A better understanding of the role of sea ice for the changing climate of our planet is the central aim of the diagnostic Coupled Model Intercomparison Project 6 (CMIP6)-endorsed Sea-Ice Model Intercomparison Project (SIMIP). To reach this aim, SIMIP requests sea-ice-related variables from climate-model simulations that allow for a better understanding and, ultimately, improvement of biases and errors in sea-ice simulations with large-scale climate models. This then allows us to better understand to what degree CMIP6 model simulations relate to reality, thus improving our confidence in answering sea-ice-related questions based on these simulations. Furthermore, the SIMIP protocol provides a standard for sea-ice model output that will streamline and hence simplify the analysis of the simulated sea-ice evolution in research projects independent of CMIP. To reach its aims, SIMIP provides a structured list of model output that allows for an examination of the three main budgets that govern the evolution of sea ice, namely the heat budget, the momentum budget, and the mass budget. In this contribution, we explain the aims of SIMIP in more detail and outline how its design allows us to answer some of the most pressing questions that sea ice still poses to the international climate-research community.
Sea ice is both a key indicator and a driver of climatic
changes on our planet. In addition, the temporal and spatial evolution of its
coverage has important implications far beyond climatic changes: polar marine
biogeochemistry and ecosystems are closely related to the existence of sea
ice
In light of this importance of sea ice from a wide range of stakeholders, it
is sobering to see to which degree simulations of its past and future evolution
differ across large-scale coupled models
To address these issues, a specific model-intercomparison project related to
sea ice, namely the Sea-Ice Model Intercomparison Project (SIMIP), has been endorsed as an official part of the 6th phase
of the Coupled Model Intercomparison Project
This contribution describes the aim and design of SIMIP in detail. In
particular, we describe the requested sea-ice-related variables that are
included in the CMIP6 data request (see
We first outline briefly some of the most pressing research questions that we hope to address through SIMIP and place them within the CMIP6 context. We then move on to discuss in more detail the philosophy behind the design of SIMIP, which then will hopefully also allow for a consistent future extension of the SIMIP protocol as sea-ice models become more and more sophisticated. In an appendix, we specify and define in detail the current list of requested sea-ice-related variables from CMIP6 simulations, giving concrete guidance on the variable definitions to model developers and modelling centres that would like to store these variables from their simulations. The diagnostic sea-ice output should be saved as much as possible in all CMIP6 experiments that include a sea-ice-model component.
The overarching aim of SIMIP is very simple. We want to understand how sea
ice works and evolves in the coupled climate system of our planet. As
discussed by
Reflecting this line of thinking, we tend to differentiate the guiding
questions of SIMIP into three distinct sets. First, why do model simulations
differ from each other? Second, why do model simulations differ from the
observational record? Third, what can we do to reduce these differences
to obtain a better understanding of sea ice in the climate system and
eventually to achieve more realistic projections of the sea-ice evolution in
both hemispheres? These guiding questions of SIMIP address aspects of all
three science questions of CMIP6
In answering the three questions, our analysis is based on the understanding
that model simulations are never able to reflect reality, but that they can
be close enough to a reflection of reality that they become useful. As
discussed by
The disagreement among state-of-the-art climate model simulations regarding,
for example, the past and future evolution of Arctic and Antarctic sea ice is
striking and has, in particular, not reduced much from CMIP3 to CMIP5
To obtain such understanding is one of the central aims of SIMIP, following
the spirit of a previous SIMIP exercise carried out about 2 decades ago
The tuning of sea-ice models (or lack thereof) is another factor that might
explain the spread of the simulated sea-ice evolution in CMIP5 model
simulations. As part of the SIMIP protocol, we hence request documentation of
the tuning procedure of the models (see
Appendix
In addition to internal variability and tuning, the forcing of the atmosphere
and ocean influence the evolution of sea ice in model simulations,
independent of the specific quality of the sea-ice model component. For
example, in the Antarctic, the influence of biases in the mean state and
trends of the atmosphere
In combining our understanding of internal variability, the tuning of
individual models, and the influence of atmospheric- and oceanic-mean state
biases on the sea-ice simulation, we will be able to assess how much these
factors can explain differences, both in the pre-industrial control
simulation of individual models and across CMIP6 experiments, in particular
the CMIP6 historical
Due to the different characteristics and mean states in the two polar
regions, the answers we obtain regarding the reasons for model differences
might be entirely different for Arctic and Antarctic sea-ice simulations, but
will in any case allow us to identify the main aspects in which large-scale
climate models need to be improved to obtain better simulations. This
knowledge will then enable us to provide guidance for the design of
large-scale experimental programmes such as the upcoming Year of Polar
Prediction (2017–2019;
In CMIP5, model simulations for both Antarctic and Arctic sea-ice extent
differ from observations
To robustly identify model shortcomings, we will use a plausibility variable
as described by
Sea-ice models, as well as all other components of Earth system models, are usually not developed as an end in themselves. Instead, they are developed to answer specific research questions, which arise, for example, out of the curiosity of a scientist, out of the aim to most robustly increase profits in economic ventures, or out of a societal interest to allow for the development of the best future policies. Independent of the motivation of any such questions, the robustness of our answers hinges on the faith we have in our models to realistically represent the main processes relevant for a specific question. As such, the understanding of model biases or the quantification of internal variability are “just” prerequisites to then eventually give answers to the truly relevant research questions that we aim to answer.
