The use of historical flood information in the English

Document Sample
The use of historical flood information in the English Powered By Docstoc
					                   Hydrological Sciences-Journal-des Sciences Hydrologiques, 47( ! ) February 2002         57

                   The use of historical flood information in the
                   English Midlands to improve risk assessment

                   ALISON WILLIAMS* & DAVID ARCHER
                   Jeremy Benn Associates, Gillow House, Broughton Hall, Skipton, North Yorkshire BD23 3AN,

                   Abstract The Easter 1998 flood was the largest flood event in the gauged record of
                   many basins of the English Midlands. Flood frequency analysis, using such gauged
                   records only, placed the 1998 event at a return period of over 100 years on several
                   basins. However a review of historical (pre-gauged) flooding on some rivers gives a
                   different perspective. Examples are given of the use of historical flood information on
                   the River Leam, the River Wreake at Melton Mowbray, the River Sence (tributary to
                   the River Soar) and the River Frome at Stroud. The cost of acquiring such historical
                   flood data is trivial in comparison to gauged data, but the benefits are demonstrated as
                   significant. In particular, historical flood data provide a better basis for risk assessment
                   and planning on flood plains through revised estimates of flood discharge and depth.
                   Key words chronology; English Midlands;floodrisk; frequency analysis; historicalflooddata;
                   hydrometric data
                   L'utilisation des informations sur les crues historiques dans les
                   Midlands en Angleterre pour améliorer l'évaluation des risques
                   Résumé La crue de Pâques 1998 a été l'événement le plus important par rapport aux
                   chroniques de débits de plusieurs bassins versants des Midlands en Angleterre.
                   L'analyse fréquentielle des débits de crue, sur la base de données jaugées seulement,
                   montre que l'événement de 1998 a une période de retour supérieure à 100 ans sur
                   plusieurs bassins. Cependant, une revue de crues historiques (avant la mise en place
                   des stations de jaugeage) sur quelques rivières donne une perspective différente. On
                   donne des exemples de l'utilisation des informations historiques de crues sur les
                   rivières, Leam, Wreake à Melton Mowbray, Sence (affluent de la rivière Soar), et
                   Frome à Stroud. Le coût de l'acquisition de telles données historiques est très petit par
                   rapport à celui des données jaugées, mais les avantages sont significatifs. En
                   particulier, les données historiques fournissent une meilleure base pour apprécier les
                   risques et pour aménager les plaines inondables au moyen d'une évaluation révisée
                   des débits et des niveaux de crue.
                   Mots clefs chronologie; Midlands anglais; risque de crue; analyse fréquentielle;
                   données historiques de crues; données hydrométriques


In 1992 the UK Department of the Environment issued a guidance note to the local
authorities on "Development and Flood Risk" to ensure that planning decisions took
account of flood risk. To implement this, the Environment Agency has initiated flood
risk surveys which are designed to define flood plains and the effect that future
development will have on the extent and severity of flooding. Specifically, there is a
requirement for the reassessment of the levels and extent of flooding in a flood event
of a 100-year return period. Conventionally such assessment requires first the hydro-
logical estimation of a 100-year flood discharge followed by hydraulic river modelling
to convert discharge to level through a channel reach.

*Now at: Babtie Group, Sheldon Court, Wagon Lane, Coventry Road, Sheldon, Birmingham B26 3DU,
 UK; alison.williams(S>babtie,com.

