Radiocarbon provides relatively poor age resolution in late Holocene marine records, primarily because of the temporal and spatial variability of the marine radiocarbon reservoir age (the apparent radiocarbon age of marine bicarbonate, typically averaging ca. 400 yr). Local estimates of marine reservoir ages and especially assessment of the temporal variability of reservoir ages are available for very few areas, so most studies correct and calibrate radiocarbon ages without accounting for the natural variability of the reservoir age. We present estimates of ca. decadal variability in reservoir ages for the last several hundred years for the North Atlantic at Nantucket Shoals, off New England, based on radiocarbon analysis of 1) known-age samples from live-collected shells of the bivalve Arctica going back to 1880 AD, and 2) older, subfossil samples whose age is estimated from aspartic acid (Asp). Interdecadal variability of the reservoir age is assessed independently of the age of the shell by comparison of variation among decadal samples within the same long-lived shell (Arctica often live 100 yr and longer). Interannual variability is assessed through analysis of series of individual bands from two individuals. These variabilities are compared to other ocean regions for which such estimates are available, such as northern Europe and the Gulf of California.

The precision of Asp racemization dating of Arctica shells is limited by the precision of calibration of the rate of racemization and by the analytical error of the method. A rate calibration is presented, based on Asp racemization analysis of 1) sample from a 110-year time series of known-age growth bands from live-collected shells, and 2) a large series of radiocarbon-dated subfossil shells. Whereas individual radiocarbon dates have low precision, as noted above, the racemization rate based on a large series of samples is determined with much better precision than any one date comprising that series. Because racemization analytical errors are proportional to the D/L value, precision for younger samples is much better than with radiocarbon.