Interview with Dean Graetz
1 November 2001



CR: So if we can start with you and how you became interested in remote sensing. That’s a little bit of your background before then, and then how you career in remote sensing developed, or your career in other things that used remote sensing if you prefer to look at it that way.


DG: OK, my reply to that question is to try and put it in a structured way. My connection with remote sensing and the beginning of remote sensing from space in Australia was in a way by coincidence. I graduated from ANU with a PhD in micrometeorology in 1972 and joined CSIRO. And I joined CSIRO in the biological side of CSIRO. It was a small group called the Rangelands Research Unit, and I was based in Deniliquin in New South Wales. And that was June the 2nd 1972 which was the year in which Landsat-1, or ERTS-1 as it was called, was launched.


Now the reason for that background was that CSIRO had been extremely active in initial negotiations with NASA with the launch of Landsat-1. Landsat-1 was the first readily-available satellite dedicated to earth observation. Before that there had been a forerunner of a strange sensor that was then called TIROS, T-I-R-O-S, TIROS-1, I think. Which later became renamed AVHRR, Advanced Very High Resolution Radiometer, which is now one of the major workhorses of remote sensing. But TIROS was around, it was being received in Canberra, experimentally, and people like Gavin Byrne did work with that.


There were two significant interests in remote sensing in CSIRO, and it does connect to the question you asked coming up, and in the 1970’s, this goes back three years before the launch, so round about 1969-70, when NASA was talking, calling for expressions of interest in this coming spacecraft and the use of the data that CSIRO was a major user of remote sensing, not necessarily from space, and had been so for some time. It was principally in the biological area through air photo interpretation. CSIRO had a long history in that in the old Division of Land Research, Land Use Research, it had various names, Land Research and Regional Survey. But it began right after the Second World War and it mapped the capability of land use for Australia. And it developed a system, by a man called C.S. Christian, after whom the laboratory is named, in the now Division of Land and Water, developed a system called Land System Mapping, which has been used all over the world.


Now, so CSIRO had a long history, experience and skilled practitioners of interpreting remotely sensed data in the form of aerial photographs. They were only just multi-spectral in the sense that they were coloured. Interpreting these and validating these on the ground. In other words, integrating ground measurements with that from space and working over very large areas of land to provide information about the nature of the landscape and its potential uses. At a cost that was commensurate with the use of the landscape. In other words it was mostly the northern part of Australia where the land use was extensive like pastoralism and so on. Now they had a long history, they were interested in potential of remote sensing to contribute to continued activities in this area and the other newly emerging area was the geoscience community within CSIRO and without CSIRO where the geologists had long been users of remotely sensed data in the form of aerial photographs.


And they were principally the structural geologists, and they looked at a photograph and their interest was in the shape, the patterns that were there looking for lineaments and various structural features whereas the landscape people were interested in the patterns and the colours. They were interested in the spectral information as well as the patterns. And the photogeologists were excited about the idea of having images from space not only because they promised to be, have a bit more sort of spectral information, the spectral dimension was greater, but they, their coverage, they were synoptic. They were up against the issue of trying to interpret the structural geological context of a great many of the mineral deposits in Australia which had been discovered by prospectors crawling all over the ground. They now wanted, sought, larger views of the earth to interpret the structural setting in which these ore bodies were found. And they were fairly sure that if you had an image from ERTS-1, Landsat-1, which was 180km by 180 that was a view of the earth that they’d never had before. That, plus the spectral dimension they could see a great contribution coming. It was mainly the structural geologists.


Right so we had two, within the CSIRO, you had two people contributing to get access to this satellite data. And in the 1970s, which, in which the mining industry was very active, there were quite a few people outside CSIRO, in the mining industry that were interested. They didn’t know terribly much about it.  And mining was very important to the Australian economy and political scene at that time and so most of the effort in remote sensing that got together was drawn together in CSIRO was well funded was on the mining side. And the, if you like, the biological side took back seat. We had a red team and a blue team and the red team won.


Now the reason why I give you that information is, that’s why people like Ken McCracken, who was a space, remote-sensing type physicist managed to get a team together and most of the initial CSIRO involvement, most of the initial national meetings and briefings were centred around the mineral research laboratories in North Ryde in Sydney. And that was really the home of things and for most Australian practitioners of remote sensing.


