Part 4 of an almost sequence - adding EO, GDPR and power to the mix
Back in the day I did an MSc in Applied Remote Sensing - yay - and for a while, working with Landsat and SPOT data to devive new insights and present information in then new ways, it seemed a potent tool, but then the music kind of died. GIS was the new kid on the block, funding of satellite borne sensors went soft, internet bandwidths (then) and imagery didn't work too well together. Today, we seem to have come full circle, and while 'remote sensing' sounds unfamiliar, the world of EO is seriously 'hot'. Among the major change agents has been the post 9/11 requirement to find out more about 'stuff' often a long way from 'home' and Moore's Law across the capacity of the related tools and technologies (storage, bandwidth, processing, compression etc).
Depending on how you classify 'flocks' of micro/nano satellites there are c.1000 live image collection sensors in orbit, providing daily coverage (or at least the potential for) the entire planet at spatial resolutions from 30cm upwards, ever increasing spectral ranges and with increasing radiometric resolution. We see it on the news, in the media so much that this isn't news. A picture may indeed be worth a thousand words; to turn that image and the visual cues and insights therefrom into something of value you extract and codify those 'words'. A series of "pictures" is worth a whole lot more and time-series is again a familiar visual media motif.
Normalised Difference Vegetation Index (NDVI, go on, look it up) was (and remains) one of the go-to algorithms that extracts, in this case, 'words' about growth and senescence in crops that the end user can deploy in studies of land cover, land use, yield, soils, planting, weather and more. In the dim and distant past I used this and other image processing techniques on multiple (and at the time costly (still dirt cheap compared to trying to acquire that information on the ground)) images from Landsat and SPOT to look at seasonal and annual studies in support of extension, planning and other interventions in Africa.
Now, with daily and in some cases free coverage from Copernicus, Planet, earth-i, Digital Globe to mention some significant players on the one hand, the leadership shown by Google (Cloud/Compute and all its EO data) and the explosion in the application of artificial intelligence techniques to imagery on the other, the study of shadows of oil storage containers in China as a proxy for economic health, the counting of containers in ports and the $bn valuation of Climate Corp when acquired by Monsanto in 2013 are just some of the headlines.
From there it is but a short step to some truly intrusive analytics. Using thermal imagery to target those with ineffective insulation? Good, right, leading to positive intervention, reduced bills, lower emissions. Using daily optical imagery to study domestic patterns? Whoa, like when cars are or aren't parked or have or haven't moved you mean? I do. Never mind the spy in the sky seeing what paper you're reading, who cares, but the postie always delivers at 11 right and no one is home again except on a Thursday (guess that's the cleaner right, small van) til 5:30 except in the school holidays (they have kids right). OK, OK, so easy to get carried away.
Much of the source of such insights is available as open data. We can all get to it, yay. Not so many of us have the expertise to set about extracting those insights and deliver value to end users (hence the Climate Corp deal). Intermediaries have always been a fundamental part of the value chain especially with the complexities of imagery though the advent of analysis ready data (ARD) may move some sector or application specific intermediaries further up the value chain. Intermediaries bring together EO data and the tools to analyse and extract information from that data with other data sources to meet end user needs. That additional data might also be open, it might be licensed, it might be the end users own data; brought together - mosaiced, triangulated, whatever - it is easy to see how property and people can be identified, associated and 'valued'.
As this article for geospatialists illustrates (and perhaps dramatises), the personal is everywhere in location data, oft-times directly but even more inevitably when you bring data sets together. The call here is to geospatial intermediaries to be mindful of and comply with their obligations under GDPR when so-doing - for effect I think the author talks of control, defusing, aggregating of a dangerous substance - data!
Hyperbole perhaps. Has the open data horse bolted, can or should the genie be put back in the bottle, should data releases be degraded, should even open data releases be tracked (per HMLR for example), how quickly does open data deteriorate in the wild, what are the incentives to data producers to continue to collect, validate and publish such data? Good and pertinent questions all, not for this post though.
But we need to be as careful in addressing these questions as we are in complying with GDPR's implications. There is a significant asymmetry in the datasphere in that much of the power to collect and assemble the data and to publish and create these insights is the opposite, residing with mega-corps and although Copernicus has it's own portal clearly not the resources of a Google to lavish on interfaces and user experience. Which gives rise to a tension. The privatisation of the digital public realm is building a parallel data realm to which 'accessible' data, open, licensed or otherwise, offers a counter-point, a choice, the absence or weakening of which would reinforce that asymmetry.
