Sunday, 20 March 2011

Nuclear power and earthquake hazard in the UK


The UK is not a country generally associated in the public mind with earthquakes. However, while the UK is nowhere near in the same league as high seismicity areas such as California and Japan, it nevertheless has a moderate rate of seismicity, sufficiently high to pose a potential hazard to sensitive installations such as nuclear power stations, dams and chemical plants.
The distribution of British earthquakes in space
The study of British earthquakes has in the past been somewhat neglected compared to some other countries, not necessarily those with more active seismicity.
Up to the 1970s, the most recent publication attempting to survey the whole history of British earthquakes was still Davison's catalogue of 1924 even though this was now 50 years out of date.

Modern instrumental monitoring of British earthquakes began around 1970 with the establishment of LOWNET by the Global Seismology Group of BGS (then IGS) which has subsequently expanded to the present country-wide monitoring network, supported by a customer group led by the Department of the Environment. This led to routine macroseismic surveying of British earthquakes from 1974 onwards, and the beginnings of reinvestigation of historical seismicity at about the same time by Roy Lilwall.

In the early 1980s, the expansion of the nuclear power programme in the UK led to increased activity in revaluating historical seismicity both from macroseismic and instrumental records, and major studies were made by several investigators independently, including BGS, Imperial College London, and private consultancies. All this work was combined and synthesised in the early 1990s to make a consistent, numerate earthquake catalogue for the UK, which was published by BGS in 1994.

A map of earthquakes in the UK

Above we see a map of earthquakes in the UK, taken from the BGS catalogue. It is clear from this map that the spatial distribution of earthquakes is neither uniform nor random. For example, in Scotland most earthquakes are concentrated on the west coast, between Ullapool and Dunoon, with the addition of centres of activity near the Great Glen at Inverness and Glen Spean, and a small area around Comrie, Perthshire, and extending south to Stirling and Glasgow. The Outer Hebrides, the extreme north and most of the east of Scotland are virtually devoid of earthquakes. For the north-west of Scotland the absence of early written records, the small population, and the recent lack of recording instruments means that there may be a data gap; for instance, there are indications that an earthquake occurred in 1925, possibly near Ullapool, with magnitude probably about 3½ ML, for which there are no first-hand reports. However, many other parts of Scotland, especially south of the Highland line, are quite well-documented, at least since 1600, and therefore the lack of earthquakes is genuine.

Further south a similar irregularity is seen. If one draws a quadrilateral from Penzance to Holyhead to Carlisle to Doncaster, most English and Welsh earthquakes will be included within it. The northeast of England seems to be very quiet; almost aseismic. The southeast has a higher rate of activity, with a number of earthquakes which seem to be "one-off" occurrences. The most notable example of these is the 1884 Colchester earthquake, a magnitude 4.6 ML event which was the most damaging British earthquake in at least the last 400 years, and yet which occurred in an area (Essex) otherwise more or less devoid of earthquakes from the earliest historical period up to the present day.
There are also important centres of activity near Chichester and Dover. The former produced a swarm-like series of small, high-intensity earthquakes in the 1830s and was active again in 1963 and 1970.

Offshore, there is significant activity in the English Channel and off the coast of Humberside. Because only the larger events in these places are likely to be felt onshore, the catalogue in the pre-instrumental period is probably under-representative of the true rate of earthquake activity in these zones. Even after the introduction of seismometers, offshore earthquakes may still have gone unnoticed on account of the distance to the nearest instruments. The Central Grabens of the North Sea are now known to be active features, only because of the improvements in instrumental monitoring over the last fifteen years.
Certain centres can be identified as showing typical patterns of activity. For example, the Caernarvon area of north-west Wales is one of the most seismically active places in the whole UK. Both large and small earthquakes, usually accompanied by many aftershocks, occur at regular intervals. The most recent of these larger events was the earthquake of 17 July 1984 (5.4 ML), which was one of the largest ever UK earthquakes to have an epicentre on land and had a very protracted aftershock sequence. Two further felt earthquakes have occurred there since, on 29 July 1992 (3.5 ML) and 10 February 1994 (2.9 ML). It is tempting to ascribe several early earthquakes of unknown epicentre (eg that of 20 February 1247) to this area just because it seems to be such a favoured site for large earthquakes.

In South Wales, on the other hand, although a line of epicentres of significant events can be traced from Pembroke (an earthquake in 1892) to Newport (active in 1974), only the Swansea area shows consistent recurrence, with significant earthquakes occurring in 1727, 1775, 1832, 1868 and 1906. (Given this periodicity it may be that a further earthquake in this area is due in the near future.) The Hereford-Shropshire area has also produced large earthquakes in 1863, 1896, 1926 and 1990, but none of these share a common epicentre.

The area of the Dover Straits is particularly significant because of the occurrence there of two of the largest British earthquakes in 1382 and 1580 (both of magnitude about 5¾ ML). Since 1580 the only earthquakes there have been much smaller, raising the question of whether there is a danger of another 1580-style earthquake in the near future. The area may be structurally continuous with a zone of activity running east through Belgium, in which case it could be argued that stress in this area since 1580 has been released further east. This does not rule out another 1580-type earthquake in the future, but it is impossible to estimate how soon it might occur.

In the north of England seismic activity occurs over a more or less continuous area from Leicester to Carlisle. The most prominent centres of repeating activity here are the upper end of Wensleydale (with significant earthquakes in 1768, 1780, 1871, 1933 and 1970) and to a lesser extent the Skipton area.

What is remarkable is the lack of correlation between this pattern and the structural geology of the UK.

