Identifying a positive correlation between 2015 VIIRS NTL data and GDP data using sub-categories of GDP in Alabama, Georgia, Florida, North Carolina, South Carolina, and Tennessee MSAs

Authors: Adam Kohl*, University of West Florida, Zhiyong Hu, Mentor
Topics: Spatial Analysis & Modeling, Remote Sensing, Economic Geography
Keywords: Nighttime lights, gross domestic product, metropolitan statistical areas, regression modeling, remote sensing,night,time,economics, statistics,
Session Type: Poster
Day: 4/5/2019
Start / End Time: 8:00 AM / 9:40 AM
Room: Lincoln 2, Marriott, Exhibition Level
Presentation File: No File Uploaded

Remote sensing offers a more cost-efficient method of estimating Gross Domestic Product than traditional census methods and allows researchers to estimate GDP in areas unmeasured by censuses. The most common method of estimating GDP through remote sensing is by analyzing the relationship between nighttime lights and GDP. Alternatively, this study explored the correlation between sub-categories of GDP and the sum of NTLs in Alabama, Georgia, Florida, Tennessee, North Carolina, and South Carolina metropolitan statistical areas for the year 2015. The objective of this study was to identify a regression model that held up across different scales and could be used in future studies to estimate the GDP of specific industries using NTL data. To achieve this, two scales were analyzed, one using the entire study area and one that splits the study area into three sub-areas. These sub-areas were the MSAs of Alabama and Georgia combined, South Carolina, North Carolina, and Tennessee combined, and Florida on its own. The exploratory regression tool within ArcMap was used to identify a common model between these two scales. One multi-variable model was identified that positively correlated the combined values of three GDP categories, farms, wood product manufacturing, and federal civilian to the sum of NTLs. The multi-variable model was evaluated using GWR and OLS regression and assessed for additional bias using Moran’s I. The model exhibited high R-squared values at both scales but was limited by the suppression of wood product manufacturing data in the South Carolina, North Carolina, and Tennessee sub-area.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login