I’m not actually sure of the date, it’s around about some where between July and September of 1972, it was the week before Landsat-1 was launched, Mineral Physics held a briefing, a national briefing, and about 400 people turned up to North Ryde to see mock-up data of what the satellite data would look like. It was, it was ah, simulated by aerial photographs, multi-spectral aerial photography taken from a U2, a spacecraft and an aerial photographic system we’d never seen in Australia before because it was still, in those days it was still a defence secret. Um, and the excitement was really quite something. Now in .. it brings you into… I make the point even though I work for CSIRO, I’m not saying it just to glorify the institution, but CSIRO was the major player in remote sensing at the launch of that and in the years before. And it has stayed a major player right through. And even though, if you like it’s been, it’s been a core area of expertise in research and application and gradually various parts of it, particularly the geoscience community has left the research field and got strictly into the application field and so the mining industry has moved away and [is] not so closely involved.


To go back, um 1972, just happened to be the year that I joined CSIRO and I had completed a PhD at the Australian National University, at that time I was in the department, the Research School of Biological Sciences with Professor Ralph Slatcher as head of department. And he is one of the great scientists that Australia’s ever produced. And he was active in a whole range of fields, particularly plant physiology in relation to agriculture. And as a consequence an awful lot of new agricultural developments would turn up. The papers, the journals, not so much the journals, special laboratory reports, would turn up on the reading list, in the tea room at RSBS and one of them was coming from [the] Laboratory for Agricultural Remote Sensing LARS, L-A-R-S which was part of um, ah, it was at Lafayette Indiana, I think. And that was one of the first dedicated remote sensing laboratories. And they were all working with colour infra-red photography, aerial photography. I used to read these, so I got a feeling for remote sensing. And when I joined CSIRO they asked me did I knew anything about remote sensing, so I said that I did and I was given the job of keeping a watching brief on this. So, um and I was working in a, doing research in an extensive land use like rangelands, pastoral industry so I was sent as the , that divisions, Rangelands Research Unit’s delegate to the remote sensing conference. And I was absolutely staggered when I saw the potential of what this data could give. None of us had any data at that stage. I think the satellite went up the following, week after. And I think the first images started coming to Australia about October., I think.


CR: That can be double-checked.


DG: Yeah, and the interesting thing also by chance, by serendipity if you like, 1972 and early into 1973 happened to be a reasonably severe El Niño drought. So for much of, much of that first series of images over Australia, Australia was cloud free. And so we had some first-class images of the state of the country as of the end of 1972 and early 1973.  And then we had one of those amazing occasions where it rained for two years, the great wet of 1974. And it was probably the wettest period in Australia in the last 20 years in Australia. But it set about, set off a lot of ecological changes in the landscape that people are still working with now. And this is captured, quite by chance, in remotely sensed data.


Well having, I was given, I was charged with keeping up with this, there was a tremendous shortage of information about this. There was a [laugh] non-existence of data. Everyone was struggling very hard. NASA was overwhelmed globally trying to supply this. And the principal form of supply was in fact in analogue form just as a negative: either a false-colour composite or a single-band black-and-white. And people were getting these images blown up to 1:250 000 scale. Um and the first one of those I saw was about February 1973. And I can remember [be]cause I asked to get data and there was often a three-month delay between satellite, the data being acquired and it being delivered. Australia had, principally I think though CSIRO, had put in an application for various sites within Australia to be flown and high on that list was Broken Hill. Broken Hill for geological reasons because the structure, the geological structures where the Broken Hill lode is also occur a little further to the north. And that relates to Kerry O’Sullivan because he was very interested and I’ll come back to the Broken Hill-based geographic information system later.


But it was on that February morning and I was working, I was at that stage doing research on the diets of sheep and cattle and a letter, a large box of prints a photo, you know a standard sort of photographic paper box appeared in my pigeon-hole at work and I opened it up, it was after morning tea, I remember it very well. I walked down into the sun and it was a whole bunch of 9-inch by 9-inch positive film and I, one of them was the Broken Hill scene.  I held it up to the light and I was just, actually overwhelmed. Just the incredible amount of information that I could interpret out of that single-band image. And when I saw the false-colour version I was truly absolutely entranced. I could see a great deal that I could interpret about the nature of the landscape and the effect of land use on it. And it took me a little while before I could finally put the words together, and over the following months, but I began to realise just how important this data source was, as most of us working around this place began to do.


And that was that here we had data that was collected by a machine so it was totally objective. And its processing, its subsequent processing, making a print or doing whatever with it was transparent. So that you could go from a purely objective, your primary data was objective. The subsequent processing was transparent, and here this thing could provide information that was readily interpreted, interpretable by a human being over 35 000 square kilometres at a song.


