0.5°C over the last century (which occurred mostly between 1910 and 1940, and again since the mid-1970s). A gradual warming of about 0.3°C at mid-depths of the subpolar North Atlantic has been observed in the last 30 years, as is noted in Levitus's paper in Chapter 3.
Changes in the variability of climate (Figure 1d) have also been documented in some of the data sets presented. For example, the interannual variance of winter temperatures in the United States increased by about 150 percent during the period 1975-1985, as described in Karl's essay. In Chapter 2, Diaz and Bradley provide evidence of a twentieth-century increase in large-scale Northern Hemisphere interannual temperature variability. Similarly, the diurnal temperature range has varied over the past 40 or so years (see Karl et al., also in that chapter). In this case, the minimum (night-time) daily temperature has risen nearly three times faster than the maximum (day-time) temperature. However, since the records in question begin in the 1950s, this change may reflect a significant anthropogenic component as well as natural variability.
We are not yet certain why these changes in climate occur. Various types of models must be used to test our hypotheses and to increase our understanding of the climate system. Models of the atmosphere, the ocean, and the coupled atmosphere-ocean-land-cryosphere system are beginning to yield insights into the causes of natural climate variations. We are beginning to realize their potential for:
Identifying the responses of key components of the climate system to changes in internal parameters, and to changes in the external forcing such as insolation or volcanic eruptions.
Explaining the sensitivity and climate signature of each of these components, internal modes of variability, and the interaction between system components (e.g., how a perturbation propagates through the system, or is attenuated or amplified by feedbacks).
Clarifying how different climate variables respond to the same change, and how certain components of the climate system influence particular time and space scales of variability.
Progress in these areas is reflected in many of the papers appearing in the earlier chapters of this volume. The Climate Research Committee considers the following results particularly noteworthy.
Recent modeling studies suggest that significant changes in the deep-water circulation may occur over time scales of decades to centuries, and that these changes may critically affect climate. The thermohaline circulation is fairly sensitive to local climate conditions in the high-latitude oceans, particularly air/sea/ice exchange in the sub-polar regions of the North and South Atlantic oceans (in Chapter 3, see the papers by Mysak and by McDermott and Sarachik; in Chapter 4, see Delworth et al.), where the deep water exchanges heat and salt with the atmosphere. Thus, relatively small changes in the climate or environment of these source regions may have profound impacts on the thermohaline circulation.
A second ocean-model finding is that the thermohaline circulation can oscillate between quasi-steady 'equilibrium' modes. This effect is apparent in both ocean models (see the papers in Chapter 3 by Barnett et al., by McDermott and Sarachik, and by Weaver) and preliminary coupled ocean-atmosphere models (see Delworth et al. in Chapter 4). Multiple equilibria might contribute to rapid climate transitions, while sustained oscillatory changes might contribute to more-or-less regular fluctuations.
Atmospheric models have traditionally led the way in modeling the earth's climate system, and satisfactory simulations of the present atmospheric circulation do exist. Earlier simulations were restricted to fixed lower-boundary conditions; current models are starting to include interaction with the underlying ocean and land surface processes. Complementary progress is being made with ocean models, and coupled models are beginning to advance as well. Despite the encouraging results described above and throughout the volume, it should be noted that satisfactory simulation of the present-day climate does not guarantee that the sensitivity of the models to prescribed changes is realistic. To properly address model sensitivity, and the realism of models in simulating decade-to-century-scale climate change, a hierarchy ranging from simple, mechanistic models through detailed process-oriented models to fully coupled ocean-atmosphere-land-cryosphere-biosphere models is needed. Many models of different construction and complexity are required to test against each other, develop better understanding of the system components, and improve parameterizations and computational efficiency.
Systematically combining observations and models, and ensuring the long-term continuity and sufficient quality of the data, will be critical to the assessment of climate variability and of the models that are used for climate simulation and prediction. The observations permit us to initialize, force, and diagnose models, providing reassurance that we are simulating the real world. As was shown in the NRC's 1991 report on four-dimensional model assimilation of data, models not only serve as the measure of our understanding and a means of prediction, but are now good enough to help guide observation, monitoring, and data-management programs. Combining theoretical and empirical evidence from models and observing systems will permit us to focus our field experiments and observational programs on those components of the climate system that