Do transform faults parallel plate motion?

A central principle of plate tectonics is that relative plate motion is parallel to transform faults. Several workers have convincingly argued, however, that transform fault valleys widen with age due to horizontal thermal contraction of the lithosphere [Collette 1974; Roest et al 1986 ; Sandwell 1986]. If so, then the transform fault zone, which is the locus of active strike-slip faulting in a transform fault, may not be parallel to the direction of relative plate motion, but parallel to the relative velocity of the walls of the transform valley. Here we apply a recent model for horizontal contraction of oceanic lithosphere as a function of age [Kumar & Gordon 2009] to calculate this bias in transform azimuth as a function of offset, spreading rate, and ridge length, with slightly different formulations for crenelate and stepping mid-ocean ridge segments. The calculated bias increases with increasing length of ridge segments between transform faults and decreases with increasing offset of ridge segments along transform faults. We used the calculated bias to estimate corrections of azimuths of transform faults on eight plate boundaries from the MORVEL data set [DeMets, Argus, & Gordon, 2009]. The largest correction is 2.3 degrees, but the median correction is merely 0.2 degrees. We express the null hypothesis (no thermal contraction) and our hypothesis (thermal contraction as predicted by Kumar and Gordon [2009]), in terms of a parameter α, which is zero for no thermal contraction and 1 for predicted thermal contraction. When the results for the 8 plate pairs are combined, we find that α = 1.1 ± 0.7 (95% confidence limits), thus excluding the null hypothesis (no thermal contraction), but consistent with the hypothesis of horizontal thermal contraction.