Innovation in computing, information, and communications technology is at the heart of nearly every large-scale socioeconomic system. Computing underlies and enables systems that affect our lives every day—from financial and health systems to manufacturing, transportation, and energy infrastructures. One important consequence is that advances in computing are critical enablers of change for addressing the growing sustainability challenges facing the United States and the world. A key finding of this report is that information technology (IT)1 will play a vital role in achieving a more sustainable future and that research and innovation in computing, information, and communications technologies are consequently critical to addressing the broad range of sustainability challenges (Box 1.1).
The critical global challenges in sustainability are deep, and solutions will require input from many disciplines. Fortunately, there are numerous opportunities to apply IT innovations in ways that will have a profound influence on sustainability efforts across many areas, including the ecological and environmental sciences, numerous engineering fields, public policy and administration, and many other areas. The National Research Council’s (NRC’s) Committee on Computing Research for Environmental and Societal Sustainability is aware that there is significant effort aimed at making IT itself “greener” and recognizes that these efforts are important.
1The committee uses the familiar acronym “IT” (information technology) to encompass computing, information, and communications technologies broadly.
A Note on the Definition of
“Sustainability” and the Focus of the Committee
An often-cited definition of “sustainability” comes from the Brundtland Commission of the United Nations (UN): “[S]ustainable development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs.”1 The UN expanded this definition at the 2005 world summit to incorporate three pillars of sustainability: its social, environmental, and economic aspects.2 This report takes a similarly broad view of the term. Although much focus in sustainability has been on mitigating climate change, with efforts aimed at managing the carbon dioxide cycle and increasing sustainable energy sources, the committee recognizes that there are numerous additional sustainability challenges that could be assisted by advances in computing and information technology and computing3 research. The committee’s focus is on addressing medium- and long-term challenges in a way that has significant and ideally, measurable, impact.
1United Nations General Assembly (March 20, 1987). Report of the World Commission on Environment and Development: Our Common Future; transmitted to the General Assembly as an Annex to document A/42/427—Development and International Co-operation: Environment; Our Common Future, Chapter 2: Towards Sustainable Development; Paragraph 1. United Nations General Assembly. Available at http://www.un-documents.net/ocf-02.htm.
2United Nations General Assembly, 2005 World Summit Outcome, Resolution A/60/1, adopted by the General Assembly on September 15, 2005.
3The term “computing” is used generally in this report and is meant to encompass information and communications technologies (ICTs). Thus “computing” and “ICTs” are used interchangeably throughout the report.
The greening of IT, through efforts such as reducing data-center energy consumption and electronic waste, should be and is an important goal of the computing community and IT industry.2 However, the focus of this report is on what could be termed “greening through IT,” the use of
2The 2010 OECD report “Greener and Smarter: ICTs, the Environment and Climate Change” (in OECD, OECD Information Technology Outlook 2010, OECD Publishing) notes that impacts from ICT life cycles (including not just use but also production and end of life) need to be considered in order to understand complete impacts. A recent McKinsey Quarterly article, “Clouds, Big Data, and Smart Assets: Ten Tech-Enabled Business Trends to Watch,” by Jacques Bughin, Michael Chui, and James Manyika, offered some cause for optimism regarding green IT: “Electricity produced to power the world’s data centers generates greenhouse gases on the scale of countries such as Argentina or the Netherlands, and these emissions could increase fourfold by 2020. McKinsey research has shown, however, that the use of IT in areas such as smart power grids, efficient buildings, and better logistics planning could eliminate five times the carbon emissions that the IT industry produces.” McKinsey Quarterly 5(3):1-14.
computing and IT across disciplines to promote sustainability in areas and systems in which advances in information and communications technology (ICT) could have significant positive impact.3
The committee believes that some of the most profound fundamentals within the field itself are suggestive of the unique contributions that computer science (CS) and ICTs can make to sustainability. For instance, the very notion of automated “queryable” structured data is at the heart of much of computer science. The scope and scale of the sustainability challenge are coupled with vast amounts of relevant data, which makes deep insights into the challenges of collecting, structuring, and understanding those data essential. Computational thinking is critical to solving almost any large problem. The committee’s focus is on problems that are intellectually challenging, grounded in IT and CS, and important for sustainability—that is, a kind of “Pasteur’s octant.” See Figure 1.1.
Despite the profound technical challenges presented by sustainability and the huge potential role for IT and CS, the committee recognizes that sustainability is not, at its root, a technical problem, nor will merely technical solutions be sufficient. Instead, solutions ultimately will require deep economic, political, and cultural adjustments, as well as major, long-term commitment in each sphere in order to put technical advancements and enablers in operation at scale. Nevertheless, technological advances and enablers can be developed and shaped to support such change, while continuing to support enduring human values in the process. Information technology can serve as a bridge between technical and social solutions
3The community has already begun addressing this challenge. Bill Tomlinson’s book Greening Through IT: Information Technology for Environmental Sustainability (Cambridge, Mass.: MIT Press, 2010) explores how IT can address sustainability challenges at scale. A 2009 article by Carla Gomes, “Computation Sustainability: Computational Methods for a Sustainable Environment, Economy, and Society” in The Bridge 39(4):5-13, provides examples of computational research being applied to domain fields (biodiversity and renewable energy sources). Gomes’s work is an important component of computational sustainability; the present report explores the broader potential for research and innovation in CS and IT to have an impact on sustainability. Additionally, the National Science Foundation’s Directorate for Computer and Information Science and Engineering and the Computing Community Consortium (CCC) jointly sponsored a workshop on the Role of Information Sciences and Engineering in Sustainability. The full report of the workshop, Science, Engineering, and Education of Sustainability: The Role of Information Sciences and Engineering, which discusses research directions for IT as it relates to sustainability, is available at http://cra.org/ccc/docs/RISES_Workshop_Final_Report-5-10-2011.pdf. This report is well aligned, in terms of research areas, with the CCC report. Additionally, the committee concurs with the CCC report Section 4, titled “The Power of Use-Inspired (Collaborative) Fundamental Research.” The present report expands on this theme in Chapter 3, especially in regard to the strength of computer science as a discipline and what it can contribute to sustainability objectives.
FIGURE 1.1 The committee’s focus is on problems at the intersection of significant intellectual merit, relevance to computer science (CS), and importance to sustainability.
by enabling improved communication and transparency for fostering the necessary economic, political, and cultural adjustments.4
Furthermore, sustainability problems are typically heterogeneous in nature—there is almost never just one variable contributing to the challenge or one avenue to a solution. Inputs, solutions, and technologies that can be brought to bear on any given problem vary themselves a great deal. Most sustainability challenges emerge in part due to interconnection—a result of multiple interlocking pieces of a system all having effects (some expected, some not) on other pieces of the system. Solutions to sustainability challenges typically involve finding near-optimal trade-offs among competing goals, typically under high degrees of uncertainty in both the systems and the goals.
In addition to noting the crosscutting nature of many sustainability challenges, it is important to recognize the emergent qualities that typify the sorts of systems being discussed here. Some projections of what might
4E. Ostrom. A general framework for analyzing sustainability of social-ecological systems, Science 325:419-422 (2009).
be accomplished with the savvy application of known technologies or near-term research are straightforward, even in systems and domains as complex as these. However, in such complex systems and domains there are likely to be emergent behaviors and properties as well—both toward and away from desired outcomes. IT practitioners have proven remarkably adept at innovating flexibly when previously unanticipated systems behaviors have demanded responses. The complexity and unpredictability of the results of unsustainable human activities require an innovative and flexible approach to solving or mitigating sustainability problems and their impacts, and IT researchers and practitioners are skilled at innovating and developing flexible solutions in dynamic environments. The committee believes that computing researchers and research approaches will be essential to grappling with current and future systems challenges in sustainability.
This report has three chapters. Chapter 1 elaborates on domains of potential impact in order to illustrate the role and the available opportunities of IT on the broader path toward sustainability. It address the question, In what ways and where can computing research have measurable, significant impact? Chapter 2 describes methods and approaches in discussing the questions, How do fundamental research questions and approaches in computing intersect with sustainability challenges, and how can problem solving and research methodologies in computing research and IT innovation be brought to bear on sustainability? In particular, the committee views one important goal of computer science in sustainability as informing, supporting, facilitating, and sometimes automating decision making—decision making that leads to actions that will have significant impacts on achieving sustainability objectives. Aimed primarily at computer science researchers, Chapter 3 articulates why the interplay between addressing sustainability challenges and computer science research merits attention, and how that interplay offers deep and compelling opportunities for progress in multiple dimensions. Appendix A summarizes presentations and discussions at the Workshop on Innovation in Computing and Information Technology for Sustainability, organized by the committee. Biographies of the committee members are presented in Appendix B.
Forward-looking IT innovations and sustained research can have significant positive impact for sustainability across many areas. For the purposes of this report, the areas are clustered as follows: built infrastructure and systems, ecosystems services and the environment, and
sociotechnical systems.5 Each of these is described briefly below. There are obvious multiple intersection points in these three distinct areas of opportunity. For example, eco-feedback devices (tools that provide instant information on environmental impact) within the home, a sociotechnical system,6 interact with the larger smart grid system, part of the built infrastructure; personal mobile devices, relying on built infrastructure and deployed in a sociotechnical context, provide data that feed into more robust modeling, a crosscutting methodology, and so on. In all of these domains, as potential solutions are deployed, careful attention will need to be paid to iterate over and evaluate solutions to ensure that progress made in one dimension of a given sustainability problem is not later offset by an unanticipated outcome or side effect in another dimension. The next major section, “Illustrative Examples in Information Technology and Sustainability,” provides crosscutting examples of domains in which IT can support and strengthen sustainability efforts.
