The Arctic is a region of increasing strategic and economic importance. Its influence spans a diverse array of stakeholders across international boundaries, including local populations (e.g., indigenous populations), natural resource industries, fishing communities, commercial shippers, marine tourism operators, national security organizations, regulatory agencies, and the scientific research community (e.g., Arctic Council, 2009). The Arctic also plays a number of roles in moderating global climate by influencing the planetary heat budget and interacting with the oceanic and atmospheric circulation systems as well as the terrestrial environment (Guemas and Salas-Melia, 2008; Lawrence et al., 2008; Deser et al., 2010; AMAP, 2011; Francis and Vavrus, 2012; Jakobsson et al., 2012; Koenigk et al., 2012; Maslowski et al., 2012; Nghiem et al., 2012).
The extent and thickness of Arctic sea ice has recently undergone extraordinary decline (Figure 1.1) that can be linked to climate changes (e.g., IPCC, 2007; Min et al., 2008; Allison et al., 2009; NRC, 2010; Kay et al., 2011; Notz and Marotzke, 2012; ). The last six summers (2007-2012) have experienced the six lowest sea ice extent minima over three decades of satellite record, and the past decade (2003-2012) has exhibited 9 of the 10 lowest minima (updated from Perovich et al., 2011). The reduction of summer sea ice extent has been greatest in the Beaufort and Chukchi seas offshore of Alaska and in the Kara and Laptev Seas north of Russia. These are regions of particular interest to stakeholders concerned with marine access to subsistence activities, shore infrastructure, marine transportation, and natural resource developments. The winter sea ice extent has also shown a downward, though less striking, trend. More notable than winter sea ice extent is the change in composition of the winter sea ice cover, associated with the reduced summer sea ice. The winter sea ice cover now includes a significantly higher percentage of thin, seasonal (i.e., first-year) ice. The fraction of the first-year sea ice in the Arctic Ocean in March increased from 38 percent in the early 1980s to 64 percent in 2010 (Stroeve et al., 2011).
Closely associated with the persistent changes in the characteristics of the ice cover are other observed changes throughout the Arctic system. For instance, in regions of sea ice loss the upper ocean is warmer and fresher (e.g., Jackson et al., 2010; Steele et al., 2011). As a result of increased open water area, biological productivity at the base of
FIGURE 1.1 Arctic sea ice extent has recently undergone extraordinary decline: March 2012 (left) and September 2012 (right). The two periods that define the annual sea ice extent cycle are March, at the end of winter, when the ice is at its maximum extent, and September, at the end of summer, when the ice reaches its annual minimum extent. The purple line indicates the median maximum and minimum ice extents in the given month for the period 1979-2000. Compared with the 1979-2000 average, the September 2012 minimum was 49 percent smaller. SOURCE: Updated from Fetterer et al. (2002), Sea Ice Index, National Snow and Ice Data Center.
the marine food chain has increased (e.g., Arrigo and van Dijken, 2011) and sea ice-dependent marine mammals continue to lose habitat (e.g., Thomas and Laidre, 2011). Increases in the greenness of tundra vegetation and permafrost temperatures are linked to warmer land temperatures in coastal regions, often adjacent to the areas of greatest sea ice retreat (e.g., Bhatt et al., 2010).
Given the expectation for a continued increase in the global temperature throughout the 21st century (e.g., Meehl et al., 2007), the declining trends in Arctic sea ice extent and multiyear ice composition are expected to continue. Associated changes will likely result in greater marine access to the Arctic (e.g., for commercial shipping and offshore natural resource development) and increased coastal erosion, as well as a range of local, regional, and hemispheric changes in the climate and ecological systems.
The current and projected conditions and activity levels in the Arctic call for an
improvement in our ability to predict characteristics of the sea ice cover. At the foundation of this challenge lies a set of specific questions related to sea ice prediction common among the broad stakeholder community: (1) Where is the ice at any given time (extent and concentration)? (2) What is it like (thickness distribution, multiyear ice versus first-year ice, packed or loose)? (3) What is its movement? (4) How and why has it changed? A major complicating factor in devising a strategy to more effectively address these questions is associated with the detailed and differing needs of the stakeholders, which shapes temporal and spatial requirements. For instance, while decadal projections of pan-Arctic ice extent and composition may serve the needs of decision makers who determine classification of endangered species or of planners who decide whether to build an ice-worthy ship, the long-range predictions will be less useful to marine operators whose main concern is the position of the ice edge. Springtime whaling and walrus hunting by coastal communities is dependent on over-ice transportation, requiring seasonal projections for the stability of the ice cover, its thickness and roughness, and the presence of open water close to shore. Detailed local information of this kind is not presently captured by monthly-to-seasonal ice forecasts and is only beginning to be targeted by experimental sea ice forecasts (e.g., Sea Ice Walrus Outlook1
Many of the needs and challenges associated with sea ice prediction depend on the timescales of interest. At shorter timescales (seasonal to interannual), internal variability will severely limit sea ice predictability. For example, summertime wind patterns are known to impact summer and autumn ice extent, yet weather patterns are essentially unpredictable beyond about 2 weeks. Additional atmospheric observations are not going to change this fundamental prediction barrier. Rather, predictive capability for sea ice on these timescales is thought to reside primarily in an adequate knowledge of the initial ice-ocean state, although admittedly little information exists on what constitutes an “adequate knowledge.” At longer (interannual to decadal and beyond) timescales, the role of trends in forcing (e.g., rising concentrations of greenhouse gases, changes in ocean mixing, increases in river discharge) is likely to provide some predictive potential, as it accounts for increasingly larger fractions of the change from present sea ice conditions.
