Sea ice assimilation into a coupled ocean–sea ice model using its adjoint

Abstract. Satellite sea ice concentrations (SICs), together with several ocean parameters, are assimilated into a regional Arctic coupled ocean–sea ice model covering the period of 2000–2008 using the adjoint method. There is substantial improvement in the representation of the SIC spatial distribution, in particular with respect to the position of the ice edge and to the concentrations in the central parts of the Arctic Ocean during summer months. Seasonal cycles of total Arctic sea ice area show an overall improvement. During summer months, values of sea ice extent (SIE) integrated over the model domain become underestimated compared to observations, but absolute differences of mean SIE to the data are reduced in nearly all months and years. Along with the SICs, the sea ice thickness fields also become closer to observations, providing added value by the assimilation. Very sparse ocean data in the Arctic, corresponding to a very small contribution to the cost function, prevent sizable improvements of assimilated ocean variables, with the exception of the sea surface temperature.

Download Full-text

Sea Ice Assimilation into a Coupled Ocean-Sea Ice Adjoint Model of the Arctic Ocean

10.5194/tc-2017-2 ◽

2017 ◽

Author(s):

Nikolay V. Koldunov ◽

Armin Köhl ◽

Nuno Serra ◽

Detlef Stammer

Keyword(s):

Sea Ice ◽

Arctic Ocean ◽

Atmospheric Temperature ◽

Substantial Improvement ◽

Arctic Sea Ice ◽

Added Value ◽

The Arctic ◽

Small Contribution ◽

Sea Ice Extent ◽

The Arctic Ocean

Abstract. High-resolution satellite sea ice concentrations (SIC), together with several ocean parameters, are assimilated into a regional Arctic coupled ocean-sea ice model covering the period 2000–2008 using the adjoint method. There is substantial improvement in the representation of the SIC spatial distribution, in particular with respect to the position of the ice edge and to the concentrations in the central parts of the Arctic Ocean during summer months. Seasonal cycles of total Arctic sea ice area show an overall improvement. During summer months values of sea ice extent become underestimated, however, it is shown that this metric is not suitable to characterize the quality of the sea ice simulation, as the root-mean-square difference to the data is reduced in nearly all months and years. Along with the SIC the sea ice thickness fields also become closer to observations, providing added-value by the assimilation. Very sparse data in the Arctic ocean, corresponding to a very small contribution to the cost function, prevent sizable improvements of assimilated ocean variables, with the exception of the sea surface temperature. The bias between simulated and observed SIC decreases mainly due to corrections to the surface atmospheric temperature, while contributions of other control variables remain small.

Download Full-text

Sea ice and atmosphere interactions and predictability: preliminary results using HadGEM3

10.5194/egusphere-egu2020-13771 ◽

2020 ◽

Author(s):

Daniela Flocco ◽

Ed Hawkins ◽

Leandro Ponsoni ◽

Francois Massonnet ◽

Daniel Feltham ◽

...

Keyword(s):

Sea Ice ◽

Initial Conditions ◽

Complete Information ◽

Arctic Sea Ice ◽

The Arctic ◽

Sea Ice Extent ◽

Sea Level Pressure ◽

Sea Ice Thickness ◽

Climate Forecasts ◽

Ice Conditions

Arctic sea ice extent has steadily declined in the past 30 years. Aside from the global impact on climate change, regional information on the sea ice presence and on its impact on oceanic and atmospheric patterns has witnessed a growing interest. There is a growing need for seasonal-to-decadal timescale climate forecasts to help inform local communities and industry stakeholders.Here we examine the influence of sea-ice thickness observations on the predictability of the sea-ice and atmospheric circulation. We perform paired sets of ensembles with the HadGEM3 GCM starting from different initial conditions in a present-day control run. One set of ensembles start with complete information about the sea-ice conditions, and one set have degraded information. We investigate how the pairs of ensembles predict the subsequent evolution of the sea-ice, sea level pressure and circulation within the Arctic and beyond with the aim of quantifying the value of sea-ice observations for improving predictions.

