Hotspots in Hindsight

Abstract:

Torsvik et al. [2006] suggest that the original locations of large igneous provinces (“LIPs”) and kimberlites, and current locations of melting anomalies (hot-spots) lie preferentially above the margins of two Large Lower-Mantle Shear Velocity Provinces” (LLSVPs), at the base of the mantle, and that the correlation has a high significance level (> 99.9999%). They conclude the LLSVP margins are Plume-Generation Zones, and deep-mantle plumes cause hotspots and LIPs.

This conclusion raises questions about what physical processes could be responsible, because, for example the LLSVPs are likely dense and not abnormally hot [Trampert et al., 2004].

The supposed LIP-hotspot-LLSVP correlations probably are examples of the “Hindsight Heresy” [Acton, 1959], of basing a statistical test upon the same data sample that led to the initial formulation of a hypothesis. In doing this, many competing hypotheses will have been considered and rejected, but this fact will not be taken into account in statistical assessments. Furthermore, probabilities will be computed for many subsets and combinations of the data, and the best-correlated cases will be cited, but this fact will not be taken into account either.

Tests using independent hot-spot catalogs and mantle models suggest that the actual significance levels of the correlations are two or three orders of magnitude smaller than claimed. These tests also show that hot spots correlate well with presumably shallowly rooted features such as spreading plate boundaries. Consideration of the kimberlite dataset in the context of geological setting suggests that their apparent association with the LLSVP margins results from the fact that the Kaapvaal craton, the site of most of the kimberlites considered, lies in Southern Africa. These observations raise questions about the distinction between correlation and causation and underline the necessity to take geological factors into account.

Fig: Left: Cumulative distributions of distances from hotspots to nearest ridge for 5 hotspot lists; heavy red curve: Distribution function for a random point on Earth’s surface. Hotspots are closer to ridges than expected at random. Right: For each list, the probability of at least as many random points being as close to a ridge. Values to right have higher significance.