Role of future scenarios in understanding deep uncertainty in long-term air quality management

Abstract

The environment and its interactions with human systems, whether economic, social, or political, are complex. Relevant drivers may disrupt system dynamics in unforeseen ways, making it difficult to predict future conditions. This kind of “deep uncertainty” presents a challenge to organizations faced with making decisions about the future, including those involved in air quality management. Scenario Planning is a structured process that involves the development of narratives describing alternative future states of the world, designed to differ with respect to the most critical and uncertain drivers. The resulting scenarios are then used to understand the consequences of those futures and to prepare for them with robust management strategies. We demonstrate a novel air quality management application of Scenario Planning. Through a series of workshops, important air quality drivers were identified. The most critical and uncertain drivers were found to be “technological development” and “change in societal paradigms.” These drivers were used as a basis to develop four distinct scenario storylines. The energy and emissions implications of each storyline were then modeled using the MARKAL energy system model. NO_x emissions were found to decrease for all scenarios, largely a response to existing air quality regulations, whereas SO₂ emissions ranged from 12% greater to 7% lower than 2015 emissions levels. Future-year emissions differed considerably from one scenario to another, however, with key differentiating factors being transition to cleaner fuels and energy demand reductions.

Implications: Application of scenarios in air quality management provides a structured means of sifting through and understanding the dynamics of the many complex driving forces affecting future air quality. Further, scenarios provide a means to identify opportunities and challenges for future air quality management, as well as a platform for testing the efficacy and robustness of particular management options across wide-ranging conditions.

Introduction

In the United States, air quality managers at the federal and state levels develop emission reduction strategies to meet air quality standards. Emission inventories are projected into the future, both with and without each candidate strategy. Air quality models are then used to evaluate the resulting changes in air quality. At the federal level, illustrative emission reduction strategies are a key component of Regulatory Impact Analyses (RIAs) (e.g., U.S. Environmental Protection Agency [EPA], 2009, 2012a 2012b, 2014). These strategies demonstrate how an air quality standard can be met and are used in estimating the resulting costs and health and environmental benefits. For states, emission reduction strategies often support State Implementation Plans (SIPs), which specify how the state will comply with air quality standards into the future (e.g., California Air Resources Board [CARB], 2007; Texas Commission on Environmental Quality [TCEQ], 2012). In RIA and SIP applications, emission projections typically stretch 5 to 15 years. In SIPs, areas that have been designated as being in extreme nonattainment are given a longer time period over which to achieve compliance, and emission reduction strategies may extend for decades (EPA, 2015).

In addition to RIAs and SIPs, a number of federal and state efforts have begun to explore the possibility for multipollutant planning (New York State Energy Research and Development Authority [NYSERDA], 2012; Rudokas et al., 2015). Multipollutant planning has the benefit of more readily accounting for important synergies. For example, increasing electricity production via wind and solar power would potentially reduce multiple air and climate pollutants simultaneously. As in this example, multipollutant emission strategies may involve transitions in energy technologies and infrastructure, and thus may require a longer modeling time horizon, possibly stretching to 2050 or beyond.

In these various applications, accurately predicting baseline emissions one or more decades into the future is difficult. Complicating these projections are uncertainties in future economic, social, and political systems that influence emissions. These drivers are complex, dynamic, and interrelated. Furthermore, unpredictable “black swan” events can affect baseline emission projections (Taleb, 2007). Recent developments that would be difficult to predict a priori include the Fukushima nuclear incident (Acton and Hibbs, 2012), the Arab Spring (Goodwin, 2011), and the recent drop in worldwide oil prices (Levi, 2014).

Much of the uncertainty regarding future air quality is systematic in nature, which makes it difficult to characterize that uncertainty using probability distributions. As a result, traditional approaches of addressing uncertainty, such as statistical and econometric methods, are challenging to apply. Alternatively, parametric sensitivity analyses can be used. In this approach, inputs to a model are varied incrementally to identify their relative effects on model outputs. Parametric sensitivity analysis, however, does not provide information about interactions among multiple uncertain factors. Nested parametric sensitivity analysis allows interactions to be examined, although such an approach quickly becomes computationally expensive if there are more than a few uncertain inputs. In cases such as these, known or anticipated sources of uncertainty are usually dealt with in a qualitative fashion (Office of Management and Budget [OMB], 2003).

An alternative to parametric sensitivity analysis is Scenario Planning (Schwartz, 1997; Ogilvy and Schwartz 1998): the use of descriptions of future states of the world, represented by realizations of the most critical and uncertain drivers, to understand the consequences of those futures and prepare for them with robust management strategies. Scenarios are “narratives of alternative environments in which today’s decisions may be played out” (Ogilvy and Schwartz, 1998, p. 2). The relevant questions about how air quality can be maintained in the future can be answered in the context of each scenario, and “what if” situations can be explored and strategies tested. By providing a context that enables planners to understand the forces that shape the future (Wack, 1985a, 1985b; Schwartz, 1997; van der Heijden, 2005), scenario analysis helps them prepare for future conditions with robust strategies that perform well under all scenarios and develop a flexible decision-making processes.

