Climate Change 2001: Working Group II: Impacts, Adaptation and Vulnerability

	Other reports in this collection

2.2.2.3. Methodological Considerations

Studies that relate observed changes in natural biota to climatic changes are necessarily correlational. It is not possible to address this question through a standard experimental approach, so direct cause-and-effect relationships cannot be established. However, the level of uncertainty can be reduced until it is highly unlikely that any force other than climate change could be the cause of the observed biotic changes. Studies can reduce uncertainty in three ways:

• Maximize statistical power
• Design to control for major confounding factors
• For confounding factors that remain, directly analyze whether they could explain the biotic changes and, if so, quantify the strength of that relationship.

Statistical power is gained by using:

• Large sample sizes (numbers of populations/numbers of species)
• Data gathered over a large region
• Studies conducted over multiple regions
• Studies conducted on multiple taxa (different families, orders, phyla, etc.)
• Selecting populations or species without a priori knowledge of changes to minimize sample bias
• Data gathered over a long time period such that bi-directional responses to opposite climatic trends may be detected.

Confounding factors can be addressed in a correlational study. Biologists know that many nonclimatic anthropogenic forces affect population dynamics, community stability, and species distributions. These forces fall largely under the main headings of land-use change, hydrological changes, pollution, and invasive species. The term "land-use change" comprises a suite of human interventions that eliminate or degrade natural habitats, leading to loss of species that are dependent on those habitats. Habitat loss can be overt destruction, as occurs with urbanization, conversion to agriculture, or clear-cut logging. Habitat degradation is more subtle; it usually results from changes in land management, such as changes in grazing intensity/timing, changes in fire intensity/frequency, or changes in forestry practices (coppicing, logging methods, reforestation strategies), as well as irrigation dams and associated flood control. Loss of habitat by either means not only causes extinctions at that site but endangers surrounding good habitat patches by increased fragmentation. As good habitat patches become smaller and more isolated from other good patches, the populations on those patches are more likely to become permanently extinct.

The main airborne pollutants that are likely to affect distributions and compositions of natural biotic systems are sulfates, which lead to acid rain; nitrates, which fertilize the soil; and CO₂, which affects basic plant physiology (particularly the carbon/ nitrogen ratio). Urban areas, in addition to having locally high amounts of sulfates and nitrates, are artificial sources of heat. Aquatic and coastal marine systems suffer from runoff of fertilizers and pesticides from agricultural areas and improperly treated sewage, as well as fragmentation from dams and degradation resulting from trawling, dredging, and silting.

These confounding factors cannot be completely eliminated, but their influences can be minimized by (Parmesan, 1996, 2001; Parmesan et al., 1999):

• Conducting studies away from large urban or agricultural areas
• Conducting studies in large natural areas (e.g., northern Canada, Alaska, areas in Australia)
• Choosing individual sites in preserved areas (national parks/preserves, field stations)
• Eliminating from consideration extreme habitat specialists or species known to be very sensitive to slight human modifications of the landscape.

If a particular confounding factor cannot be greatly minimized, it should be measured and analyzed alongside climatic variables to assess their relative effects.

Ideal target species, communities, or systems in which to look for biotic responses to climate change meet the following criteria (DeGroot et al., 1995; Parmesan, 2001):

• Basic research has led to a process-based understanding of underlying mechanisms by which climate affects the organism or community. This knowledge may come from experimental laboratory or field studies of behavior and physiology or from correlational studies between field observations climatic data.
• The target is relatively insensitive to other anthropogenic influences, so the effects of possible confounding factors are minimized.
• Short (decadal) or no lag time is expected between climate change and response [e.g., tree distributional responses may have a lag time of centuries, so they may not be ideal for looking for distributional changes over recent decades (Lavoie and Payette, 1996)].
• There are good historical records, either from being a model system in basic research or by having a history of amateur collecting.
• Current data are available (from monitoring schemes, long-term research) or are easy to gather.

Use of indicator species or communities is crucial for defining the level of climate change that is important to natural systems and for giving baseline data on impacts. However, caution is advisable when extending these results to predictive scenarios because the indicators often are chosen specifically to pinpoint simple responses to climate change. Thus, these studies purposefully minimize known complexities of multiple interacting factors, such as:

• The direct effects of CO₂ on plants may vary with temperature.
• The outcome of competitive interactions between species is different under different thermal regimes (Davis et al., 1998), and, conversely, competitive environment can affect sensitivity and response to particular climatic variables (Cescatti and Piutti, 1998).
• Different species have different lag times for response, which inevitably will cause the breakup of traditional communities (Davis and Zabinski, 1992; Overpeck et al., 1992; Root and Schneider, 1995).
• The ability of wild plant and animal life to respond to climate change through movement is likely to be hindered by human-driven habitat fragmentation; those with lowest dispersal will be most affected (Hanski, 1999).