Determining the economic value of climate and weather forecast information remains challenging (Dutton, 2002; Lazo et al., 2011; Letson et al., 2007; Morss et al., 2008; U.S. Department of Commerce, 2014). Some work indicates that a significant portion of annual U.S. gross domestic product (tens of billions or even trillions of dollars) may be sensitive to the weather (Dutton, 2002). Regardless of forecasts’ exact economic value, there is growing recognition that subseasonal to seasonal (S2S) predictions could play an important role in reducing society’s exposure to weather, climate, and other environmental variability, both in the United States and globally (e.g., Thiaw and Kumar, 2015; World Bank, 2013).
Realizing potential benefits of S2S predictions will require physical science research to advance understanding of the many complex interactions at play within the Earth system and to overcome the many technical hurdles associated with translating such research into improved S2S forecast systems (see Chapters 4 through 7). However, a crucial aspect of realizing the value of S2S forecasts involves generating and applying knowledge about the many social and behavioral dynamics, as well as the legal and equity issues that are associated with using such forecasts to improve decision-making (NRC, 1999, 2010a, 2010c). Although addressing the latter set of issues in their entirety is beyond the scope of this report, the committee believes that it is important to highlight in more detail the value proposition of S2S forecasts and to outline critical steps that the S2S research community can take to ensure that investments made in current and future S2S forecast systems are leveraged to maximize the ability to inform choice, action, and social and economic benefit.
This chapter presents the context in which S2S forecasts are or could be used by a diverse set of decision-makers, highlights some barriers to use, and presents findings and recommendations to help ensure that future S2S forecast systems and forecast products realize their potential to benefit society.
A number of federal, state, and private users presented information to the committee about how they make decisions for which S2S Earth system information is or has the potential to be a factor. Building on these presentations, the committee developed information on a large set of sectors and decisions for which S2S forecasts are or have the potential to inform decisions (Figure 3.1 and Table 3.1). Case studies later in this chapter expand upon some of these examples, including applications to water management, public health, emergency response, and national defense. As the case studies make clear, some of the potential value of S2S forecasts lies in their ability to inform decision processes that begin months or even years in advance of a potential event.
TABLE 3.1 Example Decisions from a Range of Sectors That Can Be Informed by S2S and Longer Forecasts
|Water Resources Management (see case study for more detail)||Water supply management (including flood control and drought)||Probability of heavy rainfall or runoff; probability of unusually high demand (precipitation; temperature; snowpack; runoff; likelihood of atmospheric river events)||Allocation of water supply; water transfer requests; assuring minimum flows for endangered species (accumulation of winter snowpack; timing of seasonal snowmelt; summer water demands; precipitation; temperature; snowfall; evapotranspiration)||Storage capacity and sources; conservation programs (changes in mean annual temperatures, precipitation, snowfall accumulation, runoff, evapotranspiration)|
|Hydropower scheduling||Available water supply in reservoirs; anticipated demand (lake levels; stream flow; evaporation; temperatures for demand estimates)||Probability of reaching target elevation levels in reservoir (snowmelt/inflow; evaporative loss)||Changes in demand and supply (changes in mean seasonal and annual temperature across basin and service area; changes in snowmelt patterns; changes in precipitation; changes in evapotranspiration)|
|Recreation budgeting||Reservoir/lake levels and temperature—e.g., high temperatures may increase probability of algal blooms or fish kills (inflow; evaporative loss; temperature departures)||Probability of reaching target elevation levels in reservoir/lake (snowmelt/inflow; evaporative loss)||Probability of maintaining seasonal target elevation levels (net water supply to reservoir/lake; changes in evaporative loss)|
|National Security (see case study for more detail)||Anticipating disruptive events / deployment of resources (aid, security, evacuation)||Pre-deploy resources to areas that are at greatest risk of high-intensity events (probability of disruptive events, especially flooding and drought)||Anticipate staffing and resource needs; identify timing of Arctic shipping lanes open (sea ice; probability of disruptive events including flooding and famine)||Identifying areas that may become at-risk from natural disasters—e.g., climate change, famine (regional changes in temperature and precipitation patterns; sea-level rise)|
|Food and water security||Emerging areas of food or water shortage that may require transport of large quantities of food (precipitation departures; monsoon)||Areas at risk of famine or flood during coming months to year (monthly to seasonal precipitation; drought forecasts; temperatures exceeding critical thresholds for major crop areas)||Areas undergoing desertification or decline in water quantity and/or quality (changes in precipitation patterns; salinity; changes in jet stream, monsoon, or Inter Tropical Convergence Zone [ITCZ])|
|Tactical planning||Shipping routes and operations planning (wind; wave height; sea ice; ocean currents)||Projected dates of Arctic ice breakup and thawing permafrost rendering ice roads and runways unusable (sea ice; monthly temperatures)||Inundation of coastal facilities from sealevel rise and storm surge (sea-level rise; changes in hurricane intensity)|
|Energy||Energy generation, trading, and hedging||Potential spikes in demand; availability of renewable energy (heat waves/cold outbreaks; mean daily wind speed; daily solar radiation; adverse weather impacts such as ice storms, wind storms, hurricanes)||Seasonal supplies of natural gas and renewable energy sources; trading with other producers, hedging in futures markets and over-the-counter trades; fuel adjustment clauses (winter and summer temperatures; probability of heat waves or cold outbreaks; projected runoff from snowmelt)|
|Operations and maintenance scheduling||Potential disruptive events/damage to infrastructure or supply chain (severe storms; hurricanes; floods; heat waves; wind)||Taking units offline (probability of heat waves or cold outbreaks leading to unusually high demand)||Building, upgrading, and relocating new facilities (summer and winter temperatures; sealevel rise; changes in snowpack/spring runoff)|
|Agriculture||Crop production||Susceptibility to disease; application of nutrients, pesticides, and herbicides (temperature; precipitation; wind speed; relative humidity; soil temperature)||Projected yields; food production and distribution (precipitation; soil moisture; temperature; projected dates of first/last freeze; probability of disruptive events—flood, drought, heat waves, freeze)||Types of crops that can be grown in a changing climate; trees and vine varieties (changing ecoregions; changes in monthly and seasonal precipitation; evapotranspiration; length of growing season)|
|Commodity trading in grains and other high-value crops||Protect profit; anticipate market movement||Protect profit; anticipate market movement|
|Ranching||Forage management strategies; altering stocking rates (probability of abnormally wet or dry weeks; extreme temperatures; abrupt changes in temperatures)||Herd size, pasture availability (total rainfall; vegetation health)||Long-term changes in viability of operations in semi-arid areas (precipitation; evapotranspiration; frequency of drought)|
|Fisheries||Stocking, fish kills (water temperatures; stream flow; salinity)||Migratory patterns, e.g., salmon (snowpack; streamflow)||Viability of species-appropriate habitats in lakes and rivers (temperature; water temperature; streamflow; snowpack)|
|Severe Weather/Event Management||Event management||Pre-deploy resources to areas most likely to be impacted (probability of disruptive events—storms, hurricanes, floods, fire)||Seasonal outlooks for number and intensity of hurricanes, storm outbreaks, or flooding (weekly to monthly precipitation accumulation; patterns favorable for development of storms; El Niño-Southern Oscillation (ENSO) phase and intensity)||Areas likely to become more or less at-risk from disruptive events (changes in wildfire frequency or magnitude; changes in extreme precipitation; changes in drought; changes in storm tracks; changes in hurricane frequency or intensity)|
|Risk awareness||Encouraging people to stock sufficient supplies (probability of disruptive events)||Initiate public awareness and preparedness campaigns (probability of an active season—hurricane, storm, flood)||Changes in patterns or timing of severe weather (changes in frequency or magnitude of disruptive events)|
|Wildfire management||Pre-deploying resources, wildfire management (temperature; wind; humidity)||Seasonal outlooks (precipitation; temperature; wind; fuel load)||Changes to fire susceptibility (expansion of pine bark beetle habitat; changes in seasonal water balance; changes in temperature)|
|Environmental Impacts||Oil spill||Loop currents (e.g., tracking where oil is likely to go)||Dispersion and dilution; impacts on fisheries||Changes in natural habitats|
|Coastal zone management||Hurricane / wave impacts||Beach erosion and re-nourishment||Loss of wetland habitat due to sealevel rise; Changes in shoreline habitat and wildlife (e.g., conversion of salt marshes to mangroves)|
|Transportation||Shipping and navigation||Disruptions to surface transportation systems; preparing evacuation routes for hurricanes (probability of flooding; periods of active tropical activity)||Timing of opening shipping lanes in the Arctic (sea ice; summer temperatures; streamflow on major waterways)||Susceptibility of ports to inundation; transit routes (sealevel rise; storm surge; ice-free Arctic)|