The Scenario Planning Method, as a type of scenario development, is most helpful in situations where (a) the system under consideration is highly uncertain and complex, (b) surprises would be highly costly or foregone opportunities would have been highly beneficial, (c) insufficient new opportunities are emerging, (d) planning is not sufficiently strategic with respect to changing conditions, and (e) strong differences of opinion may exist where each opinion has merit (Schoemaker, 1991; Ogilvy and Schwartz, 1998). For air quality management, these conditions apply in the following ways. Regulation can impact social and technological outcomes for subsequent years or even decades. Given the existence of technological “lock-in” of investments (investments made to meet short-term goals may have unintended consequences in the longer term), as well as learning curves and technological improvements over time (that bring new technologies and make them cheaper), there can be costly surprises or missed opportunities. This can happen when investments in innovation and technology turn out to be suboptimal in hindsight, and there are foregone benefits in terms of human and environmental health as a consequence. Air quality managers have limited time, resources, and information with which to strategize for the future and seek out new opportunities for air quality improvements. Finally, the major stakeholders in the development of air quality regulations frequently have widely differing viewpoints about what air quality regulations should require.

In this paper, we apply the Scenario Planning Method to long-term air quality management to demonstrate its use and benefits. Applied in this context, the Scenario Planning Method has the potential to help air quality managers understand key driving forces, explore how individual driving forces might behave and interact with other driving forces, estimate what the outcome would be for future air quality, and uncover robust strategies to be prepared for the future. We describe the results of an initial Scenario Planning exercise for air quality management, and demonstrate how the narratives developed to characterize alternative futures might be translated into quantitative characterizations of the U.S. energy system and emissions of several common air pollutants.

Background

Much of the uncertainty underlying emissions and air quality projections can be classified as deep uncertainty. According to the Institute of Medicine (IOM, 2013), deep uncertainty occurs under conditions such as the following: The underlying environmental, social, and economic processes are not fully understood; data collection and analysis is challenging; methods are not available to characterize the underlying processes; there is disagreement among experts about the nature of such processes; the appropriate models are not available to describe the interactions among the system’s variables; and it is not possible to know the probability distributions to represent uncertainty about key variables and parameters. Furthermore, deep uncertainty can affect decision making, as there may be no previous record to inform an effective response. We find the term deep uncertainty applicable in the context of unquantifiable and unknown future air quality and adopt it for our discussion.

Figure 1 depicts one possible air quality manager’s view of drivers and interactions important to air quality. Deep uncertainty is inherent in this process because it cannot be known with certainty how these drivers and their interactions are related, nor how they will evolve into the future. Important but uncertain driving forces could include whether population growth and migration will follow past trends, how much and what type of economic growth will occur, whether emerging technologies—such as electric vehicles and grid-scale stationary storage—will become commercially viable, whether the land use trend of suburbanization will continue or what will take its place, and the magnitude, timing, and location of climate change impacts.

Figure 1. Illustrative systems view of the underlying drivers of air quality.

Note: ** The thicker bold arrow represents the fact that policies can have system-wide impacts. For example, a fuel subsidy will ripple through an economic system as producers and consumers demand more fuel.

Furthermore, the interactions and feedback loops among components may lead to system-level outcomes that affect air quality in ways that could not have been predicted even when the behavior of individual components is known (Senge, 2006; Sussman et al., 2009). For instance, the interactions among factors such as demand for goods and services, fuel prices, demographics, land use, and their resulting impacts on vehicle miles traveled (VMT) are difficult to quantify and to project into the future. An example is the unexpected shift in VMT trends from growth to a slow-down and decline in the last 5 years (Zmud et al., 2014; Federal Highway Administration [FHA], 2013; Polzin, 2006; Cervero, 1998). Also difficult to predict are climate change, climate change impacts, and their outcome on secondary pollutant formation such as ozone and particulate matter (PM) (Jacob and Winner, 2009; Moore, 2009; Tai et al., 2012). These complexities typically force air quality managers to make simplifying assumptions. However, oversimplification and not considering the range of possible outcomes can yield costly surprises or missed opportunities.

To address the deep uncertainty in long-term future air quality management, we first turn to the literature. This uncertainty is endemic in many real-world decision-making contexts, and, in cases where the uncertainties result in a wide ranging set of possible futures, presents a challenge to public and private organizations (Millet, 1988; Kotter, 2012) seeking to make short- and long-term decisions regarding air quality management. These wide-ranging possibilities can have a great impact on the success or failure of a particular decision. For example, an unlikely combination of events is often the cause of the collapse of engineered systems (e.g., Chiles, 2008; Perrow, 1999).

The Scenario Planning process applied here involves interactions among technical experts and air quality managers to develop narratives around key drivers that affect future air quality. Through narrative development, managers might begin to ask a series of questions relevant to future air quality and the kinds of activities or resources that would be needed in each scenario in order to maintain clean air: for example, “What would emissions be like in each scenario?”; “Under what conditions could innovation lead to more effective control technologies at lower cost?”; and “What scenarios of technology breakthroughs or shifts in demands could impact the ability to reach air quality attainment?” While it is not possible to foresee specific technological innovations and economic trends that might occur, scenarios can tease out likely trends that could emerge in different futures (depending on the time frame of decision making) and their impact on air quality.

