Blue Phase at Wavelength 2018

Blue John Cavern

Last week I attended the 2018 Wavelength Conference in Sheffield. This is an annual gathering for the Remote Sensing and Photogrammetry Society (RSPSoc) and is geared towards PhD students and early career scientists. The conference aim is to provide a welcoming and constructive atmosphere to present research and progress towards PhD’s, coupled with a vibrant social programme.

This was my first experience of a remote sensing conference and the cosy nature of the common room where it was held alongside the lack of pressure of a larger event lent itself well to its ambition.

The topics covered by the research varied greatly, each with a focus on how to apply remote sensing and photogrammetry techniques in novel ways to better understand the world around us. These ranged from tracking whales to monitoring rice fields and developing systems to track small scale landslides.

One key technology which was popular among the presentations was the application of machine learning, the training of an artificial intelligence (AI) to classify images for a variety of purposes. Given it is something I’m becoming involved in at Pixalytics, every mention of AI attracted my attention. One presentation which stuck out for me was its application to track the effects of crude oil pollution in the Niger delta region. Harnessing remote sensing data and utilising the power of machine learning to sift through hundreds or even thousands of images, classify details and pick out objects of interest to monitor environmental damage is a novel approach. It provides a direct link from the science to a serious real-world issue. Whilst a localised case, the techniques demonstrated have the potential to better inform our responses to these issues which in turn will help people being affected by these disasters.

This application of science combined with the potential to one day help people resonated with me greatly. It reminded me of the work I am currently doing on the Drought and Flood Mitigation Service project which will aid the lives of Ugandan farmers.

Two keynotes were delivered during the conference, one by Dr. Alistair Graham, from Geoger Ltd, and one from the Chairman of RSPSoc Dr. Richard Armitage. Dr. Graham’s keynote was fascinating as he delivered his experiences working in a multitude of different environments from corporate to SME’s in industry to post doc positions in academia. He explained the nuances of working in each area and the possible paths for career progression open to PhD students and other early career scientists. I fall into the latter category, but the perspective he provided convinced me to keep my options open for the future. At a time when industry and academia is changing rapidly anything could happen.

Dr. Armitage’s keynote was on responsive remote sensing and his talk focused on how to use the right remote sensing data at the right time and for the right area. For the problems we come across, identifying the correct approach to take with remote sensing data is crucial.

For example, two important factors to consider for any problem are spatial resolution and data type. Some features require 5m to be visible, whereas for others the 30m resolution can show what is required. Further to consider is what type of data is best suited for the problem, optical data has its advantages but infra-red can reveal insights that optical data cannot. Having come across these points before the keynote, it served as a good reinforcement on the topic.

Blue John in the rock.

The highlight of the conference for me was the tour around Blue John cavern. Tucked away in the Peak District, surrounded by stunning views of the hills, the cavern is home to the famous Blue John stone. The tour guide was a miner who had worked in the cavern for 15 years and his knowledge on the tour was remarkable, making every stop ever more interesting.

Whilst a lot of walking and climbing was done, the colourful Blue John that spotted the walls of the cavern, together with the extremely high ceilings carved out by long gone rivers made for amazing views. If you don’t mind cramped spaces and traversing up and down a large mine, then Blue John cavern is a fantastic place to go!

For my first conference experience Wavelength 2018 was a fantastic introduction. The welcoming atmosphere, getting to see the diverse nature of remote sensing and photogrammetry research going on right now and the insightful keynotes will stick with me for a long time. I highly recommend any early career scientist or PhD student to attend the next incarnation of this conference.

Chris Doyle
Junior Software Developer
Pixalytics Ltd

EO Market Is a-Changin’

Artist's rendition of a satellite - paulfleet/123RF Stock Photo

Artist’s rendition of a satellite – paulfleet/123RF Stock Photo

Historically, if you wanted satellite Earth Observation (EO) data your first port of call was usually NASA, or NOAA for meteorological data, and more recently you’d look at the European Union’s Copernicus programme. Data from commercial operators were often only sought if the free-to-access data from these suppliers did not meet your needs.
However, to quote Bob Dylan, The Times They Are a-Changin’. NASA, NOAA and Copernicus are buying, or intending to buy, data from commercial operators.

However, as with many activities there are often precedents. For example, the SeaWiFS mission was built to NASA’s specifications and launched in 1997. It was owned by the commercial organisation Orbital Sciences Corporation and NASA conducted a ‘data-buy’. They’ve moved back in this direction last month as NASA issued a Request for Information for US companies interested in participating in the Earth Observations from Private Sector Small Satellite Constellations Pilot. The aim of this programme is to identify commercial organisations collecting EO data relating to Essential Climate Variables (ECV), and then to evaluate whether this would be a cost effective approach to gathering data rather than, or alongside, launching their own satellites.

