Illuminating Points of Interest From Space

Chris Shughrue - October 10, 2019

At StreetCred, we’re on a mission to get all the place data in the world. This presents a major challenge: starting from a blank map of the world, how do you predict where points of interest exist?

A new satellite is helping us answer that question.

Why it’s hard to measure human activity

Monitoring and modeling human activity at global scales is a fundamental scientific challenge. Increasingly, scientists are directing their attention to socioeconomic change and trends in cities, which now account for more than half of the population and 80% of global GDP.

Cities can be seen dually as both physical places we occupy (buildings, streets, and infrastructure) and social spaces where social and economic relationships connect people. When you think about the important places in your city, it’s natural to consider your daily lived experiences -- the cafe where you grab a latte on the way to work, the public park where you jog every afternoon, or the theater where you saw a life-changing performance of To Kill a Mockingbird. These experiential aspects of cities are what make places important, but also resist measurement.

Rather than focusing on complex social features that can’t be readily measured and compared, scientists (and StreetCred) have instead turned to biophysical indicators in global studies. In particular, we leverage a multitude of satellite data to monitor and model human activity over time.

What can we see from space?

Satellites are constrained by physics and processing power, so they’re generally most useful for observing human activity indirectly. Rather than measuring the movement of individual people, we look at larger-scale changes they make to the biophysical landscape, such as deforestation, changes in agricultural practices, urban growth, and depletion of water resources. In these cases, scientists might measure certain materials on the ground, such as vegetation, water, or concrete, and then classify land use based on their observations.

Figure 1. Examples of two approaches for measuring human activity with satellite data centered on Chicago, USA. One approach classifies urban land cover over time to quantify where and how much a city has changed (left). Map shows water (blue), non-urban areas (black), and urban areas from oldest to youngest (red to white). Land cover classification is the output of a model using Landsat data (data from Djikstra, Lewis, and H. Poelmann, 2014). Another approach uses near-real-time nighttime light illumination data (Román et al, 2018) to identify electrification infrastructure and energy use.

How can we directly observe human activity? Wait for it to get dark. At night, illumination from street lamps, office buildings, and residences provides a meaningful indicator of human activity. Economists have used this data to make detailed sub-national GDP models (Chen and Nordhaus, 2011), and demographers have created fine-scale population estimates (Sutton et al, 1997). Established strong linkages between human socioeconomic activity and the nighttime light data makes this data an ideal starting point for our points of interest estimates.

Until recently, studies have relied on data from a meteorological satellite (Defense Meteorological Program (DMSP) Operational Line Scan (OLS). This sensor was not originally designed to monitor human activity. As a result, it has several shortcomings for analyzing human activity: poor spatial resolution, difficulty differentiating intensity in very bright areas such as city centers, and challenges comparing images over time.

A new era of night lights

Figure 2. Nighttime lights data from VIIRS. Detailed zoom of Delhi, India shown. Data: NASA Black Marble (Román et al, 2018).

A new satellite aims to solve these challenges. In 2011, NASA launched the Visible Infrared Imaging Radiometer Suite (VIIRS) sensor as part of the Suomi-NPP satellite mission. This new sensor addresses the shortcomings of previous nighttime light data: VIIRS offers finer spatial resolution, a greater range of sensitivity, and has the potential to offer more frequent measurements.

Lighting the path forward

Recent emerging studies have already shed light on the tremendous possibilities opened up by this new satellite data. In the aftermath of Hurricane Maria in 2017, scientists were able to track loss of electricity in Puerto Rico, aiding humanitarian efforts and documenting recovery over time (Román & Stokes et al, 2019). Improved frequency and resolution have also made it possible to observe changes in nighttime lighting during the celebration of holidays such as Christmas and Ramadan, providing new insight into how energy use is tied to culture (Román & Stokes, 2015). This represents a sea change in the use of satellite data to directly measure human behavior.

Points of interest are an integral part of the complex socioeconomic landscapes we inhabit. Nighttime light data help us better understand this landscape, and ultimately identify areas most likely to contain interesting places.

This thread of study doesn’t just help us identify where and how to map all the places; it also opens a door to address fundamental questions about how people and economic services interact across the globe. For StreetCred, we’ll be able to see where our user community should direct its effort and understand more about the common characteristics of interesting places across diverse regions, from New York to Papua New Guinea. As our models and data advance, we’ll be able to measure economic complexity and services in novel ways, shining new light on the world we’re mapping.