Even drivers considered relevant but certain can be overlooked in implicit assumptions held in individuals’ conceptual mental models. This is because the current behavior of those drivers may be taken as given and may not be questioned. Scenarios exercises help expose these assumptions. For example, demographic trends might be considered certain, yet, because they are embedded in a dynamic social, economic, and political system that is in constant flux, it is possible to forget the implications of such trends. A sustained increase in life expectancy could mean more of an active senior population engaging in activities and consumption, which might change societal preferences, emissions trends, and exposures and health impacts, for example.

Scenario-based approaches have been used in a wide variety of applications (Bishop et al., 2007; Volkery and Ribeiro, 2009; Varum and Melo, 2010). In the environmental area, these include investigations of climate-related impacts (Intergovernmental Panel on Climate Change [IPCC], 2007) and climate policy (e.g., Silberglitt et al., 2003; Polasky et al., 2011; Soderholm et al. 2011). Air quality studies that apply scenarios have focused either on specific aspects of management organizational structure and functionality, or on application of alternative policy instruments, or on sensitivities to key input variables (e.g., Longhurst et al., 1996; Cofala et al., 2007; Xing et al., 2011). For example, NYSERDA, with support from NESCAUM, developed a wide range of air quality scenarios for New York based on the implementation of policies (e.g., carbon cap) and programs (e.g., electricity demand reduction or targets for wind capacity build-out), as well as key sensitivities (e.g., fuel or technology prices) (NYSERDA, 2012). The NYSERDA approach used scenarios modeled using the North East MARKAL model (NE-MARKAL) to inform regional or state/local air quality planning efforts, and demonstrated an approach to multipollutant and integrated energy and air planning. We would argue, however, that the range of scenarios could be augmented by applying the Scenario Planning Method to identify uncertain and critical drivers. Furthermore, these drivers would increase in importance with the longer time frames of analysis that may be required if multipollutant goals or climate benefits are to be considered.

In summary, we found that applications of the Scenario Planning Method (Schwartz, 1997; Ogilvy and Schwartz, 1998), which focuses on underlying driving forces and involves the development of narratives, have been limited in the air quality context. Non-air examples of the application of the Scenarios Planning Method include studies involving conservation biology (Peterson et al., 2003), water and land use (Mahmoud et al., 2009), national parks (National Parks Service [NPS], 2013), and energy (Ghanadan and Koomey, 2005).

This work couples two approaches: (1) the Scenario Planning Method and the development of qualitative narratives of the future that go beyond policies and sensitivities to explore the underlying drivers of change; and (2) the quantitative implementation of those narratives in a modeling framework (MARKAL) that captures cross-sectoral dynamics and provides information on multipollutant impacts. These approaches are complementary. By using the MARKAL framework to implement the scenarios, we provide additional information and rigor to the qualitative narratives developed during the Scenario Planning process. By using the Scenario Planning Method, we are able to develop a more challenging and diverse set of scenarios than what we would typically consider in a model such as MARKAL.

While in our discussion we focus on a time horizon that extends out over 30 years, the scenarios exercise may be of interest to planners and managers who are interested in a shorter time horizon (e.g., 10 to 15 years). For example, areas classified with attainment dates one or two decades into the future, or areas under maintenance plans, may apply scenarios for these times frames. For areas developing SIPs under shorter time frameworks, scenarios can help by highlighting the longer term implications of their near-term decisions. For example, investments in vehicle fleets, electric utilities, and supporting infrastructure (fuel charging stations, or centralized vs. decentralized grid) today may have impacts past the manager’s planning horizon. Scenarios may help managers see the consequences of shorter term investments that might help or hinder air quality efforts past the current planning horizon. Finally, consideration of the interaction between different kinds of pollutants whose impacts may be felt at different time horizons can inform current decisions (e.g., the interplay between criteria pollutants and greenhouse gases).

Methodology

The U.S. Environmental Protection Agency (EPA) initiated an effort to explore the utility of Scenario Planning in air quality management. An important milestone in this effort was a workshop that was informed in two ways. First, preworkshop interviews were conducted both with experts in wide-ranging knowledge areas and with air quality managers who could not attend the workshop. Expertise in subject areas that directly or indirectly affect air quality was included to avoid confining the thinking exercise to any specific organizational or disciplinary “culture,” and thus to encourage more “out-of-the-box” thinking. External experts in seemingly unrelated areas (such as social trends, psychology and culture, communications technology, and nanotechnology) brought unique perspectives (Schwartz, 1997; Ogilvy and Schwartz 1998, Butcher 2011). Second, the workshop was attended by EPA participants (across multiple EPA offices spanning mobile, stationary, and other sources, as well as across media beyond air) and external experts, who in conjunction represented wide-ranging knowledge with respect to air pollution source sectors, control technologies, air quality modeling, exposure and health impacts, energy, economics and policy, urban planning, and transportation, and from backgrounds in the private sector, public sector and academia.

