Gridded Population Datasets Research Articles

The impact of seasonality on multi-scale feature extraction techniques

ABSTRACTA key component in constructing a broad-scale, gridded population dataset is fine resolution geospatial data accurately depicting the extent of human activity. Analogous datasets are often developed using a wide range of methods and classification techniques, including the use of spatial features, spectral features, or the coupling of both to identify the presence of man-made structures from high-resolution satellite imagery. By using spatial and textural-based descriptors to generate high-resolution settlement layers for two dissimilar regions at the peak of seasonal disparity, this study attempts to quantify the influence of seasonality on the accuracy of a supervised, multi-scale, feature extraction framework for automated delineation of human settlement. Results generated by numerous models are evaluated against a reference dataset allowing for assessment of seasonal and feature differences in the context of accuracy. Global or regional mapping of human settlement requires the assemblage of high-resolution satellite images with variegated acquisition characteristics (season, sun elevation, off-nadir, etc.) to produce a cloud-free composite image from which features are extracted. Results of this study suggest an emphasis on imagery criteria, in particular acquisition date, could improve classification accuracy when mapping human settlement at scale.

Linking Synthetic Populations to Household Geolocations: A Demonstration in Namibia

Whether evaluating gridded population dataset estimates (e.g., WorldPop, LandScan) or household survey sample designs, a population census linked to residential locations are needed. Geolocated census microdata data, however, are almost never available and are thus best simulated. In this paper, we simulate a close-to-reality population of individuals nested in households geolocated to realistic building locations. Using the R simPop package and ArcGIS, multiple realizations of a geolocated synthetic population are derived from the Namibia 2011 census 20% microdata sample, Namibia census enumeration area boundaries, Namibia 2013 Demographic and Health Survey (DHS), and dozens of spatial covariates derived from publicly available datasets. Realistic household latitude-longitude coordinates are manually generated based on public satellite imagery. Simulated households are linked to latitude-longitude coordinates by identifying distinct household types with multivariate k-means analysis and modelling a probability surface for each household type using Random Forest machine learning methods. We simulate five realizations of a synthetic population in Namibia’s Oshikoto region, including demographic, socioeconomic, and outcome characteristics at the level of household, woman, and child. Comparison of variables in the synthetic population were made with 2011 census 20% sample and 2013 DHS data by primary sampling unit/enumeration area. We found that synthetic population variable distributions matched observed observations and followed expected spatial patterns. We outline a novel process to simulate a close-to-reality microdata census geolocated to realistic building locations in a low- or middle-income country setting to support spatial demographic research and survey methodological development while avoiding disclosure risk of individuals.

Dasymetric mapping of urban population in China based on radiance corrected DMSP-OLS nighttime light and land cover data

High spatial resolution urban population dataset is increasingly required for sustainable urban planning and management. Dasymetric mapping is an effective approach to create such dataset. However, the created gridded total population datasets usually have limitation for urban analysis in developing countries as they usually underestimate urban population because of the strong urban-rural difference. In this study, we aimed to create a dataset of gridded urban population with 1 km resolution in China in year 2000 and 2010. We proposed an index of urban nighttime light (UNTL) by integrating radiance corrected DMSP nighttime light (RcNTL) and urban land, which is then used as weight to disaggregate county-level urban population. The validation using township population in Beijing as references shows reasonable accuracy with a mean relative error of 38% and a R2 of 68%. Using only two widely available datasets (RcNTL and urban land), the proposed method is simple and computing efficient compared with methods using multiple geospatial data (e.g., land use and land cover, distance to city center, slope) and that combined with remote sensing imagery. As the used two auxiliary datasets are accessible globally, the method has great potential to produce similar urban population dataset for other developing countries where fine scale census population datasets are scarce. The produced urban population dataset is valuable for enriching our understanding of the urbanization process and designing sustainable urban planning and management strategies in China.

