Azavea has been doing interesting municipal work for more than a decade, with visualization and modeling software solutions that drive better decision-making processes. In addition to the company’s innovative work on applications ranging from urban forestry to crime analysis, they develop data analytics, and do capacity building with their Summer of Maps program, hackathons, and an EcoCamp. Sensors & Systems (S&S) editor Matt Ball spoke recently with Robert Cheetham, founder, president and CEO of Azavea about the company’s platform, applications development work, and the drive to make an impact through the power of geospatial technology.
S&S: One of your more high profile projects has been the OpenTreeMap software and services. How did you scale that up so quickly beyond just one municipality?
Cheetham: OpenTreeMap has been supported by a grant from the US Dept. of Agriculture’s Small Business Innovation Research (SBIR) program. Just as we won the first award, a group out in San Francisco released a project called Urban Forest Map that they planned to open source. So rather than build a competing system, we decided to join forces and the result is OpenTreeMap. Originally, we worked with the source code from San Francisco and made it more generic platform that could be applied to any location. Over the past year we’ve rewritten the whole platform from scratch. The re-write is aimed at a lower deployment cost, new features, and being able to build a subscription-based business model on top of it.
S&S: Is the OpenTreeMap taking hold in a lot of municipalities?
Cheetham: We’re getting interest from all over the world. We have people using it in the U.K., in Edmonton, Canada, and across several cities in the United States including San Diego, Sacramento, San Francisco, Grand Rapids, Los Angeles, and Philadelphia. And there are also cities that have implemented the software on their own. It’s open source software so we don’t charge any license fee for it, and anyone can set it up themselves. Seattle and Asheville, North Carolina are the cities we know about that have done this so far.
S&S: Was it part of the initial grant to require open source or was this part of your own plan after connecting with the San Francisco group that had that goal in mind?
Cheetham: The Small Business Innovation Research grants require you to turn the effort into a commercial product, so we had to make a case to USDA that open source software represented a viable business model.
However, the folks in San Francisco from whom we inherited the source code did have a grant requirement. They had a grant from CalFire to fund their original product and that was required to be open source.
We didn’t need to abide by that in our second version. However, we had found that the fact it was open source caused some our new customers to be more trusting that the project would be sustainable. In addition, some have been more willing to make investments in improving the software as the benefits of those improvements will be made available to all of the other cities as well.
That’s played out very well. We’re excited about the community that we’re building out around OpenTreeMap, and the decision we made to open source it.
Sharing our work is very much part of our company DNA. We’re a B Corporation (http://www.bcorporation.net/), so we operate with a social mission. Our mission is to apply geospatial technology for civic and social impact and to advance the state of the art through research. We evaluate a lot of what we do for the potential to share it and enable others to take advantage of it.
It’s not just with OpenTreeMap; we have additional open source projects for redistricting with our DistrictBuilder software, high performance geoprocessing with our GeoTrellis framework, and several other smaller efforts as well.
S&S: With this civic outreach, and a mandate to make a difference, do you find a lot of partners through that shared outlook and mission?
Cheetham: It depends on the project, but our most successful projects are ones where we have great partners. For OpenTreeMap, we worked with a small company called Urban Ecos in San Francisco. We also have partnerships with the University of Vermont, the Pennsylvania Horticulture Society, the U.S. Forest Service Research Station, and Public Laboratories TreeKIT, etc. We’re great at software engineering and user interface design, but we’re not necessarily good at building tree stewardship programs or developing tree canopy data sets. And we’re not urban ecologists, but our software gets better when we have input from those people.
At the University of Vermont they have a team that has years of experience analyzing tree canopy cover. The TreeKIT team has a ton of experience building community around a tree inventory and stewardship activities in New York City. Urban Ecos are expert urban ecologists. Each of these teams bring a lot to the table, and we bring the software engineering, user interface design and cloud computing expertise.
For our DistrictBuilder project, we have similar relationships, including with a couple of academic partners that were the principal investigators: Michael MacDonald at George Mason University and Micah Altman at MIT. They have also brought in other partners that know a lot about redistricting and this has made the product much better over time.
S&S: Are a good number of your projects tied to the adaptation and resilience to climate change? It seems we’re hearing more of those terms in an urban context, and I wonder if that mandate is driving this kind of work.
