 Well, hello everyone and welcome to another of Hydra terrors webinars Today we've got a presentation from CSI row on What's known as act or watch Australia? Which is a near real-time water quality monitoring and forecasts at national scale. That's the aspiration for it We've got several presenters. We've got Dr. Alex Held Dr. Tim Malthus and Dr. Klaus Yonk Along with myself so a big field of experts today Before I get started, I would like to begin by acknowledging that we conduct our work Across this great land and for that privilege. We would like to thank the traditional owners Hydra Terra respectfully acknowledges the Boon Wurrung people of the Kulin nation where we are located today And we pay our respects to their elders past present and emerging All right a little bit about our speakers and their organization For those of you who may not know what CSI row stands for. It's the Commonwealth scientific and industrial research organization And it exists to help the Australian economy and the Australian environment And they've been going for a long time and they do great work There's many different units within CSI row and today's presentation actually cuts across Two of those units one which is known as space and astronomy and that's where Dr. Alex Held is from and then the second unit is Environmental unit and that's where Tim and Klaus are from So we're lucky to have such a great combination of people A little bit about their expertise So Alex is one of Australia's leading earth observation experts Is the mission lead for aqua watch and director of earth observation infrastructure In his tenure at CSI row Alex has overseen the establishment of the new radar satellite Novasar one as one of Australia's newest national research facilities Leads the development of innovative science in remote sensing and drives the development of spatial technology And data analytics to support sustained earth observation and measurement of our planet's ecosystems Alex also mentioned that he's famous for having a planet named after him Not many of us can claim to have a planet. He did say it's a minor planet. It's called the held planet So that was interesting and he's been with CSI row for more than 30 years. So a great period of time to be spending in one place Dr Tim Malthus Tim is a senior principal research scientist in the environment business unit with over 30 years experience In remote sensing of water quality with a focus on optics calibration and validation field spectroscopy and sensor development Tim is a co-creator of CYROS patented hydro spectra Near-surface sensors for the measurement of water quality parameters We're going to have a bit of a look at some of that today As part of the aqua watch Australia mission he leads the insitu Sensor networks work package working with optical and electronic engineers and scientists overseeing the challenge of development and deployment of sensor networks at wide spatial scales And finally Dr. Klaus Junk Klaus is a lead scientist in the field of water quality modeling and holds the position of principal research scientist at CSIRO Environment Business Unit Klaus leads CYROS modeling water ecosystems team in addition to work package Lead for aqua watch which focuses on modeling and forecasting He's got over 25 years of experience in europe and australia and he specializes in Hydro dynamic and water quality modeling of lakes and rivers And Klaus is a bit of an example of Where you can go to many different places with science. He started off as a geophysicist Decided he didn't want to work with rocks and has since gone on a journey Using a very strong analytical brain So without further ado just a couple of things administrative We love your questions So today if you want to raise a question just use the q&a button at the top of your screen And I will read those questions out at the end of the presentation Why does Hydro Terra like to do these webinars? We like to share knowledge We like to facilitate education And we like to work with industry leaders and certainly This aqua watch program is a true leadership program You know, it's a great aspiration to be able to try and forecast water quality And it's also a great aspiration to try and link satellite data and in situ sensing in the one into the one system so Without further ado, I'd like to pass over to our CSIRO participants today Thank you. Thanks very much Richard. I might start with a few sort of high level Description of what aqua watch and division we're trying to fulfill here But I wanted to also begin by acknowledging traditional owners of the land and waters that we're meeting on today and Pay our respect to their eldest past and present Thanks next slide. So Aqua watch has been going only a year and The our ambition is to Operate it a little bit like you would see the the weather forecast On your phone and provided by the Bureau of Meteorology. So Sort of the catchphrase that become popular is the weather service for water quality Um, we're really into also very much focused on supporting Australia and many other countries contributing to their sustainable development goals and Especially those associated to water and sanitation Next slide, please Um, and these are some of the ambitious Goals and ways we want to support Three billion people worldwide that really still don't have access to clean water and sanitation And support those who are concerned about coastal water quality impacts on on industries like Tourism but also aquaculture and natural ecosystems like the Great Barrier Reef As well as supporting early warning perhaps for Problems that might cause fish kills in inland water bodies like we've seen in australian number of times But also now the developing countries looking after water quality Including potential fish kills due to harmful algae blooms that affect coastal and Fish fish populations and aquaculture quite frequently and more frequently over time So next slide, please. Um, the ambition then is to bring a lot of technologies together that Some of which and many of which have already been developed But have not been brought together into one big system And this is basically a graphic that sort of tries to explain that Where are we bringing in? the Sensor networks and situ sensors that are located across different water bodies surface water bodies and coastal water bodies already existing in australia or and new ones that we hope to add to to the list with satellite imaging technologies From where a number of our scientists and experts can derive some metrics of water quality And quantitative measurements of water quality, but at a larger sort of synoptic scale And and then provide and also bring in some more modeling and simulation capability so that The hope is we will be able to provide Forecasts and early warnings Two or three days if if not longer so that people can have The opportunity to make decisions around water quality And and doing something about the water quality Before it hits a fish farm or Goes down the river Thanks very much Next slide We launched as you can see this mission mission is like a very large program in in the CSI row nomenclature. It's a very ambitious multi business unit cross organizational program Which also aims to bring in a number of partners from outside the organization To really tackle some of these key really important challenges and problems We were launched almost a bit over a year ago when we had the un world water day Some of us were here at the at the shores of lake burley griffin with minister husic and minister plebiscite And local ministry of the act to launch aqua watch here at the same time some of us were at the united nations also Including aqua watch is one of the australian commitments to the Action agenda around water at the un As you can see a number of photos there. We included also as part of the australian delegation a strong Component of first nations representatives also interested in water Next slide please So again showing this in a Graphical form is we want to bring in together In-situ sensors dense networks of sensors find ways to relay the data from Perhaps often real remote areas across australia into a single data system as we call it the aqua watch data system But also bring in satellite information From existing satellites that we can tap into from from the us from nasa for instance or from the european