Residential scene classification for gridded population sampling in developing countries using deep convolutional neural networks on satellite imagery

BackgroundConducting surveys in low- and middle-income countries is often challenging because many areas lack a complete sampling frame, have outdated census information, or have limited data available for designing and selecting a representative sample. Geosampling is a probability-based, gridded population sampling method that addresses some of these issues by using geographic information system (GIS) tools to create logistically manageable area units for sampling. GIS grid cells are overlaid to partition a country’s existing administrative boundaries into area units that vary in size from 50 m × 50 m to 150 m × 150 m. To avoid sending interviewers to unoccupied areas, researchers manually classify grid cells as “residential” or “nonresidential” through visual inspection of aerial images. “Nonresidential” units are then excluded from sampling and data collection. This process of manually classifying sampling units has drawbacks since it is labor intensive, prone to human error, and creates the need for simplifying assumptions during calculation of design-based sampling weights. In this paper, we discuss the development of a deep learning classification model to predict whether aerial images are residential or nonresidential, thus reducing manual labor and eliminating the need for simplifying assumptions.ResultsOn our test sets, the model performs comparable to a human-level baseline in both Nigeria (94.5% accuracy) and Guatemala (96.4% accuracy), and outperforms baseline machine learning models trained on crowdsourced or remote-sensed geospatial features. Additionally, our findings suggest that this approach can work well in new areas with relatively modest amounts of training data.ConclusionsGridded population sampling methods like geosampling are becoming increasingly popular in countries with outdated or inaccurate census data because of their timeliness, flexibility, and cost. Using deep learning models directly on satellite images, we provide a novel method for sample frame construction that identifies residential gridded aerial units. In cases where manual classification of satellite images is used to (1) correct for errors in gridded population data sets or (2) classify grids where population estimates are unavailable, this methodology can help reduce annotation burden with comparable quality to human analysts.

Accuracy Assessment of Multi-Source Gridded Population Distribution Datasets in China

Population is one of the core elements of sustainable development. Quantifying the estimation accuracy of population spatial distribution has been recognized as a critical and challenging task. This study aims to evaluate the data accuracy of four population datasets in China, including three global gridded population datasets, the Gridded Population of the World (GPW), Global Rural and Urban Mapping Project (GRUMP), and WorldPop project (WorldPop), and a Chinese regional gridded population dataset, the China 1 km Gridded Population (CnPop) dataset. These datasets are assessed using a specific method based on a GIS-linked 2000 census dataset at the township level in China. The results indicate that WorldPop had the highest estimation accuracy, estimating about 60% of the total population. CnPop accurately estimated about half of the total population, showing a good mapping performance. The GPW had an acceptable estimation accuracy in a few plain and basin areas, accounting for about 30% of the total population. Compared to the GPW, GRUMP accurately estimated about 40% of the total population. The relative estimation error analysis discovered the disadvantages of the generation strategies of these datasets. The conclusions are expected to serve as a quality reference for potential dataset users and producers, and promote accuracy assessment for population datasets in other regions and globally.

High-resolution reconstruction of the United States human population distribution, 1790 to 2010

Where do people live, and how has this changed over timescales of centuries? High-resolution spatial information on historical human population distribution is of great significance to understand human-environment interactions and their temporal dynamics. However, the complex relationship between population distribution and various influencing factors coupled with limited data availability make it a challenge to reconstruct human population distribution over timescales of centuries. This study generated 1-km decadal population maps for the conterminous US from 1790 to 2010 using parsimonious models based on natural suitability, socioeconomic desirability, and inhabitability. Five models of increasing complexity were evaluated. The models were validated with census tract and county subdivision population data in 2000 and were applied to generate five sets of 22 historical population maps from 1790–2010. Separating urban and rural areas and excluding non-inhabitable areas were the most important factors for improving the overall accuracy. The generated gridded population datasets and the production and validation methods are described here.

GridSample: an R package to generate household survey primary sampling units (PSUs) from gridded population data

