Other reports in this collection
It is well established from physical, ecological, and physiological studies that climate strongly influences physical and biological systems. This section addresses whether changes in regional climate during the 20th century, documented by WGI, have resulted in measurable impacts on physical and biological systems. We also consider the potential for detecting observed impacts of regional climate change in socioeconomic systems. The objective here is to evaluate the accumulating body of evidence with regard to the following questions:
In this section, we focus on observed impacts that have been associated with regional climate changes over the past 100 years. We examine evidence in physical and biological systems in terrestrial, coastal and marine, and freshwater environments, as well as in socioeconomic systems, including agriculture, commercial fisheries, human settlements, insurance and financial services, and human health (see also other chapters in this report).
The studies reviewed document an observed impact in a physical, biological, or socioeconomic system associated with changes in one or more regional climate variables (most often temperature rise). The effects are examined with regard to the range and geographical extent of processes and species involved, their consistency with functional understanding of mechanisms or processes involved in climate-impact relationships, and the possibility of alternative explanations and confounding factors. Expected directions of change relating to regional climate warming for physical systems include shrinkage of glaciers, decrease in snow cover, shortening of duration of lake and river ice cover, declines in sea-ice extent and thickness, lengthening of frost-free seasons, and intensification of the hydrological cycle. Expected directions of change relating to regional climate warming for biological systems include poleward and elevational shifts in distribution and earlier phenology (i.e., earlier breeding, emergence, flowering) in plant and animal species.
We follow the WGI definition of climate change as a statistically significant variation in the mean state of the climate or its variability, persisting for an extended period (typically decades or longer). Climate change, as defined here, may be caused by natural internal processes or external forcings or by persistent anthropogenic changes in the composition of the atmosphere or land use.
Since 1860, the global mean temperature has warmed 0.6 ± 0.2°C; regional temperature changes have varied, ranging from greater than 0.6°C to cooling in some regions (TAR WGI Chapter 2). Annual land-surface precipitation has increased (0.5-1% per decade) in most middle and high latitudes of the northern hemisphere, except over eastern Asia. In contrast, over much of the subtropical land areas, rainfall has decreased during the 20th century (0.3% per decade), although it has been recovering in recent years (TAR WGI Chapter 2). The recent warming period began in 1976, with pronounced warming observed in northwestern North America, central northern Asia, and the southern Pacific Ocean. Detection of climate change and attribution of causes are discussed in TAR WGI Chapter 12.
Accumulation of evidence over time and space, based on numerous individual studies, is needed to detect and characterize patterns and processes of observed climate change impacts on a global basis (see Chapter 2). In many studies, changes in impact systems are compared with trends in climate variables over the same period and location. Many studies establish statistically significant trends in the observed impact and the climate variable, as well as a statistically significant association between the two (e.g., Beebee, 1995; Brown et al., 1999; Barber et al., 2000). Others refer to trends in climate documented elsewhere (e.g., Menzel and Fabian, 1999; Thomas and Lennon, 1999). When multiple species or locations are examined, cases are reported that exhibit no change, change that is consistent with understanding of climate-impact relationships, and change that is inconsistent with understanding of climate-impact relationships. This allows for assessment of whether observed changes are significantly different from random chance and are consistent with functional understanding of climate responses (e.g., Ellis, 1997; Ellis et al., 1997; Bradley et al., 1999; Pounds et al., 1999).
Individual studies that link observed impacts to regional climate change may be hampered by methodological problems such as length of time-series data of observed impacts; number of replications of populations, census sites, or species; availability of climate data to which to compare observed changes; and uncertainty about whether observed impacts and regional climate variables are measured at appropriate spatial scales (Chapter 2). In some regions, several individual studies have focused on differing aspects of a common ecosystem, providing evidence for associations between climate change and multiple responses in a given geographical area (e.g., Smith et al., 1999); in other regions, however, studies examine more isolated responses.
Because changing climate and ecological responses are linked over a range of temporal scales, long periods of study allow more accurate conclusions regarding the significance of observed ecosystem changes. Large-amplitude temporal changes usually involve large spatial dimensions, so broad-scale spatial/temporal ecosystem studies tend to be more robust. The majority of studies document trends for periods of more than 20 years (e.g., Post et al., 1997; Winkel and Hudde, 1997; Post and Stenseth, 1999); a few studies document trends for 10-19 years (e.g., Jarvinen, 1994; Forchhammer et al., 1998); and several studies analyze data from two periods with a gap between them (Bradley et al., 1999; Sagarin et al., 1999).
