Japanese people are enigmatic, and so are their earthquakes. Seismologists all around the world are endeavouring to understand the characteristics of earthquake activity in the Japanese archipelago, but it is not a simple task. Deciphering the regional pattern of seismicity is just as great a challenge to the scientific community as the underwriting of Japanese earthquake risk is a challenge to insurers.

The earth's complex pattern of seismicity emerges as a consequence of the confluence of four of the major tectonic plates that cover the globe. Japan sits uncomfortably astride one of the most significant cross-roads of the earthquake world. Apart from the major offshore sources of seismicity, the Nankai and Sagami Troughs, there are clusters of major on-land faults, many of which are capable of generating destructive earthquakes.

In the past, a massive resource invested in seismic monitoring, alongside the most technically sophisticated and lavish efforts at earthquake prediction, have given Japanese seismologists an air of self-confidence in their internal capability to deal with the earthquake threat. But in assessing this insidious peril, a seismologist has to be as vigilant and as far-sighted as a military commander, and should be prepared for the unexpected.

When the Japanese army crossed the Johor strait from Malaya to Singapore during the Second World War, the British guns were facing out to sea, in entirely the wrong direction. The bearing of attack came as a complete surprise, and the city was soon surrendered. A similar tale can be told of the devastating earthquake which struck central Japan just before dawn on 17 January 1995. It killed more than 5,000 people. The Japanese were astonished that the ruptured fault managed to reach as far as downtown Kobe.

How could this have happened? Japanese scientists naturally knew of the potential for fault rupture, but had not reckoned on a rupture crossing from the nearby neighbouring island, where it began. Needless to say, had the fault behaved as they had anticipated, the earthquake loss would have been minor. Instead it was catastrophic. But the fact that a large, damaging earthquake had occurred in the region four centuries earlier, in 1596, only raised more awkward questions about the apparent lack of Japanese seismological preparedness. As is evident with faults which exhibit new rupture behaviour, preparedness involves a degree of lateral thinking. It is the only way to contemplate sequences of events which may have no historical precedent.

Bad though an earthquake disaster might be, it is not terribly difficult to conceive of ways in which the losses might have been worse. In the 1995 Kobe earthquake, the office of the water authority was badly damaged. At the early hour that the earthquake occurred, the office was, fortunately, unoccupied, but had the earthquake occurred during regular office hours, casualties among key water officials might well have resulted. That would have set back seriously the restoration of the city's water supplies. Insurers, as well as Japanese citizens, may be relieved to know that action to minimise such potential threats is a task that civic authorities are addressing, through integrated disaster risk management programmes.

1995 was a particularly disastrous year for Japan, but 2000 was also punctuated by significant natural hazard events. There were the eruptions of Mount Usu in Hokkaido, northern Japan, and on the island of Miyake, quite a long way offshore from Tokyo. Fortunately losses from these eruptions were slight, although 4,000 inhabitants of Miyake were forcibly evacuated to Tokyo. More tragically, ten people died and 500 buildings were damaged by flooding in the Tokai district, arising from a September typhoon which generated rainfall in excess of the design capacity.

This event has been a turning point in Japanese disaster management, in so far as flood management is now recognised as requiring regional coordination, and the designation of areas at flood risk is now mandatory. A month after the September flood came the earthquake in the Tottori prefecture, which injured several hundred and damaged 20,000 buildings. The earthquake threat to Japan has been further kept vivid in the public consciousness by an earthquake on 24 March 2001, which caused some fatalities, as well as damage to 40,000 buildings in the Hiroshima, Yamaguchi and Ehime prefectures.

Damaging though these two recent earthquakes were, the prefectures that earthquake insurers are most concerned about are those in the Tokyo metropolitan area. But is Tokyo prepared? Ever since 1992, Tokyo has been on red-alert for a major earthquake. Several decades ago, the Japanese professor Hiroshi Kawasumi heightened earthquake awareness by suggesting an approximate 69-year rule-of-thumb for damaging earthquakes in the Tokyo area, based on a cursory historical survey going back to the era when Tokyo was called Edo. These past events varied in epicentre and magnitude, and their recurrence today would result in quite a wide range of potential economic losses. Thus the 1855 Edo earthquake, if it were to happen now, is estimated by RMS to cause a loss of about 20% of that of the 1923 great Kanto earthquake.

