Transcript of the using climate data for insights on future production webinar

Heather Field:

Our webinar today is on Using Climate Data For Insights On Future Production. In this webinar, Craig will summarise the source of available climate data, including historical seasonal projections, and future climate scenarios, and share examples of how to use these data sets to support insights for future agricultural production.

Heather Field:

Dr. Craig Beverley is a senior research scientist with Agriculture Victoria, and has 20 years experience in the formulation, development, and application of numerical models to simulate natural resource systems. In his current role, Craig is developing a range of both biophysical models to support economic and policy analysis of landscape systems, including an integrated biophysical catchment model capable of assessing the impacts of land management on surface hydrology, groundwater, nutrient dynamics, sediment transport, and vegetation dynamics, under current and future climates. So we're really pleased to have Craig on board. I'm looking forward to your presentation today, Craig. Over to you.

Craig Beverly:

Thank you, Heather. And thank you for the invitation to talk today. As Heather says, I'm a senior research scientist with the Landscape and Water Science group and I'm located at Rutherglen. And principally, my focus is on system modeling. So, as Heather said, the title of the presentation today is Using climate data For insights on future production. And this topic very much relates to predictive agriculture. So the aim of today's presentation is to provide an overview of some current work we were undertaking, coupling climate data with predictive models, to hopefully produce industry relevant information related to future agricultural production.

Craig Beverly:

The outline of my talk is summarised on the slide, and includes an overview of the types and sources of various climate data sets. The construct of commonly used predictive models. A summary of recently modeled paddock to landscape applications with a focus here on accounting for climate impacts. Some indicative site specific results, followed by innovation strategies and conclusions. So it's a fairly full agenda.

Craig Beverly:

So let's start with some climate data sources and analysis. So as an introduction, I'd like to briefly discuss the three types of climate data's we're typically using to assess agricultural productivity. And these include historical climate data, seasonal climate projections out to say 200 days from today, and climate change projections. In terms of historical climate, there are really two primary data sources namely raw data, such as Bureau of Meteorology, or climate station observations, or from climate stations that might be installed at field sites.

Craig Beverly:

Additionally, there's interpolated data, and specifically I refer to the patch point, daily time series on meteorological data predominantly identified at BOM climate station locations. Now, these time series data sets consist of observation data where available, and interpolated data when observation information is missing. Alternatively, there's the gridded data, which is continuous interpolated time series data. Now, these records are obtained by extracting information at specific grid resolutions, and here the grid resolution is .05 degree latitude and longitude, or approximately five by five. And information is extracted from a range of different time series gridded climate rests. And this data set is often referred to as data drill. These data sets are available from the scientific information for land owners, or CSIRO, as it's known, nine data sets hosted by the Queensland Department of Environment and Science. So data from 1899 to present, is available from that site in various formats to accommodate commonly used climate tools and models. And I'll be drawing on some of that.

Craig Beverly:

So across Victoria, there are 750 patched point data sets corresponding to, as I said, those climate station locations. And there are approximately 9000 data drill or grid data points. This figure shows the locations and influence of zones that are attributed to each of those patch point locations. And the area of influence is based on the construct of a Voronoi, I think it's pronounced, diagram, which defines regions consisting of all those points, but are spatially closer to that climate station than any other climate station location. So effectively, it's a proximity map.

Craig Beverly:

Now in our work, we further spatially interpolate this data based on topography, so as to define a continuous move climate surfaces with no edge effects. As an illustration of the varying spatial resolution of historical climate data, the patch point, this figure shows the comparison in the spatial resolution between a patch point data, that's the one on the left, and also the associated climate proximity zones, and the gridded data, or the data drill shown on the right. Both these data sets have common format attributes. Typical attributes for both patched point and the data drill include daily temperature maximum and minimum, daily rainfall, solar radiation although it's limited in terms of its confidence pre-1957, relative humidity T-max T-min at those times, vapor pressure, and potential evapotranspiration estimates of which there were quite a range, whether it's FAO Penman Monteith, or Morton potential estimates, as well as a Class A evaporation estimates of a ET.

Craig Beverly:

I'd like to now move into the seasonal climate data sets, which we're also evaluating. Seasonal forecast data is used to estimate the near future climate ranging from seven days to 200 days into the future. The primary data sets we're using include the Australian Digital Forecast data sets, or the ADFD, which contains weather forecast elements produced by the Bureau of Meteorology, such as temperature, rainfall, weather types. And this is presented in a gridded format again, covering the next seven days. For Victoria, the grid resolution is approximately three kilometers by three kilometers.