And why I say about the objectivity of the data is because the area that I worked in, part of our research was about trying to find ways of measuring the condition of the land in response to the intensity of use by pastoralism. The pastoral lands are actually owned by the states, they’re not owned by individuals, it’s leasehold land. And there’s a responsibility on all of the state governments to actually manage the land. I.e. they are to prevent overgrazing, which has been chronic in the past, and this is done by law. So anything that provides information on a scale that, and with a resolution that we’d never seen before, has to stand in a court of law. And if you have data that’s primarily objective, i.e. it’s collected, not by a biased human being, or sorry, a human being  who could be biased, you have a whole new world of doing things. Now this took a long while, not took a long while, well it took months for this to begin to sink in. The cost-effectiveness was there, The cost of this per, the cost of information per hectare was extremely low. It would empower state government land agencies who had the responsibility of managing multiple Landsat scenes worth so to speak. It would empower individual pastoralists or pastoral operations in that you could, you could really get a view of the country that was never before experienced.


Two things, there were two imperative that flowed out of that and that’s what sort of  I began working on, as well as working on the diet of cattle and sheep. One of which was how to make the most out of the spectral information. The interpretation was pretty easy. When you looked at it you could interpret what was happening, but you needed to be able to demonstrate that to the point of, not just to your fellow members of the scientific community, it had to be demonstrated to, in a court of law. And this sank into all of the resource management agencies. That eventually the decision on interpretation had to be robust in a court of law. The second thing is that we needed, we needed auxilliary information for two reasons: one of which was to help the interpretation; the visual interpretation, the spectral interpretation, the interpretation of the, get it right, the interpretation of the spectral data often depended on knowing something else about the scene that you couldn’t directly determine. In other words, this sort of spectral information meant something for soil type Y, this same spectral information if it was soil type X had a totally different interpretation. So we needed ancillary data to help in the interpretation. And we also needed ancillary data to help in the reporting. [Be]cause it was very obvious let’s say in a pastoral zone when you looked at it you could see fenceline contrasts in vegetation. You wanted to know where those fencelines were, which property you were looking at, who was the responsible people. So we needed to integrate things like: soil and vegetation maps; we needed to integrate tenure maps, boundary maps, who […] essential responsibility.


We needed some way of processing the spectral data so it went from brightness in the near infra-red to something about that was meaningful to the managers and operators and users of the pastoral lands. Something like biomass or cover or condition or soil erodedness, all of those sort of things. It took us a long while to, to sort that out, a slow realisation. And I’ll say this, the people involved in the rangelands, and there were quite a few working in the rangelands, realised this a hell of a lot faster than the agricultural people. Most of the natural resource remote sensing in Australia was focused on agriculture. Now they justified it on the basis that because they were using agricultural lands as targets, their research was going to benefit agriculture. In most cases it never did. And it never did for several reasons, one of which is they didn’t ask the right questions, the second thing is that the cost of the information they provided exceeded the value which the agriculturalists would place on it., the third thing is that it could never be timely. Back in those days we were looking at three to six months delay between acquisition and data, that all changed. And it was a slow evolution. So going back over those things that I’ve set out, they were long-term strategic developments and from 1972 until the first of those things hit the press, which was 1979, at the first Landsat meeting. It tool that long for those things to happen. The catalyst was , by about 1974…5 we  started to get the first digital data in Australia. It took a long time. When it got here, nobody knew how to process it.


And two people, Australian post-doctorals returning to Australia knew. And they were fundamental movers and shakers, one was Andy Green who went to Mineral Physics and virtually replaced a man who was there who was one of the initial fathers of remote sensing, a man called Michael James Duggin (M.J. Duggin) who was actually a pain in the arse and set things back. He was an analogue man and Andy was a digital man. The other one and they both came from the same laboratory, they were both working, they were both chemists, they were spectroscopists. So they knew about the spectral dimension. The other one was a man called Frank Honey. And Frank Honey and they both came out of working in a laboratory that was actually a spectroscopy lab moved over to remote sensing at Stamford in California. I can’t think of the name of the American who ran it, but he was rather famous. And both Andy Green and Frank honey became movers and shakers well and truly, they were real catalysts.


Now they started writing their own software and at this stage, computing people got into the game and started writing software packages to process the data. You know, packages rather than somebody’s custom-built own library.


It follows on from my comment about that around about 1975 or ’74, I think it was ’75 two Australians, post-docs, returned to Australia and began a huge catalysis of remote sensing in Australia, because they were part of the digital, the digital processing wave. This was Andy Green and Frank Honey. Frank Honey joined up with the CSIRO Division of Land Resources Management in Perth, by which stage my little rangelands group that I was in had joined up with them as well so I started working with Frank, but not for long because it was too far away. And Andy Green to Mineral Physics and he worked there with Ken McCracken.


