| With human population growing by over 80 million a year, it has been projected that within the next 50 years, the 10 billion mark will be reached. Most of this growth is expected to be concentrated in primarily urban areas in low income countries. Rapid population growth has been well documented to impact economies, environment and health of nations, which are all expected to undergo significant change. To measure impacts of population growth with high accuracy, high resolution, and contemporary data on human population distributions as well as their compositions are necessary for planning interventions and monitoring changes.;Disease burden estimation, epidemic modeling, resource allocation, disaster management, accessibility modeling, transport and city planning, poverty mapping, and environmental impact assessment have integrated spatial databases of human population. Low income regions of the world often lack relevant data or the data are of poor quality, whereas in high-income countries, extensive mapping resources and expertise are at their disposal to create such databases. The major obstacles to doing settlement and population mapping across the low income regions of the World include the scarcity of mapping resources, lack of reliable validation data and the difficulty in obtaining high resolution contemporary census statistics. Focusing in on the open-source WorldPop Project and its associated methods, within the WorldPop Project a range of open geospatial datasets are combined in a flexible regression tree framework to reallocate contemporary aggregated spatial population count data. The resultant maps, backed by statistical assessments, suggest that the resultant maps are consistently more accurate than existing population map products, as well as the simple gridding of census data. The Project's 100m spatial resolution is a finer mapping detail than has even been produced at national extents, and as the data can be integrated with household survey, microdata, satellite and other data sources, this enables the production of more diverse datasets. Population count estimates can now encompass age structures, births, pregnancies, poverty and urban growth.;The aim of this dissertation is to provide a critical eye on how the open-source population mapping pioneered by the WorldPop project can be improve to directly indicate the presence of people within its datasets more accurately. The first section is an experiment with a tool called Google Earth Engine that can rapidly analyze vast amounts of satellite imagery to extract remotely sensed data, in this case applying a Normalized Difference Spectral Vector calculation over Landsat imagery to improve the temporal resolution of the population datasets. The second section is an experiment utilizing volunteered geographic data in the form of Twitter data that is geo-located, providing a layer of information that is human volunteered as a covariate in the population mapping process. The final section is a discussion of future work in mapping population at a continuous 30 meters, as opposed to 100 meters to examine the limitations of the co-variate datasets, as well as exploring the potential future implementation of Unmanned Aerial Vehicles to validate remote sensing classifications. |
