PerspectivePALEOCLIMATE

Blowing Hot and Cold

See allHide authors and affiliations

Science  22 Mar 2002:
Vol. 295, Issue 5563, pp. 2227-2228
DOI: 10.1126/science.1069486

Hiding behind a rather dry title, Esper et al., on page 2250 of this issue, provide a new and important vision of the detailed course of changing temperatures throughout the last millennium (1). Their analysis is based exclusively on tree-ring records from 14 locations spread over much of the northern extra-tropics. Though virtually all previous Northern Hemisphere temperature reconstructions use at least some tree-ring data, the authors use many new data and a processing technique that provides a largely independent history of widespread tree-growth variations, which they scale against modern temperature observations to estimate the relative magnitude of past temperature changes.

The new record differs in several respects from that highlighted in the Synthesis of the Third Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) (2), which focused on the 1000-year reconstruction of Mann et al. (see the purple line in the figure) (3). This record has a smaller amplitude of century-to-century variability and is consistently at, or near, the upper limit of the range of alternate records produced by other researchers (4-8).

Records of past climate.

Solid colored lines indicate seven reconstructions of Northern Hemisphere climate: yellow, (4); red, (5); purple, (3); orange, (6); green, (7); blue, (8); and pink (1). All records were re-calibrated with linear regression against 1881-1960 mean annual temperature observations averaged over land areas north of 20°N, and the results smoothed with a 50-year filter. The black dotted line shows the estimate that would be made if the predictor was observed warm-season temperatures from the same region, highlighting the difference between warm-season and annual temperature changes during the observed record. Black solid line: smoothed observations, truncated in 1993 when the record of Esper et al. ends. Gray lines: annual temperature changes estimated from Northern Hemisphere borehole temperature profiles [dotted line, unweighted average of many sites (9); solid line, records gridded before averaging].

The curve from Esper et al. (pink line) shows a pronounced cold phase in the 17th century, in qualitative agreement with the other records and especially with a record of borehole temperature data (see the figure) (9), more so when the latter are first gridded to reduce bias due to regional concentrations of these records. The borehole data (and data from Mann et al.) are interpreted as indications of true annual temperatures, incorporating both warm season and cold season signals. All records in the figure have been calibrated assuming that they portray annual warmth. It is possible, however, as Esper et al. state, that their tree-growth data are more influenced by summer than winter conditions. This affects not only their own record but also a number of the tree-ring series used in other reconstructions shown in the figure.

To place their record on an absolute scale and allow direct comparison of past temperature changes with 20th century observations, Esper et al. scale their series by matching the magnitude of its multidecadal trends to those in Northern Hemisphere mean (land and marine) annual temperatures from 1900 to 1977. After smoothing to remove year-to-year fluctuations, the records can be matched closely with either the annual or summer mean temperatures, because their trends over this period are very similar.

For the early 17th century, annual temperature estimates from Esper et al. differ by about 0.7°C from those of Mann et al. [see figure 3 of (1)]. However, when we regressed the record of Esper et al. against nonsmoothed data (see the figure), this difference was reduced to about 0.4°C. Recalibrating both curves against year-by-year warm season temperatures (10) reduces this difference further to about 0.35°C.

The results of calibrating any proxy data depend on whether raw or smoothed records are used and on the chosen seasonal temperature predictand. Reconstructions of annual temperature records with predictors that are strongly influenced by summer conditions, assume stationary relationships between proxy and summer climate and between annual and summer climates (and hence between summer and winter). The relationship was stationary over the Esper et al. calibration period, but over other periods it may vary: summer warming of extra-tropical land has progressed at a slower rate than winter warming in the Northern Hemisphere (by about 0.6°C since 1860, see dotted line in the figure) and this is predicted, on the basis of climate model experiments, to continue over the next century at least.

Whatever the true degree of cold in the 17th century, a surprising aspect of the results of Esper et al. is the indication of equally cold conditions throughout the 12th, 13th, and 14th centuries, where their reconstructed temperatures are consistently well below those indicated by all other records. On the evidence of this new series, the last millennium was much cooler than previously interpreted. The warming of the 20th century is seen more clearly as a continuation of a trend that began at the start of the 19th century, not the early 20th, and an early period of warmth in the late 10th and early 11th centuries is more pronounced than in previous large-scale reconstructions. This warmth also peaks slightly earlier than could be captured in the shorter Mann et al. record and is warmer than in any previous reconstruction.

Even accepting the knotty issue of reconstruction uncertainty, the curve of Esper et al. provides evidence for greater climate swings in the last 1000 years than has yet been generally accepted. We need more independent reconstructions like this, based on improved proxy records, and we need to know why it was once so warm and then so cool, before we can say whether 21st-century warming is likely to be nearer to the top or the bottom of the latest IPCC range of 1.4° to 5.8°C (2).

References and Notes

View Abstract

Navigate This Article