Hence, while we could not answer these questions without the prerequisites just described, the success of SIMIP will eventually be measured against the degree to which it allows us to answer a wide variety of research questions. Many such questions will be posed and answered by scientists not directly involved in SIMIP work, but of course also within SIMIP we aim to provide answers from CMIP6 model simulations that are not possible without the detailed understanding of sea ice that the SIMIP protocol allows. These aims may sometimes simply be achieved by allowing us to more robustly identify model simulations that are more trustworthy than others, or by allowing a more robust understanding of individual processes, which can then guide additional research in narrowing down uncertainty for some of the most widely discussed questions related to sea ice.
For example, we currently have a limited understanding of the potential or
real predictability of the evolution of sea ice on timescales from weeks to
decades. In recent years, several initiatives have formed to address the
short-term predictability of sea ice (in particular, the Sea ice Prediction Network (SIPN),
Polar Climate Prediction Initiative (PCPI), and the Polar Prediction Project (PPP)). A dedicated international research programme focused
on polar prediction is coordinated through the Year of Polar Prediction
(YOPP;
We currently also have very little understanding of the long-term evolution
of sea ice in both hemispheres. For instance, CMIP5 RCP8.5 simulations show a
spread in the timing of seasonal ice-free conditions in the Arctic from 2005
to well beyond 2100
Our guiding philosophy during the design of SIMIP is to most efficiently prepare today for the sea-ice-related questions we might be asking tomorrow. This philosophy was born out of our shared frustration that many sea-ice-related questions that came up during CMIP5 or other similar exercises simply could not be answered because the necessary model output was not available. Hence in designing SIMIP, we compiled a consistent list of sea-ice-related variables that allow us to understand the main physical drivers that govern the evolution of sea ice. These are described by the conservation of heat, the conservation of momentum, and the conservation of mass. To make the description of the required sea-ice variables as easy to follow as possible, we divided the output variables into five groups, each describing a key aspect of sea-ice evolution. These groups are explained in more detail in this section. They are (1) sea-ice state variables, (2) tendencies of sea-ice mass, (3) heat and freshwater fluxes, (4) sea-ice dynamics, and (5) integrated quantities.
For each variable, we specify a priority that describes how crucial knowledge of this variable is for our understanding of sea ice. As a guiding principle in defining the priorities, we roughly grouped variables relative to the number of researchers most likely to use a given variable.
Priority 1 variables are those variables needed to quantify the large-scale evolution of sea ice or to understand the forcing of sea ice on either the ocean or the atmosphere. These variables will be used by many scientists, even those without a specific interest in sea ice per se. We hence strongly recommend that they are stored in any large-scale model simulation carried out within CMIP6 independent of the SIMIP data request, as they provide at least a basic description of the large-scale sea-ice state with a fairly light requirement on data storage.
Priority 2 variables are those needed to understand the detailed evolution of sea ice in response to external forcing. These variables will be used by researchers who want to understand in more detail which formulations of a model drive the bulk behaviour of the modelled sea ice. All priority 2 variables should readily be available from most modern sea-ice models.
Priority 3 variables are variables that are primarily helpful for scientists who develop sea-ice models, assess the detailed sea-ice-related budget, or carry out detailed comparisons of simulation results with field observations. These variables often are not used by the models in their standard calculation, so they need to be calculated specifically as diagnostics to understand the model behaviour. Also variables that are not available from some models because they do not (yet) include a specific process usually fall into priority 3.
The variables that we label as “sea-ice state variables” describe the large-scale state of sea ice. They are made up of the distribution of sea-ice mass and variables that allow us to assess the total heat content stored within the ice.