Copernicus program And then also add the ecosystem models that then allow us to do A lot of the processing and the forecasting that we would like to be able to do in aqua watch On the right hand side, we also have teams working on how Best do we make this information available to different types of users? All right, so you might imagine technical water managers in a water utility or in a water agency wanting the data presented in a certain way but you might also as a In the presented data In a in a different way to community groups first nations people Who want to make decisions about whether they? hope for a swim or go for fishing and things like that And so it's also about how do you make this data is is is easily Understandable, but also how is accessible as quickly as possible so people can can react if there is a water quality issue nearby Next slide just just before you do so these ecosystem models What do you mean by that is that an ecosystem as in a natural ecosystem or as an ecosystem of technologies that are I We'll go in a lot more detail into this if you want to just quickly answer this question Did we say now pass to you? Yeah the ecosystem models in the sense of process-based models which can solve differential equations on the technical side or on a non-technical side in principle models which can describe how What a quality changes over time? Driven by external factors That's in principle. What do you look at? Thanks for that Thanks Klaus So the whole data system, which we call the aqua arts data system is basically Being set up as as we speak we leveraging of An existing technology also developed in by CSIRO in in conjunction with for instance, geoscience australia Um called the open data cube and our implementation is called easy which stands for earth analytics and science innovation platform, which brings a lot of these different data streams together um, and allows us to also tap into Thousands and thousands of satellite image that sit on the cloud on amazon servers around the world So the the platform and if there's more questions happy to describe that in a bit more detail, but it runs on the cloud And it's it's like it's basically sim emulating a massive supercomputer Which then brings in all these different data sets from in-situ sensors satellite data And then runs also some of them will will run some of the modeling Technologies to then provide that data as a map interface or a data stream Next slide please One way we are really Learning by doing this because this has never been done before bringing all these different data sets together Is by setting up a number of pilot sites across australia and some internationally as well um They not only help us on a technical side how to figure out how to bring these different data sets together and play some instruments in different water bodies, but also develop partnerships with Communities or local agencies that manage water bodies around australia as well as some international partners who also Have quite unique water quality issues and problems that we want to learn about as well You I think the next slide will show you a map Where some of these sites are Apologies, we lost them the the map for australia there, but I think We have so many pilot sites you can sort of see roughly what the where those are You're still trying to have a few more in northern australia and in western australia, but There's quite a few that are both inland water bodies, but also at the coastal zone And there's a new a few new sites popping up overseas like in vietnam Uh in california in chile we're supporting the the salmon industry there for instance As well as now one in the uk Looking at some of the river flows there into into the english channel And a new one in the in italy off the coast of venez, which has some a lot of pressures on on water quality at that area If someone wants to Include a site that they've already got where they've got water quality data. Is it? How should they do it? Please reach out to us. Yes, um We're we came to collaborate with as many potential users and learn from bringing all these different datasets together As I said, eventually we want to cover the whole country It'll take us a while to be able to do that especially Be able to forecast and model every water body that Important water body across australia and I can see claus getting more and more gray hairs doing that but anyway um The idea is to eventually have the whole country covered. So, um, yeah That's happy to have a conversation about new sites. We might want to have yep Next slide, please And so, yeah, we are of course not just Implementing existing technologies, but we still have an active research role as you know in csro innovation age science agency, um with a number of teams developing new sensors new technologies and tim Uh, we'll talk about some of those the hydra spectra. For instance, but we also have teams, um helping us develop what might the A very specialized Satellite sensor need to to be able to do in space. That's just focused on water quality monitoring So next slide, please And we have a really interesting project as well overseas with with nasa and their jet propulsion lab Where we're designing A conceptual study. We don't have this funded as a satellite yet, but the conceptual study of what would A really high performance satellite Look like and what sensors we would need to have in that satellite To map at least these three major application areas Um, where we want to be able to discriminate between harmful algae blooms and non harmful blooms. For instance Look at aquatic invasive aquatic vegetation and also map The corals and coral reef a bit better than what we can do with existing satellites at the moment Next slide, please What's that? A jpl design dyson imaging It's it's a very high signal to noise Spectrometer and imaging spectrometer as it's called Which takes very detailed reflectance spectra For each pixel so each image pixel that just you have on the instrument Um, dyson is the name of the actual design of the optics and the and the sensor and the detector It's it's it's a it's a form of optical design of these sensors One version of that is already on the space station by the way called emit Um, and we're using that to to learn how to process that data and to see how much Information on water quality we can derive from that for instance And uh, just with respect to the reflectance and being able to pick the difference between Invasive species, I mean Is it really has it got to that level? You really can do species identification from space in some cases Yes, not all because I mean the plant plants look green. I mean, uh, but Uh, invasive species also the harm Algae might have some additional photosynthetic pigments that they use as part of their photosynthesis that Give them a very very narrow but very specific reflectance Signature fingerprint that we can then use to identify that species or that type of of vegetation um Yeah, I mean it's I mean, uh, the the traditional botanical um taxonomy Of course, you can't do that from space because sometimes the only difference is A pet the different size of the petals in the flower or whatever. Yeah, we can't do that yet But I think at a broad in broad scale uh, some of the major Groups of plants and species sometimes we can pick them up. So simply because they're they're different in their reflectance And so something like coral Coverage is that looking at the bleaching is that when we hear estimates of bleaching is that how they do that? They yeah, they do that also already with with some of the the Courser, I mean the the the normal multi-spectral satellites what we hope to do with this this sort of design is To discriminate much more precisely not only bleached and unbleached coral, but The different sometimes like there is a underwater you get confused between rocks and corals and and different other bentic materials like algae and other things So with this sort of design, we hope to be able to separate those different underwater features a lot better than what you can do at the moment Okay, so what resolution are you talking like? um, it's uh, it's uh, I think this one is being designed to map at 18 meter pixels um Remember that because we are we're we're doing a This I mean these satellites fly 100 kilometers an hour over over over earth in the low earth orbit