Built infrastructure and systems include buildings (residential and commercial), transportation systems (personal, public, and commercial), and consumed goods (commodities, utilities, and foodstuffs). The Climate Group’s SMART 2020 report examined the use of information and communication technology in built infrastructure in several key areas, including smart buildings, smart logistics, and smart electric grids. According to that report, these three areas alone provide a potential reduction in greenhouse gas (GHG) emissions of 15 percent of global “business as usual” emissions in 2020.7
Buildings account for up to 40 percent of energy use in industrialized countries and 40 percent of GHG emissions; in the United States they consume more than 70 percent of the electricity produced.8 Smart buildings use IT systems to make better use of energy while maintaining indoor health and comfort. The embedded IT monitors and controls environ-
5Other clusterings are of course possible. The choice of these three was inspired in part by Global e-Sustainability Initiative, SMART 2020: Enabling the Low Carbon Economy in the Information Age (2008). Available at http://www.smart2020.org/publications/.
6“Sociotechnical systems” encompass society, organizations, and individuals, and their behavior as well as the technological infrastructure that they use.
8World Business Council for Sustainable Development, Energy Efficiency in Buildings: Facts and Trends—Full Report (2008). Available at http://www.wbcsd.org/pages/edocument/edocumentdetails.aspx?id=13559&nosearchcontextkey=true. See also http://www.eesi.org/buildings.
mental and electrical systems in the building by means of computerized, intelligent networks of sensors and electronic devices.9 According to the SMART 2020 report, smart buildings could reduce carbon dioxide emissions by an estimated 15 percent in 2020.10 The sustainability of structures generally goes well beyond energy, and involves the reuse and recycling of materials, sustainable construction processes, improved indoor air quality, effective water use, and so on.11
Smart logistics use IT for more effective supply chains (those dealing with journey and load planning and with personal transportation), both in daily operational use and in long-term planning. Examples of IT contributions include better geographic information systems and design software to promote more effective transport networks, collaborative multi-institutional planning tools to lower the logistical demands associated with desired lifestyles, and better inventory-management tools. Computing innovation can also lead to better management of consumed resources. Smart electric grids use IT tools throughout the power networks to enable optimization. (Potential smart grid applications are described in greater detail in the section “Toward a Smarter Electric Grid,” below.)
In addition to reductions that can be achieved in energy consumption, smarter water- and sewage-management systems in the built infrastructure can decrease water consumption and waste. Furthermore, large-scale agriculture necessitates water and supply-chain management; advanced IT can enhance precision agriculture, including the incorporation of technologies to predict crop yields more accurately.12 (See the section “Sustainable Food Systems,” below, for more on food systems broadly.)
Transportation and city and regional planning also provide important opportunities for more sustainable development; computation and IT will be needed to enable significantly more complex planning for the optimizing of investment in new infrastructure. And, changes to manufacturing itself (which incorporates logistics, sensing, transportation, and manipulation) can help with sustainability goals by reducing environmental impacts, conserving energy and resources, and improving safety
9National Research Council, Achieving High-Performance Federal Facilities: Strategies and Approaches for Transformational Change, Washington, D.C.: The National Academies Press (2011).
11For an introduction to some of the issues related to achieving high-performance “green” buildings, see National Research Council, Achieving High-Performance Federal Facilities: Strategies and Approaches for Transformational Change, Washington, D.C.: The National Academies Press (2011).
12National Research Council, Toward Sustainable Agricultural Systems in the 21st Century, Washington, D.C.: The National Academies Press (2010).
for the individuals and communities affected by it. IT has a central role in these efforts.
Assessing, understanding, and positively affecting (or not affecting) the environment and particular ecosystems are crosscutting challenges for many sustainability efforts.13 The scale and scope of such efforts range from local and regional activities examining species habitats, to watershed management, to efforts to increase understanding of the impacts of global climate change. The range of challenges itself poses a problem: how best to assess the relative importance of various sustainability activities with an eye toward significant impact. Nonetheless, in virtually every activity related to meeting sustainability challenges, a critical role is required of data, information, and computation.
Climate science, for example, has been able to take huge leaps forward due to advances in computing research.14 Computational modeling and simulation of Earth, the atmosphere, oceans, and biota and of their many interactions have long been at the heart of understanding how changes in carbon cycles and hydrological cycles give rise to global climate change and the estimating of future impacts. Sensing, data management, and model formation connect these computational analyses to a vast body of empirical observation and to one another. Such tools allow for the continual improvement of fidelity and can help improve the basic understanding of flows of carbon, nitrogen, and other emissions of interest. These tools also improve the understanding of water and resource usage, of species distributions and biodiversity, and of ways in which human activity perturbs these. Analyses of environmental and ecosystem responses to disturbances (those from GHGs, fire, invasive species, disease) are important to meeting a range of sustainability objectives. Modeling also plays a crucial role in guiding decision makers, by connecting ecological science and research to ongoing ecosystem policy and management. For
13A recent National Research Council report “capture[s] some of the current excitement and recent progress in scientific understanding of ecosystems, from the microbial to the global level, while also highlighting how improved understanding can be applied to important policy issues that have broad biodiversity and ecosystem effect.” National Research Council, Twenty-First Century Ecosystems: Managing the Living World Two Centuries after Darwin, Washington, D.C.: The National Academies Press (2011), p. ix.
14D.A. Randall, R.A. Wood, S. Bony, R. Colman, R. Fichefet, J. Fyfe, V. Kattsov, A. Pittman, J. Shukla, J. Srinivasan, R.J. Stouffer, A. Sumi, and K.E. Taylor. The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, S. Solomon D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.), Cambridge, United Kingdom: Cambridge University Press (2007). Available at http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch1s1-5-3.html.
example, models that jointly capture the interrelationships of multiple variables and their joint uncertainty can support improved understanding and more robust decision making.
Large and long-lived impacts on sustainability will require enabling, encouraging, and sustaining desired human behavior—that of individuals, organizations, municipalities, and nation-states—over the long term. Sociotechnical systems designed to aid in behavioral assistance and reinforcement and to provide information about progress are a critical element for global sustainability efforts. Such systems and associated tools are needed at every scale and can be applied to a range of problems, from enabling effective response in times of acute crisis management, to urban planning, to promoting the understanding of behavioral impacts (sometimes referred to as footprint analysis) on carbon, water, and biodiversity.
Institutional behaviors will need to shift in order to realize continuous, long-term environmental changes. Marketing and public education initiatives are important and can contribute to individual and institutional knowledge on best practices. However, real-time information and tools can better equip individuals and organization to make daily, ongoing, and significant changes in response to a constantly evolving set of circumstances. Information dashboards accessible to key decision makers are an example of how IT can be used to collect, analyze, curate, and informatively present critical information quickly to those who need it most. For example, if the financial incentives for energy utilities shift from an emphasis on delivering more power more cheaply to an emphasis on improving the GHG emissions efficiency of a given level of service, new information will be needed. Gathering such information will require greater visibility and understanding of the dynamics of customer demand, grid capacity, and supply availability. In addition, each of the stakeholders will need more effective means of communicating needs and trade-offs. Similarly, in order for urban planning to promote, say, the reduction of liquid fuel consumption for personal transportation, the processes of street design, zoning, planting, business development, water and waste management, and public transportation need to be coordinated across multiple governing bodies and constituencies.
Personal devices, most notably sensor-rich smartphones, not only provide information and services to their users, but also can provide scientists and researchers with information that may have been missed by traditional operational networks. Furthermore, citizen scientists are increasingly engaged in scientific problem solving, for example by docu-
menting species locations, air quality, and other indicators.15 In addition, environmental challenges—those caused by damage to the environment from rising ocean water levels and temperatures or those created by the search for and extraction of materials—can be monitored, assessed, and tracked. Information about environmental challenges can also be disseminated using smarter IT. Further advances in the ability to analyze data collected by a wide array of sources will facilitate a better understanding of how environmental crises begin and how to avoid them in the future.
This section contains three illustrative examples of sustainability-related domains in which IT can have significant impact and in which there is both some current activity as well as prospects for significant progress and impact in the future. This set of examples is not meant to be comprehensive and does not reflect a prioritization. Rather, these examples were chosen to illustrate how IT—both currently understood technologies as well as new ones—could be brought to bear on sustainability challenges and also to show the range and variability of what is meant by sustainability. Each example area listed below cuts across the three broad areas outlined above.
• The smart grid. In this first example, the grid is clearly part of built infrastructure, but it also has the potential to affect regional ecosystems dramatically as new sources of renewable energy are brought online (for example, solar facilities deployed in deserts will affect the desert ecosystem). Managing the smart grid, from both the supply and the consumption side (which may not be as easily separable in any event) will require sociotechnical systems, such as data management, for humans and human organizations.
• Food systems. This second example also encompasses built environments (including the transportation system), the environment, and ecosystems (in various aspects from macro effects on watersheds to strategies for precision agriculture), and, like the smart grid, it requires sophisticated tools and data management to be most effective.
• The development of sustainable and resilient infrastructures. This third example poses crosscutting sustainability challenges, especially when considering a broad view of sustainability that encompasses economic
15W. Willett, P. Aoki, N. Kumar, S. Subramanian, and A. Woodruff, Common sense community: Scaffolding mobile sensing and analysis for novice users, pp. 301-318 in Proceedings of the 8th International Conference on Pervasive Computing (Pervasive ‘10) (May 2010).
and social issues. These challenges include planning and modeling infrastructure and anticipating and responding to increasingly frequent natural and human-made disasters.