BackgroundHousehold survey data are collected by governments, international organizations, and companies to prioritize policies and allocate billions of dollars. Surveys are typically selected from recent census data; however, census data are often outdated or inaccurate. This paper describes how gridded population data might instead be used as a sample frame, and introduces the R GridSample algorithm for selecting primary sampling units (PSU) for complex household surveys with gridded population data. With a gridded population dataset and geographic boundary of the study area, GridSample allows a two-step process to sample “seed” cells with probability proportionate to estimated population size, then “grows” PSUs until a minimum population is achieved in each PSU. The algorithm permits stratification and oversampling of urban or rural areas. The approximately uniform size and shape of grid cells allows for spatial oversampling, not possible in typical surveys, possibly improving small area estimates with survey results.ResultsWe replicated the 2010 Rwanda Demographic and Health Survey (DHS) in GridSample by sampling the WorldPop 2010 UN-adjusted 100 m × 100 m gridded population dataset, stratifying by Rwanda’s 30 districts, and oversampling in urban areas. The 2010 Rwanda DHS had 79 urban PSUs, 413 rural PSUs, with an average PSU population of 610 people. An equivalent sample in GridSample had 75 urban PSUs, 405 rural PSUs, and a median PSU population of 612 people. The number of PSUs differed because DHS added urban PSUs from specific districts while GridSample reallocated rural-to-urban PSUs across all districts.ConclusionsGridded population sampling is a promising alternative to typical census-based sampling when census data are moderately outdated or inaccurate. Four approaches to implementation have been tried: (1) using gridded PSU boundaries produced by GridSample, (2) manually segmenting gridded PSU using satellite imagery, (3) non-probability sampling (e.g. random-walk, “spin-the-pen”), and random sampling of households. Gridded population sampling is in its infancy, and further research is needed to assess the accuracy and feasibility of gridded population sampling. The GridSample R algorithm can be used to forward this research agenda.

Research on Grid Size Suitability of Gridded Population Distribution in Urban Area: A Case Study in Urban Area of Xuanzhou District, China.

The research on the grid size suitability is important to provide improvement in accuracies of gridded population distribution. It contributes to reveal the actual spatial distribution of population. However, currently little research has been done in this area. Many well-modeled gridded population dataset are basically built at a single grid scale. If the grid cell size is not appropriate, it will result in spatial information loss or data redundancy. Therefore, in order to capture the desired spatial variation of population within the area of interest, it is necessary to conduct research on grid size suitability. This study summarized three expressed levels to analyze grid size suitability, which include location expressed level, numeric information expressed level, and spatial relationship expressed level. This study elaborated the reasons for choosing the five indexes to explore expression suitability. These five indexes are consistency measure, shape index rate, standard deviation of population density, patches diversity index, and the average local variance. The suitable grid size was determined by constructing grid size-indicator value curves and suitable grid size scheme. Results revealed that the three expressed levels on 10m grid scale are satisfying. And the population distribution raster data with 10m grid size provide excellent accuracy without loss. The 10m grid size is recommended as the appropriate scale for generating a high-quality gridded population distribution in our study area. Based on this preliminary study, it indicates the five indexes are coordinated with each other and reasonable and effective to assess grid size suitability. We also suggest choosing these five indexes in three perspectives of expressed level to carry out the research on grid size suitability of gridded population distribution.

High-resolution African population projections from radiative forcing and socio-economic models, 2000 to 2100

For its fifth assessment report, the Intergovernmental Panel on Climate Change divided future scenario projections (2005–2100) into two groups: Socio-Economic Pathways (SSPs) and Representative Concentration Pathways (RCPs). Each SSP has country-level urban and rural population projections, while the RCPs are based on radiative forcing caused by greenhouse gases, aerosols and associated land-use change. In order for these projections to be applicable in earth system models, SSP and RCP population projections must be at the same spatial scale. Thus, a gridded population dataset that takes into account both RCP-based urban fractions and SSP-based population projection is needed. To support this need, an annual (2000–2100) high resolution (approximately 1km at the equator) gridded population dataset conforming to both RCPs (urban land use) and SSPs (population) country level scenario data were created.

Modelling changing population distributions: an example of the Kenyan Coast, 1979–2009

ABSTRACTLarge-scale gridded population datasets are usually produced for the year of input census data using a top-down approach and projected backward and forward in time using national growth rates. Such temporal projections do not include any subnational variation in population distribution trends and ignore changes in geographical covariates such as urban land cover changes. Improved predictions of population distribution changes over time require the use of a limited number of covariates that are time-invariant or temporally explicit. Here we make use of recently released multi-temporal high-resolution global settlement layers, historical census data and latest developments in population distribution modelling methods to reconstruct population distribution changes over 30 years across the Kenyan Coast. We explore the methodological challenges associated with the production of gridded population distribution time-series in data-scarce countries and show that trade-offs have to be found between spatial and temporal resolutions when selecting the best modelling approach. Strategies used to fill data gaps may vary according to the local context and the objective of the study. This work will hopefully serve as a benchmark for future developments of population distribution time-series that are increasingly required for population-at-risk estimations and spatial modelling in various fields.

