The
relationship between Lamb weather types and long-term changes in flood
frequency, River Eden, UK
Abstract:
Research has found that both flood magnitude and
frequency in the UK
may have increased over the last five decades. However, evaluating whether or
not this is a systematic trend is difficult because of the lack of longer
records. Here we compile and consider an extreme flood record that extends back
to 1770. Since 1770, there have been 137 recorded extreme floods. However, over
this period, there is not a unidirectional trend of rising extreme flood risk
over time. Instead, there are clear flood-rich and flood-poor periods. Three
main flood-rich periods were identified: 1873–1904, 1923–1933, and 1994–2007.
To provide a first analysis of what is driving these periods, and given the
paucity of more sophisticated datasets that extend back to the 18th century,
objective Lamb weather types were used. Of the 27 objective Lamb weather types,
only 11 could be associated with the extreme floods during the gauged period,
and only 5 of these accounted for >80% of recorded extreme floods The
importance of these five weather types over a longer timescale for flood risk
in Carlisle was assessed, through calculating the proportion of each
hydrological year classified as being associated with these flood-generating
weather types. Two periods clearly had more than the average proportions of the
year classified as one of the flood causing weather types; 1900–1940 and 1983–2007;
and these two periods both contained flood-rich hydrological records. Thus, the
analysis suggests that systematic organisation of the North Atlantic climate
system may be manifest as periods of elevated and reduced flood risk, an
observation that has major implications for analyses that assume that climatic
drivers of flood risk can be either statistically stationary or are following a
simple trend.
KEY WORDS historical floods;
trends; weather types; flood frequency; flood magnitude
1.
Introduction
Flood risk is becoming an
increasingly important issue in northwest Europe in general and in the UK in
particular. The aim of this paper is to assess the extent to which this is a
systematic trend for a large river basin in northwest England and
then to assess the extent to which the results obtained can be linked back to
large-scale atmospheric forcing.
Severe flood events have been reported for the
UK in the spring of 1998 (central England) (Horner and Walsh, 2000), autumn
2000 (Sussex and Yorkshire; Kelman, 2001; Marsh and Dale, 2002;), autumn 2004 (Boscastle;
Golding et al., 2005; Roseveare and
Trapmore, 2008), winter 2005 (Carlisle; Environment Agency, 2006), summer 2007
(central and northern England; Marsh and Hannaford, 2007; Marsh, 2008), autumn
2008 (northern England; Wilkinson et al.,
2010) and autumn 2009 (northwest England; Eden and Burt., 2010). It has been
suggested that recent decades have seen more frequent and higher-magnitude
river flow extremes (Wheater, 2006) and that we are now in a flood-rich period
(Macdonald, 2006; Lane, 2008).
The apparent increase in flood events, however,
needs to be evaluated to assess whether or not it represents a long-term trend
or simply shorter-term variability. Robson (2002)
analysed both local and UK river flood series and found that there was an
increasing trend over the past 30–50 years, emphasising that assumptions of
stationarity in flood frequency analyses need to be questioned (Milly et al., 2008). Furthermore, there seems
to be a pattern and clustering of the worst flood events rather than a random
occurrence (Wheater,
2006),
perhaps related to shorter-term climatic variability. Others have reached the
same conclusion with respect to smaller regional datasets. For example,
Scotland has seen an increased river flood frequency since 1988, with new
maximum discharges recorded for many rivers, especially in the west (Black,
1995; Black and Burns, 2002; Werritty, 2002).
In relation to Europe, Brazdil et al. (2006) notes the large number of
flood events throughout central Europe in the last two decades; Rhine/Meuse in
December 1993 and January 1995; Biescas (Pyrenees) in August 1996; Morova/Oder
in July 1997; and the Elbe in August 2002. However, although many studies have
investigated flood frequency (Kundzewicz and Robson, 2004; Lindstrom and
Bergstrom, 2004; Radziejewski and Kundzewicz, 2004; Kundzewicz et al., 2005; Svensson et al., 2005), finding statistically
significant general trends has been more difficult. This is likely to be a
consequence of the low frequency of extreme events, meaning that long records
are needed to have the required number of events to identify statistically
significant trends. There are examples of historical flood records that suggest
periods of greater flood occurrence than others. Barriendos et al. (2003) investigated the
stationarity assumption for flood records in France
and Spain .
Mudelsee et al. (2006) constructed a
500-year flood record for the River Werra in Germany . Mudelsee et al. (2004) found that Europe experienced increased flooding frequencies in the
18th century, which has been hypothesized to have been caused by the Late
Maunder Minimum period. Macdonald (2006) found that this observation was not
present in the UK .
A possible explanation for this is that this period was cold and dry and saw an
increase in the number of snowmelt and ice-dam-break flood events, which are
more common in central Europe than in the UK .
These studies aside, there have been relatively
few assessments of both the extent and timescales of flood clustering, and even
fewer assessments of what might drive them. This paper explores the extent and
timescales of flood clustering for a 2400 km2 river catchment in northwest England , the River Eden at Carlisle ,
combining the shorter-term gauged record with longer-term historical data to
construct a flood record for the last 240 years. It then tests the first
hypothesis that the flood record can be divided into relatively flood-rich and
relatively flood-poor periods.
The link between weather systems and
hydrological flows, particularly extremes (floods and droughts) has been
investigated by a few studies. Higgs (1987) investigated the link between
weather types and floods for the River Severn at Bewdley, using a 101-year
record. Zonal (westerly) weather systems were found to be associated with the
highest-magnitude floods. Rumsby and Macklin (1994) studied the flooding frequency
and magnitude of the River Tyne, considering weather types as a controlling
factor. Major floods were found to be linked to meridional circulation
(easterly weather types), while more moderate floods occurred in periods when
zonal weather systems dominated (westerlies). A possible explanation for this
was through the high-amplitude waves associated with meridional circulations,
which are linked to situations when high pressure causes blocking of
depressions, leading to long-duration, high-intensity precipitation. A further
study by Rumsby and Macklin (1996) compared the western Severn catchment, with
the eastern Tyne catchment. The west of England is more susceptible to zonal
precipitation (westerlies), while the northeast of England
is in the rain shadow of the Pennines, so receives more precipitation from
meridional (easterly) weather systems which absorb moisture over the North Sea . Grew (unpublished) used daily weather system
classifications, unlike the previous studies which used monthly or annual
categories, for 130 peak over threshold (POT) series in Scotland . Cyclonic,
westerly and southwesterly weather systems were found to trigger flood events
in Scotland .
A similar approach was taken by Longfield and Macklin (1999) for the River Ouse
in Yorkshire . Westerly, cyclonic, cyclonic westerly
and southwesterly weather systems were found to have caused 79.7% of the floods
in the flood record since 1875.
Expanding the spatial scale to include the
weather types that cause floods in central Europe
has found that similar weather types are also regionally important. However, a
different weather type classification is used in Europe ,
the Grosswetterlagen (Baur, 1944), which has 30 classes under 3 main headings
of zonal (westerly), mixed and meridional (easterly). Kastner (1997) found that
only 5 of the 30 circulation types caused floods in Bavaria ,
while Petrow et al. (2007) found that
19 of the 30 caused floods in the Mulde catchment, Germany , in the 92-year period
(1911–2002). Both these studies highlighted the importance of westerly weather
types (25% of Mulde floods), and identified the Vb circulation pattern as the
most susceptible to causing floods in Europe (Mudelsee et al., 2004; Brazdil et al.,
2005; Petrow et al., 2007). The Vb
(van Bebber) weather type is a slow moving low pressure system, which moves
northwards from the Gulf of Genoa , and therefore is characterized by a warm and
moist air mass, which leads to high precipitation in the Alps .
A continental-scale study of 488 catchments in Europe
by Prudhomme and Genevier (2010) found that the cyclonic westerly weather type
occurred more frequently before and during a flood event than the annual
average.
At a larger scale, the link between weather
types and atmospheric processes and circulations has been investigated. For the
UK ,
one of the most significant large-scale atmospheric circulation indices is
thought to be the North Atlantic Oscillation Index (Kingston et al., 2006). This is a measure of the
pressure gradient between the Icelandic Low and the Azores High (Hurrell and
van Loon, 1997). It is often used as a measure of westerly weather systems over
the UK
and it has been found that Lamb weather types correlate well with the NAO,
especially anticyclonic and westerly weather types (Jones et al., 1997). Wilby et al.
