Comment on "Long-period astronomical forcing of mammal turnover" by van Dam et al., Nature 443:687-691

Abstract

Van Dam et al. (2006) claimed 2.4 to 2.5 and 1.0 Myr turnover cycles in a Spanish rodent lineages record. However, the record variance spectrum, which is missing in ibid., shows that the varying reliability and multiple alterations of raw (gapped and unaltered) data by ibid. unnaturally boosted the 2.5 and 1.0 Myr noise cycles to the 99% confidence level, while failing to recognize a third such so called 99% significant noise period, of 0.55 Myr at 5.3 var%. Thus at least one claimed period (of 1.0 Myr) is a simple modulation of a relatively stronger noise cycle (of 0.55 Myr) overlooked by ibid. All so called 99% significant periods reach mere 5 to 6% var levels which can be hardly distinguished from noise: those periods fidelity is at staggering 1 to 2 orders of magnitude below the usual signal to noise separation marker at 12.0. Remarkably, at least ten noise periods got boosted to 95% confidence level, and some five noise periods to near 95% confidence level, as well. Even the zero padding of just 4% of data, as done by ibid., significantly suppresses (hence unreported) the strongest 99% significant period, of 7.28 Myr at 7.5 var%. Therefore, the periods claimed are due to strong noise reflection of some intermediary. As hand waving cyclic cataclysm claims start to frequent scientific journals, revision of editorial policies is called for on spectral analyses of inherently gapped long records, and of records composed mostly of natural data of significantly inconsistent reliability.