Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South

Forrest R. Stevens, Andrea E. Gaughan, Jeremiah J. Nieves, Adam King, Alessandro Sorichetta, Catherine Linard, Andrew J. Tatem

Research output: Contribution to journalArticle

Abstract

Mapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.

Original languageEnglish
JournalInternational Journal of Digital Earth
DOIs
Publication statusPublished - 23 Sep 2019

Fingerprint

land cover
Learning systems
Feature extraction
census
population density
Image processing
Availability
population modeling
image processing
urbanization
imagery
infrastructure
income
comparison
method
human population
Statistical Models

Keywords

  • built areas
  • Land cover
  • population modeling
  • remote sensing
  • settlement mapping

Cite this

Stevens, Forrest R. ; Gaughan, Andrea E. ; Nieves, Jeremiah J. ; King, Adam ; Sorichetta, Alessandro ; Linard, Catherine ; Tatem, Andrew J. / Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South. In: International Journal of Digital Earth. 2019.
@article{c0be8d666e4f4661a1a891cd7644f280,
title = "Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South",
abstract = "Mapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.",
keywords = "built areas, Land cover, population modeling, remote sensing, settlement mapping",
author = "Stevens, {Forrest R.} and Gaughan, {Andrea E.} and Nieves, {Jeremiah J.} and Adam King and Alessandro Sorichetta and Catherine Linard and Tatem, {Andrew J.}",
year = "2019",
month = "9",
day = "23",
doi = "10.1080/17538947.2019.1633424",
language = "English",
journal = "International Journal of Digital Earth",
issn = "1753-8947",
publisher = "Taylor & Francis",

}

Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South. / Stevens, Forrest R.; Gaughan, Andrea E.; Nieves, Jeremiah J.; King, Adam; Sorichetta, Alessandro; Linard, Catherine; Tatem, Andrew J.

In: International Journal of Digital Earth, 23.09.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Comparisons of two global built area land cover datasets in methods to disaggregate human population in eleven countries from the global South

AU - Stevens, Forrest R.

AU - Gaughan, Andrea E.

AU - Nieves, Jeremiah J.

AU - King, Adam

AU - Sorichetta, Alessandro

AU - Linard, Catherine

AU - Tatem, Andrew J.

PY - 2019/9/23

Y1 - 2019/9/23

N2 - Mapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.

AB - Mapping built land cover at unprecedented detail has been facilitated by increasing availability of global high-resolution imagery and image processing methods. These advances in urban feature extraction and built-area detection can refine the mapping of human population densities, especially in lower income countries where rapid urbanization and changing population is accompanied by frequently out-of-date or inaccurate census data. However, in these contexts it is unclear how best to use built-area data to disaggregate areal, count-based census data. Here we tested two methods using remotely sensed, built-area land cover data to disaggregate population data. These included simple, areal weighting and more complex statistical models with other ancillary information. Outcomes were assessed across eleven countries, representing different world regions varying in population densities, types of built infrastructure, and environmental characteristics. We found that for seven of 11 countries a Random Forest-based, machine learning approach outperforms simple, binary dasymetric disaggregation into remotely-sensed built areas. For these more complex models there was little evidence to support using any single built land cover input over the rest, and in most cases using more than one built-area data product resulted in higher predictive capacity. We discuss these results and implications for future population modeling approaches.

KW - built areas

KW - Land cover

KW - population modeling

KW - remote sensing

KW - settlement mapping

UR - http://www.scopus.com/inward/record.url?scp=85073835335&partnerID=8YFLogxK

U2 - 10.1080/17538947.2019.1633424

DO - 10.1080/17538947.2019.1633424

M3 - Article

AN - SCOPUS:85073835335

JO - International Journal of Digital Earth

JF - International Journal of Digital Earth

SN - 1753-8947

ER -