Open for discussion until I August 2002
68                            Alison Williams & David Archer

     The fundamental basis for flood estimation is the gauged record, which, in
England and Wales, has typically a 25-^4-0 year duration. Annual maximum flood
discharge for a gauged site may be used to assess flood flows of specified return
period. However, the small sample may not be representative of the population of
floods and flood frequency analysis may yield unreliable estimates of the 100-year
flood. For example, assuming a Poisson distribution of inter-event times, there is a
22% chance of a 25-year data set containing a flood of greater than 100-year return
period and a 19% chance that the largest flood does not exceed a return period of
15 years. The limitations of theoretical flood frequency and fitting techniques for a
single site are most evident when the data set contains significant outliers, but such
limitations also exist when no outliers occur in the sample.
     There are two basic approaches available to improve the reliability of flood
estimates—regionalization and the incorporation of historical flood data. Whilst
regionalization has been a component part of flood estimation procedures in the UK
since the publication of the Flood Studies Report (FSR) (NERC 1975, 1993 revised
and extended edition) and updated in the Flood Estimation Handbook (Institute of
Hydrology, 1999), the use of historical data has had a slower take up of application.
Investigations associated with the 1998 flood in particular have shown the value of
historical data in placing large recent events in perspective (Archer, 1999; Harris &
Archer 1999). The Environment Agency now recommends that historical studies be
carried out in association with flood plain surveys prepared to meet the requirements of
Section 105 of the Water Resources Act 1991.


A preliminary chronology of historical floods may be drawn up from a variety of
(a) Information already held at the Environment Agency, including reports, and (in
     some cases) surveyed levels, from predecessor organizations.
(b) A review of information contained in the publications British Rainfall (1861-
     1991) and the Meteorological Magazine (1964-1993) for dates and information on
     heavy rainfall occurrence within a basin, in some cases accompanied by
     descriptions of flooding.
(c) A search of the web site of Chronology of British Hydrological Events (Law et al,
     2000), maintained by the British Hydrological Society for flooding and flood dates
     in the target basin and in neighbouring basins.
     The dates and descriptions then provide the basis for a more thorough search of
local and regional newspapers. In many cases the description of one flood refers back
to one or more previous floods for comparison. These can then be investigated and
added to the chronology.
     In addition many libraries and County Record Offices hold a local record
catalogue under "Floods and Flooding", which may supplement the information
obtained from newspapers and extend it back in time by reference to diaries, Parish
Registers and legal documents such as Quarter Sessions Bridge Accounts. Local
publications of photographic records of towns and villages often include one or more
historical flood photographs.
                The use of historical flood information to improve risk assessment      69

    In this way a comprehensive record of flooding may be obtained on most basins
over a period of 150 years and scattered flood records for a further 100 years or more
(Archer, 1992).


The primary objective in obtaining historical flood data is to find information which
can be incorporated in flood estimation, either in combination with gauged discharge
data from a gauging station, or with respect to the frequency of occurrence of level at
ungauged sites.
    The most useful information is where the level has been observed and recorded
with reasonable precision at a site where the downstream river channel remains
unaltered (and therefore the level/discharge relationship remains the same throughout
the period). Such information may come from a contemporaneous description of the
depth of water in a building, or of the areal extent of flooding on a street. Photographs
add information in confirmation, but it must be recognized that these may have been
taken below flood peak, since, of course, approximately 50% of floods peak during the
night. Other sources are flood stones and flood marks.
     Particular attention is given to the few highest floods, where comparison of
historical events with the highest gauged flood provides a basis for inclusion in flood
frequency estimation (Archer, 1999).
     The following descriptions for the rivers Wreake, Leam, Sence and Frome (Fig. 1)
provide examples of the kind of information available, and the use of the data. In all

               Fig. 1 The investigated rivers in their regional and national context.
70                            Alison Williams & David Archer

cases historical flood data provide additional insights into the occurrence of flooding
on the basin, but the ability to incorporate these data into quantitative flood estimates
varies from basin to basin.


River Wreake

The River Wreake is a right-bank tributary of the River Soar in the main basin of the
River Trent (Fig. 2(a)). The basin area is of moderate relief and covers an area of
450 km2. A gauging station at Syston is located close to the confluence with the River
Soar and has a record extending back to 1967. The town of Melton Mowbray
(population: 25 276) is situated at the confluence of the River Eye, Scalford Brook and
Thorpe Brook, which continue downstream as the River Wreake.
     The flood event in April 1998 on the River Wreake affected many properties and
thousands of hectares of land, prompting a re-evaluation of the hydrology of the
watercourse. At Syston gauging station the peak discharge of 115 m3 s"1 was the
highest in the record, but was only marginally above the second ranked flood of
 106 m V in 1981 and the Rank 3 flood of 104 m3 s"1 in 1975. All other annual
maximum floods were of much smaller magnitude.