Being able to meet the planet’s energy needs in a sustainable fashion is fundamentally interwoven with foundational transformations in the design, deployment, and operation of the world’s electric grids. The problem is large and complicated, and the committee’s framing in this discussion is for descriptive purposes, and is not meant to be complete, to be prescriptive, or to conflict deliberately with other approaches to characterizing the problem.16 With regard to the electric grid, most analyses of potential paths to stabilizing GHG concentrations involve three interrelated advances: deep efficiency gains, electrifying the demand, and decarbonizing the supply.17 As a prime example, the United States currently consumes roughly 100 quadrillion British thermal units (Btu) (about 100 exajoules) of energy per year, with flows from supply to demand as illustrated graphically in Figure 1.2. Roughly half of the energy supply goes into the production of electricity. Of that, the largest share is provided by coal, which has the worst GHG intensity of the supplies and is the cheapest and fastest way to increase supply in developing economies. By contrast, essentially all of the renewable and zero-emission supplies also go into electricity production, but these account for a tiny fraction of the energy mix. Their share must increase substantially in order to
16For instance, a survey paper developed by IBM Research on the computational challenges of the evolving smart grid is oriented around the challenges of data, grid simulation, and economic dispatch: J. Xiong, E. Acar, B. Agrawal, A. Conn, G. Ditlow, P. Feldmann, U. Finkler, B. Gaucher, A. Gupta, F-L. Heng, J. Kalagnanam, A. Koc, D. Kung, D. Phan, A. Sing-hee, and B. Smith, Framework for Large-Scale Modeling and Simulation of Electricity Systems for Planning, Monitoring, and Secure Operations of Next Generation Electricity Grids, Special Report in Response to Request for Information: Computation Needs for the Next-Generation Electric Grid, DOE/LBNL Prime Contract No. DE-AC02-05CH11231, Subcontract No. 6940385 (2011); M. Ilic, Dynamic monitoring and decision systems for enabling sustainable energy services, Proceedings of the IEEE 99:58-79 (2011), notes the fundamental role of a man-made power transmission grid and its IT in enabling sustainable socioecological energy systems. J. Kassakian, R. Schmalensee, K. Afridi, A. Farid, J. Grochow, W. Hogan, H. Jacoby, J. Kirtley, H. Michaels, I. Pérez-Arriaga, D. Perreault, N. Rose, and G. Wilson, The Future of the Electric Grid: An Interdisciplinary MIT Study, available at http://web.mit.edu/mitei/research/studies/the-electric-grid-2011.shtml#report, aims to provide an objective description of the grid today and makes recommendations for policy, research, and data for guiding the evolution of the grid.
17California Council on Science and Technology, California’s Energy Future: A View to 2050, Sacramento (2011). Available at http://www.ccst.us/publications/2011/2011energy.pdf.
FIGURE 1.2 Current U.S. national energy flow. Roughly half of the 100 quads (1015 Btu) is lost, most coal goes to electricity, electricity goes almost equally to residential buildings, commercial buildings, and industrial processes. SOURCE: Lawrence Livermore National Laboratory (2010). Data are based on DOE/EIA-0384 (2009). Available at http://owcharts.llnl.gov/.
reduce the GHG intensity of the delivered electricity. Doing so will dramatically change the nature of the supply, however, since the availability of these resources varies with natural factors, such as wind and sun, rather than being dispatched as needed to meet demand. Furthermore, the geographic placement of these supplies is governed by natural factors, and so the points at which they attach to the grid, and therefore the pattern of flow from supply to demand and hence the power lines, stations, and devices used to convey these flows of electricity, may be quite different from the flows associated with traditional power plants. This has implications for IT, since the information-management problem for distributed energy production is fundamentally different from that for more centralized production. Managing electricity produced by a half million windmills requires advanced IT—data management, algorithms, and analytics—whereas managing a few hundred coal-fired power plants is a much simpler proposition from an IT perspective.
Already a significant fraction of the supply in the U.S. national energy flow is wasted in the generation, transportation, and conversion of this electrical energy, and of that delivered into residential and commercial buildings and industrial processes, much is wasted through inefficient or ineffective usage. Moreover, reducing the GHG emissions associated with transportation and industrial processes, which are currently dominated
by liquid fossil fuels, will involve electrification (e.g., plug-in hybrid or electric vehicles) and hence will further increase demand. Major efficiency gains, the accommodation of variable supplies, and electrification are all likely to involve change in the patterns and practices of the institutions and individuals that represent the demand, which in large part rests on access to actionable information. Innovation in IT and its use underlie all aspects of such a transformation, as described below.
Electric grids can be characterized by their key components: generation, transmission, distribution, and load. Typically, each of these components has been addressed in isolation. Although multiple approaches to transforming electric grids fit within the term “smart grid,” the fundamental change in the future will likely be to treat the key components together, as an interrelated system. Whereas other disciplines will contribute primarily to the advance of the physical components comprising the elements of the energy supply chain, IT is expected to govern how these elements behave and how the complex system as a whole functions—that is, what properties it exhibits. The section below first describes several challenges presented by smart grids and then outlines approaches to addressing these challenges, especially from an IT perspective. Finally, a discussion of the specific role of computer science research and innovation in IT is offered.
Challenges for the Modern Electric Grid
Four main challenges for the modern electric grid are discussed below:
• Increased electricity consumption and corresponding growth of the grid;
• The current model of load-following supply, in which capacity is dispatched on the basis of real-time power demand, with coarse predictive analytics deployed to ensure that enough will be available;
• The difficulty in implementing a supply-following model, in which demand is managed to better match the available supply; and
• Appropriate accounting for currently externalized costs.
Increased Consumption Increased productivity and improved standards of living correlate closely with increased energy consumption. Even in the United States, where the energy-to-gross domestic product (GDP) ratio has been steadily improving through technological improvements and efficiency measures, especially since the oil crisis in 1973, overall energy consumption continues to increase. This is an especially serious problem in recently industrialized nations, such as China. Continuing increases in consumption pose multiple challenges.
In terms of generation, rapid increase in electricity production tends to skew the supply blend toward carbon- and particulate-heavy sources such as coal, because such supplies are currently easier to bring online quickly when needed. This trend further compounds the GHG emission problem. Lower-carbon options, such as nuclear power, present other hazards, and renewable sources cannot typically be dispatched on demand, impose other environmental impacts, or are remote from areas of dense consumption. Wasteful production and manufacturing practices, especially in newly industrialized or rapidly growing economies, further compound the climate impact. IT cannot provide generation, but it can enable more effective use of generation facilities to meet increased demand, facilitate the shift toward more desirable supplies, and help manage the increasing demand.
In addition to providing adequate supply to meet growing demand—which clearly cannot continue indefinitely—it must be possible to deliver the generated energy through the transmission grid and distribution tree reliably and safely. Each power line and each piece of electrical equipment has limited capacity and lifetime.18 Steering actions or switches do not determine the amount of power transferred along each line explicitly, as is done in networks involving transportation, communications, or even water distribution. Instead, the amount of power is determined implicitly, by the underlying physics associated with a distributed collection of loads, connected to a differently distributed collection of generators, through a particular interconnection of wires and transformers. Individual consumers decide independently how much to draw at each load point, and a centralized system operator orchestrates the production at each of the generators in order to match the supply to the demand in real time within the capacity limits of each line and transformer, and within emissions limits set for each generator. This constrained optimization problem is relatively tractable if the transmission and distribution infrastructure is sufficiently overprovisioned. But, as more of its capacity is demanded, the problem becomes substantially more difficult. A network of communicating sensors is overlaid onto the grid to monitor its distributed state, and sophisticated algorithms are used to predict demand, model the flows, schedule generation, and adjust the limited set of control points that are present. Thus, the ability to meet increased demand through the physical
18For example, as more power is transferred along a line, more heat is generated, causing the line to stretch, become thinner, and sag. This increases the resistance of the line, causing it to heat further, and increases parasitic losses due to capacitance to the ground, which increases demand. All of these factors contribute to failures, which eliminate portions of the transmission or distribution infrastructure and thereby place potentially excessive demand on remaining portions.
infrastructure that exists at any particular time is almost entirely through advances in IT.
These challenges are further complicated by the changing nature of the load and the broader introduction of distributed generation. Unlike purely resistive loads, such as heating elements and incandescent bulbs, complex loads effectively cause a portion of the delivered power (called reactive power) to be returned through the grid to the generator. Historically, such “non-unit power factor” loads were predominantly induction motors, which introduce a fairly simple phase shift in the alternating current (AC) waveform. But switching power supplies, such as those used in computers, fluorescent bulbs, battery chargers, and electronics direct current (DC) adapters, introduce complex distortions on the AC waveforms. Residential grid-tie solar installations reverse the flow of electricity within portions of the local distribution tree. And the introduction of electric vehicles potentially introduces high point loads during recharging. Many of these new complex loads already possess communications and computations capabilities, and so they could potentially be a vanguard of using IT to condition demand in order to be “good citizens” of the grid.
