A deliberately integrative approach is needed to facilitate coordinated and sustained discussions and collaborations among the user, modeling, and observation communities to inform effective research activities and to set realistic expectations for predictions. Such an approach would need to take full advantage of existing infrastructure and draw from comparable efforts in other fields.
The need for sustained and facilitated communications among the many stakeholder communities is both a challenge and a significant opportunity. End-user needs are important drivers for science and research efforts and for more accurate, timely and useful sea ice forecasts. For example, regularly scheduled, iterative, and sustained discussions among end users, modelers, field scientists, and the remote sensing community could help the science communities understand user tolerances for the accuracy of particular ice characteristics. They could also help to determine the most practical approaches to obtaining the highest priority products.
Some user requirements may be beyond existing capabilities because of inherent limitations in predictability or owing to technical or practical limitations. Thus, it is incumbent upon the research element of the stakeholder community, including modelers and observers, to understand the basis of these requests, to explore alternatives, and to properly manage expectations. Increased collaboration can help overcome the challenge of communicating to the public about what can reasonably be expected from seasonal to decadal sea ice prediction.
Within the framework of organized stakeholder communication, true collaborations between the modeling and observation communities are essential. Given the reality of limited resources, these collaborative efforts are more likely to identify resource needs and shared resources. The need for integrative, sustained conversations calls for routine and frequent engagement. Although workshops and topical meetings have their place within the process, this strategy calls for a longer-term, continuous interaction. The
Examples of Sustained Conversations: ArcticNet,
Alaska Ocean Observing System (AOOS), and
Alaska Center for Climate Assessment and Policy (ACCAP)
An example of a large-scale program that facilitates sustained communication across the research and stakeholder communities is ArcticNet,a a network of Canadian institutions that aims to bring together scientists and representatives from academia, government, industry, international agencies, and northern communities. The main objective of ArcticNet is to utilize information from these sources to help Canada prepare for and adapt to climate change and increased activity in the Arctic. Through research collaborations and partnerships with corporations operating in the Arctic, reliable scientific data can be made available to a wide variety of stakeholders. Research results are shared with the community in the form of scientific publications and a data catalogue and through various media outlets. This integrated and cross-disciplinary approach to Arctic research provides unique opportunities for education, collaboration, and entrainment of the next generation of polar researchers.
The Alaska Ocean Observing System (AOOSb) is another example of a network intended to address both national and regional needs for researchers and stakeholders on coastal and ocean issues. In coordinating federal, state, local, and private needs, AOOS identifies gaps in data, helps fill those gaps when appropriate, and increases the usefulness of existing data. AOOS demonstrates the integration and collaboration necessary to enable a variety of users to obtain information and to make decisions about the marine environment in the Alaska region.
mechanisms to support and facilitate these sustained conversations would need to be deliberately identified and implemented, as opposed to relying on self-organization. This would require close and effective engagement among public, private, and academic institutions. Participants in these conversations could serve a role akin to that of a diplomat, seeking and communicating ideas and suggestions that reflect a broad viewpoint. There are excellent examples of efforts underway, on both national and local levels, to develop and facilitate interactions among relevant research communities and users (Box 3.1).
Looking more specifically at the issue of sea ice prediction, one example of an initial framework for capturing stakeholder needs in terms of variables utilized in sea ice science is provided in Table 3.1. Other frameworks can be used to organize user-science connections with the goal of advancing the utility of sea ice predictions. For this process to be successful, it is important for communities to learn each others’ language and to be aware of usage differences. For instance, the terms sea ice “memory” and “persistence” have specific meanings within the sea ice research community that may not be appreciated by
Information is made available through various data management and information products (including a website and data portal), workshops and reports, and newsletters. AOOS also makes educational resources available for teachers and interested community members. The AOOS network includes mariners, fishermen and subsistence users, search and rescue operations, scientists, coastal security operations, resource managers, and educators.
