Skip to main content
Search form
  • 00:18

    SAMANTHA LAVENDER: At one time, the issuewould have been could you find enough satellitedata to answer your question.Now the problem is can you wade through the satellite dataand extract the bit you're actually interested in.My name is Samantha Lavender, and I'm the managing director

  • 00:40

    SAMANTHA LAVENDER [continued]: of Pixalytics.Pixalytics is what we term an earth observation company.So it's really about observing the earth from afar,taking data from satellites in space,converting that information through computer programming,but then also talking to the people who's

  • 01:01

    SAMANTHA LAVENDER [continued]: going to use that data.So most people, nowadays, are aware of Google Earth,and have looked at their house or other things on it.But that data that you see in Google Earthis really only a fraction of what'sbeing collected by satellites in space.In Europe, we have that the European Space Agency.

  • 01:23

    SAMANTHA LAVENDER [continued]: In America, you have NASA and NOAA.And then also you have space agenciesin China, South Korea, India, all over the world really.And a lot of these different space agenciesare launching satellites, and many of themmake their data freely available to download.

  • 01:45

    SAMANTHA LAVENDER [continued]: And so we take that data, and we turn that into information.And it's really about trying to thinkabout what people need to know and can youmonitor that from space.One of the projects we're working onis funded by the European Space Agency,

  • 02:07

    SAMANTHA LAVENDER [continued]: and it's part of their climate change initiative.And this particular project is lookingat the color of the oceans, termed ocean color.The open ocean is dominated by phytoplankton,which are microscopic plants, and theychange the color of the water.Phytoplankton absorb carbon dioxide and breathe out oxygen.

  • 02:32

    SAMANTHA LAVENDER [continued]: So actually, the oxygen we breathe as humansis, in part, coming from the oceans and these phytoplankton.We produce large and complex datasetswith lots of different parameters, whichare very useful for the scientists,because they import that data into models

  • 02:53

    SAMANTHA LAVENDER [continued]: and write papers about it.But for the general public, it's less easy to understand.And so what we've been doing is creatingthese sort of visualizations, so people can easilysee what the data looks like.This particular visualization is looking at the parameterchlorophyll concentration.

  • 03:14

    SAMANTHA LAVENDER [continued]: So that's the main pigment that doesphotosynthesis in these plants.And it's telling you how much of thatpigment is in different parts of the oceans.And so you can see the changing concentrationsof the phytoplankton as the seasons change.What we want to be sure is that it's not

  • 03:34

    SAMANTHA LAVENDER [continued]: decreasing where we're not expecting it to,or it's increasing in areas wherethere is maybe toxic blooms.And so overall, we get a view of how healthy the ocean is.It's really important to be doing these type of projects,and it's why it's being funded because thisis how the governments across the world

  • 03:57

    SAMANTHA LAVENDER [continued]: make their decisions on what to do for the futureand how to react to the changing climate.In my role in analyzing the data,I end up most the time writing software really,and this is in a combination of programming languages.

  • 04:19

    SAMANTHA LAVENDER [continued]: So it would be both more traditional languages,such as C and also a language called IDLthat I've used since my PhD, but also, there'snow Python as well, which there are a huge number of librariesout there for analyzing satellite data being written.And so we will write software to interrogate the files

  • 04:43

    SAMANTHA LAVENDER [continued]: and look for information, and thenoccasionally, we do use off the shelf software as well, sofor example ISA, have a package calledSNAP that they've written for their satellite datathat we use.And it's that combination that to the advantageof writing your own code and runningit is you can do batch processing and process

  • 05:05

    SAMANTHA LAVENDER [continued]: hundreds of files, whereas if you're interactivelyusing a package, it's much more difficult to dealwith large numbers of files.Traditionally, you would have waited for a space agencyto launch a satellite, and you would have used their own data

  • 05:26

    SAMANTHA LAVENDER [continued]: that they're providing.But the problem with that is the development of those satellitestakes a long time, and they're big and very expensive.So there tend not to be such huge numbers of them.What the commercial side of the industryis seeing increasingly that they can launch nowwith smaller satellites, and the cost

  • 05:47

    SAMANTHA LAVENDER [continued]: of building those satellites is coming down,but also the cost of launch is coming down as well.So instead of having a few expensive very high qualitysatellites, you can now launch many more cheaper satellites.So rather than waiting for a satelliteto come around every few days over your area of interest,you might have a whole constellation

  • 06:08

    SAMANTHA LAVENDER [continued]: of satellites that are coming around every few hours,taking imagery.Another project that we've worked onwas funded by the UK Space Agencyunder their Space for Smarter Government program.The project was called OMEF, Optical and Microwave Extension

  • 06:29

    SAMANTHA LAVENDER [continued]: for Floodwater Mapping.And it was to look at whether we coulduse satellite imagery to better map flooding in urban areas.We decided to do that using both optical and radar imagery.With radar data, you're operatingin the microwave part of the spectrum

  • 06:50

    SAMANTHA LAVENDER [continued]: and the big advantage of that is that you see through clouds.So this is one particular image, and thiswas collected in January 2015 by a satellite called Sentinel 1.This is the city of York in the UK in the centerhere that was badly hit by flooding the previous November

  • 07:13

    SAMANTHA LAVENDER [continued]: and December.So this is the image after the flooding has occurredand primarily subsided.If I now load on top of that the imagewhen the flooding was occurring, then youcan see the dramatic difference.So all these very dark areas is where you have the flood water.

