FOSS4G - North America 2012: Observations

Saturday, April 14, 2012

I had the good fortune to make it to this year's iteration of the Free and Open Source for Geospatial (FOSS4G) tech conference - North American edition (other sections of it will be occurring in Germany and China). Having missed the O'Reilly festival of place, it was a great opportunity to catch up with - as one attendee put it - "The real makers in the world of maps". [Sorry, ESRI, you've actually been standing in the way of cartographic explanation for about a decade now]

While there was more afoot than I could ever have followed entirely, I took note of two items:

1.) People are really jazzed about PostGIS. Core architect Paul Ramsey kicked off what would be three days of love for the new release (version 2.0), and we were treated to all sorts of examples on how you can use it. Raster support, TileMill optimization, simple SQL strings replacing reams of jagged python for geoprocessing workflows - the possibilities seem endless.

I really like that such a great, extensible project is at the heart of open-source geographic technologies, but I still feel left out of the party. Basically, I am intimidated by the first hurdle of PostGIS: PostgreSQL at the command line. This is not that big of a deal for true developers, but I've been repeatedly burned by the wonkiness of database administration, permissions and security in postgres. Once I can get to the actual SQL, the birds start singing, the sun comes out and the world is full of possibilities, which is what makes client-focused tools like CartoDB and QGIS so enjoyable.

2.) Web maps are the home of cartographic design now. This may have been the purview of National Geographic even five years ago, but static maps are lacking the possibilities laid out - for example - by AJ Ashton and Nathaniel Von Kelso in sessions that touched on using OpenStreetMap as both art and function.

MapBox "Lacquer" Tiles

Stamen "Watercolor" Tiles
Take this with a word of warning issued by Schuyler Earle during a panel session: Software developers think they are natural designers. They're not. They need to hire designers, who can - like sober friends at 2AM - shepherd us away from bad decisions. Users will benefit if we can all get a design perspective on "I think this thing needs another button on the sidebar" before we act.

Many thanks to the organizers of FOSS4G-NA. It was a rockin' good time, full of productivity and great encounters. And for those who need a lingering shot of natural-disasterage, here is my ignite talk from the first night. Heavy on pictures, low on explanation, much the way it should be:

Read more ...

OpenstreetMap and the Bulk Battle

Wednesday, April 4, 2012

I like building footprints. As a cartographer, I find that they bring important context to a certain scale of map. In the same way that I look for patterns and paths in the contour lines of a national park, I see explanations of how we live and work in the layout of buildings in a city. Basemaps for the web applications I write always benefit from having structures as a ground reference.

Over the past few weeks I've been working to add about half a million building footprints to OpenStreetMap, mostly in the Mid-Atlantic region of the United States. The excellent data is the result of years of work - ongoing in many sites - by the Spatial Analysis Lab at the University of Vermont. They've fused LIDAR data with multispectral aerial imagery in an object-based classification of land use for their client municipalities, and they're happy to offer the results to the OSM open license.

Here's the current OSM Mapnik render in Maryland Park, just outside of D.C. -

And here's the same area with UVM-SAL building footprints added -

I've been careful to discard footprints that overlap with an existing OSM line or polygon feature, since user-traced content takes priority over bulk imported content. But having taken this data to the OSM community for review, I am now aware that it's a complicated thing to "Add data" in bulk to a crowdsourced database. Users have eloquently argued both for and against the import of these building footprints, and I find it boils down to a pair of inextricably-linked perspectives on OSM:

1.) OpenStreetMap is a community of individuals. The interaction between the user and the map is most valuable when one user "owns" their offering, and a dozen building-footprint-tracing contributors from Baltimore will feel greater ownership of their neighborhood than one guy in Vermont who uploads the whole county in one batch.

2.) OpenStreetMap is a tool. It is becoming a basemap of record in the GeoSpace precisely because that's what we wanted to do with it. OSM competes admirably with proprietary datasets, and we can use it for beautiful cartography and complex analysis. Any scale of addition is valid, as long as it is offered on the same open license.

The question that lies between these two perspectives: Can we have a quality basemap while maintaining a strong commitment to individual user engagement?

I think the answer is yes. I'm going to proceed with my buildings import on an assumption that was echoed by Kate Chapman: If a user in Bethesda finds that the buildings in her neighborhood are "already there", maybe instead of losing a sense of ownership she'll take it to the next level by adding identities to those structures.

Because I'm just adding the outlines. I can't tell you which one of them is a Bodega. There are many layers of value yet to be added.
Read more ...

WhateverConf 2012 - Notes from a Nonparticipant

Tuesday, April 3, 2012

So this O'Reilly chap evidently has some soiree planned every year this time, somewhere on the West coast. In the past, I've been able to blithely ignore the proceedings and hope that someone else would eventually point me toward anything useful that arises. Basically the same attitude I have toward SXSW.

Regrettably, this year I'm being forced to perk up from my West-coast-of-New-England-afternoon-coffee stupor and pay the hell attention. Three items caught my eye on Twitter today and yesterday:

1.) Custom cartography for Mapbox Streets - The Development Seed team clearly proceeds from day to day in a Red-Bull-fueled tornado of productivity. They've caught on that the momentum is away from the ubiquitous Google Maps API and base layers, and they're doing everything they can to offer a replacement for the whole stack. In particular, they've done Google one better here on cartographic quality. MapBox's OpenstreetMap-based tiles are uniquely beautiful, and sufficiently-varied that we users might now be able to escape the homogenization of web map styles over the past few years (Yes, Microsoft, even your basemaps look a lot like Google's). Cartographic design just won big. Here's some Tilemill-styled twitter data of mine over the new Mapbox "Lacquer" Base tiles:

2.) Foursquare has a new geocoding engine - Saints be praised, another way to escape Google. I don't yet know how solid Foursquare's placename database - called TwoFishes- will be for everyday applications. However, my current reality has me tethering whole projects to the GMaps API specifically because the Google geocoder is the only one to consistently find the correct location when a user searches for an address. I need alternatives, and they are particularly welcome from a company that is already showing a dedication to open geodata.

3.) CartoDB murders the middle of the GeoStack - The team at Vizzuality has been intimidatingly-busy as well. Though it may have been overlooked in the WhereConf hoopla, yesterday they teased out a project called VECNIK; at the moment it consists of a series of HTML5-ready libraries for fast vector styling directly in the browser. This includes pulling the parsed vector from the already-smooth CartoDB SQL API, compiling a stylesheet in the super-flexible Carto language, rendering the whole thing with Mapnik and serving it into the ModestMaps library - entirely in the browser. This happens at speeds that ten months ago would have been achievable only with cached tilesets. As with basically everything Vizzuality does, it's going to take me a few more months to grasp the full implications of this advance; I'm psyched enough for the moment though. Here's an example showing "Hand-sketched" parcel boundaries in my neighborhood (Modern browser required):

So to anyone else who's planning to drop major innovation bombs on the GeoSpace at WhereConf, $#@%in' stop. I've had enough for this week. Don't even get me started on PostGIS 2.0. Save it for FOSS4G-NA in D.C. next week, by which time I hope to have recovered enough spare neurons to take it all in. See you there.

Read more ...