Compounding all of these issues still further are the economic structures that impinge on all aspects of generation, delivery, and demand at a range of timescales. On an operational basis, collections of suppliers, consumers, and brokers typically participate in highly volatile wholesale energy markets at various granularities and timescales—a day ahead, an hour ahead, a minute ahead. Meanwhile, consumers typically experience relatively stable retail pricing. Compounding all of these issues further, utilities and the utility supply industry are still largely incentivized to produce and deliver more energy, not less. Economic or other incentives to curb growth are lacking in most parts of the world. A notable exception to this is net-metering—mechanisms that allow electricity consumers to offset their usage of electricity provided by the grid, and thus to lower their cost, by generating their own electricity on-site, typically through rooftop solar photovoltaic installation. Basically, this can be thought of as the meter spinning backward when local generation exceeds local demand. Although net-metering is comparatively common, its penetration is modest enough that it can be incorporated as offsetting demand in the neighborhood distribution tree, without appreciable impact on transmission needs. Broader, less-tangible incentives include the personal satisfaction of obtaining a zero-net lifestyle, potentially opening paths toward the decoupling of quality of life from energy usage. IT has an important role in doing the complex accounting and providing visibility into the consumption and production of otherwise invisible resources.
Current Model of Load-Following Supply Grid operation predominantly involves orchestrating a portfolio of dispatchable supplies, including baseline (nuclear, coal, and hydroelectric power), intermittent (combined-cycle gas turbine), and peaker (simple-cycle natural gas) power plants to supply precisely the real-time power demand across normal variations, spikes, and infrequent peaks in the load. Of course, the demand is not specified explicitly, but implicitly in the use of electricity. Typically, independent service operators perform day-ahead demand prediction for their entire grid, with hour-ahead and even minutes-ahead adjustments, to drive scheduling and market mechanisms while providing adequate generation capacity at all specific points in the transmission grid over time. A certain fraction of online capacity is retained as “spinning reserve” and is used to match short-term changes in demand. An imperfect matching of supply to demand manifests in degraded power quality (such as voltage sags or surges, and frequency variation). Challenges include the following:
• High cost of peak demand. Since generation and transmission capabilities must be built out to meet the peak demand, this peak drives overall investment. However, because there is significant variation in demand, a substantial portion of this investment experiences very low utilization. Fundamentally, load following relies on statistical multiplexing of independent loads; even though the individual loads are very bursty, the aggregate of many such loads is relatively smooth and predictable. However, correlations in the loads, such as air conditioning on hot summer afternoons or refrigerator compressor cycles at breaks in Superbowl action, generate very large aggregate peak demand. Means of power generation with short ramp-up times tend to have low efficiency and high GHG emissions and operating costs.
• Prediction accuracy and market volatility. A mismatch of predicted and actual demand leads to large and rapid fluctuations in wholesale energy prices. Each new broad-based usage change (for example, the increased uptake of plasma television sets, compact fluorescent lamps, electric vehicles, and so on) raises concerns of prediction accuracy. Paradoxically, by eliminating waste, energy-efficiency measures can lead to larger peak-to-average ratios and potentially lower prediction accuracy, making the grid harder to manage.
• Storage limitations. Grid-level storage exists in the form of pumped-storage hydroelectricity compressed air, thermal energy storage, batteries, and a few other possibilities, but storage capacities remain limited. Storage is typically expensive, and turnaround efficiencies (the energy extracted from storage relative to the amount stored) tend to be low. Small-scale battery storage is prevalent but expensive, and the number of
recharge cycles, and hence battery lifetime, is limited. When the effects of manufacturing and the disposal of batteries are taken into account, such storage may have a net negative environmental impact. Midscale storage (say, 1 to 100 kilowatt-hours) is almost non-existent, although flow batteries and electrolysis/fuel cell options remain in development.19
• Non-dispatchable supplies. Most renewable sources of energy, such as wind, solar, and wave, are non-dispatchable. That is, they are available only at certain times and in magnitudes determined by various environmental factors; they cannot be summoned on demand. Gross features, such as the incident solar radiation over the course of the day or the seasonal patterns in wind, are much more predictable than fine features, such as occlusion due to passing clouds or gusts and lulls, and the latter can cause very rapid changes in supply.
Much of the growth of power generation in highly industrialized nations in recent years is in renewable supplies. But the penetration of those sources is fundamentally limited in a load-following regime (i.e., one in which power output is adjusted to demand).
Many smart grid proposals focus on increasing the capacity and sophistication of the transmission system to reduce constraints imposed by transmission in matching supply to demand. These include longdistance lines, in many cases using high-voltage DC, in order to access remote renewable supplies both for increased availability and to obtain geographic decorrelation. Within a grid, especially with distributed renewable resources, there may be sufficient supply to serve the load but inadequate capacity to route the power from points of generation to points of use. Better prediction, monitoring, and scheduling seek to prevent such bottlenecks. “Smart meters,” which are currently being rolled out in many regions, provide 15-minute-interval readings, rather than monthly accounting. Their use enables more accurate prediction and more effective scheduling as well as introducing incentives, such as time-of-use pricing or critical peak pricing, to nudge the demand toward a more grid-friendly form. These efforts introduce a degree of observability into this complex system and thus open the way to decision making and action. As the IT in the grid evolves to embody monitoring, communication, embedded processing, and intelligence at various levels of the grid, it can provide a foundation for an interactive relationship between supply and demand that increases the penetration limit for renewable sources.
19For instance, Sandia National Laboratories just announced the development of a new family of liquid salt electrolytes that could lead to devices that could better incorporate renewable sources of energy on the grid. See, for instance, http://www.sciencecodex.com/sandia_national_laboratories_researchers_find_energy_storage_solutions_in_metils-86320.
Implementing a Supply-Following Model The dynamics and economics of grid operation can be fundamentally altered if the demand can be shaped to match available supply and supply-chain constraints, such as congestion or outages. This approach is referred to as supply following, in contrast to load following. Load following has developed over the past century with a great body of typically centralized, utility-side intelligence to permit consumers to use energy whenever and however they desire; however, supply following typically requires distributed, customer-side intelligence in order to manage energy demands while delivering desired services.
“Demand management” is typically taken to mean an explicit modulation of customer demand (for example, thermostat set-point adjustment, lighting adjustment, production curtailment) by the utility, according to prior arrangement, to flatten the duration curve. Numerous such measures have been employed for peak shaving and shifting load into valleys, but adoption tends to be low. Prevalent utility-driven measures involve automated voluntary adjustment to thermostat set points, especially for cooling during hot summer days. Many industry-driven proposals emphasize smart appliances, including dishwashers, dryers, and ice makers that can defer operation until less costly times of use. Plug-in hybrid or fully electric vehicles are seen as presenting a prime opportunity for the programming of demand. While naïve charging could be potentially destabilizing to the grid or even cause local, aging distribution equipment to fail (for example, if multiple electric vehicles charge on the same block), well-timed charging could provide increased stability while relieving petroleum demand.20
Dynamic variable pricing (as opposed to set schedules) introduces financial incentives for end users to shift demand so that the overall demand is more easily met. Typically, residential demand is shifted into nights and weekends away from industrial demand. Often the schedules are complex and difficult for individual users to keep track of. Charging for power according to more sophisticated pricing schedules requires more sophisticated metering, along with usable and understandable
20There are also many proposals for utilizing electric vehicle batteries as grid storage, providing power back into the grid when demand is high. While certainly attractive in concept, such usage modes present pragmatic challenges. Batteries for vehicles are optimized to be extremely light, dense, and collision-resistant, with high power density. The number of recharge cycles of the battery that could occur before a costly battery replacement is a fundamental constraint. Utilizing this precious resource to improve the management of utility capital investment and potentially having driving range unexpectedly curtailed represents adoption challenges, whereas scheduling overnight charging flattens the duration curve without such impediments. Stationary bulk energy storage need not obtain the very high level of energy density and power density demanded for the mobile case; it can potentially be designed instead for large recharge capacity and high turnaround efficiency.
controls. The modern advanced metering infrastructure rollout aims to achieve some of these possibilities.21
Peak pricing extends dynamic variable pricing to include aspects of a particular consumer’s demand and to take into account how hard it is to meet that demand. In many regions, major industrial customers are on time-of-use, peak-based pricing schedules, in which the product of usage and peak demand determines the cost over a past interval. This creates an incentive for individual users to limit their peak as well as their overall demand.
Finally, an approach known as demand response typically involves load shedding in response to a critical peak-pricing notification event from the utility. Manual demand response has been used in many non-residential markets for many years. Automatic demand response couples an Internet notification event to a preprogrammed set of demand-mitigation responses and thus involves considerably greater IT. Significant information processing, modeling, and control issues need to be addressed to carry out demand-response issues in large commercial buildings. Significant human-computer interaction issues need to be addressed to realize fully the potential of this infrastructure, particularly for residential use. For example, if new electric water heaters are equipped with an automatic demand-response facility but the default setting is to ignore notifications, then without a good interface—and consumer education—the likely result is that the default setting will not be touched and there will be no benefit. However, if the default setting is to respond to a notification by waiting until late at night to turn the heater on, when electricity demand is lower, one can anticipate large numbers of consumers wondering why their hot water systems seem to run out of hot water at unexpected times.
Peak energy reduction is extremely important for reducing the capital investment in generation and transmission assets and in reducing risk in wholesale markets. Also, peak energy has the greatest GHG emissions per unit of electric power because it is generated by less-efficient plants with shorter ramp-up times. However, reducing peak energy has limited impact on reducing the overall energy demand or impact on climate, which are dominated by non-peak demand. Reducing overall demand and reducing the impact on climate require much broader efficiency and reduction measures. In some cases, low-carbon renewable generation—for example, summer solar production—aligns well with peak demand; in other cases, such as relying on the prevalence of nighttime wind, it does not.
21For more information on the deployment of advanced metering technologies, see National Energy Technology Laboratory, Advanced Metering Infrastructure: NETL Modern Grid Strategy (2008).
