Search

Simulating sea ice drift and deformation in the Arctic Ocean is still a challenge because of the multiscale interaction of sea ice floes that compose the Arctic Sea ice cover. The Sea Ice Rheology Experiment (SIREx) is a model intercomparison project of the Forum of Arctic Modeling and Observational Synthesis (FAMOS). In SIREx, skill metrics are designed to evaluate different recently suggested approaches for modeling linear kinematic features (LKFs) to provide guidance for modeling small‐scale deformation. These LKFs are narrow bands of localized deformation that can be observed in satellite images and also form in high resolution sea ice simulations. In this contribution, spatial and temporal properties of LKFs are assessed in 36 simulations of state‐of‐the‐art sea ice models and compared to deformation features derived from the RADARSAT Geophysical Processor System. All simulations produce LKFs, but only very few models realistically simulate at least some statistics of LKF properties such as densities, lengths, or growth rates. All SIREx models overestimate the angle of fracture between conjugate pairs of LKFs and LKF lifetimes pointing to inaccurate model physics. The temporal and spatial resolution of a simulation and the spatial resolution of atmospheric boundary condition affect simulated LKFs as much as the model's sea ice rheology and numerics. Only in very high resolution simulations (≤2 km) the concentration and thickness anomalies along LKFs are large enough to affect air‐ice‐ocean interaction processes., Plain Language Summary: Winds and ocean currents continuously move and deform the sea ice cover of the Arctic Ocean. The deformation eventually breaks an initially closed ice cover into many individual floes, piles up floes, and creates open water. The distribution of ice floes and open water between them is important for climate research, because ice reflects more light and energy back to the atmosphere than open water, so that less ice and more open water leads to warmer oceans. Current climate models cannot simulate sea ice as individual floes. Instead, a variety of methods is used to represent the movement and deformation of the sea ice cover. The Sea Ice Rheology Experiment (SIREx) compares these different methods and assesses the deformation of sea ice in 36 numerical simulations. We identify and track deformation features in the ice cover, which are distinct narrow areas where the ice is breaking or piling up. Comparing specific spatial and temporal properties of these features, for example, the different amounts of fractured ice in specific regions, or the duration of individual deformation events, to satellite observations provides information about the realism of the simulations. From this comparison, we can learn how to improve sea ice models for more realistic simulations of sea ice deformation., Key Points: All models simulate linear kinematic features (LKFs), but none accurately reproduces all LKF statistics. Resolved LKFs are affected strongest by spatial and temporal resolution of model grid and atmospheric forcing and rheology. Accurate scaling of deformation rates is a proxy only for realistic LKF numbers but not for any other LKF static., DOE, HYCOM NOPP, Innovation Fund Denmark and the Horizon 2020 Framework Programme of the European Union, National centre for Climate Research, SALIENSEAS, ERA4CS, German Helmholtz Climate Initiative REKLIM (Regional Climate Change), Gouvernement du Canada, Natural Sciences and Engineering Research Council of Canada (NSERC) http://dx.doi.org/10.13039/501100000038, Environment and Climate Change Canada Grants & Contributions program, Office of Naval Research Arctic and Global Prediction program, U.S. Department of Energy Regional and Global Model Analysis program, National Science Foundation Arctic System Science program, Deutsche Forschungsgemeinschaft (DFG) http://dx.doi.org/10.13039/501100001659, https://zenodo.org/communities/sirex

Sea ice is regarded as a clear indicator of climate change in the Arctic Ocean. Landfast ice is immobile or nearly immobile sea ice in coastal regions that affects the transfer of heat, moisture, and momentum between the atmosphere and the ocean. As an extension of the land for travel and hunting, landfast ice also determines the construction of ice roads and Arctic shipping routes in the summertime. Despite the important role of landfast ice in the climate system, landfast ice is not simulated very well by current sea ice models and needs to be parameterized, for example, by a grounding scheme. Comparing landfast ice in two sea-ice simulations with different grid resolutions indicates that a higher resolution model better presents landfast ice in deep water regions, where a grounding scheme fails. The better representation of coastline details, which serve as pinning points for sea ice arches in the high-resolution model, is thought to improve the representation of landfast ice. Based on this hypothesis, a new parameterization of lateral drag as a function of sea ice thickness, drift velocity, and coastline length is presented. The results suggest that a combination of lateral drag parameterization and grounding (parameterized by basal stress) is required to simulate fast ice in most regions successfully. This work may lead to a versatile landfast ice parameterization for sea ice models in both shallow and deep coastal areas in the Arctic.