To assess sea-ice mass, the SIMIP protocol requests priority 1 variables such as sea-ice concentration in individual grid cells, sea-ice thickness, sea-ice mass per grid-cell area and variables related to the snow coverage on sea ice. At priority 3, additional variables are requested that describe more detailed properties of the simulated sea-ice cover, for example those related to the sub-grid-scale distribution of ice thickness, the amount of sea ice in ridges, or the distribution of melt ponds.
The main variable describing the amount of sea ice in large-scale model simulations used to be sea-ice volume, which usually was provided as “sea-ice volume per grid area”. Since this variable had the units “m”, it was often referred to as “equivalent thickness”, which caused confusion if researchers thought that this variable described actual thickness. However, the actual conserved quantity is sea-ice mass, which is why we have decided to prefer its storage over that of volume. In addition, we have dropped the misleading notion of “equivalent thickness”, and directly request “sea-ice volume per grid area”. This variable is partly kept for consistency with earlier CMIPs, and partly because the underlying concept of equivalent thickness remains possibly useful from an ocean perspective.
In addition to sea-ice mass, the SIMIP protocol requests the actual thickness of the simulated sea ice, averaged over the ice-covered part of the grid cell. From a sea-ice perspective, this actual thickness is geophysically more meaningful than the previous notion of equivalent thickness, as the properties of the ice cover usually depend more directly on actual thickness than on the synthetic equivalent thickness. In particular, we hope that this new definition is more intuitively meaningful and thus avoids misinterpretation of sea-ice thickness by researchers unfamiliar with the concept of equivalent sea-ice thickness.
To assess the thermal properties of sea ice, the SIMIP protocol requests at the highest priority those variables that most directly affect the atmosphere above the ice and the heat exchange between the ice and the ocean. Hence, observables such as sea-ice albedo or surface temperature are requested at priority 1. At lower priority 2, we request variables that describe in more detail the thermal state of the sea-ice cover, such as the temperature at the snow–ice interface, formally defined as “sea-ice surface temperature” by the NetCDF Climate and Forecast (CF) protocol.
Because of the central role of snow and sea-ice thickness, areal concentration, and surface temperature in describing the sea-ice evolution in coupled models, these are among the few variables that SIMIP requests at daily resolution. In contrast to CMIP5, we have added the amount of snow to the list of daily variables, since it has previously not been possible to examine the onset of surface melt at a temporal resolution of less than 1 month, which severely limited the usefulness of respective observations for model evaluation.
To understand the change of sea-ice mass in different models, SIMIP requests variables quantifying the physical cause and location of ice growth and melt. They are all requested at priority 2, and capture both the areal and the thickness evolution of the ice cover. Among others, the evolution of ice thickness through lateral growth or melt, through bottom growth or melt, through surface melt, snow–ice formation, etc., are requested as part of this variable group. Details of the processes causing changes in snow mass are also included, i.e. advection, snowfall, or snowmelt. These variables then allow one to identify at first order the physical processes that change the sea-ice and snow mass in different models. This in turn allows one to more robustly identify the underlying processes that need to be better represented by individual models to better capture the observed evolution of the ice cover.
Only quantifying where the sea-ice mass is changing does not allow us to understand why the sea-ice mass is changing the way it does. For such analysis, atmospheric and oceanic heat fluxes that affect the sea-ice cover are required, and SIMIP requests these fluxes at priority 2. In the past, individual atmospheric fluxes over sea ice have not been available, since usually only their grid-cell average, including the ice-free part of the grid cell, was recorded. This then made it difficult to assess how the fluxes developed over the sea-ice covered part of the grid cell, since a change in the net fluxes could also simply have been a reflection of changing sea-ice concentration. Therefore within SIMIP, we specifically request all fluxes over the ice-covered part of the grid cell. Individual atmospheric fluxes are often only available on the atmospheric grid, with the sea-ice model only receiving an averaged bulk flux interpolated to the oceanic grid. This averaged flux then does not allow for a detailed analysis of the heat budget. Since we do not specify on which grid the individual fluxes should be provided, modelling centres are free to simply provide the individual atmospheric fluxes on the atmospheric grid where they are readily available in any case.
In addition to the heat fluxes, an analysis of sea-ice mass changes must also take relevant freshwater fluxes into account, e.g. rainfall or freshwater flux at the bottom of the ice from sea-ice phase changes. To examine the salt release from sea ice and its possible impact on ocean circulation, SIMIP also requests storage of the sea-ice-related salt flux.
While the variables requested so far allow researchers to analyse in great detail the change of sea-ice mass in different models and in any given grid cell, the movement of sea ice is also of central importance to understand the evolution of Arctic sea ice as simulated by large-scale climate models. Hence, SIMIP requests variables that describe this movement and that allow researchers to understand its driving forces in CMIP6 simulations.
