My figure was a combination of my own calculations, similar to Squyres' ones but with slightly different inputs, and the regional observations as to the thickness of the Haematite layer, based on it filling and covering craters, and that it is eroding away. The latter info was published in a range of pre-landing papers.

I went back over the calculations and agree that simply matching the spherules at the surface to a rock volume yields a much smaller figure than I had supposed (I get 1-3 m eroded thickness. My previous calculations were based on an assumption of several cm of "gravel" rather than a soil with sparse spherules below a near-continuous surface armour.

So I guess that reduces erosion rates quite a bit, if you assume that spherules are never destroyed.

OTOH, the thickness calculations from a regional perspective suggest that tens to hundreds of metres of erosion *have* occurred, which suggests that over times the spherules wear away. There is plenty of evidence for this from the soils with perfect spheres being rare and dented, chipped, broken dominating, and highly eroded irregular remnants being also common. Over time, everything wears away.

This is probably the way to match the different estimates of erosion thickness and rate...

Another line of evidence for the amount of vertical deflation comes from the lack of craters smaller than 10m or so. These would originally have had depths of ~1m, so at least that much has been removed, over a time frame (~3 Ga) during which many hundreds per square km would have accumulated. If only 1m of erosion had occurred over those millenia, then only ~50% of the small craters would be totally removed - the remainder would be partially visible. To achieve the observed ~99% reduction in number of small craters, requires ~50m of erosion in 3 Ga.

So I agree, I can't quite reach several hundred metres, but I do get many tens of metres.

Thanks for pointing out Squyres' calculations and making me think!