Estimating small-area population density in Sri Lanka using surveys and Geo-spatial data

Area of Science:

• Demography
• Geospatial Analysis
• Population Studies

Background:

• Traditional census data, collected decennially, quickly becomes outdated due to factors like conflict, migration, and urbanization.
• Rapid population shifts at local levels challenge the accuracy of infrequent census data.
• Existing methods often rely on aging census data for population redistribution.

Purpose of the Study:

• To demonstrate a feasible "bottom-up" methodology for estimating local population density in intercensal periods.
• To develop a more accurate and timely alternative to traditional census-based population estimates.
• To provide a cost-effective approach for tracking population dynamics between official census collections.

Main Methods:

• Utilized a "bottom-up" approach combining household surveys with contemporaneous geospatial data (village area, satellite imagery).
• Applied Poisson regression models with variable selection via the Least Absolute Shrinkage and Selection Operator (LASSO).
• Estimated models on surveyed villages and generated out-of-sample density estimates for non-surveyed areas in Sri Lanka.

Main Results:

• The developed method accurately approximates census population density.
• Estimates derived from this method are more precise than those from other bottom-up geospatial studies.
• The approach effectively circumvents the issue of aging census data by incorporating up-to-date survey information.

Conclusions:

• The "bottom-up" method combining household surveys and geospatial data is a feasible and effective way to estimate local population density.
• This technique offers a more accurate and timely alternative to traditional census data, especially in rapidly changing environments.
• The proposed method is cost-effective for frequent monitoring of local population density between census years.