Craig Beverly:

Alternatively, there's the ACCESS-G, I should say, which is the Australian Community Climate and Earth Simulation Simulator, it's ACCESS, where G refers to global. And this is also made available from the Bureau of Meteorology. This data set comprises 10-day forecast at a spatial scale of 50 kilometers. And lastly, the seasonal data which is the ACCESS-S, again provided by the Bureau of Meteorology. And this provides forecast climate data ranging in scale from weekly to seasonal climate outlooks, and is provided at a spatial scale of approximately 60 kilometers, and a temporal resolution ranging from really weeks to months. The climate projections are either daily climate attributes, or else can be rainfall and temperature deciles, and I'll be drawing on some of these later.

Craig Beverly:

So the ACCESS-S, as I said, operates at 60-kilometer resolution. And that comprises 11 high incurs, which are really climate ensembles or predictions based on forecasts derived from atmospheric and ocean response models. Importantly, the ACCESS-S forecast includes some perturbations in terms of initial conditions such that they can then derive a range of different forward projections. So for instance, based on a recommendation of using nine successive days to represent the distribution of potential climate for the near future, this results in approximately 99 daily predictions for each of the future 200 days. And that gives us a forward distribution. The daily attributes summarized in a table and consistent with some of CSIRO historical data attributes.

Craig Beverly:

As part of work were undertaken, we're evaluating the accuracy of the 200-day forecast. And this slide shows a comparison between accumulative measured and predicted daily potential evapotranspiration in this top left, rainfall as well as minimum maximum temperature. And the graph here shows the envelope of response. For both, I've got here a minimum and maximum which have a dotted lines as well as the 90 and 10 percentile based on those 99-day projections for each of the future 200 days. Also shown in the blue line is cumulate evolves as measured observations, and this is for a cited Ellinbank. And as expected I guess, we can see the temperature is most accurately predicted in terms of these forecasts, whereas rainfall is least accurately predicted. So we need a means by which we can improve the reliability of this data, and or provide more meaningful confidence when using this information to producers and other stakeholders.

Craig Beverly:

So like to now move on to the third type of data, which is the climate change data sets. Now, to consider the future climate, we first need to assess the relationship between elevated CO2 and temperature. So we accept that climate projections are collated and reported by the United Nations Intergovernmental Panel on Climate Change, or IPCC. And it is important to note that there are many global climate models. In fact, there's 42 currently, and soon they'll be over 100. And these are used to estimate temperature rise due to increase in CO2 emissions.

Craig Beverly:

However, the majority of the models currently agree that for the Australian trajectory, shown on the right on this slide, it is very much the business as usual pathway which is really the worst case scenario and predicts approximately 2.4 degree increase in temperature by 2050. This trajectory is referred to as RCP8.5 where RCP refers to the Representative Concentration Pathways. In Victoria, the primary data sources we are using and converted to daily climate sequences based on these CO2 projections. The high resolution five kilometer by five kilometer downscaled Victorian Climate Projections, released mid last year, which are referred to as VCP19. other data sets include the Pattern of Change data developed by CSIRO and the BOM, and these are annualized estimates from current to 2100, based on a 50 by 50 K resolution, or else the Victorian Climate Initiative data set, specifically for 2040 and 2065. Again, at a 50-kilometer resolution. And lastly, the Victorian River Basin data set, which are applicable to just set the river basins, dominant river basins within Victorian at a basin scale. So each data set you can see has a different spatial scale, ranging from 50 kilometers to five kilometers, as well as different temporal resolution.

Craig Beverly:

Regarding the most recent VCP9 data sets. This comprises of six downscaled GCM projections to represent the different possible regional, changing climates. Now the choice of a GCM was undertaken by CSIRO based on an assessment we had that ranked each of the 42 available models to best represent Victorian conditions. Having this range of models is important when assessing the envelope of probable and possible future climate scenarios for our regional projections.

Craig Beverly:

And this table just summarizes the name of the models as well as the developer, and the country of development, and a relevance to Victoria. And you can see that some models are more applicable to different extent, whether it's Northern Victoria or Southern Victoria, and perhaps other models. For each GCN, there are four daily scenarios available based on a bass line period of 1986 to 2005 for both the RCP4.5 and RCP8.5. Now applied here at the bottom shows the climate sequence for temperatures centered on 2030, 2050, and 2070 to illustrate the fact that a data set comprises of 30-year daily series of climate information from which distribution of likely outcomes can be derived. This means that when we report against any of these periods, it is not based on a single year of simulation. It's based on a distribution of 30 years as shown on the slide, and that's really important.

