GEOB370 Final Project

Wildfire Smoke and Public Health in BC

Spatial Analysis in Exposure Assessment and Health Implications

• Abstract
• Introduction
• Data
• Methodology
• Results I
• Results II
  
• Conclusions
• Limitations
• Future study
• Acknowledgements

Methodology

Exposure Assessment
Three methods were used to assign exposure to each LHA in this project:

Method 1: Nearest monitors
Exposure for each LHA was assigned from the nearest monitoring station. Average PM2.5 levels during the fire episodes were calculated for each monitoring station and using ArcGIS tool "Near" in the toolbox "Analysis Tools - Proximity" to identify the nearest monitoring station to each LHA centroid.

Method 2: Population-weighted values from nearest monitors
Fire period average PM2.5 concentrations were firstly assigned to each DA centroid from the nearest monitoring station using "Near" tool described in Method 1. Then a population-weighted average was calculated for each LHA from DAs within it. Equation for population-weighted average is:

                                              

where xn is the concentration assigned to DA and pn is the population of the corresponding DA.

Method 3: Population-weighted values from BlueSky predictions
Fire period average PM2.5 concentrations were firstly extracted to DA centroids from the BlueSky prediction raster pixels using the tool "Extract Values to Points" in the toolbox "Spatial Analyst Tools - Extraction". Then a population-weighted average for each LHA was calculated using the same equation in Method 2.

Exposure Assessment Comparison
Maps of difference between Method 1 and Method 2, as well as Method 2 and Method 3 were produced for examination.

Health Responses Linking
The exposures assessed by the three methods were linked to two measures of health responses named:

Mean: average daily counts of prescriptions during fire period;
Sum: culmulated total counts of prescriptions during fire period.

Regression analysis
Linear regression analyses using "Ordinary Least Square" in the toolbox "Spatial Statistics Tools - Modeling Spatial Relationship" were preformed between the three exposure assessment methods (Method 1, 2 and 3) and the two health response measures (Mean and Sum). Significance (p-value), coefficient and R-squared were examined and compared.

Spatial autocorrelation
For regressions with acceptable significance (p-value<0.05), the spatial autocorrelation of residues was examined by computing Moran's Index value. This process was done using  the tool "Spatial Autocorrelation (Moran's I)" in the toolbox "Spatial Statistics Tools - Analyzing  Patterns".