Yesterday I highlighted the paper The extra-tropical Northern Hemisphere temperature in the last two millennia: reconstructions of low-frequency variability, by B Christiansen of the Danish Meteorological Institute and F C Ljungqvist of Stockholm University which showed that using a multitude of proxy samples in the norther hemisphere, that:
“The level of warmth during the peak of the MWP (Medieval Warm Period) in the second half of the 10th century, equaling or slightly exceeding the mid-20th century warming, is in agreement with the results from other more recent large-scale multi-proxy temperature reconstructions.”
Now another paper, by Esper et al published in the Journal of Global and Planetary Change, shows that not only was the summers of the MWP equal or greater than our current warmth, but that the summers of the Roman Warm Period of 2000 years ago were significantly warmer than today.
Variability and extremes of northern Scandinavian summer temperatures over the past two millennia
Jan Esper, Ulf Büntgen, Mauri Timonen, David C. Frank
Palaeoclimatic evidence revealed synchronous temperature variations among Northern Hemisphere regions over the past millennium. The range of these variations (in degrees Celsius) is, however, largely unknown. We here present a 2000-year summer temperature reconstruction from northern Scandinavia and compare this timeseries with existing proxy records to assess the range of reconstructed temperatures at a regional scale. The new reconstruction is based on 578 maximum latewood density profiles from living and sub-fossil Pinus sylvestris samples from northern Sweden and Finland.
The record provides evidence for substantial warmth during Roman and Medieval times, larger in extent and longer in duration than 20th century warmth.
The first century AD was the warmest 100-year period (+0.60 °C on average relative to the 1951–1980 mean) of the Common Era, more than 1 °C warmer than the coldest 14th century AD (−0.51 °C). The warmest and coldest reconstructed 30-year periods (AD 21–50=+1.05 °C, and AD 1451–80=−1.19 °C) differ by more than 2 °C, and the range between the five warmest and coldest reconstructed summers in the context of the past 2000 years is estimated to exceed 5 °C. Comparison of the new timeseries with five existing tree-ring based reconstructions from northern Scandinavia revealed synchronized climate fluctuations but substantially different absolute temperatures. Level offset among the various reconstructions in extremely cold and warm years (up to 3 °C) and cold and warm 30-year periods (up to 1.5 °C) are in the order of the total temperature variance of each individual reconstruction over the past 1500 to 2000 years. These findings demonstrate our poor understanding of the absolute temperature variance in a region where high-resolution proxy coverage is denser than in any other area of the world.
Discussion and Conclusions
The MXD-based summer temperature reconstruction presented here sets a new standard in high-resolution palaeoclimatology. The record explains about 60% of the variance of regional temperature data, and is based on more high-precision density series than any
other previous reconstruction. Importantly, MXD sample replication prior to the Little Ice Age, during Medieval times and throughout the first millennium AD, is much better than in any other record, and we demonstrated – based on calibration trials using reduced
datasets – that these early sections of the N-Scan record likely still contain useful climate information. This persistent climate signal allowed an estimation of temperature variability throughout the Common Era, revealing warmth during Roman and Medieval times were larger in extent and longer in duration than 20th century conditions.
According to this new record, summer temperatures varied by 1.1 °C among the 14th and 1st centuries, the coldest and warmest 100-year periods of the past two millennia. Temperatures ranged by more than 5 °C among the five coldest and warmest summers of the past 2000 years. These estimates are, however, related to the approach used for proxy transfer, i.e. figures would change, if the calibration method, period, and/or target were modified (Frank et al.,2010b). For example, variance among the 30 coldest and warmest N-Scan summers (Table 3) increases from 3.92 °C to 5.79 °C, if scaling (i.e. adjustment of the mean and variance) instead of OLS regression is used for proxy transfer. These differences between scaling- and regression-based approaches are proportional to the unexplained variance of the calibration model (Esper et al., 2005), and we suggest
smoothing the proxy and instrumental timeseries prior to calibration, as this procedure decreases the unexplained variance in all Scandinavian tree-ring records and thus minimizes the differences between various calibration methods (Cook et al., 2004).
Our results, however, also showed that these methodological uncertainties are dwarfed by the variance among the individual reconstructions.
Differences among six northern Scandinavian tree-ring records are>1.5° in 30-year extreme periods and up to 3 °C in single extreme years, a finding we didn’t expect, as the proxy records: (i) all calibrate well against regional instrumental data, (ii) partly share the same measurement series (or use differing parameters – TRW and MXD – from the same trees), and (iii) originate from a confined region in northern Scandinavia that is characterized by a homogeneous temperature pattern. Since we here calibrated all reconstructions using the same method, between-record differences are likely related to varying data treatment and chronology development methods, measurement techniques, and/or sampling strategies, as well as the remaining uncertainty typical to such proxy data. For example, splicing of MXD data on recent TRW trends as done in Briffa92 might have caused this reconstruction to appear at the lower (colder) end of the ensemble, whereas the combination (and adjustment) of novel digital MXD measurements with traditional X-ray based MXD data as done in Grudd08 might have caused this reconstruction to appear at the upper (warmer) end of the ensemble. Other differences are likely related to the combination of sub-fossil material from trees that grew in wet conditions at the lakeshores with material from living trees growing in dryer ‘inland’ sites. Also varying variance stabilization (Frank et al., 2007) and detrending techniques (Esper et al., 2003) in association with temporally changing sample replications and age distributions of the underlying data (Melvin, 2004) likely impacted the variance structure of the long-term records and consequently the absolute levels of reconstructed temperatures.
Between-reconstruction variance as revealed here represents a pending challenge for the integration of proxy records over larger regions and the development of a single timeseries that represents the Northern Hemisphere (e.g., Mann et al., 2008), for example. The composition of such records commonly relies on the calibration statistics derived from fitting regional proxy records against instrumental data (D’Arrigo et al., 2006). However, the records analyzed here would all easily pass conventional calibration-based screening procedures. Yet our analysis revealed that choosing one Scandinavian record instead of another one can alter reconstructed temperatures by 1.5-3 °C during Medieval times, for example. On the other hand, consideration of all records presented here would likely promote a less variable climate history, as the combination of diverging records tends to reduce variance in the mean timeseries (Frank et al., 2007). If such a mean is then combined with instrumental data covering the past 100–150 years, this approach might facilitate hockey stick-shaped reconstructions (Frank et al., 2010a). This seems to be a tricky situation in which expert teams including the developers of proxy records might need to be involved to help assessing timeseries beyond the typical ranking based on calibration statistics.
Our results showed that introducing an improved temperature reconstruction does not automatically clarify climate history in a given region. In northern Scandinavia, we now arrive at a situation where a number of high-resolution proxy records – all passing classical calibration and verification tests – are available within a confined region that is characterized by homogeneous temperature patterns. These records, however, differ by several degrees Celsius over the past two millennia, which appears huge if compared with the 20th Century warming signal in Scandinavia or elsewhere. We conclude that the temperature history of the last millennium is much less understood than often suggested, and that the regional and particularly the hemispheric scale pre-1400 temperature variance is largely unknown. Expert teams are needed to assess existing records, and to reduce uncertainties associated with millennium-length temperature reconstructions, before we can usefully constrain future climate scenarios.
Full paper here (PDF -link fixed)
h/t to WUWT reader Gordon Pye and Tory Aardvarrk