 What I'm going to talk to you about is the findings of a research project that's been undertaken within the last 12 months. Generously supported by Dairy Australia, the Cotton Research and Development Corporation and Grain Growers, looking at the implications of digital agriculture and by inference, big data for agriculture in Australia. So it basically involved an analysis of a lot of the issues and also a fairly close examination of some relevant case studies and I think that's quite important. So digital agriculture is a term we've used basically to describe agriculture which relies on detailed digital information about wide range of production variables which are utilised as a guide to production decisions. And we're making sure we understand that that's different from so-called big data because big data is data scale, diversity and complexity requires new architecture techniques, algorithms and analysis to manage it and extract value. The two are obviously elated because you've got to have digital agriculture if you like as part of the collection process of the raw material that becomes part and parcel of big data. So just to describe it, digital agriculture might be the use of five years of yield maps from a specific paddock by an agronomist to develop a variable rate planting and fertiliser strategy, whereas big data would be something like the analysis of all the yield maps from all the grain harvesters operating in Australia over a five year period together with weather data to identify yield and water use efficiency trends. So differences of scale and I guess digital agriculture is more about individuals and management decisions. To just give you an example and many of you will know this, if we think about it in this term, one year's yield map from a particular paddock is of pretty limited value. Five to ten years yield map in combination with an understanding of the rainfall conditions in each of those years starts to give you something you can actually make some use out of to make future decisions. And then if you had yield maps for an entire country obviously coupled with things like soil data and meteorological data, you can learn an enormous amount about a whole range of different variables. So one of the quandaries of this whole area is that by themselves and I guess the other speakers have mentioned this, small pieces of data don't actually generate much value. It's only when they're combined together in the right systems and in the right volume that you actually start to get some value out of them. What is that value? I guess that's the other point that certainly a lot of farmers would say well okay I've had a header with yield maps for the last decade. I'm still not sure I'm getting any value out of it or I've got a spray coop that I can do all sorts of things with and vary the rate and do a marvellous stuff but I'm not really generating much value out of that. I guess the experience of the developments that have occurred in the corn belt in the USA where the concepts and the practicalities of utilising much more intensive information systems to make productions decisions gives us a bit of an example. So the sort of numbers that come out of the sort of analysis that's been done in a whole range of scenarios are around about that five to ten percent productivity gain. So by moving from paddock average to square metre average management or moving from herd or flock average to individual animal management that tends to be the gain able to be achieved by these systems and obviously then that's offset if you like by the price of the system so depending on the platform you're using that would determine what the net value is but certainly the platforms that are being used in the corn belt in the USA the more sophisticated ones are charging in that region of three to ten dollars an acre and even at that level are generating quite significant returns from the use of that. So I think and it's interesting that when you look at some of the examples in the livestock industries as well that five to ten percent figure is the sort of figure that gets talked about. So certainly depending on the cost of the system obviously there is some value in pursuing them. I think the other thing to get in our minds is we're moving through a transition so if you think about agriculture and certainly fifty years ago it was a skills based industry. It was the skills of the manager and the skills of the observations of that manager that determined the success or otherwise of the business. We've slowly transitioned to a situation now if you think of reliance on data and yellow and reliance on skills in blue we've started to transition to a situation where some of the sectors are now starting to be more information based compared to what they were. So if we look at variable rate cropping systems we've moved from purely skills decision where the manager looked at the scenarios and decided what he was going to plant to one where starting to see the greater use and integration of data as part of the management system. At the other end of the spectrum if you like is perhaps the broiler poultry industry where now in Australia if you decide to have a broiler poultry set up on your property you can parachute in the technology and all the control systems and basically end up watching weight gain of the birds involved on a computer screen that controls all the inputs as well. So what we're talking about is this sort of gradual transition of agriculture from purely skills and observation based enterprise to one which increasingly integrates digital information into the system and helps that is used to make production decisions. And obviously the ultimate objective of a lot of these systems particularly in an intensive cropping system like the US is to be able to integrate all the various elements of information from yield maps and rainfall and temperature and soil type etc. through a platform utilizing an algorithm and to come up with an optimum solution for something