• Home
  • » About our work
  • » Projects
  • Pop-RF


    Census Disaggregation for Population Mapping


    Contemporary, fine spatial scale, gridded population data is critical across myriad research contexts. We use such data for understanding the ‘now’ at the intersection of hazard risk and mitigation management, health and disease modelling, and economic-, environmental-, and sustainability-related research in relation to population distributions. But to understand how human population shifts through time, and incorporate a more nuanced approach to modelling population change at fine spatial scales, novel ways of treating data and statistically describing associations of measured population counts with associated covariates are needed.



    Fig.1. Processing pipeline for population mapping

    image

    The WorldPop project, now in its twelfth year, is an effort to improve the “denominator,” or the population background against which persistence, transmission and eradication models for various diseases are developed. The high demand for sub-census level estimates of population distribution continues to increase, and with that demand comes a desire for better and more useful data for analyzing change over time. Our research

    develops leading-edge modelling methods, combining machine learning, cloud-based computing, and the best available census and ancillary data. These data and methods are used to produce 100 meter, gridded population estimates at five year intervals across the tropical and subtropical areas of Latin America, Asia, and Africa.


    New efforts supported by funders will include expanding these methods to support annual estimates with global extents. These new ensemble modelling approaches incorporate changing built area environments, multiple years of census data, and multiscalar syntheses that merge data to produce comparable gridded population data across time.

    image

    Fig. 2. Disaggregation of administrative unit based population counts (left) to a 100x100m grid representation (right).


    Contact people:

    Andy Tatem

    WorldPop, Geography and Environment University of Southampton A.J.Tatem@soton.ac.uk

    (023) 8059 2636


    Forrest Stevens

    WorldPop, University of Louisville Louisville, KY

    USA

    forrest@forreststevens.com



    Collaborators/Funders

    Bill and Melinda Gates Foundation, Columbia University, Joint Research Centre of the EC


    References

    Gaughan, A. E., Stevens, F. R., Linard, C., Jia, P., & Tatem, A. J. (2013). High Resolution Population Distribution Maps for Southeast Asia in 2010 and 2015. PLOS ONE, 8(2), e55882. doi:10.1371/journal.pone.0055882

    Gaughan, A. E., Stevens, F. R., Linard, C., Patel, N. N., & Tatem, A. J. (2014). Exploring nationally and regionally defined models for large area population mapping. International Journal of Digital Earth, (October), 1–18. doi:10.1080/17538947.2014.965761

    Sorichetta, A., Hornby, G. M., Stevens, F. R., Gaughan, A. E., Linard, C., & Tatem, A. J. (2015). High-resolution gridded population datasets for Latin America and the Caribbean in 2010, 2015, and 2020. Scientific Data, 2(150045). doi:10.1038/sdata.2015.45

    Stevens, F. R. (2015). Random Forest Population Mapping Complexity Reduction Algorithm, Data and Code. doi:doi:10.6084/m9.figshare.1494648

    Stevens, F. R., Gaughan, A. E., Linard, C., & Tatem, A. J. (2015). Disaggregating Census Data for Population Mapping Using Random Forests with Remotely-Sensed and Ancillary Data. PLOS ONE, 10(2), e0107042. doi:10.1371/journal.pone.0107042