As already noted in the introduction and references within, the ocean plays a vital role in the climate system by absorbing heat and carbon dioxide, regulating global temperatures, and influencing weather patterns. As the main source of uncertainty in seasonal to decadal (i.e., near-term) projections, the internal variability of the climate system often stems from ocean processes: ENSO, the Pacific Decadal Oscillation (PDO), the Atlantic Meridional Variability (AMV/AMO), etc. (IPCC, 2021b, Annex IV). These modes of variability involve changes in ocean heat and salt content, sea ice properties, as well as transport by ocean currents at all depths. In this context, the goal of this opportunity is to continue efforts started in previous CMIP phases to quantify the processes that drive ocean variability and change and to provide understanding and a more robust assessment of climate projections (e.g., Griffies et al., 2016; Orr et al., 2017). In addition, this opportunity aims to better coordinate modeling efforts and comparison across models, for example, through improved grid specifications, as well as investigating the impacts of changes in oceanic properties on the global climate system.

The Meridional Overturning Circulation (MOC) is an important aspect of climate and an active part of the oceanic response to climate change. The Atlantic Meridional Overturning Circulation (AMOC) is projected to decline during this century (Fox-Kemper et al., 2021a), and a potential future collapse would imply a dramatic climate shift with enormous global and regional impacts (e.g., Zhang et al., 2019; Bellomo and Mehling, 2024). Many processes and feedback that govern AMOC are still debated, for example the role of warming versus freshwater forcing (Wen et al., 2023), input from Arctic sea ice and ice sheet melting (He and Clark, 2022), and the role of deep convection in different regions (Menary et al., 2020). Furthermore, the Southern Ocean MOC also plays a key role in sequestering heat and carbon (Williams et al., 2023). Quantifying and understanding the processes governing deep water formation and upwelling requires a full-depth analysis, considering the strong mesoscale variability (Morrison et al., 2016; Hewitt et al., 2020; Jackson et al., 2020). Within this opportunity, MOC-related variables are clustered together in the variable group called ocean_meridional_overturning_streamfunctions to facilitate a better understanding of the changes in the MOC and their potential drivers.

Other processes and mechanisms prioritized in this opportunity include the relationship between changes in the ocean and polar processes and mechanisms, which remains poorly understood. For example: To what extent is ocean warming contributing to the melting of Arctic sea ice (Dörr et al., 2024) and ice shelves (Slater and Straneo, 2022)? Will deep convection migrate northward, accelerating changes to the cryosphere (Heuzé and Liu, 2024)? Will the Beaufort Gyre collapse and release its stored freshwater (Timmermans and Marshall, 2020)?

Oceanic climate change throughout the ocean basins is addressed by this opportunity. How will western boundary currents and gyres respond to climate change (Sen Gupta et al., 2021)? How will eastern boundary currents and upwelling respond (Bograd et al., 2023)? Globally, the three-dimensional dynamics of the ocean not only control thermosteric and halosteric sea level changes at regional scales (Griffies et al., 2014; Fox-Kemper et al., 2021a), but also manometric sea level changes in shallow oceans (Samanta et al., 2024; Jevrejeva et al., 2024). These sea level drivers will affect the future of human activities in coastal regions across the globe.

To achieve these goals, the Ocean Changes, Drivers and Impacts Opportunity requests the datasets necessary to analyze three-dimensional ocean processes and their time evolution, as well as their feedbacks on the atmosphere and cryosphere, going beyond those defined as baselines in Juckes et al. (2025). Most of these diagnostics and the corresponding variables were defined and some were introduced in the contribution of OMIP to CMIP6 (Griffies et al., 2016; Orr et al., 2017). The main variable groups attached to this opportunity are inherited from Griffies et al. (2016), but have here been prioritized differently based on the experience gained from CMIP6 and a desire to reduce the size of the data requested: for example, the scalar fields in Table H1 of Griffies et al. (2016) have been split into two variable groups: omip_scalar_high_priority and omip_scalars_low priority. The extensive list of variables provided by Griffies et al. (2016) covers most of the needs of this opportunity across different types of ocean models, providing continuity between CMIP6 and CMIP7 as ocean models evolve. An additional OMIP paper including further developments and refinements impacting ocean variables and the new OMIP protocols is underway – (Baylor Fox-Kemper, personal communication, 2025). A novelty in CMIP7 is the ocean_mesoscale variable group, which contains variables essential for analyzing the output of eddying ocean models. In models with a horizontal grid-spacing of 1/4° or finer, ocean eddies are no longer entirely parameterized but largely resolved, requiring additional variables for heat and salt transport to accurately compute their contribution to the heat and salt budgets. Given that mesoscale eddy activity has changed throughout the historical period (Martínez-Moreno et al., 2021) and is projected to continue evolving in the future (Beech et al., 2022), being able to better quantify these changes will be instrumental to more robustly assess climate simulations. Finally, some new variables (in the variable group int_ocean_budgets) include vertically-integrated heat and salt content, intended to more easily track the large-scale changes in the energy and freshwater cycles and compare to observations, and thereby to inform the energy budgeting of the whole Earth system. While these variables can be calculated instantaneously from the three-dimensional temperature, salinity, and grid specifications, the increasing complexity of the vertical coordinates used in modern models makes this task complex. Furthermore, because of the limits of observing technology, ranges of depth easily accessed by bathythermograph, Argo floats, and other tools have made for a standard set of layers based on hydrostatic pressure ranges in the observational literature (0–300, 0–700, 0–2000 m and total depth, where meter ranges imply their hydrostatic pressure equivalents). These new variables, calculated online, will facilitate the evaluation of ocean models using in-situ observations (e.g., Eyring et al., 2021).

