Spatial variation in the influence of the North Atlantic Oscillation on precipitation across Greenland

[1] Ice core-derived accumulation records from Greenland have been proposed as proxies for North Atlantic Oscillation (NAO) reconstruction. In a series of single-site analyses, accumulation records from ice cores drilled in western Greenland were found to exhibit the strongest linear association with NAO. In this paper, we expand on these findings by proposing a spatiotemporal statistical model to explore further the relationship between NAO and the accumulation records from 35 firn and ice cores drilled in western and southern Greenland. In particular, we propose a temporal extension of the Bayesian spatially varying coefficient regression model, which is fit using a Markov chain Monte Carlo algorithm. This model readily accommodates the irregular features of the data (i.e., variation in record lengths and irregular spacing among ice core locations) and the serial dependence within individual records. Using our statistical model, we are able to exploit the spatial dependence structure of the derived accumulation-NAO relationship to explore the regional patterns in the strength of this relationship. Our findings support previous work that identified a region in western Greenland where derived accumulation is most correlated with NAO. However, we also identify a region further inland to the east and south where the estimated strength of the linear accumulation-NAO relationship is weaker, but more certain, than in the previously identified region. Thus, our findings can be used to guide decisions regarding where to locate future drilling efforts in Greenland by weighing the trade-off between the potential strength of the accumulation-NAO relationship and its uncertainty.