Climate Change 2001:
Working Group II: Impacts, Adaptation and Vulnerability
Other reports in this collection

18.3. Future Adaptations

Predictions or estimates of likely future adaptations are an essential element of climate change impact and vulnerability assessment. The degree to which a future climate change risk is dangerous depends greatly on the likelihood and effectiveness of adaptations in that system. Studies that ignore or assume no adaptation are likely to overestimate residual or net impacts and vulnerabilities, whereas those that assume full and effective adaptation are likely to underestimate residual impacts and vulnerabilities (Reilly, 1999; Reilly and Schimmelpfennig, 1999; Risbey et al., 1999; Smit et al., 2000). Hence, it is important to have an improved understanding of the process of adaptation and better information on the conditions under which adaptations of various types are expected to occur. Such scholarship on the "how, when, and why" of adaptation is necessary to make informed judgments on the vulnerabilities of sectors, regions, and communities (Ausubel, 1991a; Kane et al., 1992a; Reilly, 1995; Burton, 1997; Smithers and Smit, 1997; Tol et al., 1998; Klein et al., 1999). Insights into processes of adaptation have been gained from several types of analysis, including listing of possible adaptation measures, impact assessment models, adaptation process models, historical and spatial analogs, and empirical analysis of contemporary adaptation processes.

18.3.1. Possible Adaptation Measures

There are many arbitrary lists of possible adaptation measures, initiatives, or strategies that have a potential to moderate impacts, if they were implemented (e.g., Benioff et al., 1996; Smith et al., 1996; Mimura, 1999b). Such possible adaptations are based on experience, observation, and speculation about alternatives that might be created (Carter, 1996); they cover a wide range of types and take numerous forms (UNEP, 1998). For example, possible adaptive measures for health risks associated with climate change listed by Patz (1996) appear in Table 18-2.

Table 18-2: Examples of multilevel adaptive measures for some anticipated health outcomes of global climate change (Patz, 1996).
Adaptive Measure
Heat-Related Illness
Vector-Borne Diseases Health and Extreme Weather Events
  • Implement weather watch/warning systems
  • Plant trees in urban areas
  • Implement education
  • Implement vaccination
  • Enforce vaccination laws
  • Implement education
    campaigns to eliminate
    breeding sites
  • Create disaster preparedness programs
  • Employ land-use planning to reduce flash floods
  • Ban precarious residential placements
  • Insulate buildings
  • Install high-albedo materials for roads
  • Install window screens
  • Release sterile male vectors
  • Construct strong seawalls
  • Fortify sanitation system
Personal behavior
  • Maintain hydration
  • Schedule work breaks during peak daytime temperatures
  • Use topical insect repellents
  • Use pyrethroid-impregnated bed nets
  • Heed weather advisories

Similarly, in coastal zone studies, comprehensive lists of potential adaptation measures are presented; these adaptations include a wide array of engineering measures, improvements, or changes, including agricultural practices that are more flood-resistant; negotiating regional water-sharing agreements; providing efficient mechanisms for disaster management; developing desalination techniques; planting mangrove belts to provide flood protection; planting salt-tolerant varieties of vegetation; improving drainage facilities; establishing setback policies for new developments; developing food insurance schemes; devising flood early warning systems; and so forth (Al-Farouq and Huq, 1996; Jallow, 1996; Rijsberman and van Velzen, 1996; Teves et al., 1996; Mimura and Harasawa, 2000). In many other sectors and regions, arbitrary lists of possible adaptations are common (Erda, 1996; Iglesias et al., 1996). In the Canadian agricultural sector alone, 96 different adaptation measures have been identified, as summarized in Table 18-3.

Table 18-3: Adaptation strategies for the agricultural sector (adapted from Smit, 1993; Carter, 1996).
Adaptation Strategy
Number of Measures
Change topography of land
Use artificial systems to improve water use/ availability and protect against soil erosion
Change farming practices
Change timing of farm operations
Use different crop varieties
Governmental and institutional policies and programs
Research into new technologies

Such lists indicate the range of strategies and measures that represent possible adaptations to climate change risks in particular sectors and regions. They show that there is a large variety and number of possible adaptations, including many with the potential to reduce adverse climatic change impacts. Many of these adaptations—especially in agriculture, water resources, and coastal zone applications—essentially represent improved resource management, and many would have benefits in dealing with current climatic hazards as well as with future climatic risks (El Shaer et al., 1996; Harrington, 1996; Huang, 1996; Stakhiv, 1996; Frederick, 1997; Hartig et al., 1997; Mendelsohn and Bennett, 1997; Major, 1998). In only a few cases are such lists of possible adaptations considered according to who might undertake them, under what conditions might they be implemented, and how effective might they be (Easterling, 1996; Harrington, 1996; Frederick, 1997; Major, 1998; Moss, 1998).