|Maintenance of highways, railroads, waterways, airports||Positioning equipment and assets, e.g., salt for roads, barges and railcars for transportation, deicing equipment and supplies for airports (probability of adverse weather, including snowfall or ice, heavy rainfall, drought)||Positioning equipment and assets for repairs of infrastructure and equipment; seasonal supplies of road salt, deicing supplies, fuel; (probability of favorable, adverse, or severe weather; number of freeze/thaw cycles; first and last frost; seasonal snowfall; ice storms)||Resizing of bridges and culverts to handle flood flows; selection of materials to handle extreme temperatures (projected number of days exceeding critical temperature thresholds; changes in maximum probable precipitation)|
|Maintenance||Positioning equipment and assets, e.g., salt for roads (probability of winter weather including snowfall or ice)||Planning for pothole repairs; seasonal supplies of road salt; possible repair of flooded roadways and bridges (probability of extreme rainfall; number of freeze/thaw cycles; first and last frost; seasonal snowfall; ice storms)||Resizing of bridges and culverts to handle flood flows; selection of materials to handle extreme temperatures (projected number of days exceeding critical temperature thresholds; changes in maximum probable precipitation)|
|Construction||Probability of weather-related delays, e.g., staffing (precipitation; temperature; snowfall; humidity; wind)||Timing of materials delivery; contract incentives and penalties (probability of disruptive events; consecutive days of hot/cold or wet weather)||Annual number of days with suitable work conditions (changes in temperature and precipitation patterns; length of frost-free season)|
|Business||Retail||Supply chain decisions, e.g., promoting products in response to weather events (probability of heavy rainfall; extreme temperatures; snowfall)||Production and purchase of seasonal items, e.g., umbrellas, outdoor activities, snow sports; possible disruption of supply chains (seasonal snowfall, number of rainy days, extreme temperatures; probability of disruptive events)||Probability of disruptive events|
|Insurance and financial management||Hedging/risk management; shifting funds in anticipation of large payouts from widespread events such as flooding or active period of extreme events such as hurricanes (probability of disruptive events)||Potential demand for energy; potential crop yields; contracts for insurance or reinsurance; setting premiums (above or below normal number of hurricanes; large-scale patterns favoring flooding or drought; extended periods of abnormally hot or cold temperatures)||Insurability of coastal property; changes in regional patterns of risk (sea-level rise; storm surge; storm patterns; frequency and intensity of hurricanes and droughts; changes in maximum daily rainfall events)|
|Public Health (see case study for more detail)||Potential disease outbreaks||Conditions conducive to development of disease vectors (temperature; precipitation; easterly waves; extratropical cyclones)||Seasons that may have above-average number of cases, e.g., meningitis, malaria (sea-surface temperatures; cumulative rainfall; temperature variability; strength of Indian Monsoon)||Changes in regions susceptible to spread of disease, e.g., areas where viruses and bacteria can survive due to warming temperatures (changes in regional temperature and precipitation patterns)|
|Extreme temperatures, heat waves, cold spells||Likelihood of a significant event (maximum and minimum daily temperatures; humidity)||Likely number of events during a season (probability of occurrence of consecutive days with temperatures above or below critical thresholds)||Changes in the frequency of extended periods of abnormally hot or cold weather (daily temperature)|
|Heat waves||Likelihood of a significant event (maximum and minimum daily temperatures; humidity)||Likely number of events during a summer (probability of occurrence of consecutive days above temperature thresholds)||Changes in the frequency of extended periods of abnormally hot weather (daily temperature)|
|Algal blooms/release of neurotoxins in water||High temperatures with relatively stagnant water and abundant sunshine (air and water temperature; precipitation; cloud cover)||Probability of extended periods (weeks) of hot, dry weather (temperature; precipitation; runoff)||Changes in conditions conducive to algal blooms (summer water and air temperatures; changes in cloud cover; changes in frequency of drought and runoff)|
NOTES: Variables needed to make these decisions are shown in parentheses. The examples are based upon presentations to the committee, examples of use solicited from the state climatologists and other climate services providers, and from published research.
Often, the long-term average or climatology of a particular phenomenon—such as assumptions for the seasonal volume of water held in a reservoir—are incorporated into decision-making as a first step. As the decision point draws nearer, adjustments are made as additional information becomes available. In this context, S2S forecasts can inform the process of adjusting decision-making between the timescale of long-term planning and short-term response to events across a wide range of sectors within the economy (Table 3.1).
Finding 3.1: S2S forecasts provide value or have great potential to provide value to society for a broad range of sectors and decision-making contexts.
Despite the wide range of potential sectors in which S2S forecasts are or have the potential be valuable, there are many challenges and barriers to their uptake by decision-makers. For example, many water managers can see the potential application of S2S forecasts to their work, and in some cases such information has provided valuable context for planning (see case water management case study below). However, the outcomes of forecast use have not always been positive; currently available products do not always fit easily into institutional decision-making frameworks, and managers are eager for forecasts of variables and at resolutions that are more directly relevant to their contexts. These points are broadly consistent with published research on applications of S2S predictions to decision-making, which to date focuses on the use of seasonal predictions in the agricultural, energy, or water management sectors (e.g., Breuer et al., 2010; Hansen et al., 2006; Lemos, 2008; Mase and Prokopy, 2014; Pagano et al., 2002). For example, Patt et al. (2007) document how use of a seasonal forecast in Ethiopia enabled an emergency management team to identify specific relief actions with months of lead time, alleviating food shortages in 2002. In contrast, seasonal forecasts prompted the restriction of credit for seed in Zimbabwe in 1997, which prevented planting and led to food shortages even though seasonal rainfall ended up at near-normal levels.
Beyond the experience of negative consequences of seasonal forecast use, one important set of documented barriers to the use of S2S forecast products relates to mismatches between currently available products and the stated needs of end users. Forecast products currently available from organizations such as the National Oceanic and Atmospheric Administration’s (NOAA’s) Climate Prediction Center (CPC) or the seasonal multi-modal ensemble (MME) forecasts from the Asia-Pacific Climate Center (APCC) (see Chapter 2), for example, are issued in the form of low-resolution depictions of the probabilities of departure from mean temperature and precipitation over a 3-month period (Figure 3.2), or as forecasts of climate indices such as the El Niño-Southern Oscillation (ENSO).
These forecasts, and text discussions that accompany them, provide general guidance on future temperature and precipitation, but they do not readily translate into operational decision support for many applications. In agriculture, S2S information could be used to assist in determining planting dates, irrigation needs, crop types, fertilization, expected market conditions, pests and disease, livestock management, and the need for insurance (Breuer et al., 2010; Mase and Prokopy, 2014). However, these decisions are dependent on the timing, magnitude, frequency, and duration of weather events within the 3-month forecast window, not departures from seasonal average conditions (Srinivasan et al., 2011; Vitart et al., 2012).
Mase and Prokopy (2014) studied seasonal forecast use in the agricultural sector and identified four barriers to the uptake of current seasonal forecasts, including mismatches between desired and available products:
- Decision contexts rarely involve direct use of information about temperature and precipitation anomalies or climate indices. Instead, tailored, sometimes derived forecast variables, have proven to be much more useful. It is not immediately clear what actions a farmer can make with the information that there is a 40 percent probability of above-normal precipitation (compared to a climatological probability of 33 percent). Variables such as the date of first or last frost might better inform the agricultural sector with specific decisions, such as determining when resources are needed to inform decisions related to harvesting or planting.
- Forecasts of average conditions or anomalies from average conditions are not always immediately useful for many types of decisions. Decision-makers often must respond to conditions that are out of the normal realm of climate variability, such as extremes of heat waves, drought, or floods, or the impacts of a volcanic eruption. A forecast of “above normal” conditions does not necessarily describe whether critical thresholds such as reservoir capacity may be ex-
ceeded or plant die-off may occur. For these types of decisions, the probability of exceeding critical thresholds is more important than the departure from the mean (see Pulwarty and Redmond, 1997; Vitart et al., 2012).