Ken was still regarded as god at this stage and had a large say over what happened within CSIRO, and he also managed a very disastrous program, not his fault, called the heat capacity mapping mission, which was one that was an absolute bummer. But, and that was purely NASA, NASA’s fault. But there was very, the geoscience community by this stage, by the mid-70s was getting very savvy about remote sensing.


Those companies who knew what they wanted and could see the benefit of this really started to get involved with Mineral Physics. Mineral Physics started doing a lot of consulting work, very close tie-up with industry and those people who sort of came into the game out of curiosity slowly drifted away. So by the time you got to the end of the 70s, ’78, ’79 those people still involved in remote sensing were pretty much the core, dedicated group. And if you look at the proceedings of the ’79 meeting you’ll see an enormous range of applications, from national parks to salinity, in fact one of the best papers, I think it was ’79, one of the best examples of remote sensing in relation to salinity, I can’t think of the bloke’s name. And I think, I think, Murray Wilson, M.A. Wilson was a co-author on it, who later worked with me.


And it was also I think about that time, this is a slight aside, is that remotely sensed data was accepted in a court of law as evidence in a serious case. And that was the first precedent for remotely sensed data being accepted as evidence. Ah! The man I was trying to think of was from the State Rivers Commission, as it was then, in Victoria. A man called, a man by the name of Don Curry, D.T. Curry. And Don was a great, He didn’t know much about remote sensing but he knew it’s, he didn’t know the technical side but he knew its value. And I think there’s a paper in there by D.T> Curry and M.A. Wilson on salinity. But Don Curry used remotely sensed data in a court case around about that time.


There was one hell of a storm, rain storm somewhere in northern Victoria. An irrigation channel burst and a woman was drowned in here house. I’m not quite sure of the circumstances. And all hell broke loose. And Don, very smartly, rang NASA, or got someone in Australia to ring NASA and ask for that scene to be acquired, the day after the flood took place. I’ll come back to the getting the scene. But he got it and as often is after a storm, there was absolutely no cloud and the sky was without aerosols and was beautiful and clear. And he used the satellite image to demonstrate that the water which had flooded this unfortunate woman’s house and killed her was turbid water, which is very easy to separate from non-turbid, clear water. It was turbid water, it was run-off from the rain storm, whereas the water in the State Rivers Commission channel is clear, and it has a totally different colour spectrally. And whilst you could see where it had come out, it was obviously not that water mass that drowned the woman. And this was accepted in a court of law. And I don’t, and it was about that time. And that was the precedent. And now satellite data it routinely accepted as a precedent, sorry as evidence in a court of law.


Just going back to that, at about that time, even though remote sensing was moving on, there were two fundamental shortages still. One of which we were going digital, and the big shortage was data and the shortage was software, available software to process it. The United States, NASA had been recording the data on board the spacecraft and relaying it back to ground. The tape recorders back then were not high-tech. They were failing, there were two on each spacecraft and they were failing and they used to flatten the batteries. So there was a limit to how much non-US data could be acquired. Over the United States they had direct-to-ground relay and I think in about one other place in the world. The United States was also inter… very keen to see what else was happening around the world and there was surveillance of Russia and… from an agricultural point of view, as far as I know. And so the United States began to, NASA began to force people to put in ground stations by restricting the flow of data. So you had to really beg to get data. And in the case of the Don Curry one, was, I think, he rang the NASA representative in Australia, which we’ve had for a long time and they put in a special request and, just by chance, they got it.

[Off-the-record anecdote]


DG: The proposal by Don Gray and others to set up an Australian Landsat receiving station in Alice Springs was put forward, went ahead and I think that we got the first images in 1978.


CR: A little later I think.


DG: ’79, March ’79. Something like that. And suddenly we had our own data and of course we had big trouble because Landsat-3, one of the following had a problem that was called the “late line start problem” which meant that the data was essentially useless. So they re-started up Landsat-2, all of these Landsats had only a design life of one year. I think Landsat ran, Landsat-2 ran for 11 years. ‘cause I was still using it in 1982. I was doing that a couple of years ago. And ACRES, oh what the… or what became ACRES set about archiving the data. So for the first time we had a digital library of the face of the ear… of the face of Australia. And all we had to do, the research people had to do was find the ways to best interpret that library, to read that library.


So that took off… the users were changing, State Government agencies began to set up. South Australia in particular was one. Western Australia has probably been the most successful state, oh it would be, Western Australia is by far the most successful applications group in remote sensing, right up to the present day in all spacecraft. But back then again under the influence of Frank Honey. In fact he, one of his apprentices from the state government department, a man called Henry Houghton, is now, I think, the director of lands. And there was a very tight liaison between research and the people who were using the data. In Western Australia, the Western Australians saw the huge value of this because they had a very large land… a very large chunk of Australia that they wanted to know something about in terms of agricultural condition, they had problems with salinity, they had problems with fire, they had enormous interest in geological investigation, mineral exploration, they had water problems; everything. And here was this data that could serve all that.