Sea ice plays a role of prime importance in the climate system because of its widespread coverage and very different physical properties compared to the ocean and atmosphere. Often compared to a white blanket, sea ice reflects solar radiation back to space and acts as a thermal insulator by drastically reducing the turbulent heat fluxes from the warmer ocean to the colder atmosphere during winter (Zampieri et al., 2024) and mitigating the warming of the ocean by absorbing heat during the polar summers (Li and Liu, 2022). Sea ice evolves because of thermodynamic processes, that change its mass and heat content, and as a result of so-called dynamic processes, responsible for its drift and deformation. Sea ice is not only a mediator of atmospheric-oceanic heat exchange but also of the transfer of momentum from the atmosphere to the ocean (i.e., stresses). The seasonal formation of new sea ice and consequent rejection of salty brine into the ocean reduces the stability of the underlying water column and contributes to forming the world's densest waters. These feed the ocean thermohaline circulation on a global scale (Fox-Kemper et al., 2021a). In contrast, the melting of sea ice freshens the surface ocean and has a stabilizing effect on its vertical stratification (Linders and Björk, 2013). Sea ice is not a continuous rigid plate, but rather composed of a dynamic ensemble of floes in non-uniform motion, with varying thickness and sizes that span over several orders of magnitude (Gherardi and Lagomarsino, 2015), making its accurate representation in climate model simulations challenging.

Changes in sea ice have significant impacts on atmosphere, ocean, marine biogeochemistry, ecosystems, and human activities. In recent decades, the Arctic sea ice cover has been declining rapidly, and the signal of a forced sea ice retreat has clearly emerged from the background noise of year-to-year variability (Notz and SIMIP Community, 2020). In contrast, the Antarctic sea ice area showed a small positive trend over 1979–2015, but this was followed by sudden decreases in recent years (Eayrs et al., 2021), which suggests that the Antarctic might be rapidly transitioning to a new, low sea ice state (Purich and Doddridge, 2023; Hobbs et al., 2024) because of thermodynamic processes (Himmich et al., 2024). Given the importance of sea ice in the climate system and the swift transitions currently occurring in the polar regions, it is essential to advance our understanding of past and future sea ice changes, their driving mechanisms, and their impacts.

Ahead of CMIP6, the SIMIP Community developed a protocol detailing a standard for sea ice model output to streamline, and hence simplify the analysis of the simulated sea ice evolution in model inter-comparison projects (Notz et al., 2016). This protocol allowed researchers to conduct process-level analysis of the three main budgets that cover the evolution of sea ice, namely the heat, momentum and mass budgets (e.g., Keen et al., 2021; Watts et al., 2021; Zanowski et al., 2021; Lee et al., 2023; Frankignoul et al., 2024; Kuang et al., 2024). Notwithstanding this massive effort and the success it has enabled in sea ice modeling studies, simulations of past and future sea ice evolution from CMIP6 models still exhibit a large inter-model spread and fail at capturing important sensitivities, including the response of sea ice area to global mean temperature change (Notz and SIMIP Community, 2020; Roach et al., 2020).

To achieve a process-based assessment of the sea ice evolution, we need variables to diagnose the state of Arctic and Antarctic sea ice, understand the mechanisms driving changes, and assess impacts (Table 3). The Sea Ice Changes, Drivers and Impacts Opportunity seeks to provide the necessary model outputs to reproduce and build on the work done with CMIP6 simulations, leading to a better understanding and, ultimately, reducing biases and errors in the simulation of sea ice. The variable groups were created mainly following Appendix D–H of Notz et al. (2016), which grouped variables based on the following categories: sea ice state variables, tendencies of sea ice mass, heat and freshwater fluxes, sea ice dynamics, and integrated quantities. The prioritization of each variable group was based on past usage/download of the variables included in the group by the SIMIP Community as well as considerations related to the amount of data produced. In addition, the seaice_gcos_ecv variable group was created to align with the seven essential climate sea ice variables defined by the Global Climate Observing System (GCOS; Lavergne et al., 2022) to facilitate the evaluation of model output against observational products. Similar to what was done in CMIP6, it was decided to not include sea ice albedo as part of this data request because of the difficulty in standardizing its calculation, and instead users are encouraged to compute the ratio of outgoing shortwave radiation over the incoming shortwave radiation as this is the only accurate way to calculate a broadband albedo that would be consistent with observation-based calculations. Finally, a few new variables were added compared to CMIP6, namely the effective melt pond fraction to allow for direct comparison with observations, as well as the integrated mass of snow on sea ice for each hemisphere to allow for a complete high-level analysis of the total mass of the cryosphere in the Earth system (see Appendix B for more detail).

