Predicting Freight Attraction with Multivariate Linear Regression and Geographically Weighted Regression using satellite Nighttime Light data

Predicting freight transportation is crucial since it is often likened to the foundation of society and a pivotal component of its progress. When access to freight data is limited in underdeveloped nations, nighttime light data could serve as a reliable proxy for assessing freight activity. This research aims to assess the reliability of Visible Infrared Imaging Radiometer Suite (VIIRS) nighttime light imagery data as an indicator of freight activity, utilizing Iran's county-level road freight transit database. The study incorporates Population (POP), Average Annual Household Income (AI), and Nighttime Light (NL) as independent variables, while the quantity of annual road freight attraction (FA) in each zone serves as the dependent variable. Two techniques, Geographically Weighted Regression (GWR) and Multivariate Linear Regression (MLR), were employed in this study. Compared to the MLR model, the GWR model's R-squared value increased from 0.68 to 0.79, indicating an enhanced model fit. The "F-test" demonstrated that the descriptive contribution of the nighttime light variable was more significant than that of other factors. The results of this study are significant for researchers and policymakers, as forecasting freight plays a crucial role in anticipating future freight traffic demands and effectively distributing transportation resources.