His GIS career started out accidentally — as most great things do. He went from being an accountant to business development for a young company that started making maps from satellite imagery.
The image was a background. It was new and innovative — fascinating and earth-shattering for Tom.
Five years later, he started Icon. He’s been at it for the last 27 years. (*bagged a Masters in GIS and remote sensing in that time, too).
Commercial imagery resolution is down to 30 centimeters.
You can identify features in pastures that otherwise would not have been visible. Counting livestock is important on several levels — for animal disease control and for inventory.
You also get a view of what’s happening with the cultivated land.
Yes. That’s our principal business; classification of land cover into categories that interest our clients. Our work supports agricultural subsidies in the validation of subsidy claims.
We also look at environmental degradation, such as nutrient problems with forestry, habitat being eroded or invaded.
It’s multispectral optical imagery broken up into seven different bands in the spectrum of light.
An optical sensor measures the reflectance of the Sun’s light from the Earth. If that’s broken up into different parts of the spectrum, you can analyze them separately. Using different band combinations, you can produce a great variety of information, or optical image sets, from the data.
We also work with radar data, which is a different product entirely — it’s the measurement of a radio signal sent out by the spacecraft. The return signal is measured and tells you altogether different but equally useful things.
No, we use optical imagery only.
We look at pixels representing the image of a cow, as seen from a great height.
Some satellites orbit the Earth at a 700 kilometers distance, which is quite some way. With a cow, we measure the spectral signature of that cow in contrast to its surrounding area.
In a perfect world, cows would all be on snooker-table-looking green grass, and they’d stand out perfectly well.
The real world isn’t like that.
We have to make adjustments for uneven pasture, shrub, rocks, and things like that — rocks, in particular, can give us false positives.
Is it a matter of taking an image and then going out to the field as soon as possible and seeing how many cows were in that field? Or is there another way of testing your results?
There is, happily.
This research has been carried out in the European Union. There is good inventory kept by the responsible authorities for who’s got what cows.
Cows are ear tagged. It’s possible to determine what a herd keeper has at any moment as an inventory.
What’s not possible from the ear tag-based tracking is to trace for food security or tell where the herd keeper has the cows. Often, farms are spread out in ways we wouldn’t think of. Spread out is different in places like Australia, Canada, and the US, because they’re big.
In Europe, spread out looks different. There might be a cluster of land parcels in one area, and then at some distance, there might be another. The purpose of this exercise was to see if we can match what we’ve observed as a point in time with what the farmer is recorded as having by way of livestock.
We add the extra dimension ofwhere, which is always important in geography.
Take Ireland with approximately 8000 farms spread over 32 sq kilometers. The average number of clusters of parcels is three.
Suppose there’s an outbreak of a notifiable disease, such as the airborne foot and mouth in Europe in 2000/2001. In that case, action must be taken quickly.
Vets need to test and isolate. If they go to the wrong place where there are no animals, that will cause a delay. It’s important to know where the animals are at any point in time to expose any required activity to the right area.
Can you determine their shape, size, state of health? Can you tell if you’ve seen that cow before?
What we do, as a sidebar module, is count the breed.
Common breeds are black and white or brown cows. Farmers have a particular breed for a specific purpose.
Another challenge, I suppose, is to see if the cow or a sheep is a mother. If the ewe is rearing a lamb, we get a slight difference in pixel count for young lambs.
It’s determinable which side of the animal the lambs are on. You’ll see some on the left and some of them on the right, but so close by, it’s a measurable entity.
Yes, we do.
In temperate climates, you get poaching — the animals gather around a water trough or at the edge of the field where they come in the morning if they’re housed at nighttime.
That leads to the churn of the vegetation cover, and you end up with exposed earth. That’s not necessarily a good sign or a bad sign.
But in context, it can tell you something else. For instance, if there are no cows in the field, and you can see evidence of poaching, it’s a sign that the field is being actively farmed.
Meteorological data is a hot topic for us at the moment.
In mapping, we speak of the skin of the Earth. It’s the same for imagery. If you get an image where you are interested in certain features, you also get the features surrounding the interest.
That means you have contextual information. Meteorological data is important because you can have two fields of precisely the same size with exactly the same soil. Still, one of them gets rained on every day, and the other one doesn’t.
There’s also Sun angle — North facing or South facing in the Northern Hemisphere can make a difference, particularly for noticeable slopes or extreme slumps.
We use all the information that we can get. We find that soil types and meteorological data history are equally important.
You look at a situation today and decide. Then, if you can look back to the archive of imagery, you may fine-tune that decision. Maybe you can spot a problem. You can see when and where the feature of interest started. You can give a much better analysis of a situation to somebody who’s looking at this as a problem.
There are massive changes taking place in the ability to process and store data. They’re having an enormous impact on the Earth observation industry.
It’s also interesting to see how parallel changes in different domains come together. That’s what brings about positive improvement.
The cost of launching a satellite is changing. It used to be a big couple of tons of metal thing launched almost exclusively by NASA. Now they’re with the cube sets at just a couple of kilos and send back data from space like the larger satellites.