And all of there government departments were centrally co-ordinated in remote sensing. NSW followed, South Australia was active and then died away, NSW died away, Queensland is now very active, the Northern Territory comes and goes. But the foundation of that and the people who were involved got their training in the late 70s and the early 80s. And while I was talking about catalysts, the Frank Honey and the Andy Green, another one is my present boss here David Jupp. I think David Jupp joined the Division of Land Research in 1976. David was a mathematician and a physicist and he brought a very different approach to digital remote sensing. And he did it in the development of algorithms, whole new ways of thinking about remote sensing. And the second thing that he did was he tried to overcome the enormous difficulties of getting enough computing and software power to process this stuff and he wished to liberate the masses, so he set about developing a micro-computer-based software processing package  which was called MicroBRIAN. And the BRIAN stands for Barrier Reef Image Analysis, that’s where it comes from. Occasionally it gets written MicroBRAIN that’s often, that’s a little unkind. But that’s where it came from and one of the,, and David won the university medal for that, sorry the CSIRO medal for that contribution.


Some of his first applications were in bathometry and using band MSS-4 or the first wave band, the green wave band, to survey water depth and Barrier Reef quality and Frank Honey was doing the same thing over in Western Australia. And this was when there was a whole lot of exploration for oil and whatever. And they could, for a couple of hundred dollars, process a scene to give bathometry of a quality that was acceptable to the users for a fraction of the cost of the old shipboard bathometry. And people were then interested in forecasting crop yield and the applications grew, as the numbers of sort of people in it to see how it worked dropped off, so the intensity and the diversity of those who stayed in the game in agriculture, in water resources, forestry and so on increased. And from then through into the 80s you see the remote sensing community sort of growing and stabilising and lots of grey-haired old people [laugh] staying around in it. But the thing about remote sensing that, and the reason that I’m still in it, is…  me personally that I’m still in it, and I’m sure that applies to the others, is that we are all true believers. We all have believed in the enormous value of this data stream to Australian society. We have all contributed in some way, some small way, part way, to making that dream, to realising that dream. We all recognise [interruption by telephone call].


CR: OK, off we go again.


DG: Right this is continuing in from the little parable of the dream. And I’ll reiterate that and say that all of the remote sensors that I know that are still in the game are still true believers, they’re not there, we don’t, none of us feel as though  we’re in the area of limiting returns for investment. We’re not, we’re not fiddling around in the area because we’re too afraid to move to other areas. All of us still believe in, that there is huge contribution still to be made, benefit to be reaped, from remotely sensed data.


Why we feel that is, two reasons: one of which is the nature of the data has changed, the nature of the problems that we’re addressing has changed, the tools that we have available to make those contributions have changed. And let’s go to the first one, the nature of the data. Landsat and SPOT and the Indian Resources Satellite, IRS and so on continue that long archive of high spatial resolution, reasonably high, moderate, let’s say moderate spectral resolution, moderate frequency of acquisition. And that stuff has been, that stuff has been terribly important.


One of the most useful projects I’ve done in the last little while has actually used the satellite data from 1982 to 1990 to calculate the rates of clearing in Australia for greenhouse gas emissions. I have used the 1972 to 1992 data to show the changing face of Australia over 20 years, which we published in 1992 as Australia’s contribution to the International Year of Space. And the point about doing that is that satellite data is very rapidly assimilated and easily assimilated by the lay person if you stick just to images using pattern and colour you very quickly learn how to interpret. And you can see, you have an eagle’s-eye view of how that landscape is changing and the reasons for that change. Now some of those reasons are good and some of those reasons are not so good. And in that way we have information flowing into the people who contribute to the political governance of Australia, the average Australian; the ordinary Australian I should say. So in that way, remotely sensed data is informing the people of Australia in a way that no other data source has done.


And I would say, from what I know of what happens in the rest of the world, that Australia is well informed in that respect. We are... the environmental consciousness, sorry I should say the respect for and concern for and willingness to do things for their environment, whatever that may mean to various people, has risen in Australia dramatically from the 70s on. It wasn’t all just satellite data, but satellite data was able to substantiate, to underwrite, to underpin, concerns that were expressed, often on local scale, let’s say the pollution in the river, to the larger scale of the condition of the whole of the catchment, what the forest industries were doing and so on. So that is still there and still with us and still relevant, still relevant.