In paleoclimate research, the integration of modeling (e.g., Sherriff-Tadano and Klockmann, 2021) and model-independent data from geologic, glaciologic, and marine climate archives (e.g., Gulev et al., 2021; Fox-Kemper et al., 2021a; Walter et al., 2023) is possible, providing an opportunity for testing the skill and robustness of climate models and improving confidence in projections of future climate change (e.g., Masson-Delmotte et al., 2013; Haywood et al., 2019; Kageyama et al., 2024). The study of paleoclimates provides a critical window into climate conditions that differ significantly from the present, including periods with higher global temperatures and different atmospheric CO₂ concentrations. For example, about 3 million years ago, a much warmer than present Arctic supported large herbivores at atmospheric CO₂ levels comparable to today (Rybczynski et al., 2013; de la Vega et al., 2020). Such paleoclimate reconstructions offer verification data, provide a test for climate model performance under warm climates bearing similarity to future projections, and highlight the exceptional nature of current anthropogenic climate change.

The goal of this opportunity is to leverage paleoclimate records and experiments to evaluate model performance across a broad range of climate states beyond the instrumental record and to reflect on thresholds in the Earth system under conditions that are extremely different from today. This includes quantifying uncertainties and model-discord (e.g., Kageyama et al., 2021) and identifying potential biases in model simulations, particularly for extreme or non-analog conditions. Paleoclimate simulations allow researchers to assess the capability of models to reproduce reconstructed climatic features such as polar amplification, flat meridional temperature gradients, or past sea ice extent, which are often poorly captured in climate models (Dowsett et al., 2013). By integrating geological and glaciological evidence with climate model outputs, model skill can be evaluated for climate states that are outside the range of modern observational conditions where models are developed and calibrated to succeed (e.g., Zhu et al., 2019). Overconfidence in model parameter calibrations can be reduced (e.g., Lohmann et al., 2022), leading to better constraints on feedback, tipping points, and the dynamics of large-scale circulation systems (e.g., Armstrong McKay et al., 2022; Wunderling et al., 2024; Brown et al., 2020; Cooper et al., 2024). The inclusion of paleoclimate benchmarks helps evaluate the stability and reliability of climate model components under diverse forcing regimes, ultimately informing future scenario projections.

To achieve these aims, this opportunity incorporates a broad suite of CMIP7 Assessment Fast Track experiments. Starting from the PMIP-sponsored CMIP7 Assessment Fast Track simulation abrupt-127k (Sime et al., 2025) that aims at process understanding within CMIP7 Assessment Fast Track (Dunne et al., 2025), with a particular focus on sea ice and climate dynamics in a warming Arctic, we include in the variable request also further CMIP7 Assessment Fast Track simulations. Our goal is that other PMIP7 simulations of climates from the distant past may also be compared with historical and future period simulations. This is particularly relevant for PMIP4 simulations, Last Interglacial lig127k and mid-Holocene simulation midHolocene that will feature again in PMIP7, and PlioMIP3 Pliocene simulation LP (Haywood et al., 2024). Yet, these are just examples. The variable request includes diagnostics of the atmosphere, ocean, land surface, and sea ice to support model-data comparisons and analysis of key climate processes across timescales (Table 4). Sea ice variables, many aligned with SIMIP (Notz et al., 2016), are included to examine cryosphere evolution under past, present and future warm climates. In addition, specific variables support paleoclimate-focused research areas such as stable water isotopes, data assimilation, and coupled carbon-cycle feedback. These outputs will enable the evaluation of model performance across a wide range of climatic states, improve understanding of processes driving climate variability and change, and enhance integration between diverse research communities within CMIP7 that focus on past or future climates.

Polar amplification – enhanced warming at high latitudes relative to global mean temperature warming – is a robust feature of global climate change identified more than a century ago (Arrhenius, 1896). While polar amplification occurs in remote regions of the planet, it has global consequences. High-latitude climate change affects sea level rise, ocean and atmospheric circulation patterns, and the carbon cycle in addition to the local impacts on ecosystems and human systems (e.g., Constable et al., 2022). The rate of Arctic climate change has implications for economic development in the region, natural resource exploration, geopolitics, and adaptation (Nanni et al., 2024). However, the uncertainty in high-latitude climate projections continues to be greater than in other regions of the globe across CMIP generations (Holland and Bitz, 2003; Hahn et al., 2021). The Causality of Polar Amplification Opportunity seeks to assess the processes driving polar amplification and enable the determination of the contributions of high-frequency processes to the inter-model spread.

At the heart of polar amplification uncertainty is the current inability to attribute causality to the polar climate changes seen in observations and simulated by models. While polar amplification is a coupled atmosphere-sea ice-ocean process, a robust, quantitative understanding of the causal factors is imperfect (Manabe and Stouffer, 1980; Previdi et al., 2021; Taylor et al., 2022). Evidence suggests that the processes and feedback between sea ice, atmosphere, and ocean that are central to polar amplification unfold at high frequencies (e.g., from days to weeks). For instance, atmospheric rivers are responsible for up to 90 % of the poleward atmospheric moist energy transport (Newman et al., 2012), and an increase in the frequency of such events penetrating the Arctic accounts for much of the sea ice decline in the Barents-Kara Sea and central Arctic during the ice growth season (Zhang et al., 2023). Such high-frequency variability and the associated interactions with the sea ice pack, atmospheric state, and ocean are posited to be central to the causality of polar amplification (Taylor et al., 2022; Parker et al., 2022; Cardinale and Rose, 2023).

