Geo4Dev 2018 Recap: How Geospatial Data can Aid Humanitarian Response
This post was written by CEGA Affiliates Jen Burney and Marshall Burke. Jennifer Burney is an Associate Professor at UC San Diego’s School of Global Policy and Strategy and Marshall Burke is an Assistant Professor at Stanford’s Department of Earth Systems Science.
On November 15th, 2018, we were honored to co-chair (along with New Light Technologies’ Ran Goldblatt and Harvard’s Patrick Vinck) CEGA’s second annual Geospatial Analysis for International Development (Geo4Dev) conference at UC Berkeley. New Light Technologies generously sponsored the event.
This year’s symposium focused on geospatial research that addresses climate- and conflict-driven migration and humanitarian response from geospatial experts in academia, government, NGOs, and the private sector. This included observation and modeling of migration and human settlement patterns (in response to climate or conflict stressors), as well as the design and evaluation of interventions for humanitarian crises, mass migration, and community resilience.
Geo4Dev is a yearly event focused on the use of novel geospatial data and analytic techniques to address issues of poverty, sustainable development, urbanization, climate change, and economic growth in developing countries and beyond. This includes a particular emphasis on the use of emerging geo-tagged big data, including satellite, social media, and CDR datasets.
As expected, the workshop was an exciting view into a lot of new work going on at the intersection of data science and international development. Building on last year’s inaugural convening, the workshop highlighted how a range of public and private organizations are increasingly integrating geospatial tools into their decision-making.
The keynote address by FEMA’s Geospatial Information Officer, Chris Vaughan, and the morning’s first panel (featuring Madeline Jones of New Light Technologies, Miguel Román of NASA, and Kumar Navalur of DigitalGlobe) focused on real-world examples of integrating geospatial data into disaster planning and response. The speakers all emphasized the need to build infrastructure for getting usable information to decision-makers in near-real time. They stressed the difference between timescales for disaster response and timescales for preparedness and development applications with geospatial data, concluding that good-enough-and-on-time information is infinitely more valuable than better-but-later data for disaster response applications.
The panel highlighted three crucial points:
- The increasing dependence of government operations on data feeds generated by the private sector, including satellite imagery (which has long been the case) but also tools like Waze and GasBuddy in the US.
- The remarkable amount of “basic” information still lacking even in wealthy countries like the US. For instance, precise data on the location of fire departments, hospitals, and schools in the US, or basic data on where people live and how we get to them in most developing countries, is often incomplete.
- How academics can engage in building better tools and analytics, as well as in vetting the increasing number of tools that governments are releasing (the disasters.geoplatform.gov site was particularly highlighted in this regard).
Beyond these take-aways, a few additional themes emerged from the day’s discussions:
Many speakers pointed to new issues of scale, in contrast to last year’s conference in which participants called for more and higher resolution data. Several Geo4Dev contributors this year acknowledged that their satellite imagery analysis is now up against geolocation limits. The analyses presented indicated that, now that the entire globe is imaged daily by multiple sources, there is a need for fast, intelligent dimension reduction. This is even truer for disaster response applications.
Finally, issues of transparency and infrastructure resilience rose to the forefront of the conference. Several speakers pointed to a need to fully understand how and why algorithms work when using machine learning in decision-making. Beyond transparency in methods, there was also an uneasy acknowledgment that a tremendous amount of work has gone into gathering new data and developing new methods, but that much of our current infrastructure for analysis remains informal. One concrete example is the incorporation of data by FEMA for its disaster response dashboards — Maddie Jones recalled realizing only by accident that one data source FEMA was scraping had changed formats without warning or communication. This did not happen during a crisis, and was discovered early, but nevertheless highlighted the need to curate both processes as well as data, and to generate institutional frameworks that can protect these kinds of data partnerships. This will be even more important when public, private, and academic institutions work together.
At the same time as we find new ways to automate and “bake in” important data flows and processes, researchers at Geo4Dev also universally expressed an excitement about this moment in time, when so many are trying to integrate geospatial data and analysis into decision-making.