               Fig. 2 Basin maps showing towns and gauging stations mentioned in the text:
               (a) River Wreake, (b) River Sence, (c) River Learn, and (d) River Frome.
                 The use of historicalfloodinformation to improve risk assessment                     71

                                                    Reduced Variate
                Fig. 3 Flood frequency plot of annual maxima for the River Wreake at Syston
                illustrating conflicting interpretations of flood magnitude using different distributions
                and fitting methods.

    A number of distributions were fitted (Fig. 3) to provide an initial indication of the
return period of the April 1998 event. The distributions and fitting methods respond
quite differently. The two-parameter extreme value distribution (EV1) gives good fit to
lower events, but treats the larger events as outliers. Different fitting methods result in
quite different return period estimates of the 1998 flood from 75 to 200 years. The
three-parameter generalized extreme value (GEV) distribution fits more closely to the
larger events, indicating a 45-year return period for the 1998 flood. In spite of the
better GEV fit, the choice between these estimates remains subjective in the absence of
a priori knowledge of the population distribution. A further problem arises from the
contrast in the upper reaches of the basin between the extensive flood damage in 1998
and minor flooding in 1975 and 1981, suggesting that the gauged basin, with double
the basin area, is not fully representative of Melton Mowbray.
    A review of the flood history was undertaken to evaluate flooding frequency at
Melton Mowbray and to place the April 1998 event in context. Local newspaper
reports were found dating back to the early nineteenth century. A key site for the
incorporation of historical flood information was the Harborough Hotel in Burton
Street, where flood depths were reported in a number of events. In addition, a board
showing maximum flood levels was maintained close to the entrance and was used for
comparisons in reports of historical floods. For example, "in [August] 1922 it was
7 inches higher than in [July] 1880". Although this board was recently removed in
hotel alterations, photographs enabled a comparison to be made with the April 1998
event showing that the recent flood can be ranked below these two historical events.
Similar comparative levels were obtained from a long-established factory.
    In addition, the mention of specific roads or premises in reports helped to make
comparisons by showing how far flood waters extended from the river. In October
1882, December 1900 and August 1922, reports stated that the floodwaters proceeded
72                             Alison Williams & David Archer

along Burton Street as far as the church and then down Mill Street. The October 1882
report in the Melton Mowbray Mercury states that the flooding was not as severe as
that experienced in July 1880. Similarly, in August 1922, the Melton Mowbray Times
stated that water reached over the railway (built 1847) for the first time. In April 1998
it is known that the water just reached the railway lines.
      It is clear that at least four events since 1800 exceeded the flood of April 1998. Of
these, the August 1922 flood was undoubtedly the greatest. Over the period from 1900,
when more comprehensive flood descriptions are available, the 1998 event was ranked
third, with August 1922 and December 1900 being greater. Over this period, the
Gringorten plotting position formula indicates a 39-year return period for the April
1998 event. From 1800 it is ranked 5, as July 1880 and October 1882 are also greater,
giving a return period of 44 years.