Accounting for Externalized Costs Another challenge to developing sustainable electric grids has to do with economic incentives and externalities. Various approaches have been articulated, including a carbon tax, fee and dividend models, and so-called cap-and-trade mechanisms (placing a cap on emissions but providing flexibility with mechanisms such as tradable permits). To truly represent the cost of externalities, any of these would require dramatically more precise accounting for environmental costs throughout the energy supply chain. In principle these offer a common metric around which optimization measures at all tiers can be integrated. Today, under a utility-centric approach, crude weightings of the energy blend appear to suffice. If a smarter grid were deployed, the task of accounting for all aspects of life-cycle costs would introduce tremendous IT challenges. The same would be true if, for instance, end users were to access information about the real-time mix in the blend to enable them to make more informed decisions about when to consume energy (assuming the presence of significant penetration of non-carbon supply).
Consideration of computer security should be integral to work on the smart grid. As just one example of the security risks, an attack that injected malicious code into smart electrical energy control systems in millions of homes might be used to manipulate demand and prices, or just to create chaos by turning on or shutting down large numbers of appliances unnecessarily. There are also legitimate potential privacy concerns with such control systems that will need to be addressed in ways that are both usable and technically sound.
Approaches to a More Sustainable Electric Grid
A forward-looking sustainable grid scenario presents a fundamentally more cooperative interaction between demand and supply and fundamentally greater transparency22 throughout the energy supply chain, with the goal of achieving deep reduction in demand and deep penetration of renewables in the supply blend.23
22For this particular goal, transparency is required for technical reasons: that is, to support a more cooperative interaction between demand and supply. However, transparency is also essential for reasons of trust, accountability, and fairness, to avoid the potential for Enron-style market manipulations to be multiplied many times with the new grid technologies.
23Today, major industrial customers with substantial in-house generation are able to take advantage of the volatility of the wholesale market by running in-house generation when prices are high, tailoring its use to meet business goals, and shutting it down completely when prices “go negative”—which they do at times in the U.S. Midwest grid, in the European Union, and elsewhere—thereby getting paid to consume power.
The introduction of storage aims to decouple generation and consumption, enabling either to take place at times that are most effective. (But even with hypothetical vast storage capacities, inefficiencies in storing and generating electricity mean that complete decoupling will not be possible.) Equally important is the ability to coordinate loads, often through the utilization of non-electrical forms of energy storage such as thermal, but also often exploiting flexibility in the actual task. In short, opportunities to improve the sustainability of the electric grid can be clustered as follows: (1) sculpt demand to match the supply of renewable power more closely; when abundant renewable power is available, use it for shiftable power needs (such as ice makers, hot water, dishwashers, electric car charging); (2) sculpt demand to smooth out spikes so that less high-speed dispatchable supply is needed; (3) reduce total demand by improved efficiency in transmission and use; and (4) apply instrumentation and modeling to measure carbon emissions as part of carbon pricing and capping policies.
Electricity is currently the most invisible of utility-supplied resources. To make progress on the opportunities available will require new forms of visibility, including visibility with respect to price and GHG emissions, consumption, and performance. IT can contribute in key ways: for example, pervasive instrumentation, monitoring, and analysis enable visibility into electric power consumption and resulting load performance.
Increased visibility is an important component of developing integrated home or building energy-management systems that can make wise decisions about how to shift energy use. These systems need flexible user interfaces and sensing systems so that they can receive information about when it is appropriate to, say, delay electric car charging, heating up a building, and so on. For example, such systems could be integrated with users’ calendars as well as connected to pricing and other generation-side signals (e.g., sun and wind forecasts). Research is needed on user interfaces, predictive models of user and appliance behavior, and perhaps auction and pricing mechanisms.
Integrated energy-management systems could also address spiking, by heating and cooling buildings more gradually or by making offsetting power-consumption choices. Making the effects of such choices visible to the user (e.g., by clarifying when the building will achieve the desired temperature, when it will be possible to turn on a particular machine, etc.) will be critical for user acceptance. Price mechanisms (e.g., charging more for sudden changes) could be explored as well.
IT research is needed for developing methods for sensing, modeling, and intelligent control of buildings. Most office buildings, for instance, exhibit chronic poor performance (some parts of the building are too hot, others too cold; some parts are poorly ventilated, others too breezy).
Improved modeling at design time and during operations has promise of reducing such problems and saving energy.24 There is a significant role for IT to play in making opportunities for waste reduction manifest and in automating its reduction in buildings and also throughout the grid as a whole.
The Role of Information Technology and Computer Science in Achieving the Smart Electric Grid
Information and data management are essential to making progress toward a smarter, more sustainable electric grid, as discussed above. Computer science research and methodological approaches will be needed at all levels to address the broad systems challenges presented by the smart grid. Initial forays into both research and applications “wins” in this area include energy efficiency and smart mini-grid and distributed energy management,25 energy-efficiency planning and building management,26 and the integration of smart grids and smart electric vehicle planning and operation.27 In many of these areas, savings of one-third to one-half in terms of overall energy consumption, with improved service and significant environmental gains, are possible. In many of these areas, a critical step is that of envisioning how the energy system could function if greater information and real-time data analysis were possible as embedded components of the system. This would require greater attention to integration of sensor technology with energy, transport, and building systems; to sensor data management; and to the role of distributed computing in processing far greater flows of information (and of forecasted performance and outcomes) than is typically the case today.
User interfaces are needed that make it straightforward for people to express preferences regarding aspects such as prices, comfort, timing, and “greenness” of their power mix. These preferences could be very complex and difficult to capture, requiring visualization techniques,
24National Research Council, Achieving High-Performance Federal Facilities: Strategies and Approaches for Transformational Change, Washington, D.C.: The National Academies Press (2011).
25C. Casillas and D.M. Kammen, The energy-poverty-climate nexus, Science 330:1181-1182 (2010).
26G. Crabtree L. Glicksman, D. Goldstein, D. Goldston, D. Greene, D.M. Kammen, M. Levine, M. Lubell, B. Richter, M. Savitz, and D. Sperling, Energy Future: Think Efficiency—How America Can Look Within to Achieve Energy Security and Reduce Global Warming, Report of the American Physical Society on the Potential for Energy Efficiency in a Low-Carbon Society, American Physical Society (2008).
27L. Schewel and D.M. Kammen, Smart Transportation: Synergizing Electrified Vehicles and Mobile Information Systems, Environment: Science and Policy for Sustainability 52(5): 24-35 (2010).
increased understanding of human behavior with regard to energy, pervasive interfaces, and so on. Information technologies may prove useful in encouraging energy consumers to shift their consumption patterns to off-peak hours, when consumption is generally more stable and comes from more sustainable sources of energy generation. In order to identify specific customers (with shiftable consumption patterns) to target with time-of-use rates, utilities could use more sophisticated predictive analysis through statistics. Information dispersal, social networking, and marketing through the Internet are other avenues that utilities are looking into, and with which computer science may be able to help. Additionally, with time-of-use rates, consumption data are increased significantly for each customer, and the complexity of tariffs and calculations with multiple tiers becomes an issue with which better analytical software could help.28 Improved statistical models and database management would be invaluable additions to the capabilities of utilities all over the country.
Improved modeling and analytical tools would help with demand forecasting that takes into account the adaptive nature of the demands (e.g., to answer questions such as: How far will people be willing to time-shift demand this Thursday?). With the help of predictive analysis and weather data, utilities could use estimated capacity to improve their consumption forecasts, which would significantly improve their cost structure.
More generally, economic mechanism design tools for designing pricing within a controlled system will be needed. Sophisticated statistical models could help validate the models through hypothesis testing. A goal might be to bring factories with energy-production capacity into the supply chain to supplement peak-hour supply, assuming that the GHG output was not made worse. This goal would also require creating an economic situation that is agreeable to both the utilities and the owners of cogeneration plants. Calculating the cost and benefit from both the utilities’ perspective and the cogenerators’ perspective, optimizing the best rate scheme to encourage sell-back, and factoring in transmission losses and efficiency present a complicated and interesting optimization problem that could be greatly aided with the use of sophisticated decision analysis tools and statistical models.
Looking further ahead, if appropriate CS research is undertaken, the ability to mitigate the intermittency of renewables through computational approaches will be greatly enhanced. These resources could be made more dispatchable without the need for 1:1 matching of renewables and
28Shwetak Patel, University of Washington, described a technology to monitor in-home electricity consumption at the committee’s Workshop on Innovation in Computing and Information Technology for Sustainability. See Appendix A for a summary of the workshop.
traditional sources or storage backup and optimized infrastructure investment. The success of the smart grid will, in part, be about the ability of the industry to shift from its current static operational, management, and planning models to a model that is increasingly dynamic—a scenario that CS research and IT are well poised to address. Whole new situational awareness tools are required to observe, monitor, and control the smart grid. The computational burden of doing this is significant, and the industry relies almost exclusively on vendors to supply solutions—vendors who typically do not invest a great deal in research and development. Switching from static balanced optimal power flow to dynamic transient analysis that can be solved in real time and at scale is not achievable today, but this ability will be a requirement for managing the future bidirectional, rapidly changing nature of the smart grid.
IT and CS approaches will have a fundamental role in aligning the temporal and spatial characteristics of resources and users and in reducing the need for the close alignment between supply and demand. Different methods and approaches will be needed for sustainable energy systems in small developing countries, for the micro-grid in developed countries, and for a continental-scale energy system. Major CS research is required to address these and related challenges.