Craig Beverly:

The next few slides show the general climate trends for both rainfall, minimum temperature and maximum temperature, extracted from the data at various locations. In this case, it's the Gippsland for a low rainfall site or moderate rainfall versus high. And the general trend is very much the same across these various regions, where rainfall is shown to be decreasing through time. Here I've got baseline 2030, '50, '70, and '90. So rainfall is decreasing, and temperatures, both minimum and maximum, are expected to increase through time. And this is consistent across Central Victoria, although to varying degrees, as it is in the southwest region of Victoria. Not surprisingly. Importantly, information embedded in these data sets are relevant to climate thresholds that also affect specific agricultural enterprises. For example, the number of days above 35 degrees in each season. So this plot shows a predicted number of days above 35 degrees. On the x axis, we've got that summer, autumn, winter, early spring, and late spring, for the range in terms of future climate predictions 2030, '50, '70, and '90. And results show that in all cases an increase frequency in the incidence of high temperature. Now this information in itself is useful, and can be packaged at either discrete points, or presented as spatial maps.

Craig Beverly:

As an aside, a useful other tool, of course, is a climate analog tool, which provides an estimate of what the future climate would be for given location. The analog explorer tool, which is available at the bottom of the screen, was developed by CSIRO, and matches the future climate of a location with the current climate experienced in another region. And these analogues include either annualized estimates, or seasonal equivalents. Here I've got spring and summer, for instance. So for example, Warrnambool in 2030 will have a spring and summer similar to Albany, whereas in 2090 the spring and summer will more likely be Wangaratta or Esperance. And that can be quite useful if you were designing new agricultural systems.

Craig Beverly:

So, in summary for climate data, we can say that historical climates are available in different formats. Seasonal forecasts vary from seven to 200 days from today, with varying levels of confidence. There are numerous climate change projections, climate analogs are useful indicators of future climate. But we know that climate scenario is poorly described extreme weather events. And that's due to the downscaling process from the global climate models. It's also important to note that climate change input data, it differs significantly from state to state in terms of spatial resolution. In Victoria, we're fortunate to have high resolution data.

Craig Beverly:

So how do we use this climate data? We can either undertake data analytics based on the climate data alone, such as the climate fresh or old information I just provided, or we can use predictive models. And given this talk, it's framed in the context of particular agricultural. I'll quickly move through in terms of the modeling. So what is a model? As it says, a model is merely a simplification or simplified representation of a physical system or process, and the level of complexity of this conceptualization can vary significantly. So broadly speaking, models can either be characterized as analytical, or empirical, or mechanistic. Analytical models are based on specific mathematical constructs that are bound by quite strict assumptions. And this means that they have limited applicability when we're talking about future climates. Empirical models or descriptive models, in contrast, are based on empirical observations, such as... And can be therefore framed in terms of probabilistic or regression equations. But lastly, the mechanistic models are those that are shown on this slide to be numerical, are based on mathematical descriptions that couple individual elements of variables such as nutrient dynamics with water, and can be described in mathematical format.

Craig Beverly:

In this talk, we're very much focused on that last type of model construct. Namely, because that type of construct is applicable beyond the observation and calibration period upon which it was built, and therefore can be used for future climate assessments. And I'm going to draw on this in terms of model selection later. So a commonly used model, often for water resources, it turns out is the simple crop factor approach. And this type of model is often referred to as the FAO56 crop cover approach, and relies on the use of defining time varying cover term effectively representing the photosynthetic dry matter response for a growth season under unlimited water and nutrient conditions. Evapotranspiration here is simply a function of a reference temperature, potentially in this main annual crop coefficient, weighted by available soil moisture and other environmental conditions.

Craig Beverly:

Alternatively, a more complex model is shown on this slide, whereby the user must specify a range of parameters defining the leaf development, and consequently biomass accumulation, by default phonology and root growth. So leaf development is represented by this S curve, up to a time where LAI is max, that's Leaf Area Index is maximum, and there alpha senescence occurs. Daily increment in leaf development is a function of accumulated thermal time, or eight units, as well as stress factors and shape parameter. In contrast, a phenological model, explicitly models pheno stages such as fertilization, vegetation for instance, and includes a variety of different environmental constructs. In this example, I'm showing a pasture phenological model which accounts for different tissue pools. In this case, growing leaves, live leaves, senescing and dead matter, as well as litter and a root pool. And tissue turnover between these pools is governed by elapsed time or environmental conditions such as soil moisture. The reason I mentioned this is because I'm going to draw on these three types of model constructs. When considering climate change, it's also important to account for the impacts of elevated carbon dioxide. And this slide shows on the bottom left a typical response function, which shows that as CO2 increases from current day, this function is shown to also add some type of an increase from a value of one in this case to a value of two. The right-hand side really shows the RCP8.5 trajectory.

Craig Beverly:

Now, depending on what vegetation model you use, elevated CO2 impacts these types of metrics, whether it's photosynthesis, optimal temperature in terms of growth, specific leaf area or radiation and transpiration use efficiencies. And again, I'm going to draw on this. So in summary, predictive models there are a range of vegetation models that are traditionally used to account for climate impacts on soil, water, plant, and animal interactions. And in general, a lot of these models poorly describe plant persistence, fruit size and fruit quality. And it's important to recognize when using them.

