This is the 4th in a four-part blog. You may want to skim through Parts 1, 2 and 3 before continuing ...
The technologies of high-resolution satellite imagery,
machine learning and large-scale data capture and storage are converging at a
rapid rate. Inevitably, the relevance of
manual addressing systems is diminishing.
Providing accurate geocoded addresses – either as precise points or
GIS-ready building footprint polygons to the half to three-quarters of humanity
who do not live around distinctive street patterns or have accurate addresses
is now a very real technical possibility and being made probable through
commercial applications. All of built
Australia – comprising 15,243,669 buildings located over the continental
expanse of 7.6 million km2, has been captured by PSMA Geoscape which
also plans to release updated datasets for at least 5,000km2 of
urban and remote community areas every year going forward. DigitalGlobe is working on a building
footprints dataset for the US and possibly internationally. Researchers from Facebook Connectivity Lab and
MIT Media Lab have developed an algorithm (RoboCode) that uses deep learning to
recognize roads from satellite imagery, and, after partitioning the road
network into neighborhoods, goes on to label regions, roads, and blocks using
proximity-based algorithms. The method
is more akin to our familiar street addressing systems than the geocoding
systems based on latitude and longitude coordinates although, arguably, the
unique addresses – such as “715D.NE127.Dhule.MhIn”, remain awkwardly unintuitive. Results on RoboCode published to date include
that human settlement maps covering 20 countries in Africa have been generated:
inside of two weeks and at a spatial resolution of 5 meters which improves
previous manually prepared countrywide data sets by multiple orders of
magnitude, although the results still need ground-truthing (verification) in
the field. Most recently, a partnership of Bing Maps, the community of Humanitarian
OpenStreetMap Team (HOT) and Microsoft Philanthropies has created building
footprint open datasets covering Uganda and Tanzania for OpenStreetMap users.
I personally believe that it is important that cloud-stored and possibly open-source available datasets of where and how I live are matched by equally robust data privacy and data protection principles, including, where appropriate, with my knowledge or consent of where and how my data is being used. Whilst exhilarating technologies now allow us to map the world and in time, geolocate every person on it and store all their details, there is risk that the parameters and limits on public use of this personal information are being left behind. The risk is compounded in that many of the agencies and staff involved in addressing and collecting sometimes deeply personal information for ‘development purposes’ do not always encrypt and adequately protect that information and openly distribute it by e-mails, memory sticks, etc. In part, this may be because managers and staff themselves are not particularly tech savvy. Awareness of the need for anonymizing shared data and developing systems and staff capabilities to do so is clearly an area needing greater attention and resources.
But for me, the most powerful lesson in addressing was one that I learnt many years ago in Epworth (even though using what today would be archaic technology!) In the face of today’s exciting, disruptive technologies we shouldn’t lose sight of the fact that sitting behind any three words/plus code/ geo-coded or labelled point or polygon on an e-map are fundamental data that identify and record who has the right to use, manage, transact with other parties, withdraw or transfer that physical space, and the rights and responsibilities of the wider community around that space. Property rights are not exogenous. In every country, these rights and responsibilities have emerged and continuously develop over time. Owner- or user-rights are a function of push-pull economic, social and political forces, and levels of societal and government enforcement that evolve in different ways in different countries and over generations of time. In a deeply esoteric way, addresses are a form of social compact between authorities and residents. As such, addresses are only of utility in poor neighborhoods if they certify to a wider recognition of what UN-Habitat has recently branded “the right to the city” by the addressee. A geo-coded or algorithm-generated spatial address is of limited use and, in some circumstances and in the wrong hands, could prove to be harmful to the urban poor (for example, by identifying them for eviction) if it is not subordinated to a ‘pink slip’. For this to happen, the massive power of evolving technologies that is being driven by suppliers might be more effective (and ward off concerns of the potential to do harm) if it were linked with and paced according to invitations by communities wanting those addresses. And the resources being invested in developing spatially accurate address labels need to be matched by equally strong efforts aimed at resolving and negotiating ‘pink slips’ - the right to live in that place, for all urban residents, including the poor. Ideally, addresses should be available to everyone – but perhaps we need to roll them out at - rather than three, only one word at a time.
 |
| Household head agreeing with community worker on the placement of her address, Tebikenikora, Kiribati (Credit: C. Butcher-Gollach, December 15, 2017) |
