Temperature Reconstructions


In this section, we describe, present, and interpret the annual and seasonal temperature reconstructions, associated uncertainties, and raw data used in calibration and verification

A. Large-Scale Trends   B. Spatial Patterns   C. Influence of Climate Forcings

 


A. Large-Scale Trends

The reconstructed annual Northern Hemisphere mean temperature series is shown in Figure 6. Based on this reconstruction, MBH98 argued that the warmth of the 1990s (3 years in particular: 1990, 1995, 1997) was unprecedented in at least the past 600 years, taking into account the self-consistently estimated uncertainties in the reconstruction back to AD 1400. Based on the most recent extensions of this reconstruction (Mann et al, 1999), it furthermore now appears that 1998 was likely to have been the warmest year of at least the past millennium.

 

Northern Hemisphere (NH) Mean Temperature Reconstruction from AD 1400-1980.

Figure 6: (pictured above)
Northern Hemisphere (NH) Mean Temperature Reconstruction from AD 1400-1980, shown with raw instrumental NH series (red) through 1998. The low-frequency trend (timescales longer than 40 years emphasized) is shown by the thick curve. The blue shaded region indicates the 2 standard error uncertainty limits in the reconstruction (see MBH98 for details).

The raw NH annual mean series used for calibration (based on the full sampling available during the 1902-1980 calibration period--see Figure 2) and verification (based on the sparser sampling available during the 1854-1901 verification interval--see also Figure 2) are also shown (Figure 7) along with the reconstructed NH series constructed from the corresponding spatial samplings (this is for the sake of comparison with the available instrumental record back in time; the NH series reconstruction discussed elsewhere is based on the full spatial sampling of the calibration period, which is implicit in the pattern reconstructions back in time). The good overall correspondence between the reconstructed NH series in both calibration and independent verification intervals visually confirms the quantitative indications of statistical skill discussed earlier.

Reconstructed Northern Hemisphere (NH) mean temperature series vs. raw instrumental NH series from 1854-1980 - click here for larger image.

View Larger Image
View Sparse Instrumental Data
View Dense Instrumental Data

Figure 7:
Reconstructed Northern Hemisphere (NH) mean temperature series vs. raw instrumental NH series from 1854-1980. For the purposes of a meaningful comparison, the NH spatial means have in this case been diagnosed in both the raw data and reconstructions from the sparse gridpoint coverage of the verification period from 1854-1901, and the dense coverage of the calibration period from 1902-1980.

 

 

We focus on the NH mean temperature series because it is the most reliable hemispheric estimate given the available spatial sampling in the surface temperature fields (Figure 2). Nonetheless, Southern Hemisphere (SH) and Global (GLB) mean temperatures can be diagnosed from the appropriate areally-weighted averages of the available spatial sampling in the pattern reconstructions. Taking into account the limitations in these latter estimates (the SH sampling is almost entirely tropical and subtropical, and the GLB estimate is necessarily dominated by the coverage in the Northern Hemisphere half of the domain), some interesting conclusions can be drawn. While the two hemispheres (NH and SH series) show similar temperature trends during the past few centuries (Figure 8), the coldness of the 19th century appears to be somewhat more pronounced for the Northern Hemisphere. The GLB series, which is dominated by the Northern Hemisphere half of the domain, shows similar character to the NH series. Only through assembling a greater distribution of both instrumental and proxy data in the Southern Hemisphere, will it be possible to calculate truly meaningful estimates of Southern Hemisphere and Global temperature variations during past centuries.

Figure 8: Comparison of reconstructed annual-mean temperature trends for Southern Hemisphere, Northern Hemisphere and Global Mean, diagnosed from the available spatial sampling(Fig. 1).

It is also instructive to examine the trends in different latitude bands. Overpeck et al. (1997) suggested that post 1850 warming was more dramatic at high northern latitudes relative to lower latitudes due to larger positive feedbacks at high latitudes. The annual mean temperature trends at high-latitudes are seen (Figure 9) to be greater than the hemispheric trends themselves. In contrast, the tropical (30S-30N) band shows less change than the entire Northern Hemisphere series.