A regional-scale example of a forum through which stakeholders work directly with scientists is the Alaska Center for Climate Assessment and Policy (ACCAPc). The goal is to enable Alaskans to be able to respond to climate changes by targeting products (e.g., National Oceanic and Atmospheric Administration [NOAA] forecasts) to address specific user needs. Using tools such as webinars, video conferences, web-based guides and maps, and social media, ACCAP reaches out to encourage dialogue between scientists and end users. These resources help convey valuable information on climate change science as well as information on uncertainty and risk management. Strategies and plans to adapt to climate change are developed in coordination with stakeholders, agencies, industries, and citizens to ensure that partnerships are built and information needs are met. The ACCAP program is part of the NOAA Regional Integrated Sciences and Assessments (RISAd) program, which supports research on complex, interdisciplinary issues that are addressed at the regional level.
others. Even commonly used terms such as “multiyear” sea ice may have different meanings that need to be clearly communicated among the relevant stakeholders.
These ongoing efforts and suggested framework offer important building blocks to advance the strategy envisioned here. However, in the committee’s view this activity will likely not be effectively facilitated without a dedicated and deliberate effort backed by sufficient resources, including designated funding. For example, an NRC report noted that maintaining some networks developed and cultivated during IPY has been difficult and many of its valued components—such as the international IPY website, its publication database, and educational and outreach efforts—have struggled to find alternative resources (NRC, 2012a). The characteristics of these sustained conversations suggest leadership from a high-level, inter-governmental office, agency, or consortium.
TABLE 3.1 Example of an initial framework that could be useful in determining key stakeholder needs and prioritizing variables to be measured by the research community to better meet those needs.
|Stakeholder Activity||Ice edge location||Ice concentration||%FY/MY||Mean FY thickness||Mean MY thickness||Ice growth rate||Ice melt rate||Floe size||% Ridged ice||% Leads mean width||Snow depth||Melt/ freeze onset||Fast ice break-up date||Ice velocity||Ice character||Local||Reg-ional||Pan-Arctic|
|Offshore resource extraction|
NOTE: The top row lists sea ice variables and measurements. Appropriate cells could be filled with check marks or indications of priority. If such a table is completed by various stakeholders, a particular variable (e.g., floe size on seasonal timescales) for which a large number of stakeholders indicate a need may be given higher priority than a variable needed by fewer stakeholders. This table also shows the breadth of stakeholders and their needs. It can be used for both seasonal and decadal timescales.
SOURCE: Generated from discussions at the workshop.
While recognizing that there are limitations in current modeling and observational techniques, the committee offers possible strategies to significantly enhance our understanding and predictions of Arctic sea ice cover over seasonal to decadal timescales. Implementation of these proposed strategies will require iterative interaction between model development and observational input, balanced by a sustained dialogue with end users.
Key Strategy: Evaluation of Existing Seasonal Prediction Methods
A coordinated and detailed comparison of the different approaches used to generate seasonal sea ice forecasts could establish baseline expectations for predictive skill and identify priority needs, setting the stage for advances in predictive capability.
As previously discussed, several methods are used to obtain seasonal forecasts of Arctic sea ice, including (1) statistical methods, (2) ice-ocean models driven by prescribed atmospheric forcing, and (3) fully coupled atmosphere-ocean-ice models. A coordinated comparison of these prediction methods will serve to inform both the science and the needs of stakeholders. It is important that the evaluations be based on regional metrics and not be limited to ice coverage. For instance, other candidate metrics are dates of ice retreat and closure of various sea routes or specific coastal locations. Various comparison approaches need to be considered, such as hindcast and so-called “perfect-model” studies.
In a hindcast model study, a retrospective assessment of past years, both initial conditions and validation data are needed. An evaluation of this kind was recently performed on forecasts of the El Niño-Southern Oscillation phenomenon (Barnston et al., 2012; Box 3.2). In the case of sea ice prediction experiments, initialization will be limited by the accuracy of key predictor variables (e.g., ice thickness), but such limitations will be common to all three approaches. Perfect-model studies can also provide useful insights to predictability. These studies treat simulation output from a coupled numerical model reference experiment as the “truth” (i.e., equivalent to observations). The different prediction methods discussed above can then be applied to forecast the conditions from this reference simulation. This can be performed for both past and future model-simulated conditions, allowing for information on how predictability characteristics may change with changes in the climate state. Such studies have the advantage of a complete knowledge of initial and time-varying conditions and provide the ability to address the implications for possible nonstationarity in statistical relationships for future climate states. It is important to remember that the results from these studies need to be considered within the context of the imperfect coupled model being used.
