Over the past year Azavea has been working with the Inter-American Development Bank (IDB) to provide technical support to Guyana’s Central Housing and Planning Authority (CHPA). IDB is one of the primary sources for development financing that supports Latin American and the Caribbean. The Bank focuses on an array of issues, including improving productivity and supporting climate friendly development. Our work with CHPA reflects this as its primary goal is to increase the institutional data and GIS capacity of CHPA. This would support their work managing the planned growth of the urban areas of Guyana and regularizing informal settlements.

Our technical support for CHPA covers tasks related to data workflows and processes, organization, and collection and creation. Open source tools and open data are a unifying thread that runs through each step of the project. In particular, our work with IDB relies on OpenStreetMap (OSM), a free and open map of the world. This focus is a goal of IDB, but also something that we at Azavea strongly support. We have a long history of work in the open source and OpenStreetMap realm, including projects that have an international focus such as DRIVER with the WorldBank to track and reduce traffic crashes, calculating the completeness of OSM building data for health analytics with DevSeed, and measuring how many people live within 2km of an all-season road in support of Sustainable Development Goal (SDG) 9.1.1. So what is the appeal of OpenStreetMap and open source in the context of international work?

Why OpenStreetMap?

OpenStreetMap is a map of the world that can be edited and used by anyone, provided it adheres with the OSM data license. OSM data can be downloaded and used by analysts or designers and is a primary data source for many of the basemaps used in today’s webmapping applications. There is a large community built around OSM data that hosts mapathons all around the world. Just check out the huge number of mapathons that were organized for the OSM Geography Awareness Week in 2018. OSM data is structured with tags and datatypes. New tags can be created if they are needed for the specific data being added to the map. This flexibility allows organizations or people to add data that is extremely specific to their geographic region.

These features make OSM an excellent resource for countries or organizations that need to supplement their internal GIS data. They can organize a local community of mappers and host a mapathon to do “armchair” mapping by tracing satellite imagery or do mapping in the field using one of the mobile editing tools available through the open source community, such as OSMAnd or OpenDataKit. Once enough data is available in OSM, an organization can download these data and use them for their GIS work. Many times, data available from OSM is extremely accurate and can represent the best resource for national level datasets, even in the United States.

How we are using OSM with IDB

IDB and CHPA wanted to supplement CHPA internal GIS data with OSM data. There are many ways to query and download OSM data but they often require a detailed knowledge of the OSM XML tagging structure. Azavea built an OSM data extraction tool on top of the Overpass API that takes the guesswork out of what data to download and allows users to preview the data that was downloaded prior to opening it in a mapping software.

The tool provides a link directly to the map extent in OSM if a user notices any incorrect data and wants to edit it right away. It also converts the files into GeoJSON format rather than the standard .osm or .pbf OSM files. GeoJSONs are a more standard format for doing GIS work. The data available to download is based on a set of predefined tags and values that was supplied by IDB.

Azavea is also conducting a machine learning pilot with IDB to identify buildings in South American cities using high resolution satellite imagery. Machine learning requires a training dataset to train the model. In, in this case we needed building footprints. We turned to OSM to acquire these building footprints and used them as labels in our open source Raster Vision workflow. Raster Vision is a framework for conducting deep learning models on satellite imagery. The model predicts a mask of buildings for Georgetown, Guyana, Paramaribo, Suriname, and Belize City, Belize. This work is ongoing so stay tuned for more information in the coming months.

Our work with IDB and CHPA is still underway but will include a few more items related to OSM and open source tools. We’ll be outlining a workflow for conflating the data accessed through the OSM extraction tool with CHPA’s internal GIS data. To support a more complete OSM map and easier and better collection of data in the field, we researched and set up a test implementation of OpenDataKit with OpenMapKit. We will demo these tools to CHPA and provide training for how to set up and implement a survey.

---

Do you have ideas or needs around using OSM to support GIS work? Reach out so we can partner together.