other than might be expected. There is no evidence of any recent dramatic or unprecedented rise in temperature.
Taken together, the reconstructions for the two hemispheres show little coherence in century-time-scale variability. The most consistent features are the cool episodes from 1550 to 1650 and again between 1880 and 1910. The early period probably represents the coldest 100-year period of the last millennium, while the latter underscores the possibly exaggerated warming trends that may be estimated by calculating hemispheric trends beginning in the 1880s rather than the 1860s (see also Folland et al., 1990).
On the basis of the instrumental record, the 1980s was clearly the warmest decade. Warmth is evident in all seasons and in both hemispheres. There is evidence, clearer in the Northern Hemisphere, that the twentieth-century warming occurred in two phases, between 1920 and 1940 and since 1975. All seasonal curves, however, exhibit large decadal-time-scale variability superimposed on century-time-scale warming.
Longer single-site and composite instrumental records confirm the warming from the late nineteenth century on. The 250-year-long records from Europe show that the 1880s were the coldest decade, at least since 1700, so part of the warming since then may reflect this unusually low starting point. In most European records the eighteenth century was warmer than the nineteenth century.
In only one (the Northern Urals, Briffa et al., 1995) of the thousand-year dendroclimatic reconstructions is the twentieth century clearly the warmest century. In all of the other reconstructions, however, this century is one of the warmest. Warmer (summer) conditions occurred in previous centuries, but never at the same time at all locations. Clearly, a greater geographical spread of long paleoclimatic reconstructions is required; types of proxy evidence that cover seasons other than summer must be employed before the apparently unprecedented warmth shown in instrumental records of the twentieth century can be placed in a longer-term context.
The authors thank Don Graybill, Ed Cook, Antonio Lara, and Ricardo Villalba for making the results of their long dendroclimatic reconstructions available. This research has been supported by U.S. Department of Energy's Atmospheric and Climate Division (Grant No. DE-FG02-86ER60397), the U.K. Department of the Environment (Contract No. PECD/7/12/78), and the E.U. Environment Programme (Grant EV5V-CT94-0500; DG12 DTEE).
THOMAS R. KARL
NOAA National Climatic Data Center
I very much enjoyed reading this paper. I should just like to call your attention to some questions that were raised in my mind as I read through it.
First, there is the question of whether or not we can really adequately measure global-scale decadal fluctuation prior to about 1900. I am primarily concerned about the spatial resolution and spatial comprehensiveness of the data. I want to show one diagram to give you an idea of the types of errors that occur when you attempt to calculate global temperature trends and do not have complete global coverage of the phenomena. Starting out with complete global coverage and systematically removing data to calculate a trend over 10, 50, 150 years—removing it in a way that matches the historical coverage of the observations over both land and sea, of course—on a 10-year time scale you get standard errors that are over 1°C. The standard errors come down to perhaps 0.4°C to 0.6°C for 50 years; when you go out to 100 years, the standard errors are much smaller. There are also some significant biases that can creep into the record. This is cause for thought.
Another question I have is related to an issue Gene Rasmusson raised: Are we optimally using the data we have? Are 5- by 10-degree boxes the best way of calculating these global numbers? Is some kind of optimal averaging preferable? You get slightly different answers depending on how you go about the averaging.
My third point is the issue that Dr. Jones raised with respect to the very cool conditions during the 1880s, especially during the summer. You wonder not only about the spatial comprehensiveness of the data at that time, but also whether there could be some exposure problems.
Another question was raised in my mind by Dr. Jones's comment that decadal-scale temperature anomalies are not hemispherically consistent. One might ask whether there is any evidence to suggest that this may be less likely to occur at longer time scales, such as centuries Can we get away