Identifying county-level regional interdependence in post-disaster recovery across three U.S. hurricanes in 2017

Regional recovery interdependence is crucial for mitigating the impacts of frequent natural disasters during the post-disaster recovery period. However, this interdependence has been understudied due to the complexity involved in identifying the impacts of inter-regional recovery. Causal discovery techniques allow researchers to infer regional interdependence from time-series datasets. This study utilizes daily points-of-interest data from the 2017 three consecutive hurricanes in the U.S. to examine regional interdependence using structural agnostic modeling, a causal discovery method, without adjusting hyperparameters of the model. The results revealed four patterns related to county and types of relationships: (1) direct recovery impacts, (2) recovery interdependence from urban to suburban and rural counties, (3) exceptions of recovery interdependence due to hurricane trajectories, and (4) the distance as an insignificant factor. These findings demonstrate the transferability of this technique in the 2017 three hurricanes in the U.S. and encourage researchers to develop inter-regional post-disaster recovery simulations and counterfactual scenarios to enhance post-disaster recovery. • Apply a causal discovery technique to daily visits to the points-of-interest dataset in three hurricane cases to reveal the regional recovery interdependence. • Reveal the recovery interdependence from urban to suburban and rural counties by major data-driven human mobility patterns. • Helps policymakers set up county-level sequential post-disaster recovery plans.