Tracking economic activities using AI and satellites

Project 1: Roads to development? Examining the Zambian context using AI

To evaluate how recent road development in Zambia affects its economy, I trained an artificial intelligence (AI) to predict road surface conditions for the core road network and built-up area in the Zambia for the last decade.

The figure shows the AI-predicted road quality in the core road network. In the top two panels, roads shown in brighter colors are predicted to have a higher probability of asphalt-paved surfaces. In the bottom panel, roads shown in brighter colors have greater changes in road surface conditions.

This figure shows the AI-predicted built-up areas for Ndola, the second largest city in Zambia. Red indicates built-up areas developed in 2009, blue indicates new built-up areas developed in 2009 and 2014, and green indicates new built-up areas developed between 2014 and 2019. The map in the upper left corner shows the ground truth of built-up areas in 2009 obtained from Google Earth, and the map in the bottom left shows the same for 2019. Comparing the area highlighted in the yellow circle, the model prediction captured in unexpected granularity the built-up areas based on the growth in economic activities on the edge of the city.

Project 2: China’s Journey to the West: Using High-Resolution Satellite Imagery to Evaluate Transportation Improvement in Central Asia

We study how the improvement of highway networks between China and neighboring countries in Central Asia has affected economic development in the region. By applying deep learning to remotely sensed data with a 5m spatial resolution, we create digital maps of enlarging road networks and expanding built-up footprints between 2009 and 2019 for a region that covers 1.3 million sq. km.

The figure shows AI-predicted changes in the primary road network that intersects with the 300 km buffer. Roads in red color did not experience upgrade in surface conditions. Roads in green experience upgrade in surface conditions. There are three types of road surface conditions: asphalt paved roads, dirt roads and no roads. Prediction for surface condition was made for year 2009, 2014 and 2019 separately before identifying upgrades.

This figure shows the predicted AI-predicted built-up for a scene that covers Khorgas. A scene is a basic image we obtain from the Planet Lab, which is a 25 km by 25km square. Area in red color indicates built-up in 2009, blue indicates new built-up developed between 2009 and 2014, and green indicates built-up developed between 2014 and 2019. One can observe growth in built-up due to the growth in economic activities due to building the dry port of Khorgas.