So again, we've stuck with, we're really moving towards ensemble-based forecasts. So we've got a 99, obviously it becomes difficult running the model many times with such a complex model. So there are 99 member ensemble members that go into the maps that we've then produced. And in terms of the forecast, there are three output variables, one being the root zone soil moisture. So that's from zero to one meter soil moisture and actual evapotranspiration and also runoff, as I mentioned before. So the inputs in blue, the access S being the Bureau's current seasonal numerical weather prediction. So that's what is used to provide the climate outlooks for rainfall and temperature for the Bureau. So we get rainfall, temperature and solar information from that system or that model. And we also use wind, the climatology as well, primarily to calculate evapotranspiration. That goes into the model itself. There is some satellite information, particularly that goes into assimilate data that's related to soil moisture as well to improve the accuracy of that soil moisture. And then there's an update that occurs to provide those three month outlooks. We do generate, we can generate these forecasts every day as we update the model, but we only issue them once a month on the first of the month at the moment. Again, that might change. Our climate outlooks are now updated. Well, we provide a two, three weekly climate outlook that's updated every two weeks. So it may well be in future that these forecasts are updated more regularly and not just once a month. Thanks, Richard. Here's an idea of what a typical map might look like. So a monthly time step, we've chosen the median of the percentiles. Again, it's an ensemble, so we have to choose a particular variable to focus on. And then we provide, as we do with our other climate outputs, is we provide an indication of where that variable is likely to sit based on what you would normally expect for a particular location at that time of year. So would you expect soil moisture in this case to be average or higher than average or lower than average? We do also provide estimates of actual soil moisture numbers. So in terms of the plant available water, for example, how full do you expect the bucket to be at this time of year as well? So there are two main outputs, I guess. Next things, Richard. So this is what the website looks like if you were to go in there at the moment. So the top left, you've got historical forecasts and projections. And so we're on the forecast at the moment. So historical, as I mentioned, is really part of the service that's been existing for a while. Both the forecast of projections is something new, so the projections provide you with climate change scenarios and emissions scenarios out for a number of decades as well. So you can look into that if that's of interest. But for this particular seasonal forecast, you can choose the month and the year that of interest of the timeline, but you'll see that then choose a particular either soil moisture runoff or ET that you're interested in. You can, along the top there, you've got, you can select by catchment by a particular river region that you're interested in, all by state or nationally to look at the data which is on the right-hand side, you'll see box plots as well. So you can aggregate that data and information in relation to each part of those websites that will guide you. And then there's also emails then too that you can ask any more detailed questions of our team. Thanks, Richard. Just to give you an idea of some of the use cases that we've discovered. So agriculture is an obvious one, particularly for soil moisture. One thing to note is that these soil moisture forecasts are on a five-by-five-k grid. So an individual farmer trying to see whether a particular paddock is gonna be wet or dry in the next three months might be a bit challenging, but it'll certainly give you an idea of what, in broader terms, what you're likely to expect across the landscape. Emergency services have expressed a real interest. So for flood risk monitoring, particularly in those areas where if soil moisture and runoff is expected to be high, the streamflow forecasts are high then. There's probably enough from me because we have a lot of questions. Of course, the early bird questions get right of way. So I'm going to do those first. So firstly, Jennifer Watson, using streamflow data to establish limits. Can you see that, Paul? Because it's undermory. Yeah, it's undermory as well. So using streamflow data to establish limits inflows into receiving water body. So I'm not sure exactly. So I probably could talk a bit more generally. So we certainly see our inflow, and we call them inflow forecasts into storages as really the high value part of the seasonal, well, both the seven-day and seasonal, but particular seasonal service. And it's interesting that, so we work with a number of water managers or storage managers, and they provide us with estimates of inflows into these catchments, into these storages. And so we don't actually calculate those inflows ourselves, but they do. So they will use streamflow measurements as well as any other water balance or that they might scale and use all sorts of techniques. But we rely on individual storage managers to provide us with a series of inflows. And so really for the seasonal, because it's a probabilistic model, as I said, all we, well, to set up the model, the most important bit of information we need is a time series of any sorts of flow. So they can be from a cage, which is the majority of our locations, but they can also be a series of inflows into a storage, which is calculated by the storage managers using gauges and water balance methods. So, yeah, that's something, and that's something that's really where we're expanding. We're talking to a lot of a number of people, water managers, to, yeah, to increase the number of forecasts that we provide forecast for. But in terms of limits, I guess, so what we've found as well is that water managers tend to be, well, certainly those that manage water tend to be rather conservative. So with DP-E, for example, in