humanitarian.info

More map madness in Nature magazine, whose 16 February issue has a commentary piece on Mapping Disaster Zones, covering work done by the Global Connection Project in response to Hurricanes Katrina and Rita and the Pakistan earthquake. The entire article is worth reading, but I’d like to focus on some of the lessons they learned:

We learned many other lessons from Pakistan, not least the need to develop tools that can adapt to local conditions on the ground. Internet connections in Pakistan were often slow and patchy, making downloads difficult. Post-disaster feedback highlighted difficulties with printing out maps and locating specific settlements among tens of thousands.

That part just brings back too many bad memories, especially the time that all the ink cartridges in our map plotter cracked in transit. It made a very big mess and we had to clean it by hand. Everybody was very unhappy.

Another obstacle was the mismatch between the local situation and the way existing geographic data is organized and represented. For example, Pakistani villages do not have a single position, but exist along a right-of-way extending from the valley to a ridge top, migrating up and down based on the changing seasons. Thus assigning a fixed position to a village is wrong most of the time.

This is an interesting problem, and one that we come up against frequently in the field. (Jishnu Das has written some more about this, using Flickr to tag his photos from his field work, based on his work with RisePak and LUMS in the Pakistan response.) Not only are settlement positions changeable in some countries but they can move or disappear, whether due to earthquake or war. So what’s the solution? The article suggests that

All mapping efforts, whether community based or UN supported, would benefit from a stronger feedback mechanism, such as a community filtering system that enables authenticated individuals to update shared geographical data online.

That would be nice, but I don’t think it will ever work. Ground truthing is absolutely indispensable for any remote sensing, and that means feet on the ground. People actually working on relief don’t have time to log on to update data – even if they have access to the internet – they just wave the map in your face, shrieking “Your map is wrong!” Which means grabbing the map, making them mark the wrongness clearly, and updating the files accordingly.

As a result, a ‘community filtering system’ isn’t likely to work; you need an authoritative reference point, preferably in the field, a focal point for maintaining and disseminating spatial data. In the absence of government capacity to do this, a Humanitarian Information Centre usually does this, with UNJLC or UNMAC handling logistics and mines respectively; in the absence of any of those, we’re a bit stuffed, so any offers would be welcome.

The full article is at http://www.nature.com/nature/journal/v439/n7078/full/439787a.html.

[Update: More from Jishnu on Kathryn Cramer’s blog at http://www.kathryncramer.com/kathryn_cramer/2006/01/a_long_interest.html; Kathryn is also one of the authors of the Nature article, and her blog is worth checking out as well.