Degree

Doctor of Philosophy (PhD)

Department

Environmental Sciences

Document Type

Dissertation

Abstract

Drought effects cross-cut meteorological, agricultural, hydrological, and socioeconomic systems and affect both natural and human communities. The drought hazard varies in time and space in its intensity and impact on human communities. An identification of the main variables that affect resilience is crucial to coping with the hazard and promoting resilience. The study applied the Resilience Inference Measurement (RIM) framework to measure the resilience levels of the 503 counties of Arkansas, Louisiana, New Mexico, Oklahoma, and Texas. In the first assessment, through k-means cluster analysis, stepwise discriminant analysis (74.7 percent accuracy, 72.8 percent leave-one-out cross-validation accuracy), and regression analysis (adjusted R2 = 0.69), four variables (significant at p < 0.05) representing the social, economic, agriculture, and health sectors were identified as the main resilience indicators. In search of the new insights into the dynamics of drought resilience, this research further assessed the temporal changes of community resilience. The counties with more affluent socioeconomic conditions and more diverse agriculture were expected to improve their resilience, while counties with poorer socioeconomic conditions and heavy reliance on agriculture were hypothesized to decrease their resilience over time, widening the temporal and regional disparity. Using shrinkage discriminant analysis, and the correlation-adjusted t-scores, 10 variables were selected as predictors (67.9% classification accuracy). The derived discriminant functions were then used to estimate the resilience levels in 2000, 2005, 2010, and 2015. The results support the hypothesis and suggest a widening gap in resilience levels among counties. To increase our understanding of the complex process underlying communities’ response to the drought impacts, a Bayesian network was built using the 10 predictors selected in the shrinkage discriminant analysis model. The resultant network identified the key links between hazard intensity, damage, and recovery dimensions of resilience and socioeconomic and food-energy-water resilience predictors. Through an evidence propagation procedure and a set of conditional probability queries, the study identified that continuing increase in drought incidence is likely to lead to higher agricultural damage, but it is unlikely to stop population growth. The findings from this study provide a useful decision-support tool for stakeholders and practitioners in the communities affected by drought.

Date

10-23-2018

Committee Chair

Lam, Nina

Available for download on Wednesday, October 30, 2019

Share

COinS