River Senee

The River Sence is a headwater tributary of the River Soar (Fig. 2(b)). It has a gauged
record from 1971, at Blaby from 1971 to 1984, and at South Wigston from 1971 to
1998, and it has a basin area of approximately 135 km2 to its confluence with the River
Soar. The lower reaches of the basin are urbanized and comprise parts of the towns of
Wigston and Blaby, which have become suburbs of Leicester (population: 300 000).
The village of Great Glen is also adjacent to the river.
     Development on the River Sence has generally avoided the active flood plain, but
there are a few properties which have been repeatedly flooded, and road and rail
crossings have inevitably been affected. However, as the scale of flooding has
generally been small in comparison to other rivers in the Leicester area, historical
reference to flooding along the Sence has been patchy and there are believed to be
occasions when roads or property have been flooded but not reported.
     Flood occurrence has been identified as far back as 1596 (Palm Sunday—
March/April), when the Parish Register of St Wistan's at Wistow reported that "the
waters did exceedingly abound" after "a very great snow". In 1618 (4 September), the
Parish Register at St Wistan's again reported that "About six of the morning the waters
so increased that they prevailed to carry the bier about the church and ran over the
communion table.". On no subsequent occasion was the church reported to have been
     Severe flooding was reported in November 1852 (when a railway viaduct
collapsed at South Wigston), and again in July 1875 and July 1880. The Greyhound
Inn is a key site in Great Glen, which was reported flooded in 1880 to a depth of 6 feet.
The remark that this was the greatest flood since 1875 or 1852 suggests that flooding
there was of even greater magnitude at the earlier dates, though not reported.
Photographs show the Greyhound Inn flooded again in August 1922 in November and
December 1954 and July 1968. There were eight occurrences in 150 years, giving a
return period of less than 20 years.
     In the case of the Sence, historical flood information, whilst useful for specific
locations, was insufficiently precise and consistent to be used in conjunction with the
gauged record in flood estimation, or to put the Easter 1998 event of 31 m' s"' into
                The use of historical flood information to improve risk assessment     73

River Leam

The River Leam (Fig. 2(c)) is a tributary of the River Avon with a basin area of
362 km2 to the confluence. It is a predominantly rural lowland basin with its main
urban area in the lower reaches at Leamington Spa, which is vulnerable to flooding.
The gauging station at Eathorpe (basin area 300 knr) has a 25-year gauged record,
with gaps, of annual maxima from 1968 to 1996.
    A review of historical flooding was carried out a few months before the Easter
floods of 1998. Comparison was made with the gauged record in which the Rank 1
event in July 1968 was accorded a discharge of 101 m3 s"1, whilst the Rank 2 event had
a discharge of 52 m J s" , a ratio of 1.95. Application of flood frequency analysis
suggested that the return period of the 1968 event was between 185 and 460 years.
    However, the historical review revealed that several events were clearly greater in
magnitude at Leamington than the flood of 1968.

    1932 (21 May) Surveyed peak levels for this event were available and were 0.67 m
higher than the 1968 flood upstream from an ancient mill weir at Leamington Spa and
0.85 m at Portobello Bridge on the River Avon just upstream from the Leam con-
fluence. Photographs and newspaper reports of flood extent on roads and in properties
confirmed the much higher level than in 1968.

    1900 (30 December) Newspaper descriptions of flood depth and extent in
Leamington Spa indicated a magnitude marginally lower than in 1932 but again
greater than in 1968. At Rock Mill on the River Leam just upstream from the Avon
confluence (but influenced by backwater), it was one and a half inches above the 1932
flood. Further downstream at Evesham, the flood was the highest in the historical
annual maximum series from 1848 to 1935 (FSR Volume 4—NERC, 1993).

     1875 (20 October) Information for the eighteenth and earlier centuries is more
sparse, but the balance of probability is that this flood and the following listed events
were also of greater magnitude than the 1968 flood. A surveyed level at Portobello
Bridge was 0.51 m higher than in 1968. The 1875 flood was reported in 1900 to be
marginally below that flood in Leamington Spa, but 17.5 inches lower on the Avon
just downstream from the Leam confluence. The flood was very widespread and on the
neighbouring larger River Trent, it was the largest flood of the nineteenth century.

     1852 (11 November) The relevant local newspaper for descriptions of this flood
was missing from the archive at Warwickshire Record Office, but floods of 1875 and
1932 were subsequently reported to be the highest since 1852. It was the Rank 3 flood
on the Avon at Evesham after 1900 and 1848. It was again a widespread flood with
serious damage on main stem and headwater tributaries of the neighbouring rivers
Trent, Severn, Thames and Wye.