Download Full-text

Presentation and evaluation of the Arctic sea ice forecasting system neXtSIM-F

The Cryosphere ◽

10.5194/tc-15-3207-2021 ◽

2021 ◽

Vol 15 (7) ◽

pp. 3207-3227

Author(s):

Timothy Williams ◽

Anton Korosov ◽

Pierre Rampal ◽

Einar Ólason

Keyword(s):

Sea Ice ◽

Initial Conditions ◽

Thickness Distribution ◽

Arctic Sea Ice ◽

The Arctic ◽

Sea Ice Concentration ◽

Sea Ice Extent ◽

Ice Concentration ◽

Forecasting System ◽

Ice Model

Abstract. The neXtSIM-F (neXtSIM forecast) forecasting system consists of a stand-alone sea ice model, neXtSIM (neXt-generation Sea Ice Model), forced by the TOPAZ ocean forecast and the ECMWF atmospheric forecast, combined with daily data assimilation of sea ice concentration. It uses the novel brittle Bingham–Maxwell (BBM) sea ice rheology, making it the first forecast based on a continuum model not to use the viscous–plastic (VP) rheology. It was tested in the Arctic for the time period November 2018–June 2020 and was found to perform well, although there are some shortcomings. Despite drift not being assimilated in our system, the sea ice drift is good throughout the year, being relatively unbiased, even for longer lead times like 5 d. The RMSE in speed and the total RMSE are also good for the first 3 or so days, although they both increase steadily with lead time. The thickness distribution is relatively good, although there are some regions that experience excessive thickening with negative implications for the summertime sea ice extent, particularly in the Greenland Sea. The neXtSIM-F forecasting system assimilates OSI SAF sea ice concentration products (both SSMIS and AMSR2) by modifying the initial conditions daily and adding a compensating heat flux to prevent removed ice growing back too quickly. The assimilation greatly improves the sea ice extent for the forecast duration.

Download Full-text

Using CloudSat snowfall rate observations to constrain and characterize the uncertainties of Arctic snow-on-sea-ice

10.5194/egusphere-egu2020-11436 ◽

2020 ◽

Author(s):

Alex Cabaj ◽

Paul Kushner ◽

Alek Petty ◽

Stephen Howell ◽

Christopher Fletcher

Keyword(s):

Sea Ice ◽

Arctic Ocean ◽

Snow Depth ◽

Arctic Sea Ice ◽

The Arctic ◽

Ice Thickness ◽

Sea Ice Thickness ◽

The Arctic Ocean ◽

Ice Model

Snow on Arctic sea ice plays multiple&#8212;and sometimes contrasting&#8212;roles in several feedbacks between sea ice and the global climate system. For example, the presence of snow on sea ice may mitigate sea ice melt by increasing the sea ice albedo and enhancing the ice-albedo feedback. Conversely, snow can inhibit sea ice growth by insulating the ice from the atmosphere during the sea ice growth season. In addition to its contribution to sea ice feedbacks, snow on sea ice also poses a challenge for sea ice observations. In particular, snow contributes to uncertainties in retrievals of sea ice thickness from satellite altimetry measurements, such as those from ICESat-2. Snow-on-sea-ice models can produce basin-wide snow depth estimates, but these models require snowfall input from reanalysis products. In-situ snowfall measurements are absent over most of the Arctic Ocean, so it can be difficult to determine which reanalysis snowfall product is best suited to be used as input for a snow-on-sea-ice model.In the absence of in-situ snowfall rate measurements, measurements from satellite instruments can be used to quantify snowfall over the Arctic Ocean. The CloudSat satellite, which is equipped with a 94 GHz Cloud Profiling Radar instrument, measures vertical radar reflectivity profiles from which snowfall rates can be retrieved. This instrument provides the most extensive high-latitude snowfall rate observation dataset currently available. CloudSat&#8217;s near-polar orbit enables it to make measurements at latitudes up to 82&#176;N, with a 16-day repeat cycle, over the time period from 2006-2016.We present a calibration of reanalysis snowfall to CloudSat observations over the Arctic Ocean, which we then apply to reanalysis snowfall input for the NASA Eulerian Snow On Sea Ice Model (NESOSIM). This calibration reduces the spread in snow depths produced by NESOSIM when different reanalysis inputs are used. In light of this calibration, we revise the NESOSIM parametrizations of wind-driven snow processes, and we characterize the uncertainties in NESOSIM-generated snow depths resulting from uncertainties in snowfall input. We then extend this analysis further to estimate the resulting uncertainties in sea ice thickness retrieved from ICESat-2 when snow depth estimates from NESOSIM are used as input for the retrieval.