COMPARATIVE EVALUATION OF SEASONAL FORECASTING METHODOLOGIES FOR ENSO
Given that the El Niño-Southern Oscillation (ENSO) affects climate and weather events (e.g., drought, flooding, and tropical storms) on a global scale, understanding and improving forecasts are critical to both the scientific community and the public (McPhaden et al., 2006). Real-time ENSO prediction capability during the 1990s was assessed to be somewhat useful (Barnston et al., 1994), with dynamical and statistical models showing comparable skills.
Although there has been progress in the prediction of ENSO during the last decades (Randall et al., 2007), numerous issues regarding its dynamics, impacts, and predictability remain uncertain. In the last two to three decades, the ability to predict warm and cold episodes of ENSO at short and intermediate lead times has gradually improved due to:
• Improved observing and analysis/assimilation systems,
• Improved physical parameterizations,
• Higher spatial resolution, and
• Better understanding of the tropical oceanic and atmospheric processes underlying the ENSO phenomenon (e.g., Guilyardi et al. 2009).
Findings from a more recent study on ENSO predictions suggest that additional advances in capabilities are likely with the expected implementation of better physics, numeric and assimilation schemes, finer resolution, and larger ensemble sizes (Barnston et al., 2012). The study evaluated real-time model predictions of ENSO conditions during the 2002-2011 period and compared them with skill levels documented in studies from the 1990s. The skills of 2002-2011 models is slightly better than that of earlier decades, with the recent decade’s dynamical ENSO prediction models outperforming their statistical counterparts to a slight but statistically significant extent. The greater power of dynamical models is largely attributable to the subset of dynamical models with the most advanced, high-resolution, fully coupled ocean-atmosphere prediction systems using sophisticated data assimilation systems and large ensembles (Barnston et al., 2012).
Questions surrounding the impact of the trend toward an increasingly seasonal Arctic sea ice cover could be addressed with the development of a highly coordinated and integrated process-based study, analogous to the Surface Heat Budget of the Arctic Ocean (SHEBA) project, focused on understanding oceanic, atmospheric, and terrestrial contributions to seasonal sea ice predictions.
The fundamental properties of the ice cover are changing as the Arctic transitions toward a seasonally ice-free state, resulting in a significant reduction in the amount of multiyear ice compared with first-year ice. In the face of this significant transition, there is the need to identify and understand whether and how key parameters (i.e., first-order effects) influence predictability. A likely outcome is the need for improved model formulations of the dynamic and thermodynamic processes governing the behavior of a sea ice cover composed of largely first-year ice.
Previous work done on the fundamental properties of first-year sea ice (e.g., Weeks and Ackley, 1982; Timco and Weeks, 2010) can inform the design of process studies that will advance our understanding of first-year ice in a predictive context. The challenge lies in developing a thorough understanding of the fundamental properties of first-year sea ice act together on a large, pan-Arctic scale to affect air-ice and ocean-ice heat transfer (ice thermodynamics) and ice pack mobility (ice dynamics, e.g., Melling et al., 2005; Amundrud et al., 2006; Barber et al., 2012). Moreover, as the sea ice cover becomes dominated by the weaker and less stable first-year sea ice, it may be more susceptible to extreme events that could have impacts lasting from seasonal to decadal timescales.
Observations of these processes and their interactions are needed to determine which aspects of existing predictive models require further development and to help determine requirements for sustained observations necessary to verify model realism as the ice cover continues to evolve. Conversely, model experiments can identify which process parameterizations are the greatest sources of uncertainty (or error) in climate model simulations. Systematic sensitivity experiments, performed with an ensemble of different models, would initiate an end-to-end process study in which the needs of models guide field experiments, which in turn feed back (via improvements in process formulations or parameter estimates) to models used for sea ice prediction.