We only have a certain amount of time during which we can collect the photons coming back to the sensor. So There is this trade-off of signal to noise and detectability and fidelity of the instrument versus the size of the area that that you can collect over pixels So this is this sweet spot of the best optimized spatial resolution versus the Discrimination capability that we need in the satellite to discriminate these different types of features So it's it's yeah, that's at the moment. That's where the technology is Is that I'm hoping that in the future As detectors and sensors get better and better over time Uh, we we can go into smaller pixels and pick up smaller water bodies and farm dams and things like that but that's the compromise from an engineering perspective at the moment that we can Pick get the signal to noise we need. Uh, but but still pick up things that are 80 meters as a minimum size Yeah, so you didn't really explain what Cone ops is I call up. Sorry the concept of operations Is sort of a it's a catchphrase in the in the space operational sites basically this this little graph shows that the the satellite is designed to Pivot itself To look at one spot for a bit longer as it flies over Right, so that it it move it always looks at the same spot That increases the signal to noise and the amount of photons we can get for that spot And therefore it allows us to discriminate things even better um, so it it Of course would be a very sophisticated sophisticated satellite that At the same time as it's collecting images. It's also Move and swinging a little bit to always point and loiter of the same point for A few seconds to get a maximum signal Yeah, like sense All right, thanks for that Sorry, so I might pass now to tim He will give you a little bit more information on some of the sensor technologies and the challenges we have to solve there And after that then like we passed to klaus to talk a bit more about the modeling in the forecasting Then good Yep, thanks very much alex Um, I think so alex has done a great job in identifying. I think what are the unique aspects of of the aqua watch mission And and it's really uniqueness lies in the fact that it's not not only just a satellite mission. It really has this Uh, this collection of multiple components Which are ultimately leading to forecasting water quality. That's the ultimate aim. So so it's including not only Not only the satellite sensors but in situ sensor information Incorporating that incorporating those into a uh, let's sort of data engine call it that or a data analytics engine and then integrating those with forecasts with forecasting models to be able to To make forecasts and that's the unique aspects of the mission And I lead the work package which is looking at in situ sensing technologies If you go to the next slide richard, so Just focusing on on australia on its own If we if we're wondering to roll out a network such such as as this across across the continent. I think australia probably perhaps Represents the biggest one of the greatest challenges that we could in terms of its landscape its remoteness Some of the geographic extremes that we have to deal with and complexities. I think it's a great It's a great continent to try and test this and if I think if we can Practice nut for australia then we can pretty much roll Something similar out all around all around the globe But there is a real challenge here on how you do Fundamentally roll out a sensor network across a A continent with with complexity With uh areas that extend essentially from the tropics to temporary ecosystems Um very large extremes in weather events wet season wet season in the north which can have Significant impacts as we've just seen this in cyclone megan um So large extremes also, you know, we want to um We want to cover a quite a range of water bodies as well So our end user indicating interests in and not only small scale things such as farm dams That was so urban systems rivers lakes reservoirs estuaries and and ultimately to coastal zones So a whole range of water bodies to to think about each with their own individual challenges And then the challenge then is to to do that in a in a means by which it's sustainable Uh, but also um is you know, to live as a robust and failure resistant system Next slide thanks ritchett. So we've developed a uh, um We sort of scoped I guess a green paper on this work Which is really identified in order in order to do this in a kind of economically viable way The internet of things is probably the the best approach for this. So a kind of Low cost approach to to sense a rollout Employing a range of different communication Uh communication infrastructures if they're available in particular areas from from things like Uh, Laura and when 5 5g networks, um or less and then satellite broadband as well So so just utilizing what's what's present but based on essentially a range of of potentially hopefully low cost Low cost sense of solutions Next slide thanks ritchett There are of course a lot of um A lot of challenges to solve in this particular area. So we've identified quite a number of particular design criteria and this is essentially we're at the stage of of working towards What might be a sort of uh, another white paper, which is essentially the design There's a design statement for what um for what this in-situ sensor might look like and but these are some of the criteria that will have to be met One are on cost effectiveness Easy to maintain It's certainly got to be robust to deliver timely data It's got to be able to deliver credible data Such that we can such that we have reliable Credible data on upon which to base our our forecasts It's designed to be open so that data will be available to to everybody Also representative of the of the general Geography of Australia and it's put it that way over water bodies. We want to want to cover and also A combination of direct communications and and internet of things Within that there are then a whole lot of uh problems to solve or challenges to overcome and I've kind of ranked these by What I consider the hardest and those which are perhaps the most solvable um, but you can see there that um clearly robustness and reliability maintainability are key things uh making sure different sensors are interoperable Flexible and and cost is always a key key factor um Our intention here is not to is not to reinvent the wheel. We're looking for technologies that essentially are That might be existing in in commercial off the shelf shelf sensors, but which are um cost effective to deploy at scale um And in order to to initiate some of that some of that work We haven't we have just recently employed external consultants who are going to undertake some Tech surveys with us. So so one study will look at a at essentially a tech survey of existing and promising low cost sensor technologies uh across both the let's say commercial uh, what's commercially available and what what's actually um In the research domain and do a kind of trl assessment of those of those different technologies such that we can sort of understand What's out there? What's promising? Um, and are there are there potentially Winners within those groups of technologies that might work in this this sort of system And then to to identify those partners who are working those areas and to work with them to encourage them to and perhaps find the funding to To uh increase the trls on those on those particular technologies The other study is also undertake and this addresses some of the some of the questions that have already been asked Ahead of the seminar We're also undertaking a survey of existing monitoring installations across the country. So certainly, you know, we're not looking to to come in and Put our sensors and in in places where perhaps sensors already exist So we're looking to to find out, you know, what is um, what are the existing installations across the country? If you if you have some we're certainly very keen to hear from you and and particularly if you are happy to make Make the data from those sensors available plus also, you know Where we might be able to piggyback some of our sensors on to onto our platforms would be really interesting as well Um, thanks Richard Just on those standards and protocols that you're going to adopt the bureau of metrology sort of metadata framework or