At the highest priority 1, the SIMIP data request includes for sea-ice
dynamics the
At priority 2, SIMIP requests more detailed variables describing sea-ice mass
transports and the integrated forces on the ice cover. Hence, at this
priority the data request includes the
To allow for an even deeper understanding of the driving forces of sea-ice movement in large-scale model simulations, at priority 3, we request the individual terms of the force balance that determine the movement of the ice. These terms include for example the sea-surface tilt term, the Coriolis force, and the internal stresses. Because we only request those variables as monthly means, a true closure of the momentum balance will not be possible. However, the monthly mean values of these terms already allow one to identify the key processes that give rise to differences in CMIP6 sea-ice model simulations, which then in turn will hopefully allow for improvements of the sea-ice dynamics in these models.
The final group of variables requested within SIMIP are primarily a service
to the research community, consisting of integrated quantities so often used
in studies examining the evolution of sea ice that we felt it useful to make
these quantities readily available: total sea-ice area, sea-ice volume and
sea-ice extent for the Northern and Southern hemispheres. In addition, we
request the areal and mass fluxes of sea ice through the four main outlets
of the Arctic Ocean, namely Fram Strait, Bering Strait, the Barents opening,
and the Canadian Arctic Archipelago. Note that for studies of sea-ice coverage, care
must be taken in using the non-linear diagnostic sea-ice extent
As outlined, SIMIP uses model simulations to understand the sea-ice evolution
of the real world. This is done by linking the model simulations to the real
world through observations. Hence, SIMIP would not be possible without a
reliable, wide range of observational records that allow one to understand if
CMIP6 model simulations capture the most important aspects of sea-ice
evolution as it also occurs in reality. For this reason, SIMIP sees itself not only as a pure model-intercomparison exercise, but also as a forum for
identifying the best possible use of observations for the evaluation and
improvement of model simulations. For this purpose, SIMIP works closely with
the National Snow and Ice Data Center (NSIDC), with other data centres and
with the observational community to maintain a detailed, up-to-date list of
sea-ice-related observational records that can be used by any researcher to
analyse the performance of sea-ice model simulations. We did not consider it
useful to provide a snapshot of this list here, given that it is quickly
evolving and expanding. For details, please see
SIMIP is a CMIP6-endorsed diagnostic MIP
All model output requested by SIMIP will be distributed through the Earth
System Grid Federation (ESGF) with digital object identifiers (DOIs)
assigned. As in CMIP5, the model output will be freely accessible through
data portals after registration. In order to document CMIP6's scientific
impact and enable ongoing support of CMIP, users are obligated to acknowledge
CMIP6, the participating modelling groups, and the ESGF centres (see details
on the CMIP Panel website at
In this appendix, we outline the SIMIP data request version 1.0 and the
related request for model documentation. To account for possible long-term
adjustments of this request, also for studies beyond CMIP6, an online version
of this request is available at
In addition to the variable request, SIMIP requests that modelling groups provide documentation of their sea-ice model that allows for a better interpretation of sea-ice simulations from individual CMIP6 models. For example, an assessment of the model simulations based on the use of thickness distributions or diagnostic salinity can provide insights into whether models that include such advanced properties generally perform better than models that do not. Basic sea-ice model documentation was requested in past CMIPs, and SIMIP is contributing additional requests to fill gaps and include requests related to the SIMIP data request for CMIP6 (personal communication, Bryan Lawrence, March 2016).
We encourage all modelling groups to provide the requested model documentation
information with as much detail as possible, and all scientists working with
CMIP6 sea-ice output to make use of the model documentation to enhance the
analysis of the sea-ice simulations. The specific documentation requirements
will be distributed through the Earth System Documentation activity (ES-DOC;
see
As groups prepare the model output following the SIMIP request, the questions below should be kept in mind so they can easily be contributed to the upcoming more detailed request for sea-ice model documentation.
Which combination of terms closes the mass and energy budgets in your model? Is your model missing any processes that relate to the requested sea-ice variables? Were any assumptions made in the calculation of diagnostic sea-ice variables we requested? If so, which ones? Does sea ice salinity impact the thermal properties of sea ice? (yes or no) Does your model use two different salinities for thermodynamic calculations and for the salt budget? (yes or no) Is the salinity used for ice–ocean exchanges variable or a constant? If constant, what is this constant? What kind of ice thickness distribution is used (if any)? How many categories and what are the category limits? How is the heat content of precipitation handled in your sea-ice model?
As discussed in Sect.