Cheetham: There are a couple ways that we’re engaged in urban sustainability efforts.
With OpenTreeMap, our next set of features are tools for planning and prioritizing where new trees get planted. We’re pushing well beyond inventory and mapping the trees toward understanding the ecosystem services values of each tree, and enabling cities and neighborhoods to understand the impact of the urban forest in terms of stormwater infiltration, air quality, and energy savings
Other climate resilience work includes the Army Corps of Engineers They are very much grappling with climate change and the velocity of that change for different locations across the United States. The Army Corps is concerned about where they should be investing in water infrastructure — dams, levees, etc. — given what we understand about changes in weather patterns and other drivers.
We also have a project with The Nature Conservancy called Coastal Resilience. This is also an application that’s been released under an open source license. We’ve created a general framework that can support a network of coastal locations, each with different functionality based on the data available in each place. Climate change, sea level rise, and coastal resilience are the central motivators behind this project.
This past spring we won another Small Business Innovation Research grant from the U.S. Department of Energy to develop prototype tools to do longer-term climate change planning at the local municipal and county levels. One of our goals is to bring some of the global climate change models from IPCC and others that give a sense of how things are changing at the global level and down-sample them to make them relevant at the local level. We’ll be combining the global climate models with other data to enable local planners to make meaningful decisions on planning for climate resilience.
S&S: You’re doing a fascinating combination of modeling at different scales.
Cheetham: We’re certainly not running the global climate change models; those are run on supercomputers by climate change modeling teams, but we are running national and even global-scale geospatial models for storm water runoff, watersheds, ecosystems, transit, and that sort of thing. The reason that we’re able to do that is the GeoTrellis framework that I mentioned earlier. It’s a software framework that enables us to work with very large datasets by distributing that work across many machines, whether in an Amazon cloud environment, or in-house on a cluster of machines. By breaking down the work and distributing it, you get much faster response times, either for real-time interactive modeling or for long-run batch processing.
S&S: So, you’re pioneering some big data modeling approaches. Would you say that geospatial is the original big data?
Cheetham: Geospatial data has been big data for a very long time. I’m not sure we’re a big data pioneer, but we’re grappling with the issues inherent in working with very large space-time datasets, and trying to come up with computing strategies that enable us to do that kind of work in a useful way. We’ve also begun applying machine learning techniques to space-time forecasting, and in particular as it relates to crime analysis and forecasting.
We have 12 years of experience, almost since the company started, with crime mapping and crime analysis. We were fortunate to receive some early funding from the Office of the U.S. Attorney as well as well as the National Science Foundation to build software for detecting changes in geographic patterns in crime, essentially an early warning system. More recently, we are able to forecast crime risk based on a number of different inputs. Our forecasting models combine weather, special events, business locations, schools and other factors. The resulting forecast enables a police captain in a precinct to make informed decisions about where to place their officers for each shift.
These are very large databases that combine complex statistics. The only way to do this fast enough to support shift planning is to break that up across many machines. When we do a forecast for Philadelphia or Lincoln, Nebraska, we have to spin up 30 to 50 machines to run all the models, and spin them back down. We could have done that computationally several years ago, but we couldn’t have done it in a cost-effective way that would be affordable without the computing infrastructure that we have today. Platforms like Google Compute Engine and Amazon Web Services give us an ability to elastically scale up very rapidly and then scale back down again so that we’re only incurring costs for computing time when we need to.
That would have been utterly unaffordable five years ago, and no police department has the budget to go buy 50 or 60 servers that will sit idle for most of the time. These new kinds of infrastructure enable us to do new things, and we’re trying to adapt our software tools to take advantage of that.
S&S: The spatio-temporal modeling has been a somewhat problematic area in the past due to the computing resources. Are you applying new algorithms and uncovering new insights thanks to the computing capacity?
Cheetham: We are certainly gaining new insights with access to this type of scalable cloud computing infrastructure. But, often, our model development process begins with desktop tools like R and ArcGIS and smaller data sets. Once we’ve developed a particular approach, we will re-architect it in order to achieve the kind of scaling we need to run models for an entire city or region or to support many usersThat’s when the new infrastructure and architectures come into play.
S&S: I could ask questions all day about spatio-temporal modeling as it’s an area that really fascinates me, but we should also touch on your history of mentoring young people. You have run the Summer of Maps for some time, and this year you ran an EcoCamp. Tell me a little about those efforts.
