About a centimetre of tidal range per kilometre of coast is what falls out of the South Taranaki Bight result: close to a metre of difference spread along roughly 90 km of bay [6][1]. The average is the least interesting part of it. Cut that same bay into the 100 m segments the method produces and you have around 900 separate tide estimates [2], and nothing requires the gradient to be evenly distributed among them. The 40 cm between Pegasus Bay and the Rakaia beaches sits inside the surroundings of one city [7], which is a shorter baseline than most regional tide models resolve at all.
Set against the alternatives, the sampling gain is roughly a hundredfold on a 10 km radar altimetry cell [3], and unbounded against a tide gauge, which reports one point and says nothing about the next headland [9]. That is the actual finding here. A model that interpolates tide between gauges is assuming the tide varies smoothly along the coast between them, and the near-metre and the 40 cm are measurements of how badly that assumption can fail at two named places, not a correction factor anyone can carry elsewhere.
It is worth being precise about what got measured. Nobody photographed the height of the sea. The images record where the waterline sits on a sloping beach, and the beach slope converts that horizontal position into a vertical level [4]. The beach is the instrument, which makes it the error term as well: a slope wrong by a tenth returns a level wrong by the same tenth [4]. No vertical accuracy for the per-segment estimates appears in the phys.org account of the study [11], and the claim of beach-scale usefulness rests on that figure being small compared with 40 cm. The mechanism also carries its own coverage limit. A cliff or a vertical seawall gives the waterline nowhere to travel, so it yields no level [5].
What the four-decade Landsat archive buys, if the accuracy holds up, is harmonics [3]. Once the repeating tidal rhythms are fitted at a location, the tide there can be run forward and backward without a gauge ever having stood on that shore [3], which is the part that matters for the stretches of coast the global network never covered [10]. Thomas Monahan of Oxford frames it as two neighbouring regions being safe or flooded in the same storm [8]. The duller version is more useful: coastal flood maps and navigation plans can now be audited at the length scale at which they get applied, on Pacific coasts, across the past forty years [5].