Craig Beverly:

At this stage, I'd also like to make a distinction between paddock scale and catchment scale modeling. This slide shows the key components of a typical farming system models which includes the influence on climate in terms of rainfall, soil evaporation, plant transpiration. And I've got water movement through the soil profile and beneath the root zone. It doesn't account for any lateral connection or any connection to groundwater. And that's also important to recognize. The most commonly used Australian models would be GrassGro, which is for predominantly beef and sheep enterprises, DairyMod for dairy pasture based systems, and APSIM for cropping systems. And we have a model here in Victoria called the CAT.

Craig Beverly:

Typical outputs from these paddock scale models, which are value to users. Elements of the entire water balance. We might have growth dynamics in terms of photosynthesis or growth rates in terms of pasture depending on the model. And nutrient cycling as well as animal dynamics if they're accounted for. In this figure is taken from the DairyMod graphical user interface, and is a valuable tool if you're looking at paddock scale dynamics.

Craig Beverly:

Regarding seasonal forecasts that I mentioned earlier, AgVic is developing a user app to estimate available pasture mass, both spatially within a farm context, or temporarily at any selected paddock within that farm. And this slide shows preliminary modeling of the predicted historical pasture mass accumulation compared to observation information, and the likely range of future dry matter banded in the quartiles for each of those 200-day forward projections that I mentioned earlier. Note that the pasture dry matter predictions show significant variations of up to 4.5 tonnes per hectare over the 200-day forecast. And that's because of that distribution in terms of climate overlay that I mentioned. But based on this information, the user can make decisions aligned to either their management or risk profile. And we're developing that further.

Craig Beverly:

I now move on to catchment scale modeling, which includes as shown on this diagram if you can see, both the unsaturated zone, the soil zone, and often also includes connection to the deeper groundwater system or saturated zone. And the climate is clearly a key driver in both those components. So key elements not accounted for with the paddock scale modeling is that catchments again modeling must account for heterogeneous landscapes. That includes spatial varying environmental conditions, whether it's climate, topography, or soil, as well as land management land practice. It also accounts generally for runoff processes. And this means it links to landscape connectivity, which accounts for water and nutrient movement across these landscapes, including run on and run off, and air, and important processes. And calibration can also be extended to include matching site specific, multiple site specific, as well as in the catchment measurement.

Craig Beverly:

So provide a sense of the range of models and application to AgVic have developed to assess the impacts of climate on agricultural productivity and water, I'd like to just give you a bit of a snapshot. First, starting with the Gippsland Basin and the Otway Basin, there's been studies, which also include calibration or multi layered groundwater models. Similar studies have been undertaken across the Victorian extent of the Murray Darling Basin, also linked to a detailed groundwater model. Recent study in northeast Victoria aimed at evaluating the impact of climate change on the six dominant industries namely viticulture, cherries, chestnuts, forestry, grazing, and cropping, and I'll refer to results of it with links to groundwater. And when this study was extended to also look at irrigated horticulture in the Mallee to assess climate change impacts on citrus almonds, table grapes, dried grapes, Shiraz, and Pinot. And lastly, the evaluation of future climates on pasture performance across the dominant Victorian dairy regions. So this means that really we've got quite an extensive statewide assessment of the impacts of climate on both agricultural productivity as well as water resources. I'm going to draw on some of these assessments in the following slides.

Craig Beverly:

But firstly I'd like to focus on some of the key outcomes that may inform your future farm management and production systems. End product is a high resolution soil moisture mapping. This slide shows predicted top soil moisture for July and December as predicted by the Bureau of Meteorology as well as comparisons with the work that's undertaken with Ag Vic. The trends are actually very similar. Yet the Ag Vic has the added value added in core explicitly incorporates farm management, operates at a finer scale rather than the national scale, and can provide other information beyond just soil moisture, multi dimensional information such as crop yield, dry matter, sediment et cetera. And that's a resource that is available.

Craig Beverly:

So focusing on the impacts of future climates, catchment modeling enables this multi dimensional assessment. By that I mean on water resources, productivity carbon, and the other model metrics. For example, in northeast CMA regional Victoria, spatially averaged prediction suggests that for 1.2 change, or reduction in rainfall in 2030, results in a 23 per cent reduction in streamflow and a 20 per cent reduction in recharge to groundwater. By 2050, you have a 4 per cent reduction in rainfall, resulting in a 34 per cent reduction in streamflow and a 40 per cent reduction in groundwater recharge. So the question is, what does this mean to agriculture?