Cheetham: Summer of Maps is in its third year. It’s been an incredibly successful program. It came out of some work we had done locally around open data and running local competitions, trying to engage local communities to help prioritize which datasets are released by local government. Summer of Maps arose partially from a realization that there’s a tremendous amount of value in GIS technology that is largely untapped by nonprofit organizations. Some of that might be because they don’t have the skills or resources, but also because they are not thinking about what kinds of questions they could be asking with geospatial technology. The second motivation was that we saw a lot of talented and smart people coming out of school that were taking entry-level jobs doing digitizing and other tedious work, and we wanted to provide a a really strong first professional experience for students. Summer of Maps combines those two concerns.
The program starts in late December each year. We open up applications for nonprofit organizations, and they submit a proposal that includes an analytical question, the data they have, and the decision they are trying to support.. An important prerequisite is that they need to come to the table with the data the project requires; we’re not going to spend the whole summer tracking down data for them. We close this initial process in late January, review the proposals, and then curate them down to about a dozen that represent the highest quality, greatest impact, and most interesting and challenging questions. We then turn these data analysis proposal around to students, and open up a student application process. The programis open to any student in the United States. They submit their qualifications and pick the projects in which they are most interested. It’s essentially a match-making process between the best students and the best projects.
In April we announce the fellowships. Each student receives a $5,000 stipend for the summer. They are paired with a mentor from our staff and work here in our office on two of the projects. The students work with the non-profits throughout the process and deliver a set of maps, data, and analysis results to each organization.
We’ve had a great deal of interest from organizations and students. This year we had 125 students that applied for only three positions. We think we’ve created a successful program, but it’s still operating at a relatively small scale. Now our challenge is to figure out how to scale it while still providing the high quality mentoring and in-office professional experience.
After our initial year, we had some inquiries about sponsorship. This year the program is being sponsored by Esri, Google, and the University of Pennsylvania’s Masters in Urban Spatial Analytics. These sponsorships help to underwrite some of the costs of the program.
We’d like to figure out how to grow it and have more impact. It has been incredibly successful for the nonprofits, some of whom have gone after funding. Last year, the Delaware Valley Association for the Education of Young Children (DVAEYC) used the maps and analytics from the Summer of Maps program to go after grants, and they had a million dollars of new funding by the end of the year.
These kinds of stories have been inspiring, and we think the Summer of Maps program has the potential for a great deal of impact for both the non-profit organizations and the students.
EcoCamp is a separate effort. Azavea has hosted hackathons, unconference, workshops, and similar events. We began a couple years ago with Hacks for Democracy just before the 2012 election, and that events brought together people interested in open data, elections and politics. In April 2013, we organized the NASA International Space Apps Challenge, where Philadelphia was the main stage location for a global event that included more than 75 cities. EcoCamp was the third such event.
We’re constantly looking for ways to improve outcomes for this type of civic hacking events. One thing that we’ve observed over time is that the primary outcome is not the software or visualizations that result, but, rather, the teams and relationships that assemble to create something together. With EcoCamp, instead of just running a hackathon over a weekend, we wanted to see if we could enhance that particular community-building effect with some new ideas.
EcoCamp took place June 20-22. On Friday, we ran a series of workshops with two tracks, one with practitioners talking to technologists about urban ecosystems, environment, and sustainability. The second was for technologists talking to practitioners about ways to apply technology to their concerns. On Saturday, in parallel with the hackathon, we hosted an unconference, a conference that that is not structured in advance; the agenda is developed by those that show up. Anyone is invited, and anyone can propose a talk.
By combining workshops, learning, presentation, discussions and coding, we hoped to enhance the long-term benefits of a hackathon and take the benefits beyond the software that might get built over the course of a weekend.
S&S: I really like your sustainability focus. Have you’ve been thinking holistically about the city for some time?
Cheetham: Our work is not just around traditional sustainability in a green sense. We work across domains of land, water and people. We believe geospatial technology has real power to help operate cities in a more effective way across a number of different domains. Urban forestry, watersheds, and energy are one part of that. But land records, homelessness, crime, zoning, and elections are also part of that. Almost everything that municipalities do has some kind of geospatial component, and we’re really focused on building tools that enable cities to leverage those capabilities in a constructive way.