An end-to-end process study in the seasonal ice zone, guided by past work, historical data and the output from sensitivity studies using current models, would enhance process understanding and simulation capability. Although process understanding is especially important for predicting the evolution of initial conditions over seasonal and interannual timescales, it also has potential for multiyear timescales, based on the apparent multiyear timescales of the ocean inflow anomalies, at least in the Atlantic sector (Polyakov et al., 2010). The nature of a process-based study of
Examples of SHEBA-Like Initiatives
There are a few initiatives under way that could be built upon and possibly integrated to conduct a SHEBA-like project in seasonal ice:
• Office of Naval Research 5-year Emerging Dynamics of the Marginal Ice Zone Department Research Initiativea (ONR MIZ-DRI). The 5-year ONR MIZ-DRI was initiated in 2011 and includes plans for a field project in 2014. The observational dataset that is generated from an integrated suite of platforms can be used to evaluate the skill of numerical forecast models.
• Ocean Observations to Improve Sea Ice Forecasting. This project was initiated during the Arctic Observing Coordination Workshop, held March 20-22, 2012, in Anchorage, Alaskab. In its very early planning stages, the project is designed to provide the necessary ocean observations to improve sea ice forecasting on daily, seasonal, interannual and decadal timescales.
• Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MoSAIC).c:. The concept of MoSAIC is to establish an international, multiyear, manned, drifting observatory in the central Arctic sea ice pack to obtain observations of atmosphere, sea ice, and ocean processes that will compose a testbed for process, regional, and global model evaluation and development.
atmosphere-sea ice-ocean coupling in the seasonal ice zone will almost certainly require a set of observations over a full annual cycle. The process study is also inherently interdisciplinary because of the broad science scope encompassing surface-to-satellite observations and models extending across the atmosphere, ocean, seafloor, and land.
The year-long Surface Heat Budget of the Arctic Ocean (SHEBA) project (Perovich et al., 1999), conducted from 1997 to 1998 in multiyear ice, serves as an excellent model for a study of this kind. SHEBA was designed to facilitate interactions between observation and modeling research communities, through the planning, implementation, and analysis phases of the project. Now that it has been over a decade since the project formally concluded, it is apparent that one of the major successes of SHEBA was the interdisciplinary teamwork that brought together a diverse group of researchers, each bringing their own particular expertise, to work on the common goals of the program (Perovich et al., 2003).
SHEBA continues to motivate cross-cutting collaborations that advance our
• Year of Polar Predictiond (YOPP): This is a joint effort from the World Weather Research Progamme and the World Climate Research Programme. The YOPP is tentatively scheduled for 2017-2018 and will include an intensive observation and modeling period to provide data for model development, data denial experiments, and predictability and diagnostic studies.
Other examples of successful studies of seasonal ice in the past 15 years include:
• International North Water Polynya Study (NOW): This program was initiated to study the North Water polynya and the mechanisms associated with its formation and biological production.
• Canadian Arctic Shelf Exchange Study (CASES): This was an international effort under Canadian leadership to understand the biogeochemical and ecological consequences of sea ice variability and change on the Mackenzie Shelf.
• Circumpolar Flaw Lead (CFL) Study: The objective of this project was to examine how physical changes affect biological processes within the flaw lead.
understanding of the processes governing sea ice thermodynamics. This significant and enduring outcome suggests that the investment in a large, focused effort, if effectively coordinated and implemented, can have a greater impact than a more diffuse approach. If followed, a key addition to the approach used during SHEBA would be the increased involvement of stakeholders outside of the sea ice research community, including a greater emphasis on the Arctic marine ecosystem, atmospheric chemistry, the coastal terrestrial system, and, more generally, end users (e.g., indigenous populations, natural resource industries, fishing communities, commercial shippers, and marine tourism operators). Some examples of SHEBA-like initiatives can be found in Box 3.3.
A comprehensive process study in the seasonal ice zone also offers the opportunity to identify, develop, and test instruments and observational platforms that can effectively and efficiently support both seasonal and decadal prediction capabilities. Measurements of sea ice motion, ice thickness, and snow cover/depth are made from a variety of sensors—including in-situ, airborne, and satellite instruments—each having different capabilities. Existing surface
networks (e.g., buoys and weather stations) will need to be sustained and extended with aircraft campaigns and new instruments. Continuous satellite observations (e.g., multispectral sensor, synthetic aperture radar, passive microwave radiometer, active microwave scatterometer, and altimeter) over decadal timescales are critical to assess the role of sea ice change in the global climate system.