- Spatial and temporal forecast scales are often mismatched to decision-making. Decisions are rarely made for a 3-month period across a large spatial domain; rather, decisions are based on discrete events occurring within a specific time frame and often for a very specific location. A product providing only above- or below-normal or normal conditions over a very large domain may not be immediately useful to a water manager trying to regulate water usage or flow for several watersheds contained within that large domain or spanning several different forecast domains (see also water management case study; Robertson et al., 2014; Srinivasan et al., 2011). For example, seasonal forecast skill is enhanced by a strong ENSO signal, but by the time forecasters have confidence of likely impacts from the event, many decisions on crops, water storage, or other resources have already been made. Furthermore, different combinations of forecast lead times and averaging period may be more or less useful in different contexts (see also Chapter 5).
- There is often a lack of understanding of and trust in forecasts. Many users do not understand the process by which forecasters reach their conclusions, and existing forecast verification metrics are often not directly relevant to users’ contexts (Morss et al., 2008). Consequently, users often have little confidence in the forecast (Mase and Prokopy, 2014). Familiarizing users with the processes by which forecasts are produced often requires direct interaction over a sustained time period. Such interactions are expensive, resulting in fewer communication pathways between the producers and users. Thus a lack of resources issue often exacerbates this issue and hampers delivery of information to end users (see also Hansen et al., 2011; Klopper et al., 2006; Lemos and Morehouse, 2005).
Such examples extend beyond agriculture. For example, forecasting events such as heat waves for the public health sector, or environmentally caused anomalous electromagnetic propagation and mirages for ocean communications applications, have the potential to provide much greater value to decision-makers than information about departures from average temperature and precipitation across a wide region (see also case studies below). Contextual factors, such as lack of trust or inflexible personal or institutional operations, also impede forecast use in water management and public health sectors. In water management in the United States, for example, some institutions may even have policies that designate certain sources as official information, posing a barrier to the use of new products (Lemos, 2008; Pagano et al., 2002; J. Jones,
personal communication, January 2015). Differences in perceptions of risk and bias and fear of bearing personal responsibility for making decisions based on probabilistic forecasts that are still less familiar to the public are also barriers to use in some cases (e.g., Suarez and Tall, 2010). Lack of resources can also influence the ability to access decision support. Although some private companies and research laboratories produce higher temporal or spatial resolution and tailored forecasts for their clients (e.g., International Research Institute for Climate and Society [IRI] tailored products, and see case study on national security and defense), these are not necessarily widely available to the public. Beyond specific barriers to use, many decision-makers (and even entire sectors) may not be fully aware of S2S forecasting efforts and the potential to apply such information to decision-making (Buontempo et al., 2014; see case study on public health below).
To summarize, the current use of S2S forecasts at present is primarily limited to general guidance, although there are emerging sectors and businesses that make more in-depth use of forecasts. Reasons for the slow adoption of and demand for S2S forecasts into operational environments include (1) a poor fit between aspects of the forecast (skill, scale, and lead time—e.g., salience and credibility [Hansen et al., 2011; Klopper et al., 2006]); (2) contextual factors such as lack of trust, inflexible operations, market fluctuations, and lack of resources; and (3) lack of awareness.
That said, the number of studies that address the role of S2S forecasts in decision-making across important sectors, including transportation, infrastructure, or health and humanitarian crises, is still limited (though studies within humanitarian and health contexts are growing—see Braman et al., 2013, and Coughlan de Perez and Mason, 2014). The relative paucity of analysis about the use (or lack of use) of S2S forecasts—particularly subseasonal forecasts—across multiple sectors and regions inhibits the understanding of the potential value of S2S predictions and the development of strategies to maximize the benefits of S2S predictions to society.
Finding 3.2: Research about the demand for and utilization of S2S predictions across multiple sectors is still limited. Studies that have been conducted often indicate significant barriers to using S2S forecasts in decision-making, including mismatches between available and desired forecast products, barriers associated with policy and practice, and lack of understanding of what could be provided.
Finding 3.3: Decision-makers generally express a need for a wider range of skillful model and forecast variables—particularly information about the likelihood of disruptive or extreme events—that are valid at finer spatial and temporal scales to inform management practices.
Some of the issues highlighted above are not new or unique to S2S forecasts. The weather forecast community, for example, faces many similar challenges, and a developing body of social and behavioral sciences research on forecast use is beginning to increase understanding about how to overcome challenges associated with increasing forecast use in decision-making (Brunet et al., 2010; NRC, 2010c). Learning from experiences on both the shorter-term weather forecasts, as well as leveraging existing knowledge about the use of seasonal forecasts, can provide guidance to maximize the use of S2S forecasts across many more sectors of society. As an example, findings about the barriers to use of seasonal forecasts above are broadly consistent with previous research on the use of weather forecasts, which identified similar types of information that users generally consider to be the most relevant to decision-making (Pielke and Carbone, 2002):
- Extreme events, including droughts, hurricanes, floods, blizzards, tornadoes, and thunderstorms (including hail);
- The benefits of good weather, meaning favorable conditions for a particular activity;
- Routinely disruptive weather, defined as not extreme, but significant enough to warrant behavioral adjustments; and
- Forecast impacts, particularly associated with misses and false alarms (including over-warning).
Finding 3.4: Building on experience related to increasing the usability and use of weather and seasonal forecasts will be important for rapid broadening of the role of S2S forecasts in decision-making.
Uncertainty and Lead Times
In addition to user-relevant forecast output variables and scales, key attributes of forecasts must exist before a prediction can make a value-added contribution to decision processes. Forecasts must be available in a timely manner, provided in a readily understandable format with known accuracy, and accepted as an available tool by the users and policy makers (Hartmann et al., 2002; Pagano et al., 2002).
Different users have different tolerances for how accurate a forecast must be before it becomes useful to them. Some users need high confidence before they can take action, while others may be more tolerant of incorrect forecasts (Lemos and Rood, 2010; Mase and Prokopy, 2014). According to a cost-loss model, if the probability of occurrence of an event exceeds the ratio of the cost of mitigation action to the losses that would be expected to occur without mitigation, then the mitigation actions are
considered worthwhile. Consequently, if the costs of action are high, then the user would require a higher level of certainty that the forecasted event will occur (Murphy, 1977). Such specific probability thresholds are usually unknown to forecasters, however, because they are situationally and geographically dependent.
Decision-makers always operate under uncertainty; even a 24-hour forecast has uncertainties. Such uncertainty is now frequently and more appropriately presented in terms of probabilistic forecasts. However, as has been shown by notable errors in weather forecasts, such as predictions of winter precipitation, when decisions have significant associated costs, people are more critical of forecast errors (Joslyn and Savelli, 2010; Roulston and Smith, 2004; Savelli and Joslyn, 2012). Furthermore, seasonal forecast information, when presented at finer spatial and temporal scales, increasingly has larger bounds of uncertainty. Thus, there may be a higher likelihood of an outcome that is different from the forecast at the particular location where the decision is made. When presented on a value or cost/loss basis, this may make decision-makers reluctant to invest in actions based upon generally lower-skill, highly probabilistic S2S forecasts. For example, a 53 percent probability of greater than normal snowfall may not justify contracting for additional snow removal availability. However, in order to make valid value decisions and actions, decision-makers need established reliability measures. If preventive action requires steep costs, and reliability measures are not well developed, then policymakers are less likely to adopt (especially experimental or new) forecast products and are more likely to resort to a wait-and-see position (Lemos and Rood, 2010). When confronted by crisis, however, such as reservoirs operating above or below their capacities, the willingness to use S2S forecast information may increase significantly (Lemos, 2008).
Thus creating more skillful forecasts does not necessarily guarantee that the forecasts will be usable or used. To be useful, a user must also have confidence in the prediction. Confidence is typically established by evaluating the success of the forecast against a large number of previous occurrences. However, S2S predictions have a number of major challenges in establishing such confidence. For example, because forecasts are typically averaged over weekly or longer time intervals, there are fewer data points against which to verify the forecasts. Furthermore, if users are interested in a prediction of a weather-driven event such as severe flooding, then multiple forecast variables are involved, including precipitation amount and intensity, soil moisture, snow pack, and temperature. Errors in any one of these variables will lead to errors in the projected outcome. The issue of developing confidence in forecasts is of central importance to users, but also links centrally to S2S forecast systems, and is therefore covered in greater detail in Chapter 5 (see the section on Calibration, Combination, Verification, and Optimization).