Download Full-text

Sea ice representation in CMIP6 simulations with EC-Earth3-Veg

10.5194/egusphere-egu2020-20372 ◽

2020 ◽

Author(s):

Torben Koenigk ◽

Evelien Dekker

Keyword(s):

Sea Ice ◽

Emission Scenario ◽

Arctic Sea Ice ◽

The Arctic ◽

Ice Thickness ◽

First Year ◽

Data Sets ◽

Sea Ice Extent ◽

Sea Ice Thickness ◽

Arctic Sea

In this study, we compare the sea ice in ensembles of historical and future simulations with EC-Earth3-Veg to the sea ice of the NSIDC and OSA-SAF satellite data sets. The EC-Earth3-Veg Arctic sea ice extent generally matches well to the observational data sets, and the trend over 1980-2014 is captured correctly. Interestingly, the summer Arctic sea ice area minimum occurs already in August in the model. Mainly east of Greenland, sea ice area is overestimated. In summer, Arctic sea ice is too thick compared to PIOMAS. In March, sea ice thickness is slightly overestimated in the Central Arctic but in the Bering and Kara Seas, the ice thickness is lower than in PIOMAS.While the general picture of Arctic sea ice looks good, EC-Earth suffers from a warm bias in the Southern Ocean. This is also reflected by a substantial underestimation of sea ice area in the Antarctic.Different ensemble members of the future scenario projections of sea ice show a large range of the date of first year with a minimum ice area below 1 million square kilometers in the Arctic. The year varies between 2024 and 2056. Interestingly, this range does not differ very much with the emission scenario and even under the low emission scenario SSP1-1.9 summer Arctic sea ice almost totally disappears.

Download Full-text

Assessment of decadal prediction skills and sensitivity to sea-ice thickness initialization in the Arctic when seasonal ice becomes dominant in the Arctic

10.5194/egusphere-egu2020-13516 ◽

2020 ◽

Author(s):

Tian Tian ◽

Shuting Yang ◽

Pasha Karami ◽

François Massonnet ◽

Tim Kruschke ◽

...

Keyword(s):

Sea Ice ◽

Time Scales ◽

Ice Cover ◽

Climate Prediction ◽

Arctic Sea Ice ◽

Added Value ◽

The Arctic ◽

Ice Thickness ◽

Decadal Prediction ◽

Sea Ice Thickness

The Arctic has lost more than 50% multiyear sea ice (MYI) area during 1999-2017. Observation analysis suggests that if the decline of the MYI coverage continues, changes in the Arctic ice cover (i.e. area and volume) will be more controlled by seasonal ice than the effect of global warming. To investigate how large and where the source of Arctic prediction skill is given a large losses of thick MYI during the last two decades, we explore the decadal prediction skills and sensitivity to sea ice thickness (SIT) initialization from the EC-Earth3 Climate Prediction System with Anomaly Initialization (EC-Earth3-CPSAI). Three sets of ensemble hind-cast experiments following the protocol for the CMIP6 Decadal Climate Prediction Project (DCPP) are carried out in which the predictions start from: 1) a baseline system with ocean only initialization; 2) with ocean and sea ice concentration (SIC) initialization; 3) with ocean, SIC and SIT initialization. The hind-cast experiments are initialized and validated based on the ERA-Interim-reanalysis for the atmosphere and ORAS5 for ocean and sea-ice, with a focus period 1997-2016. All initialized experiments show better agreement with ORAS5 than the CMIP6 historical run (i.e. the Free run) for the first winter sea ice forecast. The SIT initialized experiments show the best skill in predicting SIT (or volume) and the added value by greatly reducing errors of near surface air temperature over the Greenland and its surrounding waters. In the Central Arctic, the Beaufort and East Siberian Seas, there are only minor differences in prediction skills on seasonal to decadal time scales between the ocean-only initialized and the SIT initialized experiments, indicating that the source of predictability in these regions are mainly from the ocean; while the ocean-only initialization degrades skill with larger RMSE than the Free run, e.g. during the ice-freezing season in the GIN and Barents Seas, or at&#160; the summer minimum in the Kara Sea, the added value from the SIT initialized experiment is present, and it may have long-term effect (>4 years) probably associated with sea-ice recirculation. In all cases, the improvement from the ocean-only initialization to also including SIC initialization is found negligible, even somehow degrading the skills. This highlights the important use of SIT in predicting changes in the Arctic sea ice cover at various time scales during the study period. Therefore, the sea-ice initialization with constraint on SIT is recommended as the most effective initialization strategy in our EC-Earth3-CPSAI for present climate prediction from seasonal to decadal time scales.