like well what variable rate, what variety should I use, what fertilizer application should I use to sow the next crop. So that's the sort of the holy grail if you like and I think Alex has already mentioned a critical element and a critical competitive element in these is the algorithm the numbers the formula if you like that's used to convert all that variable information into something that's actually actionable from a management decision perspective. And I think we'll see that when we look at the example of the US systems that a lot of the participants have quickly realised that that's where the competitive advantage lies if you can get that algorithm for your system to work better than the algorithm for another system. The other interesting development is how these systems have progressed particularly in things like the cropping industry so they started off with an implement like a tractor it became a smart product it started to incorporate sensors around machine data and those sorts of things it then became smart and connected through connections with things like GPS and mobile communications technology it then started to become a product system so your tractor talked to your implement and there was a feedback system in fact I had an interesting example of that on my own family's operation at Christmas time where just prior to going away for a couple of days holiday my brothers wanted to feed some cattle and they had a brand new silage feed trailer and we spent three hours convincing the tractor that there was a silage feed trailer behind it because the only way to get the feeder to work was from the console in the tractor so that's a sort of a if you like a connected or product system and then what we're moving towards and what has also been developed is a system of systems so the machinery system talks to the irrigation system talks to the weather systems to the financial data so you've got an integration across all those different platforms and the ability to transfer information between them in a fairly seamless fashion so that's where things are quickly heading to so if we look at developments that occurred around things like the corn belt we've seen very quickly emerge quite a competitive market for the provision of digital information platforms and systems if you look at the matrix you've got the makers of precision agricultural equipment the names that would be familiar there some making the machinery some making the control modules for them and so they're the ones who are generating and capturing the data if you like some of those are also just involved in providing storage and retrieval systems for that data and then some of them are also involved in then utilizing some of that data and the insights that data back in as a delivery system to the next iteration of those decisions if you like some are just involved in providing cloud storage and data warehouses so they're not really and you'll see some familiar names there like Amazon and some of those so they're basically acting as a cheap storage facility but not doing much else a lot of the drive has been from retailers so these are the fertiliser and chemical retailers that are supplying product to the clients on the ground and SST is a very strong software product that allows the retail agronomist to come onto the farm to map out the proposed cropping regime to then estimate the fertiliser seed and the variety of requirements push a button the order goes back to head office all the paperwork's done and the retailer agronomist moves on to the next farm so the platforms that have been developed have a whole pile of benefits but the driver for it has been how do we automate this system and use the retailer agronomist time more efficiently and so they're the sort of platforms that have developed and then there's a limited development of smart models and a lot of these depend on the development of those algorithms so the organisations there need to have much more comprehensive sets of data available not just for example the weather or not just some variety results from particular corn types or those sorts of things they need to be able to put all those together to create these platforms. The uptake of the use of the systems has been quite dramatic as I said some of it driven by the fact that it solves a problem for the retailer some of it driven by the fact that for individual corn farmers who might have multiple plots of land it's a way of putting all that information together in one accessible bit and it's interesting when the first very deterministic models came out things like field scripts which was released by Monsanto the leading edge farmers didn't need a computer to tell them how to make good management decisions they were of the view that they could outperform any algorithm in an iPad but the point was made by the providers of these that in fact as some of them described it to me it was the lazy middle that became the target market so the lazy middle weren't the leading edge farmers they knew they could be better they knew they should be spending more time refining their management and upskilling but this provided an easy platform for them to avoid having to put that effort in but at the same time to gain some productivity benefits so a lot of the users of these types of systems often under the guidance of their agronomist are in fact what some of the retailers would consider not the leading edge farmers but those who come a little bit behind them so what are some of the products available probably the one that's garnered most attention is the climate corporation products they're now released under the field view label those of you familiar with the history of climate corporation might know that it was started by some ex-google employees who recognise the intrinsic value in the massive accumulated weather data from the US which is a very high density data information collected over a hundred years they realised that being able to dive into that and become much more