Mid- to late Holocene changes in tropical Atlantic temperature seasonality and interannual to multidecadal variability documented in southern Caribbean corals
Proxy reconstructions of tropical Atlantic sea surface temperature (SST) that extend beyond the period of instrumental observations have primarily focused on centennial to millennial variability rather than on seasonal to multidecadal variability. Here we present monthly-resolved records of Sr/Ca (a proxy of SST) from fossil annually-banded Diploria strigosa corals from Bonaire (southern Caribbean Sea). The individual corals provide time-windows of up to 68 years length, and the total number of 295 years of record allows for assessing the natural range of seasonal to multidecadal SST variability in the western tropical Atlantic during snapshots of the mid- to late Holocene. Comparable to modern climate, the coral Sr/Ca records reveal that mid- to late Holocene SST was characterised by clear seasonal cycles, persistent quasi-biennial and prominent interannual as well as inter- to multidecadal-scale variability. However, the magnitude of SST variations on these timescales has varied over the last 6.2 ka. The coral records show increased seasonality during the mid-Holocene consistent with climatemodel simulations indicating that southern Caribbean SST seasonality is induced by insolation changes on orbital timescales,whereas internal dynamics of the climate system play an important role on shorter timescales. Interannual SST variability is linked to ocean– atmosphere interactions of Atlantic and Pacific origin. Pronounced interannual variability in the western tropical Atlantic is indicated by a 2.35 ka coral, possibly related to a strengthening of the variability of the El Niño/Southern Oscillation throughout the Holocene. Prominent inter- to multidecadal SST variability is evident in the coral records and slightly more pronounced in the mid-Holocene. We finally argue that our coral data provide a target for studying Holocene climate variability on seasonal and interannual to multidecadal timescales, when using further numerical models and high-resolution proxy data.