To enhance predictive capability, the process study needs to have a focus on improving predictions on seasonal to decadal scales. It also needs to involve stakeholders and modeling and observational communities in its planning and implementation. If designed in this way, results from the process study can play a central role in selecting and optimizing a suite of observations that can best meet the broad requirements for more robust, sustained circumpolar and regional observations to meet different demands, including initial conditions for models, improved process understanding, model validation, and long-term prediction.
Influxes of oceanic heat to the Arctic from the Pacific Ocean and the Atlantic Ocean have likely contributed to the recent loss of Arctic sea ice (e.g., Shimada et al, 2006; Walsh et al., 2011). In both regions, the warmer water subducts and circulates below the fresh surface layer of the Arctic Ocean. The rates, locations, and processes by which these heat sources reach the overlying sea ice cover and affect sea ice anomaly evolution are poorly understood, in large part because of a lack of in situ observations. A mechanism involving reductions of ice concentration, increased responsiveness to wind forcing, and enhanced mixing has been proposed by Shimada et al. (2006), while processes of double diffusion and eddy mixing have been suggested as mechanisms by which Atlantic water heat may move upward in the water column north of Eurasia (Polyakov et al., 2011).
Compounding the uncertainty about the role of ocean processes is the varying ability of global and regional ocean models to reproduce the vertical structure of temperature and salinity in the upper layers of the Arctic Ocean. Inadequacies in process formulation as well as vertical resolution are likely sources of errors in the upper Arctic Ocean in global and regional models. Moreover, the CMIP5 decadal predictability experiments, which are targeting seasonal to decadal predictability inherent in ocean initializations, have yet to address the role of the Arctic Ocean in interannual-to-decadal prediction. This is largely because there are few data to initialize the Arctic Ocean in decadal-scale hindcast experiments.
There is a particular need at this time for a coordinated effort to design and implement a set of model sensitivity studies that will provide quantitative metrics to assess the impact of various observation types, locations, and densities on seasonal sea ice forecasts.
Observations play a critical role in seasonal and decadal sea ice prediction. There are broad requirements for more robust sustained circumpolar and regional
observations to meet different demands, for instance, initial and boundary conditions for models, improved process understanding, model validation, and long-term prediction. Sea ice model predictive capabilities have evolved and will continue to evolve. That said, to the extent that particular processes (e.g., ocean heat fluxes and cloud-radiative interactions) are found to exert high leverage on the models’ simulation of the Arctic system and sea ice prediction in particular, the diagnostic model intercomparison can point to priorities for observations and/or process studies targeted at improved model simulations and predictions of sea ice. This should build on previous work that has identified areas of systematic model bias (e.g., Sorteberg et al., 2007) and the importance of various simulated feedbacks (e.g., Winton, 2006; Kay et al., 2012).
The modeling infrastructure, especially as it pertains to data assimilation, has advanced sufficiently that one can now envision a series of sensitivity studies designed to strategically inform research investments related to observational needs for sea ice modeling and prediction.
At the seasonal timescale, observing system experiments (OSEs—testing impacts of actual observations) and observing system simulation experiments (OSSEs—testing sensitivities to hypothetical or simulated observations) can be directed at systematically investigating the effects of specific observations on prediction capabilities. Among the limited studies to date, Inoue et al. (2009) have shown that the assimilation of sea level pressure measurements from the International Arctic Buoy Program improve atmospheric reanalyses. However, impacts on atmospheric reanalyses are not equivalent to impacts on seasonal sea ice predictions (and sea ice simulations in general). Model-derived predictions of sea ice for timescales of several seasons are almost certainly affected by the initialization of ice thickness (and corresponding distributions of ice concentration), but the atmospheric initialization probably has little effect. There is large uncertainty, however, about the importance of initializations of other variables, such as snow on sea ice, ocean temperature and salinity profiles below the ice, and ocean current distributions.
The modeling tools to conduct these data-model synthesis experiments exist. For instance, coupled atmosphere-ice-ocean forecast models, such as the National Centers for Environmental Prediction’s (NCEP’s) Coupled Forecast System, routinely assimilate observational data in producing seasonal forecasts of the atmosphere, ocean, and sea ice out to a range of a year. Ice-ocean models have been developed and used to produce forecasts for the SEARCH Sea Ice Outlook1 (Figure 3.1), as well as for ice-ocean model intercomparisons. In addition, data assimilation systems have been developed for reanalyses of the Arctic atmosphere and ocean (Bromwich et al., 2010; Panteleev et al., 2011).