Improving Large Area Population Mapping Using Geotweet Densities.

Many different methods are used to disaggregate census data and predict population densities to construct finer scale, gridded population data sets. These methods often involve a range of high resolution geospatial covariate datasets on aspects such as urban areas, infrastructure, land cover and topography; such covariates, however, are not directly indicative of the presence of people. Here we tested the potential of geo‐located tweets from the social media application, Twitter, as a covariate in the production of population maps. The density of geo‐located tweets in 1x1 km grid cells over a 2‐month period across Indonesia, a country with one of the highest Twitter usage rates in the world, was input as a covariate into a previously published random forests‐based census disaggregation method. Comparison of internal measures of accuracy and external assessments between models built with and without the geotweets showed that increases in population mapping accuracy could be obtained using the geotweet densities as a covariate layer. The work highlights the potential for such social media‐derived data in improving our understanding of population distributions and offers promise for more dynamic mapping with such data being continually produced and freely available.

Taking Advantage of the Improved Availability of Census Data: A First Look at the Gridded Population of the World, Version 4

Global gridded population data are increasingly being applied to a broad range of fields. This article presents improvements made in the fourth version of the Gridded Population of the World (GPWv4) data set. GPWv4 is a minimally modeled data set that uses uniform distribution to disaggregate census data from their native input units into a thirty arc-second global grid. Due to an increase in freely accessible census data from the 2010 round of censuses, eighty-seven countries are gridded at a higher administrative level than in the previous version, with a more than fivefold increase in the total number of census units outside of the United States. The resultant decrease in unit size has greatly improved the data set accuracy. Additionally, version 4 has expanded to include grids of age, sex, and urban–rural status. Obtaining high-resolution census data and matching census geography remains a priority for improving global gridded population data sets.

Disaggregating census data for population mapping using random forests with remotely-sensed and ancillary data.

High resolution, contemporary data on human population distributions are vital for measuring impacts of population growth, monitoring human-environment interactions and for planning and policy development. Many methods are used to disaggregate census data and predict population densities for finer scale, gridded population data sets. We present a new semi-automated dasymetric modeling approach that incorporates detailed census and ancillary data in a flexible, “Random Forest” estimation technique. We outline the combination of widely available, remotely-sensed and geospatial data that contribute to the modeled dasymetric weights and then use the Random Forest model to generate a gridded prediction of population density at ~100 m spatial resolution. This prediction layer is then used as the weighting surface to perform dasymetric redistribution of the census counts at a country level. As a case study we compare the new algorithm and its products for three countries (Vietnam, Cambodia, and Kenya) with other common gridded population data production methodologies. We discuss the advantages of the new method and increases over the accuracy and flexibility of those previous approaches. Finally, we outline how this algorithm will be extended to provide freely-available gridded population data sets for Africa, Asia and Latin America.

Exploring nationally and regionally defined models for large area population mapping

Interactions between humans, diseases, and the environment take place across a range of temporal and spatial scales, making accurate, contemporary data on human population distributions critical for a variety of disciplines. Methods for disaggregating census data to finer-scale, gridded population density estimates continue to be refined as computational power increases and more detailed census, input, and validation datasets become available. However, the availability of spatially detailed census data still varies widely by country. In this study, we develop quantitative guidelines for choosing regionally-parameterized census count disaggregation models over country-specific models. We examine underlying methodological considerations for improving gridded population datasets for countries with coarser scale census data by investigating regional versus country-specific models used to estimate density surfaces for redistributing census counts. Consideration is given to the spatial resolution of input census data using examples from East Africa and Southeast Asia. Results suggest that for many countries more accurate population maps can be produced by using regionally-parameterized models where more spatially refined data exists than that which is available for the focal country. This study highlights the advancement of statistical toolsets and considerations for underlying data used in generating widely used gridded population data.