that example that I gave, they would still use the no flow scenario as their bottom limit, I guess. And in terms of providing water allocations, so they wouldn't use our forecasts for that, but they do provide our forecasts. They have provided them to their customers for additional information, yeah, in terms of developing their allocations. Look, they're probably in very porous, and it's gone off this, but look, yeah, they probably would, again, it would depend on the catchment, how much of the base flow is part of that, how much base flow contributes to stream flow. One of the things we certainly with, so in terms of the seasonal, the seasonal service, the model doesn't explicitly, or explicitly use groundwater informational base flow. So we gather that really, that's where the beauty of using stream flow comes in, is that that really captures a lot of things, and among other things, it captures the groundwater component of that catchment. It really is an indicator of how wet that catchment is, and the research that was done shows that that persistence really helps, is the biggest source of skill or performance in most of our forecasts, and that they're particularly in the Southeast mainland. So yeah, I bet they would, but again, with providing a service, we have to be able to update relatively quickly and routinely. So the best, improving performance is not the only metric we use. So we need to have an efficient service as well, and systems that can cope with providing that service and be reliable as well. But yeah, I bet in some cases, yeah, groundwater measurements could be very helpful. Not at this stage, I think. Look, I know, so we do have groundwater, so the period does deal in groundwater as well, but I think one of the issues is probably just being able to incorporate. My understanding is that we ingest groundwater data on a weekly or monthly basis. I think I don't think it's regular enough. The Bureau is also embarking on sort of a much larger project. I mentioned we use Aura to run our Australian Water Outlook. We're doing work looking the future plans and the next five or 10 years is to use a model that's used in the UK, known as Jules. So that's undergoing research at the Bureau as well. So I suspect that one will have a better groundwater component to it. And so any improvements or any corporation will, you know, into other parts of the water balance will come through that, you know, the sort of adoption of that model, I guess. Right, we better get on to the next few questions. Rainfall forecasts beyond one season. Everyone always wants to know that. Look, it's always a challenge. I think experimentally the Bureau does provide, look, you know, internally we have forecasts, I think out to nine months or something, but I really don't think there's any skill or any performance in those. And so that's a challenge, I guess. What we do have experimentally also with the seasonal streamflow forecasts that I mentioned, the catchment-based ones, we actually have experimental product where we provide an outlook out to 12 months for flows and we actually do get some skill still out to 12 months and that's because of that persistence. So particularly, you know, for places in the Southeast, on the Southeast mainland, if we were to forecast now, for example, in the higher flow part of the year, then that skill out to a year, we can have skill out to a year, but it comes from that persistence, which you just don't have in rainfall because it's pretty much chaotic for the best part. So the landscape forecast we're always going to struggle, I think. And also we find, so soil moisture is, I think has the best skill of the hydrological variables, runoff not so good, because any uncertainty and variability we have in rainfall, which is the biggest limit to providing skillful water forecasts is the rainfall. If we could do much better in rainfall. So, you know, if we look over the past, if we put in observed rainfall into our models, we can provide much better, you know, we do really well in terms of soil moisture and runoff, but it's the forecast. So it's the forecast rainfall that we fed into the model where the uncertainty is. And that uncertainty tends to get amplified when you come to the end of the water balance, because most of the water, you know, evaporation of what's left over, soil moisture and then runoff. So very small changes and that variability in rainfall gets amplified in the runoff. So runoff is a really big, and you need to be catered for the, you know, the landscape and the way that water moves through the landscape, which is a real challenge. So that's going to be the biggest challenge, I think. So Paul, do you see farmers using more the planned available water estimates or soil moisture now than rainfall? Yeah, I think they will. I think they will. Again, it's relatively early. I mean, I'm not involved in that part of the service directly. But yeah, that's certainly where the main interest, I think, from developing that water forecast and capability came from, was from the agricultural sector in relation to soil moisture. Yeah. Maybe just a point. With some of the work we've seen with DPI in New South Wales, they've got coupled pasture growth models to these sort of planned available water estimates that we're seeing. So it's quite amazing, but, you know, there are projections of pasture growth rates that can go out, you know, effectively 40 years for given areas because we now have these projections of planned available water using models like what Paul's been talking about, which I find amazing, to be honest. Apologies to a few people because we've had a couple of technical issues today. So some of the Q&A questions have dropped off. I guess it's a bit of a learning and you want to be in the early bird batch. But I do have a few that must have been put in later, so I will read those. So I have one from Jennifer Watson. I'm interested in streamflow data for a specific creek catchment area. How do I find out if data is available either via a public user or registered user to the database? So that's historical data, I'm presuming. So the first place to look at