Evaluations of Satellite-derived Spring Phenology and Implications in Public Health Applications

Climate change can alter the timing of spring phenological events, such as leaf out or flowering. Shifts in spring phenology may affect the prevalence and severity of asthma associated with pollen exposures. The health community is paying increasing attention to use satellite-derived spring phenology (start of spring, SOS) to characterize regional changes related to pollen release. Yet few studies have broadly evaluated SOS data for public heath applications. In the U.S. Northeast, we evaluated SOS metrics of MODIS Land Change Dynamic (MLCD) and eMODIS Remote Sensing Phenology (eMODIS) products by comparing them with in-situ observations and detecting their spatiotemporal patterns at the county level. Furthermore, we investigated the association of SOS with adult asthma prevalence, using changes in SOS to characterize shifts in spring phenology. In ground-based comparisons, MLCD-SOS had a higher correlation (ρ = 0.54), while eMODIS-SOS had a smaller bias (3.7 days). Thus, MLCD-SOS is more suitable to reflect interannual spring phenological trends, and eMODIS-SOS is better to be a proxy of observed leaf-out. Only MLCD-SOS exhibited a spatial gradient at the county level, with spring arrival delaying by 5 days per 1° northward, which indicates this SOS is sensitive to geospatial changes in springtime. The vegetation cover has a significant impact on estimating regional SOS, as there were high differences (22 days) between the two types of SOS in counties with low forest coverages (≤ 25%). With both types of SOS data, we did not find a significant association between changes of SOS and adult asthma prevalence when linking them at the state level. Importantly, this phenomenon reveals that the use of relative changes of SOS can contribute to robust outcomes in practical applications even though different types of SOS may provide inconsistent springtime.