Craig Beverly:

In terms of groundwater, this plot shows the change in groundwater levels for 2030 relative to 2010 conditions, where blue is a drop of greater than 10 meters. And you can see the extent of impact. By 2070, the impact is more widespread, with over 70 per cent of the region impacted. This has implications for water resource management, and for Ag production. So in terms of agricultural production, prediction suggests that the impacts of altered climate vary significantly both across regions as well as landscapes. For instance, the impacts on crop yield across the Northeast Region shows variable trends due to variation in the distribution of the modified climate, temperature, solar radiation, and PET. So assuming current practice in varieties shown on this table, some crops are predicted to increase in yield, such as canola or cherries in weighing and forestry. Whereas other crops are shown to decline in yield, such as wheat, viticulture on the lowlands, and cherries and chestnuts in Stanley. And other crops are less impacted, such as viticulture on the upland, and chestnuts in Myrtleford.

Craig Beverly:

So as an example of how these predictions can be reported and made publicly available, the North East CMA region has developed a tool with spatial vision and natural decisions, to display both the climate and climate threshold outcomes, as well as water balance and agricultural productivity in terms of yield maps. And this tool is accessible on the CMA website. And this image is actually showing a climate tool, I think they call it, specifically the change in maximum monthly temperature for the displayed region that sits behind it, for various future... Yeah, we got that summary table here, various future climate impacts. And similarly, this image it's a component of the water balance. In this case, they drainage using the same kind of format, and have a water balance estimates include rain for groundwater recharge, PET, and runoff. And lastly, here's an image of the variability in median wheat yields and the future climates. So I encourage you to look at the kind of tool sets.

Craig Beverly:

So moving west, the impact of hotter and dry future climate on growing season in the Mallee region, is summarized on this slide. The growing season is predicted to shorten by 0.6 to 1.2 days a year, and these trends are consistent with observations. For example, for oranges in the top row, producers report that the flowering is advancing by about 19 hours per year, which suggests that the model estimates of one day a year appears to be reasonable. In terms of water use efficiency, where I've define that as tons yield per megaliter irrigation applied, all modeled irrigated crops in the Mallee are shown to experience a decline in water use efficiency, mainly due to a combination of the reduced growing season, which I just showed, maintenance irrigation and crop water demand. In general, in fact, water use efficiency declines by 5 per cent in 2030, 18 per cent in 2050, and 22 per cent in 2070. So, focusing on crop yields, we're able to estimate the probability of meeting current yields on the future climates either assuming a simply current practice or modified environment or high temperature tolerant species, or combination of temperature moderation as well as genetic improvement in radiation use efficiencies. And it can be seen in a top row for instance, that for oranges, there is a 37 per cent chance of meeting current yields in 2030, and this reduces to less than 1 per cent in 2017, under current practice. And similar trends are shown for the other key crops.

Craig Beverly:

However, temperature moderation by way of say automatic sprinklers or more high temperature tolerance varieties, increases the probability of meeting current yields to varying degrees depending on the crop. For instance Shiraz in 2050, the probability increases from 10 per cent, and the current practice to 25 per cent if you were to have temperature moderation in play. Now, combined with plant breeding or tree management to enhance radiation use efficiency, yield is shown to significantly increase. And subsequently the probability of meeting current production. It can be seen that as the climate gets hotter and dryer, the impacts from genetic improvement and temperature moderation can become more pronounced and important.

Craig Beverly:

So incorporating this information into bio economic models also provides an insight into potential transitions that may occur under constrained future climates or other constraints. For instance, for the entire Northern Victorian irrigation region, the plot on the left-hand side shows the gross margins and the current, or genetically improved system, or more access to groundwater, or a combination of all three. And what we see on the current water trading and agricultural systems, improve water use efficiencies through plant breeding and water delivery, it could increase gross margins by up to 17%. Whereas access to 50 per cent of available groundwater within a sustainable diversion limits, only increases meeting gross margin by 4 per cent. Whereas a combination of both enables an increase of up to 22 per cent.

Craig Beverly:

In contrast, under limited future water, improved irrigation and genetics is expected to only increase gross margins by 4 per cent. Whereas access to groundwater is not sufficient to match current gross margins with current systems. But the combined initiative does enable potential to increase gross margins. And on the right, I'm simply showing the spatial distribution of income in this case, derived from these kinds of studies, where red in this case is less than $10 million per annum for that irrigation district, and dark blue is greater than $500 million. The top map reflects the current scenarios. And the bottom is the predicted future. And these tools are useful to inform both policy as well as industry. Certainly in terms of the development of transitional incentive frameworks, for instance.