During the workshop, participants identified key drivers of future air quality and developed sketches of four scenarios of the future. Following the workshop, a detailed narrative for each scenario was developed. The scenarios were then implemented in an energy system model and their respective impacts on future air pollutant and greenhouse gas emissions were evaluated. In this section, we describe these steps in more detail. The results of the process are presented in the next section.

In carrying out the Scenario Planning Method, we followed a structured process involving five steps described here. Figure 2 summarizes these steps: The second row beneath the arrow highlights the process, the third row highlights benefits to air quality managers, and the fourth row shows example questions relevant to air quality that could be answered in each step.

• Step 1. Defining the scope of the problem via a focal question—In general, the appropriate focal question depends on the organization’s specific needs. The question must have the appropriate level of breadth to address a wide range of fundamental drivers, but not be so general that the scenarios are not specific to the organization’s decision making.

• Step 2. Identifying relevant drivers—During brainstorming of the relevant drivers, participants share their conceptual models based on expertise and observation of previous trends, challenging their thinking on the importance and certainty of driving forces. A critical aspect of Step 2 is to distinguish certain drivers from less certain ones, and critical drivers from less critical ones.

• Step 3. Laying out the scenario framework—After drivers are identified, and relationships among drivers are sorted out, drivers are ranked according to how critical and uncertain their impact is. The two most impactful drivers form the “backbone” of a scenario matrix, each representing an axis. The resulting four scenarios thus represent combinations of diverging values of each driver.

• Step 4: Developing scenario narratives—These four scenarios serve as a point of departure for creating scenario narratives. Internally consistent, plausible stories describe how all relevant drivers can move to get from the present moment to each of the futures in the quadrants, regardless of how likely we think these storylines to be.

• Step 5: Rehearsing the future—Many Scenario Planning exercises may end here, allowing the future to be addressed in a qualitative, big-picture fashion. For others, including our air quality management application, the scenarios can be used in quantitative modeling exercises. The modeling serves to illustrate the scenarios and their relative air quality impacts, which in turn helps us understand the relevance of the scenarios for decision making. Modeling may also help verify that a scenario’s assumptions produce a coherent storyline. If not, the assumptions can be refined iteratively. The scenarios themselves can serve as starting points for sensitivity analysis and can be examined to gain further insight into the different drivers. Results may even point to unintended consequences that arise from technology or fuel shifts within or across economic sectors. Also, an important use of the scenarios and a subject of future research is the testing of different air quality management strategies for robustness: finding strategies that perform well in most or all scenarios.

  Figure 2. Highlights of the scenarios process for air quality management.

  

Results

Participants of the Scenario Planning workshop agreed on the focal question: “What is the future of air quality in the United States?” Following presentations by internal and external experts, and building upon the collective knowledge of the participants and from the interviewees, the key drivers for future air quality were identified to be (a) energy supply and demand, (b) land use development patterns, (c) economy, (d) policies, (e) climate change, (f) other environmental indicators, (g) societal attitudes, and (h) technological advancement. See Gamas (2012) for more details on the driving forces.

The group ranked the drivers by how critical and uncertain they are to air quality, and then tested pairs of these drivers to determine which pair would lead to the most widely diverging futures from an air quality perspective. For this purpose, pairs of drivers were laid out onto two axes where each axis represented diverging values of one driver. The group considered the kind of future represented by each of the four quadrants defined by the intersection of the two axes. Pairs of drivers were compared to each other; if those futures were too similar, or if another driver was found to be more overarching in its impacts, then a new pair of drivers was tested using this new information.

Among these uncertain drivers, the group collectively identified the two most important (critical and uncertain) factors to be “technological advancement” and “societal attitudes and paradigms.” Neither of these were factors that would likely be the primary focus of air quality management plans before the scenarios exercise, yet both had a significant effect on future air quality challenges and the pathways by which those challenges could be addressed. For example, technological change such as energy storage improvements, breakthrough renewable fuels, and wildcards (completely unpredictable and unexpected events, such as an earthquake, storm, scientific discovery or breakthrough) could have a significant impact on air quality by making new technologies cheaper and thus of widespread use quickly, regardless of their impact on the environment. Societal paradigms that reflect a preference for environmental protection via consumer purchases directly signal to producers a strong demand for low-environmental-impact goods and services. This could alter the production footprint and the demand for energy services.

Figure 3 shows the axes based on the two key drivers, and the resulting four quadrants and scenario names for each future. Scenario names were chosen to be descriptive and to provide easy reference to each scenario.

Figure 3. Four scenarios for air quality based on two key drivers.

Brief sketches were derived for each scenario. The following are summaries of the narratives developed from those sketches within a future horizon ranging from the present out to the year 2030. More detailed descriptions can be found in Gamas (2012).

• Conservation. Societal paradigms begin to shift toward a strongly focused effort to protect the environment, but technology does not transform rapidly. This could be the case in a future where investments in advanced technologies had not been possible and economic growth had been slow, but environmental problems had been salient and costly, moving society to prioritize environmental protection. In this future, environmental concerns likely would be addressed primarily through conservation measures and energy efficiency.