Which tuning knobs were used in tuning the sea-ice model? In what kind of simulations was the tuning done (control, transient 20th and 21st century, all of these)? What were the sea ice targets in the tuning effort?
In the following appendices, we list all variables of the SIMIP data request version 1.0, grouped into the following five categories: (1) sea-ice state variables, (2) tendencies of sea-ice mass, (3) heat and freshwater fluxes, (4) sea-ice dynamics, and (5) integrated quantities.
Regarding the spatial storage of variables, for simplicity we request all
variables to be stored on the model grid on which they actually are used
during a model simulation. For most variables that SIMIP asks for, this will
usually be the ocean grid. However, for example, individual atmospheric fluxes
over the sea-ice covered part of a grid cell are often only evaluated in the
atmosphere model, and should then simply be stored on the atmospheric grid.
Note that analogous to the CMIP6-endorsed OMIP
Fractional coverages are evaluated either relative to the entire grid cell (e.g. for sea-ice concentration) or relative to the sea-ice covered part of the grid cell (e.g. melt-pond area fraction). This is specified in detail in the following request. As a general rule, we have tried to follow the most widespread standard by which individual fractional coverages are usually represented in sea-ice models.
Temporal averages are requested for almost all variables, except for a few variables related to sea-ice dynamics that must be stored as instantaneous values at some point during the averaging period. The standard averaging period for all variables is 1 month. For the most important variables, additional daily averages are requested, as identified in the following detailed data request.
For all variables that are proportional to area fraction, i.e. extensive variables such as volume, mass, or area fraction, a zero should always be averaged in for all time steps where no sea ice is present. This is because the extensive variables naturally approach zero as area fraction approaches zero.
For all variables that are not proportional to area fraction, i.e. intensive variables such as albedo, temperature, or heat flux, SIMIP requests the area-weighted average. Hence, all time samples with non-zero sea-ice fraction are first multiplied by area fraction, then summed, and then divided by the sum of the area fractions. Ice-free grid cells at any point throughout the averaging period should be treated as missing values and the averaging should only be carried out for those periods where sea ice is present. For continuously ice-free grid cells, missing values should be reported. This is because the intensive variables do not necessarily approach zero as area fraction approaches zero.
To report grid-cell averages for multi-category models, the properties of the individual categories should be averaged to a single value for each time step by calculating the area-weighted average across all categories. The single value thus obtained for each time step should then be used for all further processing of model output. The only exception to this rule are variables that specifically ask for values for individual categories.
The most fundamental set of variables that SIMIP requests are those variables that describe the actual state of the sea-ice cover. This set of variables allows one to examine, for example, how much sea ice there actually is in a certain region, how thick the sea-ice cover is, whether there is snow on sea ice, and how densely the sea-ice cover is packed. Knowledge of the temporal evolution of these parameters then allows one to examine, for example, the seasonal cycle of the sea-ice cover or its long-term evolution. In addition, SIMIP also requests some fundamental thermodynamic quantities such as the surface temperature of the ice or the heat content of the ice cover, which allows one to close the energy budget of sea ice to leading order. Because of the fundamental nature of these quantities, many of them are requested at priority 1. At priority 2, SIMIP requests variables that are slightly less central, but still relevant for many researchers, for example surface albedo or heat content of the ice cover. At priority 3, finally, SIMIP requests variables that will only be available from advanced model formulations that include for example a melt-pond scheme or a scheme to interactively calculate the bulk salinity of the ice cover.
Fraction of time steps of the averaging period during which sea ice is
present (siconc
Areal fraction of a given grid cell that is covered by sea ice, independent of the thickness of that ice. By definition, this variable can only have values between 0 (no sea ice at all) and 1 (fully covered by sea ice). Requested as daily and monthly average.
Total mass of sea ice divided by the entire area of a grid cell. Mass is the truly conserved quantity, so we prioritise requesting sea-ice mass over requesting sea-ice volume.
Thickness of sea ice averaged over the ice-covered part of a given grid cell. This variable hence describes the actual thickness of the sea ice, which in the context of the heat budget is for example necessary to analyse the heat flux through the ice. From a sea-ice perspective, this real (or floe) thickness is a more meaningful variable to store than the so-called equivalent sea-ice thickness that was used in previous CMIPs, which is defined as the sea-ice volume divided by the area of the entire grid cell. While use of equivalent thickness (i.e. volume) is useful from the perspective of an ocean model, it is not meaningful in a sea-ice context. Indeed, in our experience users often assumed that such equivalent thickness was the actual thickness of the sea ice, which is why we request the actual sea-ice thickness as a new variable for CMIP6. It should be directly accessible in any sea-ice model, but can otherwise be calculated by dividing the total sea-ice volume by the sea-ice area. Requested as daily and monthly average.