Convenience, ease-of-use, apps and much that makes the digital consumer world relies on this asymmetry. Are we questioning of or inured to the implications of what is coming down the track and do the capabilities 'out there' at-scale demand a response in discussions about openness?
EO is back, for sure, yet the ramifications are only beginning....
Depending on how you classify 'flocks' of micro/nano satellites there are c.1000 live image collection sensors in orbit, providing daily coverage (or at least the potential for) the entire planet at spatial resolutions from 30cm upwards, ever increasing spectral ranges and with increasing radiometric resolution. We see it on the news, in the media so much that this isn't news. A picture may indeed be worth a thousand words; to turn that image and the visual cues and insights therefrom into something of value you extract and codify those 'words'. A series of "pictures" is worth a whole lot more and time-series is again a familiar visual media motif.
Normalised Difference Vegetation Index (NDVI, go on, look it up) was (and remains) one of the go-to algorithms that extracts, in this case, 'words' about growth and senescence in crops that the end user can deploy in studies of land cover, land use, yield, soils, planting, weather and more. In the dim and distant past I used this and other image processing techniques on multiple (and at the time costly (still dirt cheap compared to trying to acquire that information on the ground)) images from Landsat and SPOT to look at seasonal and annual studies in support of extension, planning and other interventions in Africa.
Now, with daily and in some cases free coverage from Copernicus, Planet, earth-i, Digital Globe to mention some significant players on the one hand, the leadership shown by Google (Cloud/Compute and all its EO data) and the explosion in the application of artificial intelligence techniques to imagery on the other, the study of shadows of oil storage containers in China as a proxy for economic health, the counting of containers in ports and the $bn valuation of Climate Corp when acquired by Monsanto in 2013 are just some of the headlines.
From there it is but a short step to some truly intrusive analytics. Using thermal imagery to target those with ineffective insulation? Good, right, leading to positive intervention, reduced bills, lower emissions. Using daily optical imagery to study domestic patterns? Whoa, like when cars are or aren't parked or have or haven't moved you mean? I do. Never mind the spy in the sky seeing what paper you're reading, who cares, but the postie always delivers at 11 right and no one is home again except on a Thursday (guess that's the cleaner right, small van) til 5:30 except in the school holidays (they have kids right). OK, OK, so easy to get carried away.
Much of the source of such insights is available as open data. We can all get to it, yay. Not so many of us have the expertise to set about extracting those insights and deliver value to end users (hence the Climate Corp deal). Intermediaries have always been a fundamental part of the value chain especially with the complexities of imagery though the advent of analysis ready data (ARD) may move some sector or application specific intermediaries further up the value chain. Intermediaries bring together EO data and the tools to analyse and extract information from that data with other data sources to meet end user needs. That additional data might also be open, it might be licensed, it might be the end users own data; brought together - mosaiced, triangulated, whatever - it is easy to see how property and people can be identified, associated and 'valued'.
As this article for geospatialists illustrates (and perhaps dramatises), the personal is everywhere in location data, oft-times directly but even more inevitably when you bring data sets together. The call here is to geospatial intermediaries to be mindful of and comply with their obligations under GDPR when so-doing - for effect I think the author talks of control, defusing, aggregating of a dangerous substance - data!
Hyperbole perhaps. Has the open data horse bolted, can or should the genie be put back in the bottle, should data releases be degraded, should even open data releases be tracked (per HMLR for example), how quickly does open data deteriorate in the wild, what are the incentives to data producers to continue to collect, validate and publish such data? Good and pertinent questions all, not for this post though.
But we need to be as careful in addressing these questions as we are in complying with GDPR's implications. There is a significant asymmetry in the datasphere in that much of the power to collect and assemble the data and to publish and create these insights is the opposite, residing with mega-corps and although Copernicus has it's own portal clearly not the resources of a Google to lavish on interfaces and user experience. Which gives rise to a tension. The privatisation of the digital public realm is building a parallel data realm to which 'accessible' data, open, licensed or otherwise, offers a counter-point, a choice, the absence or weakening of which would reinforce that asymmetry.
Convenience, ease-of-use, apps and much that makes the digital consumer world relies on this asymmetry. Are we questioning of or inured to the implications of what is coming down the track and do the capabilities 'out there' at-scale demand a response in discussions about openness?
EO is back, for sure, yet the ramifications are only beginning....
Comments
Post a Comment
Thank you for taking the time to ponder my musings and for any contribution you make. Although comments appear immediately (i.e. unmoderated) I will remove (or if possible) edit offensive comments.