(1997) identified four main phases of the NAO from pre-20th century to the
mid-1990s: (1) pre-20th century when the NAO was near zero; (2) 1900–1930 when
the NAO had a strong positive phase; (3) 1930–1960s when the NAO had a low
positive index; and (4) 1960s to the mid-1990s when the NAO had a strong
positive index (Wilby et al., 1997). Since
the mid-1990s, the strength of the positive NAO has been decreasing, and the
winter of 2010–2011 had the most negative NAO index in the 190-year record
(Osborn, 2011). Hurrell (1995) found links between shifts in the NAO and
changes in UK
temperatures and precipitation totals. Bendix (1997) highlighted the importance
of westerly weather types and an enhanced North Atlantic Oscillation in causing
floods throughout central Europe . Fowler and
Kilsby (2002) found a positive correlation between the NAO and the
precipitation quantities in the west of the UK and a negative correlation in
the east. However, the relationship does not seem to be that simple, with Wedgbrow
(2002) finding a lag between the changing NAO index and the change in UK weather. This
was hypothesized to be caused by either climatological memory effects, such as
seasonal patterns, or hydrological memory effects, for example, groundwater
levels or antecedent moisture levels. Along with the weather type
classifications, this index also has limitations for its use, as it represents
complex multivariate interrelationships very simply (Kingston et al., 2006). Thus, in this paper, we
focus upon weather types, not least because of the historical duration for
which they are available. We use these to test a second hypothesis that the
flood-rich and flood-poor periods identified in the historical record can be
linked back as a first approximation to atmospheric forcing.
2.
Methodology
2.1. The Eden
catchment and flooding in the city of Carlisle
The Eden
catchment comprises 6 major sub-catchments (Figure 1). The spatial annual
average precipitation of the Eden
catchment is 1183 mm (SAAR 1961–1990) (Environment Agency, 2008). The Eamont
sub-catchment receives the highest rainfall per year with a spatial annual
average value of 1768 mm and local annual averages in excess of 2800 mm in
areas of high topography. Â The Petteril
experiences the lowest rainfall totals with 942 mm per year, while the Lower
Eden in the city of Carlisle
receives approximately 800 mm every year.Â
The spatial differences in average annual rainfall can be explained by
the significant topographical variations within the catchment, with a total
relief of approximately 950 m. Although urbanisation has occurred in the Eden
over the last three centuries (for instance, the population of the largest
City, Carlisle, rose from 4000 in 1750 to 71773 in 2001), urban areas still
only account for less than 1% of the catchment, implying that urbanisation is
unlikely to be a major contribution to changing flood frequency at the
catchment scale. The vast majority of the catchment is rural and it remains an
unresolved issue as to whether or not rural land management, as well as river
and floodplain management, have contributed to changing flood frequency. Recent
work (Beven et al., 2008) has shown
that such effects are likely to be very difficult to detect in historical
records, not least because of natural climatic variability. Thus, whilst it
remains a possible hypothesis for changing flood frequency, and may have
contributed to those changes, it is likely that the primary driver of changing
flood frequency is a climatic one.
2.2. Short-term gauged record
It was possible to obtain a recent gauged record
of river flows for the Sheepmount gauging station in Carlisle
(Figure 1) which opened in January 1967. Digital records begin in 1976 and the
station is still operating. Here, event frequency and magnitude trends are
analysed using POT and annual maxima (Amax) series respectively for two scales
of high-flow event. The number of events per hydrological year that exceeded
the Q1 value (347 m3s–1), which was calculated from the
digitized record for 1976–2007 was determined. This threshold was chosen to
represent the full range of high-flow events in the Eden , rather than restricting the analysis to
just the overbank flood events. Analysis of the full event record will be
referred to as >Q1 events. Events were identified that were independent of
each other by requiring the time interval between floods to be three times the
duration of the typical rising limb (Bayliss and Jones, 1993), calculated from
an average of five flood events. This required events to be separated by a
minimum of four days.
To provide information on a second scale of
event, the Q1 events were also separated into high flows (347 m3s–1
<Q <500 m3s–1) and extreme flood events (Q
>500 m3s–1). The extreme flood threshold was
determined using a previous study by Smith and Tobin (1979) of long-term
flooding in Carlisle . Smith and Tobin (1979)
calculated the return period of floods in the 1800–1970 period. The return
period of the 1968 flood was found to be 42.75 years using the historical
record, which is comparable to the 38.5 years calculated by the North-West
Water Authority (Smith and Tobin, 1979). The discharge of events recorded in
the British Chronology of Hydrological Events database and Smith and Tobin were
found at Warwick Bridge (since 1959) and Sheepmount
(since 1975) gauging stations (Black and Law, 2004). This allowed the short-term
gauged record to be comparable with the longer-term record which was reported
by Smith and Tobin (1979) and which was used in compiling the long-term
historical extreme flood record.
Annual maxima series record the largest
instantaneous flood peak in each hydrological year (Svensson et al., 2005). The major advantage of
this approach is that data are easy to extract, but insignificant flows can be
included in the record, if a year was particularly flood poor. Thus, the peak
discharge of each POT event was also considered. Records were extended back to
1967 using the POT Hiflows database (www.environment-agency.gov.uk/hiflows).
2.3. Creating a longer-term
extreme high-flow record
Past research has shown the risks of concluding
the presence or absence of trends in short-term gauged records. There have been
several recommendations (Table I) as to the minimum required record length,
ranging from 10 to 50 years reflecting the problem that what can appear to be a
trend in a short-duration record may actually be shown to be fluctuation in a
longer data record (Robson, 2002; Kundzewicz and Robson, 2004; Dixon et al., 2006) and be associated with
spurious trends (e.g. Konrad and Booth, 2002). For instance, Hisdal et al. (2001), for a single station,
found significant positive and negative trends in annual flood maxima as a 30-year
moving window was applied to the record. Robson (2002) shows that shorter
record lengths are more susceptible to edge effects, when periods that have
several floods or few floods at the beginning and/or end of the record
influence the strength of the trend. Hannaford and Marsh’s (2007) benchmark
dataset for UK records had an average length of 33.7 years reflecting the fact
that much of the UK gauging station network was commissioned in the 1960s and
1970s (Lees, 1987).
Given the possibility that the 40-year record
(1967–2007) is too short to reliably detect trends in the dataset, a longer
timescale extreme flood record was constructed for the River Eden at Carlisle , using multiple sources of information. First,
the British Chronology of Hydrological Events (Black and Law, 2004) was used.
As of October 7th 2010, it listed 126 (some repeated) extreme flood records for
the River Eden. For copyright reasons, records generally cover the period
before 1931, although for the Eden ,
a record exists for the 1968 flood.
Second, newspaper reports from the Carlisle
Patriot, Carlisle Journal, Cumberland News, Evening News and Star, and the
Carlisle Directory were used to identify extreme flood occurrence. Some of the
records give specific details, such as a quotation, while others just list the
event and source. Third, extreme flood levels recorded on Eden
Bridge in Carlisle
by indentations with associated years indicate the peak flood water stage. Markings
are present for the 1822, 1856, 1868, 1925, 1952, and 1968 floods. The level of
the January 2005 extreme flood event was one metre higher than the highest
previous mark. Such marks need to be assessed for their originality, by
checking the age of the structure on which they are preserved (Brazdil et al., 2006). Eden Bridge
was built in 1815 and consists of five long arches. Therefore, all the
epigraphic markings are thought to be legitimate. However, a limitation of
using the flood levels is that the bridge width was doubled in 1932. This will
have changed the conveyance of water downstream. Water levels are controlled by
both discharge and conveyance, meaning that epigraphic markings are generally
good at indicating a flood, but are less good at indicating the magnitude of
the event.
Finally, Smith and Tobin (1979) ranked 49 major
known extreme floods at Carlisle between 1800
and 1968 according to the approximate extent of flooding. Â This was an important source of information as
it allowed the threshold for extreme floods to be standardized between the
different sources and timescales of the floods. The British Chronology of
Hydrological Events only recorded floods up until 1931, while gauged data
starts in 1959 at Warwick
Bridge and 1967 at
Sheepmount (Figure 1). Smith and Tobin’s (1979) record was used to fill the gap
between 1931 and 1959. The threshold of 500 m3s–1 was
used at Sheepmount to make the short-term gauged record comparable with the
historical extreme floods recorded by Smith and Tobin (1979). The gauged record
from Warwick Bridge was used to determine extreme
floods between 1959 and 1976. The comparable flow at Warwick Bridge
was calculated to be 460 m3s–1. Using these multiple
sources of information allowed a robust record to be compiled, whereby multiple
sources recorded the same event, along with single records allowing time
periods to be filled. The extreme flood record was developed extending back to
1770. See Table III for a complete record of the extreme floods in Carlisle showing the source of the information. The
reliability of the flood record increases with time due to better gauging and
recording of events. However, it is believed that the post-1800 record includes
most of the actual events. This is because multiple sources of information have
been used to derive the record.
2.4. Atmospheric drivers of
flood events
The weather in the UK is determined by the position,
origin, and storm tracks of air masses. Atmospheric circulation systems can be
classified (El Kadi and Smithson, 1992) and these have been used (e.g. Hess and
Brezowsky, 1977; Yarnal, 1993; Petrow et
al., 2007) to investigate the links between large-scale atmospheric
processes and regional weather and hydrology. In the UK , Lamb (1950, 1972) developed a
weather-type classification for 1861–1971. This is a classification based upon
both synoptic pressure and direction of flow, and so, Lamb weather types
describe the prevailing atmospheric pressure characteristics and hence indicate
the presence and tracks of storms over a catchment. Lamb’s original analysis
resulted in seven classes (westerly, northwesterly, northeasterly, easterly, southerly,
and anti-cyclonic and cyclonic) which were representative of weather systems
over the entire UK .