According to a five-year-old statistic, about 78% of optical data was not being used because when it was downloaded, it was cloud occluded — cloud was covering part or all of the area of interest.
Some technical changes, such as up-to-date meteorology or trends on cloud movement and leaving a certain area of interest for another day, brought the 78% down to about 55%.
It’s still a high percentage of waste, but it’s a significant improvement.
You put a chip weighing about five grams on the satellite. You task it with doing something — it’s able to determine if the area it needs to include is covered in cloud or not. If it is, it will not do that task that day — it’ll do something else and come back to it another time.
That means if you’re ordering satellite imagery, there is no waste. That does so much more for the use of the scarce resource, which is satellite time.
Five grams can solve a big problem.
Soon, we can do forestry analysis of an area for nutrition deficiency that will give us our results by pulling the image processing — essentially where the action is.
It would let scientists workon the data and work the datainto information the use case needs.
This is something we’ll be doing in the next phase of the project (currently at phase two).
The standard method of determining whether something out there is cereal is to analyze the infrared parts of the spectrum. That gives you a unique signature for barley, wheat, and oat.
People accept that — they know that shade of pink as barley, and they know the other shade of pink is wheat.
But if you stop and think for a moment, you’re not actually looking at barley — you’re looking at something that represents barley to you.
If you take that piece of logic, you realize you can’t actually see the cows either. You can’t. The human eye doesn’t allow us — there’s something in a parcel that gives you coherence with the presence of animals.
You can do that by regularly updating what you’re getting versus what you previously have available with optical imagery.
We can tell whether this is barley or wheat without actually seeing the barley or wheat. Animals can also be detected using a coherence or correlation with other data sets.
It’ll be an interesting project to look at. When we were first approached to take animal tracking on as a project, we were more skeptical than we should have been. We weren’t sure it was going to be as successful as it’s turning out to be.
You just don’t know what you’re going to find, but that’s really part of the excitement.
Tiny species, possibly not unless they’re their habitat constrained.
If you find the habitats, you’ve got some correlation that the species you’re looking for might be there.
Outside of that, if we can tell a Hershey bar with a good confidence level, any animal of similar size can be detected, unless it has a Darwinian camouflage that fits in with its surrounding environment so that it’s safe from predators.
That might obviously be a challenge.
For cattle, for example, we take in the animal and the shadow.
Doing that gives you an idea of mass, so you know it’s not a 2D painted picture of a cow. I’m saying that facetiously — it’s more information about the object of interest, and it’s another feature of the object.
Can you use the same analysis to identify animals in New Zealand? Would you have to adjust things slightly, depending on where in the world you’re looking, assuming that the same data was available?
If the same data is available and if the topography is similar, which it is for New Zealand, in many parts, yes, the model is transportable. It would need calibration of the workflow, but that’s relatively straightforward. Once the area of interest and the species are identified, the processing will depend on that context.
For example, I live in a very verdant country — the famous forty shades of rain in Ireland. We get rain every second day. Cattle are raised around the world in far more arid terrains than we have here. The surrounding vegetation will look different. To us, it would look stressed, but somewhere in Africa, it’s not stress at all. It’s just the way it is.
We would calibrate our workflow to consider the difference in the setting for where we will carry out the inventory. Once that’s done, it’s the same process.
I want to make it look like I have more livestock than what I actually have. What would I have to do? Can I buy plastic cows and put them out on the farm?
Yes, you can buy some plastic or inflatable cows and put them out on the farm.
You’d still face several challenges. You’d have to put them out early in the morning and move them around. If they were inflatable, you’d want to make sure they’re well tethered if there’s any chance of high wind. For animals that move, it would be a high maintenance solution.
Looking more than once shows if movement has taken place. If the animals are in the same spot they were in two days ago, a week ago, or some time ago, then there’s clearly a problem.
Plus, any system like this needs invigilation.
As an example, let’s take an entirely different domain. We all fill in our taxes online, and we fill them in on trust. There’s always a possibility that you might get an audit. If you get an audit and fail, then you’re in big trouble.
For subsidy schemes, there’s a high level of conformance and compliance. Farmers are honest people, and they do their best to fill out complex application forms. Most of the errors or nonconformance we’ve seen have been down to accidental errors.
Any farmer and any of the processes we carry out can be audited — annually. There’s invigilation required if there’s a payment of public funds.
What’s stopping people from putting out large plastic sheets that look like cars, so the K-mart or Amazon parking lot looks more packed than it is? There’s a lot of speculation on the correlation between parked cars and how the business is doing.
It would certainly game any system if you had an inanimate object that does not move frequently.
But cars in parking lots, say outside an office block, they’re parked in the same designated spot every day.
If you put a plastic sheet or a blow-up car there, it’s going to have the same appearance.
But the thing about forecasting commodities, or the health of an economy based on the number of cars or containers entering or leaving a parking lot or a port, is the additional data sets you refer to.