Sea-ice observations and models show zones of high deformation typical of granular medium (linear kinematic features (LKFs)). Recent high-resolution simulations feature fractures that mimic the observed pattern but with wider intersection angles. Motivated by this, we investigate the dependence between conjugate faults intersection angles and different viscous–plastic rheologies. Using an idealized uniaxial setting, the ice fracture is modeled with different confinement ratios and two different VP rheologies: one with an elliptical yield curve and a normal flow rule, and one with a Coulombic yield curve and a normal flow rule that applies only to the elliptical cap. Modeling fracture angles smaller than 30° is not possible with an elliptical yield curve in a pure compression setting. Further several modeled behaviors are inconsistent with the granular nature of sea ice : (1) the fracture angle increases with ice shear strength; (2) the divergence along the fracture lines (or LKFs) is uniquely defined by the shear strength of the material with divergence for high shear strength and convergence follow shear strength; (3) the angle of fracture depends on the confining pressure with more convergence as the confining pressure increases. With Mohr’s circle, this behavior is shown to be linked to the convexity of the yield curve. The Coulombic yield curve is able to model smaller angles but the solution is unstable because of non-differentiable corners between the straight limbs of the Coulombic yield curve and the elliptical cap. The results show that, although the fracture patterns at first appear realistic, the yield curve should be revised to take into account the nature of sea ice as a pressure-sensitive and dilatant granular material.

Most dynamic sea ice models for climate type simulations are based on the viscous-plastic (VP) rheology. New rheologies such as the Maxwell-Elasto-Brittle (MEB) rheology are usually compared against traditional VP-schemes, but the new schemes also require revisiting the validity of VP-schemes. So far, comparisons between different schemes are confounded by factors unrelated to rheology, such as grid resolution, advection schemes, forcing by atmosphere and ocean, and last but not least, by differences in numerical details of different model codes. The sea ice component of the Massachusetts Institute of Technology general circulation model (MITgcm) offers an easy-to-use testbed for comparing different sea ice rheologies and implementation that avoids any confounders because all solvers share the same code and configuration environment. VP-rheologies with different flavors of Picard (or fixed point iterative) solvers, Newton methods, and different variants of the Elastic-Viscous-Plastic solver have been evaluated in this framework. With this framework, a new implementations such as an MEB solver may be compared to these traditional solvers in idealized geometries and in realistic Arctic configurations.

Sea ice is an important component of the climate system. The thermodynamic processes of freezing and melting are generally straightforward to parameterize in a sea ice model on climate scales, but the internal dynamics in a sea ice model still rely on crude assumptions. The sea ice cover is treated as a quasi-continuous non-Newtonian fluid with a rheology describing its behavior under compression and shear. With model grid spacings of 10 to 100km, these assumptions have been used successfully to simulate the large scale properties of sea ice, but they may need to be reevaluated as numerical climate models are operated at increasingly high resolution. Still, simulations with conventional sea ice models with a viscous-plastic rheology improve with increasing resolution. Here we show two examples: fractal scaling properties of simulated sea ice at high resolution agree with satellite observations and the distribution and formation of fast ice in the Arctic Ocean improves with increasing resolution.

The Arctic sea ice deforms constantly due to stresses imposed by winds, ocean currents and interaction with coastlines. The most dominant features produced by this deformation in the ice cover are leads and pressure ridges that are often referred to as Linear Kinematic Features (LKFs). With increasing resolution of classical (viscous-plastic) sea ice models, or using new rheological frameworks (e.g. Maxwell elasto-brittle), sea-ice models start to resolve this small-scale deformation. So far, scaling properties of sea-ice deformation are commonly used to evaluate the modelled LKFs, besides other measures like lead area density. These metrics evade the problem of detecting individual LKFs by taking statistics over continuous fields like sea ice deformation or concentration. This way, they can provide specific information, but lack a comprehensive description of LKFs. We detect individual LKFs in sea ice deformation fields from satellite observations with an object detection algorithm. Combining this information with the sea ice drift fields used to derive the deformation fields, the LKFs are tracked in time. In doing so, the spatial characteristics (density, length, orientation, intersection angle, curvature) as well as the temporal evolution can be extracted from the same data-set. This algorithm can be applied to modelled sea-ice deformation and drift to enable a consistent comparison and thorough evaluation of simulated sea-ice deformation. We present preliminary results of LKFs detected in the RGPS data set and give examples of possible applications.