This subjective classification which relied on an expert basing a decision on a
synoptic chart was developed by Jenkinson and Collinson (1977) to make the
classification more objective. It has now been applied from 1881 to the present
day. It is based upon the daily mean sea level pressure, which is used to
indicate wind flow direction, shear vorticity, and flow strength (Jones et al., 1993). The Objective Jenkinson
classification has 27 classes, sub-divided by direction (N, NE, E, SE, S, SW,
W, NW), non-direction (cyclonic, anti-cyclonic), combined complex hybrid types
(CN, CNE, CE, CSE, CS, CSW, CW, CNW, AN, ANE, AE, ASE, AS, ASW, AW, ANW) and
unclassifiable (U). Jones et al.
(1993) found a strong correlation between the Lamb classification and the
Objective classification.
There are several advantages to using a weather-type
classification to investigate multivariate climatological factors: (1) The
classes are simple and easy to use; (2) The length of the record allows for
long term trends to be investigated; and (3) They are based on physical
linkages between the climate (large-scale processes) and weather patterns
(local scale). However, there are several limitations in the use of these
classifications (O’Hare and Sweeney, 1993). First, there is an issue regarding
the balance between number of classes and ease of use. The seven Lamb weather
types were thought to be too simplistic, so Jenkinson and Collinson (1977)
added another 20 classes. This allowed the UK weather to be better represented
but made the system more complex and harder to use. Second, some days
experience multiple weather types, making them difficult to classify. The
Objective Jenkinson system has an unclassified category, but this provides no
information on the specific weather types experienced. Third, the UK also
experiences different weather types in different regions. Questions have been
raised over how representative of UK
weather types these classifications are of the UK as a whole. Fourth, the Lamb
weather-type classification is subjective, although the changes made by
Jenkinson and Collinson (1977) have made it more objective. However, Yarnal and
White (1987) suggest that there are still problems in the use of objective
classifications. Fifth, there are problems associated with assigning a daily
weather type when climatological variables do not operate on daily timescales. Sixth,
the relationship between weather type and rainfall totals is not always
reliable and it has changed over the timescale of the record. Seventh, the
classifications indicate direction of origin but not the specific region, which
may differ considerable in their characteristics, including tropical, maritime,
continental air masses. Also, air masses from the same origin have different
characteristics at different times of the year. Eighth, weather-type
classifications indicate large-scale synoptic atmospheric processes and lack
detail on meso-scale frontal and orographic systems, which cause a lot of the UK precipitation.
Finally, the weather system classification scheme is inherently auto-correlated,
as when one weather type becomes more frequent, others have to decrease in
their occurrence. Despite the inherent limitations of the objective lamb
weather type classification, it still allows the link between local catchment-scale
flooding to be linked to large-scale atmospheric forcings over the historical
period. The benefits of this relatively simple classification scheme is that it
provides a daily summary of weather characteristics over a long time period,
while more detailed datasets are constrained in record length.
The methodology used aims to identify links
between the objective lamb weather types and events of different magnitude
(>Q1 events (high flows) and extreme floods) for the gauged period and
consisted of the following steps. First, the objective Lamb Weather-Type
dataset was sourced (www.cru.uea.ac.uk/cru/data/lwt),
which starts in 1880 and continues to the present day. The weather type on the
day of each >Q1 event was extracted from the dataset, along with the weather
classification on the previous two days. As the Eden is quite a large catchment
(2400 km2), the number of days of precipitation that result in a
high flow or extreme flood events downstream may be more than just the day of
the event. Grew (unpublished) stated that the number of days of precipitation
is dependent upon the specific catchment characteristics, including area and
gradient. The relative time between the peak flow in the Upper Eden (Kirkby
Stephen) and the Lower Eden (Sheepmount), has
a maximum lag of 34.5 h, and a mean lag time of 12 h. The delay between
precipitation and a peak flow occurring will increase this response time
further. Longfield and Macklin (1999) devised a method using daily rainfall
records to assess the number of days responsible for flood generation. The
previous four days were included and each day given a weighting dependent upon
the amount of rainfall. The objective Lamb weather type on the day with the most
rainfall was taken as the dominant synoptic system that caused each event. However,
we focus on the sequencing of weather types. In this study, the weather types
on the previous two days as well the day of the event are assessed. First, each
day is looked at separately; and second, the sequence of days is investigated.
 Event-generating
weather types were then identified from this dataset, as the weather types that
occur most frequently on days, and this was undertaken for both >Q1 and
extreme flood events. The weather types that occur on >Q1 events were
compared with those associated with extreme flood events. Trends in the extreme
flood-generating weather types are then investigated over the historical
timescale by calculating the percentage of each hydrological year for the
extreme flood-generating weather types both individually and combined. The
average of the 1880–2007 period was calculated, then the average was subtracted
from each hydrological year. This means that positive values represented years
which had a greater-than-average proportion of the year with these extreme
flood-generating weather types, while negative values had less than the
average. The cumulative was then calculated for the deviations from the average
and can be plotted against time. This plot is a means of visualising the
sequencing of flood-generating weather types. A period when the deviation is
trending from negative to positive suggests a greater number of
flood-generating weather types. The longer the period of this trend, the
greater the length of the period when more flood-generating weather types have
been present than average. If we imagine a flood-generating weather type as one
that may, but that does not necessarily, produce a flood, then the longer a
positive trend, the more likely it might be expected to identify a flood in the
flood series.
3.
Results
3.1. Gauged records
Figure 2(a) shows that the late
1960s and 1970s were relatively poor in terms of >Q1 flow events, with fewer
than the average number of events per year every hydrological year except 1967–1968
and 1974–1975, which were the years with the most events in the entire record. Events
>Q1 occurred in every year except 1995–1996, a year of hydrological drought.
The Pearson’s product moment
correlation coefficient
of the number of >Q1 events over time for Sheepmount is
only 0.07, which is not statistically significant (p = 0.42). Of the 138 >Q1 events since 1967, 31% were classified
as extreme flood events. The largest number of extreme flood events in a hydrological
year is 4, and occurred in 1967–1968, 1981–1982, and 2003–2004. There are also
no statistically significant trends in either non-extreme >Q1 (r = 0.01 p = 0.93) or extreme flood (r
= 0.19 p = 0.22) events over the
gauged period at Sheepmount.
Figure 2(b) shows the annual maximum flood for
the River Eden at Carlisle (Sheepmount) and
indicates a wider range in the magnitude of the annual maximum flood. The most
extreme flood was in January 2005 with a magnitude of 1516 m3s–1,
with other notable extreme floods in 1968, 1981, 1985, and 1995. Annual maximum
flows which are below the extreme flood threshold we are using (500 m3s–1)
are highlighted in grey (34% of hydrological years). The lowest AMax magnitude
for Carlisle (Sheepmount) was 291 m3s–1
in 1995–1996, although years without extreme floods seem to have occurred in
the 1970s more than at present. A second, more robust approach used to assess
the frequency and magnitude of extreme floods considers the magnitude of the
events that exceed the >500 m3s–1 extreme flood
threshold (Figure 2(c)). However, there are also no statistically significant
trends in this record. In conclusion, the short-term records do not exhibit any
statistically significant trends in either high flow or extreme flood frequency
or magnitude for the River Eden at Carlisle . Â However, as Kundzewicz and Robson
(2004) notes that a failure to identify significant trend does not necessarily
mean that there is not one, especially given the relatively short duration of
the record used here.
3.2. Historical flood record
for Carlisle
Figure 3 shows the cumulative number of extreme
floods (>500 m3s–1) since 1770. The periods on Figure
3 where the gradient of the line is steep indicate flood rich periods. Times
when the line is flatter are flood poor. It appears that there are three flood-rich
periods over the past 240 years: (1) 1873–1904; (2) 1923–1933; and (3) 1994
onwards, each separated by periods which were relatively flood poor, which have
been classified visually. The years with the most extreme floods are 1877 and
1891, with 5 recorded in these years. The period before 1850 has very few
extreme floods, which may be due to the lack of evidence for them occurring,
rather than a lack of existence. However, it is assumed that the largest events
have been recorded. The magnitude of the largest events have been estimated by
the Environment Agency (2006). Bankfull discharge at the Sheepmount station is
1434 m3s–1, and only the 2005 event exceeded this
threshold. However, floodplain inundation occurred in all the events recorded
in the British Chronology of Hydrological Events database, which has an
approximate threshold of 500 m3s–1.