Yes, yes, we haven't um, we will yeah We'll look at a number of those that there are a whole heap of um standards and protocols some around uh Certainly around metadata around calibration around quality assurance and quality control I would say we're probably relatively new in this game to compared to other groups, but but Certainly we need we need a whole stack of sort of protocols in place such to assure And essentially to ensure that that data that goes into our our ideas system uh Is is, you know, both both credible and and accurate um So so some of it links I guess to that that sort of quality assurance and quality control I hope that answered your question. Yeah, I think it's absolutely enough. Maybe the audience making my words The bureau of metrology has these publications It's brilliant hydrographers association They're up a really good And with comprehensive set of guidelines that yes, how to install etc. Yeah, yeah We're engaged with them already. Yeah, excellent. I'll move to the next slide So I thought I would just round off my small section of this the study just by by highlighting We're talking about two two Areas where we are or have been developing sensors so the first one which has already been mentioned is Is this device called hydra spectra and this essentially is a low cost Sensor which works a bit like um, what a satellite might work at work like so So it is an optical sensor It measures simply measures the spectral reflectance of water and that's very much what a what a satellite Intends to do and from that spectrum as as alex was describing We can then derive information about whatever target we're looking at be it corals or or Or perhaps elbow blooms and in water bodies um There are already scientific instruments which can do this but they're incredibly expensive You know, you could pay up to 200 000 dollars for a set of instruments, which would do the same as this device It's less than 10 000 dollars um so Hydra spectra was really designed to to measure above surface reflectances and to do so in a continuous Continuous manner so it will be permanently parked at a at a site of interests And it serves the purpose of both monitoring water quality at high temporal frequency So higher frequency than the satellites will view As such it then serves as it has a nice validation point for our satellite data so we can also We can also estimate the reliability of the satellite data plus also Way back in the first time we started developing this instrument It really was designed as an elbow bloom alert until we were really looking for a device that we could Put out on inland water bodies across Australia to alert Our engines as to when elbow blooms might be developing um The cleverness in this device is really in the spectrometer inside it and and that has been patented by CSIRO and the whole idea is that because it because it sits above the water it Suffers less of the issues we get with biofouling and and other issues of sensors in the water so so Its maintenance requirements are are somewhat lower than those those which actually sit in the in the water And The uh, I should just explain what we have here is a number of it has a number of fields of view the device itself The head of the device is is what you can see on the top right there um, it has a number of fields of view which which um We measure the spectrum of light and from that we derive a fairly accurate estimate of of Reflectance, we also have some cameras on board the device one looking up at the sky and one looking horizontally Out to give us an idea of the changes in water color And this is just one example from an installation on the Fitzroy River Up in up near Rockhampton Just to show you some examples of what those cameras deliver and already, you know Those cameras deliver quite useful information just in highlighting changes in water quality as you can see in the differences in color Of course by differences in tidal state This is an estuarine part of the river um And just the data on the bottom right are just a time series of of data showing showing a comparison between um Chlorophyll estimated using the hydro affected device versus an institute measured Chlorophyll, we have a pretty good agreement and this is for Another installation. This is in the Spencer Gulf Down down towards Adelaide Next slide. Thanks Richard So As Alex mentioned, we have a number of pilot sites Around around Australia and around the world. This is just an example of some of those deployments and and showing the deployment of hydro spectra um So typically, you know, we can put them on uh fixed platforms such as the pylons. You can see that's the one on the on the Fitzroy estuary um Or we can put them on boys the image top left is is on The uh is on the coven is on coven sound Typically if we put them on a boy, we will also deploy instrumentation um In situ instrumentation as well. So so on this boy There's not only a hydro spectra on the top which you can see but there is also um, we've also deployed in situ sensors Which you can't see of course Those are coastal installations. The bottom right one shows that shows an installation on on a lake Graeme sound dam in New South Wales We've been working with hunter water um Just shows a typical example and where the boys again on a boy, but the boys are uh less perhaps less robust and um Then compared to the coastal ones Typical spectra you can see on the top left there So this is just an example from the installation of coven sound just for one one month of one month of data um and From that we can then begin to generate this is one of our longest deployments Um, and you can see a time series of chlorophyll estimates And here in coven sound on the bottom left And a comparison to estimates the red dots are estimates from landsat satellite data And the uh, the fine point data is is daily estimates of chlorophyll from the hydro spectra sensor that you can see that Generally over most of the time the the two data sets correspond quite well Um, thanks very much ritchard And our last very early stage low trl development is is that we've been working also on a low cost nitrate sensor So this is really taking the concepts which are already already proven of optical detection in the uv of um Of nitrate sensing Some of you are probably fairly familiar with some of the instruments that that are already available commercially that can do this um But trying to do so in a in a device that is that is uh much cheaper So we're exploiting led technologies Um, and it simultaneously trying to account for some of the other signals that interfere with those nitrate detections Be it dissolved organic matter and and turbidity as well as temperature. So um We have a talented talented postdoc working in this area Um, and he has developed already a product. He's developed a prototype which you can see sort of on the on the bottom right Um, and we're currently scoping up a field deployable instrument And we're and he has also done quite a considerable amount of sort of theoretical development of the device Uh such that we're we're ready to go uh Yeah, we can we can um be confident that what we've got is a is a device that will reliably measure um Measure nitrate in the presence of those other interfering compounds and And once we have the field prototype developed we'll be um Uh, we can retrieve nitrate quite quite accurately Um, and thanks. I think the next slide I think what's your aspiration on At the size of the sawned end of the day we're going to be able to get down small enough to Go into monitoring wells for groundwater applications as well or Potentially we're we're currently sort of developing it for a housing which um I guess is as I'm sort of guessing here is about 10 Uh 10 centimeters in diameter and and perhaps 20 20 to 25 centimeters long Um, that's a housing that houses the the the electronic components. You can see top right there plus also the the battery and and Other things I think there is probably the potential for further miniaturization of this um once once we get beyond the our our plan is to actually field test it in um cotton In a cotton application where essentially, you know, urea is being applied When nitrate fertilization is being um undertaken quite regularly and the concentrations are quite high It's a nice. No, they're sort of nicely controlled Um studies that is where we're