Finding 3.5: Assessing tolerance for uncertainty and developing user-oriented verification metrics are important to building confidence in the use of forecasts among decision-makers. At the S2S timescales this aspect has been generally underdeveloped.
Given the current barriers to use highlighted in the section above, this section highlights potential avenues for increasing forecast uptake into decision-making. Decision-makers have a range of capabilities, from those able to apply statistical techniques to extract useful information from forecasts, to those with less ability to modify or interpret probabilistic forecasts. Tailored and interpreted forecast information has the potential to increase the value of S2S forecasts by expanding the range of forecast variables and outputs that are available to specific users. Some tailored product variables may be obtained through the development of statistical models or correlation fields for relating existing forecast variables and spatial scales, such as regional temperature and precipitation averages, to other, more useful variables (Mase and Prokopy, 2014). For example, extension agents in the agricultural sector prefer derivative information over simple climate predictions (i.e., forecasting impacts instead of seasonal departures of temperature and precipitation). Here, forecasts of the probability of receiving sufficient precipitation during crop maturation or sufficient soil moisture for seed germination can help the agricultural sector anticipate the amount and timing of irrigation needs.
Creating some tailored products is possible using currently available forecast output. Developing other tailored forecast variables will require advances in the S2S forecasting system itself, particularly through an expansion of the capabilities of coupled Earth system models that enable, for example, new forecast variables such as the occurrence of unusual surf or extreme waves, mean cloud cover, and likelihood of harmful algal blooms. Similarly, some progress on predicting the less probable but high-impact events, so-called extreme events, can be made through tailoring existing forecasts. Such events may include a very hot week in an otherwise near-normal summer. Because these are part of a continuous distribution, there is potential to predict the probability of such events occurring within a forecast period. However, improvements in coupled model forecast systems are likely also needed to meet user demands for predictions of such extreme or disruptive events, especially since the noteworthy nature of extremes is typically sector specific. For example, uncharacteristically low winds might not represent a problem for water, transportation, or agriculture sectors, but would be significant for the wind energy and air quality sectors.
Qualitative interpretation of forecasts can also increase their uptake and value. For example, simplified forecast discussions that accompany National Hurricane Center public advisories provide the explanations behind the forecast that increases many users’ trust of such forecasts. Thus increasing the use of such simplified forecast discussions in other routinely issued forecasts could yield almost immediate benefits. However, careful thought needs to be given to how such information is integrated into decision-making processes (NRC, 2009).
The multitude of potential applications, driven by a multitude of different decision-makers with different needs requiring different formats, increases the complexity of production and dissemination of forecast information. To some extent, private-sector providers may develop products that meet this need, but on a large, advisory scale it is likely that the producers of information will need to consider multiple formats, along with broader scale efforts to develop tailored, sector-specific products.
Finding 3.6: Many forecast products that have the potential to provide greater benefit to society could be developed from existing modeling technology. Developing other important forecast variables and uses will require advances in modeling technology. These variables are likely to be sector or decision-specific and their provision is likely to involve derivative products and/or other decision support.
The Need for Social and Behavioral Sciences
As highlighted in the paragraphs above, developing a system that supports use of S2S information requires more than increased understanding of sources of predictability and improved prediction skill. Advances in use and value require consideration of the decision-making context, which often requires complementary research in the social and behavioral sciences. Specifically, social science research can help to address many of the barriers to use previously highlighted in this chapter, including increasing understanding of users’ confidence in the accuracy of forecasts, users’ decision-making contexts and how to best integrate forecast information, and decision-making in contexts of high uncertainty and limited skill (see also NRC, 2010c). This includes research into how users react to false alarms and the costs associated with incorrect forecasts, and how probabilistic information can be better communicated to fit into users’ operations (R. Morss, G. Eosco, S. Jasko, J. Demuth, personal communication, March 2015). Social science research on perceptions of quality—for example, at which point users will make an investment of time and resources to integrate forecasts in their operational contexts and what types of products are needed to mesh into existing decision-making infrastructures—will also be important to understanding the potential value
of S2S forecasts. As much of S2S information is probabilistic, research will be needed on the interface between probabilistic forecasts and decision-maker applications to determine new ways of translating forecasts to mesh with common usage of other Earth system information. This may involve setting more nuanced decision limits, particularly around low probability predictions (e.g., 51-55 percent). Additional research is also needed on the role that social networks may play in the dissemination of information and practice (e.g., Mase and Prokopy, 2014).
Finding 3.7: Understanding decision contexts for a wide array of users in both sectoral applications and technical capacities is essential for increasing use of S2S forecasts. Such understanding cannot be advanced without social and behavioral sciences research.
Integrating Users into the Process of Developing Forecast Products
Perhaps even more critical than improving forecast products and access is building trust in the S2S forecast process. Scientists and operational forecasters who create the information are often disconnected from how that information is being applied, at least outside of agency operations (Lemos et al., 2012). Broader use of S2S forecasts will be encouraged by creating systems of integrated actors and organizations that initiate, modify, import, and diffuse science and technology, identifying information pathways, relating new information to prior experience of the users, and creating toolkits to enable application of information to various decision contexts (Lemos et al., 2012; Vitart et al., 2012). This requires integration of users in decisions relating to the research and development process, from defining relevant research questions to the process of production and dissemination of products (Lemos and Morehouse, 2005). Developing a mechanism for integration may be informed by existing mechanisms used for integration of numerical weather prediction (NWP) forecasts with a variety of users, but will need to be adjusted for different circumstances related to the production of S2S forecasts. Model developers and operations in NWP interact with a variety of forecast centers around the country and with the media in annual meetings and routine correspondence. To the extent practical, expanding this existing stakeholder network to include mechanisms for collaboration with S2S model developers and operations would be beneficial as compared to developing a separate network from scratch.
The discussion of opportunities and limitations involved in producing S2S forecasts highlights potential trade-offs and risk of using the products that enhances the users’ confidence in adopting new products or practices (Lemos et al., 2012). Such discussions are facilitated, in many cases, by boundary organizations, such as NOAA’s
Regional Integrated Sciences and Assessments (RISA) Program and IRI. These organizations conduct interdisciplinary research on decision-making processes with the goal of better coupling the production and use of climate information (Goddard et al., 2014; Lemos and Rood, 2010; Pulwarty et al., 2009). The growing experience of these organizations points to the considerable effort and long-term relationships that are needed to help users understand what types of forecasts are available and the process of producing the forecasts, and to engage them in the design of forecast products and aligned decision-making frameworks to take advantage of currently available forecast technology. For example, IRI has found that seasonal climate forecasts of relatively low skill can still be successfully applied to water management problems in Brazil and Chile, but only through coupling climate forecasts with streamflow projections and working with managers to explicitly link such tailored forecasts to their reservoir management decision-making process (Robertson et al., 2014). Similarly, malaria early warning systems based on S2S forecasts have been successful through extensive efforts to forge relationships between end users and physical and social scientists who manage the technical aspects of designing forecast products (see case studies below). Growing experience among interdisciplinary researchers has resulted in a similar set of conclusions relating to the benefit of “co-producing” forecast products and information together with the end users of such information (Meadow et al., 2015).
Finding 3.8: An ongoing, iterative process between the developers, providers, and potential users facilitated by the relevant social science researchers improves the use and value of S2S predictions.
Ongoing engagement between decision-makers and scientists involved with producing forecasts can facilitate the development of iterative or multi-step decision-making processes, such as the “ready, set, go” framework. Here, warnings, preparation, and action are keyed to increasing probabilities of adverse events (e.g., Coughlan de Perez and Mason, 2014). Preliminary planning may be initiated when an extended S2S prediction indicates the possibility, perhaps at a small probability, of a significant event. This would be followed by preparatory actions (such as pre-positioning emergency supplies) if subseasonal predictions indicated increasing probabilities of the event. Finally, action (such as evacuation) would be initiated based on a deterministic or short-range ensemble prediction with a high level of certainty. This scenario assumes a reliable transition of the predicted probabilities between the seasonal climate system, the subseasonal system, and the short-range deterministic or ensemble systems—all with their own statistical characteristics and skill levels. Blending the probabilities on these diverse timescales (and possibly spatial scales as resolution improves) into a coherent chain of predictions for a user is a difficult post-processing challenge.