Agriculture in the United States and other parts of the world over the past century has been characterized by a dramatic increase in productivity, resulting in relatively affordable and available food. Some of the driving factors for this increase include the relatively inexpensive availability of fossil fuels and abundant fertilizer and water; concentration and specialization in farm production, including the increased use of automation and robotics in meat processing; the increased mechanization of farming and the availability of new technologies; advances in plant breeding; government programs and subsidies; and the expansion and commercialization of markets. These developments have allowed for food to be produced at unprecedented volumes and have supported significant population growth.29 The increase in output from agriculture, unlike that in many other industries, has not been associated with a similar increase in inputs. For example, the acreage of cropland used in 2005 was comparable to
29Indeed, a recent issue of The Economist carried a “debate” about whether computing was the most significant technological advance of the 20th century, and the “anti” side, articulated by Vaclev Smil, argued that the transformation of agriculture enabled by the production of sufficient nitrogen for fertilizer with the Haber-Bosch process of nitrogen fixation was much more significant. See http://www.economist.com/debate/days/view/598.
the acreage used in 1910. Thus, there has been a tremendous increase in productivity in U.S. agriculture, with more food being produced with significantly less capital, land, labor, and materials.30
However, current agricultural practices pose challenges to the sustainability of the food system, as well as to the broader social, economic, and environmental systems within which they are embedded. Much of the focus of agriculture has been on maximizing production to satisfy human food, feed, and fiber needs while secondarily considering environmental and societal impacts. There is a growing concern regarding the negative consequences of current trends in agricultural productivity and a concern that these trends cannot continue indefinitely. Increases in agricultural productivity have spawned hypoxia in coastal and inland waters around the world because of increased concentrations of nitrogen and phosphorus, altering the planet’s biogeochemistry. A sustainable food system will be key to ensuring that the world’s population receives necessary nutrition without contributing additional damage to the environment and society. As with the electric grid, the opportunities for IT seem most salient in the systems issues in sustainable agriculture.
The recent report of the National Research Council on sustainable agriculture defines a “sustainable agriculture system” as one that (1) satisfies human food, feed, and fiber needs and contributes to biofuel; (2) enhances environmental quality and the resource base; (3) sustains the economic viability of agriculture; and (4) enhances the quality of life of farmers, farmworkers, and society as a whole.31 The first point naturally requires both sufficient food production and a population sized appropriately for the food that can be produced. The American Public Health Association, in its policy statement on sustainable food systems, builds on these ideas, defining a sustainable food system as one that “provides healthy food to meet current food needs while maintaining healthy ecosystems that can also provide food for generations to come with minimal negative impact to the environment.”32
A sustainable food system will need to address simultaneously all four of the objectives listed above rather than optimizing over any individual dimension. A 2012 NRC report of two workshops provides a broad exploration of food security, agriculture, and related sustainability chal-
30National Research Council, Toward Sustainable Agricultural Systems in the 21st Century, Washington, D.C.: The National Academies Press (2010).
31National Research Council, Toward Sustainable Agricultural Systems in the 21st Century, Washington, D.C.: The National Academies Press (2010).
32American Public Health Association, “Toward a Healthy, Sustainable Food System” (2007). Available at http://www.apha.org/advocacy/policy/policysearch/default.htm?id=1361.
lenges. It notes that “neither the modern food systems nor the traditional systems assure long term food security for all” and examines availability, access, and utilization as well as barriers to expanding production (without damaging future capacity) and policy, technology, and governance interventions that could help.33
This section brie y explores the challenges facing the creation of a sustainable food system that promotes public health and identifies potential areas in which IT and research in computer science can have an impact.
Challenges to Developing a Sustainable Food System
A few of the many challenges to creating a sustainable global food system are highlighted below. They include increasing demand, environmental impacts, and public health impacts.
Increasing Demand The U.S. population increased by approximately 9 percent from 2000 to 2009, and total consumption of food has increased in parallel. In addition to the increase in the total amount of food consumed, the composition of the nation’s diet has shifted toward an increased consumption of meat beyond levels recommended by federal guidelines. Since 1960, there has been an increase in the use of grains for livestock feed, and so a shift toward meat consumption produces a greater strain on the agricultural system.34 Rising incomes in emerging markets such as Mexico and China have produced greater demands on U.S. agricultural exports, and such demand is likely to increase in the future. The emerging biofuels and bioenergy fields have also placed further demands on agriculture to provide materials for alternative energy production. In 2007 and 2008, 23 percent of the U.S. corn harvest and 17 percent of the soybean harvest were used to produce ethanol and biodiesel. The various demands on agriculture today increasingly strain the natural resources of land and water that are required to satisfy global food needs.
Environmental Impacts Agriculture is a contributor to greenhouse gas (primarily methane and nitrous oxide) emissions through various soilmanagement activities and livestock operations. Through biomass burning and windblown dust, farms also serve as sources of air pollutants, such as particulate matter. Conventional industrial agriculture applies
33National Research Council, A Sustainability Challenge: Food Security for All: Report of Two Workshops, Washington D.C.: The National Academies Press (2012), p. 2.
large amounts of nitrogen-based fertilizers in order to replenish nutrients in the soil and substantial quantities of herbicides and pesticides to control both plant and insect pests. The increased use of such fertilizers and pesticides leads to runoff during floods or heavy rains, which pollutes rivers, streams, and bays. Tilling of the soil contributes to land degradation, and farming in dry regions consumes water resources for irrigation. It is estimated that in the United States, 80 percent of available potable water is used for agricultural irrigation, and overdrafting of underground aquifers (when the rate of extraction exceeds the rate of natural recharge) threatens agricultural activity in a large swath of the U.S. Midwest.35
Public Health Impacts In addition to the environmental impacts of modern agriculture, there are public health impacts from current agricultural practices. Air and water pollution from farms damages not just the environment but also the health of the individuals living and working in or near the damaged environment. Factory farming of food animals has also increased the risk of foodborne pathogens, in part due to the close quarters of animals kept in confined animal feeding operations (CAFOs). The interplay between supply and demand of highly processed foods has health implications as well, even as debate continues about the specific mechanisms and contributors to diseases such as diabetes and heart disease.
Approaches to Developing Sustainable Food Systems
Creating a more sustainable global food system will not be easy. This section outlines several approaches, none of which is sufficient alone, although each contributes to increased sustainability and could benefit from the contributions of computer science and IT. The approaches include taking a systems view; developing methods for measuring the costs, benefits, and impacts of different agricultural systems; the use of precision agriculture; information for informed consumption; and the development of social networks for local food sourcing.
Taking a Systems View Overall, there is a need to take a systems view of agriculture (much like taking a systems view in other areas of sustainability, such as the smart grid, described previously) in order to understand and analyze the total impact of agriculture on the environment, economy, and society. A systems perspective is relevant at all points in the sys-
tem. At the farm itself, individual farms can combine crop and livestock production so as to reduce the need for synthetic fertilizers. Traditional agriculture has focused on controlling the farm ecosystem by simplifying it (e.g., through monoculture) and applying external inputs (e.g., water, fertilizer, pesticides, and herbicides). The systems view seeks to reduce or eliminate those external inputs (and their associated carbon and pollution emissions) by, for instance, designing and managing a more complex ecosystem involving a larger variety of species.36 A systems view can provide guidance on how to develop ways to make the system as a whole more sustainable. For instance, rather than viewing a farm (or farms) in isolation and having inputs and outputs, one could view the entire cycle of food production and consumption as providing natural resources for growing food that is consumed by people. This cycle includes land, water, and other farm inputs, crops, transportation, processing, retailing, consumption, and recycling or waste. At each stage, there are effects on the environment and society; thus it is important to consider the connections between farms, the ecosystem, and communities (local, regional, and global). An important role for IT is to enable farmers to manage these more complex systems through mechanisms such as sensing, predictive modeling, and precision machinery.
Methodology for Measuring Costs, Benefits, and Impacts There is a substantial need for the development of methods and tools to measure the total costs, benefits, and impacts of different agricultural systems. For example, comparative studies of GHG emissions from different field-management practices for animal wastes would allow for better quantification of the environmental impacts of agricultural systems and, just as with the smart grid scenario, allow for prices to reflect costs and value better. In general, evaluating different farming systems will require assessing how each system balances productivity and efficiency with environmental and societal impacts and will require analyzing the behavior of complex high-dimensional and highly interactive systems. In addition to the technical challenges of developing such measures, there are also significant challenges in helping them to be seen as accurate and legitimate by both producers and consumers. Novel visualization techniques, explanation facilities, interactive simulations, and other techniques may help here.
36The control of pests provides an example of moving from a traditional view to a systems view. The traditional way of controlling pests is to apply pesticides, which requires little knowledge of the pests. A more sustainable approach may be to use benign control measures, which require an understanding of the pest’s life cycle and its interaction with other parts of the farm ecosystem.
Precision Agriculture The use of information and computing technology in agriculture has greatly increased in the past 50 years. It has allowed farmers to assess variation within fields and to generally maintain or increase yields while reducing inputs (particularly water, nutrient, and pesticide application). Technologies used here include the Global Positioning System, real-time kinematics, and geographic information systems, especially satellites. IT already plays a substantial role in this area and will continue to play a critical role in the future. There is also a connection with methodologies for measuring costs and benefits: if the cost of water for agricultural use reflects its true cost, there may be much more incentive to use precision agriculture to reduce the consumption of water.
