[Image of a split circle appears with photos in each half of the circle flashing through of various CSIRO activities and the circle then morphs into the CSIRO logo]
[Image changes to show a photo of a crop in a paddock and Guobin can be seen in the top right talking to the camera and text appears: Digiscape FSP Seed-Funded Project (2020-21), Forecasting surface and groundwaters availability, David Robertson, Guobin Fu, Olga Barron, Geoff Hodgson, Andrew Schepen, 22 March 2021]
Guobin Fu: The presentation I just acknowledge all these co-authors for this piece of work.
[Image changes to show a new slide showing a map of Australia, a bar graph and text: Background and Objectives, Water use in Australia, Agriculture uses ~70%, ~15-20% groundwater, Conjunctive use of surface and groundwater, Seasonal use influenced by availability, cost, quality, Climate and hydrological variability, Predictability, Seasonal forecasting services for, Climate, Streamflow, is it possible to produce seasonal groundwater forecasts?, Objective – Develop a proof-of-concept system for forecasting surface and groundwater availability at seasonal to intra-annual time scales]
So the background for this project there is the water use in Australia above 70% is by agriculture and within this 70% probably 15-20% from the groundwater. So really, we need to know the water availability from both surface water and groundwater. So the right figure is where this irrigation are in the Australia. So as some of you know, the CSIRO Land and Water working with the Bureau of Meteorology have developed one of the forecasting model in last decades.
So, if you go to Bureau’s webpage you can see the seasonal forecaster for the rainfall and the streamflow within three to twelve months in advance, so that’s, that’s the team study. So the simple objective of this work is can we extension this model to the groundwater domain?
[Image changes to show a new slide showing a line graph showing rainfall from the years 1970 to 2020 and several formulas and text appears: Preliminary Analysis – Understanding groundwater process, Regression between groundwater level and lagged accumulated rainfall anomalies]
So before this, let’s do so most of the groundwater forecasting in the literature trying to link the groundwater level to the, some kind of variables such as this model with the rainfall. You see it’s not really bad to start with, so the red line is accumulated rainfall and the black one is the groundwater level. You can see that it’s not bad model performed but suddenly something has change. Next, please.
[Image shows another line graph appearing on the bottom of the slide showing a streamflow and groundwater levels plotted and Guobin can be seen talking in the top right]
So the reason is, because the year 2011, there’s a huge flood event. Yeah, the rainfall is good, but if some flood event happen that model will fail so that means we need a link, while we do a groundwater level forecast, we also need to look at a surface streamflow. Next, please.
[Image changes to show a new slide showing a flow diagram of how the data gets to the Ensemble Streamflow Forecast below a text heading: Developing/Modify forecasting system]
So this is the framework for this project. So the first two lines, basically is an existing CSIRO model developed by Di Roberts group so you get some coupling from the GCM, so get the rainfall forecast and you put the rainfall forecast onto the surface rainfall runoff model, get the streamflow forecast. So the really new for this project is a surface line, so how we link the output of the surface hydrologic model to the groundwater storage and to forecasting the groundwater level.
[Image changes to show a new slide showing a groundwater level prediction graph and Guoubin can be seen in the right talking to the camera and text appears: Groundwater level prediction and uncertain…]
So, this is very simple to test to try to link the streamflow simulation component to the groundwater level. So after we’ve fixed the model, we can run the model 10,000 times to quantify the uncertainty.
So the red data there really is observed ground water level. So the dark blue there is the 50% range our 10,000 times. So the light blue one is a 90% of quanta range our 10,000 graphs. You can see the red give a, the model performed very good. Next please.
[Image shows a new line graph appearing on the right of the screen]
So this is a sim, we just plot to the one-to-one and see how the observes compared to the model performed. So the left one is a time series but this one is just a time just from the life model large, so it’s the same thing here, next please.