Sea ice deformation localizes along Linear Kinematic Features (LKFs) that are relevant for the air/ocean/sea-ice interaction and for shipping andmarine operations. At high resolution (< 5km) viscous-plastic sea ice models start to resolve LKFs. Here, we study the short-range (up to 10 days) potential predictability of LKFs in Arctic sea ice using ensemble simulations of an ocean/sea-ice model with a grid point separation of 4.5 km. We analyze the sensitivity of predictability to idealized initial perturbations, mimicking the uncertainties in sea ice analyses, and to growing uncertainty of the atmospheric forcing caused by the chaotic nature of the atmosphere. The similarity between pairs of ensemble members is quantified by Pearson correlation and Modified Hausdorff Distance (MHD). In our perfect model experiments, the potential predictability of LKFs, based on the MHD, drops below 0.6 after 4 days in winter. We find that forcing uncertainty (due to limited atmospheric predictability) largely determines LKF predictability on the 10-day time scale, while uncertainties in the initial state impact the potential predictability only within the first 4 days.

The impact of a subgrid-scale ice thickness distribution (ITD) and two standard ice strength formulations on simulated Arctic sea ice climate is investigated. To this end, different model configurations with and without an ITD were tuned by minimizing the weighted mean error between the simulated and observed sea ice concentration, thickness, and drift speed with an semiautomatic parameter optimization routine. The standard ITD and ice strength parameterization lead to larger errors when compared to the simple single-category model with an ice strength parameterization based on the mean ice thickness. Interestingly, the simpler ice strength formulation, which depends linearly on the mean ice thickness, also reduces the model-observation error when using an ITD. For the ice strength parameterization that makes use of the ITD, the effective ice strength depends strongly on the number of thickness categories, so that introducing more categories can lead to overall thicker ice that is more easily deformed.

Many climate models use a rheology of the viscous-plastic type to simulate sea ice dynamics. With this rheology, large scale velocity and thickness fields can be realistically simulated, but the representation of small scale deformation rates and Linear Kinematic Features (LKF) is thought to be inadequate. However, at high resolution (< 5 km) the rheology starts to produce lines of localised deformation rates. In this study we use results from a 1-km Pan-Arctic model to investigate the influence of these deformation features on the scaling properties of sea ice deformation. For evaluation the EGPS satellite data set of small-scale sea ice kinematics for the Central Arctic (successor of RGPS) is used. The modelled sea ice deformation shows multi-fractal spatial scaling and, in this sense, agrees with the satellite data. In addition, the temporal coupling of the spatial scaling is reproduced as well. Furthermore, we examine the regional and seasonal variations of spatial scaling properties and its dependence on the ice condition, i.e. sea ice concentration and thickness, which are in agreement with previous RGPS studies.

Linear kinematic features (LKFs) in sea ice, potentially important for short-term forecasts and for climate simulations, emerge as viscous-plastic sea ice models are used at high resolution (~ 4.5 km). Here we analyze the short-range (up to 10 days) potential predictability of LKFs in Arctic sea ice using an ocean/sea-ice model with a grid point separation of 4.5 km. We analyze the sensitivity of predictability to idealized initial perturbations, mimicking the uncertainties in sea ice analyses, and to growing uncertainty of the atmospheric forcing caused by the chaotic nature of the atmosphere. For the latter we use different members of ECMWF ensemble forecasts to drive ocean/sea-ice forecasts. For our analysis, we diagnose LKFs occurrence and investigate different sea ice characteristics. We find that forcing uncertainty (due to limited atmospheric predictability) largely determines LKF predictability on the 10-day time scale. When it comes to metrics, we demonstrate that spatial correlation, although a useful metric to measure some aspects of deformation field similarity, fails to detect LKF similarity when LKFs are only slightly shifted in space. The Modified Hausdorff Distance (MHD) appears to be a more appropriate metric, but it can be misleading if the LKF density is very high, for example due to artificial LKFs caused by spurious small-scale perturbations of the sea-ice initial state.

