Data are presented from a survey by airborne scanning laser profilometer and an AUV-mounted, upward looking swath sonar in the spring Beaufort Sea. The air-snow (surface elevation) and water-ice (draft) surfaces were mapped at 1 x 1 m resolution over a 300 x 300 m area. Data were separated into level and deformed ice fractions using the surface roughness of the sonar data. The relation (R = d/f) between draft, d, and surface elevation, f, was then examined. Correlation between top and bottom surfaces was essentially zero at full resolution, requiring averaging over patches of at least 11 m diameter to constrain the relation largely because of the significant error (similar to 15 cm) of the laser instrument. Level ice points were concentrated in two core regions, corresponding to level FY ice and refrozen leads, with variations in R attributed primarily to positive snow thickness variability. Deformed ice displayed a more diffuse "cloud," with draft having a more important role in determining R because of wider deformed features underwater. Averaging over footprints similar to satellite altimeters showed the mean surface elevation (typical of ICESat) to be stable with averaging scale, with R = 3.4 (level) and R = 4.2 (deformed). The "minimum elevation within a footprint" characteristic reported for CryoSat was less stable, significantly overestimating R for level ice (R > 5) and deformed ice (R > 6). The mean draft difference between measurements and isostasy suggests 70 m as an isostatic length scale for level ice. The isostatic scale for deformed ice appears to be longer than accessible with these data (>300 m).