Five sound-speed models spread the Challenger Deep's maximum depth across 18 metres

An echosounder measures time, not distance. Depth comes out of a multiplication: the mean sound speed along the path times half the round-trip travel time, and sound speed in seawater changes with temperature, salinity and pressure ^[4]. At the Challenger Deep that round trip takes roughly 15 seconds ^[5]. Divide the preferred 10,927 m by 7.5 seconds and the implied mean sound speed is about 1,460 m/s ^[19]. So the 18 m disagreement between the five models is a disagreement of about 2.4 m/s in that average ^[20], near 0.16 percent ^[21]. At this level of precision the instrument is not the limit; the temperature and salinity profile through 11 km of water is.

The preferred model is a splice. The upper-ocean XCTD casts from Hakuho-maru's 2023 cruise reach only to about 1,900 m, so the team joined them to a full-depth CTD profile the same vessel collected in 1992, 31 years earlier ^[7][25]. Their stated reason is that the 2023 casts capture upper-ocean conditions closest in time to the survey while the older cast supplies sound speed through the rest of the column ^[7].

Within that one model, the three basins came out at 10,926 m in the west, 10,912 m in the centre and 10,927 m in the east ^[8]. The east exceeds the centre by 15 m ^[22], less than the 18 m spread across models ^[6]. The comparison between basins is the sturdier of the two results, because all three were processed identically. The absolute number is what carries the model uncertainty.

Then there is what a single sounding covers. The EM124's 2 degree by 2 degree beams correspond to a footprint roughly 400 m across in both directions at 11,000 m, under a flat-seafloor approximation ^[10]. A figure like 10,927 m is therefore an estimate over an area, assembled from many soundings, and the study warns that narrow depressions or sharp relief may not be fully represented when the footprint gets that large ^[11].

Against that, the 10,920 m value from the 1984 Takuyo survey ^[2] sits 7 m shallower than the new preferred figure ^[23], comfortably inside the model spread. Modern high-precision measurements of this spot still differ from each other by several metres to more than 10 m ^[3]. The paper's stated question is what controls a depth estimate when the seafloor lies nearly 11,000 m below the surface ^[18], and its answer is that metre-scale comparisons need the sound-speed model, vertical reference, observing conditions and processing method attached to every number ^[9].

Ship handling showed up in the data yield. About 87% of soundings survived editing on the 4-knot east-west lines, against 80% on the north-south lines at the same speed, 84% at 8 knots and 78% at 15 knots ^[13]. Changing heading at the same speed cost 7 percentage points; going from 4 knots to 15 knots cost 9 ^[24]. The north-south lines varied most where the ship crossed rapid changes in seafloor topography ^[14].

The raw EM124 files, processed soundings, bathymetric grids, sound-speed models, sensor configuration and processing workflow are all public ^[16]. Anyone who later obtains better full-depth oceanographic information, or a more precise vertical reference, can reprocess the 2023 observations, and the grids can be checked against pressure measurements from deep-submergence vehicles ^[17].