Craig Beverly:

Now the ability of integrating climate with these predictive models is also advantageous, because you're able to evaluate trade to landscape outcomes in terms of agricultural productivity. And this slide shows the performance in terms of grain yield difference maps for two types of wheat varieties of varying transpiration efficiencies when applied across the wheat growing districts of southern Australia, under two different elevated CO2, carbon dioxide concentrations. And this work is based on field studies undertaken by AgVic as well as internationally. And results show that under elevated future CO2, transpiration efficiency can increase crop yields from between 38 to 78 per cent. Again, this information is very valuable to plant breeders, as well as farmers, and hopefully a broader industry.

Craig Beverly:

In a similar example, the performance of pasture varieties is also been evaluated. And this refers back to that phenological model I described earlier. Results show that these images are actually showing the estimated live dry matter across different regions and climates. And here you won't be able to see the scale, but red is less than one tonne per hectare for early spring on a available, yellow is three to four ton, and the darker colours are greater than seven tonnes. But what's important to see is the actual shift in the colour ramping. And the figures show decline in available live pasture mass, but also very much the influence of landscape position as well as soil type. Similar images can be developed in terms of digestibility, where we're seeing a decrease under warmer and drier conditions, where the dull green is digestibility of 70 to 75 per cent, and the pale green is 60 to 65 per cent. And again, this information is really valuable to provide insights for plant breeders. And the platform also to undertake trade to landscape genotype assessments in the future.

Craig Beverly:

So in summary, we've been able to undertake statewide paddock to catchment scale simulations of agricultural production on the various climate regimes. We've had the capability to look at and evaluate dryland as well as irrigated agricultural enterprises. And when you couple that to bio-economic constructs, it gives you a quite a good insight in terms of developing up scenarios related to transitions or into incentivize perhaps different land practice. And hopefully, we've developed a range of products that help inform producers, industry, and policy.

Craig Beverly:

As a final observation now, I'd like to highlight the importance of model selection when evaluating the impact of climate on assessing agricultural production. So in the following example, I'm going to compare the predictions derived using that earlier simple crop factor approach, and the degree day or thermal time approach using exactly the same climate data. In the case of irrigated crops, I'm comparing the predictions for oranges as well as almonds using a two different models. Now, the top table shows the predicted water use, and the bottom table is the water use efficiencies. And a crop factor approach predicts increasing water use with hotter and drier climates. In contrast, the degree-day approach predicts a declining water use through time. In terms of water use efficiency, the crop effector estimates are relatively constant as shown on the bottom table. Whereas the degree-day approach predicts a declining water use efficiency. These results consistent with the model constructs, because the degree-day approach takes account of a shorter growing season, whereas the crop factor approach assumes of fixed growth span. So caution must be used when using a crop factor approach, which is often used, for instance, for water resource management. And based on these observations, such assessment would likely over predict the water use demands and the future climate.

Craig Beverly:

Innovations in agriculture. The World Health Organization has identified three innovations in terms of future agricultural productivity, and that's for food security, efficient utilization of available water, as well as the sharing ideas and knowledge. And some innovations relevant to AV are listed on the slide, and include this capability in terms of trade to landscape to assess the likely product agricultural production. There's also information we're providing under different climates for the genotype evaluation, as well as disease and pest spreads. In terms of efficient water use and utilisation, clearly we can evaluate that using the tool, coupling the predictive models with the climate scenarios. And importantly, also, we're sharing the ideas in terms of digital agriculture, the notion of coupling these predictions with seasonal forecastings, and making those available to users, is a really strong incentive as as well as a work program that we're currently undertaking. So all of these key focus research areas can be informed by covering the climate scenarios with predictive models that I've described.

Craig Beverly:

So in conclusion, can I say that climate change predictions are complex, and include assumptions based on our best information today. Future climate predictions are shown to have varying levels of confidence. And therefore we need to present predictions in terms of likelihood. And the understanding of predictive tools and all the assumptions in is prudent when interpreting the model predictions, and that refers to the model choice that I just mentioned. But importantly, modeling is enhanced when linked with industry and producer networks, to create those opportunities to engage and discuss with both decision makers and users, as well as to help make sense of what the predictions mean for each of our stakeholders.

Craig Beverly:

And that's where I'll conclude. Thank you very much.

Heather Field:

Fantastic. Thanks, Craig. Really great presentation, and definitely highlighting the various data sources available, and the models and tools that can be used to provide some valuable information for future decision making for farmers, and those on the land. So that's been terrific. We have had some comments come in about what a great presentation, very informative. And we do have some questions as well. So I'll make a start on those because we've got about 15 minutes, which is great.

Heather Field:

So our first question from Graham is around when you're working on modeling agricultural production under climate change, what are some of the key things that make some projects more fruitful than others, and any tips for what makes your job a bit easier?