A critical gap of uncertainty remains regarding the timescale between these two regimes, in which the potential predictability is low. Blanchard-Wrigglesworth et al. (2011) addressed these various prediction timescales in a numerical modeling study and suggested that a time frame of minimal predictability (for total Arctic Basin ice area) occurs from about 2 to 4r years. Although this study represents a state-of-the-art assessment of sea ice predictability, the results are based on one model (Community Climate System Model [CCSM4]) and are contingent on the ability of that model to capture the spectrum of oceanic, atmospheric, and terrestrial variability that affects sea ice. Hence the robustness of this
result across climate models is uncertain, and the timescales are likely to vary based on the sea ice property, simulated atmospheric and oceanic forcings, and region of interest. Nevertheless, the existence of a minimum in predictability at interannual timescales of several years is plausible and likely. For this reason, the report focuses on seasonal and decadal predictability, with the understanding that improvements in predictions over these two timescales may eventually extend into the interannual timescale of relatively low predictability.
The particular challenges confronting the prediction of the character and behavior of sea ice in the seasonal time frame are compounded by the increasingly urgent need for this information by a variety of stakeholders. As noted above, coastal regions and the marginal ice zone are an important focus of these needs because sea ice affects access to shore infrastructure, marine transportation, resource extraction, and fishing activities. At these timescales stakeholders require rapid access to the information, and errors in that information tend to have immediate and often serious consequences. Furthermore, the regions and sea ice properties of interest will likely continue to experience fundamental change accompanying long-term variations in climate.
On decadal timescales, recent work has highlighted the considerable variability of Arctic sea ice. Kay et al. (2011), for example, found that different ensemble simulations from a single climate model subjected to the same changing external forcing (e.g., greenhouse gases and solar variability) exhibited a considerable spread in the simulated late 20th century Arctic ice loss. Moreover, some simulations for the 21st century revealed decadal-scale periods of gains in ice extent when internal variability counteracted greenhouse gas-forced trends. However, the chaotic internal variability can also reinforce the trend in ice loss, leading to instances of very rapid sea ice retreat (e.g., Holland et al., 2006). This large internal variability provides an inherent limit to predictability at these timescales, and as such, any decadal-scale predictions are necessarily probabilistic in nature. To date, however, little work has been done to assess the inherent limits on decadal predictability for different sea ice properties, at different times of year, and in different regions.
Some global coupled climate models are able to realistically simulate the past behavior of Arctic sea ice (e.g., Jahn et al., 2012). Figure 1.2 compares a single realization by the Community Climate System Model Version 3 (CCSM3) with observations of the actual ice cover, demonstrating the model’s success in capturing not only the decadal-scale pace of ice loss, but also realistic interannual variability. Of particular note is the ability of the modeled ice extent to undergo a decade of recovery within the inexorable downward trend. Climate models that were discussed in the 4th Intergovernmental Panel on Climate Change (IPCC) assessment were compiled under the Coupled Model Intercomparison Project, Phase 3 (circa 2006). Simulations with newer models (circa 2011) that are informing the 5th IPCC assessment have just recently become available in CMIP5. As the models have progressed over this time,
FIGURE 1.2 The 20th to 21st century September ice area in the Northern Hemisphere from a single CCSM3 ensemble member (black line) compared with the observed time series of September ice area from Fetterer et al. (2009, red line). It demonstrates the model’s success in capturing the decadal-scale pace of ice loss and realistic interannual variability SOURCE: Adapted from Holland et al. (2011).
the realism of simulated sea ice area in response to anthropogenic and natural forcing has improved in comparisons with satellite observations available since 1979 (Massonnet et al., 2012; Stroeve et al., 2012).
