Identification of Causalities in Spatio-temporal Data

TL;DR

This paper presents a generic Granger causality-based method for identifying causal relationships in complex spatio-temporal data, validated on urban models and applied to real-world transportation and socio-economic data in Greater Paris.

Contribution

It introduces a novel, robust approach to detect causality regimes in coupled spatio-temporal processes, with validation on urban morphogenesis models and real case studies.

Findings

Causality regimes identified in urban growth models

Link established between territorial dynamics and network development

Method validated for real-world transportation data

Abstract

This paper contributes to the understanding of strongly coupled spatio-temporal processes by describing a generic method based on Granger causality. The method is validated by the robust identification of causality regimes and of their phase diagram for an urban morphogenesis model that couples network growth with density. The application to the real case study of Greater Paris transportation projects shows a link between territorial dynamics, more particularly of real estate and socio-economic, and the anticipated network growth. We finally discuss potential extensions to other temporal and spatial scales.