Area fraction of the sea-ice surface that is covered by snow. In many models that do not explicitly resolve an areal fraction of snow, this variable will always be either 0 or 1.
Total mass of snow on sea ice divided by the entire area of a grid cell. This then allows one to analyse the storage of latent heat in the snow, and to calculate the snow-water equivalent.
Thickness of snow averaged over the snow-covered part of the sea ice. It hence describes the actual thickness of the snow, which in the context of the heat budget is for example necessary to analyse the heat flux through the ice and snow. This thickness is usually directly available within the model formulation. It can also be derived by dividing the total volume of snow through the area of the snow. Requested as daily and monthly average.
Mean surface temperature of the sea-ice covered part of the grid cell. Wherever snow covers the ice, the surface temperature of the snow is used for the averaging, otherwise the surface temperature of the ice is used. Requested as daily and monthly average.
This is also known as the equivalent thickness of sea ice, which is calculated by dividing the volume of sea ice by the entire grid area. This measure used to simply be called ice thickness in previous CMIPs, which gave rise to some confusion for users expecting this variable to describe actual thickness. Since ice mass is more general than volume, this variable is somehow obsolete and primarily part of SIMIP for these historical reasons.
Reports surface temperature of ice where snow thickness is zero.
Reports temperature at interface, NOT temperature within lowermost sea-ice model layer.
Age of sea ice since its formation in open water.
Mean surface albedo of entire ice-covered part of grid cell.
Mean height of sea-ice surface (i.e. snow-ice interface when snow covered) above
sea level. This follows the classical definition of freeboard for in situ
observations. In the satellite community, sometimes the total height of sea
ice and snow above sea level is referred to as freeboard. This can easily be
calculated by adding
Heat content of all ice in grid cell divided by total grid-cell area. This
includes both the latent and sensible heat content contribution. Water at
0
Heat-content of all snow in grid cell divided by total grid-cell area. This
includes both the latent and sensible heat content contribution. Snow-water
equivalent at 0
Area fraction of grid cell covered by each ice-thickness category (vector with one entry for each thickness category starting from the thinnest category, netcdf file should use thickness bounds of the categories as third coordinate axis).
Actual (floe) thickness of sea ice in each category (NOT volume divided by grid area; vector with one entry for each thickness category starting from the thinnest category, netcdf file should use thickness bounds of categories as third coordinate axis).
Area fraction of grid cell covered by snow in each ice-thickness category (vector with one entry for each thickness category starting from the thinnest category, netcdf file should use thickness bounds of the categories as third coordinate axis).
Actual thickness of snow in each category (NOT volume divided by grid area; vector with one entry for each thickness category starting from the thinnest category, netcdf file should use thickness bounds of categories as third coordinate axis).
Total mass of all salt in sea ice divided by grid-cell area. Sometimes, models implicitly or explicitly assume a different salinity of the ice for thermodynamic considerations than they do for closing the salt budget with the ocean. In these cases, the total mass of all salt in sea ice should be calculated from the salinity value used in the calculation of the salt budget.
Area fraction of sea-ice surface that is covered by melt ponds.
Mass of water in melt ponds divided by area of sea ice.
Volume of refrozen ice on melt ponds divided by melt-pond-covered area.
Area fraction of sea-ice surface that is covered by ridged sea ice.
Total volume of ridged sea ice divided by area of ridges.
Mean sea-ice salinity of all sea ice in grid cell. Sometimes, models implicitly or explicitly assume a different salinity of the ice for thermodynamic considerations than they do for closing the salt budget with the ocean. In these cases, the mean salinity used in the calculation of the salt budget should be reported
While the sea-ice state variables already allow one to calculate how much the amount of sea ice in a certain grid cell is changing from one averaging period to the next, such an estimate does not allow one to infer why precisely the sea-cover is changing the way it does, i.e. whether a given mass change is driven by dynamics or by thermodynamics. Therefore, SIMIP requests a rather detailed list of variables that describe where the amount of sea ice is changing. These variables are all requested at priority 2 since they are usually used for the in-depth analysis of the simulated sea-ice cover. These variables should all be readily available from any modern sea-ice model. All tendencies are negative for decreasing mass.
Total change in sea-ice area fraction through thermodynamic processes.
Total change in sea-ice area fraction through dynamics-related processes (advection, divergence, etc.).
Total change in sea-ice mass from thermodynamic processes divided by grid-cell area.