Craig Beverly:

A very good question. Thanks, Graham. And it really does need to be stressed that what was the most productive was actually the engagement with producers as well as discipline experts. So the Mallee example, significantly increased their modeling capability, because producers gave us remarkable insights into yield penalties that they had observed. So their life experience in terms of yield penalties due to climate threshold, which we weren't able to identify in literature, let's say. So having that interaction with both the local producers, who then have a vested interest in helping to guide the modeling, as well as disciplinary experts, by way of the irrigation offices, really value added significantly, both the modeling and the robustness of the modeling, but also helped us to shape the model product into something that was more meaningful and relevant to the actual end users. And that is such a valuable combination that we perhaps in the past have ever really failed in being able to deliver and utilise.

Heather Field:

Thanks, Craig. We have a question from Nick, from earlier on in your presentation. And Nick wants to know, a data drill grid points, those access by a MetI.

Craig Beverly:

Sorry, I don't know. Nick, I'm very sorry. I don't know what you're... Are you asking does MetI access the data drill? Is that the question?

Heather Field:

I think it is. It's wanting to know if the MetI it's where you get the data from. Yeah.

Craig Beverly:

Certainly for our work, we use both. But I would expect that certainly the data drill is now of a premium data set that's promoted by CSIRO. Because you can easily get a solution at each of the grid points. Whereas using a patch point, traditionally, you actually have to undertake interpolation between regions in order to get that continuous smooth climate surface and information. So I suspect data drill would more likely be what's used by third parties.

Heather Field:

Okay, great. Thanks, Craig. From Peter, and Peter wants to know, the distribution of climate stations in Alpine regions is sparse. Data... Sorry, let me start that again. The distribution of climate stations in Alpine regions is sparse. And does this present challenges for protection of water catchment inflows?

Craig Beverly:

Most definitely. Yes, the spatial resolution does have a significant particularly in elevated locations. So where there's large topographic gradients, you have to account for those in terms of the model constructs and input data. But without having those Met data's sources in the elevated, it's more likely we've got lower confidence in terms of the inflows in those areas. So it's a very good point. We do need, perhaps to have more site-specific information in those elevated regions in order to inform the models.

Heather Field:

Thanks, Craig. From Cam. Cam would like to know how can you create or get access to the five-kilometer grid climate data future predictions for use in APSIM and GrassGro?

Craig Beverly:

That's not a problem we're currently seeing on our website. If you look at VCP19, they're hosted by, I think it was a climate change website. And I'm not sure who... I should know. But whether CSIRO, I think they're managing that data on behalf of Victorian Government. But it is available as I understand publicly. The only issue I have is, I'm not sure about the baseline data is publicly available. And whenever you're doing a comparison between future scenarios, it's advised to use the baseline information, not the historical. And there's good reasons why. Because the baseline information is based on historical, but it is a smooth data set. And it's that data set that's been used in their forward projections. So you do need access to the baseline. But to answer your question, it should be publicly available. If not, please contact me. And in terms of reformatting it for both GrassGro or DairyMod, that's quite a simple task. And I think a number of us, whether it's ourselves and DELWP are probably in the process of making that publicly available as well, in those three different formats. Please contact me if you need assistance with that.

Heather Field:

Right. Kim Cam has just added an extra bit. Was it... Hopefully you know what this means. Was it PCP?

Craig Beverly:

VCP. Victorian Climate-

Heather Field:

Sorry. Yes. Victorian Climate Projection.

Craig Beverly:

... Projections 19.

Heather Field:

19. Yeah. Right. Thank you. And we've got another question from Graham. For the North East CMA Forest future growth, important to note that production modeling doesn't account for specific issues, risks such as bushfires. Can you discuss how extremes don't always get captured by modeling, and ways this could be improved in the future?

Craig Beverly:

Good question. So to answer the first one, you're right. We did not take account of bushfire risk. We simply, we looked at different species and different stocking rates, et cetera. we modeled a tree system within that environment. We didn't have an overlay in terms of those other risk factors. It's up to the harvesters to do that, plantation owners. In terms of a downscaling, what the problem is when we're looking at future climate, to try to explain is that, the downscaling from the global climate models which are very coarse resolution when you downscale to five K by five K, and all you're applying that downscaled model to these elevated environments, you do miss the extremes. They get smoothed out. So there's a lot of work to try and develop up better downscaled algorithms in order to capture the extremes, which currently are not being captured adequately in any of these future predictions.

Craig Beverly:

The forecast data is a probabilistic derivation. So it probably should be able to capture the range or envelope of those extremes. But in terms of climate change, no. They do poorly match the current climate extremes. And we can easily show that in some of the work we've done. So-

Heather Field:

Thanks, Craig.

Craig Beverly:

... more work needs to be done. Yeah. Sorry.

Heather Field:

Thanks, Craig. A question from another Nick. Is change in irrigation area, and this is in relation to the example you gave in the Mallee, higher because it assumes water is too expensive, IE dry land pastures?