Contributions from high-frequency variability (e.g., atmospheric rivers and cyclones) to the inter-model spread in polar amplification cannot be accurately assessed with monthly mean model outputs. In the absence of sub-monthly data, the atmospheric energy transport by transient eddies can only be calculated as the residual between the total atmospheric energy transport and the transport by the mean meridional circulation (e.g., Donohoe et al., 2020). Furthermore, there is increasing recognition that the character (moist versus dry, e.g., Graversen and Burtu, 2016) and vertical structure (e.g., Cardinale and Rose, 2023) of the atmospheric energy transport matter more than the total amount. To enable studies that advance our understanding of the causal mechanisms of polar amplification, this opportunity contains the necessary model outputs from sea ice, atmosphere, and ocean components to assess the contributions of high-frequency processes to polar amplification and the inter-model spread in projections of polar climate change (Table 5). Key sea ice variables include thickness, concentration, and surface energy budget diagnostics. Atmospheric variables such as wind, temperature, specific humidity, and geopotential height at all model levels and surface pressure are needed (Cox et al., 2024). For the ocean, mixed layer depth and thermal energy transport are most useful in studying polar amplification.

Ocean extreme events are by definition uncommon and intense occurrences, often impacting marine life and coastal environments. Marine heatwaves are an example of short-term extreme oceanic events. These rare events occupy the tail of the upper ocean temperature distribution – typically defined as being in the 90th percentile of the climatology or a similar threshold – and can persist for days to months (Hobday et al., 2018; Oliver et al., 2021). They primarily affect the mixed layer which overlaps with the euphotic zone where most ocean photosynthesis occurs (Smith et al., 2024; Smale et al., 2019), and they have a large impact on reef-forming corals that provide critical ecosystem services such as coastal defense, subsistence fisheries, and nursery habitats for commercially important fish and shellfish (Gomes et al., 2024). Other ocean extremes, including anomalous subsurface oxygen or pH, salinity changes, or extreme sea level events, can occur independently or as compound events with marine heatwaves (Gruber et al., 2021; Burger et al., 2022; Ren and Rudnick, 2021; Han et al., 2022). Mesoscale eddies, fronts, or other anomalies in surface velocity or vorticity are often associated with these compound events. The Ocean Extremes Opportunity aims to investigate and address the impacts of extreme ocean conditions such as marine heatwaves, hypoxic zones, extreme salinities, sea level extremes and storm surges, ocean acidification, and compound events.

As the climate changes, extreme conditions are becoming more frequent and severe in many regions, posing significant risks to marine ecosystems, livelihoods (e.g., fisheries), and coastal communities and infrastructure (Gruber et al., 2021; Fox-Kemper et al., 2021a; van de Wal et al., 2024; Smith et al., 2025). Studying these characteristics requires high-frequency surface data, built up over the whole scenario time series to establish climatological ranges and capture events and changing likelihoods (Table 6). Coastal hazards in the form of extreme sea levels cause billions of dollars of damage globally, and are projected to increase in frequency (Fox-Kemper et al., 2021a). Extreme sea levels are caused by the complex interplay of multiple contributors, including astronomical tides, storm surges, waves, and sea level variability (Idier et al., 2019; Melet et al., 2024), which vary on sub-daily to interannual frequencies, as well as climate change trends. Storm surges and extreme waves are caused by prevailing atmospheric surface pressure and wind conditions. Their magnitude and the extent of their impact on a coastline are influenced by bathymetry, coastal morphology, tidal amplitude, and tidal cycle, and are further exacerbated by sea level rise (Bernier et al., 2024). Additionally, coincident pluvial and fluvial flooding can compound the severity of inundation and erosion, especially during extreme weather events. These factors highlight the need for comprehensive climate data, including winds, atmospheric pressure, precipitation, and variables associated with ocean circulation (currents, temperature, salinity), as well as other environmental variables, to better understand current and future coastal hazards risk. Long-term assessment of ocean extremes is important for identifying vulnerable regions, assessing infrastructure and ecosystem viability, designing coastal protection to appropriate levels, and enabling various adaptation measures to be considered and tested via sensitivity testing. In the coming decades, ocean extremes and associated floods and erosion are likely to remain a leading cause of natural disasters because of the increasing frequency and intensity of extremes combined with escalatingcoastal development associated with greater exposure (Bernier et al., 2024).

To better understand the transient nature of ocean extreme events, the Ocean Extremes Opportunity seeks to capture the statistics of these events and how they compare to observed statistics during the historical period, their changing likelihoods, and their sensitivity to anthropogenic forcing through a wide range of scenarios. Studying the characteristics of ocean extremes requires high-frequency surface data, built up over the whole scenario time series to establish climatological ranges and capture events and changing likelihoods. Furthermore, high-frequency subsurface marine heatwave events are also under study and have been found to be more frequent under climate change (Sun et al., 2023). By including surface values of temperature, salinity, velocity, sea level, pH and a minimal amount of subsurface information (200 m depth only), including also oxygen concentration, this opportunity will allow for a better understanding of the mechanisms and stratification conditions associated with these extremes. Additionally, variables needed to quantify storm surges, which exacerbate extremes, are included. These data will be useful in understanding such events, but they will also enable the study of their correlation with modes of climate variability such as ENSO, with changes in ocean currents and stratification, as well as the evaluation of potential impacts on vulnerable ecosystems and coastal regions. Surface fluxes, which are part of many other opportunities, can also be used for causal inference about specific events.

