Comparison between a multi-variate nudging method and the ensemble Kalman filter for sea-ice data assimilation

Abstract

Increasing ship traffic and human activity in the Arctic has led to a growing demand for accurate Arctic weather forecast. High-quality forecasts obtained by models are dependent on accurate initial states achieved by assimilation of observations. In this study, a multi-variate nudging (MVN) method for assimilation of sea-ice variables is introduced. The MVN assimilation method includes procedures for multivariate update of sea-ice volume and concentration, and for extrapolation of observational information spatially. The MVN assimilation scheme is compared with the Ensemble Kalman Filter (EnKF) using the Los Alamos Sea Ice Model. Two multi-variate experiments are conducted: in the first experiment, sea-ice thickness from the European Space Agency's Soil Moisture and Ocean Salinity mission is assimilated, and in the second experiment, sea-ice concentration from the ocean and Sea Ice Satellite Application Facility is assimilated. The multivariate effects are cross-validated by comparing the model with non-assimilated observations. It is found that the simple and computationally cheap MVN method shows comparable skills to the more complicated and expensive EnKF method for multivariate update. In addition, we show that when few observations are available, the MVN method is a significant model improvement compared to the version based on one-dimensional sea-ice concentration assimilation.

Impact of assimilating sea ice concentration, sea ice thickness and snow depth in a coupled ocean–sea ice modelling system

“…The effect is much stronger for the SMOS rim SIT assimilation, indicating that a large portion of the original sea-ice volume overestimation is located in the MIZ. This is a consequence of too much ice in the control model, causing the observed MIZ to be located deeper in the Arctic compared to the model, as noted by Fritzner et al (2018).…”

“…In general, similarly to that found by Yang et al (2014) the SMOS observations were found to have a relatively small impact on the SIC and the SIT far from the ice edge. Fritzner et al (2018) assimilated SMOS observations into a stand-alone sea-ice model with the EnKF. This study showed that, due to the correlation between SIC and SIT, the SMOS observations were found to have a positive effect on the modelled SIC.…”

Impact of assimilating sea ice concentration, sea ice thickness and snow depth in a coupled ocean-sea ice modeling system

“…The effect is much stronger for the SMOS rim SIT assimilation, indicating that a large portion of the original sea-ice volume overestimation is located in the MIZ. This is a consequence of too much ice in the control model causing the observed MIZ to be located deeper into the Arctic as compared to the model, as noted by Fritzner et al (2018).…”

Assessment of High‐Resolution Dynamical and Machine Learning Models for Prediction of Sea Ice Concentration in a Regional Application

“…(Caya et al, 2010;K. Wang et al, 2013;Sakov et al, 2012;Buehner et al, 2013;Yang et al, 2014;Posey et al, 2015;Shlyaeva et al, 2016;Xie et al, 2016;Mu et al, 2018;Fritzner et al, 2018Fritzner et al, , 2019. Common for many of the Arctic sea-ice models used in these studies is that the model resolution is typically coarse, on the order of 10-20 km.…”

“…(Lisaeter et al, 2003;Sakov et al, 2012;K. Wang et al, 2013;Buehner et al, 2013;Posey et al, 2015;Fritzner et al, 2018Fritzner et al, , 2019. Sea-ice concentration (SIC) is by far the most used variable in sea-ice data assimilation studies, however other types of observations have become available in recent years.…”