3.3. Weather types for instrumented period
floods
Using the Objective Lamb Weather Types, it was
found that 11 of the 25 weather types have caused extreme flood events in the
gauged periods (1976–2007 at Sheepmount), of which 5 (cyclonic = 27.3%, westerly
= 15.9%, southwesterly = 15.9%, cyclonic southwesterly = 6.8%, cyclonic westerly
= 15.9%) accounted for 81.8% of the extreme flood events (Figure 4). These
results are similar to the findings of Longfield and Macklin (1999) for the
Yorkshire Ouse Catchment, where four circulation types (W, C, CW and SW)
accounted for 79.7% of all events (and the same 5 weather types caused 82.6% of
flood events in the Ouse record). These particular weather types highlight the
importance of both cyclonic weather types and weather systems from a westerly
and southwesterly direction to both high flows and extreme floods occurring in Carlisle . Cyclonic weather systems are likely to cover a
greater spatial area and lead to a more coherent catchment response. Furthermore,
as they are not prescribed a direction, this means that they are often blocked
by other air masses, meaning they are stationary, resulting in a prolonged
rainfall event. Several studies have found that these weather systems are of
notable importance in accounting for precipitation in this region (Malby et al., 2007) and the UK in general
(Sweeney and O’Hare, 1992). Figure 4 shows that the other objective Lamb
weather types are not important in causing >Q1 events, with only 14 >Q1
events being caused by the other 20 weather types, 8 of which are extreme flood
events.
Figure 5(a) indicates that the common weather
types on the two preceding days are the same as the ones on the day of the
event itself for extreme events. However, the order of importance of the five
dominant weather types is different for the preceding days than the day of the
extreme flood itself. While cyclonic weather systems are the most common on the
day of the event, weather systems from a southwesterly (38% on previous day,
25% on two days before) and westerly (20.4% on the previous day, 25% on two
days before) direction are the most common on the two preceding days. Cyclonic
weather systems are less common on the days previous to an event occurring
(15.9% on previous day, 9.1% on two days before). Furthermore, cyclonic weather
systems from a westerly and southwesterly direction are also less common on the
days prior to an event.
The sequencing of the weather types may also be
important in causing extreme floods, as they control the antecedent conditions
of the catchment. This has been assessed in terms of whether or not the
previous two days and the day of the event were classified as an event-generating
weather type (C, W, SW, CW, and CSW). Table II shows that 47.7% of extreme
floods have had event-generating weather types on both the day of the event and
the previous two days, while a further 27.3% of extreme flood events occurred
on days with both the day of the event and the day before classified as an
event-generating weather type. Only one extreme flood since 1976 occurred when
none of the three days were classified as one of the extreme flood-generating weather
types. No extreme floods occur in sequences where just the day of the event is
an event-generating weather type (1 0 0).
None of these analyses takes account of the
proportion of the year associated with each weather type. Therefore, Figure 6
shows the percentage of the 1976–2007 period classified as each weather type. Anti-cyclonic
and cyclonic weather systems dominate, accounting for 20.7 and 13.8%,
respectively, for the whole period, and 21.1 and 13.0%, respectively, of the
last 40 years. Weather systems from a southwesterly and westerly direction also
have a high frequency individually, as well as for anti-cyclones and cyclones.
The likelihood of a particular weather system
causing an event can be determined by dividing the number of events occurring
on days of a particular weather type by the total number of days of the same
weather type over the same period. Figure 7 shows that the most likely weather
type to cause an extreme flood in Carlisle is the Cyclonic Westerly, with a
2.6% chance of an event occurring on a day with this weather system over the UK . This is
because it is the least common of the event-generating weather types over the
40-year period in terms of occurrence, but has still caused 7 events. Cyclonic
synoptic events have a 0.7% chance of leading to an extreme flood occurring, as
although most events occur on cyclonic days, these weather systems occur most
often. .
3.4. Comparison of high flow and extreme flood
event-generating weather types
Figure 4 shows that 10 of the objective Lamb
weather types have occurred on days of high flows, and the same 5 weather types
account for 93.6% of the days when high flows occur (cyclonic = 34.0%, westerly
= 19.1%, southwesterly = 17.0%, cyclonic southwesterly = 12.8, cyclonic westerly
= 10.6%) A Chisquare test showed that the weather types that cause extreme
floods and smaller-magnitude high flows are statistically similar (p = 0.32). The weather types on the
preceding two days are also similar for both high flows and extreme flood
events (Figure 5(a) and (b)). This is significant because it means that weather
types cannot be used to distinguish between the magnitude of the event that
might occur: whether a high flow or an extreme flood. Furthermore, the
sequencing of weather types show no significant (Chi-square test p = 0.99) difference for events of
differing magnitude, with high flows and extreme floods showing similar
percentages for each sequence.
3.5. Weather types for the historical
period
The relationship between weather systems and
extreme flood frequency will now be investigated over a longer timescale. A few
previous studies have looked into how weather type frequency has changed over
approximately the last 100 years (Lamb, 1972; Jones and Kelly, 1982; Briffa,
1990; Sweeney and O’Hare, 1992; Fowler and Kilsby, 2002; Malby et al., 2007). Many of these
investigations reported a decrease in the number of westerly days since the
1950s, while cyclonic and anti-cyclonic weather systems have become more common
since the 1980s. The focus here is those weather types are found to produce
most of the extreme flood events in the recent gauged period.
Figure 8 shows the cumulative number of extreme
floods (>500 m3s–1) since 1880, superimposed upon the
cumulative deviation of flood-generating weather types. The periods in Figure 8
where the gradient of the line is steep indicate flood-rich periods. Times when
the line is flatter are flood poor. In relation to flood-generating weather
types, it is clear that there are a number of scales of variability. At the
largest scale, there are two periods of generally positive trend, with the
exception of short breaks in this trend: 1902–1938 and 1983–2007; before 1902
there was a dominant negative trend; and between 1938 and 1983 there were
shorter periods of negative and positive trend. Thus, Figure 8 shows that flood-generating
weather types are not randomly located in time but clustered into periods when
there are generally more than average and generally fewer than average. The
association between these weather type patterns and the cumulative flood record
is interesting. For both the periods of 1902–1938 (until 1931) and 1983–2007,
the cumulative flood records are weak positive exponentials suggesting that as
the duration of generally positive deviations becomes longer, and the number of
flood-generating weather events in the positive sequence becomes greater, so
there are more floods. This does not require a mechanism like land use change
or groundwater recharge (which for the Eden is important but not that much so –
the Base Flow Index is 0.498), but is a result of clustering of
flood-generating weather types, which in turn increases the probability that
one of these weather events becomes a flood, and so, leads to an increase in
the number of floods. Similar overall trends are shown in historical rainfall
records (e.g. for Lockwood Reservoir; Fowler et al., 2002). Of course, it is also possible that the objective
Lamb weather-type classification misses some climatic signals, such as
precipitation intensity or quantity, as it is only a broad categorical system.
Figure 9 shows how the proportions of individual
weather types per year change over time. Firstly, the cyclonic-westerly (Figure
9(a)) weather system does not vary significantly from the average, with only a
range of 4.9% (0.6–5.5%). Also periods with more cyclonic-westerly weather
systems do not correlate well with the periods of increased extreme flood
activity in the Eden .
The cyclonic southwesterly (Figure 9(b)) weather type varies by 4.7% (0.6–5.2%)
and seems to match the flood-rich and flood-poor periods visually quite well. Pre-1918,
the proportion of the year classified as a cyclonic southwesterly weather type
decreased, while extreme flooding had a low frequency. Between 1919 and 1955,
the proportion of the year categorized as cyclonic southwesterly increased,
which occurred simultaneously with the 1923–1933 flood-rich period. Since the
mid-1950s to 2007, the proportion of cyclonic southwesterly types per year has
stayed quite constant, although there has been a slight increase since the
mid-1980s. The cyclonic (Figure 9(c)) weather system has varied by 17.8% (5.5–23.3%)
in terms of the proportion of the year classified as this weather type over the
last 140 years. During the pre-1923 flood-poor period, this weather type was
decreasing in terms of the proportion of the year classified as it. It then
increased during the 1923–1933 flood-rich period. It has also increased since
the mid-1970s, although specific years have had less than the average proportion
of the year classified as cyclonic. The westerly (Figure 9(d)) weather system
has varied by 9.6% (5.2–14.8%) throughout the whole period. This weather system
does not seem to match the flood-rich periods well, with a decline in the
proportion of the year of the westerly weather type since the mid-1990s, which
coincides with the start of the flood-rich period. Finally, the southwesterly
(Figure 9(e)) weather system has varied by 11.5% (3.6–15.1%). This weather type
has the highest level of agreement with the extreme flood frequency, with the
proportion of the year classified as southwesterly increasing from 1900-mid
1930s, falling significantly from 1960 to 1980 and then increasing again in the
current flood-rich period.
4.
