If you visit the National Museum of Natural History in Washington, D.C., one of the exhibits you’ll come across is a map of the Earth, which shows lights detected by satellites at night. With even a cursory look, it’s clear the lights pick out spatial patterns of urban and economic development. Look at the USA, and you see the coasts are brightly lit, whereas the country’s interior is much less so. Look at the Korea peninsula and you see that whilst South Korea is almost ablaze with light, the North is noteworthy for its almost complete absence of light.
The potential ability of night-time lights imagery to detect spatial patterns of urban and economic development has been known in the remote sensing community since the late 1970s. However, it has only recently been brought to the attention of economists following a paper by Vernon Henderson, Adam Storeygard and David Weil entitled “Measuring Economic Growth from Outer Space .” This paper alerted economists to the strong correlation between a country’s rate of GDP growth and the growth in intensity of its night-time lights, and the fact that the lights represent (for economists) a relatively untapped dataset with global coverage and a time-series dating back more than twenty years.
In the South Asia Urban and Water unit of the World Bank, we have been working with the night-time lights data for the past eleven months. In particular, as part of the region’s Urbanization Flagship, we have been using the data to examine patterns of urban growth across the region over the sample-period 1999-2010. Below are our takeaways on this unique and exciting data set:
Patterns of urban physical expansion: Urban areas tend to be more brightly lit than rural areas. From this simple observation, lights data can be used to measure a city’s spatial footprint and examine its growth over time. If the intensity of light exceeds a certain “Digital Number” (DN) level then an area is classified as part of a city, whilst areas with lower levels of light intensity are classified as rural. While night-time light monitoring may not be as accurate as that based on other types of satellite imagery such as MODIS and LANDSAT, it’s less time consuming and more cost-effective. The night-time lights data offers relatively low-cost results, providing a good overview of patterns of urban physical expansion at a national or regional scale.
Patterns of urban economic growth: Economic growth is conventionally measured using GDP. However, even in developed countries, national statistics offices don’t tend to publish the data at the level of individual urban settlements. As argued by Henderson et al., a key strength of the night-lights data is that it can be used to measure economic growth at refined spatial levels in the absence of GDP data. For instance, in our work, we are using the data to compare rates of economic growth across all South Asian cities with a population greater than 100,000. The definition of these cities is based on their spatial extents as observed in the lights data.
Intensive versus extensive growth: The night-lights data also can be used to compare cities in terms of patterns of growth. Because the data measures not only the presence of light but also its intensity, it can be used to assess the degree to which a city is growing intensively (increasing levels of economic activity and density within its pre-existing footprint) versus extensively (outward expansion of its footprint over time). So far, our results for South Asia, show considerable variation in this respect. Some cities, such as Kabul and Thimphu, exhibit both intensive and extensive growth; others, such as Gwadar in Pakistan and Mysore in India, appear to be growing outwards while their cores stagnate.
Single cities versus multi-city agglomerations: One of the most striking preliminary results for South Asia is the existence and emergence, over a 10 year period, of multi-city agglomerations; continuously lit belts of urbanization which consist of two or more cities, each with a population in excess of 100,000. The ability to distinguish between single cities and multi-city agglomerations is another strength of the lights data. Growing numbers of multi-city agglomerations suggest increasing “connectedness” throughout a country’s urban system.
The intriguing case of dimming cities: Based on our analysis so far, the night-lights data indicate that while most cities in South Asia have grown over time, there are a non-negligible number whose lights have dimmed over time. Although cities in advanced countries have been known to go into severe and absolute economic decline as a result of the more general process of deindustrialization (think of Detroit), it’s harder to imagine that such cities exist in the rapidly urbanizing environment of South Asia. We’re still investigating the causes of dimming lights in these cities; one possibility is that they may be driven by increased power outages over our sample period. This highlights the importance of using the lights data in combination with other sources of data to obtain a full understanding of patterns of urban development.
- Measurement error: As with most sources of data, the night-lights data is not free of measurement error. In particular, measurement error arises from several sources, including the deterioration of satellite sensors over time and the ways in which the data is acquired and then transformed into a usable format. Most notably, there is a well-known “blooming” problem where the light diffuses over an area beyond the true extent of lit area. This leads the “raw” lights data to overstate a city’s spatial extent (not to mention its level of GDP). Imposing a threshold “Digital Number” to distinguish urban from rural areas helps mitigate this problem. However, it’s unlikely that the threshold that helps accurately delineate urban extents in South Asia can accurately delineate urban extents in, for example, Africa (where population densities are much lower) or even urban extents in South Asia in the future.
Overall, the night-time lights data represents an intriguing data set for analyzing spatial patterns of urban expansion and economic growth. However, whilst we’re finding the data particularly useful in countries like Afghanistan, where alternative data sources are comparatively thin, it should not be considered a panacea to all data problems. It’s important to triangulate it with other sources of information wherever possible. Nevertheless, as we’re discovering in the South Asia region, it can, when properly used, shed new and exciting light on detailed spatial patterns of urban growth for both development practitioners and policymakers.