Wind waves are ocean surface gravity waves generated by the action of the wind blowing across the ocean surface over a certain distance known as fetch (Young, 1999; Holthuijsen, 2007). Understanding how wind wave climate evolves on global and regional scales is essential for predicting coastal hazards, erosion, and other wave-related impacts, as well as for supporting marine operations and climate adaptation strategies such as renewable energy activities (Casas-Prat et al., 2024). However, despite the many roles surface waves play in the coupled climate system (Cavaleri et al., 2012), studies on wind wave climate projections are often hindered by significant uncertainties (Morim et al., 2019), particularly at the extremes (Meucci et al., 2020) which are of the utmost importance for the safety of offshore and coastal activities. Currently, most global climate models participating in the CMIP effort do not include an active, two-way coupled wave component (Casas-Prat et al., 2024), and those that do often use a coarse wave component resolution to lower computational costs (e.g., Li et al., 2017). To enhance our understanding of wind wave climate and to improve future projections, high-resolution data on ocean surface wind speed and sea ice concentration are crucial to drive offline wave simulations (typically, subdaily fields on the ESM grid are necessary for this task, see Table 7). The purpose of this opportunity is to enable high-resolution and flexible modeling of ocean surface wind wave climates over an ensemble of ESM outputs, in support of global and regional risk and impact assessments, building on the internationally coordinated effort of the Coordinated Ocean Wave Climate Project (COWCliP). By offering computational efficiency and spatial detail in comparison to what is presently possible using a full-physics online wave model as an ESM component (see the next opportunity), this method provides actionable data for stakeholders involved in climate adaptation and marine planning.

The societal benefits of improved wind wave modeling extend far beyond academia. Coastal hazards driven by wind waves pose significant risks to coastal communities, economies, and ecosystems. Currently, around 15 % of the global population lives within 10 km of the coast, which equates to more than a billion people (Cosby et al., 2024). In terms of economic risk, coastal areas are home to major cities, critical infrastructure, and industries like shipping, fisheries, and tourism. Coastal cities and agglomerations have increased in number by 4.5 times since 1945 (Barragán and de Andrés, 2015) and the population living in proximity to coasts is projected to continue to increase, further intensifying the vulnerability of coastal areas to hazards such as storm surge, flooding, and erosion. As climate change intensifies, the frequency of extreme wave events is likely to rise (Meucci et al., 2020; Lobeto et al., 2021; O'Grady et al., 2021), putting more lives and assets in danger.

Furthermore, the growing demand for sustainable energy sources highlights a critical need for comprehensive assessments of wave energy, offshore wind energy, and other types of renewable energy that can be deployed on or near coastlines. These assessments depend on the potential of high-resolution coastal wind wave information, which is fundamental to evaluating the feasibility of wave and offshore wind energy as renewable resources (Kulkarni et al., 2018; Jung et al., 2024). Such data help identify optimal locations for wave and wind energy farms by analyzing long-term wave climate patterns and understanding climate variations. Large-scale atmospheric and oceanic variability, such as the ENSO and the North Atlantic Oscillation (NAO), climatic trends, and connected regional, coastal wind and wave variability at high resolution are all important for long-term planning (Tseng et al., 2024). The harsh marine environment poses strong currents, high waves, and corrosive conditions, which can impact the safety, longevity and reliability of infrastructure. Detailed risk assessments help identify potential hazards, inform mitigation strategies, and ensure compliance with international safety standards, safeguarding both workers and the environment.

For effective offline wind wave climate modeling, high-resolution temporal and spatial data are essential. Wind patterns can vary significantly over short time scales and distances, and the interaction between wind, waves, and ice requires detailed data to capture these dynamics. Similarly, sea ice concentration data is critical for understanding how ice attenuates or reflects wave energy, as well as how a changing ice cover due to warming temperatures alters wave patterns. Without high-resolution data, models may miss localized phenomena such as extreme wave events or changes in coastal wave energy distribution, which are crucial for understanding and predicting coastal erosion and other hazards. By incorporating finer-scale data into wave models, we can better understand the spatial variability of wave energy, identify vulnerable coastal zones, and assess the future risks posed by changing wave climates.

Traditionally, ocean surface wind wave climate studies, such as those highlighted in the previous opportunity (ID 24), rely on high-frequency output from ESMs for offline execution of a comprehensive suite of statistically or dynamically downscaled wave climate ensembles for past and future conditions (e.g., Hemer et al., 2013; Morim et al., 2019; Meucci et al., 2020, 2024; Casas-Prat et al., 2024). However, numerous studies have demonstrated that ocean waves play a critical role in regulating Earth's climate system by influencing the exchange of energy, momentum, and mass between the ocean and atmosphere (e.g., Babanin, 2006; Belcher et al., 2012; Cavaleri et al., 2012; Qiao et al., 2013, 2016; Li et al., 2016, 2019; Li and Fox-Kemper, 2017; Fox-Kemper et al., 2021b). Some of these feedback mechanisms have been recognized by meteorological institutions, where two-way wave-atmosphere interactions are now incorporated into weather prediction models (e.g., Janssen, 1991; ECMWF, 2024). Ongoing research seeks to clarify to what extent ocean surface waves influence the climate on longer timescales. The goal of this opportunity is to support research on the influence of ocean surface waves on the climate system, including their feedbacks within ESMs and their long-term role as climate drivers. Unlike ID 24, this is not part of an established coordinated modeling effort, but rather aims to advance understanding and foster new developments in surface waves that are coupled online as a component within ESMs.