Information for Informed Consumption Increasing the information available to individuals regarding the nature of the food that they buy and how it was produced can assist them in making sustainable choices about food. Already there is an emerging market for foods that have been produced in a sustainable manner.37 An important method by which such information is currently conveyed is through the development of standards, certifications, or other eco-label programs. Each of these programs outlines a set of criteria for food producers and distributors in an effort to address various environmental, sustainability, or health goals.38 Perhaps the most well-established food standard in the United States is the organic agriculture certification, which focuses primarily on health and environmental goals and does not address the broader goals of sustainable agriculture. Food-labeling requirements in the United States provide some information, for example, on the country of origin of meats and fruits, but general information about sustainability and food transport (which has implications for fossil fuel usage) is not available. Current standards and certifications are typically communicated using logos or other print labeling on food packaging. However, potential exists for providing much richer information regarding sustainability and information to help consumers sort through the proliferation of eco-labels in the market. The wider adoption of smartphones may allow for easier dissemination of this information, as users could search for sustainability information at the point of purchase. One example of empowering individuals with information is the Monterey Bay Aquarium’s Seafood Watch guide39 that
37National Research Council, Toward Sustainable Agricultural Systems in the 21st Century, Washington, D.C: The National Academies Press (2010). Chapter 6.
39The Monterey Bay Aquarium’s Seafood Watch guide is available at http://www.montereybayaquarium.org/cr/seafoodwatch.aspx. The guide provides a list
provides detailed and up-to-date information about what types of seafood are caught or farmed in a sustainable manner. In addition to a web site, the guide also is available as a smartphone application so that consumers can have portable access to its vast database.
Social Networks for Local Food Sourcing IT could be used to increase networking among individuals and organizations, encouraging locally and regionally sourced food consumption. Community-supported agriculture (CSA) already benefits from the organizing power of online networks to distribute relevant information, create markets for local farm producers, make it easier to place orders, and help connect consumers with local food. Generally, IT could be used to help make a more effective market for local foods.40 Beyond efficiency, there is little argument that humans have emotional connections to food; techniques to strengthen the farmer/consumer connection could also be valuable. IT could also be useful for gathering information on regional surpluses or deficits, allowing fresh foods to be allocated to areas where they are most needed and diminishing reliance on processed foods with longer shelf lives.
The Role of Information Technology and Computer Science in Achieving a Sustainable Food System
As with the smart electric grid, information and data management are essential to making progress toward a smarter, more sustainable, global food system. Computer science research and methodological approaches will be needed at all levels to address the broad systems challenges—encompassing the environment and ecosystems, social and economic factors, and personal and organizational behaviors—affecting food production, distribution, and consumption. Three critical areas are described briefly below: information integration; education and reform; and systems modeling, prediction, and optimization.
Information Integration Information integration can help individuals and organizations on both the demand and the supply side of the food system
of sustainable choices and the least sustainable choice of fish to consume. Legal Sea Foods has questioned the value of the guide. See http://www.nrn.com/article/legal-sea-foods-defies-aquarium%E2%80%99s-watch-list.
40As one example of an effort in this area, see http://www.urbaninformatics.net/projects/food/ regarding a project exploring “ubiquitous technology for sustainable food culture in the city.” Another example is LocallyGrown.net, which seeks to provide an online infrastructure and organizational capacity for local farmers’ markets and CSAs, particularly small-scale growers with few or no employees.
make sustainable choices regarding the production and consumption of food. Providing consumers with information about the sustainability of food production, in addition to other aspects of sustainable food systems such as health and environmental impacts, will require the sharing and integration of information across producer and consumer platforms. Developing and optimizing the infrastructure and architecture for such information integration will be an important contribution of IT. More generally, areas in which IT could be of substantial help include the creation of databases of information and the maintenance of the currency of that information as well as connecting farmers and consumers through social networks and the Internet. The development of analytical software for optimizing sustainable food purchasing choices for both consumers and large-scale purchasers (such as supermarkets) is another rich area of IT contributions.
Education and Reform Tools are needed to help both consumers and policy makers understand the trade-offs posed by the global food system and to navigate those trade-offs toward increased sustainability. The role of IT here is not just in providing information on availability and techniques, but also in allowing access to communities of individuals with similar interests. There are numerous opportunities to effect change through demand-side modification of food consumption.41 Efforts to encourage the preparation and even the growing of food at home could have a significant impact on overall distribution needs. Increasing the availability of fresh, healthful foods in certain communities (e.g. low-income communities) would also help. Additional challenges exist in predicting the information that purchasers and individuals will need, displaying information that will encourage more sustainable consumption habits, educating consumers about sustainable choices without overwhelming them, and so on.
Systems Modeling, Prediction, and Optimization Improving the efficiency of the food system in general will require modeling a complex and interactive system and methods for predicting food shortages and surpluses in order to help ensure that food is available in different regions at
41Andrea Grimes, Martin Bednar, Jay David Bolter, and Rebecca E. Grinter, EatWell: Sharing nutrition-related memories in a low-income community, Proceedings of the 2008 ACM Conference on Computer Supported Cooperative Work (2008); Andrea Grimes and Richard Harper, Celebratory technology: New directions for food research, Proceedings of the Twenty-Sixth Annual SIGCHI Conference on Human Factors in Computing Systems (2008); and T. Aleahmad, A. Balakrishnan, J. Wong, S. Fussel, and S. Kiesler, Fishing for sustainability: The effects of indirect and direct persuasion, Extended Abstracts from Conference on Human Factors in Computing Systems (2008).
reasonable costs. In addition, the transportation of food to various markets could be optimized according to sustainability cost functions if a comprehensive model of the food system were available. Given a model of the food system, one could also assess the costs and benefits of various agricultural and farming strategies, the design of food sheds, and distribution systems.
The resilience of the nation’s societal and physical infrastructures poses deep and crosscutting sustainability challenges, especially when one takes a broad view of sustainability that encompasses economic and social issues. For example, although transportation is a major source of GHG emissions and urban sprawl consumes open space and farmland, competing incentives in the realm of societal sustainability include the need for workers to commute to jobs, for people to have access to whole foods, and for available space that allow businesses to change and adapt over time. Contributing to the challenges of resilience of societal and physical infrastructures is the increasing risk of natural and human-made disasters. Sustainability concerns related to climate change, resource consumption, and land use are closely linked to natural and human-made disasters.42 There will inevitably be more disasters, and enhancing society’s resilience and ability to cope with them will contribute to sustainability. Even apart from climate and resource consumption, the sheer magnitude of the world’s population means that crises, when they happen, will be at larger scale. This section examines the sustainability challenges around planning and modeling infrastructure and anticipating and responding to increasing disasters and the ways in which information technology can assist with developing sustainable and resilient infrastructures. The section focuses on cities as centers of large human populations, but many of the issues discussed apply generally.
Challenges to Developing More Sustainable and Resilient Infrastructures
Cities are highly complex, evolving systems, involving the interaction of numerous people and processes, as well as natural and built infrastructure, legal and regulatory frameworks, and much else. The diversity of use within the systems adds another level of complexity. Each building’s use and design are unique within a particular city; each city’s infrastructure has distinctive characteristics. The heterogeneity of structures within any
42National Research Council, Adapting to the Impacts of Climate Change, Washington, D.C.: The National Academies Press (2010).
given city poses challenges—and because cities are often quite different from one another, the extrapolation of lessons learned is also challenging.
Just as cities are increasingly complicated, the challenges of coping with disasters are compounded by their heterogeneity. There are acute natural disasters (such as hurricanes, earthquakes, and floods), acute engineering and other human-made disasters (such as the 2010 Deepwater Horizon Gulf oil spill), as well as “slow” or chronic disasters (such as droughts, refugee crises, and rising sea levels). In addition, whether acute or chronic, there are the ongoing processes of cleanup and recovery from disasters. Many situations are best described as combinations of natural and human-made disasters with both acute and chronic time frames.43
The problems associated with the resilience of societal and physical infrastructures have complicated time lines. For instance, urban, suburban, and rural areas are developed over long periods of time and are almost constantly being shifted into new uses. These long time lines create legacy systems that may not be compatible with newer systems or that could be costly to update. Planning becomes increasingly complicated as new infrastructure, often costly and time-consuming to implement, must anticipate the future needs of a particular area.
Similarly, the time needed and the ability to prepare an area for potential emergencies vary and depend not just on characteristics of the area, but also on the anticipated types of disasters and crises. Some disasters, like hurricanes, come with at least some advance warning, and others, like earthquakes, strike at unpredictable times. Some events cause intense damage only in limited areas, while others affect enormous geographical regions. An additional challenge is that the frequency of disastrous events is such that recovery after one event (itself a major sustainability challenge) may well not be complete before the next major disaster strikes—either in the same region, as happened with Hurricane Katrina and the Deepwater Horizon oil spill, or different regions competing for resources and attention, as in the earthquake in Haiti in 2010 that was followed by severe flooding in Pakistan.
The Role of Information Technology in
Developing Sustainable and Resilient Infrastructures
Information and communications technologies offer a range of methodologies, approaches, applications, and tools that will be integral to the
43Author Bruce Sterling coined the term “Wexelblat Disaster” to refer to disasters caused by the interaction of natural disasters and failures of human-engineered technology. The 2011 earthquake and tsunami that destroyed a nuclear power plant in Japan leading to core meltdowns is an example.
development of sustainable and resilient infrastructure and to coping with disasters when they occur. Several such technologies are highlighted below.