To interest NASA the companies need to have a constellation of at least three satellites in a non-geostationary orbits, and the ECV dataset will need to include details of both instrument calibration and processing techniques used. Initially, NASA plans to provide this data to researchers to undertake the evaluation. According to Space News, 11 responses to the request had been received. Discussions will take place with responding companies over the next month and it’s anticipated orders will be placed in March 2018.

NOAA is another US agency looking to the private small satellite sector through their Commercial Weather Data pilot programme. To supplement their own data collections they’ve already purchased GPS radio occupation data and are planning to buy both microwave sounding and radiometry data.

Not everyone is aware that the Copernicus Programme also purchases data from commercial sources as part of its Contributing Missions Programme. Essentially, if data is not available for any reason from the Sentinel satellites, then the equivalent data is sought from one of 30 current contributing missions which include other international partners such as NASA, but also commercial providers.

Whilst part of the drive behind this approach is to ensure data continuity, in the US the backdrop has a more long term concern with President Trump’s intention to move NASA away from EO to focus efforts on deep space exploration. It’s not been fully confirmed yet, but there is due to be a Congress budget discussion later this week and if approved it could mean the loss of the following four NASA missions:

• Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) satellite
• Orbiting Carbon Observatory-3 (OCO-3)
• Climate Absolute Radiance and Refractivity Observatory (CLARREO) Pathfinder
• Deep Space Climate Observatory (DSCOVR)

Whilst buying data from commercial providers may offer opportunities, it also has a number of challenges including how to buy this whilst maintaining their commitment to free-to-access data, and with the shorter lifespans of small satellites the increased pressure on calibration and validation work.

It’s clear that things are evolving in the EO market and the private sector is coming much more to the fore as a primary data supplier to researchers, national and international bodies.

Unintended Consequences of Energy Saving

Black Marble 2016: Composite global map created from data acquired by VIIRS in 2016. Image courtesy of NASA/NASA’s Earth Observatory.

Last month a report in Science Advances got a lot of publicity as it described the increase in global light pollution following research using satellite data. Even more interesting was the fact that one of the key drivers, although not the only one, was the switch to LED lights which have mainly being bought in due to their increased energy efficiency.

Recently there has been a lot of night-time imagery released as photographs taken from the International Space Station, and we’ve used them in our blogs. However, night time imagery has also been collected from the uncalibrated Operational Linescan System (OLS) on the Defense Meteorological Satellite Program (DMSP) satellites for a number of years. This was followed by the Suomi National Polar-orbiting Partnership (Suomi NPP) research mission in 2011 that carries the Visible Infrared Imaging Radiometer Suite (VIIRS) which had a planned life expectancy of around five years, however it is still in orbit and continues to collect data. Much more recently, on the 18th November 2017, a second VIIRS instrument was launched aboard the NOAA-20 satellite (previously called JPSS-1).

The role of LED lights in the increase in light pollution was described in detail in the paper ‘Artificially lit surface of Earth at night increasing in radiance and extent’ by Kyba et al which was published on the 22nd November 2017. The paper was based on satellite data collected between 2012 and 2016 from the Visible Infrared Imaging Radiometer Suite (VIIRS) aboard the Suomi National Polar-orbiting Partnership (Suomi NPP) satellite and one of the key drivers behind the new research is that VIIRS offered the first calibrated and georeferenced night time radiance global dataset. Within the 22 spectral bands the instrument measures is a day/night panchromatic band (DNB). This band has a 750 m spatial resolution and operates on a whiskbroom approach with a swath of approximately 3,000 km which means it provides global coverage twice a day, visiting every location at 1:30 pm and 1:30 am (local time).

The team from the GFZ German Research Centre for Geosciences who did the research concluded that outdoor light pollution has increased by 11% over 5 years. However, for us, the really interesting part was that new LED lights are linked to this increase in light pollution.

Over the last decade within the UK, a lot of local Councils have switched to using LED streetlights mainly due to the energy, and associated cost, savings. However, there was also a message that this would reduce light pollution as they would direct light downwards and reduce nightglow. This is coupled with the fact that businesses and consumers have also been pushed to move towards this type of light for the same reasons. This was brought home to us recently as a firm opposite our home installed new outside LED lights. It has made a significant different to the amount of light in our room and even in the middle of the night it is never completely black.

What the research team found by comparing VIIRS images from 2012 and 2016 was that:

  • The lower cost of LED lights has actually led to more lights going up, mainly on the outskirts of towns and cities. A 2010 paper by Tsao et al published in Physics Today indicated that we tend to purchase as much artificial light as possible for around 0.7% of GDP and so as lighting becomes cheaper, the quantity increases.
  • Flat composite global map created from data acquired by VIIRS in 2016. Image courtesy of NASA/NASA’s Earth Observatory.

