Using geographic information system (GIS) and remote sensing techniques for estimating populations for polygons overlaid on United States census geography is common practice. These methods are used by many groups, both governmental and private, to estimate the number and characteristics of people living in specific target areas based on a known value for the source area. This information is desired for such diverse reasons as estimating the population affected by natural and human disasters to comparing data for noncompatible zones over some period of time. However, the accuracy of population estimates using these procedures has not been determined with any degree of certainty. The purpose of this research was to test the accuracy of various methodologies of population estimation using GIS and remote sensing techniques.
Population estimates were made for selected census blocks that were partially inundated during the flooding of the Missouri River in August 1993. Eight different methods were used to estimate population for each of the twenty-five partially flooded blocks selected as study sites. The selected block groups for the study are located in the central Missouri towns of Franklin, New Franklin, Rocheport, and Hartsburg. The true flooded population for each block was determined by door-to-door population counts. People living in the selected census blocks were questioned to verify if they or their neighbors were flooded during 1993. The results of each estimation methodology are compared to the "true" counts to determine accuracy. The results of this testing led to a quantified ranking of each method's accuracy in predicting the true value for a given areal unit. The utility and appropriateness of using specific methodologies, as well as the possibility of assigning error values for a given method are discussed.
Steven V. Muller
Geographic Resources Center
Dept. of Geography UMC
17 Stewart Hall
Columbia, MO 65211
Telephone: 314-882-1404
Fax: 314-884-4239
Email: c568053@mizzoul.missouri.edu