18.3.2. Impact Assessment Models

Estimates of likely future adaptations are essential parts of climate change impact models. Integrated assessment models also include assumptions about adaptations in the impact components (Leemans, 1992; Rotmans et al., 1994; Dowlatabadi, 1995; Hulme and Raper, 1995; West and Dowlatabadi, 1999). Some early studies of impacts assumed no adaptation (Tol et al., 1998), invoking the so-called "naive" or "dumb farmer" assumption. The "dumb farmer" assumption—which is not unique to agriculture—is a metaphor for any impacted agent that is assumed not to anticipate or respond to changed climate conditions but continues to act as if nothing has changed (Rosenberg, 1992; Easterling et al., 1993; Smit et al., 1996). By ignoring autonomous and planned adaptations, such studies do not distinguish between potential and residual net impacts and are of limited utility in assessing vulnerability.

An alternative approach that is common in more recent impact modeling has been to assume levels of adaptation. Applications include Nicholls and Leatherman (1995) for coastal zones, Mendelsohn et al. (1994) and Rosenzweig and Parry (1994) for agriculture, Sohngen and Mendelsohn (1998) for timber, and Rosenthal et al. (1995) for space conditioning in buildings. These studies demonstrate that adaptive measures have the potential to significantly alleviate adverse impacts of climate change and to benefit from opportunities associated with changed climatic conditions (Helms et al., 1996; Schimmelpfennig, 1996; Mendelsohn and Neumann, 1999). The models of Rosenzweig and Parry (1994) show that, with adaptations assumed, food production could be increased under climate change in many regions of the world. Stuczyinski et al. (2000) conclude that climate change would reduce Polish agriculture production by 5-25% without adaptation; with adaptation assumed, production is estimated to change by -5 to +5% of current levels. Downing (1991) demonstrates the potential of adaptations to reduce food deficits in Africa from 50 to 20%. Mendelsohn and Dinar (1999) estimate that private adaptation could reduce potential climate damages in India's agriculture from 25 to 15-23%. Reilly et al. (1994) estimate global "welfare" losses in the agri-food sector of between US$0.1 billion and 61.2 billion without adaptation, compared to +US$70 to -37 billion with adaptation assumed. These studies indicate potential rather than the likelihood of adaptation to alleviate damages (or benefit from opportunities) associated with changes in climatic mean conditions (rather than changing conditions that include variability and extremes of climate).

Impact models invariably are based on climate scenarios that focus on adaptation to changed average conditions, with little attention given to interannual variations and extremes. Limited research suggests that the potential of adaptation to cope with changes in average conditions is greater than its potential to cope with climate change-related variability. For example, Mendelsohn et al. (1999) show that, assuming adaptation, increases in average temperature would be beneficial for U.S. agriculture, but increases in interannual variation would be harmful. West and Dowlatabadi (1999) demonstrate that considering variability and extremes can lead to estimates of "optimal" adaptation and damages that differ considerably from those based on gradual changes in mean climatic conditions. The importance of considering variability, not just mean climate, when estimating adaptation is widely recognized (Robock et al., 1993; Mearns et al., 1997; Alderwish and Al-Eryani, 1999; Alexandrov, 1999; Luo and Lin, 1999; Murdiyarso, 2000).

In numerical impact models, assumptions about perception and adaptation are more commonly arbitrary or based on principles of efficiency and rationality and assume full information (Yohe et al., 1996; Hurd et al., 1997; Mendelsohn et al., 1999). As Tol et al. (1998), Schneider et al. (2000), and others have noted, however, actual and assumed behavior do not necessarily match. In an analysis of global food production, Parry et al. (1999) assume farm-level and economic system adaptations but recognize that the "adoption of efficient adaptation techniques is far from certain." In addition to questions relating to rationality principles, adaptation behavior is known to vary according to the amount and type of information available, as well as the ability to act. Hence, rational behavior that is based on assumed perfect information differs from rational behavior under uncertainty (Yohe et al., 1996; Yohe and Neumann, 1997; West and Dowlatabadi, 1999). Replacing the "no adaptation" model with one that assumes rational, unconstrained actors with full information replaces the "dumb farmer" assumption with the "clairvoyant farmer" assumption (Smit et al., 1996; Risbey et al., 1999). Reilly (1998) questions the ability and hence the likelihood of agents to detect and respond efficiently to the manifestations of climate change. Tol (1998b) also questions whether perfect foresight and rational behavior are realistic assumptions for predictive models. Schneider (1997) explores further the assumptions that underlie equilibrium approaches (ergodic economics), including the equivalence of temporal and spatial variations.

Numerical impact assessment models tend to use, rather than generate, information on adaptations to estimate future impacts of climate stimuli, after the effects of adaptation have been factored in. They indicate the potential of human systems to adapt autonomously and thus to moderate climate change damages.

Other reports in this collection