It is noteworthy that whereas all of the forecasts represented in Figure 3.1 projected a 2012 September minimum ice extent that
FIGURE 3.1 July 2012 Pan-Arctic Sea Ice Outlook. Ice-ocean models have been developed and are one method used to produce forecasts for the SEARCH Sea Ice Outlook. This figure shows 21 community estimates of the expected minimum of sea ice for 2012. The projected Arctic sea ice extent median value for September 2012 was 4.6 million square kilometers. However, on September 16, 2012, Arctic sea ice reached a minimum extent of 3.41 million square kilometers, the lowest seasonal minimum extent on satellite record. SOURCE: Adapted from ARCUS and SEARCH.
was well below the long-term (1979-2007) average, they all also overestimated it. This result suggests that although there is a skill in the seasonal forecasts relative to climatology (albeit a changing climatology), there remains the need for a concerted effort to improve sea ice predictions—the need that motivated the present report.
In the case of seasonal predictions, the existing capabilities in data-model synthesis have yet to be exploited in the Arctic because prediction capabilities at these timescales are relatively new and because the Arctic observational and modeling communities have tended to be distinct. Therefore, as a first step in designing OSEs and OSSEs, there is a particularly urgent need for a coordinated effort by these communities to design a set of experiments that will provide quantitative metrics of the impact of various observation types, locations, and densities on seasonal sea ice forecasts, as well as the accuracy and temporal resolution that are required.
Possible observational variables for inclusion in these experiments are ice thickness distributions, ice extent and concentration, snow on sea ice, and the upper-ocean profiles of temperature, salinity and current velocities. The latter category of observations includes ocean measurements not only from under the ice, but from the surrounding open ocean. In addition to these state variables, consideration should be given to measurements focused on the exchange of energy between the air-ice and ice-ocean boundaries, which drive ice dynamics and thermodynamics (e.g. radiation, sensible heat, moisture, and momentum) The OSEs and OSSEs would address impacts of measurement errors as well as varying distributions of measurements.
These types of experiments may be regarded as prerequisites for the design of an Arctic observing network (NRC, 2006), and seasonal sea ice prediction can provide a compelling focus for such experiments. Further, many of the variables listed above (e.g., ice thickness, snow on sea ice, and under-ice ocean profiles) are observational challenges in their own right. The logistics and expenses involved in obtaining these measurements adds to the urgency of OSEs and OSSEs to justify, for sea ice prediction and for other applications, future investments in the observations.
Key Strategy: Enhanced Numerical Model Capabilities
Enhancement of model-based predictive capabilities will require coordinated experiments to (a) identify which variables and processes are critical to simulating a realistic ice cover, (b) investigate the source of climate model drift, and (c) guide decisions regarding high-priority model development needs and the expansion of models to include additional capabilities and variables of interest to stakeholders.
HURRICANE FORECAST IMPROVEMENT PROGRAM
The NOAA-led Hurricane Forecast Improvement Program (HFIP) is a 10-year program started in 2007 with the overarching goal of improving hurricane forecast skill, with an emphasis on hurricane intensity and structure.a Hurricane forecasts have improved significantly in the last 10 years, but rapid intensification remains a significant challenge.
The specific goals of HFIP are to: 1) improve the accuracy of hurricane intensity and track forecasts and 2) increase the forecast confidence of customers and decision makers, especially those in the emergency management community (NOAA, 2010). To help facilitate these goals, the National Oceanic and Atmospheric Association has partnered with other federal agencies, the National Center for Atmospheric Research, universities, and the Naval Research Laboratory (NRL). These collaborations help address the challenges associated with transitioning new forecasting research and technology into operations. Furthermore, there has been an effort within HFIP to ensure open access to the data involved (NOAA, 2010).
differences among models and, in some cases, the component(s) of the model that is the source of those differences. To date, however, MIP efforts have had limited success in pinpointing the variables or processes that are the root cause of simulation errors, and any conclusions have often not efficiently fed back to the model developers to improve the models.