    1848 (1 October) At Leamington Spa, newspapers reported the flood to be "by far
the highest in remembrance" and at Stratford on Avon "we can positively state the
flood has been the highest that has taken place since the flood that occurred 53 years
ago" (1795). At Evesham, it was the Rank 2 flood after 1900. Descriptions of flood
extent and property affected suggest an event of greater magnitude than 1968.
74                            Alison Williams & David Archer

     1795 (10 February) No specific information was available at Leamington, but it
was higher than 1848 on the Avon just downstream from the Leam confluence. It was
referred to then as "about 50 years since there was one much higher when boats were
made use of to convey parties between New Bridge and the town of Warwick". It was
also said to be the highest on the River Trent in 200 years (NERC, 1975).

   1735 (5 September) This preceded newspaper accounts, but was referred to in a
1795 diary account of W. Harrison of Warwick as 7 inches higher on the Avon at
Warwick than the flood of 1795 (Warwickshire Record Office CR 556/688/1).

     In addition, several events were probably of lower magnitude than 1968, but
definitely greater than the Rank 2 event in the recorded data series, including March
1947, October 1939, April 1920, October 1882, M y 1875 and March 1867.
     As shown by Archer (1999), historical evidence suggests a return period of
approximately 40 years for the 1968 flood, compared with the estimate from the
gauged record alone of over 180 years.
     The event of Easter 1998 provided further evidence of the notable, yet
unexceptional return period of the 1968 event. If the 1998 event, with a peak discharge
of 116 mJ s~', is appended to the gauged annual maximum series, the resulting single-
site estimate of the return period is of the order of 1000 years. However, although it is
probably the largest (or equal with 1932) in the historical record, the inclusion of
historical floods reduces the estimated return period to about 175 years.

River Frome

The River Frome (Fig. 2(d)) has a very permeable Cotswold basin, with underlying
rocks predominantly of Oolitic limestone. It is gauged at Ebley Mill, with a basin area
of 198 km2, and has a record from 1969 onward. A review of historical flooding was
carried out in 1998, though not specifically with respect to the Easter floods, which
were not significant on the Frome basin and did not even achieve annual maximum
     In spite of the limited occurrence of damaging floods during the period of gauged
record, numerous references were found to very severe flooding, which had occurred on
the basin, dating back to 1820. Considerable variation was found in the frequency of
flooding on different tributaries. Reports of flooding were frequent on the Nailsworth
Stream and on the River Frome below the Nailsworth Stream confluence, but infrequent
on the Painswick Stream and absent on the upper Frome above Chalford.
     The most widespread and frequent flooding appears to have resulted from
sustained winter rainfall, which causes a slow build-up of groundwater flow and a very
prolonged recession. Events in November 1875, December 1900, December 1929 and
December 1965 resulted in severe and prolonged flooding from Stroud to Saul, near
the confluence with the Severn, as well as on some tributaries.
    On very rare occasions, severe flooding has also resulted from intense summer
thunderstorms, which have mainly affected the Nailsworth Stream and the River Slad,
with the most severe events occurring in October 1882, July 1907 and August 1931. In
the last of these, the water was reported to have risen 10 feet in 15 min at Nailsworth.
In all these events, the water level receded rapidly and, with lower flood volumes, the
                The use of historical flood information to improve risk assessment     75

effects diminished downstream. Events of this origin and magnitude do not appear in
the gauged record.
    In the case of the River Frome, a modelling approach was used to extend the
gauged record and this resulted in an estimate of the 100-year return period flood of
26.6 m J s"1, which was 23% higher than that obtained from the Ebley Mill gauged
record alone (JBA, 1998).