As mentioned above, the decision-makers will be concerned with some measure of risk and consequence based on a combination of the evolving probability of the adverse event and the costs of mitigating compared to those of not mitigating. The more quantitative the model of risk and consequence, the more meaningful will be the estimates derived from the evolving probabilities of the adverse event. Not all organizations will necessarily participate in all three stages. Certain organizations may enter the process at different points and levels of certainty.
The set of actions taken and the probability thresholds that act as triggers are usually dependent upon users’ unique circumstances and institutional landscapes. Developing models for applying S2S information in these types of scenarios represents an opportunity for growth in the private sector. For example, applications requiring acquisition of resources, such as power poles in advance of an expected ice storm, may require more lead time than others for which resources are already available, such as frost protection for an orchard. Such decisions are not static; they are revised as new information becomes available, including reduced uncertainty as the forecast lead time shortens. Decision processes on weather timescales could be instructive, such as how public safety officials change their decisions from the timescale of outlooks issued several days before an event, to watches issued hours before an event, and then warnings issued minutes before an event.
Finding 3.9: Successfully aiding users with a multi-step decision model for mitigating the effects of adverse events is a difficult challenge and one not yet considered carefully in the S2S prediction community.
Resources Required to Encourage Use
Developing interactive, transparent processes is a time-consuming and expensive process. It is constrained by limited resources of research, forecasting and many user communities. As forecasting capabilities improve, the demands users place upon providers of forecast information will only increase. Forecasters and researchers need to be careful not to overpromise the capabilities of improved systems. Especially if they are unable to also address the translation process, the demand for services and interpretation of products may exceed the level that can be met, resulting in disenchantment and abandonment of forecasts (Meinke et al., 2006).
As mentioned above, boundary organizations can play an important role in facilitating transparent dialogs and processes that can help overcome many barriers to forecast use. Many existing structures are engaged as boundary organizations at the weather
and climate change scales, including the NOAA RISA Program, IRI, National Weather Service Forecast Offices, the Department of Interior’s Climate Science Centers, the Department of Defense Climate Services, the emergent U.S. Department of Agriculture (USDA) Climate Hubs, and other programs within academia. All of these programs and offices engage with decision-makers and possess expertise in social science methodology coupled with a physical understanding of weather or climate. They often work in interdisciplinary teams and with those intermediaries who ultimately reach the individual decision-makers.
Finding 3.10: Growth in the use of S2S products will place more demands upon operational agencies and boundary organizations to explain reasoning employed in producing forecasts, and in developing a suite of products that meet the needs of a diverse user community.
Water Management in the Western United States
Improved forecasting capability on S2S timescales is an oft-stated goal of water managers, especially in the drought-prone basins of the western United States (e.g., NIDIS Program Implementation Team, 2007; WGA, 2008). Federal, state, and local water managers in California, for example, seek improved forecasts in order to stretch their ability to balance the needs of 38 million residents (representing 12 percent of the U.S. population), the needs of an agriculture sector that farms more than 9 million irrigated acres and leads the nation in production, and demands for hydroelectricity from the state’s 13,765 MW of capacity.1 The state’s recent, severe drought (e.g., Griffin and Anchukaitis, 2014) has only heightened awareness of the challenges associated with meeting these needs (Figure 3.3).
In California, winter precipitation makes up the majority of the annual water budget, and estimates of spring run-off from winter snowpack are currently used to help manage and coordinate the in- and outflow from the state’s vast system of reservoirs, aqueducts, and groundwater storage facilities. A primary tool used for making decisions about reservoir levels, water allocations, and water transfers is the California Department of Water Resource’s forecast of total April-July river runoff. These forecasts are issued beginning in February and are updated monthly through May. Water-year-based (i.e., September-August) indices are also issued based on historical analogs.
In other western states, similar forecasts are issued by the USDA Natural Resources Conservation Service in partnership with NOAA/National Weather Service (NWS). The skill of these forecasts is currently derived entirely from a sparse network of snowpack measurements, which means they are not readily disaggregated by month.
In contrast to forecasts derived from observations of the winter snowpack, water managers have not relied heavily on the current array of operational S2S weather and climate forecast products. When used, the forecasts tend to be assessed qualitatively and used as a “tie-breaker” in higher stakes, scarcity situations (M. Crimmins, personal communication, March 2015). There are a number of barriers to use of currently available S2S forecast products. First, users may not be aware of times when forecast products have higher skill, such as during a strong ENSO event. This potential variability of skill from month to month is critically important in the California context. Second, the spatial resolution of current S2S forecasts is often inadequate for quantitative use. Finally, institutional barriers can sometimes limit the use of experimental information. For example, state and federal water managers are sometimes restricted to using only forecasts that are operationally issued by federal agencies in their decision-making process (J. Jones, personal communication, January 2015).
Water managers are in broad agreement that better S2S forecasting could improve the basis for a number of their decisions. On the subseasonal timescale, efforts to incorporate quantitative precipitation and temperature forecasts from existing numerical weather and climate predictions may help improve the temporal resolu-
tion of river run-off forecasts and allow for better decisions about, for example, flood control. For example, anticipating atmospheric river events with several weeks’ notice would allow managers to assure capacity to contain excess run-off. During dry winters, the likelihood that drought will persist into late winter and spring is information that is also consistently sought, but is not yet reliably available. This might again be achieved through better anticipation of the likelihood of extreme precipitation events associated with atmospheric rivers arriving from the Pacific (e.g., Dettinger, 2013), which have recently been shown to contribute the majority of annual precipitation and snowpack in California (Dettinger et al., 2011; Guan et al., 2010). Such capacity is likely to occur through improving advance knowledge about the state of the Madden-Julian Oscillation (MJO), Pacific/North American teleconnection pattern (PNA), or Arctic Oscillation (AO) (e.g., Guan et al., 2012, 2013).
More accurate seasonal forecasts of winter precipitation, issued in the previous summer and fall, could substantially improve decision-making about allocations to water project contractors, planning for reservoir water and power operations, anticipation of the number of water transfer requests, and planning for emergency flows for endangered species management. Developing more accurate seasonal forecasts may also spur the development and implementation of novel economic instruments such as reliability contracts to spread risk and improve allocation choices during drought (e.g., Hartmann, 2005; O’Donnell and Colby, 2009).
On seasonal to longer lead times, information about the likelihood of drought continuing for multiple years is needed to inform funding decisions for drought response, conservation programs, and the initiation of programs such as water banking. Improved understanding of natural climate variability could improve analog year analysis. A focus on producing forecasts at the scale of California and Colorado River Basin through statistical or dynamical MME could be particularly useful. More generally, there is a desire for tailored forecasts that fit individual water project/agency location and timing needs. Managers would benefit most from focused research driven by policy that is aligned with their specific needs (e.g., Hartmann et al., 2002).
It has long been recognized that variability in weather, climate, oceanic conditions, and vegetation can influence the emergence of epidemic diseases (e.g., Kelly-Hope and Thomson, 2008; Kuhn et al., 2005). Yet the use of climate information to inform decision-making in the public health sector remains relatively limited (Jancloes et al., 2014). Currently, barriers to unlocking the potential of climate-inclusive frameworks for
disease prevention and control relate less to a lack of climate information and more to a need for sustained, multidisciplinary research efforts and institutional collaborations between climate information providers and the public health community (Jancloes et al., 2014; Thomson et al., 2014; Roger Nasci, personal communication, March 2015). However, a few emerging efforts provide a glimpse into how improvements in Earth system forecasting could enable important advancement in the management of disease and other public health risks.
