Available via license: CC BY 4.0
Content may be subject to copyright.
Contents lists available at ScienceDirect
Development Engineering
journal homepage: www.elsevier.com/locate/deveng
Geospatial data for research on economic development
It is estimated that in 2016 nearly 38 per cent of workers in the
world's least developed countries were living with their families
below the poverty line (approximately $1.90 per person per day) (UN,
2017), often without access to schools, healthcare, electricity, safe
water and other critical services. Yet, accurate data on the number of
poor people and their living conditions are scarcely existent, espe-
cially in developing countries. Census counts comprise the primary
source of this type of data, but are published infrequently, varying in
resolution and precision, due in part to limited available resources for
data collection.
The gap in the availability of data between the developing and the
developed world is, at least partially, the result of technological ad-
vances and innovative solutions for data collection. Novel sources of
data facilitate the development of new approaches for demographic
modeling and measurement in wealthy nations, while in developing
countries, fewer sources of big data exist (Blumenstock et al., 2015). For
example, while nearly half of today's world population has internet
access, more than 4 billion people - mostly in developing countries - still
don't have access to the internet.
1
The result is an increasing gap in the
availability of data for developing and developed countries. Consider,
for example, the utilization of data gathered from online social media
for demographic research. Social media platforms such as Facebook,
Twitter and other collaborative virtual spaces are increasingly har-
nessed to track the digital traces (Bruno, 2007) of society, whether for
monitoring stocks of migrants (Zagheni Emilio et al., 2017) or for
generating taxonomies of happiness levels in US states (Mitchell et al.,
2013). This type of data only provides insights about those who parti-
cipate in such virtual spaces and does not always fully capture the social
complexity of the ever-changing developing world.
The lack of accurate and timely data on the distribution of poverty
also often forces policy-makers to make decisions based on limited (and
sometimes inaccurate) data (Blumenstock, 2016). For example, devel-
opment programs and policies are typically designed to change out-
comes (e.g. to raise incomes, to improve learning, or to reduce illness).
An impact evaluation is intended to evaluate the outcome of a program
by answering one specific question: What is the impact of a specific
program on an outcome of interest? (Gertler et al., 2016). Answering
this question requires to track and monitor the impacts of the program -
and this can only be done by a continuous and frequent collection of
data from multiple sources. Such data provides the basis for any evi-
dence-based policy making.
But poverty and other societal characteristics also hold a spatial
dimension. Poverty can often be identified visually and reflected in the
density of residential structures, patters of roads, the material of roofs
or the distribution and extent of green spaces. Many of these features
are observable from space. From the perspective of social science,
capturing this data may provide meaningful information about the
context that shapes social phenomena (Council and Stern, 1988). There
are more than 1400 satellites continuously collecting an unprecedented
amount of data from Earth at ever-improving spatial, spectral and
temporal resolutions, over increasingly short time intervals. Data col-
lected by these satellites and other sensors are progressively shifting the
way we measure and understand our world. These data allow re-
searchers to study and map the physical environment and infer about
social processes in near real time, in almost every corner of the world.
The heightened availability of satellite data has revolutionized our
ability to map human settlements and understand the distribution and
characteristics of the population on Earth. New methods for machine
learning, predictive analytics and remote sensing are reforming the way
we measure and address societal challenges, detect poverty and steer
efforts to reduce it. Researchers are continually discovering new ways
to convert remotely sensed data into meaningful information about the
nature and pace of change of human and physical landscapes.
Over the last few decades, researchers have been working to detect
and map poverty and wealth from space, primarily with nighttime light
data. Based on the assumption that poorly lit areas with high popula-
tion counts signify higher percentages of poor people and lower wealth,
researchers have been able to estimate the number of individuals living
in poverty, globally (Elvidge et al., 2009). Nighttime lights are also
being extensively used to map urbanization. According to this ap-
proach, a pixel's nighttime light value exceeding a specified threshold
(which may vary across regions or countries) signifies urban develop-
ment. Moreover, because the intensity of light emitted at night is clo-
sely related to economic development, this proxy is also leveraged to
predict national economic growth (Henderson et al., 2012) as well as
regional economic development (Doll et al., 2006). However, inference
of human activity with Nighttime lights is often inaccurate, particularly
in low-density urban areas (Zhang and Seto, 2013). DMSP-OLS sensors,
which until recently (with the availability of data collected by the
Visible Infrared Imaging Radiometer Suite (VIIRS)) have been the pri-
mary source for nighttime light data can exaggerate the extent of urban
areas (Henderson et al., 2003) while overlooking small or developing
https://doi.org/10.1016/j.deveng.2019.100041
1
https://www.weforum.org/agenda/2016/05/4-billion-people-still-don-t-have-internet-access-here-s-how-to-connect-them/.
