GI for Sustainability

GI & Sustainability II

5608.2 - Assessing the Dynamics of Resilience using a Bayesian Network in the Lower Mississippi River Basin

Wednesday, July 5
3:10 PM - 3:30 PM
Location: Maryland C

Introduction: Substantial effort has been made to identify metrics and variables to measure community resilience. However, how these various variables interact as a system that may affect the final resiliency, in other words, their dynamical linkages, have rarely been studied. Understanding the complex nature-human system dynamics, the cause-and-effect relationships are crucial for pre-disaster preparation, post-disaster recovery, and establishment of mitigation plans. It is difficult for classical statistical tools to model systemic changes of all the components involved in a natural-human system. Bayesian network, which represents interdependencies among variables in a graph in a probabilistic manner, could be a powerful tool for representing and quantifying the causal relationships among resilience factors while expressing the model uncertainty in the form of probability distribution.
Objectives: This research aimed to develop a Bayesian network model to understand and represent the relationships between key resilience-related nature and human components, and then use the model to simulate the systemic changes in the Lower Mississippi River Basin under different scenarios. The target variable to be modeled was population change, which has been commonly used to represent the resilience of a coupled natural-human system in the disaster resilience literature. The study was conducted at the census block group scale, with a total of 2,086 block groups covering the study area.
Methods: Firstly, a comprehensive set of resilience-related indicators based on a thorough literature review and data availability was prepared. This step resulted in 36 variables, including variables on exposure to coastal hazards, property damage sustained, socioeconomic capacity, and environmental conditions. 11 statistically significant variables were selected through preliminary statistical analysis. Secondly, a Bayesian Network was developed to illustrate the relationships between population change and the resilience variables. Since nature-human systems are complex where the structural relationships among different components are often difficult to decipher, the genetic algorithm approach was used to optimize the structure of the Bayesian network. Thirdly, the influence of each factor on population change was explored through belief updating. Then, the spatial population change patterns under two types of scenarios – varying the exposure to hazards and community characteristics – were simulated by computing the posterior marginal distribution of population change.
The above steps yielded an optimized Bayesian network model with the highest fitness value (cross-validation accuracy) of 66.9% through a 906-generation genetic algorithm. The scenario simulation resulted in maps that reveal which parts of the study area could withstand higher exposure to disasters by not suffering population decline, as well as how the spatial patterns of population vary if certain strategies were applied (e.g. raising employment rate and elevation). In addition to contributing to the resilience dynamics literature, this study demonstrates the need for spatial analytical techniques in addressing important societal problems.

Heng Cai

Ph.D Candidate
Louisiana State University

Heng Cai is currently a Ph.D candidate in Environmental Sciences, Louisiana State University. Her research interests include spatial statistics and modelling, community resilience measurement, nature-human system dynamics modelling. Heng received B.E. degree in 2010 from China University of Geosciences in Beijing, China and M.S degree in 2013 from the Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing, China.

Presentation(s):

Send Email for Heng Cai

Daniel M. Lopez Lopez

Master Degree
Centro de Investigacion en Geografia y Geomatica Ing. J. L. Tamayo A. C.

Agrologo (Soil scientist), Magister of Science, International Institute for Aerospace Survey and Earth Sciences, 1992, Enschede, The Netherlands. Magister in Geomatic, CentroGeo, 2006, México. Lecturer on Landuse planning and Land evaluation - Geographic Institute Agustin Codazzi, Bogotá, Colombia (1989-2000). Researcher CentroGeo (2000-present). Research fields: Process and analysis of DEM for analyses and modelling environmental problems, analysis and assessment of vulnerability -Resilience to global environmental change.
Projects: Digital image processing for land use-land cover and geomorphologic cartography. Spatial analysis and modeling for land use planning and Vulnerability-Resilience to Global environmental change.

Presentation(s):

Send Email for Daniel Lopez Lopez


Assets

5608.2 - Assessing the Dynamics of Resilience using a Bayesian Network in the Lower Mississippi River Basin



Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Assessing the Dynamics of Resilience using a Bayesian Network in the Lower Mississippi River Basin