Download Full-text

Rapid decline of Arctic sea ice volume: Causes and consequences

10.5194/tc-2019-2 ◽

2019 ◽

Cited By ~ 2

Author(s):

Jean-Claude Gascard ◽

Jinlun Zhang ◽

Mehrad Rafizadeh

Keyword(s):

Sea Ice ◽

Surface Air Temperature ◽

Arctic Sea Ice ◽

The Arctic ◽

Ice Thickness ◽

Sea Ice Extent ◽

Sea Ice Thickness ◽

Ice Volume ◽

The Past ◽

Arctic Sea

Abstract. The drastic reduction of the Arctic sea ice over the past 40 years is the most glaring evidence of climate change on Planet Earth. Among all the variables characterizing sea ice, the sea ice volume is by far the most sensitive one for climate change since it is decaying at the highest rate compared to sea ice extent and sea ice thickness. In 40 years the Arctic Ocean has lost about 3/4 of its sea ice volume at the end of the summer season corresponding to a reduction of both sea ice extent and sea ice thickness by half on average. From more than 16 000 km3, 40 years ago, the Arctic sea ice summer minimum dropped down to less than 4000 km3 during the most recent summers. Being a combination of Arctic sea ice extent and sea ice thickness, the Arctic sea ice volume is difficult to observe directly and accurately. We estimated cumulative Freezing-Degree Days (FDD) over a 9 month freezing time period (September to May each year) based on ERA Interim surface air temperature reanalysis over the whole Arctic Ocean and for the past 38 years. Then we compared the Arctic sea ice volume based on sea ice thickness deduced from cumulative FDD with Arctic sea ice volume estimated from PIOMAS (Pan Arctic Ice Ocean Modeling and Assimilation System) and from the ESA CRYOSAT-2 satellite. The results are strikingly similar. The warming of the atmosphere is playing an important role in contributing to the Arctic sea ice volume decrease during the whole freezing season (September to May). In addition, the FDD spatial distribution exhibiting a sharp double peak-like feature is reflecting the Multi Y ear Ice (MYI) versus First Year Ice (FYI) dual disposition typical of the Arctic sea ice cover. This is indicative of a significant contribution from the vertical ocean heat fluxes throughout the ice depending on MYI versus FYI distribution and the snow layer on top of it influencing the surface air temperature accordingly. In 2018 the Arctic MYI vanished almost completely for the first time ever over the past 40 years. The quasi complete disappearance of the Arctic sea ice is more likely to happen in summer within the next 15 years with broad consequences for Arctic marine and terrestrial ecosystems, climate and weather patterns on a planetary scale and globally on human activities.

Download Full-text

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

Geoscientific Model Development ◽

10.5194/gmd-10-1009-2017 ◽

2017 ◽

Vol 10 (2) ◽

pp. 1009-1031 ◽

Cited By ~ 12

Author(s):

Petteri Uotila ◽

Doroteaciro Iovino ◽

Martin Vancoppenolle ◽

Mikko Lensu ◽

Clement Rousset

Keyword(s):

Sea Ice ◽

Thickness Distribution ◽

The Arctic ◽

Ice Thickness ◽

Polar Regions ◽

Intermediate Water ◽

Sea Ice Extent ◽

Sea Ice Thickness ◽

System Sensitivity ◽

Ice Model

Abstract. A set of hindcast simulations with the new version 3.6 of the Nucleus for European Modelling of the Ocean (NEMO) ocean–ice model in the ORCA1 configuration and forced by the DRAKKAR Forcing Set version 5.2 (DFS5.2) atmospheric data was performed from 1958 to 2012. Simulations differed in their sea-ice component: the old standard version Louvain-la-Neuve Sea Ice Model (LIM2) and its successor LIM3. Main differences between these sea-ice models are the parameterisations of sub-grid-scale sea-ice thickness distribution, ice deformation, thermodynamic processes, and sea-ice salinity. Our main objective was to analyse the response of the ocean–ice system sensitivity to the change in sea-ice physics. Additional sensitivity simulations were carried out for the attribution of observed differences between the two main simulations.In the Arctic, NEMO-LIM3 compares better with observations by realistically reproducing the sea-ice extent decline during the last few decades due to its multi-category sea-ice thickness. In the Antarctic, NEMO-LIM3 more realistically simulates the seasonal evolution of sea-ice extent than NEMO-LIM2. In terms of oceanic properties, improvements are not as evident, although NEMO-LIM3 reproduces a more realistic hydrography in the Labrador Sea and in the Arctic Ocean, including a reduced cold temperature bias of the Arctic Intermediate Water at 250 m. In the extra-polar regions, the oceanographic conditions of the two NEMO-LIM versions remain relatively similar, although they slowly drift apart over decades. This drift is probably due to a stronger deep water formation around Antarctica in LIM3.

