Spatial-Temporal Distribution of Population At-Risk: Comparison of Gridded Population Data, Census Data and VIIRS Data

Wednesday, 17 December 2014
Bandana Kar, University of Southern Mississippi, Stennis Space Center, MS, United States and T. Edwin Chow, Texas State University, Geography, San Marcos, TX, United States
The U.S. coastal communities today face a higher risk of experiencing economic and social disasters than before due to their exposure to an increasing number of coastal hazards and also, because they have experienced a dramatic increase in population and property values over time. Given the continuous increase of coastal population, identifying at-risk communities will help mitigate coastal hazard impacts and prepare for future coastal hazard events. This study fulfills this goal by determining the effectiveness of two global data sets Landscan (gridded population data) and NOAA’s Night Time Imagery from the Visible Infrared Imaging Radiometer Suite (VIIRS), and the U.S. census data at the census block level that are used for demographic analysis and monitoring population dynamics in the U.S. For this study, the global data sets for the years 2011 and 2012, the census data for 2010 were used in conjunction with parcel boundary data (finest resolution data that can be used as a proxy for population data) available annually for Florida.

Through comparative analysis, the following research questions were explored: (1) which data set can accurately determine population density and its spatial distribution? (2) how much error is associated with each data set and what is the spatial distribution of error? The study is conducted for the coastal counties of Florida. A preliminary analysis using parcel level data as reference data set revealed that almost 25% of population was lost by both Landscan and VIIRS data sets in comparison to parcel data. Further analysis will be conducted to determine the spatial distribution of error, to identify the correlation between data sets, and identify the data set that is less erroneous in estimating population and its spatial distribution.