Holocene period 14C dates can be calibrated by means of dendrochronology (Stuiver et al. 1993). At present, the tree-ring curve based on absolutely dated trees and a matched floating German pine chronology (Kromer & Spurk 1998), and a new recommended curve INTCAL98, has been constructed to c. 10,200 BP or 9900 cal BC (Stuiver & van der Plicht 1998). This new curve is extended through the Late Glacial using measurements of corals, dated by both U-series isotopes and 14C (Bard et al. 1993).
There is no recommended curve for the Glacial Period; nevertheless, datasets are available which contain calibration information [ILLUSTRATION FOR FIGURE 1 OMITTED]. Firstly, the U-series/14C-dated coral series has been extended by a few isolated measurements (Bard et al. 1993). Secondly, another U-series/14C dataset for speleothems is available, albeit measured with large error bars (Vogel & Kronfeld 1997), and low in time resolution. Thirdly, a 100,000-year-old varved sediment from Lake Suigetsu, Japan has produced more than 250 14C samples, which were measured by AMS, yielding a high time resolution dataset to about 45,000 BP (Kitagawa & van der Plicht 1998). These latter data were also presented at the 16th International Radiocarbon Conference in Groningen, June 1997, where it was decided which datasets would be included in INTCAL98.
The Japanese varved record is not included in this new curve because
a it is floating and has to be matched to the tree ring curve for the Holocene part; and
b there are possible problems with the chronology of the older parts (in particular before 30,000 BP, as will be discussed below) probably caused by missing varves and/or varve-counting errors.
The speleothem data has also not been included because of potential problems such as the unknown initial 14C content, and the possibility of detrital Th contamination. Indeed, the 'speleothem curve' and 'varve curve' deviate markedly [ILLUSTRATION FOR FIGURE 1 OMITTED]. Additionally, the coral data are obviously a marine dataset; a marine reservoir age must be subtracted in order to construct a calibration curve. This reservoir age is taken as c. 500 years for INTCAL98.
Recently, van Andel (1998) presented both a calibration curve back to 45,000 cal BC, based on values of the geomagnetic field strength as measured by Laj et al. (1996), and made comparisons with the GRIP ice-core chronology, in particular the Denekamp (IS-8) and Hengelo (IS-12) interstadials, in terms of the frequency of 14C-date occurrence. Two problems arise with this treatment of calibration. First, the magnetic data used for deduction of a calibration curve are not the most suitable. Second, the chronologies carry uncertainties that do not yet justify an exact calibration of the interstadials used; a precise and accurate calibration is necessary for the deduction of calibrated frequency distributions. These two points will be discussed in detail below.
The geomagnetic record used exhibits large fluctuations of magnetic grains (Laj et al. 1996). It is based on only three cores from the same Azores area, which are not dated by 14C. It is preferable to use the palaeointensity curve reconstructed from compiling records from 18 different locations (Guyodo & Valet 1996). This record averages both local mineralogical and sedimentological problems and also regional heterogeneous effects of the geomagnetic field. Together with the global 10Be flux reconstructed by Frank et al. (1997), it was also used as input for a carbon cycle model to calculate time-dependent 14C variations (Bard 1997). This geomagnetic record explains the general trend observed in the varved chronology from Lake Suigetsu. These data are plotted in FIGURE 2 as [Delta]14C (see footnote). Note that 1 per mil [[Delta].sup.14]C corresponds to 8 BP; thus 500 per mil [[Delta].sup.14]C means that 4000 BP is the difference between radiocarbon years and calendar years (deviation from the Libby line; compare [ILLUSTRATION FOR FIGURE 1 OMITTED]). …