Breaking waves, for instance, contribute to the generation of sea spray, which has significant implications for cloud formation (Veron, 2015; DeMott et al., 2016; Brumer et al., 2017; Deike et al., 2022). Additionally, wave-driven processes modulate air-sea gas exchange, particularly affecting CO₂ uptake (Deike and Melville, 2018; Woolf et al., 2019). Ocean waves play a significant role not only in the modulation of the atmospheric boundary layer, but also influence ocean mixed layer depths through both breaking and non-breaking motions that induce turbulence via Langmuir instabilities or other mechanisms (Qiao et al., 2013, 2016; Li et al., 2016, 2019; Li and Fox-Kemper, 2017). However, fully-coupled interactions between waves, mean flow, and turbulence still have many unknowns (Kantha and Clayson, 2004; Suzuki et al., 2016; Wu et al., 2019; Fox-Kemper et al., 2021b).

In the polar regions, ocean waves play a crucial role in shaping sea ice dynamics, particularly in the marginal ice zone (e.g., Roach et al., 2018). As waves propagate through this region, they can fracture the ice, breaking it into smaller floes. This process enhances both lateral melting (melting from the edges of ice floes) and basal melting (melting from beneath the ice) by increasing the ice's exposure to warmer ocean waters (Alberello et al., 2022). Additionally, waves alter air-sea-ice flux exchanges, influencing how heat, moisture, and momentum are transferred between the ocean and the atmosphere (Bennetts et al., 2024). Furthermore, the reduction of sea ice due to storm-driven wave activity can have far-reaching consequences (Kohout et al., 2014; Blanchard-Wrigglesworth et al., 2024). When sea ice is absent or significantly weakened, ice shelves become more exposed to powerful ocean swells, leading to ice shelf disintegration and an acceleration of the loss of polar ice masses (Massom et al., 2018).

Given the increasing recognition of ocean waves as key components of the Earth system, several modeling centers have started incorporating wave coupling into ESMs and Global Climate Models (GCMs; Qiao et al., 2013, 2016; Li et al., 2016, 2017; Reichl and Li, 2019; Bao et al., 2020; Danabasoglu et al., 2020; Brus et al., 2021). This coupling should allow for a more accurate representation of air-sea interactions, improving the simulation of weather and climate phenomena (Fox-Kemper et al., 2019, 2021b). By integrating wave processes into climate models, researchers aim to enhance predictions of extreme events and long-term climate variability as well as to better understand the roles of waves in the climate system and the feedback processes they entail.

High temporal and spatial resolution outputs of wave parameters from coupled climate models, including significant wave height, mean and peak wave periods, and mean wave direction, are particularly valuable for comparing wave climate simulations performed with the traditional stand-alone spectral wave models, which rely solely on atmospheric forcing (Table 8). By analyzing wave fields from coupled climate models alongside conventional wave climate model outputs, we can start assessing the added value of wave coupling in representing long-term wave climate variability and extreme wave events. In addition, this opportunity includes some key ocean and sea ice variables so that the impacts of parameterizations and wave effects mentioned above can be assessed. Since surface waves affect multiple aspects of the climate system, this opportunity has the potential to improve the results and findings from other opportunities (e.g., ID 13, ID 24, ID 49, and ID 73). This opportunity will enable more comprehensive evaluations of wave climate projections, leading to an improved understanding of wave-driven feedback and better-informed applications in climate science, coastal hazard assessment, and marine operations.

The CMIP Rapid Evaluation Framework (Hoffman et al., 2025) was created to evaluate and benchmark the newly available CMIP7 Assessment Fast Track simulations as soon as they are uploaded to the Earth System Grid Federation (ESGF), providing metrics and diagnostics that are available through different open-source evaluation and benchmarking tools. This opportunity contains the set of variables (Table 9) that are needed for the planned diagnostics and metrics for the REF (CMIP Model Benchmarking Task Team, 2024). The selected metrics and diagnostics for the REF to be available for all CMIP7 Assessment Fast Track experiments were intentionally chosen for very basic evaluations and are not expected to require highly specific variables. The exact selection of variables was also made consistent with the model evaluation diagnostics in Chap. 3 of the latest IPCC report (Eyring et al., 2021). Due to the fixed timeline for the CMIP7 Assessment Fast Track simulations, there is only a short period for the technical implementation of the REF, and therefore the available metrics and diagnostics in this first version of the REF will be limited to a temporal resolution of monthly mean data and about five metrics/diagnostics per realm. Implementation will be based on a selection made by the community. The realms were chosen specifically to be consistent with the realms used for the data request. Find more information about the REF Opportunity in Hoffman et al. (2025).