Total change in sea-ice mass through dynamics-related processes (e.g. advection) divided by grid-cell area.
The rate of change of sea-ice mass due to sea ice formation in supercooled
water (often through frazil formation) divided by grid-cell area. Together,
The rate of change of sea-ice mass due to vertical growth of existing sea ice
at its base divided by grid-cell area. Note that this number is always
positive or zero since sea-ice melt is collected in
The rate of change of sea-ice mass due to transformation of snow to sea ice divided by grid-cell area. Always positive or zero.
The rate of change of sea-ice mass change through evaporation and sublimation
divided by grid-cell area. If a model does not differentiate between the
sublimation of snow and sea ice, we recommend to report
The rate of change of sea-ice mass through melting at the ice surface divided by grid-cell area. This number is independent of the actual fate of the meltwater, and will hence include all sea-ice meltwater that drains into the ocean and all sea-ice meltwater that is collected by a melt-pond parameterisation. Always negative or zero.
The rate of change of sea-ice mass through melting/dissolution at the ice
bottom divided by grid-cell area. Note that this number is always zero or
negative since sea-ice growth is collected in
The rate of change of sea-ice mass through lateral melting/dissolution divided by grid-cell area (report 0 if not explicitly calculated thermodynamically). Always negative or zero.
Mass of solid precipitation falling onto sea ice divided by grid-cell area. Always positive or zero.
The rate of change of snow mass through melt divided by grid-cell area. Always negative or zero.
The rate of change of snow mass through sublimation divided by grid-cell
area. If a model does not differentiate between the sublimation of snow and
sea ice, we recommend to report all sublimation within
The rate of change of snow mass through advection with sea ice divided by grid-cell area.
The rate of change of snow mass due to transformation of snow to sea ice divided by grid-cell area. Always negative or zero.
The rate of change of snow mass due to wind-driven transport into the ocean.
To understand the drivers of the sea-ice tendencies introduced in the previous section, SIMIP requests storage of the actual heat fluxes that gave rise to the tendencies in sea-ice mass. In contrast to previous CMIPs, all fluxes are to be evaluated over the ice-covered part of the grid cell. In earlier CMIPs, the fluxes were usually just provided as a grid-cell average, including the fluxes over open water. This made it impossible to analyse the heat budget of sea ice, which is why SIMIP now requests these fluxes over sea ice. In many coupled models, the individual fluxes over sea ice are only available on the atmospheric grid since only integrated net fluxes are passed on to the ocean model. In this case, the fluxes over sea ice should simply be stored on the atmospheric grid.
In addition to the analysis of the heat fluxes, also the freshwater fluxes are an important driver for the interaction of sea ice with the climate system of the Earth. To understand the magnitude of these freshwater fluxes, SIMIP requests storage of both the salt flux and of the freshwater flux from sea-ice melt or growth.
The sign convention is generally positive downward. However, to remain consistent with the NetCDF Climate and Forecast (CF) convention, upward fluxes that carry the term “upward” in their name are positive upward, as detailed in the following.
The downwelling shortwave flux from the atmosphere to the sea-ice surface. Always positive or zero.
The upward shortwave flux from the sea-ice surface to the atmosphere. Always positive or zero.
The downwelling shortwave flux underneath sea ice, i.e. the amount of shortwave radiation that penetrates the sea ice. Always positive or zero.
The downwelling longwave flux from the atmosphere to the sea-ice surface. Always positive or zero.
The upward longwave flux from the sea-ice surface to the atmosphere. Always positive or zero.
The net sensible heat flux over sea ice. Positive for a downward heat flux.
The net latent heat flux over sea ice. Positive for a downward heat flux.
The net sensible heat flux under sea ice from or to the ocean. Per sign convention, heat from the ocean is counted as negative since it describes an upward heat flux.
The net heat conduction flux at the ice surface, i.e. the conductive heat flux from the centre of the uppermost vertical sea-ice grid box to the surface of the sea ice. Positive for a downward heat flux.
The net heat conduction flux at the ice base, i.e. the conductive heat flux from the centre of the lowermost vertical sea-ice grid box to the bottom of the sea ice. Positive for a downward heat flux.
Mass of liquid precipitation falling onto sea ice divided by grid-cell area.
If the rain is directly put into the ocean, it should not be counted towards
Total flux of salt from water into sea ice divided by grid-cell area; salt flux is upward (negative) during ice growth when salt is embedded into the ice and downward (positive) during melt when salt from sea ice is again released to the ocean.
Total flux of fresh water from water into sea ice divided by grid-cell area; This flux is negative during ice growth (liquid water mass decreases, hence upward flux of freshwater), positive during ice melt (liquid water mass increases, hence downward flux of freshwater).
