We are given two irregular picked seismic horizons of 578 samples each, shown in the right panel of Figures 1 and 2 after nearest-neighbor binning into a 40x40 grid. To obtain these horizons, an interpreter picked ``iso-velocity surfaces'' in time from a 3-D cube of stacking velocities, then converted the surfaces to depth via the same velocities. Note that the bounds of the data do not cover the entire grid - a disturbing problem which we wish did not exist, but will rectify shortly.
20 wells reach the shallow horizon, while only 17 reach the deeper one. The left panel of Figures 1 and 2 display the depth to the actual subsurface horizons in which we're interested, as measured by the well logs. Due to a further subwindowing of the data on my part, only about 15 wells are included in the calculations.