Many climate models still simulate an Arctic ice pack at odds with observations. Figure 1.3 displays dozens of model runs from CMIP5 under two different future forcing scenarios termed “representative concentration pathways” (RCPs; Moss et al., 2010) including RCP4.5 and RCP8.5, which reach 4.5 W.m-2 (watts/square meter) and 8.5 W.m-2 radiative forcing by 2100, respectively. Most simulations capture a long-term reduction in ice extent that is driven fundamentally by external forcing— primarily increasing concentrations of greenhouse gases. The spread among the runs, however, has not decreased appreciably since the IPCC-4 generation, suggesting that basic challenges remain to be overcome. Because so many variables affect the ice evolution, identification of the causes
An ad hoc committee will plan and conduct a public workshop that outlines the current state of Arctic sea ice research, discusses knowledge gaps, and identifies emerging or important new science questions for the coming decade. Through invited presentations and discussion, participants will examine current observations and modeling efforts of sea ice, and identify (but not prioritize) areas of research and technology advances needed to better understand current and future changes. The committee will examine Arctic sea ice prediction, with a particular emphasis on seasonal to decadal timescales. The workshop will be designed to bring together polar scientists and agency representatives to explore whether there are new capabilities and infrastructure available to study sea ice in different ways that might shed new light on emerging research questions. This information may provide the context for future planning and policy development for sea ice research activities. The outcome of this activity will be a consensus report of the committee that builds on workshop presentations and discussions to provide conclusions on the following topics:
• What key scientific questions remain and how can we improve our understanding of the coupling between oceans, atmosphere, and sea ice (e.g., on what processes should observations be focused)?
• What systems of monitoring and observations are needed to better understand and predict the connection between changes in Arctic sea ice and its impacts on climate?
• What aspects of coupled sea ice models do we understand the best and in what ways can models better utilize existing observations and monitoring of sea ice to enhance our understanding of processes and future changes, and improve sea ice prediction?
a This report is sponsored by NASA, ONR, and the intelligence community.
of discrepancies in a particular model is often elusive, and sometimes improvements to one process reveal problems in other processes that were originally masked because of compensating errors.
As discussed previously, Arctic sea ice prediction has inherent limitations due to the chaotic nature of the climate system that may severely limit the possible predictive power. However, these limitations are poorly understood, especially across the full range of timescales and variables of interest to stakeholders. Our ability to realize the inherent predictability that does exist is further hindered by a limited understanding of the coupled and complex interactions between Arctic sea ice, oceans, and the atmosphere. Advances in understanding and seasonal to decadal predictive capabilities require enhancements of our theoretical,
FIGURE 1.3 Many climate models still simulate an Arctic ice pack that differs from observations. This figure shows simulations of September sea ice extent from 1900 to 2100 by 23 global climate models participating in the 5th IPCC assessment. Each thin colored line represents one ensemble member. The thicker solid red line depicts observed ice extent, based on the Hadley estimate; except for September 2012 which is from NSIDC. The wide spread among the runs has not decreased appreciably since the 4th IPCC assessment, and because of the complexities associated with sea ice evolution, there are challenges that remain to be overcome. SOURCE: Wang and Overland, 2012. Reproduced/modified by permission of American Geophysical Union.
observing, and modeling capabilities. Evidence of the high level of concern about these limitations and the challenges involved in addressing them is demonstrated by the many recent studies that have focused on identifying key questions and recommendations related to Arctic sea ice predictability (see Appendix A for a summary of recent efforts). This report seeks to build on these efforts, with a specific emphasis on improved integration between the diverse community of stakeholders with a keen interest in and significant requirement for improved sea ice
Couplings—Two-way interaction between different subsystems (e.g., atmosphere, cryosphere, hydrosphere, etc.).
Decadal scale (4 to 30 years)—The term “projection” is commonly used when referring to this timescale.
Feedbacks—A sequence of interactions that determines the response of a system to an initial disturbance.
First-year ice—Floating ice of no more than 1 year's growth developing from young ice; thickness from 0.3 to 2 m (1 to 6.6 ft) where level; ridges of much thicker ice, to 30 m (98 ft), can form where floes are fractured by pressure, and these are rough and sharply angular.
Forcings—External data input into models that drive variability and change (e.g., solar radiation, greenhouse gas concentrations, volcanic aerosols).
Interannual scale (1-4 years) - The term "prediction" is generally used, although predictions for this timescale presently show little skill relative to climatology or persistence of trend.
Internal variability—Interactions within the climate system, as opposed to those forced externally (e.g., by changing greenhouse gas concentrations, solar variability, volcanic aerosols). Examples of internal variability include El Niño, the Arctic Oscillation, and the enhancing or dampening effects of feedbacks.
predictions on the seasonal to decadal timescale.
