Date of Award
Doctor of Philosophy
Mark D. Schwartz, Woonsup Choi, Christopher De Sousa, Le Wang
Population Estimation, Remote Sensing, Urban Environment
Knowledge of detailed and accurate population information is essential to analyze and address a wide variety of socio-economic, political, and environmental issues and to support necessary planning practices for both public agencies and the private sector. However, such important data are generally only available once every decade through the National Census. Moreover, populations in some rapidly-developing areas may increase quickly, such that this ten-year frequency does not meet the needs of these areas. Therefore, a cost-effective method for population estimation is necessary. To address this issue, this research integrated geographic, sociological, and demographic theories and exploited remotely sensed imagery and geographic information system (GIS) datasets to derive better population estimates at the census block level, the finest level of the national census.
Specifically, three new approaches have been proposed in this dissertation to assist in the improvement of small-area population estimation accuracy. First, existing remotely sensed and GIS data have been adopted to estimate two major components of a demographic framework, including the redistribution of newly built dwelling units from the aggregated geographic level to the census block level and the estimation of persons per household (PPH) at such a fine scale. Second, in addition to the use of existing data, new urban environmental indicators were also extracted and employed to improve population estimation. In particular, to implement the automatic enumeration for individual housing units, a new spectral index, biophysical composition index (BCI), has been proposed to derive impervious surface information, a desirable urban environmental parameter. Third, using the extracted high-resolution urban environmental information and GIS data, a new bottom-up method was developed for small-area population estimation at the census block level by incorporating these high-resolution data into the demographic framework.
Analyses of the results suggest three major conclusions. First, existing GIS spatial factors, together with demographic information, can assist in improving the accuracy of small-area population estimation. Second, the BCI has a closer relationship with impervious surface area than do other popular indices. Moreover, it was shown to be the most effective index of the four evaluated for separating impervious surfaces and bare soil, which consequently might assist in more accurately deriving fractional land cover values. Third, the use of the new environmental indicators extracted from remote sensing imagery and GIS data and the integration of demographic and geographic approaches has significantly improved the estimation accuracy of housing unit (HU) numbers, PPH, and population counts at the census block level.
Therefore, this research contributes to both the remote sensing and applied demography fields. The contribution to the remote sensing field lies in the development of a novel spectral index to characterize urban land for monitoring and analyzing urban environments. This index provided more significant separability between impervious surfaces and bare soil than did other existing indices. Moreover, three major contributions have been made in the field of applied demography: 1) the generation of accurate HU estimates using high-resolution remote sensing and GIS datasets, 2) the development of a model to derive an accurate PPH estimate, and 3) the improvement of small-area population estimation accuracy through the integration of geographic and demographic approaches.
Deng, Chengbin, "Small-Area Population Estimation: an Integration of Demographic and Geographic Techniques" (2013). Theses and Dissertations. 678.