Development Engineering 4 (2019) 100041
Available online 27 February 2019
2352-7285/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
T
settlements. In addition, the extent and intensity of lit areas cannot
directly delimit urban regions due to the “blooming” effect (the iden-
tification of lit areas as consistently larger than the settlements with
which they are associated) (Imhoff et al., 1997) and “over saturation” of
pixels (i.e., pixels in bright areas, such as in city centers, reach the
highest possible intensity and no further details can be recognized)
(Hsu et al., 2015). As a result, there has been a significant move towards
mapping poverty with daytime satellite data at higher spatial resolu-
tions. For example, Engstrom et al. (2017) showed that object and
texture features extracted from high-resolution satellite imagery over
Sri Lanka (e.g. the number and density of buildings, number of cars,
density and length of roads, roof material, etc.) could explain nearly 60
percent of poverty headcount rates and average log consumption,
compared to models built using nighttime lights that could explain only
15 percent of the variation in poverty or income. Jean et al. (2016)
combined nighttime maps with high-resolution daytime satellite images
to estimate consumption expenditure and asset wealth. The authors
showed, that a convolutional neural network can be trained to identify
image features that could explain up to 75% of the variation in local-
level economic outcomes in Africa. Other studies develop novel
methods designed to map slums and informal settlements in developing
countries by capturing some of their unique spatial features (e.g.
building density, building heights, and sizes), differentiating between
informal from formal settlements (Taubenböck and Kraff, 2014). With
around 1 billion people living today in slums (UN-Habitat, 2015)
lacking access to drinking water, sanitation facilities and durable
housing conditions, accurate and up-to-date information on the extent
of slums and the number of slums dwellers is fundamental for the
preservation of a sustainable human society.
These new sources of geospatial data are transforming our under-
standing of the modern world. Novel machine-learning algorithms are
combined with satellite imagery from multiple sources to estimate road
quality in developing countries, predict consumption expenditure,
measure wealth and inequality, monitor small-holder farmers' yields in
regions affected by drought, and monitor electricity provision in rural
areas. Parallel computing platforms with large storage capacities are
becoming increasingly available, allowing for the extraction of mean-
ingful information about the developing world and encouraging new
design of environmentally responsible approaches to measuring and
monitoring economies.
This special issue of Development Engineering highlights 5 case
studies that illustrate novel use of technology and new sources of big
geospatial data to derive insights pertaining to societal processes,
primarily in developing countries. Augusto Zagatti et al. utilize na-
tionwide de-identified Call Detail Records (CDR) to investigate
nighttime and daytime population densities and commuting patterns
in Haiti. The authors construct origin-destination matrices of
weighted connections between home and work locations to identify
fragmented labor markets in the country. Eitzel et al. evaluate the
potential of ‘medium-tech’ digital mapping technologies as solutions
for empowering disenfranchised communities in rural Zimbabwe. The
authors demonstrate that digital collaborative mapping with rela-
tively simple technology allows spreading information among rural
farmers in Zimbabwe. This case study demonstrates the potential of
geospatial data as a tool to address poverty and sustainability and
improve the well-being of the most vulnerable populations on our
planet. Marty et al. apply a Geographic Simulation and Extrapolation
(GeoSIMEX) approach to mitigate the spatial imprecision inherent in
the geoparsed data when evaluating the impact and success of in-
ternational aid and interventions projects. The authors present ana-
lysis of the impact of Chinese aid on vegetative land cover in Rwanda
and Burundi and find that Chinese investments have slowed the loss
of vegetation in Rwanda. Rivera Ballesteros et al. utilize satellite
imagery collected by Sentinel-2 satellite for supervised image classi-
fication to estimate the extent of tree cover in twelve gazetted forests
in semi-arid Burkina Faso. The authors show this source of imagery
allows to detect patches of forest cover in arid and semi-arid regions
that could not be identified with lower resolution imagery and eval-
uate the extent and characteristics of deforestation in these regions.
Finally, Goldblatt et al. discuss the challenges of measuring urbani-
zation in developing countries that lack the resources and infra-
structure needed to produce reliable data. The authors highlight the
potential use of publicly available remotely sensed data for mapping
changes in the built-up LC/LU in Ho Chi Minh City, Vietnam, in the
period between 2000 and 2015.
The invention of computerized Geographic Information Systems
(GIS) in the 1960s revolutionized how we collect, map and analyze
spatial data and give meaning to the world around us. With advances in
computer processing power and data storage, increased availability of
personal computers, the invention of the internet (the World Wide
Web) and the growing popularity of mobile and “smart” phones, the
volume of data we generate and consume is growing exponentially.
Devices on satellites, airplanes and UAVs collect massive amounts of
imagery that capture the characteristics of Earth. All these studies
featured in this special section illustrate novel research methods to
analyze these unique sources of data to achieve a better understanding
of the nature of developing countries.