The prioritization process, both in terms of selecting, combining, and sometimes rejecting opportunities, and in terms of prioritizing among variables and variable groups, is inevitably imperfect. While the team met frequently, discussed the selected opportunities in detail, and then reviewed the variables collectively and individually for errors or oversights, there are far too many variables involved to keep a clear perspective on all that is required and most urgent. Building this team from a group of scientists, few of whom had previously worked together, required effort, as did familiarizing oneself with the procedures, protocols, and software selected by the IPO and CMIP Data Request Task Team for the development of the CMIP7 data request. Added complexity and challenges stemmed from the need to coordinate with other author teams in the data request development process, and doing so strictly remotely. The data request surely would have benefited from a slower, longer process, where opportunities to meet in person, such as at large conferences, could have been taken. On the other hand, the design of the CMIP7 data request had to be implemented within the tight schedule of the overall CMIP7 Assessment Fast Track and IPCC cycle.

While consistency with the CMIP6 data request is insufficient to ensure a successful CMIP7 outcome, it is at least reassuring that the groundwork laid out in CMIP6 was maintained, particularly in the two “Drivers and Impacts” opportunities (ID 47 & 73). Due to specifics of requests from opportunity teams, some variables, for example intpn2, were only requested in a regional context (30 to 90° S); modeling centers may provide the global version of these variables instead or in addition to the regional data. Many of the other opportunities were designed by reference to specific single-model studies that were successful, and thus the multi-model intercomparisons enabled by this data request have a stable foundation.

Despite the large number of variables, increases in output frequency over CMIP6, and the inevitably vast quantity of data that this request will trigger to be captured, there are still identifiable gaps in the resulting data request that will make certain processes remain in the shadows, at least in a multi-model sense. Sub-daily data remains extremely rare, but in many of the processes highlighted in these opportunities (e.g., those involving surface waves or processes that depend on particular phases of the diurnal cycle), that frequency is requested for study. While emphasis on extreme events is mostly on ones that persist over a matter of days, such as heat waves, their peak intensity is of even shorter duration and potentially more impactful. Many of these high-frequency variables are only requested for ocean surface or near-surface levels, which reduces their data volume substantially. One issue that was not resolved in this data request is how to describe the ocean surface as a coordinate designation – a depth of 0 m is not wholly accurate, as many modeling centers provide the uppermost gridcell average, which is not centered on 0 m. However, no consensus on an improved designation was found in time for this data request. Ocean tides are rarely included in ESMs, but in some prototype simulations they have shown significant climate impacts that persist in averages (Arbic, 2022); in most of the world, a semi-diurnal tide dominates, and thus, sub-daily output is needed for tidal coastal dynamics (e.g., Ahmed et al., 2025).

Although this data request also applies to the models in the HighResMIP simulations, and appreciating that some CMIP models will be fairly high-resolution for all applications, reproduction of real-world spatial detail and heterogeneity remains a challenge for the representation and study of many Earth system processes that observations reveal to be important (Hewitt et al., 2020, 2022). While parameterizations can be made to capture some aspects of these unresolved or poorly-resolved phenomena in the Earth system, a one-size-fits-all data request, such as that produced for CMIP7, cannot meaningfully capture the variety of parameterization mechanisms or their diagnostics so as to compare them across models. This means that it remains painful to carry out careful multi-model parameterization and high-resolution intercomparison studies, although with diligence and focus they still occur (e.g., Chassignet et al., 2020; Uchida et al., 2022; Li et al., 2019).

In a world of increasing automation and artificial intelligence, the process of constructing these data requests was still frustratingly manual and required human intelligence. One simpler idea would be just to request the most popular variables from the CMIP6 request. While download statistics about previous generations of CMIP models were collected by the Earth System Grid Federation (which maintains and curates the repository), key flaws in the collected statistics limit their utility. Not every download was used; some downloads – especially of large-output variables – were downloaded much less often than they were used (i.e., they were shared among colleagues at modeling centers after being downloaded only once), and many routine analyses (e.g., ocean heat content using multiple equations of state, McDougall et al., 2021) ended up requiring the download of much more data than necessary to obtain key data subsets. So, while our team did consult these download statistics in prioritizing and confirming key variables, CMIP7 Assessment Fast Track missed a chance to profoundly improve the data request, analysis tools, and data access, while simultaneously reducing data storage requirements. The goal is to make the most of CMIP data worldwide, but that is a lofty aspiration. With added foresight, the data request process could have been mostly automated, and then our team could have focused more on facilitating new scientific insights instead of onerous cross-checking and debugging.

Another aspect of automation and artificial intelligence is the growing use of emulators and machine learning parameterizations. The data requested here are not intentionally suited to such purposes, except in that they are the variables for processes with impacts and effects that are important. Their importance suggests that they are likely to be included in machine learning approaches (e.g., ID 22 & 80 of the Impacts and Adaptation data request; Ruane et al., 2025). However, in fields outside of climate science, many recent successes in machine learning have resulted from carefully curated benchmark training datasets (e.g., Wu et al., 2018; Hu et al., 2020). Although there are limits to how much a benchmark-trained system applies to the real world (Raji et al., 2021), at present, the bigger issue for climate science is the lack of benchmark climate datasets designed specifically for machine learning applications. A future potential application of machine learning would be to provide diagnostic emulators that can infer many variables from a few saved variables, thereby greatly reducing the data storage and data request complexity in favor of a more complex emulation (with the accompanying issues in making these emulators trustworthy).