Discussion
If the latter part (post-1965, the start of the
gauged record) of Figure 3 is analysed, then it could be concluded that there
is a unidirectional trend of increasing extreme flood frequency in Carlisle,
that starts with very few floods during the late 1960s, a rising number of
floods until the early 1990s, and a very rapid rise after that into the flood-rich
period from 1994 to 2007. However, when put into the historical period context,
it becomes clear that there is not a unidirectional trend and that the period
since 1994 has been flood rich, but so have other periods. Several other
studies have identified flood-rich and flood-poor periods in historical flood
records (Grew and Werritty, 1995; Werritty et
al., 2002; Macdonald, 2006; Macdonald et
al., 2006; McEwen, 2006). These examples, along with the River Eden,
indicate that there are flood clusters throughout the historical period. However,
a conclusion from Macdonald (2006) was that these flood-rich periods are not
nationally synchronous, which indicates that regional climatic variability and
catchment-specific characteristics are important in controlling flooding
frequency. Possible reasons why flooding may not be recorded as regionally
synchronous may be that; (1) There is an absence of extreme flood event
recording in the documentary evidence; or (2) There are different causal
mechanisms for extreme flood generation in different river catchments. However,
it has been found that high-magnitude floods can transcend catchment
boundaries, depending on the precise forcing mechanisms and antecedent
conditions. The 1771 floods occurred in both the Rivers Eden and Tyne (Macdonald (2006), and there were floods on both the
River Severn and River Trent in 1796. Furthermore, along with the flood event
on the River Eden in January 2005, the River Tyne also exhibited flooding
(Archer et al., 2007a, 2007b). The
storm event which caused this flooding extended from the 6th to 9th January
2005 and affected northern England ,
southern Scandinavia , Germany and the Baltic Region
(Carpenter, 2005).
The presence of flood-rich and flood-poor
periods throughout the historical period has implications for the concept of
stationarity. This assumption states that natural systems fluctuate within an
unchanging range of values. However, as Milly et al. (2008) stated it has been compromized by human disturbance,
natural climate change, and variability. They go on to say that the assumption
of stationarity is dead and cannot be revived. Through analysing historical
flood records, it could be concluded that stationarity in flood frequency has
never existed, as floods have tended to cluster within certain periods
separated by flood-poor periods. Therefore, the range of behaviour of the
system depends on the timescale over which it is analysed. If the post-1960
period is analysed without the context of the historical period, then it could
be concluded that the behaviour that we are seeing today is unprecedented. However,
if the whole period is analysed then it can be concluded that today’s flood
frequency is not out of the range of past flood occurrence, suggesting a more
stationary behaviour. However, the presence of flood clustering means that
there is a variable chance of a flood occurring in every year of the record.
This raises a problem for flood frequency studies, as using short records will
not capture flood variability (Bardossy and Pakosch, 2005), while using longer
time series does not reflect stationary conditions, invalidating the
assumptions of standard frequency analysis (Khaliq et al., 2006).
This study has highlighted the importance of
cyclonic and westerly weather types, along with previous studies (Longfield and
Macklin, 1999). However, a recent study by Macdonald et al. (2010) for Wales
found that the link between westerly (SW, W, NW) weather systems and flood
occurrence only is statistically significant for northern Wales . However,
there is no link between cyclonic weather types and flooding for Wales . Therefore,
for Wales ,
weather types seem to be a poor predictor of flood frequency. This builds on
the findings of Dixon et al. (2006)
who found a strong east–west gradient in stream flow trends for western Britain .
Possible explanations for why flood frequency and weather types can exhibit
contrasting trends in different river systems may be the role of rain-shadows
of orographic rainfall, the influence of oceans (distance) (Dixon et al., 2006) and catchment aspect. Macdonald
(2010) explained these weak associations by commenting on the dataset
limitations, with the need for concentration and distribution of the days of
each weather type to be considered instead of just the annual totals. This has
been addressed in this study by including the weather types occurring on
extreme flood days and the preceding days and found that the sequencing of
weather types is important in explaining extreme flood occurrence. It has been
found that the same five weather types that occur on extreme flood days are
also more likely to have occurred on the previous two days compared to the
other objective Lamb weather types. However, it has been found that the
likelihood of a certain sequence of weather types in causing an extreme flood
is not significantly different to it causing a high river flow. This indicates
that while weather types can be used as an indicator of how likely a high
flow/flood is, they cannot be used to predict the likelihood of an extreme
event occurring in advance.
In addition, weather types and their link
to flood risk have been investigated over longer timescales. Brazdil et al. (2006) states that knowledge of
synoptic patterns for recent flood events can help explain past flood event
occurrence. It is this assumption that this study has made, using the last 40
years to define extreme flood-generating weather types and then looking at
their frequency over the historical period. This study has established a
possible link between the proportion of the year defined as an extreme flood-generating
weather type and extreme flood frequency. However, this analysis assumes that
the weather types that cause extreme flooding has not changed over time. The
weather types that have caused floods during the last three decades (1978–1987;
1988–1997; and 1998–2007) have been analysed to assess this assumption. Figure
10 shows that the weather types that result in extreme floods have changed from
the first decade, when two forms of anticyclonic weather types caused extreme
floods (A, AS) to CSE and CS in the last decade. In terms of the high-flow
events, the importance of the cyclonic weather type has nearly doubled from the
1978–1987 to the 1998–2007 period. However, the dominant weather types (5
extreme flood-generating weather types) have not changed over the 30-year
gauged period. Furthermore, it has been shown that the weather types that cause
both high flows and extreme floods have not changed statistically over the
decades (ANOVA p = 0.816). Malby et al. (2007) found that for the Eden
catchment, southwesterly and
westerly weather systems contributed the most to the decadal precipitation
totals. Furthermore, winter rainfall delivered by these weather systems has
increased over the last 30 years. Specifically, the precipitation associated
with each westerly weather system has increased between the 1970s and the 1990s
for 5 rainfall gauging stations in the Eden catchment. The quantity of rainfall
supplied by southwesterly weather systems was highest in the 1980s. Jacobeit
et al. (2003) found that a broader
range of circulation modes are important if studies are extended back into
historical periods.
The periods of a greater proportion of the year
than average of the five extreme flood-generating weather types correlate with
the flood rich periods. Furthermore, they match the periods identified by Wilby
et al. (1997) as periods when the NAO
was in a strong positive phase. Jones et
al. (1997) found a strong correlation between the NAO index and westerly
weather systems, which is one of the extreme flood-generating weather types. Wedgbrow
(2002) found a lag between the changing NAO index and the change in UK weather.
This was hypothesized to be caused by either climatological memory effects,
such as seasonal patterns, or hydrological memory effects, for example,
groundwater levels or antecedent moisture levels. This study has also found a
lag between the increase in the proportion of the year classified as one of the
five extreme flood-generating weather types and the increase in flood
frequency. This would seem to be expected due to the ‘chain of causality’
(Lawler et al., 2003) whereby the
link between large-scale atmospheric forcings, such as the NAO, are spatially
scaled down to their catchment effects, through the weather types and the
amount of precipitation which occur.
5.
Conclusion
Various event frequency and
magnitude indices have been used to investigate trends in both high flows and
extreme floods. First, the gauged record at Sheepmount, Carlisle (1967–2007),
was used to define the threshold of a high-flow event, which was taken to be
the Q1 value of 347 m3s–1. Extreme floods were defined by
a threshold of 500 m3s–1, to allow compatibility with the
longer-term Smith and Tobin (1979) study. The gauged record showed that,
although the 1960s and 1970s seemed >Q1 event poor in comparison to more
recent decades, there were no statistically significant trends over time. This
was also the case when the non-extreme >Q1 and extreme records were
considered separately. The annual maximum series showed that two thirds of
years experienced an extreme flood. There were more annual maximum events which
were less than 500 m3s–1 (high flows only) during the
1970s than at present, although again these trends were not statistically
significant. Multiple sources of documentary and epigraphic evidence were used
to compile an extreme flood record from 1770 to 2007. This showed that extreme
flood events have clustered in time, and 3 flood-rich periods were defined as
1873–1904; 1923–1933, and 1994–2007.
The Lamb weather types that occurred on the days
of extreme flood days (1976–2007) were extracted, and it was found that 11 out
of the 25 weather types have caused extreme floods in the gauged period. Of
these, 5 have caused 81.8% of extreme floods. These were cyclonic, westerly, southwesterly,
cyclonic-westerly, and cyclonc wouthwesterly. It was shown that there was no
statistically significant difference in the weather types that occur on days
with either high flows or extreme floods. The same five weather types were more
likely to occur on the previous two days before extreme flood events. The
sequencing of weather types was also found to be important, with ~50% of
extreme floods and high flows occurring after sequences of 3 days of the 5
event-generating weather types. However, the sequencing of days was not
statistically significant in determining whether an extreme flood or high flow
occurred. This means that weather types cannot be used to distinguish between
the magnitude of the event which might occur: whether a high flow or an extreme
flood.
The proportion of each hydrological year of the
five extreme flood-generating weather types was calculated. It was found that
there are two periods when the proportion of the year is less than the average
for a sustained period; 1902–1938; and 1983–2007. These were shown to correlate
with the flood-rich periods, although a lag existed between the increase in the
extreme flood-generating weather types and flood frequency increasing. These
periods also match with Wilby et al.