[Image changes to show a new slide showing three graphs showing the Rainfall, Streamflow, and Groundwater, and dates appear across the bottom and text appears: Initial Products – One Issue Time]
So the output of this project is use any time slides we trying to forecast the rainfall which is the up panel and the streamflow which is the middle panel and the groundwater is the broad band. So the left arm, orange colour is really observed up to the time. So the right panel is the future forecast. So as before the dark blue one is 50 percentile quanta of the range and the light blue is the 90 percentile and the red data as we already have observed we put there how the model performed. So if you look carefully you can see the groundwater forecasting is slightly better than the surface and the rainfall and also where your time goes.
[Image changes to show the same three graphs moving in an animated way through the months and text appears: Forecasting Products, Animation]
For example if you forecast next month, two months in advance, probably it really is better than your forecast next ten, 11 months in advance. Yeah, so the model can start in any of the point, so another 40, 50 years date. We start every month to forecasting the rainfall, streamflow, and the groundwater in 12 months in advance. So it’s one, two, three, four, yeah. So this is really, is our final product for this project.
[Image changes to show three diagrams showing the Rainfall, Streamflow, and Groundwater level and text appears: Forecasting Skills, Continuous Ranked Probability Skill Score, Blue – Forecast better than historical distribution of observations, Red – Historical distribution of observations better than forecast]
We look at the forecasting scale so the colour, if it’s a blue colour, it means our forecasting model is better than the historical distribution of the observation. So, if you don’t have any model, the simplest one is that you use a historical distribution. For example, if you want to know July rainfall, for example, for the region, you look at last 50 years July’s rainfall. So you pretty much get a historical distribution. So our blue one means our forecast model is better than the simple historical distribution and the red one is worse as it means the historical distribution is much better.
So if you look at the rainfall, we pretty much have some skill forecasting one month in advance but if you go to two or three or four month the skill is disappeared. But for streamflow, we probably can do like three months to four months forecasting in advance, and for ground water we’re pretty much confident up to twelve months in advance. It’s a little strange for the streamflow starting from October or November, that’s because it is the region’s seasonal rainfall.
[Image changes to show a new slide and Guobin can be seen in the top right talking to the camera and text appears: Summary & Next Steps, Conjunctive use of surface and groundwater important for agriculture and towns, Approach to joint forecasting surface and groundwater availability, For case-study, Surface water forecasts skilful to ~3 months, Groundwater forecasts skilful to long lead times, Understanding and modelling process important, Additional data likely to improve performance, Groundwater abstraction, Calibrating model to groundwater storage, Next steps, Impacts (e.g. journal paper, conference/meetings, federal/state government, etc.), Evaluation on other catchments, Customer discovery (Market size, Products, etc.)]
So the summary of this study is, yes, we can do the conjunctive use of groundwater and surface water because they are important for agriculture yields and also for towns and for other water use. So this one is demonstrated our concept, it’s proved our concept that we can extension the existing CSIRO model to do the surface water extension to the groundwater domain. So for our key studies then, our surface water up to three months we have the red here with good skill, but for the groundwater we have a long bigger time up to 12 months to have a good skill. But the model can be improved in several ways.
For example, we still don’t fully understand the physical process inside. So because, as I said, this just a very short six months quote FTE project. So we find this interesting result, so we plan to write the paper, journal paper this one and also this one, method is the model only applied to one of the catchment, so it’ll be interesting to see how the model performs for other catchments and also exploring potential customers like is the State Government planning interest, or is the farmer planning interest, or is the town water user planning or management interesting for this forecast. So we are, we are we’re looking for the potential customer to apply this model on and approve this method. Next one, please.
[Image changes to show a new slide and Guobin can be seen talking in the top right and text appears: Acknowledgment, Groundwater advice – Matthias Raiber, Sreekanth Janardhanan, project scope and management – Andrew Moore (Digiscape FSP), Graham Bonnett (Drought Mission), Roger Lawes (A&F), Thank you for your attention]
Yeah, before I can answer any questions we get several colleagues help, so from the groundwater domain, also from the Project Management and Scope especially mention the Drought Mission because we’re trying to see how this model can be linked or extension used of the Drought Mission. So that’s all my presentation. Thank you.
[Image changes to show a white screen and the CSIRO logo and text appears: CSIRO, Australia’s National Science Agency]