One challenge that confronts teams such as ours trying to optimize the data storage effort is that some variables can of course only be calculated online during simulation, while others can be calculated offline when given the correct input data. Indeed, standard analyses that may be centrally prepared, as considered for example in the CMIP7 Rapid Evaluation Framework (Dingley et al., 2025), are ideal to limit data storage load and to speed up exploration by avoiding preparing the same analyses over and over again. Two oceanic examples that our team considered were mixed layer depth, which can only be calculated at full accuracy online (see Treguier et al., 2023, for a quantification of the errors when calculating after the fact), and ocean heat content (phcint), which can be calculated after the fact from 3D temperature, salinity, and grid specifications (areacello, volcello). Treguier et al. (2023) find that mixed layer depth errors in calculating after the fact were somewhat smaller than the errors in CMIP6 due to discrepancies in the definition of mixed layer depth used by multiple modeling centers. In this data request, we seek to eliminate these inter-model definitional discrepancies, which then leaves after-the-fact calculation as a leading error, which justifies our request for the mixed layer depth as calculated online. The ocean heat content (and salt content) can be calculated accurately from 3D fields, but this calculation is done so often and is so expensive in data recall that we ask modeling centers to do it in advance, especially as this avoids users needing to download 3D temperature data in order to find a 2D field. Furthermore, the use of new vertical coordinates makes these calculations complex and sensitive to how the grid specifications are used (including missing ingredients needed to account for temporal rectification effects). Providing the calculated ocean heat content necessarily increases the number of variables requested from modeling centers, with some redundancy in many cases, but as these variables are used widely, require downloading large data volumes, and are prone to error in calculation we deem this addition to be justified. Another example is the request for modeling centers to provide hemispherically-integrated sea ice variables (such as sea-ice area, extent, volume, etc.), as we know from previous CMIP phases that these are the most downloaded sea ice variables, and calculating them offline after regridding can lead to errors compared to doing it online. In general, there tends to be a tradeoff between the number of variables requested (which can actually reduce the data volume needed to be downloaded and stored when they are, e.g., integrated, as ocean heat content is, or 2D, as mixed layer depth is) and the volume of data requested (e.g., higher frequency 3D fields reduce the rectification errors in after-the-fact calculations in variable vertical grid models). We only anticipate these issues will become more complex as model configurations diversify in vertical grid specifications and unstructured horizontal grids. The OMIP protocol and diagnostics are under discussion (Baylor Fox-Kemper, personal communication, 2025), and this forthcoming paper will contain further details and discussion.

There are a few potential issues with baseline and depth-related ocean variables that require further specification. Both conservative temperature and potential temperature are available as variables, and modeling centers can choose to upload only the variable corresponding to the equation of state used by their model, or to upload both temperature variables after performing a conversion. It is important for the metadata for each model to note which equation of state is used and which uploaded variable is the native state variable and the derived state variable. Similarly, practical salinity or absolute salinity can be uploaded or both (after a similar conversion). Unlike in past data requests, there is expected to be diversity of equations of state used in this CMIP round, and these variables differ substantially enough that model-observation comparisons require apples-to-apples comparisons to avoid the spurious appearance of model bias. For the depth-integrated variables (those that begin with “integral_wrt_depth”) that are part of this data request, it is intended that these variables be relevant for ocean energy and salt content calculations and for comparisons to hydrographic observations by CTD sensors (which detect depth by a pressure sensor). So, in the case of models making the Boussinesq approximation, depth from the free surface should set the deeper integration bound (thereby making heat content change less dependent on the choice of reference density) while in non-Boussinesq models pressure from the free surface is the relevant integration bound. Where the ocean is ice covered, the depth/hydrostatic pressure of the liquid ocean should be used (this convention can be accounted for when comparing to, e.g., ice-tethered profilers). Finally, we request that the ocean mixed layer depth be calculated as suggested by Treguier et al. (2023) as to be directly comparable to the SEANOE observational product: as the depth at which (Conservative or potential) density differs by 0.03 kg m⁻³ from the (Conservative or potential) density found at the 10 m depth level. The choice of density should correspond to the equation of state used by the model.

Inevitably, future model generations will involve more data, more complex data structures (e.g., unstructured grid meshes, parameters describing hybrid ML-dynamical models), and more and new questions about how the Earth works and what we can do to conserve and protect its bounty. The length of assessment reports, the number of papers cited in the assessments, the size of data repositories, and the number of variables requested have all been monotonically increasing. The CMIP enterprise has become central to the study of the Earth system, but if care is not taken to seek out unifying ideas, streamlined processes, and opportunities to take advantage of automation, it may crumble under its ponderous weight (Stevens, 2024). Sharing our climate science challenges through collaborations with data scientists and computer scientists is one pathway to lightening the burden, but this data request did not fully accomplish a handshake with those communities through benchmark data.

Finally, the essence of science is prediction of what may occur. Scrutinizing such predictions in the light of observations and experiments is how they are evaluated and improved. The continual race to higher spatial and temporal resolution in climate model output is paired with observations that are increasingly high-resolution and sophisticated. This data request design process involved mostly scientists who have experience in evaluating models versus observations rather than in collecting observations. We foresee that there is a potential to evaluate models in a much more intricate manner than just decimating and interpolating observations collected by other scientists onto our model grid and then calculating simple biases and errors, as it has often been done in the past. To this end, we suggest that future data request designs could benefit from increased involvement of observationalists. It is a life's work to become an expert in the limits and advantages of particular observations. Their suggestions during the data request design would have offered new insights.