Modeling and Simulation Urban regions can be modeled with varying degrees of spatial detail and behavioral realism. For a highly disaggregate, behaviorally realistic model,44 the process of modeling a new region is time-consuming, often requiring person-years of effort. A major factor is difficulties in collecting and readying the needed data. Further, problems of missing data—common in U.S. metropolitan regions and even more so in developing countries—make the task much more challenging. Modeling the development of cities over periods of 20 or more years, under different alternatives, can provide important information to inform public deliberation and debate about alternate plans and possible futures. Transportation modeling, and more comprehensively integrated modeling of urban land use, transportation, and environmental impacts, have a substantial history and are in operational use in many regions. Nevertheless, there are major limitations in current knowledge, and new research is needed to address the coming challenges adequately. In addition to the scientific challenges of the modeling itself, it is important to consider how the modeling work fits into the larger political and organizational process of making major decisions (often a contentious process), and to shape the technology to respond to these contextual challenges.
Turning from simulations of long-term development to immediate support for coping with disasters: during a disaster copious amounts of information can be collected; however, more does not always mean better or more helpful information.45 Sorting out how to manage and use IT capabilities at hand most effectively and, perhaps even more importantly, the vast amounts of data that can be made available by those capabilities, is a non-trivial exercise.46,47
45Bruce Lindsay, Social Media and Disasters: Current Uses, Future Options, and Policy Considerations, Congressional Research Service (2010). Available at http://www.fas.org/sgp/crs/homesec/R41987.pdf.
46See “Disaster Relief 2.0: The Future of Information Sharing in Humanitarian Emergencies,” available at http://www.unocha.org/top-stories/all-stories/disaster-relief-20-futureinformation-sharing-humanitarian-emergencies, for an early assessment of crowdsourcing information and data flows in a humanitarian crisis. In this case the Haiti earthquake of 2010 was a primary example.
47Dan Reed, vice president of Microsoft Research, discussed some of the computational challenges posed by the 2010 Gulf oil spill, noting that the disaster stemmed from a “complex multidisciplinary system with emergent behaviors across a wide range of temporal and spatial
In addition to modeling the effects of current disasters, IT offers opportunities for in-depth simulation of potential disasters and for individuals to exercise and manage a given organization’s response to a crisis to hone and refine their skills and approach.
Communication IT provides the communications capabilities before, during, and after a crisis for coordinating activities and for delivering alerts and warnings to affected populations. IT provides critical capabilities for the other phases of crisis response as well, such as modeling and simulation to predict likely consequences or to contribute to the understanding of the effectiveness of particular mitigation measures. As discussed in a 2007 NRC report, IT provides capabilities that can help people make better sense of information, grasp the dynamic realities of a disaster more clearly, and help them formulate better decisions more quickly. IT provides the tools to capture knowledge and share it with disastermanagement professionals and the public. IT can help keep better track of the myriad details involved in all phases of disaster management.48
The Role of Information Technology and Computer Science Research in Developing Sustainable Infrastructure and Fostering Resilience
Advances will be needed in IT and computer science research and methodological approaches to enable better simulations and better understanding of the uncertainties associated with achieving more sustainable development that is also more resilient in the face of disaster. Advances are also needed in the areas of encouraging citizen participation, developing indicators of resilience and future outcomes, and improving IT infrastructures themselves.
Performance Running a simulation for a high-end, behaviorally realistic model for a major metropolitan region is a slow process, currently often requiring days, even on today’s fast computers. Similarly, the process of constructing a new scenario (i.e., a package of infrastructure improvements, zoning changes, tax incentives, and perhaps such things as tolling
scales.” He described some of the challenges in modeling such a system: “we lack the software engineering and programming methodologies to assemble, test and verify an integrated solution … the computational demands of an integrated, fully multidisciplinary, parametric simulation study of the oil spill and its effects would make accurate climate modeling seem like child’s play on an abacus by comparison.” Dan Reed, Lessons from the Gulf of Mexico (2010), available at http://www.hpcdan.org/reeds_ruminations/2010/08/lessons-from-the-gulf-of-mexico.html.
48National Research Council, Improving Disaster Management: The Role of IT in Mitigation, Preparedness, Response, and Recovery, Washington, D.C.: The National Academies Press (2007).
or congestion pricing) can require months of work by experts in transportation, land use modeling, and other disciplines. Creating far more efficient algorithms for modeling systems that are increasingly complex presents an important challenge. One natural approach is parallelizing the algorithms. In a number of cases this is quite feasible: for example, when modeling residential location choice (where will people decide to live, given characteristics both of the household and the possible dwellings), one can use massive parallelism, with each household making its decisions independently. One must then undo some assignments if two households attempt to move into the same place simultaneously (perhaps mirroring what happens in real life with several people all trying to rent or buy the same dwelling). However, new or improved algorithms are likely a richer source of performance gain, which will be important because many of the applications envisioned require huge performance increases (for example, using a simulation in real time in a meeting, or running a simulation many, many times to compute information about uncertainty). The precomputing of key scenarios and interpolating among the results (when the changes are smooth rather than abrupt), rather than computing the results from each scenario from scratch, should also be investigated. In terms of algorithms, one class of new algorithms that should be investigated is multiscale models, in which the simulation is first run at a relatively coarse grain (e.g., a zonal level), and the results from this are fed to further simulation runs within each zone, and so forth. (See Chapter 2 for more on modeling.) In this case the reason for using a multiscale model is performance. Heterogeneous models are also relevant for urban simulation—for example, coupling UrbanSim (a regional-scale model) with statewide freight mobility models. This could be further optimized by simulating only within zones that have changed significantly from the prior simulation period or that are of particular policy interest, and otherwise remaining at the coarser level.
Managing Uncertainties Urban modeling is rich with uncertainties on many levels, including future population, global economic conditions, the price of energy, the impact of climate change, and many others. There have been some successes in propagating uncertainty through the modeling process and capturing it in the indicators that the system produced,49 but much more needs to be done in terms of both statistical techniques and effective presentation of the results.
49Hana Ševcíková, A. Raftery, and P. Waddell, Assessing uncertainty in urban simulations using Bayesian melding, Transportation Research Part B: Methodology 4:652-659 (2007).
Citizen Participation To date, citizen science (or citizen information gathering) is being used for such activities as open mapping projects, but much of this type of activity has not been integrated with modeling work. Harnessing the energies and interests of citizen scientists has strong potential, both as a source of additional data and as an avenue for public participation and the legitimation of the modeling activity. Leveraging existing technology (such as mobile applications, cloud services, mapping and location services, microcommunications platforms, social media, and so on) offers numerous opportunities to improve approaches to emergency and disaster management.50
Some organizations are experimenting with gathering situational awareness from citizens, and in particular citizen use of social media.51 At the same time, there are significant challenges with regard to data quality, coverage, and institutional acceptance, among other things. Technical approaches here may include reputation systems that let staff at institutions build up confidence in particular observers, and ways to correlate data from multiple observers and to detect outliers.
During disasters, more attention should be paid to the information and resources held by the public because members of the public collectively have a richer view of a disaster situation, may possess increasingly sophisticated technology to capture and communicate information, and are an important source of volunteers, supplies, and equipment. Again, the information provided by the public will not always be correct; further, making full use of it may require considerable changes to existing practices. It is likely that the development of new, automated, and mixed-initiative techniques to manage and process the potential flood of information will be needed. Another important factor is how to engage the entire population, given the existence of groups with cultural and language differences and other special needs.
Indicators of Future Outcomes Simulations already produce indicators of such outcomes as GHG emissions, consumption of open space, and comparative measures of compact versus low-density development, all for multiple years and under different scenarios. However, as discussed above, it is also necessary to anticipate disruptions and potentially even disasters, due to climate change, mass movement of refugees, and other
50National Research Council, Public Response to Alerts and Warnings on Mobile Devices: Summary of a Workshop on Current Knowledge and Research Gaps, Washington, D.C.: The National Academies Press (2011).
51Sarah Vieweg, Amanda Hughes, Kate Starbird, and Leysia Palen, Microblogging during two natural hazards events: What Twitter may contribute to situational awareness, Proceedings of the 2010 ACM Conference on Computer Human Interaction, pp. 1079-1088.
factors. A research challenge is to develop indicators of community resilience in the face of such events.52 These might include the percentage of electrical energy generated locally (or that could be generated locally if need be), the redundancy of the transportation system and the food supply chain and their ability to cope with a sharp increase in fuel prices or even rationing, the ability to cope with sea-level rise (if relevant), the ability to walk to the most significant destinations if need be, the availability of food produced nearby, and so forth. These indicators need to be accepted by decision makers and the community to be useful in the political process. More abstract and much more difficult, if not impossible, to incorporate into a predictive model (but nevertheless important) are the civic capital and connectedness of the community.
IT Infrastructure Improvements Large disasters upset physical infrastructure, such as the electric grid, transportation, and health care—as well as IT systems. IT infrastructures themselves need to be more resilient; IT can also improve the survivability and can speed the recovery of other infrastructure by providing better information about the status of systems and advance warning of impending failures. Finally, IT can facilitate the continuity of disrupted societal functions by providing new tools for reconnecting families, friends, organizations, and communities.
IT and computer science could have a major impact in a wide diversity of sustainability challenges. The examples above illustrate some of the efforts that are needed. Individual problems are highly multidimensional, requiring innovation in different areas of computing as well as deep domain knowledge.
FINDING: Although sustainability covers a broad range of domains, most sustainability issues share challenges of architecture, scale, heterogeneity, interconnection, optimization, and human interaction with systems, each of which is also a problem central to CS research.
The next chapter explores more specifically the potential for computing and IT research and innovation to help address these challenges.
52An example of this is the Climate Change Habitability Index. For a description, see Yue Pan, Chit Meng Cheong, and Eli Blevis, The Climate Change Habitability Index, Interactions 17(6):29-33 (2010).