In light of the evidence that mantle plumes appear to have drifted Bernhard Steinberger (Steinberger et al. 2004) proposed a global model for the mantle which allowed for plumes to 'drift' in global convection cells, sometimes refered to as "Mantle Wind" This model was also devised to try and explain the Hawaii-Emperor bend event (43Ma) which is not consistent with observed plate motion of the Pacific
The devised model could indeed be adjusted to fit the aparent southward drift of the Hawaiian hotspot (and a few others in the Pacific and Indian oceans) and the overall track up to the 43Ma bend but using a mantle flow model based on reconstruction by connecting Pacific and African plates through Antarctica places the Hawaii track too far west of where it is now.
Using a different mantle flow model based on reconstruction through Australia and Lord Howe rise the track is predicted up to 65Ma. Steinberger comments that these mistfits show our incomplete understanding of intraplate deformation and that there may be more to the explanation of hotspot tracks
The idea that the mantle convects is not new and is generally thought to occur to some degree. Calculated Rayleigh and Prandtl numbers indicate that the mantle is very likely to convect in a non chaotic manner but the modelling of this
convection is very complicated and observations of mantle convection are scarce at best. Steinberger's convection driven drift is an elegant solution to the problem of hotspot tracks but seems to require too much adjustment to fit current observations - i.e. it does not provide solutions on its own
Mantle plume drift history for the Hawaiian plume Accross a north-south cross section at 155°W.
Violet line is a fixed source drift history and the red line is a moving source history
a)120Ma b)90Ma
c)60Ma d)30Ma
e)Present
This is a diagram from Steinberger's paper showing some different models and the resulting computed track for various island chains
Tracks show movement with ticks every 10My
Red lines (model 1, shown for Hawaii and Louisville), and the purple line (model 2,
shown for Hawaii), are computed for absolute plate motions such that the fit to Tristan and
Re´ union hotspot tracks is optimized for the two plate motion chain models as indicated;
optimization parameters are African plate rotations 0–47 and 47–83 Myr ago.
Green lines
(shown for Re´ union and Tristan) are for optimizing the fit to Hawaii and Louisville for plate
motion chain model 1; parameters are Pacific plate rotations 0–25, 25–47, 47–62 and
62–83 Myr ago.
Black (model 1) and blue (model 2) lines are for optimizing jointly to all
four tracks for the two plate motion chain models as indicated; parameters are African
rotations 0–47, 47–62 and 62–83 Myr ago.
Dotted lines, hotspots assumed fixed; shortdashed
line, only hotspot motion on African hemisphere considered (shown for Hawaii);
long-dashed line, only hotspot motion on Pacific hemisphere considered (shown for
Hawaii); continuous lines, all hotspot motions considered.
It is obvious that all of the models are placing the tracks too far west of present day tracks. Steinberger states that to fit observations they would need a strong westward component of plume drift to explain the Emperor bend but their evidence for circulation from palte reconstruction shows no evidence for this.
Although the model has a poor fit before 43Ma the later tracks are very well explained, the case may well be that some unknown plate deformation played a major role before that which may be evident presently.