specific about whether risks was a major advantage for insurers and they would sell that service to insurers Monsanto came along with its masses of variety data from from corn variety and soy breeding in the US put those together with the very detailed soil maps that are also publicly available in the US and started to come up with platforms that take away or simply simplify a lot of the management decisions one of the most striking ones was what they call the nitrogen adviser so once all that work's been done once that mapping's been done and the information's available one of the key questions for a corn farmer in the US is how much nitrogen to apply and when and this tool created quite simple predictive models based on all that information behind to help make that decision in the first year this particular product was released it was produced on 70 million acres in one year and so I think there's some interesting lessons there in that it solved a simple problem it didn't try and manage the whole farm and so the use and uptake of it was was accelerated by the retail agronomist providing that product and helping farmers get used to it and then the fact that it solved a simple problem not the whole not running of the whole farm another interesting example is the farm link product and the farm link started life as a leasing company providing finance for contract harvesters they realised that all those contract harvesters that they financed were all generating yield maps they developed some technology to do two things one standardised the calibration of those yield maps and secondly transmit that yield map information from the harvester to a single storage site they then were able to combine that information with the soil maps again with the detailed rainfall data available in the US and develop up what's essentially amounts to a benchmarking service so in other words it allows a comparison of the performance of the yield on a particular field with a very similar yield some very similar field somewhere else so they sold that service to the agronomist who are then able to sit down with their clients and say well look here's how we compared with these other fields that were harvested that were very similar conditions maybe we should have done this different or that different then the secondary outcome of that was they realised that in fact these yield maps were a very useful calibration for satellite imagery so they knew the actual yield they had the satellite imagery they could therefore calibrate the satellite imagery for that particular field and then extrapolate that calibration out to all the other fields that they weren't actually harvesting so they believe that that's given them a very low cost and highly accurate ability to predict crop outcomes and in fact they believe it'll produce a better result than the USDA is able to once it's developed a bit more so that I guess is some sense of the sort of developments that are occurring in the US of course the my John Deere platform some of you may be familiar with most of the high end John Deere equipment now comes out with the ability to transmit in real time to the my John Deere platform one of the things that's developed a lot of these services were what you might call loyalty plays so in other words if you bought a blue coloured tractor your data was specific to the blue coloured tractor and the longer you kept retaining that data the more you were more or less committed to that blue blue coloured tractor because to move away from that particular variety of tractor would mean that you no longer had access to that data that quite quickly changed it's considered two reasons one the providers of those platforms realised that it was creating a real problem for them in that they would have to be able to manage the requirements and expectations of their entire customer base and so they would have to develop more and more complicated software products if they were going to retain that information that digital information as proprietary to their system so what happened was the development of yes climate corporate involved in the development of this platform they also can use it in conjunction with precision planting which happens to be owned by Monsanto and the Monsanto seed group but they've also created an opportunity for the ag gateway and the open ag data alliance compliant systems to actually store their data on that and case IH to store their data on that and then you can have multiple users out the other end and all using proprietary or competing software but all interoperable so that one bit of data residing on that platform can go to a multiple different uses so I guess the analogy is the Apple iOS system where Apple contains the iOS platform but you can have a very competitive market for applications and software all that can use that platform so that's what's fairly quickly happened in terms of developments in the US and there's a very competitive market developed among software providers for the different sorts of systems that farmers might want so as I said it was initially a product loyalty platform for big players now for see a new revenue source by providing a service that also has the open access arrangements for other users the system costs for the high end system costs for US farmers in the region of three to ten dollars per acre and companies like John Deere expressed the view that they see it as a way of smoothing out what is a very seasonal revenue flow for that company that by providing an ongoing data maintenance and management service they're creating another source of revenue that helps them to smooth out their cash flow the lessons from the development of these systems is that ways to reduce costs are much easier sell to farmer than the promise of increased yield I guess intrinsically we probably all recognise that as a logical thing that if you can show a farmer the way to use these tools to reduce costs it's a lot easier sell than promising them that they'll be increased