References
Blumenstock, J.E., 2016. Fighting poverty with data. Science 353, 753–754. https://doi.
org/10.1126/science.aah5217.
Blumenstock, J., Cadamuro, G., On, R., 2015. Predicting poverty and wealth from mobile
phone metadata. Science 350, 1073–1076. https://doi.org/10.1126/science.
aac4420.
Bruno, L., 2007. Beware, your imagination leaves digital traces. Times High. Lit.(6).
Council, R.R., Stern, P.C., 1988. Linking remote sensing and social science: the need and
the challenges. In: People and Pixels: Linking Remote Sensing and Social Science.
National Academies Press, pp. 1–27.
Doll, C.N.H., Muller, J.-P., Morley, J.G., 2006. Mapping regional economic activity from
night-time light satellite imagery. Ecol. Econ. 57, 75–92. https://doi.org/10.1016/j.
ecolecon.2005.03.007.
Elvidge, C.D., Sutton, P.C., Ghosh, T., Tuttle, B.T., Baugh, K.E., Bhaduri, B., Bright, E.,
2009. A global poverty map derived from satellite data. Comput. Geosci. 35,
1652–1660. https://doi.org/10.1016/j.cageo.2009.01.009.
Emilio, Zagheni, Weber, Ingmar, Krishna, Gummadi, 2017. Leveraging Facebook's ad-
vertising platform to monitor stocks of migrants. Popul. Dev. Rev. 43, 721–734.
https://doi.org/10.1111/padr.12102.
Engstrom, R., Hersh, J., Newhouse, D., 2017. Poverty from Space: Using High-Resolution
Satellite Imagery for Estimating Economic Well-Being (Policy Research Working
Paper No. 8284). Poverty and Equity Global Practice Group, World Bank,
Washington DC.
Gertler, P.J., Martinez, S., Premand, P., Rawlings, L.B., Vermeersch, C.M., 2016. Impact
Evaluation in Practice. The World Bank.
Henderson, M., Yeh, E.T., Gong, P., Elvidge, C., Baugh, K., 2003. Validation of urban
boundaries derived from global night-time satellite imagery. Int. J. Remote Sens. 24,
595–609. https://doi.org/10.1080/01431160304982.
Henderson, J.V., Storeygard, A., Weil, D.N., 2012. Measuring economic growth from
outer space. Am. Econ. Rev. 102, 994–1028. https://doi.org/10.1257/aer.102.2.994.
Hsu, F.-C., Baugh, K.E., Ghosh, T., Zhizhin, M., Elvidge, C.D., 2015. DMSP-OLS radiance
calibrated nighttime lights time series with intercalibration. Rem. Sens. 7,
1855–1876. https://doi.org/10.3390/rs70201855.
Imhoff, M., Lawrence, W.T., Stutzer, D.C., Elvidge, C.D., 1997. A technique for using
Development Engineering 4 (2019) 100041
2
composite DMSP/OLS “City Lights” satellite data to map urban area. Remote Sens.
Environ. 61, 361–370. https://doi.org/10.1016/S0034-4257(97)00046-1.
Jean, N., Burke, M., Xie, M., Davis, W.M., Lobell, D.B., Ermon, S., 2016. Combining sa-
tellite imagery and machine learning to predict poverty. Science 353, 790–794.
https://doi.org/10.1126/science.aaf7894.
Mitchell, L., Frank, M.R., Harris, K.D., Dodds, P.S., Danforth, C.M., 2013. The geography
of happiness: connecting twitter sentiment and expression, demographics, and ob-
jective characteristics of place. PLoS One 8, e64417. https://doi.org/10.1371/
journal.pone.0064417.
Taubenböck, H., Kraff, N.J., 2014. The physical face of slums: a structural comparison of
slums in Mumbai, India, based on remotely sensed data. J. Hous. Built Environ. 29,
15–38. https://doi.org/10.1007/s10901-013-9333-x.
UN, 2017. Progress towards the Sustainable Development Goals: Report of the Secretary-
General (Economic and Environmental Questions: Sustainable Development).
(United Nations).
UN-Habitat, 2015. Slum Almanac 2015/2016: Tracking Improvement in the Lives of Slum
Dewllers. Participatory Slum Upgrading Programme, UN-Habitat.
Zhang, Q., Seto, K.C., 2013. Can night-time light data identify typologies of urbanization?
A global assessment of successes and failures. Rem. Sens. 5, 3476–3494. https://doi.
org/10.3390/rs5073476.
Ran Goldblatt
∗
, Madeline Jones, Brad Bottoms
New Light Technologies Inc., 1440 G Street Northwest, Washington, DC
20005, USA
E-mail address: ran.goldblatt@nltgis.com (R. Goldblatt).
∗
Corresponding author.
Development Engineering 4 (2019) 100041
3
