Sie sind hier: Startseite Publikationen und Wissenschaftliche Arbeiten FGH Abstracts Freiburger Geographische Hefte, Heft 69

Freiburger Geographische Hefte, Heft 69

Dirk Riemann (2012): Methoden zur Klimarekonstruktion aus historischen Quellen am Beispiel Mitteleuropas.

Against the backdrop of the current debate on the size of the anthropogenic contributions to a range of natural processes, the task of gaining a comprehensive understanding of the global climate system assumes an especial urgency. By exploiting information derived from long-term, high-resolution time series representing a range of climate parameters, the predictive power of climate models can be improved. Moreover, high resolution analyses enable improved estimates of the regional effects of climate change. In this context, the discipline of Paleoclimatology has played an increasingly significant role in recent times.

The ubiquity and quality of text-based historical sources in Central Europe enables a very fine-timescale description of local weather conditions for the last 500 years (comprising, for example, thermal and hygric information), almost without exception down to a monthly resolution. The sources in this period are sufficiently rich in content to permit a summary into seven temperature classes, ranging from -3 (extremely cold) to 0 (nor- mal), to +3 (extremely warm). For the earlier period 1000-1500, the data is more sparse and only a yearly resolution is possible in general; nonetheless, individual sources can sensibly be assigned one of the three index values [-1, 0, 1] (corresponding to „cold“, „normal“ and „hot“). This collection of index values was integrated by GLASER (2008) into a continous time series, which represents the fundamental data source of the present work. Although the investigations are principally devoted to temperature reconstruction, the methods developed herein are quite general, and can equally be used to reconstruct other weather parameters from corresponding index values.

One of the initial questions of this thesis was whether such historical sources can provide an adequate medium- and long-term description of climate change. In the index series derived by GLASER (2008), one observes long-term behaviour clearly corresponding to certain well-known climatic phases, such as the Mediaeval Optimum, Little Ice Age and modern temperature increase. But why are such historical sources are at all in a position to describe long-term signals? If anything, one would intuitively expect a loss of long-term signal, on the assumption that the perceptions of chroniclers would adapt to the prevailing temperatures of the day (Thus for example, a winter recorded as „warm“ in a cold period may be in terms of average temperature equivalent to a winter described as „cold“ in a warm period). This expectation, however, is based on a strictly statistical approach. In periods in which the average temperature value deviates from the total average over the whole millennium, the statistically defined boundaries of the temperature classes that are described by one index step may be shifted on average. If, however, one considers the absolute temperature values, then it turns out that the shift in average temperature value for these time intervals is on average less than 1°C. Given both the high interannual and average monthly temperature variabilities (STDs of up to 3°C), we can assume that the shift in average temperature was not always noticed: A cold winter has always been percived as a cold winter throughout the millennium.

In fact a change in outlook is necessary in order to view fluctuations in average temperature as a retrospective construction. If one perceives medium- and long-term climate changes as a consequence of fluctuations in the frequency of unidirectional temperature anomalies over a longer time period, then uncertainties in the index values of particular months or years are qualified. Statistical uncertainty is further diminished by the fact that historical sources, in contrast to other climate proxies, frequently contain explicit references to all individual months in a given year, which multiplies the amount of data available. The extensiveness of the data set is further enhanced by the multitude of sources it contains, as well as their diverse provenance - the values attached to individual years and months are frequently derived from several independent sources. Moreover, even from individual sources it is often possible to derive rough objective estimates of temperature from descriptions of physical events such as the freezing of a body of water, or phenological details. It is therefore concluded that the climate descriptions contained in the historical sources are sufficiently finely differentiated that one can characterise them with index values in a reasonably systematic manner. This assertion is especially valid for those periods in which numerous sources from a wide spread of regions are available. We can moreover assume that the inhabitants of the region under investigation experienced within a relatively short time span more or less the full gamut of weather conditions, and were therefore aware of the high variability in average temperature.

An important finding of the present investigation is the realisation that one can describe medium- and long-term climate fluctuations with only three seasonally resolved index classes. In this context, a suitable method for the calibration of index data was used: specifically, the index variability over longer-term periods was scaled to match measured variability for the same periods, a process known as vriance scaling. A general mathematical rule was derived, which in turn also allows a quantitative estimate of the temperature anomalies related to one index step. The uncertainty arising from this transition to index classes can be estimated by means of MC-simulations. This estimation demonstrates that the long-term signal remains clearly recognisable, even in the presence of spurious index values for individual years or months. Due to the very high monthly and seasonal variabilities of the region of investigation, as well as the high interannual variability, erroneous index values resulting from adapted perceptions of the chroniclers will apply perhaps to individual years and months, but by no means to all data points over a long time period.

The result of the calibrated index series is a temperature reconstruction for Central Europe for the last 1000 years, in which the major climate trends can be recognised and quantified. For example, it is evident from this reconstruction that the yearly average temperature in the region since 1980 has consistently exceeded the average across the millennium. It is however noteworthy that this increase is primarily attributable to a growth in winter temperatures. The summer temperatures over the last thousand years exhibit a consistently high variability, a pattern which is also to be found in the period since 1980. Nonetheless, summer temperature anomalies contribute significantly less than winter temperature anomalies to the overall signal, owing the lower amplitude of their variability. Measured with respect to the magnitude of temperature fluctuations, the temperature increase since 1900 is unparalleled. The size of the observed short- and medium-term fluctuations are however by no means unique to modern times.

The quantitative estimation in °C of the information given by individual indices opens up the possibility of developing a statistical model to regionalize index-based climate information for Central Europe, reflecting the spatial distribution of the data sources. Moreover, the accuracy of the reconstructed signal for the entire area can be substantially increased, because the reconstructed fields are a more faithful reflection of realistic climatic patterns. In this way it is possible to realistically incorporate information from different regions and thereby more precisely determine the temperatures of selected regions. Therefore, the model can be also used to verify the statements made in the historical sources using climatological criteria. At the time of writing, the HisKliD data-set does not allow the construction of a continuous index series. Although such a model could in future play a role in bridging the various subdisciplines of climate reconstruction, this aspect is not considered due to the present limitations of the database.

The viability of index-based climate reconstruction is examined by means of a multi-proxy study, in which several sets of dendrochronological and historical source-based data sets are each reduced to a time series consisting of three-valued, seasonally resolved indices. A common signal is then extracted from the time series by means of a majority principal: it is assumed for example to be warm when most of the time series agree this is so. In this manner a continuous climate signal was derived for Central Europe, based on a broad data set comprising several different climate indicators.

Indices represent a sufficiently variegated source of information to describe not only long-term climate change on a region-wide basis, but also accurately represent localised weather patterns. This result has possible ramifications for paleoclimatological research for time periods of millennial length: armed with this information as a basis, research efforts can be shifted away from the accrual of new data to construct continuous time series, to more targeted investigations aiming to verify or discredit assertions derived from time series about weather at particular points in time. Taken collectively, such distinctive individual events enable the derivation of a long-term climate signal. Distinctive descriptions of seasonal conditions in a chosen year, like for example the presence of droughts or very cold winters, are very helpful for climate reconstruction, while much of the rest of the signal is simply noise.

Benutzerspezifische Werkzeuge