Presentation on theme: "Population Modeling Methods and Projects: Implications for Data Use Greg Yetman CIESIN, Columbia University GEO Meningitis."— Presentation transcript:
Population Modeling Methods and Projects: Implications for Data Use Greg Yetman CIESIN, Columbia University firstname.lastname@example.org GEO Meningitis Environmental Risk Consultative Meeting, September, 2007 World Data Center for Human Interactions in the Environment
Outline Population Surfaces: Two Cases Population Modeling Methods and Projects Implications of Methods and Inputs for Data Use Derivative Data Products
Population Surfaces: Census and Ambient Populations Where census is the only input or considered truth, population model shows where people live. (GPW, GRUMP, Accessibility Model) Ambient populations include day and night population locations (highways, airports, commercial/industrial land use)
Simple Allocation Proportional distribution of population across grid cells in administrative units Simple masking (water bodies, permanent ice) to remove uninhabited areas GPW data product uses simple allocation
Gridded Population of the World (GPW) Version (pub)GPW v1 (1995)GPW v2 (2000)GPW v3 (2003) Estimates for19941990, 19951990, 1995, 2000 Input units19,000127,000 400,000
Accessibility Modeling Populated places (points) and roads used as population centers Optionally, city stable lights (satellite- derived) used to add detail to accessibility surface Administrative unit population reallocated within units using surface UNEP, CIESIN Population surfaces
Weighted Interpolation Multiple input layers used to interpolate population surface. Landscan from Oak Ridge National Laboratory reallocates administrative population esimates using: –Lights at Night –Elevation/Slope –Distance to Roads –Distance to Rivers –Land Cover
Global Rural Urban Mapping Project (GRUMP) Points Polygons Gridded surface 1 km Hybrid approach that uses administrative data as truth and reallocates based on city lights and points.
Detailed Urban Extents from Satellite Data Automated and semi-automated extraction of urban extents from moderate or high- resolution imagery (Landsat, SPOT, Quickbird, Ikonos) or non-visible data (RADAR) Oxford University malaria mapping project is using Landsat and Radarsat data to define detailed urban extents in parts of Africa
Implications of Methods and Inputs for Data Use Variable dependence –Know the population model inputs to avoid endogeneity Night vs. day populations –Estimating day vs. night populations with existing population surfaces is problematic Variable spatial inputs for most projects create issues in inter-regional comparison Where practical, running the analysis with multiple population surfaces will provide an assessment of the variability (range) of population estimates.
Population Migration Little systematic data on global or regional basis –Landscan re-modeling by Oak Ridge based on newspaper reports in war-torn countries Potential Approaches: –Update administrative boundaries based on survey data or reports and produce model output(s) appropriate for application –Re-allocate population surfaces using weighted interpolation based on population movement estimates –Model population as a network, similar to hydrologic modeling with sources and sinks