looking to really try out a an initial uh deployment of this device Um, we haven't really got to the point of considering considering ground waters But I think I think there is scope. I could certainly come back to you. Um, I can get to ruin The post office working on this to come back to you and talk about the scope for miniaturization That'd be great. We've got a lot of sites where we're monitoring the like water authorities their Batch, you know raw water before it gets treated And uh, it's quite good because you can get that journey of um You know your nitrogen, etc. And then also pick up the algal blooms, but it's a fairly controlled space, you know, it's a structured That might be a good place to try yeah um to ruin has already Um, you know trial just a deployment not not so much of that device but of of normal night, you know Typical commercial nitrate sensors from drones as well. That's another it's another interesting interesting development, too So with that, thanks. I'll hand over to klaus for his part on on forecasting Thanks, tim. So Now we have all our data gathered from the satellite and from in-situ sensors and we we have to take Make something out of that model the system how it develops over time And predict that in the future. That's our aspiration here Um, when we come came to to a to aqua watch, we first had a quite quite a number of end user workshops to figure out um What the needs of different end user groups are and that's word cloud and principle shows you A whole bunch of things, but it's turns out that at the end they want to have Modeling and forecasting for water quality as as an outcome for them to be able to well To have developed mitigation options for the system. They're looking at and seeing If there's any early warning Possibility so they they they can do it something in advance Or plan into the future So water quality modeling and forecasting is one of the Main issues they they have and they want to Get out of of aqua watch as a result Next slide please So with this Predictive data modeling for real-time decisions We would then be able to reduce as I said negative impacts of poor water quality And that's in favour for existing industry environment and community That's our goal. We want to go with the modeling and forecasting next slide The forecasting itself, uh, it's a bit Opening a The door to to everything forecasting of what water quality is There's no water quality per se what a quality can be everything Starting with temperature can be also physical process hydrology its salinity Its nutrients its pathogens You name it Everything is water quality. So we can't really do Everything in this in this project. So we have to focus a little bit on specific water quality parameters Which are of high importance At the moment and then widen our aspects what we what we try to To use When I've mentioned here on these pictures in principle four of those major types which we in inland water Which we directly can can see from the satellite But where we also have data available because these are the two essential things. We need this the satellite data We need the in-seater data to do our modeling forecasting. So we have um, the clockwise sense for example black water here lake victoria Which is a high concentration of dissolved organic carbon, uh, which tints the water black And that has the effect that also microbial interaction will just reduce the the the The dissolved oxygen. So here we have another example of a constituent which we can measure The doc in principle of the blackness of water and we can then Infer another water quality parameter, which is dissolved oxygen, which we cannot measure by the satellite On the next one is the the colorful one is an harmful agribloom in 2016 the mega blooms in legume Something which we Probably do in a standard way meanwhile from satellite to see how Bloom formation is what we want to do here is also coupled up with modeling and see how does that bloom Form in different parts of the lake or if the lake is small or in the river How does it spread? over time The next one, uh, to the lower right. That's a bushfire event. So that's um, gipsel and lakes Well, we had after the bushfires a high quite high intense Rainfall which washed The the mobilized sediments into the river into the river and then into receiving water bodies That has of course quite significant impact of ecosystems Along the river but also in in the likes And that's not only a one-off Just during the flood but or during the runoff but also Over longer times because the sediments accumulate and can lead to other problems later on and finally a little bit like the the runoff from bushfires is the normal flood conditions washing sediments into that case like alexanderina and the out into the Into the into the coastal seas next slide Are you integrating with all the sort of? hydrologic flood models and that sort of thing Now that's the aspiration we have to do that especially for the bushfire part We we try to to do that because we need to know The load the same is valid for plume development In the coastal regions where we look how plumes form in the in the gbr To do that we need to know what's the catchment hydrology and how does sediment is in principle Uh transferred from the catchment into the receiving waters So hydrology plays a big role hydrology also plays a big role in the problem space Which we have to tackle the modelling and forecasting. That's the We need to monitor across different time scales If you look at hydrology that that's a day to week where things change metrology change on a daily or sub daily scale But for example, internal reaction mechanisms can be very fast Um, so half-legal blooms. We need to know what's the algal physiology? How do they do they grow under different light temperature and nutrient conditions that can change of course very rapidly? um in all of an hour we we need to to be able to To model that And it's becoming even worse if you look into mixing processes in in the water column These are on a minute time scale so lake models usually work on time scales internally of a few minutes to resolve turbulent interaction so And that makes of course a quite big problem in in the modelling to to get all these aspects together and in the data necessary availability for this for this next slide Another problem is is the spatial uh resolution um That comes into uh Two ways one way is the modelling itself We have to to use different types of models to resolve. Let's say a big reservoir like like you on the on the left side Or a very shallow urban lake like Tunganong in camera Where we probably don't need a real three-dimensional model a one-dimensional would do But we also want to go into a smaller scale systems like what a treatment lagoons which are on the size of 100 meters if they are open um, and here you see for example, how that looks like in the terms of Chlorophyll A estimation across these these systems And it's not very uniform So they're they're local effects. These local effects are very hard to model in principle if you don't have enough data available And it's getting even harder if you go to very small riverages because The hydraulics of those riverages are can be very complex Especially if they are meandering like here in this case. That's the darling at Meninde And it also is another another aspect that data gathering from remote sensing in these very small systems So a very narrow rivers is sometimes not possible on the low flow conditions. They might be smaller in width than the satellite resolution is so we lose the capability to observe them from space or only in Only have the capability to have that spatial resolution from the we are damn or Next slide please So what we can do at the moment is for here an example for Lake Yume We have other levels on the left hand side, which we derive from remote sensing We have a 3d 3d models what we can have 3d models or 1d models for hydrodynamics in the lake Driven by metrology the symmetry That means lake level in an outflow the hydrodynamics the physical system is very well known You can calculate the currents mixing certification and then we put on that on top of that an algal model The algal model now it makes it starts making it complicated because we don't know exactly If we if you go there what sign of bacteria species do we have so what physiology what