[an error occurred while processing this directive] DEADLIEST EVER EARTHQUAKES [an error occurred while processing this directive]



A short discussion by
Roger Musson

Probably most people are familiar with the USGS NEIC list of deadliest ever earthquakes at http://wwwneic.cr.usgs.gov/neis/eqlists/eqsmosde.html; the top ten read as follows (slightly edited) -

1) January 23, 1556 China, Shansi 830,000

2) July 27, 1976 China, Tangshan 255,000* 8.0 Estimated death toll as high as 655,000.

3) August 9, 1138 Syria, Aleppo 230,000

4) May 22, 1927 China, near Xining 200,000 8.3

5) December 22, 856+ Iran, Damghan 200,000

6) December 16, 1920 China, Gansu 200,000 8.6

7) March 23, 893+ Iran, Ardabil 150,000

8) September 1, 1923 Japan, Kwanto 143,000 8.3

9) October 5, 1948 USSR 110,000 7.3 (Turkmenistan, Ashgabat)

10) December 28, 1908 Italy, Messina 70,000-100,000 7.5

I had occasion to look more closely at this list recently and was surprised to see how much of it crumbles on closer inspection. I circulated my comments to USENET and relevant mailing lists because I hoped people would point out anything I had overlooked, but I had no response. However, some people have asked for access to those comments, hence this web page.

What follows is the result of only a very cursory review using material easily to hand. Some key sources are hard to get at.

I'll start by comments on each of the events above:

1) No problem. Note that 830,000 is a minimum figure (recovered bodies). Shaanxi is the preferred spelling these days.

2) No problem except that the estimate of 655,000 is pretty well discredited now. The figure 242,700 also appears in sources, and may be more accurate.

3) The date of this earthquake should be 15 October 1138 not 9 August 1138. Translating dates from early Arabic sources is fraught with difficulty. The death toll is given by Ambraseys (BSSA 1962) as 130,000 rather than 230,000. Which is correct?

4) The epicentre was not near Xining; the quake wasn't even felt in Xining to any extent. The epicentre was in Gansu province, near Gulang. I cannot find any evidence that the death toll was so high, and this figure seems to come from Båth (1979) which is not exactly a primary source. The actual death toll seems to have been about 41,000.

5) No problem. Damghan is the town, Qumis the region.

6) The epicentre is in Ningxia, not Gansu.

7) Totally fake. This is a confusion with the 24 Dec 893 Dvin earthquake which killed 30,000.

8) No problem, but I confess to not having checked this one. Kanto is the preferred spelling now.

9) Where does the death toll for this event come from? The Kondorskaya-Shebalin catalogue gives "many tens of thousands" dead AND injured; Ambraseys (QJEG 1978) gives 19,800 dead.

10) The best estimate of the death toll seems to be about 80,000.

So that leaves us with:

1) 1556 Shaanxi 830,000

2) 1976 Tangshan 255,000

3) 856 Qumis 200,000

4) 1920 Ningxia 200,000

5) 1923 Kanto 143,000

6) 1138 Aleppo 130,000

7) 1908 Messina 80,000

So we need to bump up the reserves. The next three:

11) September, 1290 China, Chihli 100,000

12) November, 1667 Caucasia, Shemakha 80,000

13) November 18, 1727 Iran, Tabriz 77,000

Comments on these:

11) The date was 27 September, the epicentre was onshore, near Ningcheng in Liaoning province, not in the Gulf of Chihli (Bohai). Death toll was 7,220.

12) Correct date seems to be 4 January 1668 (see Ambraseys and Adams, Eos 1989). The death toll is OK. The epicentre is in what is now Azerbaijan.

13) Fake. It is a confusion with the 26 April 1721 event which killed over 40,000.

That puts the 1668 event in at number 8. Next two:

14) November 1, 1755 Portugal, Lisbon 70,000 8.7

15) December 25, 1932 China, Gansu 70,000 7.6


14) Sources readily to hand give very widely varying figures for the death toll, but 70,000 is credible.

15) I can find no reliable source indicating that the death toll from this quake even hit four figures, never mind five.