Improved subseasonal (in particular for 2-4 weeks) forecasts of relative humidity have the potential to improve response to meningococcal meningitis epidemics, particularly those in the so-called meningitis belt of central Africa (Pandya et al., 2015; Thomson et al., 2006b). Meningitis, a bacterial infection, has a history of devastating impact in this region, with large outbreaks affecting hundreds of thousands of people. Untreated infections lead to death 50 percent of the time (WHO, 2012). Epidemics in the Sahel region emerge during the dry season, and correlations between low humidity and meningitis cases were first noted more than 30 years ago (Greenwood et al., 1984). Further research revealed strong relationships between meningitis and dusty, dry conditions (Sultan et al., 2005; Thomson et al., 2006b) and abrupt cessations of epidemics with increases in humidity (Molesworth et al., 2003). However, environmental conditions are only one of many factors, including demographic, behavioral, and ecological conditions, which can precipitate infection. Specifically, in many regions, lack of disease surveillance limits the potential to develop accurate disease transmission models of any kind. This can make translating correlations between disease and environmental conditions into actionable information very challenging (Pandya et al., 2015; M. Hayden, personal communication, March 2015).
The MERIT (Meningitis Environmental Research Information Technologies) initiative was launched in 2007 by the World Health Organization (WHO) as a multi-sector partnership between climate and environmental scientists, social scientists, and the public health community to encourage collaboration and the development of innovative solutions for controlling meningitis epidemics (García-Pando et al., 2014; Thomson et al., 2013). Years of subsequent data collection, climate data and forecast output analysis, and collaboration with public health officials in Ghana has led to the development of relative humidity thresholds that can readily be incorporated into existing public health frameworks. Forecasts of relative humidity and storminess up to 2 weeks in advance, coupled with the observed 2-week lagged relationship between humidity and meningitis, have led to a prototype decision-support tool that issues
meningitis predictions at lead times of up to 1 month—enough time to influence positioning of vaccines (Pandya et al., 2015). The end of the dry season is paced by the annual northward migration of the Intertropical Convergence Zone, but rainfall events can modulate the timing of seasonal change on local-to-regional scales (Figure 3.4). Knowledge generated in developing the prototype early warning system is now driving research into the dynamics of west African monsoon onset and retreat that is specific to meningitis-prone regions (e.g., Broman et al., 2014). Rainfall events are usually associated with African easterly waves, equatorial Kelvin Waves and Rossby Waves, extratropical cyclones, and/or the MJO (Mera et al., 2014). Better representation of these phenomena in forecasts may thus increase predictability of the end of the dry season and of meningitis risk.
Malaria is the most widespread parasitic infection in humans, with approximately 500 million cases and more than 1 million deaths yearly (Greenwood et al., 2005). A substantial subset of these cases is linked to malaria epidemics (as opposed to endemic infections), and development of early warning systems to reduce the incidence and impact of these epidemics is a key goal of the world health community (WHO, 2001, 2015). In vulnerable semi-arid and highland areas of Africa, malaria prevention programs target the control of malarial mosquitoes through spraying of pesticides during epidemics, prophylactic drug therapy for target groups during malaria season, and
positioning of resources to ensure timely and effective medical care for infected individuals. Current early warning systems of epidemic malaria rely primarily on disease surveillance, with epidemic alerts issued and repositioning of resources triggered by thresholds in weekly caseloads. There is the possibility for developing more advanced warnings through increased knowledge of climate and oceanic conditions ahead of the malaria season.
Transmission and infection rates of malaria have been linked to rainfall and temperature variation. In Botswana, for example, epidemics usually emerge in the 2 months after the November-February rainy season, and research linking emergence of epidemics to December-February seasonally averaged rainfall and sea surface temperatures has allowed for forecasts of epidemic risk with 1-month lead time (Thomson et al., 2005). Although this represents a potentially large improvement in time available for public health officials to mount a response, early warning systems may be most useful when case numbers can be predicted 2 to 6 months ahead of risk—enough time to allow for tactical positioning of resources (Myers et al., 2000). A growing body of research suggests that integrating monthly to seasonal forecasts of sea surface temperatures, cumulative rainfall, and temperature variability—especially from MMEs shown to have skill in regions of interest—will likely allow for the development of malaria early warnings at longer (4-6 month) lead times (Jones and Morse, 2010, 2012; Lauderdale et al., 2014; MacLeod et al., 2015; Thomson et al., 2006a; Tompkins and Di Giuseppe, 2015). In the short term, skill and geographic reach of such malaria predictions would increase with improved representation of ocean-atmosphere processes associated with the West African and Indian Monsoons, along with the better representation of the Indian Ocean Dipole and ENSO, both of which influence rainfall amounts and temperature variation in critical areas.
Sixteen additional climate-sensitive diseases have been identified as targets for research and other investments to promote the development of climate-inclusive early warning systems (Kuhn et al., 2005). This list contains some of the world’s most devastating diseases, such as cholera, dengue fever, and West Nile virus. In many cases, nontraditional forecast output variables, such as ocean nutrient forecasts in addition to ocean temperature information for cholera (e.g., Jutla et al., 2011), daily temperature fluctuations instead of average temperature for dengue virus and malaria (Lambrechts et al., 2011; Paaijmans et al., 2010), and extreme rainfall and temperature events instead of average conditions for a variety of other public health disasters (Coughlan de Perez and Mason, 2014), are likely to be important. Close collaboration between the
forecasting and applications communities is critical to develop research agendas that will support the identification and development of new forecast products that maximize benefits to the public health sector (Buontempo et al., 2014; Morse et al., 2005).
National Security and Defense
One specific area where S2S forecasting could prove particularly beneficial on a routine basis is global ocean and ice predictions, particularly as the Arctic warms. In addition, disaster preparations in advance of catastrophic tropical cyclones and other severe events where the military may be called to respond could benefit from pre-staging of relief efforts around the world. Food and water security will be important areas where S2S prediction can contribute key information to national security, and having insight into possible famine due to drought or flood conditions will be crucial to economic and stability concerns in the future.
The important decisions made by the defense sector regarding military operations that involve advance warning of environmental conditions on S2S timescales include vessel routing, military exercise planning, war games, tactical planning, disaster relief, and search-and-rescue advance planning (e.g., in the Arctic). In addition, there are serious threats posed by extreme events to military facilities in vulnerable locations. For example, military facilities on the remote Indian Ocean island of Diego Garcia house the Air Force Satellite Control Network, which serves as an essential global positioning system (GPS) command and control hub (Vedda, 2011). Other installations vulnerable to extreme weather and ocean events include Bahrain, Guam, Eglin Air Force Base, Florida, and Norfolk, Virginia.
The Department of Defense (DOD) currently uses standard public climate data sets, forecast model products, and specialized data sets, models, and methods developed by DOD (i.e., Fleet Numerical Meteorology and Oceanography Center [FNMOC], Naval Oceanographic Office [NAVO], Air Force Weather Agency [AFWA]). DOD has, and continues to develop, advanced and tailored products to aid decision-making. These include predictions about performance of equipment and people/organizations given environmental conditions. DOD has many downstream decision-support tools into which the predictions feed. For example:
- Commander Third Fleet ship operation planning in the eastern North Pacific utilizes seasonal forecasts of northeast Pacific winds and waves, by month based on both standard climatologies and statistical predictions derived from multiyear model reanalyses as a basis to revise/update the timing of the operations.
- For tropical cyclone/hurricane predictions, 2- to 4-week timescales and below are essential for avoiding adverse impacts to sea operations before, during, and after cyclone passage. Military exercises, supply chains, and ship movements are vulnerable to tropical cyclones. DOD currently issues monthly tropical cyclone formation probability forecasts based on dynamical-statistical ensemble forecasts (Navy statistical module attached to the NOAA dynamical model output Climate Forecast System version 2 (CFSv2), which are 1- to 2-week CPC-issued forecasts of above- and below-average probability of cyclogenesis and rainfall).
- Piracy activity predictions (Figure 3.5) are based on forecasts of wave height and surface winds in the Indian Ocean, and a statistical module relating operations/behavior to environmental conditions and similar models for disaster relief operations. Versions of this methodology are also used by agencies involved in migrant and drug interdictions.
- Forecasts of beach and amphibious landing conditions for planning are based on high-resolution air, ocean, wave, and surf model predictions from historical reanalyses and statistical/analog techniques.