They help to prove or disprove.
If you take any single data set and blindly follow it, regardless of what it is, you’re going to be error-prone.
You’ll need other data sets to validate what you have. You’ll also need validation processes — spot-checking or using other pieces of data, such as discovering that the Kmart car park has more Ford cars in it than Ford actually produced.
I am interested in monitoring all the cows in Denmark. I want to carry out this livestock inventory. What are the data requirements? Do I have to wait for seven or eight sunny days? What would the output of the inventory look like?
If you had a vector shape of each field, you’d give us the field details. We’d tell you whether there were animals there — the animal of interest to you — and how many of them we’d estimate to be there.
If there’s one cloud-free day for an area the size of Denmark, the imagery will be captured in one, possibly two days. Then it’s a processing issue. Happily, processing these days with the way processing capability has gone is a variable. You can expand your processing capability to meet the demand.
Is it a simple inventory of 15 cows in this field, 30 cows in that field, and no cows in the other field?
Do you need to know something else?
Do you need to know how many cows are in a particular area, a cadastre, in Denmark?
Do you need to know the change between this year and last year, or between seasons?
What are the questions you need to answer?
There are many. An interesting one is monitoring the catching of tuna and whether somebody was fishing illegally and not.
If someone has a license to fish to catch one standard net of tuna — a big net ring — they swoop up the tuna and bring them back onshore.
If somebody is fishing illegally, they’ll have a second net. It’s possible to tell from satellite imagery if the net is full by looking at its shape.
If it’s full, it’s oval. If it’s empty, it’s round.
You might think, what’s the big deal about an extra fishnet.
With tuna, each one of those can be worth about two and a half million dollars. Plus, the damage to stocks, not paying any taxes, and violating environmental rights.
Satellite imagery helps to detect that because you can see from space whether there are only two or possibly three nets and whether they’re full or empty.
Another use case that I came across outside of agriculture was radar imagery discriminating the difference between ice and water.
Two guys developed a system which they then subsequently sold to cruise ships. People who go on cruises to the Arctic areas and the Antarctic want to see ice. They want to see big sheets of ice.
Their system delivered a prediction of where ice flows in relation to the cruise ship’s course. The captain can safely adjust his route slightly, in the right direction, so everybody gets their photo op.
Brilliant idea. Simple and well-executed.
I’ve been doing this since 1987.
Things have changed an awful lot. It’s still exciting and I still find it interesting. I learn new things every day.
Find something you love doing and you’ll never have to work a day in your life.
I’ve been very fortunate because I am interested in the geography business that finds new things to do every day. It’s a great time in the industry with the explosion of processing power, the proliferation of new spacecraft, and the latest applications and youthful energy.
It’s a mixture of the two.
We work primarily with governmental agencies. Our clients are official bodies. But we have a research arm looking for new ideas, new applications. We’re fortunate to have a good relationship with the European Space Agency, which supports new ideas.
One thing about that type of research is that no is an acceptable answer.
If you look at something and it’s not possible, then that’s an acceptable answer. It’s been looked at and the examination of the issue has value.
It can also mean that when somebody looks at it again in five years, it becomes possible.
Research is fundamental to who we are and what we want to be. The European Space Agency’s support is vital for us to continue.
They’re a great font of knowledge and very good at making sure we follow things logically.
They’re a vital resource for us.
We’ve developed a repeatable method of doing that.
I divide up what we do into two areas.
One is the fulfillment of a statutory obligation that has to be done by law. Things that must be checked, like farmers growing crops and claiming subsidies.
Then, the other stuff is extra education.
We say, “Here’s a good idea we have. Would you be interested in looking at that?”
Our approach is customer service discovery. We start off with the basic questions and we continue to ask those questions in a multiple-choice environment, so the potential client thinks about what they might like and what they might not like.
We then tell them what we think wecan do and what wecan’t do.
The client ends up with a well-designed specification for what might or might not be done.
This approach has served us well. Often, with technology that’s outsourced because it’s not available in-house, the business doesn’t understand the limitations or the possibilities.
That’s where they need our expert guidance.
We bring them on a journey.
I have a confession to make.
My son joined the business two-and-a-half years ago.
He’s making excellent progress. It’s an industry that has a long way to run in innovation, excitement, and possibilities. When we look at environmental issues, food security, and security for people, the industry plays a role in keeping the world a little saner than it might otherwise be.
An inquiring mind that can think logically.
Those two things appear to be opposite to each other. Sometimes an inquiring mind seems to think of illogical things.
But if you have an idea and you say, “Let’s test and have the fun in that,” you’re using your logic and your inquiring mind.
Be sure to subscribe to our podcast for weekly episodes that connect the geospatial community.
For more exclusive content, join our email. No spam! Just insightful content about the geospatial industry.
Crafting a quality application for a job you really want takes time, so you do not want to spread yourself too thin. When constructing your CV, it is important to keep your audience in mind. Realistically, the first set of eyes will likely be a computer algorithm, scraping the submitted CVs for certain keywords.