(1997) periods of a strong positive NAO index. Thus, the analysis suggests that
systematic organisation of the North Atlantic climate system, which drives the
weather types experienced by the UK, may be manifest as periods of elevated and
reduced flood risk, an observation that has major implications for analyses
that assume that climatic drivers of flood risk can be either statistically
stationary or are following a simple trend.
Acknowledgements
Ian Pattison was funded by the Environment
Agency and United Utilities as part of the EU Interreg IVB project ALFA (http://www.alfa-project.eu)
and acknowledges additional support from Durham University and the Eden Rivers
Trust. Two anonymous reviewers and the editor (Glenn McGregor) provided
valuable and constructive comments on an earlier version of this manuscript.
References
Archer DR, Leesch F, Harwood K. 2007a. Assessment of severity of the
extreme River Tyne flood in January 2005 using gauged and historical
information. Hydrological Sciences
Journal 52(5): 992–1003.
|
Archer DR, Leesch F, Harwood K. 2007b. Learning from the extreme River
Tyne flood in January 2005. Water and
Environment Journal 21: 133–141.
|
Bardossy A, Pakosch P. 2005. Wahrscheinlichkeiten extremer Hochwasser
unter sich andernden Klimaverhaltnissen. Wasserwirtschaft
7–8: 58–62.
|
Barriendos M, Coeur D, Lang M, Llasat MC, Naulet
R, Lemaitre F, Barrera A. 2003. Stationarity
analysis of historical floods in France and Spain (14th-20th Centuries). Natural Hazards and Earth System Sciences
3: 583–592.
|
Baur F, Hess P, Nagel H. 1944. Kalander der Grosswetterlagen Europas
1881-1939, Bad Homburg.
|
Bayliss AC, Jones RC, 1993, Peaks over threshold database: Summary
Statistics and seasonality, Institute of Hydrology Report No. 123, Institute
of Hydrology, Wallingford.
|
Bendix J. 1997. Naturliche und antropogene Einflusse auf den Hochwasserabfluss
des Rheins. Erdkunde 51: 292–308.
Beven. 2008. Analysis
of historical data sets to look for impacts of land use and management change
on flood generation. Department for Environment, Food and Rural Affairs,
London.
|
Black AR. 1995. Major flooding and increased flood frequency in
Scotland since 1988. Physics and
Chemistry of the Earth 20(5):
463–468.
|
Black AR, Burns JC. 2002. Reassessing the flood risk in Scotland The Science of the Total Environment 294: 169–184.
|
Black AR, Law FM. 2004. Development and utilisation of a national
web-based chronology of hydrological events. Hydrological Sciences Journal 49(2): 237–246.
|
|
Brazdil R, Kundzewicz ZW, Benito G. 2006. Historical hydrology for
studying flood risk in Europe. Hydrological
Sciences Journal 51(5): 739–763.
Brazdil R, Pfister C, Wanner H, von Storch H, Luterbacher J. 2005.
Historical climatology in Europe - the state of the art. Climatic Change 70(3):
363–430.
|
Briffa KR, Jones PD, Kelly PM. 1990. Principal component analysis of
the Lamb catalogue of daily weather types: Part 2, Seasonal frequencies and
update to 1987. International Journal
of Climatology 10: 549–563.
|
Carpenter G. 2005. Windstorm Erwin, CAT-i Catastrophe information.
|
Dixon H, Lawler DM, Shamseldin AY, Webster P. 2006. The effect of
record length on the analysis of river flow trends in Wales and central
England. Climate Change and Variability
Hydrological Impacts. IAHS Publication: 308, 490–495.
|
Douglas EM, Vogel RM, Kroll CN. 2000. Trends in floods and low flows
in the United States: impact of spatial correlation. Journal of Hydrology 240:
90–105.
Eden P, Burt S. 2010. Extreme rainfall in Cumbria, 18-20 November 2009.
Weather 65(1): 14.
|
El Kadi AK, Smithson PA. 1992. Atmospheric classifications and
synoptic climatology. Progress in
Physical Geography 16: 432–455.
|
Environment Agency. 2006. Cumbria floods technical report: Factual
report on the meteorology, hydrology, and impacts of the January 2005
flooding in Cumbria, Environment Agency, 2008: Hiflows, 167.
|
Fowler HJ, Kilsby CG. 2002. Precipitation and the North Atlantic
Oscillation: A study of climatic variability in Northern England. International Journal of Climatology 22: 843–866.
|
Gan KC, McMahon TA, Finlayson BL. 1991. Analysis of periodicity in streamflow and rainfall data by
Cowells indexes. Journal of Hydrology
123(1): 105–118.
|
Golding B, Clark P, May B. 2005. The Boscastle flood: Meteorological
analysis of the conditions leading to flooding on 16 August 2004. Weather 60(8): 230–235.
|
Grew H, Werritty A. 1995. Changes in flood frequency and magnitude in
Scotland 1964- 1992, Proceedings of the 5th National Hydrology Symposium,
British Hydrological Society, Wallingford, 3.1–3.9.
|
|
Hannaford J, Marsh TJ. 2007. High flow and flood trends in a network of
undisturbed catchments in the UK. International
Journal of Climatology 28(10):
1325–1338.
|
Hess P, Brezowsky H. 1977. Katalog der Grosswetterlagen Europas
1881-1976, third revised edition, Berichte des Deutschen Wetterdienstes no.
113, Offenbach, Germany.
|
Higgs G. 1987. The role of prevailing atmospheric circulation on long
term discharge records: A British example. Swansea Geographer 24:
7–12.
|
Hisdal H, Stahl K, Tallaksen LM, Demuth S. 2001. Have streamflow
droughts in Europe become more severe or frequent? International Journal of Climatology 21: 317–333.
|
Horner MW, Walsh PD. 2000. Easter 1998 floods. Water and Environment Journal 14(6): 415–418.
|
Huh S, Dickey DA, Meador MR, Ruhl KE. 2005. Temporal analysis of the
frequency and duration of low and high streamflow: years of record needed to
characterise streamflow variability. 310:
78–94.
|
Hurrell JW. 1995. Decadal trends in the North Atlantic
Oscillation-Regional temperatures and precipitation. Science 269: 676–679.
|
Hurrell JW, Van Loon H. 1997. Decadal variations in climate associated
with the North Atlantic Oscillation. Climatic
Change 36: 301–326.
|
Interagency Advisory Committee on Water data, 1982, Guidelines for
determining flood frequency: Bulletin 17B of the Hydrology Subcommittee, Office
of Water Data Coordination, US Geological Survey, Reston Va, 183.
|
Jacobeit J, Glasser R, Luterbacher J, Wanner H. 2003. Links between flood events in central Europe since AD 1500 and
large scale atmospheric circulation modes. Geophysical Research Letters 30(4):
1172–1176.
|
Jenkinson AF, Collinson BP. 1977. An initial climatology of gales over
the North Sea, Synoptic Climatology Branch Memorandum No. 62, Meteorological
Office, Bracknell.
|
|
Jones PD, Hulme M, Briffa KR. 1993. A comparison of Lamb circulation
types with an objective classification scheme. International Journal of Climatology 13: 655–663.
|
Jones PD, Jonsson T, Wheeler D. 1997. Extension to the North Atlantic Oscillation using early
instrumental pressure observations from Gibraltar and southwest Iceland. International Journal of Climatology 17: 1433–1450.
Jones PD, Kelly PM. 1982. Principal Components Analysis of the Lamb
catalogue of daily weather types:1: annual freqencies. Journal of Climatology 2:
147–157.
|
Kastner W. 1997. How do Grosswetterlagen become flood producing
weather situations in Bavaria? DGM 41(3): 107–112.
|
Kelman I. 2001. The Autumn 2000 floods in England and Flood Management.
Weather 56(10): 346–348.
|
Khaliq MN, Ouardo TBMJ, Ondo JC, Gachon P, Bobee B. 2006. Frequency
analysis of a sequence of dependent and/or non-stationary
hydro-meteorological observations: a review. Journal of Hydrology 329(3):
534–552.
|
Kingston DG, Lawler DM, McGregor GR. 2006. Linkages between
stmospheric circulation, climate and streamflow in the northern North
Atlantic: research prospects. Progress
in Physical Geography 30(2):
143–174.
|
|
Konrad CP, Booth DB. 2002. Hydrologic trends associated with urban
development for selected streams in the Puget Sound Basin, western Washington,
US Geological Survey Water Resources Investigations Report 02-4040, Tacoma,
Washington.
|
Kundzewicz ZW, Gracyyk D, Maurer T, Pinskwar I, Radziejewski M,
Svensson C, Szwed M. 2005. Trend detection in river flow series: 1. Annual
Maximum flow. Hydrological Sciences Journal
50(5): 797–810.
