Recent technological advances in temporal geographic information systems (TGIS) include the Bayesian Maximum Entropy (BME) method, which accounts for the composite space/time variability and the wide variety of soft data characterizing many environmental and health processes. However, there are still several unaddressed implementation issues in the application of BME in environmental and health studies. In this work, the BME approach is applied to an air and a water environmental exposure assessment study where several unaddressed implementation issues are addressed.

First, a moving-window implementation of the BME method was numerically implemented and applied to the assessment of long-term exposure to ambient PM_2.5 across the contiguous U.S. Results for this work indicate that the moving-window BME method provides an efficient framework to account for the non-stationarity of the air pollutant variability and for the incompleteness of daily PM_2.5 measurements, which leads to estimates that are about 10 to 20% more accurate than those of classical approaches.

In a second study a two-stage estimation framework is implemented to estimate the concentration of E. coli across the tubewells in Bara Haldia, Bangladesh. The first stage of this framework consists in a latrine hydrological regression model, while the second BME stage of this estimation framework rigorously accounts for the uncertainty associated with the Most Probable Number (MPN) estimation of the density of microorganisms using data from multiple dilution series. The findings of this work indicate that latrines are a potential source of contamination of tubewells and thus have a significant impact on the spatial distribution of E. coli across tubewells.

Both applications show that the estimation framework based on the BME method successfully reduces estimation error compared with conventional geostatistical methods and provide highly informative maps.