The Arctic is highly sensitive to changes in the global climate. One of the most prominent examples in this context is the shrinking sea ice cover over the past decades. The impact of the change in ice coverage on the global climate requires a thorough understanding of ice dynamics, as drifting sea ice transports salt and heat and in this way influences ocean dynamics. Sea ice motion is mainly driven by wind, ocean currents and internal ice stress. Convergent ice motion causes features such as ridges and rubble fields, which change the momentum exchange between atmosphere, ice and ocean. Openings in the ice due to divergent motion increase the exchange of heat and matter between ocean and atmosphere. On time scales of days to weeks, linear kinematic features, such as leads and ridges, evolve on spatial scales ranging from meters to tens and hundreds of kilometers. These features also emerge in numerical sea ice models when the resolution of the simulations is increased to a few kilometers. While plausible, their realism in the simulations is yet unclear and requires a detailed evaluation, with the help of (in our case satellite-based) observations. We use Arctic-wide MITgcm simulations for 2006 at a spatial resolution of approximately 4km for a regional comparison with microwave satellite observations, e.g. from Synthetic Aperture Radar (SAR). We derive sea ice displacement from a sequence of satellite images by measuring the offset between matching patterns in different images. Discontinuities in the resulting velocity field indicate regions of instantaneous deformation that occur at some point in the time interval between the acquisition of the two SAR images used for displacement retrieval. The obtained quantities of deformation by divergence, shear and vorticity are scale-dependent. As a consequence, they depend on the spatial resolution of the SAR images and differ from the quantities calculated by the model. Hence, the comparison between model simulations and results of retrievals from remote sensing data is not straightforward. Therefore, we pay special attention to the spatial and temporal scales of the observed / modelled processes and introduce appropriate statistical tools. By evaluating the kinematic features in the results of high-resolution sea ice models based on the microwave remote sensing data we expect to be able to assess and improve the ice rheology in these sea ice models.

Most climate models use a rheology of the viscous-plastic type to simulate sea ice dynamics. With this rheology, large scale velocity and thickness fields can be realistically simulated, but the representation of small scale defor- mation rates and Linear Kinematic Features (LKF) is thought to be inadequate. Here, the spatial grid spacing of a traditional VP sea ice model is gradually reduced to 1 km in order to investigate how LKFs emerge with increasing resolution and to explore spatial and temporal scaling laws for sea ice deformation. Increasing the spatial resolution localizes the strain rates along the LKFs. At 1 km grid spacing, the distributions of strain rates have power-law tails that clearly deviate from the basin of attraction of Gaussian distributions and, in this sense, agree with satellite observations and results obtained with the elasto-brittle rheology. Increasing the resolution of the wind forcing leads to more small scale strain rate events in the ice and improves the agreement with observational spatial scaling laws.

Linear kinematic features (LKFs) in sea ice, potentially important for short-term forecast users as for climate simulations, emerge as viscous-plastic sea ice models are used at high (

In spite of its comparatively small volume sea ice plays a major part in the Arctic climate system, because the interactions with the ocean and the atmosphere lead to many feedbacks in the coupled system. Unfortunately, these interesting properties are also one reason why modeling sea ice is still not mature. Many physical processes, most prominently maybe the formation and evolution of leads, are still only poorly represented in numerical models. The cause of those misrepresentations is often very hard to pinpoint because many factors play a role. One possibility to address this problem is an adjoint model. Such a numerical tool takes one objective function out of the huge output of a climate model and calculates the gradients and thereby the sensitivities to all variables that are modeled. In a first step one property of the sea ice model such as the ice transport through a strait in a certain time span or the minimal summer sea ice extent is defined. The result of the adjoint model then gives directly the influence of all modeled variables of the ocean, sea ice and the atmospheric forcing on this property, resolved in space and time. This level of detail is in no way feasible to arrive at via traditional sensitivity analysis by parameter perturbation. With this information we want to investigate the role of different components of current sea-ice-ocean-models in the simulation results. For this we will focus for one part on the sensitivities of modeled sea ice distribution to the boundary, initial and forcing conditions prescribed to the model. The results can be used to inform future choices for additional measurements to improve model output. The other focus will be on the modeling framework itself, and the influence the rheology and the different physical parameterisations used in the sea ice models have on those calculated sensitivities.