Sixty-nine years on from the great Tokyo earthquake of 1923 was the year 1992, which, fortunately, was not the ominous year it might have been had Kawasumi's rule been cyclically precise. And while few seismological phenomena are really rhythmic, because earthquake processes are dynamically chaotic, time never stops ticking in the build-up of seismic strain.

Fire, of course, was the great curse of the Kanto earthquake of 1923. The flames may have been fanned by typhoon winds, but part of the fire tragedy was man-made. As far back as the 18th century, a fire protection buffer zone had been created in the old city of Edo. However, this fire break was conveniently, if recklessly, built over during the later urban development of Tokyo. Furthermore, voluntary fire brigades had been effective in limiting fire disasters in Edo. They were a nice example of human self-organisation to deal with common societal risks. Both the effectiveness of fire suppression and the presence of fire breaks are key elements of simulation models of fire loss following an earthquake.

Fire prevention cannot be separated from urban planning, and integrated urban disaster risk management is becoming a watchword in Japan, with an institute in Kyoto focusing on strategies for improvement. Not just civic authorities, but corporations too are appreciating the need to address the issue of disaster risk management, and especially to protect the critical nerve centres of their businesses. Efforts to mitigate consequential business interruption are being made, through decentralization of offices and industrial facilities away from the most vulnerable parts of metropolitan Tokyo.

Notwithstanding the increasing local efforts at earthquake loss reduction around Tokyo, large Japanese insurers are prudent in attempting to diversify their book of business away from excess concentration on Tokyo earthquake risk. One efficient way of achieving this objective is to swap Japanese earthquake risk with Californian earthquake risk. In the jargon of the theory of competitive games, this is a ‘win-win' strategy, because both parties to the swap gain a benefit, even though they may otherwise be business competitors. Tokio Marine and Fire Insurance Company has recently swapped $150m of Japanese earthquake risk with Californian earthquake risk held by Swiss Re. Transactions such as this can be engineered so that the risk characteristics of the swap will be similar to both parties. But, inevitably, the surprise element of earthquake risk will affect the loss volatility.

The Kobe earthquake surprised the Japanese in various ways, but it is not unusual for earthquakes to pack a few bombshells; the Northridge, California, earthquake, which occurred exactly one year before the Kobe quake, caused several. Post-earthquake surveys in California and elsewhere have been conducted for the very purpose of learning from earthquakes' unwelcome surprises. At Northridge, the fault which ruptured was not mapped as active by the US Geological Survey. The pattern of ground shaking showed notable directional variations, which were not anticipated. The failure of steel welds resulted in much more damage to steel-framed buildings than structural engineers had envisaged.

Many Tokyo citizens may have little option over their earthquake exposure, but a risk-averse Tokyo resident of reasonable financial means would choose to live in an apartment of recent construction. With improvements to the Japanese building code, the vulnerability to earthquake shaking deteriorates notably with age of construction.

Is a well-to-do Tokyo resident living in a modern apartment safer than his counterpart in Los Angeles? After all, an earthquake in Tokyo might occur while a resident is working in an old office, walking in a narrow, fire-prone street, or travelling in a crowded train. But then the nightmare scenario for catastrophe insurers is a rupture of the Newport-Inglewood fault, which cuts right through the Los Angeles area. Which is more volatile, the estimate of the property/casualty loss in Tokyo or Los Angeles? The answer to this question may eventually be reduced to fine issues such as fire susceptibility. The spacious boulevards of Los Angeles have no parallel in modern Tokyo, although, centuries earlier, the wise authorities in Edo planned broadways for their value in fire mitigation.

The surprise element in earthquake loss is one of the elusive factors of earthquake underwriting that insurers would do well to appreciate, even if it is not reflected explicitly in premium rates. A surprise can arise from some hitherto unknown or poorly understood seismological phenomena, or from an exceptional sequence of contingent events, but underwriters inclined to a jaundiced view of current premium rates can always take comfort from knowing that surprises over earthquake loss occurrence need not necessarily be negative. New initiatives by Japanese authorities in disaster risk management may prove to be successful in mitigating future losses. Reinsurance underwriters should wish them luck.