FORECASTING ALLERGENIC POLLEN LEVELS THROUGH SPATIALLY-BASED REGRESSION MODELING
Sassard, Andrea Marie
MetadataShow full item record
The south-central United States has many allergenic plants that cause allergic respiratory disease. Current pollen forecasts methods have short-term predictive ability and fail to capture environmental variation. Advanced knowledge of pollen occurrence and levels allows allergy sufferers and medical professionals to better prepare. A geographic information system (GIS)-based predictive model, combining the distributions of reactive tree, grass, and weed species with localized environmental variables, was evaluated. Pollen data from monitoring stations in the region from 2003 to 2013 were analyzed and correlated to potential predictors, such as average meteorological parameters and land-use/ land-change data. For all stations, the mean total daily concentration was 336 grains/m<sup>3</sup>. Moderately strong correlations exist between concentration and temperature and latitude. Data were interpolation for use in regression modelling. The distribution and number of monitoring locations impacted interpolation accuracy. A pixel-wise spatial regression model was developed using normalized difference vegetation index (NDVI), but had no predictive power. Other variables, such as temperature, have potential for use in future modelling. The outcomes of this project are better understanding of pollen distributions and timings based on a regional view of existing spatial data sets.