will be on the Bureau's website. We do have a national database of streamflow information and it's called Water Data Online. So if you go to, probably do a Google search on Water Data Online and Bureau, that will take you to that particular website where there's a national database. Or at founding that, I suspect if it's not there, it might not be, but you could also look on the state agencies have their own databases as well. Okay, a question from Anthony Mazalik. Can we access daily runoff estimates for all 99 ensembles for use in our own water balance, water supply models? So now, look, that's the other thing I didn't mention. So I think that's going to be a challenge so we don't provide daily. If something you'd have to ask, I wouldn't know if there was daily information, certainly not public. I don't know if it would be technically available either. So we had as a separate project, that's something that might come down the pipeline as well as part of the Australian Water Outlook was that we have been looking at, in addition to the seasonal forecast, we had been looking at providing forecast for zero to nine days using the same model. So technically it is possible, but we haven't got there yet. So that is on the cards, that is something the Bureau is seriously looking into. And then I would have thought, yes, then in that case I would have said yes, but I'm not quite sure with the current seasonal service whether that's possible. But I suspect have a look at the website and follow up contact the Bureau through the website to ask that question. Okay, so Jason Pan has asked for the new technology to do the water forecast and monitoring. Does UAVs and USV have an involvement in it? So I think he's looking at that, is there an application for that technology to help you to improve it? Probably not at this stage. One of the challenges that we have is that we provide national services so we need to have frameworks and to set up to ingest national data. That's one of the issues we had with the soil moisture as well, how do you calibrate and parametrize a national model where you don't really have good soil data. If you look at the country of being isolated areas where you have really good soil moisture data, for example, to test the model against. So that's one of the challenges we have that we have to make some assumptions as well. But we also have to, again, it's a real balance of trying to improve the model and provide the best performance, but how can we do that operationally? So it's not just about providing the best forecast, but providing it in a way that is valuable to users in a way that they can access and understand and at a time step that makes sense as well. So it's a challenge, but we're always looking into better ways. And as I said, that new modeling framework that we're looking into will also get us to ask those sorts of questions again, how can we better? So we do have data assimilation and we do have teams that look into that, those areas. All right, well, we're just about where we are out of time. Paul, you've given us an extra six minutes for free, which is very kind of you. Pleasure. It's been so nice. It's been put into the Q&A, so actually into the chat. So I'm gonna have a quick go with those. Paul Webb, have to go. Very interesting and informative. I will be watching the projections, the evolution to support natural resource management, stakeholder awareness. Locally in South Queensland, the climate mates are doing great work making Bureau of Met product info and explanations available for land managers. So that's a good response. I do think this is something that is an area to look at how to collaborate is just readiness to find what one may want out of the Bureau. You know, that sort of interface between maybe the application of it versus what's obviously fantastic resources inside. I think there's a bit of a theme that's come out of a couple of questions there and comments. Irrigators use plant available water capacity plus RF plus streamflow forecasts and outlooks. So I guess RF is rainfall. Yeah, look, that's a good point too, that a lot of it, and that's I think something the Bureau, so the Bureau has undergone a big transformation the last two years and we're still undergoing that transformation. So I think one of the important parts of that as well is to bring together a lot of our services to understand how customers use different parts of the Bureau. I think in the past services developed sort of independently to a large degree. So there will be changes in the coming years to the Bureau's web interface to better reflect that sort of how users engage with the Bureau and how they use different services. So to improve that experience as well. So it's a good point. Thanks for raising that. Oh, it's always good to have a bit of a deep poll. Oh, always. Yeah. I think that concludes all the questions. So I'd just like to say thanks very much, Paul. That has been excellent. And I really do think you guys are doing a fantastic job. I think the sort of nirvana of maximizing utility of this data is going to be some degree facilitated through these sort of software providers who integrate the Bureau's data into their services. And I think sort of maximizing that interface is probably key to getting to market. It is. Thanks for that. That's important, Richard, because we can only do so much. So we will draw the line, I guess. And there's an expectation that, yeah, that people will make use and add a lot of value to our, what we produce and the data we produce. It is. I mean, it's, people look at all these different forecasting services, right? And they don't realize that sitting behind it, 99.9% of the time is the Bureau's data. Yeah. They're going through some other set of algorithms. It may or may not be most appropriate. So anyway, look, Paul, great to catch up. And many thanks for everyone who have been involved in today. It's been a great audience and many thanks for all those questions. It's been really good to reconnect, Paul, and catch up again soon. All right, thanks very much, everybody. Thanks for having me. That's a pleasure. See you. Bye-bye.