Data application of the month: Population and Settlement data

How are population and settlement data used in disaster risk reduction and response efforts?

Human population data is essential for disaster management efforts to assess how many people are potentially exposed to a hazard or reside within an already affected area. An exposure map shows this assessment. They are generated by merging geospatial population data with hazard related data and interpreting the overlaps. Information on the number/density of people exposed aids preparedness efforts in anticipation of a major event such as a hurricane, as high exposure areas can be prioritized for evacuation efforts in order to potentially save a significant number of lives. Exposure maps can also be used to investigate the likely impact caused by a natural hazard, as estimations can be made regarding the extent of the damge and casualties if population data and disaster magnitude are known.

While census data conveys how many people are living in an administrative unit, Earth Observation (EO) satellite data shows where people live and provides information about population dynamics. For exposure maps to be optimally useful, data sets need to be up-to-date, of high quality, and easily accessible. Satellite-based population data can contribute to monitor target 2 of the Sendai framework for Disaster Risk Reducyion 2015-2030, i.e. to aim for "...a substantial reduction in numbers of affected people" with respect to disasters. In addition, population and settlement data can contribute to monitor some of the Sustainable Development Goals and targets, for example target 11.3: "By 2030, enhance inclusive and sustainable urbanization and capacity for participatory, integrated and sustainable human settlement planning and management in all countries."

Satellite-based population and settlement data provides useful information for exposure and risk mapping. It can also contribute to monitor some of the indicators of the Sendai Framework for Disaster Risk Reduction, as well as those related to the Sustainable Development Goals. A key advantage regrding satellite imagery is that it is monitored on a daily basis in contrast to census data, which is only updated every decade or less depending on funding or interest. Satellite imagery is capable of monitoring changes in population and settlement dynamics on a more frequent basis.

How are urban settlements mapped from space and how does satellite imagery contribute to the estimation of population data?

There are many ways to map the human population. An urban settlement map shows where people live and the dynamic of the built environment. Census data can be converted geospatially for a view of how many people actually live in a certain area.

An new data set called the Global Urban Footprint (GUF) has recently drawn attention. It is a global map of the location, size, and shape of the built environment now in an unprecedented spatial resolution of 0.4 arc seconds (~12m). It is the most detailed and consistent map of the built environment so far. The map encompasses both large urban conglomerates as well as dispersed small-scale built areas in rural regions. It's availability will be essential for a myriad of urban and disaster management initiatives.

The radar satellites TerraSAR-X and TanDEM-X are responsible for providing information for GUF data. The German Aerospace Center (DLR) launched these satellites in 2007 and 2010. Post-mission processing of the satellite imagery is what ultimately produced the GUF. Using Synthetic Aperture Radar (SAR), human settlement areas can be delineated because they show a characteristic small-scale heterogeneity of local wave backscatter. Essentially, settlements display double bounce effects and direct backscattering associated with certain structural or geometric features, particularly from vertical walls (buildings, bridges, traffic signs, etc.). SAR is advantageous for mapping buildings because it works on an all-weather basis. Moreover, the structural properties of urban settlements that produse unique backscatter are similar all over the globe because nearly an man-made structure shows vertical walls.

It is important to note that roads, parking lots, or airport runways are impervious man-made features that are not included in GUF because they do not have a vertical component. False positives were most common in rugged terrain: large rocks en masse, steep embankments, etc. False omission occurred when buildings were very close together or surrounded by high trees that caused the vertical walls to not be illuminated by the radar beam. These errors were fixed by including additional reference layers derived from multispectral satellite data; this leads to a corrected final GUF product (Figure 1).

GUF Image
Figure 1: A correction workflow for the final GUF product. The top image shows initial identification, the images in the middle show thematic reference layers and the image at the bottom shows the final GUF map. Image: The Global Urban Footprint.

It is also important to estimate how many people are residing in an area. Population census data is originally collected as detailed data at the household level, then spatially aggregated to the level of administrative units such as a municipal district (top image of Figure 2). This means that each pixel within an administrative boundary is assigned the same population value, no matter where it is an urban area or a forest. It is obvious that this is not representative of real world population congregation. Therefore, to create a meaningful exposure map within an administrative boundary, it is necessary to disaggregate the census data. The disaggregation is possible using satellite imagery.

Census Data


Census Data

Figure 2: Census data that are aggregated by administrative units (top image) can be spatially disaggregated with the help of satellite-based land cover map and satellite-based lights at night map. The bottom image shows the result of the random forest modelling approach. Images: WorldPop.

Facebook hosts the finest resolution of global population distribution data available at 30 metres. The social media company combines machine vision AI with satellite imagery and census information to create population density maps (Figure 3). In cooperation with the Center for International Earth Science Information Network (CIESIN) at Columbia University, Facebook utilizes Convolutional Neural Networks on high resolution satellite imagery in order to identify buildings from freely available mapping resources (Maxar Imagery Basemaps) and validates these findings against available census data (Gridded Population of the World). Moreover, disaggregating this data with regard to available and highly-detailed demographic information allows for the consequent introduction of the data into pre-existing demographic data sets and a comprehensive overview of demographic measurements, specifically related to age and gender. This demographic information for a total of 40 specific categories is then integrated into the population density maps. Collectively, these maps provide information on both the location and the demographic of a population in a certain country.