Tectonic sketch map of the UK

This map shows the major crustal subdivisions in the UK. The boundaries between areas of moderate or high seismicity and areas of very low seismicity do not correspond to any major structural feature; for instance the sharp dividing line running SE from Inverness. And the major boundaries shown above are not clearly reflected in the pattern of seismicity either as dividing lines between zones of differing rates of seismicity nor as lineations marked by earthquakes. It seems likely that the pattern of seismicity may be influenced by the distribution of ice during the last glaciation - certainly for Scotland this appears to be the case.

The distribution of British earthquakes in time

log N = a - b M  where N is the number of earthquakes per year exceeding a given magnitude M. The constant a reflects the absolute level of seismicity in an area, and the value of b has generally been found to be consistently close to 1.0.

Magnitude frequency plot for the UK

 It has long been realised that larger earthquakes occur less frequently than smaller earthquakes, the relationship being exponential, ie roughly ten times as many earthquakes larger than 4 ML occur in a particular time period than earthquakes larger than magnitude 5 ML. This can be expressed by the Gutenberg-Richter formula

This holds true for the UK. The graph above shows an analysis for the area 10o W to 2o E and 49o N to 59o N. This deliberately excludes the northern North Sea area which is of high seismicity and completely under-represented in the catalogue before 1970 because of the impossibility of detecting smaller events in this area before that date. A least-squares regression to this data gives the relationship log N = 3.82 - 1.03 M

Also shown is an alternative doubly-truncated exponential model which gives a curved fit ot the data at the higher magnitude end.

One can therefore draw the following conclusions about average recurrence - the UK may expect:

an earthquake of 3.7 ML or larger every 1 year
an earthquake of 4.7 ML or larger every 10 years
an earthquake of 5.6 ML or larger every 100 years.

Seismic hazard calculations

Seismic hazard calculations in regions of low seismicity, such as the UK, are generally based on probabilistic methodology. Probabilistic seismic hazard assessment (PSHA) uses a combination of interpreted geological and seismological data to calculate the probability that a certain level of ground motion will be exceeded, or not exceeded, in a given period of time.

This methodology can be divided into three principal components as follows:

(i) Definition of a set of seismic source zones which define the geographical variation of earthquake activity. These source zones are based on the distribution of observed seismic activity together with geological and tectonic factors and represent areas where the seismicity is assumed to be homogenous; ie there is an equal chance that a given earthquake will occur at any point in the zone.

(ii) An understanding of earthquake recurrence with respect to earthquake magnitude, as described in the previous section.

(iii) An attenuation relationship is required which defines what ground motion should be expected at Location A due to an earthquake of known magnitude at Location B. The rate at which the strength of shaking decreases with distance from an earthquake's epicentre varies regionally and has to be calculated or estimated. Peak ground acceleration (pga) is the measure of earthquake shaking most used by engineers in this country. However, it has two disadvantages - firstly, the attenuation of pga in the UK is very poorly known, and secondly, pga is actually not a particularly good measure of the actual expectation of damage. A useful alternative is intensity, which is an expression of ground shaking in terms of its effects. The attenuation of intensity in the UK is very well documented, and intensity is directly proportional to damage, making it a very meaningful parameter.

The intensity attenuation model used here is expressed by the formula

I = 3.32 + 1.44 ML - 3.34 log R,  where ML is local magnitude and R is hypocentral distance in kilometres.

Seismic hazard studies in the UK in the past have been mostly single-site studies for particular installations. The first attempt to look at hazard for the UK as a whole using the PSHA methodology was conducted by Ove Arup around 1991. This study calculated hazard at eleven representative sites in the UK. Following this, a study to produce contour maps of UK seismic hazard was commissioned by the Department of Trade and Industry, and was carried out by BGS and AEA Technology. In this study the computer code SUNMIC was used, which allows a "logic tree" model to be applied to the hazard, by which uncertainty in input parameters can be modelled by the inclusion of multiple choices each with a weighting value).

Seismic hazard results

Here is a sample hazard map of the UK, based on the study made for the DTI.

Intensity hazard map for the UK

The map shows intensities that are 90% likely not to be exceeded in 50 years - equivalent to a return period of 475 years. For guidance, a simplified equivalence of the intensity values (on the European Macroseismic Scale) is as follows:

3 - Felt by few
4 - Felt by many indoors, windows and doors rattle
5 - Felt by most indoors, small objects fall over
6 - People run out in alarm, slight damage to buildings (plaster cracks)
7 - Moderate damage to buildings (chimneys fall, cracks in walls)

As might be expected, the areas of highest hazard parallel the areas where earthquakes have been most common in the past, but particularly those places where repeated earthquake activity has been highly localised - this localisation has a pronounced effect on the hazard calculations compared to areas where the seismicity, while high, is more diffuse and less repetitive. The zones where hazard is higher than average encompass the W Highlands of Scotland, an arcuate zone running from Carlisle to Pembroke, NW Wales and W Cornwall. The places in the UK with lowest seismic hazard are Northern Ireland (especially the western counties) and outlying parts of Scotland, including the Orkneys and Outer Hebrides.

The actual values of hazard are not particularly high, since the predicted intensity for the higher zones is only 6 EMS. In other words, even in areas of relatively high exposure to earthquakes in the UK, if a facility has a life of 50 years there is only a 10% chance that it will experience shaking equivalent to intensity 6. Moving briefly from hazard to risk, if we take as a guideline that probably less than 5% of buildings of normal construction (eg conventional brick houses) will be damaged in a place when the intensity there is 6, the probability of damage for a single house in 50 years is therefore less than 0.5%.

Roger Musson

25 June 2003