This report was developed from insights and information gained during a workshop. The goal of the workshop was to foster a dialogue among stakeholders (e.g., Arctic indigenous residents, polar scientists, agency representatives, and commercial interests) to explore current major challenges in sea ice prediction, and to identify new methods, observations, and technologies that might advance seasonal to decadal sea ice predictive capabilities through improved understanding of the Arctic system. The most prominent theme to emerge from the workshop was the idea of a committed and deliberately integrative approach to Arctic sea ice prediction that would require a
Numerical models—Numerical models solve systems of equations describing the fundamental physics, fluid motion, and thermodynamics of an Earth system component. These models can include single Earth system components (i.e., sea ice, ocean, land, or atmosphere) or can include multiple components that are coupled through the exchange of heat, water, and momentum (i.e., ice-ocean models, global climate models). Biogeochemistry, chemistry, and other aspects can also be incorporated through the inclusion of additional coupled equations or parameterizations.
Marginal ice zone—A band of pack ice 100 to 200 km (62 to 124 mi.) wide that forms a buffer between open seas and dense interior pack ice; here, waves, swells, and eddies have strong impacts that affect the ice, creating highly variable ice conditions.
Multiyear Ice—Ice that has survived at least one melt season; the thickness of multiyear ice floes can range from 2 to 20 m (6.6 to 66 ft) thick.
Predictability—The extent to which future states of a system may be predicted based on knowledge of current and past states of the system. Predictability is inherently limited because knowledge of the system’s past and current states is imperfect and future variations of the external forcings are not exactly known.
Seasonal scale (21 days to 1 year)—The terms “prediction” and “outlook” are commonly used when referring to this timescale.
Statistical models—A model based on statistical relationships between different variables in past behavior of the system to be modeled.
Weather scale (1 hour to 10 days)—The term “forecasting” is commonly used when referring to this timescale.
sustained and coordinated conversation among the user, modeling, and observation communities. It was noted that this approach needs to go beyond ad hoc workshops and demands long-term, continuous, two-way interaction. This theme, which is discussed in more detail in Chapter 3, drove many of the key challenges and strategies laid out in the report.
The report addresses Arctic sea ice prediction over the seasonal to decadal timescales as a driver of the need for improved understanding of sea ice variability (Box 1.1). The committee's focus was on ice conditions during all seasons within the whole Arctic marine environment (i.e., Arctic Ocean and the subpolar seas, including the seasonal sea ice zone).
Although the Statement of Task does not explicitly mention stakeholders, it was the committee’s view that a report on needs in sea ice prediction would be seriously deficient if stakeholders were not a prominent part of the underlying discussion. A similar sentiment was also raised in a recent NRC report: “IPY-related predictive modeling will continue to play a crucial role in helping commercial enterprises, individuals, and governments assess the regional and global risks associated with ongoing melting ice, sea level rise, permafrost degradation, and other effects of rising polar temperatures in a warming world” (NRC, 2012a).
Further, the committee and workshop participants observed that the motivational questions posed in the Statement of Task were not unique to this activity. Rather, they are questions that are often asked of researchers involved in observing and modeling the Arctic sea ice cover. This realization led the committee to consider additional, more overarching questions in the preparation of this report: (1) Given the significant investments and the progress that has been made in observing and modeling the Arctic sea ice cover, why are we not further advanced in the ability to predict its condition on seasonal to decadal timescales? (2) How can we apply the tools and insights we have developed in a systematic way to more effectively address the questions posed in the Statement of Task?
After presenting a series of key science questions, Chapter 2 identifies gaps and challenges related to understanding and predicting Arctic sea ice evolution. It begins with a set of overarching challenges that are foundational, including issues related to the Arctic environment and its stakeholders.
These overarching challenges are followed by challenges and gaps that are more specific to sea ice predictions, laid out as a function of timescale from seasonal to decadal. Chapter 3 presents possible strategies to significantly advance our understanding and predictions of Arctic sea ice over seasonal to decadal timescales. The organization of Chapter 3 is designed to generally follow the order of key challenges, though there is not a direct correspondence between the highlighted points made in Chapter 2. Examples of recent and ongoing activities are provided throughout Chapter 3 to demonstrate successful approaches that have been designed and implemented to address related issues. Key challenges and strategies are denoted in gray boxes throughout Chapters 2 and 3. Chapter 4 concludes with summary comments. Definitions of terms used throughout the report are provided in Box 1.2. This report does not make specific recommendations because of the reliance on the workshop in developing the ideas put forward in this report and the relatively short tenure for deliberations and analysis. The report does not include extensive background information. The interested reader is encouraged to utilize the numerous references and website links provided throughout the text.