Evaluating the skills of isotope-enabled general circulation models against in situ atmospheric water vapor isotope observations

The skills of isotope-enabled general circulation models are evaluated against atmospheric water vapor isotopes. We have combined in situ observations of surface water vapor isotopes spanning multiple field seasons (2010, 2011, and 2012) from the top of the Greenland Ice Sheet (NEEM site: 77.45°N, 51.05°W, 2484m above sea level) with observations from the marine boundary layer of the North Atlantic and Arctic Ocean (Bermuda Islands 32.26°N, 64.88°W, year: 2012; south coast of Iceland 63.83°N, 21.47°W, year: 2012; South Greenland 61.21°N, 47.17°W, year: 2012; Svalbard 78.92°N, 11.92°E, year: 2014). This allows us to benchmark the ability to simulate the daily water vapor isotope variations from five different simulations using isotope-enabled general circulation models. Our model-data comparison documents clear isotope biases both on top of the Greenland Ice Sheet (1-11% for δ¹⁸O and 4-19% for d-excess depending on model and season) and in the marine boundary layer (maximum differences for the following: Bermuda δ¹⁸O=~1%, d-excess = ~3%; South coast of Iceland δ¹⁸O=~2%, d-excess = ~ 5%; South Greenland δ¹⁸O=~4%, d-excess=~7%; Svalbard δ¹⁸O=~2%, d-excess=~7%). We find that the simulated isotope biases are not just explained by simulated biases in temperature and humidity. Instead, we argue that these isotope biases are related to a poor simulation of the spatial structure of the marine boundary layer water vapor isotopic composition. Furthermore, we specifically show that the marine boundary layer water vapor isotopes of the Baffin Bay region show strong influence on the water vapor isotopes at the NEEM deep ice core-drilling site in northwest Greenland. Our evaluation of the simulations using isotope-enabled general circulation models also documents wide intermodel spatial variability in the Arctic. This stresses the importance of a coordinated water vapor isotope-monitoring network in order to discriminate amongst these model behaviors.