Climate Trends: The various studies of observed impacts of recorded regional temperature change over the past century, which include the recent warm decades of the 1980s and 1990s, often differentiate responses to mean, minimum, and maximum temperatures. Regional precipitation changes and periods of droughts and floods are much more variable in observed records and more uncertain with regard to future predictions and are not the primary focus here. Studies also have considered possible observed responses to the rising atmospheric concentrations of CO2 over the past century, but these studies are not included in this review.
To the extent that periodicities or trends are found in the climate record, nonzero autocorrelations are to be expected on the interannual time scale. Their importance depends on the percentage of variance associated with the periodicities and the magnitude of the trend relative to interannual noise. Often the periodicities represent only a small proportion of the total variance; this is especially true on a local level, where the noise is likely to be higher than at broader spatial scales. A nonzero autocorrelation does not automatically mean the year-to-year ecological impact is not meaningful because if year-to-year climate variability is associated with a periodic or steadily increasing climate forcing, so too would be the ecological response.
Processes and Mechanisms: Beyond statistical association, an important aspect of many studies is comparison of documented changes to known relationships between climate and impact systems. For example, under regional warming, retreat of glaciers is expected because of shifts in the energy balance of glaciers, as is poleward expansion of species' ranges when temperatures exceed physiological thresholds. If documented changes are consistent with known processes that link climate and the impact system, confidence in the associations between changes in regional climate and observed changes is enhanced.
Multiple Causal Factors: The presence of multiple causal factors (e.g., land-use change, pollution, biotic invasion) makes attribution of many observed impacts to regional climate change a complex challenge at the individual study and meta-analysis levels (e.g., Prop et al., 1998; Körner, 1999). Some of the competing explanations for observed impacts themselves could have a common driver that would make them strongly correlated; identifying these drivers is a methodological challenge. Studies seek to document observed climate change impacts by ruling out other possible contributing causative factors, ecological or anthropogenic, through study design and sampling techniques (e.g., Parmesan, 1996; Menzel and Fabian, 1999; Parmesan et al., 1999), statistical analyses (e.g., Prop et al., 1998; Reading, 1998), or expert judgment (De Jong and Brakefield, 1998; Brown et al., 1999). Sometimes, different studies offer alternative explanations for observed impacts (e.g., Körner, 1999).
Signals of regional climate change impacts may be clearer in physical systems than in biological systems, which are simultaneously undergoing many complex changes that are not related to climate, including land-use change and pollution processes such as eutrophication and acidification. Observed impacts in high-latitude and high-altitude physical systems, such as melting of glaciers, may be more straightforward to detect, whereas biological responses to climate tend to be more complex and may be masked by the presence of the aforementioned multiple causal factors. To deal with these ecological complexities, confounding factors often are minimized by conducting studies away from large urban or agricultural areas, in large natural areas (e.g., northern Canada, Australia), or in preserved areas.
Signals of regional climate change impacts probably are most difficult to detect in socioeconomic systems because such systems are strongly affected by simultaneous trends in population and income growth and urbanization and because of the presence of adaptive capacity (see Chapter 18). Observed climate change impacts in socioecosystems may be adaptations in many cases, such as farmers sowing crops earlier in response to warmer spring temperatures.
An example of these methodological complexities in climate change impact detection may be drawn from the human health sector. Although climate is known to influence many disease vectors (such as the range of anopheline mosquitos that carry malaria), the presence or absence of sanitation systems, vaccination programs, adequate nutritional conditions, animal husbandry, irrigation, and land-use management also influences whether the presence of a disease in wild vectors leads to disease outbreaks in human populations (see Chapter 9).
Evaluating Patterns of Change: Grouping individual studies to evaluate patterns and processes of change on larger spatial scales reduces the influences of study-specific biases and local nonclimatic factors. Comparing expected geographical patterns of responses to regional climate changes and to changes that are not related to climate helps distinguish among multiple possible causations. For example, regional warming would be expected to skew the distribution of insect extinctions to be greater at the southern boundaries rather than at the northern boundaries; land-use change, in contrast, would be expected to cause approximately equal extinctions at both range boundaries (Parmesan, 1996; Parmesan et al., 1999). Care must be taken to ensure that the sample of studies is representative across time and space, is not biased in reporting, and uses appropriate statistical tests. Spottiness of evidence in other regions may indicate that observed impacts of regional climate change are not occurring, have not yet been detected, or are being masked by other changes, such as urbanization.
Some studies of observed impacts have used a "fingerprint" approach, based on the definition of expected biological changes arising from regional climate change (e.g., Epstein et al., 1998). This approach is similar to that used in detection of climate changes (see TAR WGI Chapter 12) but differs in that fingerprint studies of ecosystem impacts use selected data and that long-term monitoring of changes in ecosystems generally is lacking at regional or global scales.
|
Other reports in this collection |
|