    There has been a shift in the spectra of artificial light within cities from the yellow/orange of the old streetlights to the white of LED’s.

  • The majority of countries of the world had seen an increase in light pollution. Although, perhaps surprisingly some of the world’s brightest nations such the US, UK, Germany, Netherlands, Spain and Italy had stayed stable; which may suggest there is a point of saturation of outdoor lighting. The only countries that had less light pollution were areas of conflict or whether there was issue with the data, such as Australia where there were significant wildfires when the first data was collected.

Light pollution has a negative impact on flora and fauna, particularly nocturnal wildlife, and there is increasing evidence that it is also negative for humans. This is an example of why we have to be so careful with the concept of cause and effect. Decisions made for improved energy efficiency look to have had unintended consequences for light pollution.

Sentinel To Be Launched

Sentinel-2 Image of Plymouth from 2016. Data courtesy of Copernicus/ESA.

Sentinel-2B was launched at 01:49 GMT on the 7th March from Europe’s Spaceport in French Guiana. It’s the second of a constellation of optical satellites which are part of the European Commission’s Copernicus Programme.

Its partner Sentinel-2A was launched on the 23rd June 2015, and has been providing some stunning imagery over the last eighteen months like the picture of Plymouth above. We’ve also used the data within our own work. Sentinel-2B carries an identical Multispectral Imager (MSI) instrument to its twin with 13 spectral bands:

  • 4 visible and near infrared spectral bands with a spatial resolution of 10 m
  • 6 short wave infrared spectral bands with a spatial resolution of 20 m
  • 3 atmospheric correction bands with a spatial resolution of 60 m

With a swath width of 290 km the constellation will acquire data in a band of latitude extending from 56° South around Isla Hornos, Cape Horn, South America to 83° North above Greenland, together with observations over specific calibration sites, such as Dome-C in Antarctica. Its focus will be on continental land surfaces, all European islands, islands bigger than 100 square kilometres, land locked seas and coastal waters.

The satellites will orbit 180 degrees apart at an altitude of 786 km, which means that together they will revisit the same point on Earth every five days at the equator, and it may be faster for parts of southern Europe. In comparison, Landsat takes sixteen days to revisit the same point.

With all Copernicus data being made freely available to anyone, the short revisit time offers opportunities small and micro Earth Observation businesses to establish monitoring products and services without the need for significant investment in satellite data paving the way for innovative new solutions to the way in which certain aspects of the environment are managed. Clearly, five day revisits are not ‘real-time’ and the spatial resolution of Sentinel data won’t be suitable for every problem.There is joint work between the US and Europe, to have complementarity with Landsat-8, which has thermal bands, and allows a further opportunity for cloud-free data acquisitions. Also, commercial operators provide higher spatial resolution data.

At Pixalytics we’re supporters of open source in both software and imagery. Our first point of call with any client is to ask whether the solution can be delivered through free to access imagery, as this can make a significant cost saving and allow large archives to be accessed. Of course, for a variety of reasons, it becomes necessary to purchase imagery to ensure the client gets the best solution for their needs. Of course, applications often include a combination of free to access and paid for data.

Next’s week launch offers new opportunities for downstream developers and we’ll be interested to see how we can exploit this new resource to develop our products and services.

Differences Between Optical & Radar Satellite Data

Ankgor Wat, Cambodia. Sentinel-2A image courtesy of ESA.

Ankgor Wat, Cambodia. Sentinel-2A image courtesy of ESA.

The two main types of satellite data are optical and radar used in remote sensing. We’re going to take a closer look at each type using the Ankgor Wat site in Cambodia, which was the location of the competition we ran on last week’s blog as part of World Space Week. We had lots of entries, and thanks to everyone who took part!

Constructed in the 12th Century, Ankgor Wat is a temple complex and the largest religious monument in the world. It lies 5.5 kilometres north of the modern town of Siem Reap and is popular with the remote sensing community due to its distinctive features. The site is surrounded by a 190m-wide moat, forming a 1.5km by 1.3km border around the temples and forested areas.

Optical Image
The picture at the top, which was used for the competition, is an optical image taken by a Multi-Spectral Imager (MSI) carried aboard ESA’s Sentinel-2A satellite. Optical data includes the visible wavebands and therefore can produce images, like this one, which is similar to how the human eye sees the world.

The green square in the centre of the image is the moat surrounding the temple complex; on the east side is Ta Kou Entrance, and the west side is the sandstone causeway which leads to the Angkor Wat gateway. The temples can be clearly seen in the centre of the moat, together with some of the paths through the forest within the complex.