A new strategy for model intercomparison is needed that will identify specific, key processes of importance to sea ice prediction; incorporate lessons learned from model sensitivity studies; and collaborate closely with model developers to identify approaches to resolve unrealistic model behavior. Regional models and ice-ocean coupled systems will likely be an essential part of the strategy, given the greater control achieved in these approaches by prescribed (e.g., observationally- or reanalysis-derived) lateral and/or surface forcing of the Arctic. Interestingly, one outcome of these studies, along with the identification of factors that influence sea ice prediction skill, may be to realize simplifications that can be applied to coupled models. This result would allow for models with reduced complexity to be used for seasonal-scale sea ice prediction.
At the decadal timescale, where predictions are largely influenced by forcing, model sensitivity studies explore and quantify the impact of a range of parameters or representations of physical processes on predicted model outputs. These experiments show how a particular scenario may be affected by multiple parameters. Performing
targeted sensitivity studies with new parameterizations can reveal weaknesses of other parameterizations. These simulations make it possible to analyze the sensitivity of simulation results to some of the decisions made in model development. The weather community may be able to contribute lessons learned for conducting sensitivity studies (Box 3.4).
As the Arctic ice cover transitions to a predominantly thinner state, the decadal model sensitivities among variables affecting ice growth, melt, and movement may change relative to those of the past. Targeted sensitivity studies would help identify which variables and processes are critical to simulating a realistic ice cover in the new state, including the mean climatology, simulated variability of and response to changes in external forcing, and where uncertainties inherent in model parameterizations have the largest influence on simulated sea ice processes. Through collaborative efforts with the model development community, this should feed back into improvements in the physics of the models.
A particularly important issue to address via the model sensitivity studies and intercomparison activities is climate model drift from an observationally-based initialized state, which contaminates predictions on seasonal to decadal timescale. This drift results from systematic biases in coupled climate simulations. Improvements in model simulations are required to address this issue. Research on data assimilation methods and alternative methods of model initialization, for example, by using anomaly fields, will need to be considered. Additionally, various mechanisms to “de-drift” predictions need to be assessed in retrospective studies to determine their utility in realizing useful predictive skill.
Related to the call for targeted model sensitivity studies, there is a need for enhanced capabilities in numerical models to provide useful information on key variables of interest. For example, many climate models do not distinguish between land-fast ice and other sea ice, yet the behavior of land-fast ice is of keen interest to a variety of stakeholders. The date of spring breakup is a particularly influential event for coastal infrastructure and operations, but most models lack sufficient resolution and the specific processes that govern evolution of land-fast ice. Many of these requirements call for predictions that have sufficient spatial detail to resolve the highly varying ice characteristics that occur near the coastline. The nature of seasonal sea ice prediction demands accuracy within a few hundred kilometers of shore and within the marginal ice zone. In addition, forcing from tides and ocean waves may play an important role in sea ice evolution on seasonal to decadal timescales. These factors are not typically considered in large-scale numerical models. Model enhancements that incorporate these and other relevant processes would allow for investigations of their role in sea ice prediction and ultimately result in better predictive skill and more useful information for stakeholders.
One way in which model capabilities can be enhanced is by finer resolution. Recent studies have shown that models with higher horizontal and vertical resolution are able to more realistically simulate certain processes
in the atmosphere (e.g. Byrkjedal et al., 2008; Girard and Bekcic, 2005) and ocean (e.g., Fieg-Rudiger et al., 2010) Higher resolution information may also be necessary to meet certain stakeholder needs. However simply increasing model resolution is not a panacea, because enhanced computational and storage costs need to be considered in light of the relevant benefits for sea ice prediction. Moreover, model parameterizations that have been developed for coarse resolutions may not be ideal for considerably finer spatial-scales (e.g., Lipscomb et al., 2007; Girard et al., 2009) and may need to be revisited, requiring further model developments. Nevertheless, with computational resources increasing and likely benefits in terms of simulation quality, increased resolution in global and regional models together with regionally refined and adaptive model grids need to be explored in the context of benefits for sea ice predictive capability.
Key Strategy: Improved Information and Data Management
Given the vast amounts of disparate data on Arctic sea ice and the numerous stakeholders who use these data, there is a need for a coordinated and centralized information hub for Arctic datasets that facilitates timely access to observational and modeling results and encourages sustained communication among stakeholders.
