Adjustment of regional climate model output for modeling the climatic mass balance of all glaciers on Svalbard

• Published in: Journal of Geophysical Research: Atmospheres
• Main Authors: Moeller, Marco, Obleitner, Friedrich, Reijmer, Carleen H., Pohjola, Veijo A., Glowacki, Piotr, Kohler, Jack
• Format: Article in Journal/Newspaper
• Language: English
• Published: Uppsala universitet, Luft-, vatten- och landskapslära 2016
• Subjects: Earth and Related Environmental Sciences; Geovetenskap och miljövetenskap; Svalbard; Svalbard Archipelago; glacier
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-303139; https://doi.org/10.1002/2015JD024380

Large-scale modeling of glacier mass balance relies often on the output from regional climate models (RCMs). However, the limited accuracy and spatial resolution of RCM output pose limitations on mass balance simulations at subregional or local scales. Moreover, RCM output is still rarely available over larger regions or for longer time periods. This study evaluates the extent to which it is possible to derive reliable region-wide glacier mass balance estimates, using coarse resolution (10 km) RCM output for model forcing. Our data cover the entire Svalbard archipelago over one decade. To calculate mass balance, we use an index-based model. Model parameters are not calibrated, but the RCM air temperature and precipitation fields are adjusted using in situ mass balance measurements as reference. We compare two different calibration methods: root mean square error minimization and regression optimization. The obtained air temperature shifts (+1.43 degrees C versus +2.22 degrees C) and precipitation scaling factors (1.23 versus 1.86) differ considerably between the two methods, which we attribute to inhomogeneities in the spatiotemporal distribution of the reference data. Our modeling suggests a mean annual climatic mass balance of -0.05 +/- 0.40 mw.e.a(-1) for Svalbard over 2000-2011 and a mean equilibrium line altitude of 452 +/- 200m above sea level. We find that the limited spatial resolution of the RCM forcing with respect to real surface topography and the usage of spatially homogeneous RCM output adjustments and mass balance model parameters are responsible for much of the modeling uncertainty. Sensitivity of the results to model parameter uncertainty is comparably small and of minor importance.