False springs and spring phenology: Propagating effects of downscaling technique and training data

Projected changes to spring phenological indicators (such as first leaf and first bloom) are of importance to assessing the impacts of climate change on ecosystems and species. The risk of false springs (when a killing freeze occurs after plants of interest bloom), which can cause ecological and economic damage, is also projected to change across much of the United States. Given the coarse nature of global climate models, downscaled climate projections have commonly been used to assess local changes in spring phenological indices. Few studies that examine the influence of the sources of uncertainty sources in the downscaling approach on projections of phenological changes. This study examines the influence of sources of uncertainty on projections of spring phenological indicators and false spring risk using the South Central United States. The downscaled climate projections were created using three statistical downscaling techniques applied with three gridded observation datasets as training data and three global climate models. This study finds that projections of spring phenological indicators and false spring risk are primarily sensitive to the choice of global climate models. However, this study also finds that the formulation of the downscaling approach can cause errors representing the daily low-temperature distribution, which can cause errors in false spring risk by failing to capture the timing between the last spring freeze and the first bloom. One should carefully consider the downscaling approach used when using downscaled climate projections to assess changes to spring phenology and false spring risk.