The second thing is the nature of the, oh the other thing is that one of the, I talked about TIROS, which was 1979 and then in 1981 they launched an updated version of that, a very simple sensor onboard a series of spacecraft run by NOAA, and so the National Oceans and Atmospheres Administration a United States agency, quite separate from NASA. And they put this sensor up which was designed just to distinguish between ocean and land, and cloud and ice. And it was called, rather magnanimously, Advanced Very High Resolution Radiometer, AVHRR. Now this data, we get global coverage on a daily basis. The data is squirted down from the satellite, any one with the appropriate dish or a piece of wet string can collect the data. And it was designed really for looking at and throwing away. But in an amazingly short time, and again I had a very small contribution to make here, it was very small, Gavin Byrne used the TIROS data and said we should, he published that and said we should do something about it.


Then the AVHRR data came along and a few people in CSIRO did it. I was fortunate enough to actually be at a remote sensing meeting in Egypt in 1981 the year Anwar Sadat was assassinated I think, and I met a man from NASA called Jim Tucker, C.J. Tucker. And Jim was using this data and he had ways of geographically warping it and straightening it out and because it was daily data, very high temporal frequency, coarse resolution, either 1 kilometre or 4 kilometre resolution, fairly coarse spectral resolution it only had 5 bands. But nonetheless he had a poster at this meeting in Cairo that showed the greening of the Nile delta. And I  stood there, I remember standing there and looking at that and thinking “Holy Shit!” here is another new dimension, a totally new dimension, even though we were getting snapshots, with Landsat we were getting if you like high resolution snapshots every 16 days.


Here was something that was global on a daily basis. So you could address problems that were actually about the functioning of landscapes, the functioning of systems, rather than the condition of systems and change. You could look at… you had such high frequency coverage that you could actually see the sorts of physio…, the changes that were related to the physiological functioning of landscapes. And Jim Tucker subsequently proselytised the world to that. And in the early 80s he was working in Africa, desertification was a big issue right through the 70s, and suddenly AVHRR data was being used all over the world. You could show the greening of the global land surfaces and so on. Frank Honey was the first person in Australia really keen to get that going. And in about 1981 or 2 Frank assembled a very simple receiving station on the roof of one of the university buildings, the University of Western Australia on an old gun carriage I think. And they used to try and track the satellite following the signal with an oscilloscope. And they collected some of the early NOAA data for Australia. In fact in the great howling drought of 1982-83 which is the most severe drought for the last 100 years they actually have some high resolution NOAA data of Australia. And they also have, and it’s been used widely, NOAA coverage of the Ash Wednesday fires. And suddenly when this was circulated, and I think that was actually acquired, I think New Zealand for some reason had NOAA reception, they do now and they have, but I think it was actually from New Zealand that they captured the Ash Wednesday fires and that little image got plastered up [everywhere?].


DG: Right, just to pick up with the comment on the NOAA data and the realisation the Ash Wednesday fires in particular, which was 1983, the Ash Wednesday fires how a whole new user group came out of the woodwork to start using AVHRR. And there were dedicated AVHRR users and there were people who saw this as a complement to Landsat data and so on. And a whole lot of new problems could be addressed, in particular drought and fire.


I remember it rather well because the Ash Wednesday fires and all those sorts of things were in early 1983, it was an election campaign, and Malcolm Fraser, who we mentioned before, that was where he lost power, he also made a very unfortunate comment after attending a church service about the Ash Wednesday victims which was 78 people who died in the wildfires. I was beginning to film a television series, that actually took satellite data to, via television to people, it was a series the ABC made called “Heartlands”. It was the first set of Heartlands, it was made by ABC Country, Rural Television which was the group who used to make “A Big Country” and this was our contribution to the bicentennial. And it looked at what I thought was the, well I say “I”, because I was the talking head, what was the great conservation issues. And my argument was that it wasn’t the warm and furries, we can fix those up no trouble, its in fact the conservation of our core agricultural lands and water.


They were not well managed, for various reasons, and these were our heartlands, these were the ones that would feed us and so hence the title. But in that I used, I used NOAA data showing changes in the Australian continent going into the drought which I got from the United States. And when was that? 1983. It went to air in 1984, and what are we? 16 years later? 17 years later. I’m still working using NOAA with my colleagues here. We are trying to remove all of the non-surface influences so that we have the most perfect record of how the continent has behaved over the last 20 years, from 1981 to now. Because that’s about half the time that the current forecasts for climate change are, that they are about 40-50 years before they have confidence that we’ll see these sorts of changes. We want this data to calibrate those models, to validate those models, and to say, this is how the continent has changed just due to the circumstances, or the factors that we can interpret. Like we, in that time, 1983-84 one of the most severe droughts for the last 100 years, 1988-89 was one of the wettest years on…  89-90 was one of the wettest years on record, it was the, massive flooding in northern NSW and so on. So we have a history there that we want to interpret for the scientific community, particularly within CSIRO, but also we plan to write it up for the Australian public so that they are informed of the nature of this.