To summarize, for this type of decision-making, DOD has a well-developed capacity to utilize and ingest the type of tools/specialized forecasts that users often demand. However, areas for improvement abound. In the short term, improvements in observations for setting initial conditions (especially in areas with insufficient observations such as oceans and geographic regions including Africa and the Western Pacific) and in forecast skill covering the global oceans are especially needed (see Chapter 5). On slightly longer time horizons, improvements in S2S predictions will be vital for future tactical and operational planning under climate change. For example, there is great need for better-integrated predictions of sea ice in a changing Arctic. The Navy and Coast Guard have focused attention on the Arctic via their 2014 Roadmap and 2013 Strategy, respectively (U.S. Navy Task Force Climate Change, 2014; USCG, 2013). Indeed, Arctic installations are some of the most vulnerable.
The combination of thawing permafrost, decreasing sea ice, and rising sea level on the Alaskan coast have led to an increase in coastal erosion at several Air Force radar early warning and communication installations. According to installation officials, this erosion has damaged roads, utility infrastructure, seawalls, and runways. . . . As a result, only small planes or helicopters are able to land in this location, as opposed to larger planes that could land on the runway when it is fully functional.
Daily operations at these types of remote radar installations are at risk due to potential loss of runways, and such installations located close to the coastline could be at risk of radar failure if erosion of the coastline continues. Air Force headquarters officials noted that if one or more of these sites is not operational, there is a risk that the Department of Defense early warning system will operate with diminished functionality. (GAO, 2014)
For the Coast Guard, there is an acute need to engage in medium-range (subseasonal) response planning in the event of an accident in Arctic seas (e.g., an oil spill or a cruise ship evacuation). Navy forces are also much more likely to be engaged in the Arctic to assist Coast Guard search and rescue and other civil support operations (U.S. Navy Task Force Climate Change, 2014). Within this context, there is a need for better forecasts now, but the need will be particularly great as climate change begins to initiate ice free passage through the Arctic Ocean and activity related to shipping/tourism and energy extraction begins to increase.
The Deepwater Horizon Oil Spill
On April 20, 2010, an explosion aboard the Deepwater Horizon (DWH) drilling rig killed 11 workers and triggered a massive spill of oil and natural gas into the deep waters of the Gulf of Mexico that lasted for 87 days before it was finally capped off (Graham et al., 2011; Lubchenco et al., 2012). For the purposes of this report, DWH provides an instructive case study regarding the response on S2S timescales to a large, unexpected forcing event.
Oil spill trajectory forecasting systems in 2010 were well-suited for responding to surface spills, even beyond the scale of the 1989 Exxon Valdez oil spill, and worked well for predicting where surface slicks moved for the 72 hours after the event, during which time weather forecasts were sufficiently accurate. However, the vast volume and depth of spilled hydrocarbons associated with the DWH, the extended duration of the spill, and the spatial extent of its impacts presented unique challenges for projecting the consequences of the spill to an alarmed public.
To help meet this challenge, the relevant scientific community mobilized to document the event observationally and understand its consequences (e.g., Lubchenco et al., 2012). This response included model forecasts and predictions, with three additional near-term ocean forecast systems being added to the initial three-member ensemble used for 72-hour surface slick predictions (Lubchenco et al., 2012). For S2S timescales, the National Center for Atmospheric Research (NCAR)/Los Alamos National Laboratory (LANL) (Maltrud et al., 2010) and the NOAA/Geophysical Fluid Dynamics Laboratory (GFDL; Adcroft et al., 2010) adapted existing regionally eddy-permitting (1/10° and 1/8° horizontal resolution, respectively) global Earth system models to explore the long-term transport and dilution of hydrocarbons or the resulting oxygen drawdown from the spill.
Despite the mobilization of the climate and ocean modeling community to address the consequences of the DWH, a frenzy of popular media activity during the spill created a particularly challenging environment for clearly communicating credible scientific guidance regarding what could be expected on S2S timescales. There was at times a particular focus by the official sources on defending the government’s scientific integrity, especially after the first official estimates placed the flow rate at a minimum of 5,000 barrels per day based on observed surface slicks, but subsequent analysis by academics revealed the true rate to be an order of magnitude larger (McNutt et al., 2011). During the event, scientists from a wide range of backgrounds speculated about the implications of the spill, often going directly to the media without first passing through peer review. For example, one widely covered NCAR press release on June 3, 2010, with vivid animations (Figure 3.6A) was based on scientifically correct ocean
model simulations, but was widely misinterpreted in the popular press as suggesting the impending arrival of harmful concentrations of oil along the entire East coast of the United States (McNutt et al., 2011). About the same time, NOAA scientists developed projections of the regional spreading and dilution of dissolved oil and the possibility of significant oxygen drawdown in the deeply submerged plumes of oil that accounted for the estimated spill rates and the biological consumption of oil (Adcroft et al., 2010). These projections were based on a prototype high-resolution climate model and extensive input from NOAA’s oil-spill projection team. An animation from this study (Figure 3.6B) correctly depicted the localization of dissolved oil from the spill to the northern Gulf of Mexico, but as a new government product, it was only released to the public by NOAA after it appeared in a peer-reviewed journal, about 2 weeks after the well was capped.
These two early studies illustrate some of the specific challenges of using innovative modeling tools to provide insights during an emergency. The particular challenges of communicating across scientific disciplines and using scientific expertise to inform the public during a high-profile incident such as DWH has led to calls for the development of a “community of disaster science” (McNutt, 2015), with expertise that can be applied to responses to a wide range of high-profile events.
Sufficiently accurate and well-validated S2S forecasting of the weather and ocean currents could have helped to better target the response to DWH. For example, the likelihood of DWH oil impacting the beaches along the west coast of Florida and in the Florida Keys was unclear using the climate models described above. Specifically, these projections did not account for the position and strength of the Loop Current, which exerts strong control on the direction of ocean waters. As a result, the projections overestimated the geographic extent of impacted shoreline. Had there been a skillful and validated forecast of the Loop Current structure on S2S timescales during the summer of 2010, it may have been deemed unnecessary to deploy as many Local Incident Command Posts and as much oil response equipment to Florida. Some of these resources were never used and, with more skillful S2S forecasts, could have been sent to alternate locations where they would have been of much greater value (D. Payton, personal communication, June 2015).
There is a high cost to incorrect projections of the direction of an oil spill, and in an emergency there is little time to identify the sources of incorrect current directions or other model biases. For use in official government guidance, therefore, it is important that models are observationally validated or their quality otherwise established. For an unprecedented forcing event, observational validation may be impossible, and scientific journals’ peer-review and embargo may be inconsistent with the time constraints
of the situation. Thus alternate approaches may be required for establishing legally required quality assurance.
As result of the challenges in coordinating the broad participation of the scientific community to DWH, several of the leaders in the response to DWH founded the Scientific Partnerships Enabling Rapid Response (SPERR) to promote rapid communication and coordination of efforts and sharing of expertise during disasters. This 1-year pilot project ended in 2015, but such efforts provide a forum for developing the expertise and frameworks necessary to build capacity on forecasting the consequences of unanticipated events. This is an important step toward developing a “community for disaster science” in advance of an incident (McNutt, 2015).
Catastrophic/unprecedented (unusual/infrequent) events include nuclear power plant accidents that could distribute radioactive material over wide areas and affect many nations (e.g., Chernobyl and Fukushima) and intentional nuclear detonations that could affect large populations (e.g., Nagasaki and Hiroshima). Accidental radiological release events have the potential to impact the air/sea over S2S timescales. Linked modeling systems encompassing the Earth system components can provide important benefits in forecasting the scope of impacts (Pullen et al., 2013).
Nuclear weapon scenarios in the current geopolitical context include limited nuclear exchanges. The range of possible scenarios transcends the “mutually assured destruction” envisioned during the Cold War where nuclear winter was an assured outcome (NRC, 1985). Recent simulations of a regional nuclear exchange (100 15-kt yield) with a comprehensive Earth system model including atmospheric chemistry, ocean dynamics, and interactive sea ice and land components have revealed significant impacts at S2S timescales (Mills et al., 2014). Under this scenario, black carbon injected into the stratosphere would deliver global ozone losses of 20-50 percent, reduced sunlight, and catastrophic effects on global crop yields.
