On an increasingly hungry planet, it is important to know where our food is grown, and where the additional food we will need is going to be grown. This knowledge depends on good maps of farmland, yet the ones we rely on are often inaccurate, particularly in Africa, where farmland will grow the most this century.
Our project aims to improve on existing maps of African farmland. It starts from the idea that humans are better at interpreting images than algorithms, and is enabled by 1) the recent availability of publicly accessible, comprehensive, high resolution Earth imagery, and 2) a global internet-enabled workforce. Our methods connect Google Earth imagery to amazon.com’s Mechanical Turk platform via a PostGIS database, using python’s boto package and OpenLayers to provide a mapping interface. We use R and its packages rgdal and rgeos to assign mapping locations and assess worker and overall mapping accuracy, with pprepair cleaning map data to ensure spatial integrity.
We will present a detailed overview of the methods behind our crowd-sourced mapping project and our initial results. Ultimately, this project will be integrated with image pattern-recognition methods, feeding high quality training and validation data to a system that continually updates and refines crop field maps. Our goal is to create the “gold standard” for land cover maps.
Lyndon’s primary interest lies in understanding how global change (climate change and food demand) shapes agricultural land uses, and how these in turn impact terrestrial ecosystems and species. He uses remote sensing and spatially-explicit simulation models as his primary tools of investigation. Further details on his research are here.
For exhibition and sponsorship opportunities, contact Susan Stewart at email@example.com
For information on trade opportunities with O'Reilly conferences email mediapartners
For media-related inquiries, contact Maureen Jennings at firstname.lastname@example.org
View a complete list of Strata + Hadoop World 2013 contacts