We investigate the sensitivity of sub-ice-shelf melt rates under Pine Island Ice Shelf, West Antarctica, to changes in the oceanic state using an adjoint ocean model that is capable of representing the flow in sub-ice-shelf cavities. The adjoint code is based on algorithmic differentiation (AD) of the Massachusetts Institute of Technology’s ocean general circulation model (MITgcm). The adjoint model was extended by adding into the AD process the corresponding sub-ice-shelf cavity code, which implements a three-equation thermodynamic melt-rate parameterization to infer heat and freshwater fluxes at the ice-shelf/ocean boundary. The inferred sensitivities reveal dominant timescales of 30–60 days over which the shelf exit is connected to the deep interior via advective processes. They exhibit rich three-dimensional time-evolving patterns that can be understood in terms of a combination of the buoyancy forcing by inflowing water masses, the cavity geometry and the effect of rotation and topography in steering the flow in the presence of prominent features in the bedrock bathymetry. Dominant sensitivity pathways are found over a sill, as well as ‘shadow regions’ of very low sensitivities. To the extent that these transient patterns are robust they carry important information for decision- making in observation deployment and monitoring.

Past climates provide a means for evaluating the response of the climate system to large perturbations. Our ultimate goal is to constrain climate models rigorously by paleoclimate data. For illustration, we used a conceptual climate model (a classical energy balance model) and applied the so-called “adjoint method” to minimize the misfit between our model and sea-surface temperature data for the Last Glacial Maximum (LGM, between 19,000-23,000 years before present). The “adjoint model” (derivative code) was generated by an “adjoint compiler”. We optimized parameters controlling the thermal diffusion and the sensitivity of the outgoing longwave radiation to changes in the zonal-mean surface temperature and the atmospheric CO2 concentration. As a result, we estimated that an equilibrium climate sensitivity between 2.2 degC and 2.5 degC was consistent with the reconstructed glacial cooling, and we were able to infer structural deficits of the simple model where the fit to current observations and paleo data was not successful.

So-called complete gravity field models together with their full error variance/covariance information have recently been designed to be integrated into geo-scientific process models. In our case, the ocean's mean dynamic topography (altimetric mean sea surface referenced to the geoid) is used to improve estimates of the general ocean circulation in the context of stationary ocean models. We want to combine complete gravity field models with altimetric data for which a full error propagation is also implemented in the processing. Thus we derive estimates of the ocean's mean dynamic topography with a regular covariance matrix. The goal of this project is to assess the effects of this data combination on improving ocean models. Preliminary results already show that geoid models developed from GRACE data are, while accurate on very long scales, hardly yet accurate enough for that purpose. We anticipate that the increased accuracy, especially on shorter scales, of gravity measurements from GOCE will contribute to a more realistic description of ocean currents as well as mass and heat transports.

As part of an ongoing effort to obtain a best possible, time-evolvinganalysis of most available ocean and sea ice data, a dynamic andthermodynamic sea-ice model has been coupled to the MassachusettsInstitute of Technology general circulation model (MITgcm). Icemechanics follow a viscous-plastic rheology and the ice momentumequations are solved numerically using eitherline-successive-over-relaxation (LSOR) or elastic-viscous-plastic(EVP) dynamic models. Ice thermodynamics are represented using eithera zero-heat-capacity formulation or a two-layer formulation thatconserves enthalpy. The model includes prognostic variables for snowand for sea-ice salinity. The above sea ice model components wereborrowed from current-generation climate models but they werereformulated on an Arakawa C grid in order to match the MITgcm oceanicgrid and they were modified in many ways to permit efficient andaccurate automatic differentiation. This paper describes the MITgcmsea ice model; it presents example Arctic and Antarctic results from arealistic, eddy-permitting, global ocean and sea-ice configuration; itcompares B-grid and C-grid dynamic solvers and the effects of othernumerical details of the parameterized dynamics and thermodynamics ina regional Arctic configuration; and it presents example results fromcoupled ocean and sea-ice adjoint-model integrations.