  • 07:36

    SAMANTHA LAVENDER [continued]: As we process the data, we produce another layerthat is just maps of that flood water.And so by having the different shades of blue,you see how the flood has changed over time.We're looking now at taking what we've done in this project

  • 07:59

    SAMANTHA LAVENDER [continued]: and making this a commercial service.So when we have flooding in the future,instead of us doing this research after it's happened,actually, these images are processed in real time,and people can look at these flood maps as they're created.

  • 08:20

    SAMANTHA LAVENDER [continued]: Data quality and accuracy is critically important.The data is never going to be perfect.There are always going to be issues with it.Particularly contracts we do with the European Space Agency,where we're quality controlling data before it goes outto the scientific community, we can spend months, if not years,

  • 08:41

    SAMANTHA LAVENDER [continued]: looking at data sets, trying to find issues with itto make sure that what the scientists don't interpretisn't something that's an anomaly in the data,rather than something in the real world.Now, we have programs like Copernicusthat is being run by the European Union that

  • 09:03

    SAMANTHA LAVENDER [continued]: has an increasing number of satellites being launched,and you can be downloading gigabytes, if not terabytesof data a day over the internet to actually usefor your own applications.And so that's what is now becoming the huge gamechanger in what we're doing.

  • 09:25

    SAMANTHA LAVENDER [continued]: One of the problems we're increasingly facing now isbecause satellites and also drones, to some extent as well,are creating so much data, actually the internet cannotcope now with downloading this data.And what the space agencies are starting

  • 09:46

    SAMANTHA LAVENDER [continued]: to think about is, instead of getting everyone downloadingthe data and then having to find somewhere to store it,but also taking up so much bandwidth by doing that,it would be better if people couldstart transferring their code and their processingto where the data is.So it's really changing from whatwould have been traditionally FDP, where you got whole

  • 10:08

    SAMANTHA LAVENDER [continued]: [INAUDIBLE] to more an interactive system,but there are questions to be answered over that.There is the security of your own softwarewhen you're putting it on someone else's server.In my early days, when I sat up and ran a processing systemfor one particular satellite, every day

  • 10:31

    SAMANTHA LAVENDER [continued]: as the data was being downloaded in processes,I would look at it to just check it was correct,and nothing had gone wrong in the processing.Nowadays, we're pulling data from lotsof different satellites.So what we have to do is increasingly writesoftware that tries to do this, and thisis where machine learning has a roleto play in that you can maybe spot anomalies in your data,

  • 10:54

    SAMANTHA LAVENDER [continued]: but also, more traditional techniques,as well, where you're setting thresholds,or you're doing statistics on the data,and you're testing the files to seewhether the parameters are in the range as you would expect.Looking at the data is critically important,

  • 11:16

    SAMANTHA LAVENDER [continued]: because if you write a piece of software,it will produce an answer, but that'snot necessarily the right answer, and especially, as youstart dealing with more complex data,and you're trying to simplify what'shappening in the real world.You have to make assumptions.And quite often, those assumptions

  • 11:37

    SAMANTHA LAVENDER [continued]: will go wrong, or the model you have is not hugely accurate.You have to look at the images to see if they make sense,but you also have to make plots and graphs and statisticsand look at those as well.And it's really that sanity check,and it's very difficult to write softwarethat does that sanity check as easy as a human can do.

  • 12:01

    SAMANTHA LAVENDER [continued]: And so even in projects where we've got softwareanalyzing millions of images, actually, we stillhave an element of that, where we look at them.And we can be looking at maybe 10,000 images,and we just create what we call quick looks into almosta video, and you just watch it.

  • 12:23

    SAMANTHA LAVENDER [continued]: What you find is more you get lookingat a particular set of image, and I was doing this yesterdayfor a data set I produced, the more you can spot with your eyevery quickly something that's wrong.So there's definitely still that very important human component,and that's certainly trying to be addressed with techniques,

  • 12:43

    SAMANTHA LAVENDER [continued]: like machine learning and artificial intelligence,but it will be a long time and maybenever before they can get to what the human brain can do.So over the last couple of years,we've been developing something we're currently

  • 13:03

    SAMANTHA LAVENDER [continued]: calling the Pixalytics portal, and the idea is that someonecan log on to that website, they can pick an area that they'reinterested in, and they can download it there and then,and then they can take it into their own software,or they can view it, just as a pretty picture if that's

  • 13:23

    SAMANTHA LAVENDER [continued]: what they want, or it might be a graph of data and use it.And then that makes data really easy for people to use.One of the things we did in Pixalyticsactually was we wrote a book to try and helppeople who just have a computer or they'reinterested in remote sensing, download data,

  • 13:46

    SAMANTHA LAVENDER [continued]: and use it themselves, using both this free to access dataand free to access software.The really exciting thing is alongside the satellite data,we have lots of other data now, increasingly being collected.So you have data from social media.You have data from people's smartphone.

  • 14:07

    SAMANTHA LAVENDER [continued]: You have data from internet of things devices.And so really it's increasingly becomingnot about using one source of data,but using all that data to get the understanding that youwant.


Samantha Lavender, PhD, managing director of Pixalytics, discusses the practice of taking data from satellites, processing it, and making it available to users. Covered are aspects of mapping, optical and radar imagery, software, and the need for human verification of the quality and accuracy of this data in the age of data overload. Pixalytics is introduced as a centralized portal to access observational Earth data .

Looks like you do not have access to this content.

Working with Satellite Data for Earth Observation: Pixalytics

Samantha Lavender, PhD, managing director of Pixalytics, discusses the practice of taking data from satellites, processing it, and making it available to users. Covered are aspects of mapping, optical and radar imagery, software, and the need for human verification of the quality and accuracy of this data in the age of data overload. Pixalytics is introduced as a centralized portal to access observational Earth data .

Copy and paste the following HTML into your website