To the south-east are the outskirts of Siem Reap, and the square moat of Angkor Thom can be seen just above the site. To the right are large forested areas and to the left are a variety of fields.
In addition to the three visible bands at 10 m resolution, Sentinel-2A also has:

  • A near-infrared band at 10 m resolution,
  • Six shortwave-infrared bands at 20 m resolution, and
  • Three atmospheric correction bands at 60 m resolution.

Radar Image
As a comparison we’ve produced this image from the twin Sentinel-1 satellites using the C-Band Synthetic Aperture Radar (SAR) instrument they carry aboard. This has a spatial resolution of 20 m, and so we’ve not zoomed as much as with the optical data; in addition, radar data is noisy which can be distracting.

Angkor Wat, Cambodia. SAR image from Sentinel-1 courtesy of ESA.

Angkor Wat, Cambodia. SAR image from Sentinel-1 courtesy of ESA.

The biggest advantage of radar data over optical data is that it is not affected by weather conditions and can see through clouds, and to some degree vegetation. This coloured Sentinel-1 SAR image is produced by showing the two polarisations (VV and VH i.e. vertical polarisation send for the radar signal and vertical or horizontal receive) alongside a ratio of them as red, green and blue.

Angkor Wat is shown just below centre, with its wide moat, and other archaeological structures surrounding it to the west, north and east. The variety of different landscape features around Angkor Wat show up more clearly in this image. The light pink to the south is the Cambodian city of Siem Reap with roads appearing as lines and an airport visible below the West Baray reservoir, which also dates from the Khmer civilization. The flatter ground that includes fields are purple, and the land with significant tree cover is shown as pale green.

Conclusion
The different types of satellite data have different uses, and different drawbacks. Optical imagery is great if you want to see the world as the human eye does, but radar imagery offers better options when the site can be cloudy and where you want an emphasis on the roughness of the surfaces.

The cost of ‘free data’

False Colour Composite of the Black Rock Desert, Nevada, USA.  Image acquired on 6th April 2016. Data courtesy of NASA/JPL-Caltech, from the Aster Volcano Archive (AVA).

False Colour Composite of the Black Rock Desert, Nevada, USA. Image acquired on 6th April 2016. Data courtesy of NASA/JPL-Caltech, from the Aster Volcano Archive (AVA).

Last week, the US and Japan announced free public access to the archive of nearly 3 million images taken by ASTER instrument; previously this data had only been accessible with a nominal fee.

ASTER, Advanced Spaceborne Thermal Emission and Reflection Radiometer, is a joint Japan-US instrument aboard NASA’s Terra satellite with the data used to create detailed maps of land surface temperature, reflectance, and elevation. When NASA made the Landsat archive freely available in 2008, an explosion in usage occurred. Will the same happen to ASTER?

As a remote sensing advocate I want many more people to be using satellite data, and I support any initiative that contributes to this goal. Public satellite data archives such as Landsat, are often referred to as ‘free data’. This phrase is unhelpful, and I prefer the term ‘free to access’. This is because ‘free data’ isn’t free, as someone has already paid to get the satellites into orbit, download the data from the instruments and then provide the websites for making this data available. So, who has paid for it? To be honest, it’s you and me!

To be accurate, these missions are generally funded by the tax payers of the country who put the satellite up. For example:

  • ASTER was funded by the American and Japanese public
  • Landsat is funded by the American public
  • The Sentinel satellites, under the Copernicus missions, are funded by the European public.

In addition to making basic data available, missions often also create a series of products derived from the raw data. This is achieved either by commercial companies being paid grants to create these products, which can then be offered as free to access datasets, or alternatively the companies develop the products themselves and then charge users to access to them.

‘Free data’ also creates user expectations, which may be unrealistic. Whenever a potential client comes to us, there is always a discussion on which data source to use. Pixalytics is a data independent company, and we suggest the best data to suit the client’s needs. However, this isn’t always the free to access datasets! There are a number of physical and operating criteria that need to be considered:

  • Spectral wavebands / frequency bands – wavelengths for optical instruments and frequencies for radar instruments, which determine what can be detected.
  • Spatial resolution: the size of the smallest objects that can be ‘seen’.
  • Revisit times: how often are you likely to get a new image – important if you’re interested in several acquisitions that are close together.
  • Long term archives of data: very useful if you want to look back in time.
  • Availability, for example, delivery schedule and ordering requirement.

We don’t want any client to pay for something they don’t need, but sometimes commercial data is the best solution. As the cost of this data can range from a few hundred to thousand pounds, this can be a challenging conversation with all the promotion of ‘free data’.

So, what’s the summary here?

If you’re analysing large amounts of data, e.g. for a time-series or large geographical areas, then free to access public data is a good choice as buying hundreds of images would often get very expensive and the higher spatial resolution isn’t always needed. However, if you want a specific acquisition over a specific location at high spatial resolution then the commercial missions come into their own.

Just remember, no satellite data is truly free!