• iSustainability. Societal paradigms begin to shift toward a focused effort to protect the environment, and technology advancements are both rapid and transformative. These outcomes could be the case if there were early investments in technological advancements and environmental problems had raised public concern sufficiently to change preferences and behaviors. The result could be a society where technology is transforming in a direction that supports environmental protection.

• Muddling through. Protecting the environment is not a sustained priority for the population. Furthermore, technology development is relatively stagnant. This could happen if concerns other than the environment were prevalent and investments and other conditions had not led to the development of advanced technologies. The result could be a society where no moves are made to improve air quality and air quality improvements do not occur as a co-benefit of other technological developments.

• Go our own way. Society does not consider environmental quality to be a high priority. Technologies advance rapidly, however, driven by other factors. This could be the case if geopolitics were such that energy security and independence were a top priority and technological innovation ensued to meet these goals. The result could be a society where no deliberate moves would be made toward achieving environmental quality, but there could be environmental co-benefits associated with other changes.

Steps 4 and 5 occurred following the workshop, given time constraints during the workshop itself. In Step 4, we fleshed out the scenario sketches into detailed narratives that would be internally consistent, plausible, and that would challenge our current thinking. The objectives of developing the fuller narratives included ensuring that the scenarios would be very different from each other and providing information that could support quantitative modeling of each scenario. While the two drivers that formed the axes—technological advancement and societal attitudes and paradigms—drove the core differences among the scenarios, it was important to also consider the changes and trends in the other key drivers that would be coevolving in a consistent manner with them.

Each of the important driving forces identified during the workshop was “tracked” in each scenario narrative. Storylines were examined to ensure they were internally consistent and plausible (meaning, in this context, possible, albeit not necessarily highly probable). For example, it was important to track the changes in key drivers such as energy (e.g., energy efficiency, renewable energy) and development patterns (travel patterns, residential vs. office and commercial space). In Conservation and iSustainability society would be mindful of its environmental footprint, and thus, energy efficiency, telework, and renewable sources of energy would be preferred. An increase in telework could reduce vehicle miles traveled, reduce the size of office space, but increase the size of residential work space. In Muddling through, the status quo in terms of energy use would be preferred so the motivation for reducing the environmental footprint would be absent. In Go our own way, all sources of energy would be preferred if they were domestically produced, so demands would adjust accordingly.

In Step 5, we used the MARKet ALlocation (MARKAL) energy system model (Loulou et al., 2004) and the 2014 version of the EPA MARKAL nine-region database (EPA, 2013) to evaluate the scenarios and gain further insights into the role of different drivers and their consequences for air quality. The database includes representations of the Clean Air Interstate Rule (CAIR), the Mercury and Air Toxics (MATs) rule, Tier III on-road vehicle emission standards, industrial New Source Performance Standards (NSPSs) and state-level Renewable Portfolio Standards (RPSs). The database and its earlier versions have been used in a variety of applications related to air pollution as well as climate change (Akhtar et al., 2013; Loughlin, 2013; Loughlin et al., 2013; Loughlin et al., 2011).

MARKAL was selected for this exercise because it is uniquely suited to illustrate the scenarios: It includes a detailed representation of the full energy system, which extends from extraction of fuels to their use in meeting end-use energy demands; it tracks resulting energy system air and GHG emissions; and its relatively fast run time (typically less than 1 hour on a desktop computer) supports iterative “what if” modeling.

The scenarios were implemented into the MARKAL framework by modifying several categories of the model’s inputs, including (a) technology-specific hurdle rates that reflect different consumer and business preferences, (b) the future costs of advanced technologies, such as electric vehicles and solar photovoltaics (PV), (c) the availability of specific technologies, consistent with the narratives, and (d) shifts in energy-service demands to reflect the impacts of assumptions regarding telework and mass transit. Through these parameters, we change underlying factors that drive MARKAL’s evolution of the energy system in ways unique to each scenario. We did not dictate how each scenario will evolve; instead, we adjusted the conditions under which it evolved. The resulting scenarios differ in both their baseline projections and how they respond to additional stimuli, such as the addition of a constraint on pollutant emissions.

A subset of modeling results is shown here to illustrate the scenarios and demonstrate how this process can help air quality managers visualize and understand different resulting emission impacts, and the use of the MARKAL model in this context. First, for each scenario, we show results for three major aspects of the energy system through changes in primary energy consumption, electricity generation mix, and light-duty vehicle mix.

Figure 4 shows the primary energy mix (i.e., “mined” fuels and resources, including fossil equivalents for renewable technologies, at the beginning of the fuel supply chain) for each scenario. In Conservation, overall demands are lowest relative to the other scenarios due to energy conservation and efficiency, while in iSustainability society is moving away from fossil fuels but has growth in nuclear and renewables to meet a still growing demand for energy, particularly electricity. While by today’s standards this growth in nuclear might seem difficult given the Nuclear Regulatory Commission (NRC) approval process, license renewals have resumed within the last year (Energy Information Administration [EIA], 2014) and technology advancements in iSustainability could mean that small modular reactor technologies could play a role in this scenario. Increased renewables in iSustainability are driven by cost breakthroughs for solar power. Natural gas is dominant in Muddling through, particularly in the absence of growth in renewables and retirements of nuclear. Both natural gas and coal are used extensively in Go our own way, although end-use efficiency improvements slow the rate of demand increase.

