I'm here from Orbica which is a company in Christchurch and we do spatial data but we also do a couple of passion projects and this is one of my passion projects. Now we're going to start this with the most stupid question of the day. Who does remote sensing in GIS? Put your hands up. Actually that's good so let's reverse it. Who doesn't do remote sensing in GIS? That's perfect.

This tool which is a design for people to use to look at Sentinel-2 satellite data came around from a couple of things. One is we've had clients talk to us and say we want tools for not GIS operators but environmental monitoring consents operators to look at changes in landscape before and after so someone

with no skills in that field and my dad. My dad is obsessed with that where on earth show that's on Discovery where they show a couple of satellite images and randomly talk about completely different interpretations of them. So we kind of, he loves this stuff, he loves looking at this stuff so

we kind of started thinking about how we could start building an application that sort of addressed that. And when I was teaching one of the sort of the major things we always show to people is the Ural Sea on the border between Uzbekistan and Kazakhstan and now over the 20 or 30 years since water had been diverted how the landscape had completely changed

from a rich ecosystem, it supported fishing fleets, to a highly eutrophic wasteland pretty much which so it's pretty much been recognized as one of the largest man-made ecological kind of disasters on the planet. And this is a fantastic way to view it and this is looking at something in a spatial

context but also looking at in a temporal context what's happening with changes over time. And this is a question you know as geographers, geologists, physical geographers, we all want to know this temporal aspect. How does landscape change or how does human habitation change over time and how can we view it really easily. And satellite imagery is sort of one of

the key things for this. It's not really a fun process if you're doing it in the back end, you have a satellite that beams data down to a ground station, this has some sort of spatial resolution on the ground, it has some sort of time-based resolution and some sort of spectral resolution, the

wavelengths of data it's connecting in different slices. When you get that data and there's petabytes and petabytes of this data floating around, you can start thinking about how you process it. If you really want to do some really nice spectral analyses these have to be corrected, you have to

correct them for the water absorption in the atmosphere and also correct them on the ground for the geometry of them and basically make sure these are georeferenced. Then if you're going beyond small chunks of data, we're going to remove clouds, want to make sure that the data we're looking at is actually cloud-free and in New Zealand that is, I can't even go into how painful it is,

cloud-free New Zealand does not exist, land to the long white cloud constantly and mosaicing data from individual tiles into larger areas and how we serve these. So if we're thinking of this concept of people being able to access data without doing any of this stuff, we pretty much just want to go from some

sort of satellite imagery to the desktop for a user and the easiest way to do that is pay for it, but it gets really really expensive really really fast. If we look at our WorldView 3, this is kind of the

multi-spectral and 30 centimeter panchromatic and we use the panchromatic band to sharpen, you're looking if you want a task, it's about 30 US dollars a square meter and if you want that without cloud that's basically another 15 US dollars on top of that for under 5% cloud and you're

going to buy a minimum of 100 square kilometers. So these prices ramp up really fast, you can't buy them in small chunks, so if you're looking at small farms wanting to look at crop changes, governmental organizations wanting to look at how you can look at small chunks of land that may have

changed land use type, especially for consenting, it really puts it out of the game park for a lot of people and when we talk to clients like city councils they say we want high resolution as possible in near real time to the desktop for free and it's ridiculous how many times I've heard

this and of course that's completely impossible you know especially when you add free into it, it really throws a spanner in the works. So we kind of came up with this idea for this application that someone who was working at a desktop, a consent officer could look at something quickly to make quick sort of observational changes. No really flash tools in it, just something

really simple they can look at it before and after an event. So the data requirements for this is obviously number one, we're all Kiwis, we love free data and the Australians I presume love free data, everyone loves free data, so highest resolution possible based on the caveat of number one, short

revisit times based on the caveat of number one etc etc etc. Application, got to be free, everyone loves free and I call it the dad level of being able to use it. If I can put it in front of my father and he can run it that is the perfect target of operations. Anything more complicated than two or three

buttons it breaks. So to do that we also needed to have all of our data processing, pan sharpening, band creation, indices creation has to be on the server side and we wanted things that worked relatively fast so people could just jump on the desktop and it's kind of that seamless experience like a

Google Maps. As I'll show you today that didn't happen because the service is now playing up and everything's loading really really slow. Whenever you do a tech demo this seems to happen, it seems to be some law. So we'll show you some of the data but it's a bit of pre-K stuff. This application is for

anyone to use, you can jump on the website, have a look at it, have a play with it because we're interested in feedback of how people use it and how people interact with it. This is a very alpha version so it's something that we're still developing. So when we think about data and we think about free data, Landsat and the Copernicus program are sort of the the gold standards of what we can get for free. Landsat's roughly around 30 meters

a pixel, it's been collecting Landsat 5 since about 1980 something, 84, 85 and Copernicus, especially the Sentinel-2 which we use for our sort of RGB imagery has been running since 2016. So we've got some quite good free datasets out there at various scales that people can look at for free.

