Impact of surface wind biases on the Antarctic sea ice concentration budget in climate models

Abstract

Highlights• The ice concentration budget (ICB) is applied to several forced and coupled models.• Ice drift and 10 m wind vectors are evaluated against observations and reanalyses.• Biases in winds are largely responsible for biases in the ICB in free drift regions.• Errors in winds and sea ice model physics are equally important closer to the coast.• The method provides a new tool for climate model intercomparisons. AbstractWe derive the terms in the Antarctic sea ice concentration budget from the output of three models, and compare them to observations of the same terms. Those models include two climate models from the 5th Coupled Model Intercomparison Project (CMIP5) and one ocean-sea ice coupled model with prescribed atmospheric forcing. Sea ice drift and wind fields from those models, in average over April-October 1992-2005, all exhibit large differences with the available observational or reanalysis datasets. However, the discrepancies between the two distinct ice drift products or the two wind reanalyses used here are sometimes even greater than those differences. Two major findings stand out from the analysis. Firstly, large biases in sea ice drift speed and direction in exterior sectors of the sea ice covered region tend to be systematic and consistent with those in winds. This suggests that sea ice errors in these areas are most likely wind-driven, so as errors in the simulated ice motion vectors. The systematic nature of these biases is less prominent in interior sectors, nearer the coast, where sea ice is mechanically constrained and its motion in response to the wind forcing more depending on the model rheology. Second, the intimate relationship between winds, sea ice drift and the sea ice concentration budget gives insight on ways to categorize models with regard to errors in their ice dynamics. In exterior regions, models with seemingly too weak winds and slow ice drift consistently yield a lack of ice velocity divergence and hence a wrong wintertime sea ice growth rate. In interior sectors, too slow ice drift, presumably originating from issues in the physical representation of sea ice dynamics as much as from errors in surface winds, leads to wrong timing of the late winter ice retreat. Those results illustrate that the applied methodology provides a valuable tool for prioritizing model improvements based on the ice concentration budget-ice * Corresponding author

Comparing sea ice, hydrography and circulation between NEMO3.6 LIM3 and LIM2

“…This implies that changes in windiness are likely to be a major mechanism driving the SH sea-ice extent increase. Notably, the LIM modelled trends are larger than the observed ones, which may indicate a too sensitive ice drift response to increasing windiness, a too fast moving model sea ice and a too far northern winter sea-ice edge, as also supported by earlier studies Lecomte et al, 2016).…”

Drivers of Antarctic Sea Ice Volume Change in CMIP5 Models

“…The other models all have smaller sea ice mass transport, associated with negative mean sea ice biases, as discussed above. Biases in the speed and direction of sea ice velocity, particularly at the ice edge, substantially limit model capability to accurately reproduce rates of sea ice advance and retreat and should therefore be a key focus for future model improvements (Lecomte et al, ).…”

“…Casting equation in terms of the discretized CMIP5 output variables, for a given model grid cell with western and eastern bounds ( x 1 , x 2 ) and southern/northern bounds ( y 1 , y 2 ), gives The residual term is interpreted as thermodynamic change from both atmospheric and oceanic sources (Lecomte et al, ; Uotila et al, ). All three terms were then divided by the individual grid cell area and multiplied by the number of seconds in a year to give units of meters per year.…”

“…In this study, we investigate dynamic and thermodynamic contributions to the sea ice volume budget in global coupled climate models with different mean states and trends in Antarctic sea ice, by decomposing modeled sea ice budgets into advective and thermodynamic freeze/melt components. Although it has been suggested that detailed analyses of sea ice drivers and processes in individual models are warranted to identify the origins of model biases (Lecomte et al, ), there have been few such intermodel comparisons of sea ice advective and thermodynamic changes. Previous model studies have examined only sea ice concentration, not the entire sea ice budget.…”

“…In this model, the dynamic influence of strengthening westerly winds may cause increased upwelling of warmer subsurface water and an enhanced thermodynamic response in the simulated sea ice (Haumann et al, ). A study of sea ice concentration in the Community Climate System Model and Institut Pierre Simon Laplace Low‐Resolution model (IPSL‐CM5A‐LR), alongside an ocean‐sea ice model forced by reanalysis winds, showed that mean errors in the central and outer sea ice pack are largely due to inadequate representation of wind strength and velocity, while errors in simulated sea ice near the coast may be equally influenced by the model ice rheology (Lecomte et al, ). Overestimated and underestimated Antarctic sea ice transport in global coupled climate models has substantial implications for trends of freshening or salinification of the ocean surface and stratification of the ocean column, with further consequences for the simulation of deep water properties and mode water formation to the north (Abernathey et al, ; Bitz et al, ; Haumann et al, ).…”