Figure 4. Primary energy mix.

Figure 5 shows how electric generation varies from scenario to scenario. Wind and, to a lesser extent, solar play prominent roles in Conservation; solar power and nuclear expand in iSustainability; natural gas-fired electric power dominates in Muddling through; and coal and natural gas are prominent in Go our own way. All four scenarios show an increase in use of electricity over time, albeit at different rates. Conservation exhibits the lowest demand for electricity as introduction of energy efficiency and conservation measures slows demand growth. Go our own way and Muddling through rely on fossil-fuel-based electricity to a much greater degree. Muddling through includes a relatively high quantity of electricity produced through industrial combined heat and power (CHP) plants.

Figure 5. Electricity generation mix.

Figure 6 illustrates technology penetrations in the light-duty vehicle fleet for each scenario. Conservation and iSustainability, in the upper quadrants, have lower VMT as a result of conservation and increased telework. In terms of vehicle technology mix, Muddling through sees little change away from conventional internal combustion engine (ICE)-powered light-duty vehicles. In contrast, E85-compatible ICE vehicles are phased in over time in Conservation and Go our own way. Technology advancements in iSustainability and Go our own way allow stronger growth in electrified vehicles, such as plug-in hybrid electric vehicles with 20- and 40-mile electric range (Plug-in 20 and Plug-in 40, respectively). iSustainability has the greatest market penetration of electric vehicles, which reach a market share of approximately two-thirds by 2040. For the purposes of air quality management, reliance on ICEs and continued increases in VMT in Muddling through could lead to future emissions increases. In contrast, iSustainability would reduce light-duty emissions by electrifying vehicles, but this would increase electric sector emissions.

Figure 6. Light-duty vehicles mix.

By tracking emissions across the energy system, MARKAL indicates how these various storylines affect system-wide emissions. For example, Figure 7 shows national-level NO_x and SO₂ emission trajectories for each scenario, projected through 2040. In the figure, gray trajectories indicate low technology development, while black ones indicate high technology development. Solid lines represent adoption of new societal paradigms, and dashed lines represent stagnant behaviors.

Figure 7. Illustrative result showing the NO_x and SO₂ emissions trajectories.

Despite the wide-ranging differences in assumptions regarding technology development and behavior, all four scenarios indicate that the general trend of decreasing NO_x emissions will continue, largely driven by air quality regulations. There is a substantial drop in emissions between 2010 and 2015 due to the issuance of light and heavy-duty vehicle rules, as well as the Clean Air Interstate Rule or CAIR (EPA 2001, 2005, 2011 and 2012b). By 2040, NO_x emissions across the scenarios are below 2015 emissions by approximately 27% in Muddling through and 45% in iSustainability. There is more uncertainty related to future SO₂ emissions, which range from approximately 12% greater (Muddling through) to 7% lower (Conservation) than 2015 emissions by 2040. It is important to note that in modeling the scenarios, it is not just the decision variables that change (e.g., investments in technologies and fuels), but also the underlying parameters that represent preferences and behaviors (e.g., technology-specific hurdle rates that emphasize or deemphasize energy efficiency and changes in end-use energy demands in response to fuel prices). Also, given the uncertainties in the system that is being modeled, differences across scenarios have the potential to provide greater insights than the absolute values themselves.

Discussion

The application of the Future Scenarios Method in the context of long-range air quality planning provides one approach to address the deep uncertainty that air quality managers face. We discuss some of the lessons learned, insights and challenges associated with each step in the process, including the development of the scenarios, and their implementation in an energy system model in order to quantify potential difference in air emissions outcomes.

The focal question should be driven by organizational needs, to be broad enough to be useful to stakeholders with multiple perspectives and needs, but not so general that the answer is too vague to be useful for its intended application to specific regions, industries, or pollutants. For example, a question such as “What is the future of the environment?” is too broad and may make the problem intractable in topic and geographic scope. In contrast, the question “What are the future challenges and opportunities for multi-pollutant management in the Southwest U.S.?” opens the door for an investigation of the factors driving emissions and of the management options that are available.

The fact that the focal question is tailored to a specific organization and set of issues makes salient that the scenarios themselves are developed to respond to these specific needs. Using scenarios created to answer another focal question or problem, or for another organization, would defeat the purpose of the exercise. Those would have been created for a different application, at different time scales or to address different issues. Thus, those scenarios might not address the most critical and uncertain drivers for air quality management or for the specific organization.

The critical drivers that were chosen were broad in scope and captured two key dynamics in terms of both societal and technological trends. The group considered additional drivers beyond those already listed. Some drivers were found to be “signals” or symptoms of other underlying drivers or causes. For example, a change in VMT signals a change in driving preferences and land use patterns. Changes in oil prices signal changes in geopolitics or disruptions to supply infrastructure due to unforeseen weather events such as storms.