So 1755 occupies the number 9 slot and we need another one. Next quake:

16) May 31, 1970 Peru 66,000 7.8

Estimates of fatalities vary from 50,000 to 67,000, so I'll accept the figure above. So we now have as top ten:

1) 1556 Shaanxi 830,000

2) 1976 Tangshan 255,000

3) 856 Qumis 200,000

4) 1920 Ningxia 200,000

5) 1923 Kanto 143,000

6) 1138 Aleppo 130,000

7) 1908 Messina 80,000

8) 1668 Azerbaijan 80,000

9) 1755 Lisbon 70,000

10) 1970 Peru 66,000

But having weeded out the events that shouldn't have been on the list, what about the ones that should have been and weren't? The old chestnut of the 1737 Calcutta quake (300,000 dead) is thankfully missing (wasn't an earthquake). Also, the 1202 earthquake in the Levant which was supposed to have killed more than a million is not there either, since latest estimates are about 30,000 for this event. But there are some others that need to be considered.

Candidate #1: 13 Dec 115 Syria. Destroyed Antioch and said to have killed 260,000. I've no idea whether this death toll is grossly inflated; it may well be. My source is Willis (BSSA 1928) and the death toll is repeated by Tiedemann (Swiss Re 1991); Willis's source is Arvantakis (1903) which I can't check. The earthquake appears in Mallet's catalogue with only the year, and no description of casualties. If the death toll is true, this is the new number two.

Candidate #2: 25 Sep 1139 Azerbaijan - killed 230,000 according to Ambraseys & Adams, op. cit.

Candidate #3: 17 Sep 1303 Shanxi - I don't know how this one was missed. Estimates of fatalities go from 100,000 to 400,000, with 200,000 the preferred figure.

And it wouldn't surprise me if I've missed something, if I can pick these three up just in an afternoon spent poking around my office. Another Internet list compiled by Fujita (http://www.glg.msu.edu/geophys/fujita/intensity.html) has some more candidates:

1876 Andaman Islands - 215,000; I don't believe it. It doesn't appear in any of the Indian catalogues that I've seen, and I'm suspicious as to why it has no day and month if it occurred as recently as 1876. I suspect a typhoon. If anyone knows better, please say.

1779 Tabriz - 100,000 to 200,000; this is an error for the 8 January 1780 Tabriz earthquake which killed about 50,000.

29 Nov 533 Syria-Turkey - 130,000; this is our friend from 1138 rendered in the Moslem dating system; note the same casualty number.

1201 Aegean Sea - 100,000; most vaguely dated events with big casualties around this period are confusions with the 20 May 1202 event and I think this is no exception.

So none of these four go through, and that gives us as top ten the following list:

1) 1556 Shaanxi 830,000

2) 115 Syria 260,000

3) 1976 Tangshan 255,000

4) 1139 Azerbaijan 230,000

5) 1303 Shanxi 200,000

6) 1920 Ningxia 200,000

7) 856 Qumis 200,000

8) 1923 Kanto 143,000

9) 1138 Aleppo 130,000

10) 1908 Messina 80,000

However, if 115 is exaggerated, it drops off and 1668 Azerbaijan comes in at number ten. I've reordered the events that are equal at 200,000 to reflect my best guess as to which is most likely to be underestimated.

My question: what am I missing and what is wrong with the above analysis? Of course, it's clear that estimation of death tolls for many early earthquakes is educated guesswork, considering how uncertain the death tolls are even for some modern events. There are several early events I can think of that must have had very high death tolls but there is simply no way of telling just how many lost their lives. Equally, it is easy to find early writers who clearly inflated the number of deaths in their histories, the 1202 event being one which has been discovered. I'm concerned that the 115 event in Syria may be another such in which the evidence that shows the casualties to have been exaggerated is simply lacking. I also find it suspicious that, if one were to accept the USGS death toll for the 1138 Aleppo event of 230,000, then one would have two earthquakes with deaths of 230,000 one year apart, one in Syria in 1138 and one in Azerbaijan in 1139, and I'm wondering if these two events have been confused in some way. Also, I may be finding fault where none exists; perhaps there is a reliable source that confirms 110,000 dead in Ashkabad in 1948, which I didn't find, and if someone could direct me to it I would be most grateful.

There will also be those who say that this sort of ranking exercise is not science, and inherently futile. However, people like lists, and they like to know the biggest of everything, and I think we owe it to the public who fund our work through their taxes to give them the best answers we can to these questions.

Roger Musson
7 March 2001