Now um, to stop all that, that was about new data, new problems and new ways of working with it, sorry, new data and new problems. The second thing is the image processing software. That was a great limiting for a long time, along came MicroBRIAN, another one called DISIMP which was written by the Division of Computing Research, a man called John O’Callaghan led that. John O’Callaghan is one of the great leaders in this. It was John O’Callaghan who led me, and his name is one those early papers in 1979, and said look we’ve got to integrate this, we’ve got to integrate this with the ancillary data. How are we going to do it? He wrote the software. There were no GIS systems then, we had to write them. So it was, we wrote the software which first geographically, or geometrically corrected the satellite data, then we wrote the software so that we could integrate other ancillary data sets. We had to write the software to digitise it, all that sort of stuff. I didn’t write the software, that we John O’Callaghan’s group. And that was a huge contribution.

Right, OK just on a, on a sort of people note, the area that sort of has reached maturity in remote sensing is really its application to rangelands and its application to agricultural areas. There is still lots more work to be done in the application of remote sensing to water quality, which is a big issue. And here we are in need of different sort of spacecraft. We really need the hyperspectral, in other words 256 channels, that sort of data is only just coming available now, from space. In fact, part of being a sterile helper at the nest, that’s a biological term.


I’m one of those now and I’m quite happy to that. We’ve been, And my colleagues down the hall, we’ve been involved in calibrating, absolutely calibrating a hyperspectral satellite, the first one in space called Hyperion, and we’ve been using the brightest spot on the Australian continent, which is Lake Frome, and the darkest spot, which is Lake Argyle. It’s dark because the water is unbelievably clear, so it virtually absorbs everything in the near-infra-red on. So we have the brightest and the darkest spots in Australia pressed into service to calibrate a satellite. And if you calibrate a satellite in Australia you can use it for the rest of the world. That’s the global thing that took a little while to realise as well.


And I can talk briefly here about some of the consulting work that I’ve done. With other people. A couple of them stand out. Following that theme, the inverse of that theme is that you can survey the world from an office anywhere in Australia. If you like, you can have an office in Oodnadatta and monitor the wheat yield in Canada. Because now, you can do that, because the data are so readily available, and you can, they are timely, almost immediately available. The world has become incredibly interconnected. And the software and the intellectual property needed to do that analysis, are not terribly.. they’re available and not restricted in any way.


In the late 90s I worked, I had the privilege, and it really was a privilege of working with Elders Agribusiness which is now defunct, for reasons not of business. And Elders Agribusiness were the leading agribusiness people in Australia. they, and they were very far-sighted as they still are today in different forms, and they gave me a consultancy to demonstrate to them, to explore for them the many different ways in which remote sensing data could help agribusiness. And it was in international trade, it was in lending money—the condition of property, selling real estate, selling property, it was in supply and distribution of merchandise that they well, it was, how can I put it, it was agricultural surveillance. They dealt in futures and they wanted to know coming yields of crops and whatever. And I was absolutely amazed at just how far-sighted and forward thinking they were. For various reasons, a lot of these things were not acted upon, [laugh] maybe my work wasn’t good enough. They ran into financial troubles because of the futures market around about that time and there was a collapse which was unfortunate, because they were … But I noticed the other day, looking at a real estate brochure for a property in, a pastoral property, there was a satellite image of the property. And that was what I was proposing to them back in 1989, and they were quite happy to do that.


I’ve worked for mining companies, using satellite data for environmental management, and they have been the best people I’ve ever worked with. I’ve never met a mining company that was not trying to do the right thing. They… mining companies got a bad press unjustifiably because they were often at the forefront of disputes of exploration licences. Whereas they were only acting lawfully, the state government agency that had given them the licence to explore were the ones that the greens, the environmentalists should have git stuck into. It wasn’t the mining company, they were acting lawfully. And I used to spend a lot of time explaining this to green audiences and getting bread rolls thrown at me, but the mining companies they were always… they were much better than government agencies. The government agencies were the ones that I had to work with, because they had the responsibility for land management.