Download Full-text

Improving the WRF model's simulation over sea ice surface through coupling with a complex thermodynamic sea ice model

Geoscientific Model Development Discussions ◽

10.5194/gmdd-8-10305-2015 ◽

2015 ◽

Vol 8 (12) ◽

pp. 10305-10337 ◽

Cited By ~ 1

Author(s):

Y. Yao ◽

J. Huang ◽

Y. Luo ◽

Z. Zhao

Keyword(s):

Sea Ice ◽

Wrf Model ◽

Arctic Sea Ice ◽

The Arctic ◽

Ice Thickness ◽

Cold Bias ◽

Sea Ice Thickness ◽

The Impact ◽

Ice Model ◽

Snow And Ice

Abstract. Sea ice plays an important role in the air–ice–ocean interaction, but it is often represented simply in many regional atmospheric models. The Noah sea ice model, which has been widely used in the Weather Research and Forecasting (WRF) model, exhibits cold bias in simulating the Arctic sea ice temperature when validated against the Surface Heat Budget of the Arctic Ocean (SHEBA) in situ observations. According to sensitivity tests, this bias is attributed not only to the simulation of snow depth and turbulent fluxes but also to the heat conduction within snow and ice. Compared with the Noah sea ice model, the high-resolution thermodynamic snow and ice model (HIGHTSI) has smaller bias in simulating the sea ice temperature. HIGHTSI is further coupled with the WRF model to evaluate the possible added value from better resolving the heat transport and solar penetration in sea ice from a complex thermodynamic sea ice model. The cold bias in simulating the surface temperature over sea ice in winter by the original Polar WRF is reduced when HIGHTSI rather than Noah is coupled with the WRF model, and this also leads to a better representation of surface upward longwave radiation and 2 m air temperature. A discussion on the impact of specifying sea ice thickness in the WRF model is presented. Consistent with previous research, prescribing the sea ice thickness with observational information would result in the best simulation among the available methods. If no observational information is available, using an empirical method based on the relationship between sea ice concentration and sea ice thickness could mimic the large-scale spatial feature of sea ice thickness. The potential application of a thermodynamic sea ice model in predicting the change in sea ice thickness in a RCM is limited by the lack of sea ice dynamic processes in the model and the coarse assumption on the initial value of sea ice thickness.

Download Full-text

Parameter calibration and uncertainty analysis for snow depths from the NASA Eulerian Snow On Sea Ice Model and derived sea ice thickness from ICESat-2

10.5194/egusphere-egu21-13791 ◽

2021 ◽

Author(s):

Alex Cabaj ◽

Paul Kushner ◽

Alek Petty

Keyword(s):

Sea Ice ◽

Snow Depth ◽

Arctic Sea Ice ◽

The Arctic ◽

Ice Thickness ◽

Blowing Snow ◽

Sea Ice Thickness ◽

Ice Growth ◽

Arctic Sea ◽

Ice Model

Snow on Arctic sea ice plays many, sometimes contrasting roles in Arctic climate feedbacks. During the sea ice growth season, the presence of snow on sea ice can enhance ice growth by increasing the sea ice albedo, or conversely, inhibit sea ice growth by insulating the ice from the cold atmosphere. Furthermore, estimates of snow depth on Arctic sea ice are also a key input for deriving sea ice thickness from altimetry measurements, such as satellite lidar altimetry measurements from ICESat-2. Due to the logistical challenges of making measurements in as remote a region as the Arctic, snow depth on Arctic sea ice is difficult to observationally constrain. The NASA Eulerian Snow On Sea Ice Model (NESOSIM) can be used to provide snow depth and density estimates over Arctic sea ice with pan-Arctic coverage within a relatively simple framework. The latest version of NESOSIM, version 1.1, is a 2-layer model with simple representations of the processes of accumulation, wind packing, loss due to blowing snow, and redistribution due to sea ice motion. Relative to version 1.0, NESOSIM 1.1 features an extended model domain, and reanalysis snowfall input scaled to observed snowfall retrieved from CloudSat satellite radar reflectivity measurements. In this work, we present a systematic calibration, and an accompanying estimate in the uncertainty of the free parameters in NESOSIM, targeting airborne snow radar measurements from Operation IceBridge. We further investigate uncertainties in snow depth and the resulting uncertainties in derived sea ice thickness from ICESat-2 altimetry measurements using NESOSIM snow depths.

Download Full-text