Craig Beverly:

Yes, that's a good question. So in terms of that, a series of models we did. One was the production models, which we didn't look at the change in the land area, per se. But when we brought it into the buyer economic model, we did look at the change in irrigated area that may occur due to limited future water, for instance. And that was because we were also simulating the competition with a range of different enterprises, from mixed cropping systems, to dairy systems, to irrigated horticulture systems. And once you start to bring those other systems into play, you can look at that transition that might occur due to the competition in terms of water availability and price. And that's the real value of having those bio economic type models, as opposed to a static land use model, where you make assumptions about changing land area. So not sure if that answered it. But there's two very different approaches you can use to try and estimate the impact of water scarcity and price on land use and enterprise.

Heather Field:

Thanks, Craig. Question from Marion. Not 100 per cent sure which part of your presentation this relates to, but she's asked predictions for changes in freight.

Craig Beverly:

No, interesting, you mentioned that. We had hoped to bring freight into the buyer economic model, because that's one of the drivers in terms of profitability. So it can be incorporated into those kinds of structures. But in the modeling I've presented to you today, there's there's no freight. We were very much interested in looking at freight infrastructure in terms of climate future. Because if we've got an estimate of likely yields, what impact does that have on those other particularly freight, I guess, considerations and investment. But no, we haven't made that link yet.

Heather Field:

Thanks, Craig. Now, we are nearly at time, but we do have two or three more questions. Are you happy to stay on for a few more minutes, Craig to answer those?

Craig Beverly:

Definitely.

Heather Field:

Excellent. All right. We've got a question from Peter. Whilst noting the long term climate predictions, we also have shorter term events such as the current La Nina period, which is expected to increase the incidence of warm, wet, and humid conditions over the coming three to six months with implications for pests and disease pressures, as well as crop growth. How is increasing climate variability factored into forward predictions?

Craig Beverly:

Yes. In terms of the seasonal forecast, which is at near future, so it's seven to 200-day projection, that distribution, that does definitely look at the variability, the likely variability could occur within that time period. So as I said, for any one day simulation for the future forecast, is 99 climate scenarios that are undertaken to give you that, so that you can simulate the extremes, and account for various extreme events. So the forecast if you couple it to the model, should capture that envelope of response. Not so with the future climate in terms of climate change. All we have the six different model types that we would use, and hopefully that provides the range of likely outcomes. But the seasonal forecast, the near future does definitely account for that range and variability.

Heather Field:

All right. Thanks, Craig. And another one from Peter on the Mallee Project. Given 75 per cent of Australian olive production occurs in Victoria, mostly in the Mallee, it would be good to see olives included in future climate modeling work. Is this possible?

Craig Beverly:

Definitely. And I hope there is future work in that space. You'll find that in two-weeks time that why there was a decision to select the crop types they did. That wasn't a deliberate decision to exclude some rather than others. It was identified through workshop of what were seen to be the likely future dominant irrigated systems.

Heather Field:

Thanks for that. And a last question. If anyone is interested to investigate, engaging in your work and Agriculture Victoria to do some climate change modeling, how best should they approach Agriculture Victoria and yourself?

Craig Beverly:

Either directly. I suggest also to Graham Anderson, who's probably on the line, he's a very good advocate. So there's two ways. I would probably approach myself and Graham, and we could work out the channels that might need to be advised or engaged in order to undertake that work. I'm hopeful that in the near future there will be more work that's done through our department regardless. But if there's any specifics, by all means, feel free to contact myself or Graham, and we'll work that through with you.

Heather Field:

Fantastic. That's great to hear. Because it looks like there is a lot of interest out there. And this is about being able to provide them a good overall look at what's available, and how it potentially could be used. So thank you for your presentation today, Craig. And we have had some really good feedback come through about a great presentation. And people have found the information quite informative. So thank you for that.

Heather Field:

So we we will wrap up. And on your screen there you can see that our next webinar will be in a couple of weeks time, on the 22nd of October. And we will be going into some specific examples of applying climate change modeling for irrigated horticulture in the Victorian Mallee. So building on what we did here today, and going into a little bit more detail. So we'll hear from the project team. Don Arnold from the Mallee CMA, a lead consultant, Anna Roberts, and Natalie Nelson from Agriculture Victoria. So that should be a good webinar to follow today's.

Heather Field:

So everyone who has registered for this webinar will receive an email with some details closer to the date. So make sure you check your inbox for an email from Agriculture Victoria. So I will wrap it up there. And thank you again for your valuable time, Craig, and all those online today. We did have just over 80 people tuning in live, which is terrific. Thanks everyone, and have a good afternoon.

Craig Beverly:

Thanks Heather.

Page last updated: 21 Apr 2021