temperature dependence do these species have on the functional class and here comes in aqua watch or the ambition of aqua watch to be able to To to discriminate between species which would allow us to draw back information about this physiology and Get the the uncertainty of our models smaller And the red box there risk analysis for science and that's of course where we want to go But that's the next step on top of it because we can't Control science or see signed up inside of a kid sign of toxins from space It's very hard to get Consistent long-term data sets to do the modeling directly on it when it's not very the causal Relations between toxin production and environmental factors are also not very well known But anyway, we are on the way there from green yellow to the red to solve those next slide, please And on that path we use Different system we use in-situ data. That's the colorful dots here off in this case chlorophyll a cell counts of cyanobacteria and ligand We use modeling to calculate and simulate those cell counts. That's the black line And we use hydrospector data to calibrate our model and see how well the hydrospector Um shows us What we model and what's measured these are the blue and black dots and for like you these these are remarkable remarkably good In terms of we can even see daily cycles Of bloom formation Reacting directly on the internal mixing dynamics of a water body next slide And that's a blown up Um Sumo version of that one where we can see the daily cycle of uh, how much is that a month 30 days um The dots again are hydrospector derived cell cones the lines are no um The simulations Where we know in implement forecasts Based on hydrospector derived data. So we take one day hydrospector point Throw it into the into the forecast model and it simulates one of those colorful lines and then next day we do the same um And that results in this variation, uh, which we see here So that gives us a sense of uncertainty which we produce By using a simulation model How well can we estimate cell counts a harmful algal bloom development? In such a like system, which is relatively well monitored Then you have to think okay. What is necessary? To do so if you go to other systems We don't have that many data to calibrate your your model and go ahead Next slide please The problem is if you want to take those models and scale up So we can't really throw these models in every lake and try to solve them because we can't calibrate them So we need to have other means how to do that One of those means is we need to have those data and use machine learning for example And combine them with the process models to avoid this calibration step However, if the red hand the red side on the on the right hand side, you see um available data for different water quality parameters across Australia Um and their their value Their mean value, um And you see the distribution is anything else than uniform It's very concentrated on coastal region inland. There's Nearly nothing. Well, there are data available, but they are not continuous They're not really long term which we would need to set up Forecasting models So that's that's our problem Um field now we need now to see How can we use those models those and those data's which are available from our water agencies? What are utilities? um And how can we then general generalize our? um Simulation models in a way that we can deploy them and transfer them from one point to the next and even outside of of the monitored region and we start that With a very simple water parameters, which we which are very very well recorded. That's water temperature the mutual means also stratification mixing uh Electric conductivity turbidity tss So we have values for those and then it goes to to less frequent measured data like chlorophyll eye pigments new trends And potentially toxins and pathogens I'll show you some examples on the next slide what we what we want to do. Um So that's just a Summary again We look at river water temperature as a basic parameter for ecosystems At chlorophyll eye as a proxy of harmful algal blooms in rivers and lakes And at the moment also on dissolved oxygen um Which is derived as a From persistent certification. Um Now we use that to determine potential fish kills next slide So the water temperature we already have relatively good models available For lakes we saw that on the previous slides, but also for rivers. Um, um, where we have enough data We can calibrate them here. That's um That at the kiva Just before it goes into the Murray um The blue line in the middle. That's the observed data and the yellow line is the model data So pretty pretty pretty nicely matching We can use that for forecasting that and that model in principle takes nothing else than that air temperature and discharge To forecast that and air temperature and discharge is usually available on a large scale So we are able now to do that for or have trained that model for 300 stations in the Murray Darling basin And we're now looking at generalizing that with machine learning methods to be able to To in principle transfer those from the gaging stations to any stream section in the river network next slide For chlorophyll a we try the same thing, but it's getting harder because we don't have that many data available Um, however, we have quite a few data Many data available from remote sensing. So here what we do here is in principle instead of using a simulation model um process model and in-situ data using remote sensing data and in-situ data To derive a machine learning model to forecast chlorophyll a in rivers next one And that's an example How we do that on the left hand side if we have a river station and grab samples on a Hopefully weekly basis Then we need to derive from the satellite the chlorophyll a value at that point However, we can't really use a single point because that would Give you quite a big uncertainty. So we use along the river a line where we extract the The info information from the satellite over 300 meters upstream of the monitoring point um And we use the center line because we have also the problem that we need to Specify that system in a way that it works under high flow as well as under low flow conditions Um, and we always using water pixels and not then pixels by chance And if you do so we get the picture on the right hand side. So the the red dots are um Seven years of data That's swan hill in the merry Seven years of data of chlorophyll i on the red dots and the black Black crosses i'm principal war one day ahead four casts one week ahead four casts of the chlorophyll i using remote sensing data and the trained machine learning model and the gray Is the uncertainty of that? So we are getting there We can do that only at the moment for for the for the merry and tributaries But we're trying to expand that to other areas where we have or find enough chlorophyll a data to do so Next slide please There's a summary. Um We have a lot of water quality risk Problems which are across large scales and And currently they only treat localized on an invent base scale and aqua is trying with these models And these systems to to broaden our knowledge and pass them on transfer that or the transfer learning pass them on to other Positions other states other lakes other river segments across the continent and even go on a Entitled and entirely continental scale And for that we're using so-called hybrid models. So we we don't Only use machine learning Um and guess something but we constrain that with process process understandings the process models and remote sensing Data which gives us the ability to look on a wider scale And then we can have that holistic view when we're and why things are happening for some water quality parameters and that gives us then The possibility to go to water agencies to give them That for for their planning We have we can base early warning on that Which leads to risk minimization for water utilities water intake When do they take the water in and how do they manipulate that? Mitigation strategies can be developed because we we're not only looking at the single point but also On a catchment scale probably Uh Yes, I think that was everything I had to tell you for our system At the moment if it's Only trained on our pilot systems But as you can have seen we are already starting to Expand that to to a larger regional at least