Generation of fine-scale population layers using multi-resolution satellite imagery and geospatial data

A gridded population dataset was produced for Pakistan by developing an algorithm that distributed population either on the basis of per-pixel built-up area fraction or the per-pixel value of a weighted population likelihood layer. Per-pixel built-up area fraction was calculated using a classification and regression trees (CART) methodology integrating high- and medium-resolution satellite imagery. The likelihood layer was produced by weighting different geospatial layers according to their effect on the likelihood of population being found in the particular pixel. The geospatial layers integrated into the likelihood layer were: 1) proximity to remotely sensed built-up pixels, 2) density of settlement points in a fixed kernel, 3) slope, 4) elevation, and 5) heterogeneity of landcover types found within a search radius. The method for weighting these layers varied according to settlement patterns found in the provinces of Pakistan. Differences in zonal population estimates generated from the 100-meter gridded population layer resulting from this study, Oak Ridge National Laboratory's LandScan (2002), and CIESIN's Gridded Population of the World and Global Rural Urban Mapping Project (GPW and GRUMP) are examined. Population estimates for small areas produced using this paper's method were found to differ from census counts to a lesser degree than those produced using LandScan, GPW, or GRUMP. The root mean square error (RMSE) for small area population estimates for this method, LandScan, GPW, and GRUMP were 31,089, 48,001, 100,260, and 72,071, respectively.

Large-scale spatial population databases in infectious disease research

Modelling studies on the spatial distribution and spread of infectious diseases are becoming increasingly detailed and sophisticated, with global risk mapping and epidemic modelling studies now popular. Yet, in deriving populations at risk of disease estimates, these spatial models must rely on existing global and regional datasets on population distribution, which are often based on outdated and coarse resolution data. Moreover, a variety of different methods have been used to model population distribution at large spatial scales. In this review we describe the main global gridded population datasets that are freely available for health researchers and compare their construction methods, and highlight the uncertainties inherent in these population datasets. We review their application in past studies on disease risk and dynamics, and discuss how the choice of dataset can affect results. Moreover, we highlight how the lack of contemporary, detailed and reliable data on human population distribution in low income countries is proving a barrier to obtaining accurate large-scale estimates of population at risk and constructing reliable models of disease spread, and suggest research directions required to further reduce these barriers.

The effects of spatial population dataset choice on estimates of population at risk of disease

BackgroundThe spatial modeling of infectious disease distributions and dynamics is increasingly being undertaken for health services planning and disease control monitoring, implementation, and evaluation. Where risks are heterogeneous in space or dependent on person-to-person transmission, spatial data on human population distributions are required to estimate infectious disease risks, burdens, and dynamics. Several different modeled human population distribution datasets are available and widely used, but the disparities among them and the implications for enumerating disease burdens and populations at risk have not been considered systematically. Here, we quantify some of these effects using global estimates of populations at risk (PAR) of P. falciparum malaria as an example.MethodsThe recent construction of a global map of P. falciparum malaria endemicity enabled the testing of different gridded population datasets for providing estimates of PAR by endemicity class. The estimated population numbers within each class were calculated for each country using four different global gridded human population datasets: GRUMP (~1 km spatial resolution), LandScan (~1 km), UNEP Global Population Databases (~5 km), and GPW3 (~5 km). More detailed assessments of PAR variation and accuracy were conducted for three African countries where census data were available at a higher administrative-unit level than used by any of the four gridded population datasets.ResultsThe estimates of PAR based on the datasets varied by more than 10 million people for some countries, even accounting for the fact that estimates of population totals made by different agencies are used to correct national totals in these datasets and can vary by more than 5% for many low-income countries. In many cases, these variations in PAR estimates comprised more than 10% of the total national population. The detailed country-level assessments suggested that none of the datasets was consistently more accurate than the others in estimating PAR. The sizes of such differences among modeled human populations were related to variations in the methods, input resolution, and date of the census data underlying each dataset. Data quality varied from country to country within the spatial population datasets.ConclusionsDetailed, highly spatially resolved human population data are an essential resource for planning health service delivery for disease control, for the spatial modeling of epidemics, and for decision-making processes related to public health. However, our results highlight that for the low-income regions of the world where disease burden is greatest, existing datasets display substantial variations in estimated population distributions, resulting in uncertainty in disease assessments that utilize them. Increased efforts are required to gather contemporary and spatially detailed demographic data to reduce this uncertainty, particularly in Africa, and to develop population distribution modeling methods that match the rigor, sophistication, and ability to handle uncertainty of contemporary disease mapping and spread modeling. In the meantime, studies that utilize a particular spatial population dataset need to acknowledge the uncertainties inherent within them and consider how the methods and data that comprise each will affect conclusions.