Total flux of fresh water from sea-ice surface into underlying ocean. This combines both surface meltwater that drains directly into the ocean and the drainage of surface melt ponds. By definition, this flux is always positive.
The variables that SIMIP requests for sea-ice dynamics are needed to understand how and why sea ice moves horizontally. The most important parameters there are obviously the actual velocities, which are requested on the native model grid. In addition, the sea-ice speed is requested, which allows one to account for possible back-and-forth movement of the ice during the averaging period. Because of their importance, these variables are requested at priority 1 and at daily resolution.
At lower priority, SIMIP requests primarily those variables that allow one to examine the various forces that are responsible for the actual sea-ice movement. These are at priority 2 the total atmospheric and the total oceanic stress and the strength of the sea-ice cover. At priority 3, further details on the actual forces are requested.
Note that four variables are requested as instantaneous values. These are the divergence and the maximum shear of the sea-ice velocity field, and the average normal stress and the maximum shear stress in the sea ice. These variables can only usefully be analysed if stored simultaneously as instantaneous values at some point during the averaging period.
Speed of ice (i.e. mean absolute velocity) to account for back-and-forth
movement of the ice during the average period. Such change of direction
during the averaging period may reduce the individual vector quantities
Includes transport of both sea ice and snow by advection.
Includes transport of both sea ice and snow by advection.
Computed strength of the ice pack, defined as the energy (J m
Divergence of sea-ice velocity field (first shear strain invariant). Requested as instantaneous value.
Maximum shear of sea-ice velocity field (second shear strain invariant). Requested as instantaneous value.
Atmospheric drag coefficient that is used to calculate the atmospheric momentum drag on sea ice.
Oceanic drag coefficient that is used to calculate the oceanic momentum drag on sea ice.
Average normal stress in sea ice (first stress invariant). Requested as instantaneous value.
Maximum shear stress in sea ice (second stress invariant). Requested as instantaneous value.
Much of the analysis of sea ice in climate research is concerned with
integrated measures such as total hemispheric sea-ice area or sea-ice volume.
Within SIMIP, we hence consider it useful to have these measures directly
available, and request them at priority 2. Note that care should be taken in
the use of sea-ice extent, because its non-linear behaviour can cause
substantial artefacts both regarding the spatial and the temporal evolution
of the sea-ice cover
Total area of sea ice in the Northern Hemisphere.
Total area of sea ice in the Southern Hemisphere.
Total volume of sea ice in the Northern Hemisphere.
Total volume of sea ice in the Southern Hemisphere.
Total area of all Northern Hemisphere grid cells that are covered by at least 0.15 areal fraction of sea ice.
Total area of all Southern Hemisphere grid cells that are covered by at least 0.15 areal fraction of sea ice.
Net (sum of transport in all directions) sea-ice-mass transport through the
following four passages, positive into the Arctic Ocean. Note that the
definitions of the passages are for SIMIP purposes just meant as default
values as given by the physical ocean MIP described in Fram Strait: (11.5 Canadian Arctic Archipelago: (128.2 Barents opening: (16.8 Bering Strait: (171
Net (sum of transport in all directions) sea-ice-area transport through the
following four passages, positive into the Arctic Ocean. Note that the
definitions of the passages are for SIMIP purposes just meant as default
values as given by the physical ocean MIP described in Fram Strait: (11.5 Canadian Arctic Archipelago: (128.2 Barents opening: (16.8 Bering Strait: (171
Net (sum of transport in all directions) snow mass transport through the
following four passages, positive into the Arctic Ocean. Note that the
definitions of the passages are for SIMIP purposes just meant as default
values as given by the physical ocean MIP described in Fram Strait: (11.5 Canadian Arctic Archipelago: (128.2 Barents opening: (16.8 Bering Strait: (171
The following table contains a summary of the variable request. It lists for
each variable the requested averaging method (intensive, extensive, or
instantaneous), the priority and the output frequency. A possible updated
version of this table is available at
We thank the international sea-ice community for their input and guidance during the design of SIMIP, most notably during the Sea Ice and Climate Modelling workshop in Reading on 26th September 2014. We are very grateful to Martin Juckes for his tremendous help and support throughout our work on the design of SIMIP, and to David Bailey, Helmuth Haak, and Alex West for their helpful comments and feedback. We thank WCRP CliC for their logistical and financial support, which was essential for SIMIP. The article processing charges for this open-access publication were covered by the Max Planck Society. Edited by: S. Valcke Reviewed by: D. Bailey and one anonymous referee