system With the aspiration to go in continental this and including other water quality parameters as those relatively simple ones Which I have shown. Yeah, thank you So cloth just before we move to the early bird questions Have you done any work with um the rivers institute and professor david hamilton? I did So because they've got that global lake monitoring network that he presented on I would have thought they'd be fantastic baseline data for you to use and that was Across hundreds and hundreds of lakes globally Yeah, so that's glion So we're working together with david There's also a little bit more Going on probably in the future together with uh a project david And also sorrow and wolf with in forecasting water quality parameters which is based in Virginia tech university funded by the national science foundation of the us um They take that into account so that that's As you said that that's an invaluable Source of information data information where we can train those those models Yeah All right. Well, thanks very much for that. We better move to questions um We've only got a fairly short amount of time for questions today because a few people have to leave Um in about 15 minutes So let's get into it Question number one as a driller installing dozens of monitoring wells a year How can we get involved and suggest sites? It gets about groundwater this one. You're not really into groundwater yet. Are you? No, not yet. Uh, I think um having having said that I think we are also getting quite a bit of interest from uh remote rural communities who have to rely on groundwater for their drinking water and um We might bring that into aqua watching due course, but we have to find a way to to put sensors and be able to predict that into those water um drilling tubes and the and the The the tubes that come out of and provide that the groundwater for drinking as well So, um, yeah at the moment, it's mostly surface water and coastal water. We're focusing on but Yeah, certainly that's something we can think about in the future as well I I think I should have probably also mentioned that I mean you've seen All the issues we still have to solve and the complexities of of the system. We're trying to model We're not going to jump and provide Aqua watch service right away. So I mean we our goal is to have A number of the pilot sites well, uh Tested and and doing some of the modeling in about 2026 um and hopefully Gradually increase the coverage of aqua watch To the whole continent from then onwards. So we're we're being realistic in in the fact that we're Really trying to do something very very complicated, but hopefully With big impact if we make it work So slowly um, so I think you've answered question to during the presentation Um, was there anything further you wanted to add to that? Um, yeah, yeah, go ahead Tim. Sorry. Yeah, I think I thought I could see something here I think certainly the plan is to be able to offer information at different levels as alex indicated, you know, there are a number of Sort of apps or applications that we envisage some of the outputs of this datron and some of those would be to a level um, let's say of community interests or Um, you know, perhaps the individual who wants to know, you know, I want to want to take my family water skiing on this Reservoir is there an elbow bloom present? Yeah, et cetera. You know something like that I think another way we're really keen to engage is through the development of citizen science Um, so and that's a really good way of way of engaging a community um, not only in contributing data to this um to aqua watch itself, but also in Erasing their awareness of water quality issues in their local areas You're talking about water watch Would water things like things like that? Yes. We we are developing a uh, another app which is called eye on water which uses The mobile phone camera to measure the color of water and from that color we can relate that to essentially uh A relationship with water quality. Let's put it that way So that's another way everybody carries a mobile device these days So it's a very easy way for us to capture data and to encourage people to be thinking about You know, what does this water body? What does this water look like? You know, does it look does it look swimmable or Or uh, is it green or is it turbid? Um, as a brown et cetera. Yeah Excellent question three. How does the forecast perform? What is the data requirement in developing this? forecasting system Very hard question to answer um Of course as many data as possible because the more data you have The smaller the answer it becomes um There are some some basic requirements For modeling sis for for pure modeling. Uh, you you need of course several seasons of of data to um to calibrate your your model if you then go um to To scale up um to try to transfer that to other systems Then you have to imagine a single season will not be uh, uh enough because there will be Different geoclimatic regions where you want to apply that. Uh, so you have to expand that Over let's say multiple seasons multiple locations And then that increases of course the necessary necessity of of more data for that not only by Editing up different stations, but also increasing the size the length of those stylus To to be able to forecast that on a more global term Question number four. Is this approach likely to be able to accurately check on inshore turbidity in tropical areas? Satellite versus in situ sensing Yeah, I can have a crack at that. Uh, yes, this is certainly an aspiration So, you know CSIRO has a has a sort of uh 20 25 year history in the development of Water quality monitoring tools using satellites for for water quality particularly focused on the Great Barrier Reef So so our algorithms are we're confident our algorithms deliver very very good data for that We are not far off the production of a kind of continental scale Turbidity or suspended sediment map for Australia that's going to be at fairly coarse scale based on On modus data, of course, I mean one kilometer data It's a it's a machine learning algorithm that we are just trial. We're going to roll out as a trial essentially So, you know in our plan is our intention is to really do this for all for all of Australia In terms of in-water sensing we we don't have much north of North of the Fitzroy. We have an installation near Townsville at Lucinda Jetty Um, so there is a lot of the lot of the north that could yet be instrumented in in coastal areas. That's certainly an aspiration Uh, but but yes, uh, you know ultimately we we want as uh, we want aqua to be able to produce Outputs at as high a frequency as possible and as high a spatial resolution as possibly for the entire for the entire coastal Coastal coast of Australia why is it um Why is it so coarse this one kilometer grid? um it Yeah, initially it's it but you're rolling something out at high spatial resolution across something You know something with a 36,000 kilometer coastline is is a lot of data basically and so This is this is really just a trial And this this data is very is Motorstater is just available from nasa and it's a it's a 20 year time series from 2002 um So it's a very good one to develop these tools initially on to see you know Is is is this actually going to work? Is it reliable? Is it accurate? It gives us a test and then we can begin to think about how do we roll it out to a higher higher spatial resolution sensors it'll be interesting to see um how this gets applied in the future like Trying to look at the effectiveness of land use change and catchments and things. Yeah Number five, how will could this link in with existing real-time monitoring already in place? so um I could have a crack at this and this relates to questions five six and nine. I think are all Uh are all quite similar as I stated in my my section of the talk, you know, we're very interested to hearing from different groups who have have existing real-time monitoring in place um That we could contribute to that um to that survey we're currently undertaking Um of of existing networks and so we've been very keen to hear from people Uh who have those in place? um And as I said our our intention is not to Not to reinvent the wheel here. Certainly not not planning to come in ourselves and and install sensors when when when there's perfectly adequate sensing technologies and probably