Population detection profiles of DMSP-OLS night-time imagery by regions of the world

One emerging application of night-time light imagery focuses on estimating levels of access to electricity globally (Doll & Pachauri, 2010; Elvidge et al., 2010). A central consideration of such studies is the population density which can be consistently detected from night-time light imagery. Whilst numerous studies have addressed the relationship to light and population statistics in order to predict population, the use of spatially explicit population databases allows for a more detailed description of these relationships. This paper reports the variation of different detection profiles of two publically available gridded population datasets. These are disaggregated by region to reveal a vast contrast in what we may assume to be observable population in different parts of the world. A dynamic trend emerges with respect to levels of development with the most developed nations hypothesized to be the theoretical minimum observable population density. Beyond contributing to the analysis of areas of the world without access to electricity, more fundamentally, this analysis addresses a basic question about night-time lights and how it relates to population globally and in particular the relative merits of two commonly used population databases.

Assessing the use of global land cover data for guiding large area population distribution modelling

Gridded population distribution data are finding increasing use in a wide range of fields, including resource allocation, disease burden estimation and climate change impact assessment. Land cover information can be used in combination with detailed settlement extents to redistribute aggregated census counts to improve the accuracy of national-scale gridded population data. In East Africa, such analyses have been done using regional land cover data, thus restricting application of the approach to this region. If gridded population data are to be improved across Africa, an alternative, consistent and comparable source of land cover data is required. Here these analyses were repeated for Kenya using four continent-wide land cover datasets combined with detailed settlement extents and accuracies were assessed against detailed census data. The aim was to identify the large area land cover dataset that, combined with detailed settlement extents, produce the most accurate population distribution data. The effectiveness of the population distribution modelling procedures in the absence of high resolution census data was evaluated, as was the extrapolation ability of population densities between different regions. Results showed that the use of the GlobCover dataset refined with detailed settlement extents provided significantly more accurate gridded population data compared to the use of refined AVHRR-derived, MODIS-derived and GLC2000 land cover datasets. This study supports the hypothesis that land cover information is important for improving population distribution model accuracies, particularly in countries where only coarse resolution census data are available. Obtaining high resolution census data must however remain the priority. With its higher spatial resolution and its more recent data acquisition, the GlobCover dataset was found as the most valuable resource to use in combination with detailed settlement extents for the production of gridded population datasets across large areas.Electronic supplementary materialThe online version of this article (doi:10.1007/s10708-010-9364-8) contains supplementary material, which is available to authorized users.

A high resolution spatial population database of Somalia for disease risk mapping

BackgroundMillions of Somali have been deprived of basic health services due to the unstable political situation of their country. Attempts are being made to reconstruct the health sector, in particular to estimate the extent of infectious disease burden. However, any approach that requires the use of modelled disease rates requires reasonable information on population distribution. In a low-income country such as Somalia, population data are lacking, are of poor quality, or become outdated rapidly. Modelling methods are therefore needed for the production of contemporary and spatially detailed population data.ResultsHere land cover information derived from satellite imagery and existing settlement point datasets were used for the spatial reallocation of populations within census units. We used simple and semi-automated methods that can be implemented with free image processing software to produce an easily updatable gridded population dataset at 100 × 100 meters spatial resolution. The 2010 population dataset was matched to administrative population totals projected by the UN. Comparison tests between the new dataset and existing population datasets revealed important differences in population size distributions, and in population at risk of malaria estimates. These differences are particularly important in more densely populated areas and strongly depend on the settlement data used in the modelling approach.ConclusionsThe results show that it is possible to produce detailed, contemporary and easily updatable settlement and population distribution datasets of Somalia using existing data. The 2010 population dataset produced is freely available as a product of the AfriPop Project and can be downloaded from: http://www.afripop.org.