describes the geoid rather well:
http://en.wikipedia.org/wiki/Geoid.But to elaborate, the geoid of a perfectly homogeneous sphere is also a sphere, and would be defined on the surface of said sphere. However, Earth is internally "lumpy", containing so-called "masscons" (mass concentrations) and g also varies as 1/r^2 from the centre of the geoid, so you need a reference radius: in this case, the time averaged (10 yrs, I think) mean sea level at a specific point in Marseilles. The "près" is because sea level changes, not only by the second, but also by the year, decade and century.