Illustrating the nonlinear effects of urban form factors on transportation carbon emissions based on gradient boosting decision trees.

While prior studies have examined the effects of urban form on transportation carbon emissions, the exploration of nonlinear influences remains limited. This study presents an approach that transcends simple quantification of urban form's impact on transportation carbon emissions by also identifying the threshold range over which urban form variables exert their influence. Using 282 Chinese prefecture-level cities as the sample, this study employs gradient boosting decision trees to identify the nonlinear effects and the relative importance of urban form factors on transportation carbon emissions. We find that urban form factors jointly account for 31.32 % of the predictive power in estimating transportation carbon emissions after controlling for transport facilities, socioeconomic, and demographic factors. Urban polycentricity and transportation carbon emissions generally exhibit an obvious and complex nonlinear relationship. In addition, polycentricity, urban dispersion, the number of (sub)centers, and population density all have clear threshold effects on transportation carbon emissions. We further identified their effective ranges to guide urban development and land use planning.