Assessing the Spatial Transferability of Freight (Trip) Generation Models across and within States of India: Empirical Evidence and Implications for Benefit Transfer

Abstract

This paper analyses and compares transferability of freight production (FP) and freight trip production (FTP) models to provide guidance on when, when not, and how to transfer models from one spatial context to another. Separate sets of model parameters are estimated for various industry sectors in two regions within Kerala, a Southern State of India and are transferred to Jaipur, a Northern State of India. Model performance in five different transfer combinations are assessed to understand the degree of transferability of FP and FTP models which included intrastate transfers within Kerala and interstate transfers to Jaipur. Two approaches of updating model parameters using locally collected data are evaluated – Bayesian updating and combined transfer estimation – with metrics that measure transferability improvement. The relative performance of transferred models revealed that the degree of transferability of freight models vary widely across industry sectors and depends on: (i) measure of business size and the (ii) metric of measuring freight activity. That is, employment-based models show better transferability than area-based models and FP models are more transferable than FTP models. The transferability results also suggested that the interstate transferability is higher than interstate transferability. Overall, the study findings will assist planning agencies to: (i) identify the locational characteristics that restrict the transferability of freight models, (ii) develop modelling strategies that focus on the appropriate metric of measuring freight and business size indicators, and (iii) reduce costs and resources in regions where there is lack of institutional capacity to develop freight demand model systems.