Assessing Multi-centennial Spatial Coherence Among Atlantic Landfalling Tropical Cyclones

Paleo-proxy reconstructions of tropical cyclone impacts along the US and Caribbean coastlines suggest centennial patterns of local landfall variability that may not be reflected at synoptic scales. For example, sediment-based records of past hurricane activity from the northeast United States show a decreased (increased) susceptibility to storm strikes from ~1100 to1400 CE (~1400 and 1600 CE), while reconstructions from the eastern Yucatan peninsula exhibit the opposite trend during those same intervals. This indicates prolonged intra-basin patterns of tropical cyclone behavior unobserved in the modern era. Verifying these trends and assessing the underlying driving mechanisms would improve our understanding of tropical cyclone activity over longer timelines. However, the limited number, variable temporal resolution, and sparse spatial distribution of paleo-proxy records impede our ability to adequately interrogate these larger patterns of broad low-frequency spatial coherence. Resolving these uncertainties requires a tropical cyclone dataset more reflective of long-term basin-wide activity, which the current network of paleo-proxy reconstructions is unable to provide. Here we present an assessment of the regional distribution and coherency of landfalling tropical cyclones using 1150 years of synthetic Atlantic storm tracks (consisting of 100 storm per year). This dataset was statistically downscaled using initial and boundary conditions from the AM 2.1 atmospheric model coupled to sea surface temperatures from version 2.1 of the Last Millennium Reanalysis (LMR), which incorporates paleoclimate reconstructions to more accurately simulate Common Era climate conditions. The results show that the frequency of tropical cyclones that have impacted discrete coastal sectors within 1) the eastern United States, 2) the Gulf of Mexico/Yucatan Peninsula, and 3) the southern Caribbean over the past millennium were predominately internally correlated. However, comparisons between these larger domains reveal degrees of coherency that differ from those at smaller spatial scales. Strikingly, the centennial variability of storms impacting the eastern coast of the United States is strongly antiphased with those impacting the Gulf/Yucatan domain between ~1150 and ~1700 CE (r = -0.49, p < 0.001), which mirrors the trends observed in the sparse paleo-reconstructions from those regions. Further work is needed to resolve the mechanisms driving these coherent shifts in landfalling locations, which may be related to low-frequency alterations of the synoptic climate modes previously shown to influence storm tracks on shorter timelines (e.g., ENSO and meridional displacements of the Intertropical Convergence Zone). This work is the first to examine long-term variability in the distribution of landfalling tropical cyclones in the Atlantic over the past millennium, demonstrates the utility of leveraging paleoclimate data assimilation (i.e., LMR) products to support paleo-hurricane analyses, and helps us identify regions most at risk from storm impacts along timelines commensurate with future climate change scenarios.