This content is current only at the time of printing. This document was printed on 22 April 2019. A current copy is located at https://apvma.gov.au/node/19451
You are here
Transcript for Chris Lee-Steere
The runoff risk assessment framework
This presentation was delivered at the APVMA’s science feature session on 15 October 2015. The full video is available on our YouTube channel.
Thank you and thanks to the APVMA for the opportunity to do this. I'm talking about the new run off risk assessment framework. It's recently been out for consultation by the APVMA. I think the consultation period is finished but I haven't seen any comments yet so I'm just going to run through the background to this and try and show some methods about how it can be applied. It had its genesis in that magical review of diuron and I'm sorry to the regulatory affairs manager from DuPont if he's still here but it really did help us develop some good stuff. There he is.
Essentially it's a three‑step process and I'm calling them steps rather than tiers. The first part is actually a low tier and it's running a generic model, but it's a model to give an edge of field concentration that gets put into our standard water body. The second step and the third step start looking at likelihoods and probabilities. The second part really is focused on rainfall, while the third part is focused on characterising the receiving waters within the regions we're looking at.
I guess one of the first things to note is there's no actual international harm minimisation in this area on the use of a particular model. Conceptually we all do the same thing but the Americans have been using their GENEEC program for a long, long time. The Europeans developed FOCUS a while ago and they continue to revise that and bring in their various scenarios. We use a model based on the outcomes of an OECD workshop, I think back in about the year 2000. The Department of Environment brought in the various equations from this workshop to develop this front‑end model.
There were limitations to that and when we were looking at it through the diuron review we started to address those limitations to try and get a bit of extra flexibility. The main one at that time was to start to use look‑up tables to get values for rainfall versus run off from a couple of different soil types, wet versus dry soils or covered versus bare soils. This is still a front‑end model and at this point it's difficult to build in too much context. The results we get from this are very much screening level that give us an indication of whether we can stop an assessment or then have to start to bring in additional context.
This is the primary equation in the front‑end. To me the most critical component is the run‑off value and the amount of rainfall or the rainfall value and the amount of that that will leave the site as run‑off. The model itself is quite good because it's used in a national context so that the smaller the number of input variables we can have, the less reliance we have on default values. This one takes into account only a few chemical‑specific values and a couple of geographic values so we can account for the slope and we can account for the half‑life, absorption potential and we take an eco‑toxin point for the chemical.
This run‑off over rainfall ratio is much more difficult to address and the department, when they set up this model, set that at a 20ml run‑off based on a 100ml rainfall event. There was no flexibility in that. Looking at the look‑up tables that were available from the OECD workshop we could start to develop different ratios for different rainfall volumes for different soil types. There are still problems with that. It only looks at soils classed as sandy or loamy and it doesn't allow us to look at soils heavier than loam so that it potentially underestimates the run‑off from that and that was an issue raised again through the diuron review but we couldn't really address it at the time because we didn't have the information.
This was one of the main problems I guess with that front‑end model. It assumed a run‑off event happened three days after application, which is perfectly fine but the outcome of it made the assumption that all risk was equal so if you're applying to a wheat field or a banana crop or if you're applying in the height of summer or the height of winter the risk was deemed equal. Clearly that wasn't the case but we didn't have the ability at that time to build in the additional context.
I've been getting rather side‑tracked with this for a long time and I might just skip to another slide. After this paper got put out for comment, GRDC and others published through a website called soilquality.org.au a series of results on soils in grains‑growing regions. One of those components they looked at was the clay content. I got quite excited about this because the USDA apply curved number approaches to their run‑off calculations and they have four classes of soil for each use pattern. Those soil classes are based on clay content. Now with this data we can group our own soils from the different dryer land regions to marry up with the USDA curve numbers. They essentially give us the rainfall run‑off relationships for all the different curved numbers of which they've got about 100.
If I go back to this one, now we can get composite curve numbers for the different states based on their own soil characteristics and develop run‑off curves for different uses within overall scenarios. This one just looks at dry land and there's a number of situations there than can be assessed depending on what crop you're looking at or what stage it is. This will be the stage where we can bring in additional scenarios. I was talking to Rowan Rainbow at lunch and he was talking about limited traffic on fields and how that helps reduce run‑off and that would be the type of situation where if we could get additional data we could certainly bring in the equations into the model at this stage.
I've shown you the WA one, very, very sandy‑dominated. There's Queensland, very heavily clay‑dominated. The run‑off for everything else being equal is always going to be higher in Queensland situations than WA but that's not the end of the story. This is still all that front‑end model. What I wanted to show here was the edge of field concentration, not what goes into the standard water body which is our one hectare surface area and 15cm depth. Looking at the 100ml rainfall and 20ml run‑off default scenario that has been applied, the edge of field concentration is predicted to be a lot lower than these new revised scenarios. What you've got to realise, I guess, is that doesn't translate necessarily to higher receiving water concentrations because if you look at the rainfall values down here, these are the rainfall values that the model says gives you the peak concentration in the standard water body and this is the amount that's running off, and the run‑off levels are much, much lower than they are in the standard scenarios. Even though these values are much higher, you're getting less water over the 10 hectare field running into the standard water body. In actual fact, when you put it all out the receiving water concentrations turn out to be not too different at this part.
