Forecasting pollen aerobiology with Modis EVI, land cover, and phenology using machine learning tools

Grass pollens are a major source of aeroallergens globally, inducing allergic asthma and hay fever in up to 500 million people worldwide. Pollen forecasting research and methods are site-dependent and tend to be empirically derived composites of expert knowledge and weather data. In this study we utilize satellite-based information of landscape conditions and phenology to better discern and predict grass pollen evolution. We employed machine learning approaches to formulate and better understand relationships between landscape phenology and seasonal flowering-induced pollen concentrations. We show that machine learning approaches significantly improved pollen prediction capabilities and provided key information to better attribute changes in pollen counts driven by shifting ecological landscapes from climate change drivers.