Kundzewicz ZW, Robson AJ. 2000. Detecting trends and other changes in
hydrological data, World Climate Programme- Data and monitoring, World
Meteorological Organisation, Geneva.
|
Kundzewicz ZW, Robson AJ. 2004. Change detection in hydrological
records- a review of methodology. Hydrological
Sciences Journal 49(1): 7–19.
|
|
Lamb HH. 1950. Types and spells of weather around the year in the
British Isles: annual trends, seasonal structure of the year, singularities. Quarterly Journal of the Royal
Meteorological Society 76: 393–429.
|
Lamb HH. 1972. British Isles weather types and a register of daily
sequence of circulation patterns: 1861-1971, Geophysical Memoirs, 116, HMSO,
London.
|
Lane SN. 2008. Climate change and the Summer 2007 floods in the UK. Geography 93: 91–97.
|
Lawler DM, McGregor GR, Phillips ID. 2003. Influence of atmospheric
circulation changes and regional climate variability on river flow and
suspended sediment fluxes in southern Iceland. Hydrological Processes 17:
3195–3223.
|
Lees ML. 1987. Inland water surveying in the United Kingdom - A short
history, 1985 Yearbook: Hydrological data UK series, Wallingford: Institute
of Hydrology, 35–47.
|
Lettenmaier DP, Wood EF, Wallis JR. 1994. Hydro-climatological trends
in the continental United States 1948-1988. Journal of Climate 7:
586–607.
|
Lindstrom G, Bergstrom S. 2004. Runoff trends in Sweden 1807-2002. Hydrological Sciences Journal 49(1): 69–81.
|
Lins HF, Slack JR. 1999. Streamflow trends in the United States. Geophysical Research Letters 26(2): 227–230.
|
Longfield SA, Macklin MG. 1999. The influence of recent environmental
change on flooding and sediment fluxes in the Yorkshire Ouse basin. Hydrological Processes 13: 1051–1066.
|
Macdonald N. 2006. An underutilised resource: historical flood
chronologies a valuable resource in determining periods of hydro-geomorphic
change, Sediment Dynamics and the Hydromorphology of Fluvial Systems,
Proceedings of a symposium in Dundee, IAHS Publication 306, 120–126.
|
Macdonald N, Phillips ID, Mayle G. 2010. Spatial and temporal
variability of flood seasonality in Wales. Hydrological Processes 24:
1806–1820.
Macdonald N, Werritty A, Black AR, McEwen LJ. 2006. Historical and
pooled flood frequency analysis for the River Tay at Perth, Scotland. Area 38(1): 34–46.
|
|
Malby AR, Whyatt JD, Timmis RJ, Wilby RL, Orr HG. 2007. Long term
variations in orographic rainfall: analysis and implications for upland
catchments. Hydrological Sciences
Journal 52(2): 276–291.
|
Marsh T. 2008. The 2007 floods in context, BHS 10th National Hydrology
Symposium, Exeter.
|
Marsh TJ, Dale M. 2002. The UK floods of 2000-2001: A
hydrometeorological appraisal. Water
and Environment Journal 16(3):
180–188.
|
Marsh TJ, Hannaford J. 2007. The Summer 2007 floods in England and
Wales - a hydrological appraisal, Centre for Ecology and Hydrology, p.32.
|
McEwen LJ. 2006. Flood seasonality and generating conditions in the Tay
catchment, Scotland from 1200 to present. Area
38(1): 47–64.
|
Milly PCD, Betancourt J, Falkenmark M, Hirsch RM, Kundzewicz ZW,
Lettermaier DP, Stouffer RJ. 2008. Stationarity is dead: Whither water
management. Science 319: 573–574.
|
Mudelsee, Borngen M, Tetzlaff G, Grunewald U. 2004. Extreme floods in
central Europe over the past 500 years: Role of cyclone pathway Zugstrasse Vb.
Journal of Geophysical Research -
Atmospheres 109: D23, D23101.
|
Mudelsee M, Deutsch M, Borngen M, Tetzlaff G. 2006. Trends in flood
risk of the River Werra (Germany) over the past 500 years. Hydrological Sciences Journal 51(5): 818–833.
|
O'Hare G, Sweeney J. 1993. Lamb's circulation types and British
weather: an evaluation. Geography 78: 43–60.
Osborn TJ. 2011. Winter 2009/2010 temperatures and a record breaking
North Atlantic Oscillation index. Weather
66(1): 19–21.
|
Petrow Th, Merz B, Lindenschmidt KE, Thieken AH. 2007. Aspects of
seasonality and flood generating circulation patterns in an mountainous
catchment in south-eastern Germany. Hydrology
and Earth System Sciences 11:
1455–1468.
|
Prudhomme C, Genevier M. 2010. Can atmospheric circulation be linked
to flooding in Europe? Hydrological
Processes DOI: 10.1002/hyp.7879.
|
Radziejewski M, Kundzewicz ZW. 2004. Detectability of changes in
hydrological records. Hydrological
Sciences Journal 49(1): 39–51.
|
Richter BD, Baumgartner JV, Wigington R, Braun DP. 1997. How much
water does a river need? Freshwater
Biology 27: 231–249.
|
Robson AJ. 2002. Evidence for trends in UK flooding. Philosophical Transactions of the Royal
Society of London: A 360: 1327–1343.
|
Roseveare N, Trapmore G. 2008. The sustainable regeneration of
Boscastle, BHS 10th National Hydrology Symposium, Exeter.
|
Rumsby BT, Macklin MG. 1994. Channel and floodplain response to recent
abrupt climate change: The Tyne basin, Northern England. Earth Surface Processes and Landforms 19: 499–515.
|
Rumsby BT, Macklin MG. 1996. River response to the last neoglacial
(the little ice age) in northern, western and central Europe, 217-233. In Global continental changes: the context of
palaeohydrology, Branson J, Brown AG, Gregory KJ (eds), Geological
Society Special Publication No. 115, Geological Society, London.
|
Smith K, Tobin GA. 1979. Topics in Applied Geography: Human Adjustment
to the Flood Hazard, Longman: London.
Smith K, Tobin GA. 1979. Topics in Applied Geography: Human Adjustment
to the
Flood Hazard, Longman: London.
|
Svensson C, Kundzewicz ZW, Maurer T. 2005. Trend detection in river
flow series: 2: Flood and low flow index series. Hydrological Science Journal 50(5):
811–823.
|
Sweeney JC, O'Hare GP. 1992. Geographical variations in precipitation
yields and circulation types in Britain and Ireland, Transactions of the Institute of British Geographers 17: 448–463.
|
Wedgbrow CS, Wilby RL, Fox HR, O'Hare G. 2002. Prospects for seasonal
forecasting of summer drought and low river flow anomalies in England and
Wales. International Journal of
Climatology 22: 219–236.
|
Werritty A. 2002. Living with uncertainty: climate change, river flows
and water resource management in Scotland. The Science of the Total Environment 294: 29–40.
|
Werritty A, Black AR, Duck RW, Finlinson W, Thurston N, Shackley S,
Crichton D. 2002. Climate change: flooding occurences review. Central
Research Unit Scottish Executive, Edinburgh.
|
Wheater HS. 2006. Flood hazard and management: a UK perspective,
Philosophical Transactions of the Royal Society of London: A. Mathematical, Physical and Engineering
Sciences 364: 2135–2145.
|
Wilby RL, O'Hare G, Barnsley N, 1997, The North Atlantic Oscillation
and British Isles climate variability, Weather, 52, 266-276.
Wilkinson ME, Quinn PF, Welton P. 2010. Runoff management during the
September 2008 floods in the Belford catchment, Northumberland. Journal of Flood Risk Management 3(4): 285–295.
|
Yarnal B. 1993. Synoptic climatology in environmental analysis,
Belhaven Press: London.
|
Yarnal B, White DA. 1987. Subjectivity in a computer assisted synoptic
climatology 1: classification results. International
Journal of Climatology 7: 119–128.
|
Figure 1.          Map of the Eden Catchment.
Figure 2. Gauged records from the Eden at
Sheepmount in Carlisle (a) POT; (b) AMax; (c) Magnitude of POT events.
Figure 3.          Cumulative
number of extreme floods as a function of time.
Figure 4.          Percentage of floods since 1978 which
have occurred on days of particular Lamb weather types.
Figure 5.          Percentage of floods which have
occurred on days and preceding days of particular Lamb weather types: (a) high
flows; (b) extreme flood events.
Figure 6.          Percentage of the year classified as
each Lamb weather type for the 1976–2007 period.
Figure 7.          Probability
that a day with a particular Lamb weather type will also have an >Q1 event
occurring.
Figure 8.          Plot
showing how the proportion of the year classified as the five extreme flood-generating
weather types and extreme flood frequency have changed over time (dashed line
represents the cumulative extreme flood record, solid line represents weather
types).
Figure 9. Plot showing how individual extreme
flood-generating weather types have changed over time: (a) CW; (b) CSW; (c) C; (d)
W; and (e) SWÂ (dashed line represents
the cumulative extreme flood record, solid line represents weather types).
Figure 10.        Percentage of floods since 1978 which
have occurred on days of particular Lamb weather types: (a) 1978–1987; (b) 1988–1997;
(c) 1998–2007.
Table I.           Summary of record length
requirements from the literature for identifying trends in flood records.