The narratives and their underlying drivers provided insights regarding how air quality managers would “rehearse the future” by understanding future challenges and opportunities associated with each scenario. There are a qualitative component and a quantitative component to this, both of which can provide insights. Drawing from the qualitative narratives, we might expect emissions to decrease in Conservation and iSustainability and increase or decrease less in Go our own way and Muddling through because conservation measures and technologies in the first two scenarios are reducing emissions deliberately. However, in Go our own way, if society is trying to maximize the use of domestic resources, it might invest heavily in energy efficiency to do so. This might be a somewhat surprising result if society has been assumed to place a low priority on environmental quality. In this scenario, environmental quality is a co-benefit of the reduction in energy demand driven by energy security preferences. Furthermore, the scenarios and the use of MARKAL specifically also help us see the cross-impacts that some technologies can have over others. For example, in iSustainability, widespread electric vehicle adoption must also impact electricity production, emissions, and other technologies that are competing for electricity.

The two main drivers have different ways of improving air quality. Societal change can push technology (possibly reducing emissions even as consumption increases) and lower consumption of fossil fuel energy (reducing emissions by reducing output). Technology change can reduce emissions via the development of emissions controls, but can also improve efficient production processes, which can reduce emissions without additional controls.

Emissions and air quality outcomes for the scenarios are not predefined. One of the benefits of the scenario process is strategically thinking about how the complex combination of drivers could ultimately affect air quality. This is accomplished through the development of scenario narratives based on a structured scenario development process. In addition, the application of the scenarios in the MARKAL model proved insightful in uncovering potential inconsistencies in the qualitative storylines. For example, in Conservation, we initially expected that the model would predict significantly reduced fossil fuel consumption in the electric sector. However, given the lack of interest in nuclear power and the lack of resources to adopt the more advanced renewables and emerging nuclear power technologies, the market share of natural gas for electricity production increased dramatically. This was an outcome that was not anticipated during narrative construction, but that was important in terms of understanding the implications of the scenario on energy use and emissions.

In terms of translating the scenarios into a model like MARKAL, the challenge was to translate big-picture concepts to fairly detailed values and parameters in the model. We went through several iterations in arriving at our final implementation. Earlier iterations attempted to model every detail of the original scenario narratives. The resulting model was highly constrained, however, leaving little flexibility to respond to any additional technology or policy assumptions to be tested across the scenarios. In the current implementation, we sought to focus more on changing technology availability and preferences, but leaving the model with sufficient flexibility to respond to additional stimuli in a way that made sense according to the underlying narrative. For example, we would expect the model to respond very differently from one scenario to another for a stringent carbon reduction target. In Conservation, advanced technologies are not available, so the model would likely transition to displacing coal with additional natural gas and to placing carbon capture and sequestration (CCS) on the remaining coal plants. In contrast, a carbon reduction solution for Go our own way might include technologies with carbon capture, and an increased transition of the light duty fleet to ethanol, natural gas, and other domestic fuels.

In evaluating these results and potential uses of Scenario Planning, it is important to remember that scenarios are not predictions, projections, or sensitivities. Predictions and projections are based on observed past trends of known drivers such as population, economic growth, and consumption patterns, often relying on business-as-usual or “most likely” futures. In sensitivity analyses, high and low values are assigned to certain relevant variables or sets of variables (symptoms), but the analyses may not assess the impact of drivers (causes) on those variables. Even the most sophisticated air quality, statistical, and economic modeling techniques require sets of initial assumptions and parameters, usually based on observed past trends, previous experience, and expert judgement. By assuming that past trends will continue, underlying drivers, potential surprises, and opportunities can go unrecognized under more traditional approaches.

Expert judgment can provide an educated guess about how likely it is for a specific set of future conditions to materialize, but cannot be expected to anticipate the unexpected, especially from driving forces that go beyond the expert’s area of expertise. Probabilistic exercises may be more appropriate when the system under consideration is smaller, is less complex, and/or when more data are available to perform them. Attempts to provide quantitative estimates of overall uncertainty may neglect the very uncertainties that matter most, or provide a misleading sense of confidence about the results, if applied under conditions of deep uncertainty (Craig et al., 2002; Koomey, 2002; Koomey et al., 2003; Scher and Koomey, 2011).

Future Directions

Scenario narrative construction and revision is an ongoing, iterative process that can and should be revisited. Different combinations of drivers may be tested on the axes until a manageable number of scenarios are found that are sufficiently divergent, challenging and plausible. While four axes may render four scenarios, it may also be possible that one of the scenarios is not useful because it is not divergent enough, in which case the exercise may ultimately render only three useful scenarios. It is also possible that in thinking about the narratives for each future, the specific sets of values that we think would result for other drivers may not be internally consistent (which in and of itself is useful information to know). It may also be the case that a fifth scenario is generated to represent a separate state of affairs to understand how events would unfold toward a specific desirable or undesirable future. As events unfold over time, scenarios should be updated through new scenario exercises.

