Diaz, H. F., and G. N. Kiladis, 1995: Climatic variability on decadal to century time scales. In Future Climates of the World: A Modelling Perspective, A. Henderson-Sellers (Ed.), Elsevier, 191-244.


INTRODUCTION

Variability is an intrinsic property of the Earth's climate system. It is evident in the vast glaciated landscapes of North America and northern Europe, the product of massive continental ice sheets that have periodically waxed and waned during the Pleistocene epoch (approximately the last two million years of the Earth's history). It is inferred from fossil records which indicate that, tens of millions of years ago, warm habitat species thrived in high-latitude areas which, today, are too cold to support them (see Chapter 3 by BARRON). MITCHELL (1976) developed a synthesis of the spectrum of climatic variability as a function of frequency and characteristic spatial scale. He also listed three categories of potential sources of climatic variability: (i) internal stochastic mechanisms, (ii) external forcing mechanisms and (iii) instabilities or resonant modes of the climate system.

The first category refers to processes operating within the system, defined here as the atmosphere, hydrosphere and cryosphere. Stochastic mechanisms involve chaotic feedback effects (processes which amplify or diminish an initial perturbation of the system) that arise in large part from the non-linear nature of the governing equations. The second source of climatic variability derives from processes independent of the climatic system, and involve mechanisms which are external to the state of the system, such as solar radiation changes due to orbital variations, tectonic movements, etc. The third category involves some form of resonant amplification of internal modes of the system forced by recurrent (or periodic) external mechanisms. The quasi-periodic reversal of wind direction in the lower stratosphere, known as the quasi-biennial oscillation, is an example of this type of climatic variability. The El Niño/Southern Oscillation phenomenon could be considered a less regular example of this type of feature.

On the time-scale of climatic variability of interest here (decadal to century scale variations), the variance spectrum associated with various long-term climatic indices is essentially red, that is, the amplitude of the variations increases with increasing period of the oscillation (KUTZBACH and BRYSON, 1974). Another way of stating this is that temporal persistence in long-term climatic series inflates the variance contribution of longer-period oscillations relative to those at higher frequencies (GILMAN el at., 1963). This property is illustrated in Fig. 1, where the theoretical spectra of random time series with different levels of temporal persistence are shown. The lag-one autocorrelation of a time series, which samples some continuous process such as temperature, expresses quantitatively the degree of temporal persistence between adjacent samples or time intervals of that particular process. A zero-valued lag-one autocorrelation coefficient indicates that successive realisations of the time series process are effectively independent of each other. By contrast, high values of the lag-one autocorrelation indicate that successive values are not statistically independent, by that the process, in effect, retains a "memory" of previous realisations or values. The spectrum of time series with the latter characteristics tends to exhibit higher variance at lower frequencies compared to the former class of processes (known as white noise).

Within the frequency internal of interest here (~10-100 years), there are certain periods which tend to exhibit, more of less consistently, an excess of variance when compared to higher or lower frequencies (MITCHELL, 1976; STOCKER and MYSAK, 1992). A likely source of climatic variability on these time-scales may be related to changes in the ocean's thermohaline circulation (see a review by HELD (1993)), which can modify the large-scale exchanges of heat between the atmosphere and the oceans and between high and low latitudes. Other possible sources of natural variability are volcanism, solar variability, and changes in biogeochemical cycles. Changes in the solar flux are totally external to the Earth's climate system. Volcanism may also be considered to be largely independent of climate. In the past there have been large changes in the Earth's biochemistry arising from varying combinations of climatic and non-climatic forcings. Since the Industrial Revolution, humans have entered the picture as a possible cause of regional and global scale climatic variation.

In the sections that follow, we consider each of these possible sources of climatic variability on the decadal to century dime-scales. In this chapter, information about regional and global climates during the last 1,000 years is reviewed, along with a broad evaluation of the types of data upon which much of our knowledge about the climate during the pre-instrumental period is based. Since the large-scale ocean-atmosphere phenomenon known as El Niño/Southern Oscillation (ENSO) represents the largest source of interannual variability in the modern climate system, it is also necessary to review some aspects of the low frequency behaviour of ENSO, in addition to providing an overview of the structure and evolution of these events. Finally, a brief appraisal is made of possible future changes in climate, including anthropogenically induced changes, and uncertainties in the projections of such changes derived mostly from so-called general circulation models. We also examine possible "surprises", that could arise from potentially rapid changes in the thermohaline circulation of the North Atlantic Ocean.