Reference
|
Record length needed to
identify trends
|
Interagency Advisory Committee on Water Data 1982
|
10 years
|
Richter et
al., 1997Â
|
20 years
|
Lettenmaier et
al., 1994; Lins and Slack, 1999; Douglas et al., 2000Â
|
at least 30 years
|
Gan et al., 1991; Huh et al.,
2005
|
at least 40-year records
|
Kundzewicz and Robson 2000, 2004
|
at least 50 years
|
Table II.        Percentage of events of each sequence
(day of event, day before, 2 days before) of event-generating weather types. 1
= event-generating weather type (C, W, SW, CW, CSW); 0 = day with another
weather type.
Sequence
|
% of extreme floods
|
% of high flows
|
% of all events
|
1
1 1
|
47.7
|
53.2
|
51.4
|
1
1 0
|
27.3
|
21.3
|
23.2
|
1
0 0
|
0.0
|
8.5
|
5.8
|
1
0 1
|
6.8
|
10.6
|
9.4
|
0
1 1
|
4.5
|
3.2
|
3.6
|
0
1 0
|
2.3
|
1.1
|
1.4
|
0
0 1
|
9.1
|
0.0
|
2.9
|
0
0 0
|
2.3
|
2.1
|
2.2
|
Â
Table
III. A complete record of the extreme floods in Carlisle showing the source of
the information.
|
BCHE
|
Newspapers
|
Epigraphic
|
Smith & Tobin
|
Warwick Bridge
|
Sheepmount
|
1771
|
Y
|
|
|
|
|
|
1773
|
Y
|
|
|
|
|
|
1781
|
Y
|
|
|
|
|
|
1783
|
Y
|
|
|
|
|
|
1794
|
Y
|
|
|
|
|
|
1794
|
|
Y
|
|
|
|
|
1803
|
Y
|
|
|
|
|
|
1804
|
Y
|
|
|
|
|
|
1808
|
Y
|
|
|
|
|
|
1809
|
Y
|
Y
|
|
Y
|
|
|
1809
|
Y
|
Y
|
|
Y
|
|
|
1815
|
Y
|
Y
|
|
Y
|
|
|
1815
|
Y
|
|
|
|
|
|
1818
|
|
Y
|
|
|
|
|
1821
|
|
Y
|
|
|
|
|
1822
|
Y
|
Y
|
Y
|
Y
|
|
|
1851
|
Y
|
Y
|
|
Y
|
|
|
1852
|
Y
|
|
|
Y
|
|
|
1856
|
Y
|
Y
|
Y
|
Y
|
|
|
1858
|
|
Y
|
|
|
|
|
1868
|
Y
|
|
Y
|
Y
|
|
|
1868
|
Y
|
|
|
|
|
|
1870
|
Y
|
|
|
|
|
|
1874
|
Y
|
|
|
|
|
|
1874
|
Y
|
|
|
|
|
|
1874
|
Y
|
|
|
Y
|
|
|
1875
|
Y
|
Y
|
|
|
|
|
1876
|
Y
|
|
|
|
|
|
1876
|
Y
|
|
|
|
|
|
1876
|
Y
|
|
|
|
|
|
1877
|
Y
|
|
|
|
|
|
1877
|
Y
|
|
|
|
|
|
1877
|
Y
|
|
|
|
|
|
1877
|
Y
|
|
|
|
|
|
1877
|
Y
|
|
|
|
|
|
1878
|
Y
|
|
|
|
|
|
1878
|
Y
|
|
|
|
|
|
1880
|
Y
|
|
|
|
|
|
1881
|
Y
|
|
|
|
|
|
1881
|
Y
|
|
|
|
|
|
1881
|
Y
|
|
|
|
|
|
1882
|
Y
|
|
|
|
|
|
1882
|
Y
|
|
|
|
|
|
1883
|
Y
|
|
|
|
|
|
1883
|
Y
|
|
|
|
|
|
1883
|
Y
|
|
|
|
|
|
1885
|
Y
|
|
|
|
|
|
1885
|
Y
|
|
|
|
|
|
1890
|
Y
|
|
|
|
|
|
1891
|
Y
|
|
|
Y
|
|
|
1891
|
Y
|
|
|
Y
|
|
|
1891
|
Y
|
|
|
Y
|
|
|
1891
|
Y
|
|
|
Y
|
|
|
1891
|
Y
|
|
|
Y
|
|
|
1892
|
Y
|
Y
|
|
Y
|
|
|
1894
|
Y
|
|
|
|
|
|
1895
|
Y
|
|
|
|
|
|
1896
|
Y
|
|
|
Y
|
|
|
1898
|
Y
|
|
|
Y
|
|
|
1899
|
Y
|
|
|
Y
|
|
|
1899
|
Y
|
|
|
|
|
|
1900
|
Y
|
|
|
|
|
|
1903
|
Y
|
|
|
Y
|
|
|
1903
|
Y
|
Y
|
|
Y
|
|
|
1903
|
Y
|
Y
|
|
|
|
|
1904
|
Y
|
|
|
|
|
|
1907
|
|
Y
|
|
|
|
|
1914
|
Y
|
|
|
Y
|
|
|
1914
|
Y
|
|
|
Y
|
|
|
1916
|
Y
|
|
|
Y
|
|
|
1918
|
Y
|
Y
|
|
Y
|
|
|
1921
|
Y
|
|
|
Y
|
|
|
1924
|
Y
|
|
|
Y
|
|
|
1924
|
Y
|
|
|
Y
|
|
|
1925
|
Y
|
Y
|
Y
|
Y
|
|
|
1926
|
Y
|
Y
|
|
Y
|
|
|
1926
|
Y
|
Y
|
|
Y
|
|
|
1927
|
Y
|
|
|
|
|
|
1928
|
Y
|
|
|
Y
|
|
|
1928
|
Y
|
|
|
Y
|
|
|
1929
|
Y
|
Y
|
|
Y
|
|
|
1929
|
Y
|
|
|
Y
|
|
|
1930
|
Y
|
|
|
Y
|
|
|
1931
|
Y
|
|
|
Y
|
|
|
1931
|
|
|
|
Y
|
|
|
1932
|
|
|
|
Y
|
|
|
1933
|
|
|
|
Y
|
|
|
1933
|
|
|
|
Y
|
|
|
1938
|
|
Y
|
|
|
|
|
1941
|
|
Y
|
|
Y
|
|
|
1945
|
|
Y
|
|
|
|
|
1947
|
|
Y
|
|
Y
|
|
|
1947
|
|
Y
|
|
|
|
|
1952
|
|
|
Y
|
|
|
|
1954
|
Y
|
|
|
Y
|
|
|
1954
|
|
Y
|
|
Y
|
|
|
1954
|
|
Y
|
|
Y
|
|
|
1954
|
|
Y
|
|
Y
|
|
|
1956
|
|
Y
|
|
|
|
|
1958
|
|
Y
|
|
|
|
|
1962
|
|
Y
|
|
|
Y
|
|
1964
|
|
Y
|
|
Y
|
Y
|
|
1965
|
|
|
|
|
Y
|
|
1966
|
|
Y
|
|
|
Y
|
|
1967
|
|
Y
|
|
|
Y
|
Y
|
1967
|
|
Y
|
|
|
Y
|
Y
|
1968
|
Y
|
Y
|
Y
|
Y
|
Y
|
Y
|
1972
|
|
Y
|
|
|
Y
|
Y
|
1979
|
|
|
|
|
Y
|
Y
|
1979
|
|
|
|
|
Y
|
Y
|
1982
|
|
Y
|
|
|
Y
|
Y
|
1985
|
|
|
|
|
Y
|
Y
|
1987
|
|
Y
|
|
|
Y
|
Y
|
1990
|
|
Y
|
|
|
Y
|
Y
|
1991
|
|
|
|
|
Y
|
Y
|
1991
|
|
|
|
|
Y
|
Y
|
1995
|
|
Y
|
|
|
Y
|
Y
|
1995
|
|
|
|
|
Y
|
Y
|
1997
|
|
Y
|
|
|
Y
|
Y
|
1997
|
|
|
|
|
Y
|
Y
|
1998
|
|
|
|
|
Y
|
Y
|
1999
|
|
|
|
|
Y
|
Y
|
1999
|
|
|
|
|
Y
|
Y
|
2000
|
|
|
|
|
Y
|
Y
|
2000
|
|
|
|
|
Y
|
Y
|
2000
|
|
|
|
|
Y
|
Y
|
2002
|
|
|
|
|
Y
|
Y
|
2002
|
|
|
|
|
Y
|
Y
|
2003
|
|
|
|
|
Y
|
Y
|
2004
|
|
|
|
|
Y
|
Y
|
2004
|
|
|
|
|
Y
|
Y
|
2004
|
|
|
|
|
Y
|
Y
|
2005
|
|
Y
|
|
|
Y
|
Y
|
2005
|
|
Y
|
|
|
Y
|
Y
|
2006
|
|
|
|
|
Y
|
Y
|
2006
|
|
|
|
|
Y
|
Y
|
2006
|
|
|
|
|
Y
|
Y
|
Add a comment