This gives an example of an output at this first part of the assessment. This is just a hypothetical here but looking at a dry land situation, application to grain, I've put poor hydrological condition for the soil, just a bit of a default, and you can see these rainfall values designed to give the peak concentration, these concentrations area a lot less than the edge of field. That's because as I say they're being distributed into 1.5 megalitres of water in the standard water body and in this particular case we're getting risk quotients exceeding one at step one.
This is kind of where we were before the development of the framework where we could fiddle around perhaps with slopes, maybe try and refine the ecotox value but it became very difficult to mitigate risk further once we got to this level. The framework now allows us to do that and the second part of it, as I said, really is a likelihood of rain and it is essentially looking at two variables, a wet day versus a dry day, and if we do get a wet day how much rain happens? I've given an example here of two extremes just to demonstrate the point but what we're trying to do is say, find out the rainfall value you need before you get an unacceptable risk in the standard water body and then work out the likelihood of getting it.
In this particular example we're looking at a 12mm a day rainfall event, looking at Hay in the New South Wales wheat belt and Innisfail u in the wet tropics. I've taken the data from the three wettest months of the year for each area, the winter months in Hay. This is based on over 100 years of daily rainfall data. You've got about a 27% chance of getting 0.1mm or more compared to over 50% chance of getting a wet day in Innisfail. That's the first part of it.
The second part is to look at accumulative frequency distribution of the positive rainfall. Here we've got 100% chance of exceeding zero millimetres because we're only looking at positive rainfall. What's the chance then of exceeding 12mm a day? In Hay it's only about 7% on the days that it rains you don't exceed 12mm in every 7% of the time. Again, in Innisfail you exceed it about half of the time. You end up being able to, I guess, get an overview of whether you've got a risk of the rainfall event happening and if I was doing an assessment and this was the outcome, I would conclude here that the run‑off risk in Hay was acceptable and I wouldn't continue with the risk assessment but in Innisfail the risk would be too high.
Now in the discussion paper the trigger value is a 10% combined probability. I guess ultimately it's a policy decision but it was put forward at the moment as the trigger to move up to the in‑stream analysis.
In‑stream analysis now starts to look at characterising the receiving waters in the environment of the regions we're looking at. We're moving completely away from default scenario assessments. It marries up the real world stream flow data which had been taken from monitoring stations with at least 10 years worth of daily data. Ten years might seem a lot but by the time you separate it, I do it by season for dry land, I do it by month for the tropics and the sub‑tropical areas, you end up getting a much reduced data set so 10 years or more was the cut off. The long‑term rainfall data get developed for each region. They get married up so we can look at probability of rain compared to how long a particular stream's in flow for and work out base flow components versus how much of the flow rate is coming from run off. We just use those then to develop our theoretical in‑stream concentrations and we get a distribution of those theoretical in‑stream concentrations for any particular region.
They are theoretical because while we know what the flow rates are, we know what the rainfall is, we don't know which if those streams is going to be exposed in any particular farming situation. All we know is that particular activity occurs in this catchment and this is what the receiving water looks like that it could go into.
That's a map of the dry land cropping areas. I still don't know what dry land they grow up there. If anyone can tell me I'd be grateful to hear it but this is front the Australian Bureau of Agricultural Resource Economics Data. I get a lot of information from then on slopes and growing regions and catchments. That's just an overview of the amount of data we could actually get for dry land and there's quite a bit, lots of catchments in the different states. There wasn't really sufficient data from South Australia, and the Victorian stream flow data are probably a useful surrogate because the soils are quite similar but then we use the State‑specific rainfall data. There's well over 100 different stream flow values or monitoring stations in Queensland and Victoria, over 240 in New South Wales; not so much in WA but there's just really not that much water there.
This just kind of brings it all in together. There's still a lot of conservatism built into this level three. We assume that the rainfall value happens over one hour even though it's a daily value. That maximises the edge of field concentration, which gets distributed into the stream flow. We do use specific slope values and the important thing about the slope values is the AB data, you can get slope values for both a region and for the growing area within that region. They can be quite different so I always take the data related to the growing areas in the region. We run the front‑end model and get the edge of field concentration. That gets distributed into each flow rate from each station. If each one of these stars is an individual monitoring station, we can use that to build up our distribution of in‑stream flow rates. Then it's a case of obtaining your regulatory acceptable concentration and working out from there what percentage of receiving waters could potentially have in‑stream concentrations exceeding that.
This is just an example then of the tier three output. I'll go back. There's the tier one. Each state had a risk here or a potential risk but by the time we bring in all our actual stream flow data and rainfall data and different percentile flow rates in this particular case we can demonstrate that we're never going to have a potential of 10% or more of receiving waters impacted. Importantly this doesn't say that everything is protected. It says anything on this side of the intersect can potentially have receiving water concentrations exceeding your allowable concentration but we can start to quantify now just how much of a risk a chemical might be. It becomes quite a robust tool for looking at spatial context and temporal context because now we can look at times of the year where it's probably not useful to apply a chemical or we can look at regions where you've got higher risks than other regions. Whereas in the past we were struggling to do that and an unacceptable risk just meant that we couldn't support the chemical.
Sorry, that was quite quick, but I just wanted to show you how we can bring in all this context now and start to come up with some fairly robust and informed decisions on run‑off risk. Thank you.
Errors and omissions excepted; check against delivery.