their yield and the other point about a lot of these is that the main delivery model is via crop advisors so it's not farmers themselves sitting down and going through the detailed process of establishing the maps and inputting all the information it's actually their advisors using the platform to put the farmers information in and then allow the farmer to use that once it's set up so I think that's quite an important lesson in terms of the adoption of these sort of systems that it's not likely that there'll be a mass rush of farmers to buy these products and and put all their farm data for the last 10 years on it that in fact often it's facilitated by advisors data ownership and security most US vendors are now starting to adhere to the data gateway or open ag data alliance standards and encourage application software developers so so that doesn't mean all the data is in a standardized format it just means that anyone who wants to develop a software product to work on a particular set of data can obtain the metadata that explains the structure of the data that they're working with so it basically means that you can transfer data from from one platform to another so if you get sick of blue tractors and decide you want to go to yellow then you can translate your data across to the new platform when you change products so that's quite an important developments most guaranteed data confidentiality there's some exceptions to that for example John Deere retains the right to machine data so in other words the performance of the engine and temperature and running and those sorts of things and believes that's proprietary to them and allows farmers to set access rights to production data so you can store your production data on the JD platform but you can say what uses that data can be put to and and there's a range of those on Santa climate Corp ensure the data remains secure and is not provided to third parties so again they've recognized that for farmers to be confident and to use these systems you have to provide that assurance around data and recognition of data ownership all that said of course in the event of a legal situation any of this data is accessible under subpoena and that's no different to the situation of a farm notebook or a paddock record of some sort there's no change there so the ultimate level of security in the case of a legal case for example is no different to anything else so that certainly needs to be kept in mind so can these systems are developing quite quickly in the U.S. and becoming utilized and uptake is quite strong will we sort of see these sort of things in Australia I think it's important to realize the U.S. systems are developed on a few things that we don't have so very high density rainfall and climate data both weather stations and Doppler radar that gives very localised rainfall not rainfall extrapolated between two points a hundred kilometres apart so I think that's quite important very high resolution national soil map so that one to twenty five thousand national soil map is is is readily available and accessible to anyone in the U.S. so a lot of these systems use that even though they recognise that that's not highly accurate but at least gets them in the ballpark of understanding what the situation is extensive public GPS and mobile phone network and GPS correction system so right across the corn belt certainly that is is a given so that that sort of accessibility is already there and also we need to remember that in fact the systems been developed offer a monoculture high input crop so very high inputs which brings with it the ability to be more efficient and a single crop focus so you've got a very lot of research gone into the performance of a single crop which makes predictions and algorithms a bit more stable than than where there's been less intense research but also importantly a large and competitive U.S. market for software applications so agri tech as they call it so there's a whole investment population floating and and trying out and developing software applications for agri data so not public research institutions but actually commercial providers funded by venture capitalists and and developing these platforms and I think that's been quite a major and so when you go to a a conference on agri data in the U.S. you might have a couple of hundred vendors they're all displaying their wares and showing how competitive the situation is so that's quite important as well just gives you an example of some of the soil data that's available so what's what do we like to see in Australia I think the Australian cropping industry will be a recipient of the technology spillings from the U.S. and we're already seeing that the a couple of the major retailers use the SST software platform which is used by their retail agronomists and you can see that there's already been the use of that in fairly extensive areas throughout the cropping belt so the the sort of stuff we're talking about is already here but predominantly being used by retailers as a way of managing the sourcing and supplying of inputs there's no developments in the U.S. in terms of systems incorporating livestock and crop rotations and Alex touched on this earlier I think that's the big difference and that's the big challenge for Australia that the real value particularly in the southern systems will be the ability to integrate information about livestock enterprises and cropping enterprises on the one sort of platform and there's none of that sort of development obvious in the U.S. and there's virtually no development of off the shelf livestock systems in the U.S. Most of the systems are in fact proprietary systems that operate in the big feed yards or the big pigory or poultry or dairy operations where they develop their own system but there's not really those sort of commercial off the shelf type systems being used in the livestock systems there. The relevant algorithms for probabilistic decision support in Australian crops will need further