To subscribe for email updates from eWater, please enter your details below.
To subscribe for email updates from eWater, please enter your details below.
By Tony Weber BMT WBM Pty Ltd
Recently, I had to deliver a workshop to some senior staff from a major river conservation commission in China. The workshop set out to cover which models were currently available in Australia and what models might suit some particular issues that the Chinese staff wanted to explore.
It was an interesting day, because it really made me think about the reasons for modelling and how modelling results could best be used to solve their particular environmental issues.
Fundamentally, modelling is part of a decision-making exercise, and in order to make that decision, we need to know what the actual question is (sounds a bit like Hitchhiker’s’ Guide to the Galaxy doesn’t it?). But the question, or issue, is rarely expressed in a form that fits how models work. For example, one of the problems raised by the Chinese staff was poor water quality in their river. What model should we use for that? There are literally hundreds of water quality models out there, and they can model a wide range of constituents. The choice depends on the actual question needing answers.
In selecting the model, we need to define what question we want answered or what issue we need more information about. Its fine to want more information about water quality, but what is really needed - in other words, what particular water quality problem needs addressing? Is it algal blooms? Turbid waters? Sand slugs? Bank erosion? Toxicants like pesticides or heavy metals? Bacteria and viruses? Each one of these problems would require a different model or group of models to facilitate more detailed understanding. This is why modelling reaps best rewards when sufficient time is spent defining the modelling question.
Another recent example is some work I was involved in on the Tamar Estuary and Esk Rivers in Tasmania (see Volume 6 p26 H2O Thinking). In this project, the key issue that needed to be better understood was sediment in the Tamar Estuary. Previous studies had showed that some of the sediment in the estuary was coming from the catchment, but what was the source of this sediment, whereabouts in the catchment was it being generated, could anything be done about it and what was going to happen with an increase in agricultural activity in some parts of the catchment?
In the end, this became an ideal application of a catchment model, as we were able to use the spatial and empirical relationships provided by eWater’s Source Catchments to show probable causes and the areas of the catchment that were likely to be contributing sediment.
What the modelling also gave us was a much better understanding of the catchment itself, and a guide to some further questions that still needed answering. For example, while the modelling showed the steeper areas of the catchment which experienced higher rainfalls had the potential to generate more sediment, we weren’t really able to show how much of that sediment was likely to reach the estuary. The major reason for this was the lack of data about sediment movement in the rivers and the ability of dams to trap that sediment. While we could have simply referred back to the literature, without data we were unable to determine the actual quantity of sediment that was leaving the catchment, even though we could show the amount likely to be produced.
This brings up two big points. Firstly, often the reason to model is not necessarily to produce the magic answer (42, if you’re a Hitchhiker’s Guide to the Galaxy fan!), but rather to improve our understanding of the system being modelled and to gather more information about it to help in the decision-making process.
Secondly, often it is data, or the lack of it, that really dictates how far we can take the modelling task. We are very lucky in this country to have a lot of data relatively easily accessible. In my recent visits overseas helping to adapt eWater’s music (model for urban stormwater improvement conceptualisation) for the UK, it really became apparent how much we take this access for granted. For example, rainfall data can be readily (and freely) downloaded direct from the Bureau of Meteorology (BoM) website for many locations across Australia. Try doing the same in the UK or France for example and firstly you need to cough up significant amounts of cash (please don’t get any ideas BoM) and secondly, you’ll quickly find attempting to get the raw data rather than some derived product is extremely difficult.
That being said, there is still a dearth of some data in this country, especially if your field is water quality modelling. We used to have significant resources devoted to this in many state agencies, but sadly, very few new monitoring sites are commissioned each year and I’ll wager that there are more being closed than opened. Thankfully, due to the efforts of the Bureau’s Water Division we have seen a significant investment in water resource monitoring in the last few years; however what is now needed is a similar level of investment in ecological water quality and monitoring.
A good example of this is the recent flooding across the country. After a recent trip through the catchments in South East Queensland to look at the impacts of the event, I was amazed at the changes in the catchment. There is only a fleeting opportunity in which to collect data after such events and fundamental questions about the nature of extreme events mean a much wider monitoring effort in several catchments would help us better understand them. While the focus is quite rightly on flood recovery, if we wish to be able to model such events in the future, we need to invest in data collection now.
The same effort needs to be made around chronic issues, such as the gradual decline in ecology in some of our largest rivers. We simply can’t go out into the field over a few months and hope to gather enough information to be able to model such declines; the data collection needs to occur over long periods to give us enough information to start understanding the processes that may be involved. If we don’t have the data to make such a start, how can we even know which model to use if we want to improve our understanding?
Finally, we really need to think about what it is we want to do with the modelling results and the models themself once we have produced them. Do we want to communicate the actual outputs, and if so, who will be impacted or influenced by the numbers? In using such results, those who have built and run the model have a responsibility to communicate how useful the model was in producing the results. How good was the data behind it? What was its uncertainty? How useful is the model at reproducing what we see in real life? Often I have seen models built and run, a few PowerPoint presentations made (us consultants are good at PowerPoint’s!) and then the model is left to stagnate on someone’s computer because no one has really considered how the model is to be used in the future. Given some of the dollars invested in producing these models, such behaviour can sometimes make for some very expensive PowerPoint presentations.
In the end, this is really a call for a better understanding of the need to build a model. Modelling is an extremely powerful way of improving our understanding of the way systems work and we have some excellent tools (and the computers to run them) with which to work. If we think a little bit more about why we want to model, we will be able to properly capitalise on our investments in using them.
If you are looking for more information on choosing the right model, there are several excellent papers on this, including on the Toolkit site at http://www.toolkit.net.au/modelchoice/. eWater is also producing guidance on best modelling practice that will be very helpful when undertaking the actual modelling.
See our website for a new guide to Best Practice Modelling.