In the immediate aftermath of such rare events, emergency response–focused simulations would be conducted through the Department of Homeland Security (DHS) Interagency Modeling and Atmospheric Assessment Center (IMAAC) utilizing the DOD Defense Threat Reduction Agency (DTRA) for national consequences. (Separately, the Department of Energy [DOE] National Atmospheric Release Advisory Center [NARAC] and DOD DTRA consequence assessment assets can be mobilized for international events, as in Fukushima.) However, these simulation tools were not designed to provide forecast information on S2S timescales.
The DHS National Exercise Division engages in major emergency exercises every year—gaming out the response to such events as pandemics and earthquakes. It utilizes the National Infrastructure Simulation and Analysis Center (NISAC) at Sandia National Laboratory as an extensive tool set to examine the impacts from catastrophic events. However, these tools do not encapsulate real-time prediction out to S2S timescales. The nation needs to be prepared to anticipate, prepare, and react to such impactful events at the appropriate timescales. A framework that could encompass multiple temporal scales of impacts, while exercising state-of-the-science models, could prepare agencies to collaborate and respond to a catastrophic/unprecedented incident. Incidents that could produce sustained regional-to-global impacts beyond several weeks’ duration include volcano eruptions and forest fires/biomass burning.
Although user needs are not fully articulated at this point and reside mostly in case studies and anecdotal information, the potential value in developing S2S forecasts is clear. However, decision-makers need a wider range of model and forecast variables than what is generally available at present. Producing forecast products that are valid at finer spatial and temporal scales, and generating event- and impact-based information, are some of the most commonly expressed needs. For example, knowing the probability of receiving an abnormally high number of heat events during a summer can help local and state emergency management officials prepare resources to minimize impacts.
Advances in S2S predictions, which will be described in Chapters 4 and 5, will provide opportunities to advance from basic products and provide finer spatial and temporal resolution and additional variables that are not presently available to most end users. Some of these specialized products will continue to be created off line, through techniques such as downscaling of S2S forecasts, coupling S2S forecasts with sector-specific dynamic or decision models, such as in the case of predictions of piracy activity (Figure 3.5), or seasonal hydrology forecasts to inform water management. Such specialized products already provide some of the needed capabilities, but access to such products is limited, because development of derived products can be costly.
Decision-makers may also benefit from an ability to run on-demand simulations to respond to unanticipated events. These could be useful in responding to an event or in generating scenarios to be used in a planning context. The capability to use models similar to those used to make operational predictions can provide decision-makers with the flexibility of generating a range of scenarios. For example, forecasts of a range
of ocean currents and surface winds can help decision-makers anticipate where to deploy resources to contain an oil spill or the possible outcomes of a volcanic eruption.
However, advancing effective use of S2S information requires both “knowledge of ” and “knowledge in” the process (c.f., Lasswell  discussion of the policy process). Knowledge in the process focuses on mechanisms that promote the use of S2S information, including understanding of product requirements as described above, evaluation metrics, and integration into operational systems. Knowledge of the process includes research on decision-making processes and contexts that promote or inhibit the use of information. These can roughly be thought of as a need to understand use and a need to promote use.
An essential first step is improving understanding of what stakeholders view as actionable information. S2S information has transformative potential, but the path toward application of such information is unclear because of a lack of study and synthesis of existing information. More generalized information on what end users want, which is important to help inform advances in S2S forecasts, requires more than ad hoc case studies of particular decision contexts, such as water managers’ use of information in a particular basin. Case studies should be broadened to include more sectors, the interaction between sectors, and more regions of the country, in order to develop a more systematic assessment of user needs across sectors. Such an assessment will be particularly important for developing forecasts of variables such as sea ice or harmful algal blooms that are not as readily available at the present time. Furthermore, quantification of the value of use of seasonal forecasts is needed to establish a baseline against which improvements in use can be measured.
In addition, the S2S community should learn more about decision-makers’ tolerances for use of products with limited skill. Under what conditions are decision-makers willing to accept limited skill and still use the products? Are they more willing to accept limited skill in the context of general guidance as compared to concrete decisions? How do they respond to failed forecasts? A more thorough examination of decision makers’ tolerance for forecasts with lower skill will reveal some sectors or applications that can make effective use of present and near-term products, as well as longer-term opportunities and targets for new products as skill improves.
S2S information may be used to consult, consider, incorporate, or engage in dialog about risks (A. Ray, personal communication, March 2015). Examining the decision processes in each of these applications will enable development of new products, as well as drive improvements to the dissemination of such products by both the public and private sectors. It is likely that targeted products will need to be developed for a range of different sectors and decision-making contexts.
Recommendation A: Develop a body of social science research that leads to more comprehensive and systematic understanding of the use and barriers to use of seasonal and subseasonal Earth system predictions.
- Characterize current and potential users of S2S forecasts and their decision-making contexts, and identify key commonalities and differences in needs (e.g., variables, temporal and spatial scale, lead times, and forecast skill) across multiple sectors.
- Promote social and behavioral sciences research on the use of probabilistic forecast information.
- Create opportunities to share knowledge and practices among researchers working to improve the use of predictions across weather, subseasonal, and seasonal timescales.
Realization of the full value of improvements in S2S predictions will also require engagement of end users throughout the process of developing and disseminating forecast products. Just as the retail sector places consumers at the center of its research and development, decision-makers who are the likely consumers of S2S information should be integrated into the research and development process. Integrating developers, providers, and users in the context of strategic planning for the S2S enterprise assures the growth of S2S applications and helps to push the boundaries of the science of S2S prediction. An iterative engagement with users is required in part because the diversity of applications of S2S forecasts is large, and the science of S2S forecasting is rapidly advancing. Ongoing work will be necessary to continually match and integrate what is technologically feasible with what is most actionable for decision-makers. In particular, it will be important to (1) understand what variables and timescales provide the most value and opportunities; (2) understand how decision-makers might operate within the context of limited skill or high uncertainty predictions; and (3) determine the formats and message content for products in partnership with those using those products. Such iterative engagement will also provide guidance to the operational community on the critical research challenges, such as forecasting extreme events, and the way in which information can be most effectively delivered.
As with weather forecasts, most decision-makers are likely to acquire information via an intermediary (Breuer et al., 2010; Lemos and Rood, 2010; Mase and Prokopy, 2014; Pagano et al., 2002). There are opportunities to utilize existing programs, such as NOAA’s Regional Integrated Sciences and Assessments, which actively engage decision-makers in co-production of knowledge related to needs for climate information and services. Numerous academic programs promote interdisciplinary research related to the use of climate and
scientific information in societal applications. These present existing avenues that should be built upon to examine decision-making, generate decision-support tools, and provide guidance on future S2S research priorities, operational forecast products, and services.
As demand for S2S products grows, there will be new opportunities for research and applications, necessitating changes in the workforce. Blended research between the physical and social sciences will facilitate the transfer of knowledge between forecasts, outlooks and predictions of the physical environment, and their social applications. Growing the number of “extension agents” or other boundary roles and institutions should also be considered to improve the outcomes of S2S forecast use, and to better integrate decision-makers into the process of developing S2S forecasts. Changes in the structure of the workforce are discussed further in Chapter 7.
Although it is important to bolster the capabilities of operational centers to produce useful forecasts, it is also important to not neglect the private sector’s role in delivering new products. S2S forecasts offer an obvious opportunity for private-sector providers to transform forecasts of conventional variables into new, value-added products focused on user needs and preferences. An emerging private sector is already providing detailed analyses needed for specialized applications. Thus private-sector providers should be closely involved in any program for engaging stakeholders and should be informed of the results and conclusions of such efforts. The U.S. Small Business Innovation Research (SBIR) program is one such mechanism for the weather and climate research agencies to engage the private sector in improving operational commercial offerings and to more effectively target specific user groups. Continued growth of both the private sector and the array of products in the public sector are thus required to meet the growing demand for services.
Recommendation B: Establish an ongoing and iterative process in which stakeholders, social and behavioral scientists, and physical scientists codesign S2S forecast products, verification metrics, and decision-making tools.
- Engage users with physical, social, and behavioral scientists to develop requirements for new products as advances are made in modeling technology and forecast skill, including forecasts for additional environmental variables.
- In direct collaboration with users, develop ready-set-go scenarios that incorporate S2S predictions and weather forecasts to enable advance preparation for potential hazards as timelines shorten and uncertainty decreases.
- Support boundary organizations and private sector enterprises that act as interfaces between forecast producers and users.
This page intentionally left blank.