The dynamics of coupled natural and human systems are complex and vary across time, location, and organizational unit. So different social groups may be unequally affected by degraded environments, which in many cases do not randomly occur over space. Therefore, the concept of environmental justice was introduced to address this issue. One environmental resource that has been a focus point for environmental justice is freshwater, and stream health based environmental justice studies are the most recent approach used by researchers to describe these phenomena. However, many of the previous studies were only performed at the census tract level and no proper spatial level was defined for environmental justice studies with respect to stream health. On the other hand, due to computational limitations for stream health indices estimation, only a few water quantity and quality parameters were used to develop stream health predictive models. Therefore, the purpose of this study is to address the following knowledge gaps in the area of stream health based environmental justice by: 1) determining the role of spatial level of socioeconomic factors on stream health based environmental justice studies and 2) assessing the relative importance of parameter estimation in stream health based environmental justice modeling.To address the first knowledge gap, three Bayesian Conditional Autoregressive (CAR) models (ordinary regression, weighted regression and spatial) were developed for four common stream health measures based on 17 socioeconomic and physiographical variables at three census levels of county, census tract, and block group in the Saginaw River Basin in Michigan. This watershed was an ideal place to perform this study since it was identified as an area of concern in the Great Lakes region while having one of the most diverse populations in the state. For all stream health measures, spatial models had better performance compared to the two non-spatial models at the census tract and block group levels. In addition, multilevel Bayesian CAR models were also developed to understand the spatial dependency across three levels. Results showed that considering level interactions improved the predictive power of the environmental justice models. Residual plots also showed that models developed at the block group and census tract (in contrary to county level models) were able to capture spatial variations, which is an important aspect of environmental justice studies.To address the second knowledge gap, first ecologically relevant streamflow and water quality indices were used to improve the performance of the stream health predictive models. The outputs (fish and macroinvertebrate indices) from newly developed stream predictive models were then used to develop similar CAR models. Results showed that incorporating the more accurate stream health indices improved the spatial dependencies at the census tract and block group levels compared to county level. In addition, the multilevel models had better performance than single level models. Finally, the modified stream health indices improved stream health based environmental justice models’ performance by reducing redundancies in independent variables. This research finding provides a valuable tool to target vulnerable communities with respect to access to clean water, which enables water resources policymakers to allocate resources in a way that reduces environmental inequality.
Read
