Okay, enough dinosaurs. Back to earthquakes. This is a story from a book called Kojishin 古地震 ("Earthquakes of antiquity"), edited by Hagiwara Takahiro 萩原尊禮, which combines earth science and historiography to awesome effect.

So, in Taihō I (701 CE) there was an earthquake on the north coast of central Japan. This is uncontroversial. More controversial is the rumor that this earthquake sank an entire island, Atlantis-style. This is what the Tango fūdoki zanketsu 丹後風土記残欠 ("Remnants of the Tango fudoki"), the main source for this rumor, has to say about the matter (sample source for text, and Kojishin also quotes this part):

凡海郷者往昔去此田造郷万代浜四十三里（三字虫食い）三拾五里二歩四面皆属海壱之大島也所以称其凡海者故老伝曰往昔治天下当大穴（七字虫食い）到坐于此地之時引集海中（三字虫食い）小島（三字虫食い）凡枯以成壱島故云凡海矣（三字虫食い）宝元年三月巳亥地震三日不已此郷一夜蒼（四字虫食い）海漸纔郷中之高山二峯与立神岩出海上今号常世島亦俗称男島女島毎島有神祠所祭者天火明神与日子郎女神也

Ōshiama 凡海 in ancient times was 43 ri from Bandai Beach 万代浜 in Takuri 田造 [now within Maizuru, Kyoto] [...] 35 ri and 2 bu [...] a single large island surrounded by the ocean on all sides. The name "凡海" has been passed down from long ago. In ancient times, when Ōnamochi-no-Mikoto 大穴持命 and Sukunabiko-no-Mikoto 少彦名命 arrived here to rule the land, they gathered together all the small islands in the ocean [...] made them into one island. This is why it is called "凡海" [sort of "entire ocean"]. On the day of the Yin Earth Pig (己亥) [the 26th], in the third month of Taihō I, an earthquake began and did not stop for three days. This land was lost to the sea in one night [...] Finally all that was left above water were the two peaks of a tall mountain that had stood in the middle of the land and a rock on which stood a shrine. The peaks are now called "Tokoyo no shima" 常世島, or by the common people "Ojima" 男島 [also written 雄島, either way "Man island", known as Kanmurijima, today] and "Mejima" 女島 [also written 雌島]. There is a shrine on the islands, dedicated to Ame-no-Hoakari-no-Kami 天火明神 and Hikoro[?]-no-Megami 日子郎女神 [apparently identified with Hoyahime-no-Mikoto].

Terrifying stuff, but not true in the slightest. Yamamoto Takeo 山本武夫 (the individual author credited with this chapter in Kojishin) presents a range of historiographical evidence suggesting that the events described did not take place.

First, he notes that the Shoku Nihongi 続日本紀, completed in 797, doesn't mention anything of the sort in its entry on the earthquake:

丹波国地震三日。

Earthquake in Tamba Province, three days.

Brevity is a well-known feature of these chronicles, but you would think that "half of the province" sinks would be recorded if it had indeed happened.

He then turns his attention to the Tango fudoki zanketsu itself, arguing persuasively that it is at least in part an Edo-period forgery. One datum probably of interest to No-sword-reading types is the presence of the place name 田造, "Takuri," in the text quoted above. This corresponds to 田辺, "Tanabe," in (verified) older texts. How do we know? Because the earliest known instance of the 田造 spelling is a write-o introduced in the first printed edition of the Wamyō ruijushō 和名類聚抄, published in 1617. Ergo, it seems likely that the Tango fudoki zanketsu, or at least the part that uses 田造, was written in the 17th century by someone referring to a post-1617 edition of the Wamyō ruijushō.

Okay, so it looks like an Edo-period forgery. That doesn't mean that the information is wrong. Maybe the sunken island story slipped out of the Shoku Nihongi during the editorial process. Maybe the place-name issues ("Tango Province" wasn't an official province until 713 CE; there are places called 凡海 in the area that are part of the mainland) are also just garbled transmission. The key question to my mind is, did an island in fact sink into the sea?

The answer is, again, simple: it did not. People and paper lie, but broken earth abides, and the second half of the Kojishin chapter on the Taihō I earthquake relates an expedition to the area by a team of geologists (the authors, basically) to debunk the lost-island story.

First of all, the water surrounding the two islands that are alleged to be all that remains of a sunken super-island is 60-70 meters deep. This means that the super-island would need to have sunk 60 meters in a single night. This is simply not credible. Even a 10-meter shift as a result of an earthquake is a rare, dramatic event, and 20 meters is basically unheard of. Furthermore, the islands are at the mouth of a bay, and if the surrounding sea floor were 60 meters higher the result would not be an island but a peninsula.

Okay, so it didn't happen overnight. Maybe it took place over a longer period of time, with the earthquake still as the ultimate cause? Again, no. Erosion doesn't work fast enough to hollow out a sixty-meter deep seabed in the timespan required, even if given a slight head-start by an earthquake, and there's no sign of any rockslides or anything else that would change the shape of the seabed fast enough either.

Science always wins.