better sensing technologies employed um already This is already Yes, that too Richard um sorry in terms of Last track of my question actually apologies. I've got no one um The next so you've sort of gobbled up a couple of questions there Discuss the CSIRO's ADIAS platform if possible um Yeah, as I mentioned before it's it's it's um ideas or ADS system as we're calling it now acquired data system is basically the The the the big cloud computing infrastructure platform that's bringing all these different data sets together to prove and eventually Also run the modeling the forecasting Uh using ingesting a lot of satellite data as well as in-situ data that's hopefully coming in um Almost in near real time into the system as it's it's it this is churning the the data and doing the forecast um We run the system at the moment using amazon web services. So it's massive cloud computing infrastructure um where the system uh is been implemented using um python Notebooks if for those of you who are into into sort of big data analytics, it's it's the language that is used a lot for this It's most of it is open source um and uh It allows you also to visualize the results at the same time with the right routines on when you write these programs um What else can I say it does cost us to run the system because we have to pay the The computing costs to to the cloud computing provider but our analysis And those of others before us Suggests that running a lot of this big data ingestion and crunching On the cloud seems to be much more cost effective than spending Tens of millions of dollars on a supercomputer that you May have to replace Every five or ten years Or upgrade whereas the cloud is constantly being upgraded right and we the other thing is the benefit of this system is It can run on any Part of the world effectively because it runs on the cloud And it taps into global data that we can access from nasa and the european space agency so Tomorrow we might have a project funded by the gates foundation in africa or something like that And it's quite quite easy to Set up a system that does just that So that gives us that flexibility But if people are really really interested in sort of deep deep tech uh cloud computing we can definitely set up a separate webinar just on that topic With our data experts Okay There's a question there about groundwater quality in mining areas Is there anything that's um specific to mining that you haven't already covered? So Just water quality in general rather than groundwater, I suppose Yeah Yeah, go ahead Klaus um So alex already mentioned um if it's water quality, it's about the water quality sensors which you would need to to have There to to do any significant water quality modeling forecasting however, um There are several There's the possibility that you can monitor groundwater surface water interactions by Looking at changes in the in the water quality in there In the color of the water you can see and especially in the mining in this sector there is If you're not looking directly at let's say metal contamination In terms of drinking water concentrations but in terms of What's the the effect of let's say the through flow from Ground water through tailings and then into the river which brings let's say metal Components into the river which oxidize and this oxidation products They're really colorful and you can see them from space. So you can could monitor turning dams relatively easily Um, but you also could look at let's say the potential Breaks of turning dams Because you could see the color changes in the downstream rivers So that that that's a possibility which we can already do or could already do with the existing aqua watch system or the remote sensing Yeah, and I think there is quite a bit of interest in You developing similar sort of sort of novel sensors. I think for heavy metal detection in the field But I guess we are not there yet, but it'd be fantastic to partner with Some some of those online If if there is an interest in in working with us on some of those applications I think there'd be a huge amount of interest in that from the People involved with compliance monitoring alike um Before we go any further you were going to There's a you're having a meeting to bring together Potential partners and things. I think mark wanted us to Just mention that Yes, we do we do because I think um While and then something that perhaps wasn't clear in our presentations We we the ambition is to make a lot of this data available like you have data from the bureau of meteorology Freely available. It's a public as a public funded service, right? But the we see a lot of potential for The data to be value added further for specific industry sectors or specific applications where where we probably think that private industry Should go or is already involved with so we want to bring a lot of those Companies and and potential users and value adding Data service providers into a meeting That we're also coordinating In trying to set up in the next few months Just to have that conversation. How can we then Make the data available and what what can people use it? Is it useful? That sounds good. So how do they find out about the the meeting if they want to attend? We'll we'll how about we um, we forwarded the invites to you Richard and Hydro terra and you then distributed to to those who might be interested is would that work? I think so. So we could circulate it out to the people who've Yeah, hold in this webinar as as a starting point and then If the people listening feel there's other people who might be interested they could forward it on to them Yeah, and we'll let and yeah mark mark boundary would be the point of contact for that and I think he's online as well Okay Now I was conscious that a couple of you had to leave About 10 minutes ago We've got 12 questions left. I think we're Probably going to just have to do a lucky dip on say one of them um And respond to these other questions by email. I would suggest is the best way to go um So maybe we can pick Goodness, maybe the first one. Casey price Is their plans to expand the location of in-situ sensor packs? I'm sure they are expensive But if not, is there an opportunity for local river lovers to fundraise for an in-situ sensor in their local river? We are really seeing the power of local communities who care for their rivers right now And I believe this would be possible if you would be open to expanding the network. That's a good idea Yeah, good idea. Um, I think um, definitely, um, we have sort of an optimal sensor pack list of the sort of sensors that that That we would need and that klaus for instance would need for his modeling um Some of which are useful for validating the satellite signal, but also Measuring things we can't see from space, of course as klaus was saying so um, we would be happy to at least pass on that list and you know, the the data analytics people the ideas team is Constantly trying to figure out how do you stream sensors from different vendors into aqua white? So we're not We're not going to be recommending a type of brand necessarily, but the type of measurement that would be useful um And uh, happy to yeah have that conversation and tim and his team probably would be very interested in Collaborating with you on that and creating a much bigger network of sensors across australia. Yeah I think if you can formalize that Specification from the bundle of sensor types Yeah, and you know the accuracy etc that you're needing That would be really powerful. We'd love to be involved in sort of helping to promote that. That's for sure Okay. Yeah, we'll pass that on definitely um I need to go to catch an aeroplane And others need to leave the mentor of left. Um, I suggest we respond to these other questions by email but Thank you very much to everyone who's Attended today. It's been fantastic and thank you very much to our presenters really impressed with what syros Taking on here. It's a very big challenge, but a a fantastic one to be taking on So really appreciate three of you presenting today. It's been Really informative and great to see so many people here as well So thank you to you and sorry we couldn't get through all the questions today Thanks very much for the invite and thanks everybody for the questions and the engagement. I look forward to talking to you a bit more Thanks guys. Thank you