development there is some already and of course all on the back of the APSIM model developed by the CSRO some time ago and further developed and we've got production wise being operated by grain growers and things like yield profit etc providing a basis so we've started down the track certainly we're not anywhere near to the development that the U.S. corn industry is at the moment but there's progress being made. So what would it take to facilitate the development of these sort of systems in Australia? I guess the big challenge is is it feasible to improve the knowledge of soil types by for example improving the quality of soil testing and to aggregate that information together that would obviously require private individuals to allow their soil test data to be aggregated and and used in some of these big applications I mean that may be one way to get around the sort of limitation we have in terms of the specifics of soil data available in the U.S. Farmers and industry should probably develop and commit to open access data standards and privacy products because that's how you're going to encourage a competitive software market. Farmers certainly don't want to be locked into proprietary systems they want to be confident that if they do go down the track of investing in these sort of systems that they will be able to be choosy about their future suppliers and that's certainly something that seems to have helped in terms of the development of these sort of products in the U.S. Farmer ownership of data and control over use requires agreement so this isn't something that's legislated in the U.S. in fact in both the U.S. and New Zealand the industry collectively farmers software developers technology providers have all come together and negotiated a standard agreement about the nature of ownership and use of the data and that certainly seems to have created a degree more confidence in the application of these. Is it feasible to incentivize private weather data systems to supplement the existing BOM system? I was forecast based up at Narrabri is already doing that so you can buy a fully automated highly precise weather station for around about five thousand dollars install it it gets integrated into the system and then you can retrieve information from the web that includes both the BOM stations and private stations and get much more accurate where the data relevant to your particular crop as someone said to me this extrapolated rainfall between two points a hundred kilometers apart is all right for virtual models but it's not much good for growing a real crop having that much more localized weather data is quite important to have these systems working well so I think there is some feasibility of filling in the gaps if you like with that private weather data system and this is a point Alex made and others have made as well the other algorithm development needs data analysts not plant or animal scientists in fact climate corp told me last year they put on about 90 data analysts and one plant physiologist so it's it's it's not that we don't understand the systems it's just there's a lot of work involved in getting all that data integrated and put together in a way that creates a useful product out the other end so what are the issues clarity regarding data ownership and control over use is certainly something we need cropping and livestock digital systems both private and government I think need to be open access so that for example NLIS can be tapped into by the by the developers of for example livestock systems and integrated so you don't have to have here's your livestock management platform that you're using on the farm and I'll by the way when you're selling cattle you've got to move to a different system to generate your national vendor declaration and upload the data around the the livestock tag so so having a single system rather than multiple systems would obviously be appealing but that requires open access data and that should apply to research as well because a lot of the gains that have been made in the US have been based on the ability to look at very large stores of data from research trials and to use that to develop some of the algorithms so USDA at the moment is starting to move towards having all its internal research trial data transferred to an open access system and I think that needs to be thought about seriously here in Australia because a lot of the IP restrictions that are placed on them never eventually generate any revenue so you wonder whether in fact we'd be better off with an approach that basically said that data is accessible to anyone who wants to play around with it and to try and come up with new products or new services that could be based on it and I think the final thing that's probably needed in Australia is in fact a forum or a way of bringing the various different parties together so you've got data analysts and computer specialists, software developers, researchers, farm retailers and suppliers and farmers themselves all have got an interest in this area but have no common meeting points so I think there's some quite merit in thinking about how do you bring all those various groups together and get that interaction that invariably ends up bringing together or bringing forward some of these new developments and I think ultimately the dilemma is this that the more Australian farmers collect and freely share production data the more likely it is that digital agriculture will deliver productivity gains but that's in the future so getting the right frameworks and the right thinking around this is important now so that by the time some of these developments occur the data is accessible and able to be used and provides use and generates those productivity gains a quick ad we've got a conference coming up to discuss these issues in much more detail 2nd or 3rd of June in Sydney and as always the Institute is generously supported by a wealth of industry organisations and we always like to acknowledge them so thank you