Observational Evidence

While the variations of climate over time scales of ten thousand to a hundred thousand years in all likelihood are caused by variations in solar irradiation over the year due to orbital effects, the so-called Milankovitch effect (Milankovitch, 1920, 1941; Berger,

1988), climate variations on shorter time scales are still rather mysterious. The most spectacular of these variations, at least as interpreted from ice-core measurements, appear to have amplitudes of several degrees Kelvin and were particularly common during the last glaciation (e.g., Alley et al., 1993). Occurrence of such extreme events during the Holocene, at least for the last 8000 years or so, has so far not been reported. Also, the less extreme climate fluctuations are of considerable importance to society. There have been numerous reports of climate variations over the last several hundred years, including a period of relatively warm climate, at

Observed surface temperature correlated with the global average temperature

Observed surface temperature correlated with the global average temperature

180 120W 60W 0 60E 120E 180

FIGURE 1 (a) Observed pointwise correlation of the annual surface temperature with the global averaged temperature based on observations from the period 1950-1995. (b) The same for a 300-year control simulation with ECHAM4/OPYC3 coupled model. Note the area of slight negative correlation in the North-Atlantic Greenland area both in the observations and in the model results. Similar patterns are found in the model also when averaged over longer periods, for at least until 50 years means.

180 120W 60W 0 60E 120E 180

FIGURE 1 (a) Observed pointwise correlation of the annual surface temperature with the global averaged temperature based on observations from the period 1950-1995. (b) The same for a 300-year control simulation with ECHAM4/OPYC3 coupled model. Note the area of slight negative correlation in the North-Atlantic Greenland area both in the observations and in the model results. Similar patterns are found in the model also when averaged over longer periods, for at least until 50 years means.

least in Europe, during the 11-13th centuries, and a relatively long period of cold climate, the so-called little ice age from the 14th to the end of the 19th century.

Available observational records, instrumental as well as indirect information on past climate, are spatially rather restricted. Before the end of the 18th century they are available mainly from Europe and central China, together covering only some 3% of the earth surface. Furthermore, available data as well as model simulation studies show that the patterns of surface temperature anomalies have rather distinct signatures, with some areas of the earth in fact being negatively correlated with the global average temperature (Fig. 1). A notable region is the Atlantic-Arctic sector, including parts of Northern Europe, which in fact is slightly negatively correlated with the global average temperature. This has the surprising effect that Iceland, Greenland, and Northern Scandinavia generally are colder than normal when the average temperature of the earth is higher than normal. The reverse can be seen over the tropical part of the Pacific and the Indian ocean and is strongly correlated with the global averaged temperature. It is interesting to note that climate models are capable of reproducing this particular pattern rather well (Fig. lb). Model experiments also suggest (e.g., Fig. 8 in Bengtsson, 1997) that climate anomalies over large geographical regions can continue over several decades due to internal low-frequency variations in the climate system. This means that it may be quite misleading to rely too heavily on observational information which is restricted by geography and time when we wish to draw general conclusions on climate events in the past and relate such events to specific external forcing mechanisms such as variations in solar irradiation or atmospheric changes due to volcanic eruptions.

Mann et al. (1998, 1999) have addressed this problem in a commendable systematic and comprehensive way. By combining available instrumental and palaeo-data at annual resolution, they have produced a continuous record of the annually averaged surface temperature of the Northern Hemisphere for the period 1000 until present. The method is based on the determination of the characteristic empirical orthogonal functions (EOFs) for the present climate and then the projection of the available palaeo-data onto these modes (Fig. 2). For information before the middle of the 18th century, one must rely on palaeo-data such as those from ice cores, tree-rings, and corals. Before 1450 even such data at an annual resolution are sparse, so the reconstructed temperature evolution has large error bars. An important aspect of the methodology used in the study of Mann et al. is that such error bars follow a priori. The reduction in the size of the error bars with time reflects the steadily improved data set and its geographical coverage, making it possible to determine more EOFs.

Three important aspects in Figure 2 need to be highlighted. First, there is an indication of a general ongoing cooling on the order of 0.1 K until ca. 1900. (This cooling trend is more clearly seen in Mann et al. (1999), where the record is extended over the whole period 1000-1998.) It is concluded that this cooling trend is in broad agreement with the Milankovitch forcing. Second, there are characteristic temperature fluctuations from year to year but with

0.78/1998

-0.8-0.8 1400 1500 1600 1700 1800 1900 2000

1860 1998

Reconstructed 2 s error bars

Cl 50 year lowpass filter ■ NH temperature (Parker, 1999)

FIGURE 2 Reconstructed surface temperature from 1400 until present (after Mann et al., 1998). Observed surface temperature data from Parker (1999, personal communication) have beeen inserted.

typical low-frequency variations of several decades. These fluctuations extend over the whole record. Third, there is a pronounced warming from the early part of the 20th century, reaching large values in the past few years. The warmest decade of the last 1000 years is the 1990s, with 1995, 1997, and 1998 being the warmest years in the whole record, with more than 3 standard errors than any year back to 1400. There are two circumstances in Figure 2 which require a more substantial analysis. These are the low-frequency temperature variations and the steep temperature warming taking place during the last century. Mann et al. (1998) have offered a set of explanations based on a simple correlation with the assumed solar variations, a volcanic index, and a simplified expression for the greenhouse gas forcing.

I will discuss this in more detail using some recent climate simulation experiments (Roeckner et al., 1999; Bengtsson et al, 1999) as tools in such an evaluation. However, first I will discuss the possible mechanisms responsible for the variation of the Northern Hemisphere temperature.

0 0

Post a comment