Reply to comment by F. Speranza et al. on “Recent eruptive history of Stromboli (Aeolian Islands, Italy) determined from high‐accuracy archeomagnetic dating”

Abstract

[1] We are grateful to Speranza et al. [2005] (hereinafter referred to as SSM) for the opportunity to clarify the respective merits of the ‘‘large sample’’ and ‘‘core-drilling and stepwise demagnetization’’ methods to determine paleomagnetic directions. The former was initiated in the ‘30s by Thellier and made popular in the fields of archeomagnetic research. SSM’s criticism can be summarized as seven points. [2] The first concerns laboratory cleaning procedures, use of demagnetization diagrams, and rock magnetic investigations. SSM feared that we might have used an inappropriate blanket af demagnetization. The Saint Maur paleomagnetic laboratory has decades of experience and full demagnetization with both af and thermal techniques, use of vector demagnetization diagrams and principal component analysis have been applied to thousands of samples. We have checked on a selected subset (more than 20 samples) from the Stromboli study that af demagnetization from NRM to 40 mT produced single component remanent vector going through the origin (e.g., sample S1-5, Figure 1a). Detailed thermal demagnetizations do not show any change in direction larger than 1–2 (e.g., sample S1-9, Figure 1b). We are using only single component samples with recent magnetizations that have not been overprinted. The minicore method has its own drawbacks: possible heterogeneity of magnetization on a larger scale, lower accuracy of orientation, risk of producing a secondary drilling induced remanent magnetization, or DIRM. Such DIRM [e.g., Genevey et al., 2002] may result in a curved demagnetization diagram up to large af (60 mT) and lead to larger dispersion and different paleodirections as observed by SSM in their samples. Such DIRM cannot occur in our case. Magnetic mineralogy (hysteresis, susceptibility, etc.) may be different from one sample to another, but we have repeatedly checked that this has no bearing on the stability of the paleodirection. Thus, the large samples result in good control of potential small-scale heterogeneity in magnetization. Therefore, SSM’s understandable concern is unsubstantiated upon analysis. [3] The second comment regards the number of discarded samples which SSM find too large, resulting in artificially small confidence intervals. In only three sites have we rejected more than one sample, and this is by no means arbitrary. For lava spatters lying on unstable ground, it is almost inevitable that some parts of the outcrop have been displaced after cooling, and Speranza et al. [2004] (hereinafter referred to as SP2004) themselves envision this possibility for three of their sites. But they give only the ‘‘number of cores giving reliable directions’’ in their Table 1, so that the effect of final sample selection cannot be evaluated in their own case. We illustrate our most criticized site, S1 (a95 = 1.0 , k = 1507) in Figure 1c. Five of the 17 samples were clearly displaced: 1) their paleodirections do not change during af demagnetization, and 2) they lie far away from the McFadden [1982] rejection circle, which contains the other 12 samples used in the average paleodirection. These 12, widely distributed, kg-size samples can hardly be considered as ‘‘undersampling’’ compared to 9 mini-cores taken from the same site (Str04 given by SP2004, with a95 = 4.3 , k = 147). Further evidence that our accuracy is not due to ‘‘arbitrary rejection of data’’ is provided by the more steady lava flows (sites S2 and S5 of Arrighi et al. [2004, Table 1] (hereinafter referred to as AR2004)), where all 22 and 11 samples, respectively, were used for the average paleodirection, giving a95 of 1.5 and 1.8 . We recall that over the last three millennia, secular variation in Europe is contained within a cone with a 12 half-angle, which requires uncertainties no more than 20 per cent of this, or about ±2 , in order to draw a high-resolution master curve. SSM are right to note two S1 samples whose declinations should be ‘‘of 3 and 7 , consistent with the mean declination of site Str04’’: this was due to an EastWest sign inversion in several figures by Arrighi [2004], the true values being negative (samples S1-5 and S1-16 in Figure 1c). We thank SSM for spotting this misprint. [4] As noted by SSM in their third comment, the geomagnetic field itself could be too distorted in volcanic terrain to properly retrieve the true geomagnetic direction; this potential problem equally applies to SP2004. The importance of this effect can be limited by adequate sampling, as shown by a detailed experimental study [Tanguy and LeGoff, 2004]: in 11 out of 12 archeomagnetic sites distributed over the whole area of Mt. Etna, the field GEOPHYSICAL RESEARCH LETTERS, VOL. 32, L23305, doi:10.1029/2005GL023768, 2005