No single organization or agency has adequate resources to systematically undertake the task of robust field observations, data synthesis, and environmental modeling. Collaborative efforts and data sharing are therefore essential. Moreover, data continuity is a fundamental imperative so that long-term Arctic sea ice trends can be ascertained and provided to stakeholders for reliable planning. Sharing data also enables researchers to communicate and collaborate more effectively.
Given the vast amounts of disparate data on Arctic sea ice, background information, model results, observational date, etc. can be difficult to find for the numerous stakeholders who use these data, particularly for new users. The committee acknowledges that a more centralized framework could improve information management (Parsons et al., 2011). Although there are numerous data repositories for climate-relevant data, they tend to be scattered and inconsistently cross-linked. Rich measurement datasets are often reduced to their basic parameters with a loss of important information. Also, data transfer and data transformation at data centers add additional layers of complexity and data latency.
In the committee’s opinion, the main purpose of a centralized information hub is to serve as a primary launching pad for searches aimed at gaining access to this wide array of information. The intention is not to recreate existing and diverse resources, but to facilitate the ease of their retrieval. A central information hub would unify the various databases, providing a seamless and consistent system for information and data
DATA SHARING: TWO CASE STUDIES
Some progress has been made in data management. Case 1: As of 2011 Shell, ConocoPhillips and Statoil have signed an agreement with the National Oceanic and Atmospheric Association to formally share scientific information that the companies have acquired individually and jointly in the Arctic region.a
This collaboration leverages the complementary strengths of NOAA’s scientific expertise and the industry’s significant offshore resources. Scientific datasets for the Arctic region are shared, including weather and ocean observations, biological information, and sea ice and seafloor mapping studies. NOAA’s ability to monitor climate change and provide useful products and services that inform energy exploration activities in the Arctic will likely be improved through the sharing of high-quality data. The integration of these data could also provide a greater national capacity to effectively manage and respond to environmental disasters in an area where limited personnel and facilities exist. Data and information are shared with the public through NOAA’s existing outlets. Quality control on all data provided is conducted by NOAA before it is incorporated into its products and services.
Case 2: Since 2011, all National Science Foundation proposals must include a supplementary “data management plan,” which is subject to peer review. The primary goal of this new data sharing policy is to “assure that products of research help NSF achieve its mission to promote the progress of science and engineering.”b The plan should outline the types of data to be produced, the standards to be used, the policies for data access and reuse, and the plans for archiving.
discovery and access. Recent Web standards provide distributed databases that appear uniform and singular to the user. Therefore it is not necessary to create new archives, but rather to leverage existing infrastructure. A key characteristic of the central information hub and the individual components that lie behind it is the timeliness of the available resources. This is particularly critical for applications related to seasonal sea ice predictions, which require real-time data access for model output, in situ observations, satellite and aircraft survey data, etc. The centralized hub could also serve as an integrating resource, providing access to information on the various elements of the Arctic sea ice system (e.g., ocean, atmosphere, and sea ice.)
A separate, but related issue is long-term data storage limitations. In the climate modeling community, the push toward high-resolution and complex models coupled with diverse stakeholder needs has resulted in a rapid and increasing demand for data storage, analysis and distribution (NRC, 2012a). Thus many climate modeling
groups are increasingly limited by the cost of long-term data storage.
Consistent data storage protocols need to be adopted that preserve ancillary data and original sample rates along with interoperability standards for data interchange. Indexes need to be based on agreed metadata vocabularies (e.g., Marine Metadata Interoperability Project5) for search and reference efficiency. Data density and timely accessibility of information to and from nonfederal sources are some obstacles to be resolved, although there has been slight progress (Box 3.5).
It is hard to dispute the widespread desire for a central information hub and the value that could be derived from it. The major challenge lies in its initiation. There is no shortage of candidate organizations well suited to facilitate the design and implementation of a central information hub, though effective data management requires resources. Because permanence is a critical characteristic, there would be an expectation of long-term and stable support for this important activity. Given the broad and critical nature of these needs, which reaches beyond the issue of sea ice predictability, it may be most appropriately addressed by a high-level, cross-cutting entity.