I suppose if you wanted to, if I had to find a word to say, words to say where I’ve worked, I say I’ve worked in the application of remotely sensed data for land management. And principally pastoral land management but also aboriginal lands and mining and so on. I you read, “Looking Back” you’ll see the sorts of things that I’ve contributed to, in a very small way. But it’s, in most cases my contribution has been, my principal, my most important contribution I feel has been the articulation of the problem. When you’re dealing with land management you are dealing with a social system as well as the natural system. I’m a natural scientist, and most of the resource management problems fall down, the research on it falls down because they concentrate on the natural sciences and they forget that what they’re doing is supposed to interact with the social system. So they either solve the wrong problem or they come up with a solution that’s totally unusable. And it still goes on. I went out of my way, which is at great [laugh] which you don’t get any praise for, to work with social scientists and try to find about out how things work in the social system. Now I did that by actually working with a couple of… one very enlightened state government agency, the South Australian Pastoral Board.


And we developed a image-based resource information system and we stuck an “L” in it for “Lands” so it became “Land Image-Based Resource Information System”, LIBRIS. It was to be the library of knowledge. And that was based on the software that was written by CSIRO, by John O’Callaghan, and we worked in the early 80s, it started in 1981 with this whole idea of data integration and it flowed from the early-80s to the mid-80s and now everybody does it with commercial software, it’s not a problem, GIS is everywhere.


But before that, the first great big data integration task done in Australia was an absolute killer, it made a lot of enemies. It was called, we called it BIBGIS, Broken Hill Image-Based Geographic Information System. And what we did for the Broken Hill scene, we put together every data set we had, and that included tenure, vegetation, slope, soil type and we got hold of all of the radiometric data, thorium, potassium all those sort of doovers that were the geophysical data sets, flown by BMR then, now AGSO, that were all languishing away on tape. We got hold of those and we warped them, we geographically corrected them and enhanced them and overlaid them and put them all together. And the driver of that, no not the driver the, one of the catalysts was Kerry O’Sullivan from CRA. CRA had a major commercial interest in that area, CRA put up a very small amount of money and CSIRO out up a much larger amount of money, but anyway… but CRA through Kerry could interpret, he knew what he wanted. Kerry always used to say the deeper we want to go, the higher we’ve got to go to see it. And that, BIBGIS was about 1980 or 81 I think, 81 perhaps. And the data was released only as slides, as images because it was, quite a lot of it was commercial in confidence we couldn’t re-sell it or give it away or whatever. But it was an absolute eye-opener for the geologists because they could suddenly do all sorts of new modelling. They had a data dimension, data dimensions richer than they’d ever had before.


CR: It must have been quite overwhelming.


DG: It was, it was. And the at this stage we realised there was actually another little niche to fill and that was that suddenly we were dealing very large data sets, many of which were correlated and the statisticians came out of the woodwork [interruption by telephone call].


DG: Just picking up on the growth of data and all that sort of thing and suddenly there was a realisation that there was a niche to be exploited and there was a movement into remote sensing by several key statisticians who really helped with, dealing with the,, sort of data, data reduction techniques. Norm Campbell, N.A. Campbell, in Western Australia was one of them but there were many others. And Norm Campbell is still in remote sensing, he’s just retired from CSIRO but he’s still involved. But dotted through the literature you’ll see the names of people who were statisticians. And they were an essential part of remote sensing. They were, in a way they sort of sharpened the swords. Because most of us came at this from an applications point of view. We had to learn the image processing. We had to learn to understand the spectral and the spatial dimensions of the data and the whole data reduction techniques.


And I mention this because it’s important and also personally one of the few contributions I’ve made was that we actually began a change in perception in remote sensing with a statistician, whose name escapes me just at the moment. We, in the early 80s we began the work on what is called “mixture analysis”. And what this was, facing up to the fact that almost every pixel that you had in an image was a mixture of land surfaces’ contributions. And to run around under the delusion that you had pure samples and therefore you could just classify it into pure samples and the world would fall nicely out into pigeonholes, which everyone was doing then, and some of the are still doing to this day, an alternative way is to say “the landscape, as captured by satellite data is made up, every sample, i.e. every pixel, is a mixture, and a spectral mixture.” And by various techniques you can untangle that mixture and you come out with your pure spectral types. And your job then is to identify what they are, not to delude yourself that you’ve got this lovely cloud of bees, sorry you can imagine a volume of bees and that’s what it’s like and there are no little blobs in spectral space, nearly all one big cloud, a people force that through a cheese grater and say “look I have all these different types”. Our job was to find the extremes and then to build, to represent a pixel as a mixture of the following things. And we started in a small way with a colleague Roger Peck and a statistician, whose name I still can’t damn well think of, and we were the first, not the first globally to use mixture analysis, mixture modelling, but we were the first in Australia, as far as I know. And that’s now a commonly used routine.


And… what else?


CR: I think we’ve covered nearly enough for the moment.  And I do appreciate it. So perhaps we might finish here