An increasing number of data providers are today combining information from censuses with satellite-derived geospatial features to redistribute populations and produce gridded population datasets.
The COVID-19 pandemic, which originated from Wuhan, China has created a war-like situation in the world. More than 170 countries are affected; people are mostly confined to their homes, if not battling for their lives or recovering in hospitals; and nations are struggling to limit the number of those infected. The situation is even more challenging in underdeveloped, or remote parts, where the administrations don’t even know the population count, let alone providing people with food, medicines and other essentials in this time of crisis.
Importance of population data.
Identifying the size, structure and distribution of a population is essential for planning development related works. Without knowing where people are located, governments and policymakers cannot improve/ expand access to health, transportation, energy and other services.
SDSN TReNDS Manager Maryam Rabiee, who has complied a report titled Leaving No One off the Map: A Guide for Gridded Population Data for Sustainable Development, says, “Population and the environment are constantly changing and to ensure that we leave no one behind, we need reliable population data. A number of SDG indicators are related to population, and they measure access to basic services and facilities.”
As population scientists have expanded the range of topics they study, increasingly focusing on the relationship between population and social, economic and health conditions, there have been enhancements in data collection and emergence of new data collection techniques and procedures. This has also led to the evolution of the concept of gridded population data.
After creating a base map of geographic cells based on satellite imagery, each cell can be viewed as the representation of an area on the surface of the Earth, typically defined by its latitude-longitude coordinates. The collection of these cells, rows and columns defines a grid area. The population data derived by this process is called gridded population data.
There are various methods of data collection, the most common and reliable among them being the census, or the complete enumeration survey. Governments all over the world bank on this method to collect data concerning population, housing, agriculture, etc. A population census is considered highly accurate because data from each household is collected and studied before drawing any conclusions.
However, the method has its limitations. For example, census data is only collected once in ten years and thus may not be fully accurate after a certain period, as a lot can change in just a few years. In some countries, the gap between two census is even longer. Another limitation of this methods is the inability of the enumerators to access certain locations, especially conflict or disaster-hit regions, as well as areas where local language presents a communication barrier. In such cases, the population residing in an area is either miscounted or is completely left uncounted.
Integrating geospatial for accuracy
The advancements in geospatial technology and remote sensing have paved the way for the production of more frequent population data which is more accurate and helps governments to design development plans “leaving no one behind”.
An increasing number of data providers are combining information from censuses with satellite-derived geospatial features to redistribute populations and produce gridded population datasets. “Gridded population datasets can be used in a wide range of application areas, such as disaster response, health interventions and survey planning, and they can offer us more spatially refined estimates,” explains Rabiee.
The integration also allows redistributing population data within different geographic boundaries to identify and characterize settlements and built infrastructure, manage resources, urban and rural planning, risk management and disaster response. Satellite imagery is significant to this method of producing population estimates because it does not face geographical and temporal limitations of traditional data sources and allows for more frequent population estimates.
Challenges of gridded population data
While gridded population data offers great opportunities in both science and immediate response applications, it is not perfect. For datasets based on census data, users need to be aware of the age and quality of the censuses. Further, the datasets can be dramatically different for different countries.
Highlighting these challenges, Hayden Dahmm, SDSN TReNDS Manager who worked with Rabiee on the report, says, “They (gridded data) are not universally accurate and uncertainty is inherent in the estimations. For example, there can be issues with mistaking rock formations for houses, overlooking settlements in heavily forested areas, or incorrectly distributing populations over uninhabited areas. Additionally, coastlines are a source of greater uncertainty, and different models can produce highly different estimations of the local population. If we apply these models without recognizing their unique characteristics, there is a possibility that people will be overlooked.”
“Despite the progress around gridded population data, there is still considerable ambiguity with regard to the datasets. Each of the datasets has been developed for different purposes and each has its advantages and disadvantages. In our research with stakeholders, it was apparent that most policymakers and other data users lack the time and technical expertise to understand different characteristics, applications, limitations and potential of gridded datasets,” explains Rabiiee.
Overcoming data problems
To address most of these challenges, POPGRID was established. Led by the Center for International Earth Science Information Network (CIESIN) at Columbia University, SDSN TReNDS and the Global Partnership for Sustainable Development Data (GPSDD), POPGRID is a “data collaborative” which aims to accelerate the development and use of high quality georeferenced data on human settlements, infrastructure and populations by convening and drawing on the expertise of an international, interdisciplinary community of data developers and users from both public and private sectors.
“Our main focus at POPGRID is to do a better job sharing, accessing and documenting these kinds of data, and in particular to work with the stakeholder and user communities to have a better sense of their priorities and needs and see if we can match the provision of data with the demand and use for gridded population data,” says Dr. Robert Chen, SDSN TReNDS Co-Chair, Director of CIESIN, and Manager of NASA’s Socioeconomic Data and Applications Center (SEDAC).
To bridge the knowledge gap around gridded population data and help improve its accessibility and understanding among policymakers and other users, SDSN TReNDS, on behalf of the POPGRID Collaborative, has complied a report. “The report drawn from an extensive literature review and interviews with key data providers and users in POPGRID, presents an overview, analysis and recommendations for the use of gridded population datasets in a wide range of application areas. The document also presents a comparison assessment of the use of different datasets and their varying outputs, and will address many misconceptions around gridded population data,” explains Rabiee.
The report was written with two overarching questions in mind — how can gridded population data supplement current population data sources and support users from the sustainable development community to make timely, informed decisions; and which gridded population dataset is the most suitable for a the user?
Dahmn says, “Gridded population data are to complement, not substitute census data. The datasets featured in this report rely on information derived from national censuses to produce estimates with higher frequency and/or granularity.”
The report also suggests that gridded population data can be a valuable supplement to traditional data sources, but it is not error-free. Although these datasets address some of the limitations of traditional sources, they do add their own sources of uncertainty. More validation work is needed to compare gridded population data estimates against authoritative data on population location. There is a critical need for a more systematic analysis and objective validation of these products to further refine methods and improve their accuracy and utility.