Identifying statistical trends for environmental quality based on archival convenience databases.

摘要

Modern society greatly impacts the environment in complex, threatening, and not completely understood ways and the need for monitoring and assessment of environmental phenomena has become critical. The major objective of this work is to conduct two case studies in using convenience databases with large spatial or temporal extents to inform environmental assessment, in order to identify guidelines for how such databases can be appropriately used in environmental evaluations.; The first case study addresses correlations between environmental quality and diabetes with state-wide databases. Statistical techniques clarified intriguing correlations between diabetes and air pollution emissions. Calculation of the correlation of diabetes with Toxic Release Inventory (TRI) emissions confirmed the significant association found by previous research. In contrast, a multivariate regression found that state-wide diabetes rates are not significantly associated with TRI emissions, indicating that the bivariate correlation between diabetes and air pollution resulted from confounding.; In the second case study, statistical methods are applied to routine ground water monitoring data. Correlations of arsenic and several anionic constituents in the Hueco Basin are positive and significant (p-values between 0.001 and 0.05), suggesting that competitive desorption from hydroxide solids plays an important role in mobilizing arsenic. To augment the archival information, two cost-effective cuttings experiments were designed and performed to test the local origin and desorption mechanism suggested by the water analyses as factors associated with arsenic contamination in water in the region. Fifteen well cuttings were analyzed for arsenic, iron and total organic carbon Also the cuttings were leached in pH 9 and 10 solutions and the leachates were analyzed for dissolved arsenic. The important role of solid-phase iron in controlling the dissolved arsenic concentration was strongly supported. Significant associations between dissolved arsenic and solid-phase iron (R square 0.71 p-value 0.01), and significant associations of arsenic leached from the cuttings with dissolved and solid arsenic (R square 0.62 and 0.66, p-value 0.05 and 0.01) were found. The correlation between arsenic leached and solid iron was positive but not significant. These cuttings results are consistent with competitive desorption of arsenic which was suggested by the statistical analyses of the archival data.; These case studies show both the promise and limitations of archival, convenience databases. The first case study demonstrates that statewide databases may be too highly aggregated to use for associating environmental exposures with health outcomes but may be useful for understanding the most important factors driving large-scale variations in disease prevalence. The second case study shows the potential of using archival data in hypothesis development which can serve to target future data collection efforts.