This updating process should leverage both the earlier narratives that were developed and the quantitative modeling of potential alternative storylines, surprises, and unintended consequences that may occur. In future air quality scenario planning exercises, experts in other environmental media could also be included, as this will enhance understanding of potential cross-media impacts and co-benefits (Mackenzie, 2003). Stakeholders outside the organization should also be included to incorporate their views and form a broader understanding of driving forces. The organization’s specific needs and goals should dictate the kind of expertise to include in the exercise.

The scenarios exercise laid out here has drawbacks. The scenarios exercise, while insightful, is a high-level qualitative exercise—thus the use of a model like MARKAL to illustrate the outcomes. Air quality and land use modeling could also help inform the decision maker to further simulate the conditions that the scenarios represent in tangible and quantitative ways. Furthermore, once the scenario has been modeled, it is not easy to track how an individual component might be having an impact because all of the drivers are being simulated at one time. Monte Carlo-type simulations on the scenarios could be a useful next step.

Now that the scenarios have been outlined and modeled, a useful subsequent step is to apply those scenarios in a range of research investigations. For example, we plan to examine regional emissions under each scenario and identify the potential for emerging air quality emissions challenges. In addition, air quality modeling of the emissions trajectories could shed light into further implications of these futures, including air quality management challenges for nonattainment areas, predicting which locations may be future nonattainment areas, and highlighting important synergies for multipollutant management (e.g., simultaneously addressing criteria, greenhouse gases, and toxics), and in examining health, environmental, and sustainability impacts.

Additionally, in this application, we incorporated only environmental rules that have been finalized. An extension to this work potentially could evaluate the efficacy of candidate single- and multipollutant management strategies across the range of scenarios.

To address location-specific impacts, the scenario outputs may need to be downscaled to a finer spatial resolution. Tools such as NE-MARKAL provide state-level resolution. Alternatively, approaches such as the Emission Scenario Projection method translate regional growth factors into county-level, pollutant-specific and source category-specific factors (Loughlin et al., 2011; Ran et al., 2015) that can be used in detailed air quality modeling exercises. In estimating impacts, it may also be advantageous to integrate models and methods that site new power sector and industrial facilities based upon heuristics, consistent with other scenario assumptions (e.g., Hobbs et al., 2010; Kraucunas et al., 2015).

The scenarios may be used to explore the robustness of current and future air quality management strategies, identifying where otherwise unforeseen challenges may arise and where there may be technological synergies between air quality and climate goals. Future research should also include understanding the impact and cost of “lock-in” (short-term investment decisions over a longer term) relative to more holistic and robust decision making that considers longer time horizons. Another topic of interest would be to explore the implications of climate change across the scenarios. One such implication involves the seasonality of emissions. For example, a warmer climate would potentially increasing summer air conditioning demands, both in magnitude and by lengthening the season over which space cooling is necessary. The strategy generating the necessary electricity would differ from one scenario to another, as would the resulting emissions. Furthermore, the pollutant mix would be different across the scenarios, which would have implications on the photochemistry of pollutant formation. In contrast, a warming climate would likely decrease wintertime emissions.

Conclusion

At the time of this study, we found no explicit applications of Scenario Planning for air quality management. Our application of this method highlights how it may support improved understanding by air quality managers of deep uncertainty by helping them sort through the complex and dynamic interactions between air quality and the economic, social, and political systems that affect it. Air quality managers can identify the critical and uncertain driving forces most relevant to air quality and build scenario narratives around those driving forces. For example, our initial efforts and application of the scenarios in the MARKAL framework suggest that regardless of the scenario, current regulations on air quality will likely lead to reductions in emissions of NO_x, but that the magnitude of those NO_x reductions and the future direction of SO₂ emissions may be affected by how technology advances and how social attitudes are realized. Additional work on how to translate scenario narratives into quantitative modeling will continue to improve our understanding of the impacts of these types of deep uncertainty.

Disclaimer

While this work has been cleared for publication by the EPA, the views expressed in this article are those of the authors, and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency.

Acknowledgments

The authors acknowledge Ron Evans and Bryan Hubbell (OAQPS, EPA), who were instrumental in supporting and contributing the scenarios development; scenario-related discussions with workshop participants and EPA staff. Carol Lenox, Ozge Kaplan, and William Yelverton, who support the MARKAL model database development (ORD, EPA); and Jim Butcher and Erik Smith of the Global Business Network.

ORCID

Daniel Loughlin

http://orcid.org/0000-0002-5102-3507

Additional information

Notes on contributors

Julia Gamas

Julia Gamas, Ph.D., is an environmental protection specialist at the U.S. EPA Office of Air Quality Planning and Standards.

Rebecca Dodder

Rebecca Dodder, Ph.D., is a physical scientist with the U.S. EPA Office of Research and Development.

Dan Loughlin

Dan Loughlin, Ph.D., is a physical scientist with the U.S. EPA Office of Research and Development.

Cynthia Gage

Cynthia Gage, Ph.D., is retired from the Office of Research and Development, U.S. EPA.