Basal melting of marine‐terminating glaciers, through its impact on the forces that control the flow of the glaciers, is one of the major factors determining sea level rise in a world of global warming. Detailed quantitative understanding of dynamic and thermodynamic processes in melt‐water plumes underneath the ice‐ocean interface is essential for calculating the subglacial melt rate. The aim of this study is therefore to develop a numerical model of high spatial and process resolution to consistently reproduce the transports of heat and salt from the ambient water across the plume into the glacial ice. Based on boundary layer relations for momentum and tracers, stationary analytical solutions for the vertical structure of subglacial non‐rotational plumes are derived, including entrainment at the plume base. These solutions are used to develop and test convergent numerical formulations for the momentum and tracer fluxes across the ice‐ocean interface. After implementation of these formulations into a water‐column model coupled to a second‐moment turbulence closure model, simulations of a transient rotational subglacial plume are performed. The simulated entrainment rate of ambient water entering the plume at its base is compared to existing entrainment parameterizations based on bulk properties of the plume. A sensitivity study with variations of interfacial slope, interfacial roughness and ambient water temperature reveals substantial performance differences between these bulk formulations. An existing entrainment parameterization based on the Froude number and the Ekman number proves to have the highest predictive skill. Recalibration to subglacial plumes using a variable drag coefficient further improves its performance., Plain Language Summary: In a world of global warming, the melting of glaciers terminating as floating ice tongues into the oceans of Arctic and Antarctic regions allows those glaciers to flow faster and hence to make a considerable contribution to global mean sea‐level rise. Underneath the ice‐ocean interface, turbulent currents of the order of 10 m thickness (so‐called plumes) develop that transport the melt water from the grounding line where the glacier enters the ocean toward the calving front that marks the seaward end of the glacier. At its base, ambient relatively warm and salty ocean water is mixed into the plumes and is vertically transported toward the ice‐ocean interface, where the melting is increased due to the additional heat supply. Understanding these processes is essential for their incorporation into computer models for the prediction of such melt processes. In this study, an accurate simulation model for the water column is constructed that is able to consistently reproduce these processes. The algorithms developed here are proven to provide reliable results also for models with only a few grid points across the plume and can therefore be implemented into climate models with surface‐following coordinates to more accurately simulate future scenarios of sea level rise., Key Points: A vertically resolving model with second‐moment turbulence closure has been constructed for subglacial plumes. Convergent numerical formulations for the ocean‐to‐ice fluxes of momentum, freshwater and heat have been derived from an analytical model. Model results are consistent with bulk parameterizations for the entrainment of ambient water., Bundesministerium für Bildung und Forschung (BMBF) http://dx.doi.org/10.13039/501100002347, https://doi.org/10.5281/zenodo.6203838

The effects of anthropogenic climate change are most drastic in the Arctic. This amplification of climate change signals is strongly connected to the sea ice in the Arctic Ocean. This thesis presents an analysis of the sea ice cover in numerical ocean a sea ice models with a focus on two different parameterizations: an active ice thickness distribution and an ice strength parameterization that is based on this additional thickness information. The research questions are: (1) can the parameterizations improve the reproduction of Arctic-wide sea ice observations? (2) Do the parameterizations actually reproduce physically observed behavior? (3) How can the parameterizations and their use in basin-scale models be improved further? In a first step, model quality is assessed by a quantitative measure of the reproduction of satellite observations of sea ice concentration, thickness and drift. Including a full ice thickness distribution in each grid cell instead of only two ice categories clearly improves the model results. At the same time, a strength parameterization based on a two-category approach produces better model results than a multi-category strength parameterization. In a next step, the two parameterizations are evaluated in more detail. The ice thickness distribution parameterization reproduces local observations in the Arctic to a large degree and simulates faithfully regional and seasonal differences found in observed distributions. The poor performance of the multi-category ice strength parameterization is explained by the physical assumptions that were made in its original derivation and that do not agree with the current understanding of the ice cover. In conclusion, using an ice thickness distribution improves model performance, but a multi-category parameterization of the ice strength should be avoided. In future work, a new ice strength parameterization could be derived from the physical properties of the ice pack that are demonstrated in this work.