And again it's that process of how does someone who doesn't know about remote sensing or know about GIS access this data. It's free, people can jump online, they can start grabbing it themselves, they can do all the processing themselves but if you don't have that background knowledge it becomes far far harder. So we're really concentrating on Sentinel-2. Sentinel-2

is very spectrally rich, you can do a lot with it, it's 12 bands of data that come through at various resolutions from 10 meters up to 60 meter resolution but it's for free data it's incredibly rich. So we started building a website and I started building a website and it was awful, Benika at the

back and Santosh they took my disastrous code and refactored it into something useful and so we basically used all these open source technologies but pretty much the main guts of it are hanging off leaflet as the mapping side of it and we use a service called Sentinel Hub. Sentinel Hub is a service

run out of Czechoslovakia, not Czechoslovakia, Slovenia, there we go, it was either Slovakia or Slovenia, great guys and they released this product where you can basically use their really cool back end with API's to do stuff. We thought about doing our own data cube but for

lazy old me this was a really cool way of getting stuff prototyped really quickly and so Sentinel Hub kind of works in this way where they have a whole bunch of data up there so they've got Landsat for all of Europe unfortunately not New Zealand but Sentinel data comes down all the time from late 2015 everything comes down to the AWS server which you can access

the cool thing is that they have at this server in this sort of ability to write your own scripts so for us wanting to build a pathway to a rest endpoint we can start saying let's code up some back end stuff we can do some indices we can do pan sharpening and that's basically directly handed out as

a service to a website and especially when we can do indices we can just code as many indices as we want so in the back end I think we have 50 something indices or something ridiculous like that because you can just keep on programming them and they just appear as services. The other cool thing with Sentinel Hub is it also has a full Python API so we can do machine

learning we can do build our own data cubes from it which gives us a lot of power to do stuff in-house and on the desktop as well and the reality is that's a really cheap service for what you get a commercial account with a low amount of hits is like a hundred euro a month it makes it quite fun for

us to develop stuff in-house and play with stuff and see how we can sort of develop these products so the product we came up with is something called Explorer it is designed around time it's designed around virtually no controls dad-level controls date and some pull downs so what I'll show you now is

hopefully I'll show you this is the bit where I hold my breath so this is how the app looks it's really simple you log in and it allows you to basically go through and choose various functions now I'm drunk driving over

here basically it allows you to choose different sort of late levels of complexity and we've got a few number of different indices up here and two

dates and two times so not only can you compare the same time with different data you can compare different times with the same data and so this is really good for sort of various little sort of tools that we use this here is an area of Christchurch this is large amounts of irrigation that's gone on and to look at the difference between where's my eyes

when I need them 2019 and 2018 we can just slide between them and so it's designed so you can easily look at landscape change at this level we're adding a bunch of other tools into this one is a statistics feature service where a user can come along draw a little polygon around it and say

what's happened in this time and it will plot up a difference so this is a normalized difference water index now moisture index for this one so it's very sort of very simple for people to get access to these really simply so here's another example Coffs Harbour we're looking at the fires at the moment so

this imagery is from five days ago Sentinel captures imagery every five days so it automatically updates to the latest imagery and we're looking at one of the shortwave infrared bands and you can see where the active burn fronts are

really clearly so again really simple I can send this to my dad and go hey look at this this is kind of interesting and you get things like that so we have a bunch of different things we have vegetation indexes on there a number of them we have the swir we've got I think they are the main two

examples I had up so we have a bunch of different things that allow people to really simply go in there and and look at data like those indexes which they wouldn't have been able to do before so yeah so as I said this is an alpha stage we've still got lots of work to do on the API and even the UX that's

basically a passion project we're working on on the side statistics service is really key for us Sentinel hub completely supports it basically you click on a point and it just generates the time series of that point with a really nice API change notification is really important to us where we can

start going how is an area changed over time can we get a notification from it again where if we think about the council level where they're looking at a tiny area and saying has this been deforested draw an area around it and look at the changes over time dad also ties into more data sources our