 

Reconstructed (NH) mean temperature series (full latitudinal coverage  vs. the average reconstructed series for the extratropical latitude) - click here for larger image.

Raw Latitude Band Data
Recon Latitude Band Data
Raw Tropical Data:
Annual, Cold Season, Warm Season
Reconstructed Tropical Data:
Annual, Cold Season, Warm Season
Raw Extratropical Data:
Annual, Cold Season, Warm Season
Reconstructed Extratropical Data:
Annual, Cold Season, Warm Season

Figure 9:
Reconstructed Annual Mean Northern Hemisphere (NH) mean temperature series based on full latitudinal coverage (red) vs. the average reconstructed series for the extratropical latitude 30N-70N band (blue).

 

Comparison of reconstructed NH (red) and North American (blue) regional temperature variations during past centuries - click here for larger image.

Reconstructed European Data: Annual, Cold Season, Warm Season
Raw European Data:
Annual, Cold Season, Warm Season
Reconstructed N. American Data:
Annual, Cold Season, Warm Season
Raw N. American Data:
Annual, Cold Season, Warm Season

Figure 10:
Comparison of reconstructed NH (red) and North American (blue) regional temperature variations during past centuries.

It is also instructive to compare hemispheric trends with other more regional temperature trends. In Figure 10, we show the reconstructed NH series along with an areal mean reconstruction over the North American region, during the past few centuries. It is clear that the 19th century was especially cold in North America (approximately 0.6 degrees C colder than the entire NH mean), and the subsequent warming trend of the 20th century accordingly more dramatic (i.e., approximate 1.2 C vs. 0.6 C). The fluctuations are significantly greater on almost all timescales for the North American series, which is simply a consequence of the spatial sampling statistics of a smaller region. It is thus clear that one would be remiss in drawing conclusions regarding hemispheric-scale temperature changes from such highly-variable regional temperature estimates, underscoring the importance of drawing inferences from the largest-scale mean trends in which regional "noise" is dampened and certain types of signals (e.g., the influence of climate forcings--see later section) are more clearly detected. Seasonal distinctions are also clearly important. For example, it is clear that interannual fluctuations in European cold-season temperatures are considerably greater than those during the warm season (Figure 11). This observation is consistent with the impact of the large year-to-year variability in the predominantly cold-season NAO phenomenon (see e.g. Luterbacher et al, 1999; Mann, 2000; Cullen et al, 2000). The impact of the NAO, and detailed regional inferences, are discussed in detail in the subsequent section.


Figure 11 Comparison of European Temperature Trends Back to 1750 for cold and warm half-year seasonal windows.

 

 

We also provide the time histories of the first 5 reconstructed principal component (RPC) series (along with their raw counterparts from 1902-1993). Note that the RPCs are available for different lengths back in time owing to the decreasing spatial degrees of freedom resolved by the multiproxy network back in time (see MBH98 for a discussion).

Reconstructed Principal Component (RPC) series for the first 5 eigenvectors - click here for larger image.

View Larger Image   View Annual RPC #1 Data   View Annual RPC #2 Data

View Annual RPC #3 Data   View Annual RPC #4 Data   View Annual RPC #5 Data

View Warm Season RPC #1 Data   View Warm Season RPC #2 Data   View Warm Season RPC #3 Data

View Cold Season RPC #1 Data   View Cold Season RPC #2 Data   View Cold Season RPC #3 Data

View Cold Season RPC #4 Data   View Cold Season RPC #5 Data

Figure 12:
Reconstructed Principal Component (RPC) series for the first 5 eigenvectors (see Figure 3) back in time, along with their 20th century instrumental counterparts. [Reprinted with permission from Mann et al (1998). Nature (London), 392, 779-787. Copyright (C)1998 Macmillan Journals Limited.]


On to... Spatial Patterns