Emerging economic geography of urban restaurants as freight generators: Logistics policy implications for managing dark kitchens and food trucks

The rapid rise of app-based food delivery platforms has redefined how restaurants shape urban space. However, little is known about how these evolving restaurant types cluster and interact with urban land use. Using spatial analysis involving Ripley's-K and Moran's-I and predictive models involving decision trees, random forest, and multinomial logit models, this study attempts to explain the location choices of restaurants based on their relative distance to the city centre, rent, population density, and night-time light (NTL) intensity. Analysis results reveal that dark kitchens exhibit the tightest clustering, often in low-rent, high-density zones, while in-person dining is concentrated in high-rent, high-NTL areas. Among the models tested, random forest outperformed decision trees and multinomial logit models in predicting restaurant types, with night-time light emerging as the strongest spatial predictor. The clustering patterns observed in emerging urban restaurant types differ significantly from traditional brick-and-mortar establishments; study findings highlight the urgent need for adaptive freight planning and zoning policies to address the growing logistical footprint of digitally mediated food establishments. While based in Indian cities, the framework and insights of this study are transferable to other global contexts where on-demand food delivery and mixed-use zoning intersect in urban areas.