key at the moment is building an open data cube with landsat 5 that goes back to 1985 there are lots of interest around the emissions trading scheme forestry pre 1990 post 1990 so this gives us an awesome tool to be able to say to people how does this change over time what areas are you interested in so yeah that landsat data cube is sort of one of the big things and

we're going to use the open data cube that Alex and Caitlin have been working on so yeah and the other thing is doing animated gif exports everyone loves them just being able to quickly choose an area and saying produce an animated gif that covers you know very short time periods or

very long time periods cloud relatively cloud free yes so that's pretty much it again it's a it's a tool we want people to go ahead and use and sort of experience and dare I say it break it and try to figure out how we can do better things with it we have a thing downstairs with a little stall downstairs with a couple of computers that are running so people can

jump on it now otherwise go to explorer dot orbiter dot world you sign up no we're not going to spam you with anything but it's just a way for us to collect users and collect some feedback from you at the end cool that's pretty much me any questions yeah we have plenty of time for

questions oh hey how short am I you were 14 minutes so we have any questions from the audience all right well I will happily kick things off so I really love that you tested this you know you were saying it had to be dad

friendly how often did you test it with your dad quite a bit okay I do something I do a change and say can you can you operate this and he'd be like what does this do how I don't understand so there was kind of different iterations and I put up things like NDVI what does that mean

well it looks at plant health by looking at near-infrared absorption blah blah blah and it I don't so so then and that's why you go it doesn't come up as NDVI it says plant health you know and then there's a little drop-down that says this is an NDVI so it was that you know it is an end user

focused product where we've kind of thought about the kind of target audience who will use it and then worked our way back instead of going we're a bunch of GIS remote sensing nerds we can do all these cool tools that are just going to completely just destroy people if they don't have that not background knowledge behind it so it's something

you know in the focus on the end user is something we're really keen on at Orbica and that's kind of our development almost runs backwards because of that can you add third-party data sources yeah yeah again

it's we're just pulling and the Sentinel hub data is a resting point WMS that's a tiled WMS basically so we are looking at when we move to the descent when we move to the Landsat 5 stuff that's got to be separate because Sentinel hub give Landsat 7, Landsat 5, Landsat 8 but they've only

cached it for the northern hemisphere and so we can't use it so we're starting to look at how we can add other data sources in there because it's driven on leaflet we can we it's easy enough to start using a leaflet side of it just to pull in feature services etc etc so we're only

using the Sentinel hub as one part and our sort of long-term plan is we'll move away from that and start doing our own data cube if and when in New Zealand wide data cube arrives I know there's lots of discussion about that at the moment you know being able to tap into that pull information on again it's it's about what the end user wants we can start just loading that

data up you know we all know there's a million different free data sources around the place you know you can throw it all up there in my database you know so it's about it's about avoiding that situation and really sort of you know as I say this is this is for GIS not for GIS people you almost need to

trim that GIS part of your brain out and think about you know I'm interested in landform change I'm interested in how cities change what do I need to see that at a simple level and this is why it is so simple and why we haven't gone overboard and all these different tools I'm developing this

product from receiving the data from the satellites and then having that web HTML server what will you main choke point so what really took your time in developing so you had this idea and then what really took time I haven't

coded in like 20 years was probably the biggest stumbling point so I was kind of learning CSS I was learning JavaScript kind of on the fly looking for examples and kind of mashing things together and that's why when Benika jumped on board and took over my code I think I think we basically

started from scratch again again it's it's all it's a bunch of simple tools and you know you go to GIS teams and you say well we've got this thing that slides between two different lot of data layers and they're like that's not interesting at all you know we can do that been able to do that for years and not GIS but putting it in the context of satellite data and

putting it in the context of change for end users who don't have experience so that was kind of the big thing you know it's how the probably one of the hardest thing for us in the development is what we didn't put in we had all these cool ideas and it's like well we can't really put that into something that someone's not ever going to use we just have to keep on

stripping it back until it's you know the bare bones of what it is all right we've got one time for one more question at the back so it's good that it's the last question because I have to ask your favorite imagery you've played with oh we've you know I did lots of work in Antarctica and we use worldview free down there and it is spectacular but we did some really cool

stuff looking at we were playing around with visualizations and we were looking at the irrigated land through Canterbury and actually doing weird time series where each of the RGB bands each of the colors was represented by a different time and a different date and it was like these pure pastel

driven LSD crazy you know I really want to take some up and blow them big up on my walls because they looked amazing and it was just because we're taking that temporal data and using it in a really different way it just produced these amazing results that was yeah really really interesting