Facebook Population Map
Figure 3: Example of a Population Density Map by Facebook. Image: Facebook Connectivity.

LanScan hosts global population distribution data available at approximately 1 kilometer resolution. The LandScan algorithm uses spatial data and a multi-variable dasymetric modeling approach to disaggragate census data within an administrative boundary. Based on the spatial data and the socioeconomic and cultural understanding of an area, cells are preferentially weighted for the possible occurrence of population during the day. Whithin each conutry, the population distribution model calculates a "likelihood" coefficient for each cell and applies the coefficients to the corresponding census data. The total population for any given area is then allocated proportionally to each cell. Since no single population distribution model can account for the differences in spatial data availability, quality, scale, and accuracy as well as the differences in cultural settlement practices, LandScan population distribution models are tailored to match the data conditions and geographical nautre of each individual country and region.

Landscan Distribution Data Set
Figure 4: LandScan population distribution dataset. Image: Landscan.

The above examples provide estimates on how many people are residing where. Using archived satellite imagery it is also possible to provide estimates regarding when people have resided where. Understanding and measuring settlement dynamics is important to update exposure and risk maps, especially when urban fringes with vulnerable peoples expand into hazard-prone areas.

A newly available data set on human settlement has gained recent publicity: The World Settelement Footprint (WSF) 2015 (Figure 5). Launched by the German Aerospace Center (DLR), the WSF offers valuable insight into human settlement on Earth in 2015 at a resolution of 10 metres. In particular, the map allows for a more granular look at various aspects of human settlement in the recent past: migration flows, at-risk populations and consumption information are all visible through WSF. The satellites Landsat-8 and Sentinel-1 (A/B) are responsible for providing optical and radar satellite imagery from 2014-2015 for WSF data. Landsat-8 was launched by the National Aeronautics and Space Administration (NASA) in 2013. Landsat-8 uses a Thermal Infrared Sensor (TIRS) and a Operational Land Imager (OLI) to measure surface temperature and access high-resolution imagery. Sentinel-1 A and  Sentinel-1 B were launched by the European Space Agency (ESA) in 2014 and 2016, respectively. Both use C-Band Synthetic Aperture Radar (SAR) to acquire imagery regardless of the weather. Landsat-8 and Sentinel-1 (A/B) satellites complement eachother as they pertain to different on-ground structures, human-made and built-up, respectively. The WSF2015 data set is available to the public free of charge.

World Settlement Image
Figure 5: A portion of Western Africa in WSF2015. Image: DLR.

An innovative approach to map population dynamics even during day and night can be carried out through the use of mobile phone data as shown in the video below. For this aproach, a dense network of cell towers is necessary, i.e. the approach is not applicable in all regions of the world. Even though this is not a space-based application, we include this approach here since it is showing what is possible to monitor this extremely important data.

Video 1: Dynamic population mapping using mobile phone data. For more information read the paper by Deville et al. (2014).

How can I access population and settlement data?

There are several data sets based on remote sensing that can be used to assess population. Below are links to population estimates, urban extent, urban land use, human settlement layers, socioeconomic data, and urban hazards.

Population data

  • The Global Urban Footprint (DLR) (Link)
  • Gridded Population of the World (NASA) (Link)
  • LandScan Global Population Distribution Dataset (ORNL) (Link)
  • WorldPop (Link)
  • Socioeconomic data and applicatons (NASA) (Link)
  • Data Discovery Portal (WASCAL) (Link)
  • Cartographic section of the United Nations (Link)
  • Urban hazard map web platform (ISCGM) (Link)
  • Urban Atlas (EEA) (Link)
  • Global human settlement layer (JRC) (Link)
  • Africa High Resolution Population Density Maps (Facebook) (Link)
  • Population Density Maps (Facebook) (Link)
  • World Settlement Footprint (DLR) (Link)

How are the datasets used for disaster risk management and emergency response?

An operational example of using satellite-based population data for emergency response can be found in the Copernicus Emergency Mapping Service (EMS). The Copernicus EMS portfolio for rapid mapping includes reference maps and delineation maps. Refrence maps provide a quick updated information on the territory and existing assets using data collected before the disaster. By combining the reference information with population data, Copernicus provides an estimate of people exposed within the area of interest (AOI). Exposure is an integral part of any disaster risk profile. Delineation maps provide an assessment of the geographical extent of the disaster event. Combining the geographical extent polygon with population data, Copernicus estimates the number of people affected.

This was put into practice during the floods in Malawi in January 2015. Heavy rains during several weeks lead to severe flooding across the country. The floods caused the displacement of thousands of people, extensive damage to crops, livestock, and infrastructure. Some areas were inaccessible, impeding on the ground conduction of assessments regarding damages and losses. Copernicus EMS was activated by the European Commission's Humanitarian and Civil Protection Department on 25 January 2015 (activation website). The first reference delineation maps were published (Figure 6), giving a more precise estimate of the number of people affected within the flooded area (cf. Table 2).

EMS Reference Map
Figure 6: EMS Reference Map. Image: Copernicus.
Table of Reference Map
Table 1: Exposure table of reference map. Table: Copernicus.
EMS Delineation Map
Figure 7: EMS Delineation Map. Image: Copernicus.
Table of Delineation Map
Table 2: Exposure table of Delineation Map. Table: Copernicus.


Zircon - This is a contributing Drupal Theme
Design by WeebPal.