Simulation of δ18O in Svalbard precipitation by the regional circulation model REMOiso: a comparison with GNIP and ice core data

The results of a regional circulation model REMOiso fitted with water isotope diagnostics are compared with three isotope series from western Svalbard. The regional (approx. 50 km resolution) model was forced at the lateral boundaries of the model domain by the global model ECHAMiso using the SST of the ERA-40 reanalysis for the period 1958 to 2001. The model domain encompasses the North Atlantic to the north of ca. 30°N and the adjacent areas including Greenland, part of the Canadian Arctic and Svalbard Archipelago.We examine two isotopic (δ18O) records from the ice cores drilled on Svalbard ice caps in 1997 (Lomonosovfonna, 1250 m asl) and 2005 (Holtedahlfonna, 1150 m asl) and monthly GNIP series from Ny-Alesund. The REMOiso skill in simulating the climate of Svalbard is evaluated through comparison of the modeled data with longer instrumental series of meteorological observations from Longyerbyen and Ny-Alesund. Analysis demonstrates that the model captures the seasonal cycle and the interannual variations of surface air temperature (SAT) and precipitation generally well. The elevation for the ice core sites in the model is approximately 500 m lower than the actual altitudes. However, the simulated multiannual mean precipitation-weighted δ18O values lie in a good agreement with both the ice core and GNIP data. The interannual variability of the annual mean δ18O in the model is reproduced very well for the GNIP station series and Holtedahlfonna ice core (r=0.4 for the common period of 1958-2001). Generally poor year to year agreement with the Lomonosovfonna data could be due to a higher contribution of precipitation at this core site formed by the easterlies. Location of Svalbard at the north-eastern flank of the model domain conditions in this case the presence of boundary effects between the much coarser ECHAM and regional REMO grids. Both the model and instrumental data demonstrate that intermittence and variations in intensity of precipitation events in the study area place a natural limit on the accuracy of past SAT reconstruction from the ice core isotopic series. The precipitation-weighted SAT is proposed, in turn, as a better target for paleoclimate reconstruction. The results encourage a further analysis of the REMOiso data. The model can successfully be used to improve our understanding of the processes driving the variability of water isotopes in Svalbard precipitation. This will boost the skills of the ice-core based climate reconstructions for Svalbard, which now extend back to the year 1100.