The rivers investigated illustrate the value of including historical information in flood
estimation, but this information has been applied more successfully to flood estimation
in the cases of the rivers Wreake and Leam than those of the rivers Sence and Frome.
     However, even where historical information cannot be directly incorporated into
flood estimation procedures, it can reveal characteristics of flooding on a basin which
are not evident from a gauged record, even of 30 years duration. Such is the case of the
infrequent summer floods on the River Frome.
     Conventional hydrometric data are acquired at considerable expense through the
operation of streamflow networks. A significant component of the benefit of such
networks is perceived in providing flood information relevant to flood design and
alleviation. Significant additional benefit is achieved with the use of historical data,
which can be acquired at little expense compared with standard streamflow networks.
A survey for a typical basin, such as described above, can be carried out in
approximately one week. There is therefore a strong economic justification for a
national programme of collation of such data.
     Incorporation of historical data into design flood estimation increases the
reliability of flood plain mapping and ultimately provides a better basis for planning
decisions on flood plains. This may be done either by incorporation of historical
information in the gauged flood record for flood frequency analysis, or by direct
reference to the frequency of historical levels at a site.
     However, it should be borne in mind that the methodology does have a number of
pitfalls which hamper its use. "Quiet periods" in the flood record, such as between
1618 and 1852 for the River Sence, may not correspond to periods lacking flood
events. The analysis is dependant on the events being newsworthy at the time. Events
will generally only be reported in the press if they have affected human life or
property. In addition, other news may be deemed more important, so flood events may
not be reported at all. During the Second World War, there was little news in the local
papers other than the national news and the effects of the war on the local community.
     Changes in land use in the catchment will affect the river regime and should be
borne in mind when undertaking historical reviews. Over a period of 200 years, the
behaviour and course of a river may change considerably. Research should be
undertaken to ensure that all years in the period of record are comparable.
     Using solely level data to compile a chronology of flood events may be
problematic at some locations. The relationship between stage and discharge at any
river section is unlikely to be stable, due to natural erosion and deposition effects and
changes in the river cross-section by human interference. Most sites where flooding
becomes an issue are urbanized areas where the effects of channel and flood plain
modifications are likely to be greatest.
76                                       Alison Williams & David Archer

    Historical records may provide a more realistic assessment of the risk of extreme
flooding than short periods of gauged flow, which may be both unrepresentative of the
flood population and of comparative severity at other locations within the basin. The
application of historical records should be undertaken with care and in relation to
catchment changes.

Acknowledgements The flood histories were prepared by Jeremy Benn Associates as
part of broader flood studies on behalf of the Environment Agency (Midlands Region),
with whose permission this paper is published. The support of Allison Shaw and
Rachel Spence is particularly acknowledged.

Archer, D. R. ( 1992) Land of Singing Waters: Rivers and Great Floods ofNorlhmnbria. Spredden Press, Stocksfield, UK.
Archer, D. R. (1999) Practical application of historical Hood information to flood estimation. In: Hydrologica! Extremes:
       Understanding, Predicting, Mitigating (ed bv L. Gottschalk, J.-C. Olivrv, D. Reed & D. Rosbjere) (Proc.
       Birmingham Symp., July 1999), 191-199. IAHS Publ. no. 255.
Harris, E. & Archer, D. R. (1999) Floodplain mapping of the River Learn and River Itchen. J. Chartered Insln Wat.
       Environ. Manage. 13(4), 280-285.
Institute of Hvdroloev (1999) Flood Estimation Handbook (5 volumes). Institute of Hvdrologv, Wallingford, Oxfordshire,
JBA (Jeremy Benn Associates) (1998) Reassessment of the River Frome hydrology. Unpublished report for the UK
       Environment Agency.
Law, F. M., Black, A. R., Scarrott, R. M. J. & Miller, J. B. Chronology of British Hydrological Events.
NERC (1975, 1993 revised and extended edition) Flood Studies Report (5 volumes) Natural Environment Research
       Council, London, UK.

Received 27 September 2000; accepted 27 August 2001