Changing patterns of heavy rainfall in upland areas: a case study from northern England



There has recently been a widespread shift in the pattern of UK rainfall towards more heavy falls of rain in winter and fewer in summer. Here, this change is examined in the context of orographic enhancement for a transect of rain gauges running across northern England from coast to coast and including both the Lake District and Pennine uplands. Gauges have been selected where very long records of daily rainfall exist; where data are missing, these have been infilled using data from nearby gauges. The very long records for Armagh and Durham are also included to provide additional context in time and space. For the upland gauges, the increase in total winter rainfall in recent decades and the simultaneous decrease in total summer rainfall are reflected in the number of heavy falls of rain, as defined using two threshold indices. The 1990s saw record numbers of heavy falls in winter and an almost complete absence of heavy summer rainfall in the uplands, in marked contrast to lowland gauges. Comparison of the rainfall record with the Lamb Weather Catalogue suggests that increased winter rainfall is related to an increase in the rainfall provided by westerly weather types. Decreased summer rainfall is related to a reduction in rainfall associated with cyclonic weather types. The results presented here underline the value of long-term monitoring and the maintenance of records from key historic sites. Copyright © 2010 Royal Meteorological Society

1. Introduction

There is now clear and unequivocal evidence of global warming since the mid-19th century (Trenberth et al., 2007), very often reflected at regional scales. For example, Hulme and Jenkins (1998) identified warming of 0.5 °C in the UK Central England Temperature record (Manley, 1974; Parker et al., 1992) during the 20th century. Since increased air temperatures will lead to increased evaporation and higher water vapour content in the atmosphere, precipitation must inevitably increase too, although it is recognised that the spatial and temporal response will be complex (Meehl et al., 2007). Most climate models predict increases in both rainfall frequency and intensity at high latitudes in the Northern Hemisphere (e.g. McGuffie et al., 1999; Semenov and Bengtsson, 2002; Meehl et al., 2007), including northern Europe, where increases in winter rainfall (predominantly from frontal systems) are likely to contribute to higher annual rainfall totals, although total summer rainfall (mainly convective) may decrease somewhat (Christensen et al., 2007).

Fowler and Kilsby (2007) note that downscaled model output for northern England indicates an increase of 20–30% in mean monthly rainfall in winter (November to March) with a reduction of up to 50% in summer (May to September), and use these results to project changes in annual runoff regime in eight catchments in northwest England. Of course, a reduction in summer rainfall totals does not necessarily mean that intense rainfall events will become less frequent; indeed, convective storms might become more common in a warmer atmosphere (Lane, 2008). Fowler and Ekström (2009) provide multi-model ensemble estimates of the impact of climate change on UK seasonal precipitation extremes. Extreme precipitation seems likely to increase across the UK in winter, spring, and autumn in the range 5–30%. In summer, there is low confidence due to poor model performance; both increases and decreases in extreme precipitation are predicted. Regional results for the 1-day 5-year return period for northeast England (NEE) and northwest England (NWE) mirror national results with significant percentage increases of about 20% in both regions in winter, but only for NWE in summer (∼10%).

Any study of rainfall regime in hilly terrain—past or future—must include consideration of orographic enhancement. Whilst high-resolution numerical models are now regularly used in climate studies (e.g. Bromwich et al., 2005), attempts to downscale rainfall predictions from regional climate models do not currently include explicit adjustment for orographic effects; this is usually addressed by the use of statistical downscaling techniques. As the resolution of these techniques improves, it will be necessary to develop modelling approaches to estimate the increase in lowland rainfall rate at upland sites, similar in scale to the UK Meteorological Office's FRONTIERS system (Browning, 1979; Collier, 2002), for example. The particular needs of flood prediction will require consideration of short-term rainfall intensities as well as longer-period amounts. Any predictive methods must be tested against the context of the historical observational record. This issue is addressed here by studying historical changes in heavy falls of rain in upland areas as indicated by daily totals. This is too long a time scale for local floods, but in terms of flood generation in medium to large catchments, daily totals are a reasonable aggregation. General trends in the contribution of heavy rainfall-to-precipitation totals in the UK have been studied by Osborn et al. (2000), Osborn and Hulme (2002), Burt and Horton (2007) and Maraun et al. (2008). For daily totals in the UK, the clear climatic signal over recent decades has been a widespread shift towards more frequent and greater contribution from heavy falls in winter, and towards less frequent and lower contribution from heavy falls in summer. Osborn et al. (2000) noted that the upward trend in mean winter precipitation in the UK over the period 1961–1995 is strongest in the western and upland regions and that the change might be partly driven by increased westerly zonal flow as measured by the North Atlantic Oscillation (NAO). Several papers have since shown how changes in the pattern of UK rainfall at the decadal timescale relate to changes in the atmospheric circulation. Murphy and Washington (2001) use a simple index representing the North Atlantic surface-level pressure field to explain changes in 20th century UK precipitation, including a tendency for wetter winters and drier summers in the last two decades. Hannaford and Marsh (2008) show that high-flow indicators for many northern and western UK river basins are correlated with the NAO, with recent changes probably caused by a more positive NAO since the 1960s. Mayes (1996) explained the geographical distribution of monthly rainfall fluctuations in terms of interactions between atmospheric circulation and topography, a theme explored in more detail by Malby et al. (2007). Fowler and Kilsby (2002) found strong correlations between winter rainfall and the NAO, especially for a gauge in the Pennine Hills (Moorland Cottage) just to the south of the transect examined in this paper.

Orographic rainfall enhancement on the hills of northern England has been studied for many decades, whether in relation to topographic controls on the distribution of rainfall (e.g. Chuan and Lockwood, 1974; Burt, 1980), links with atmospheric circulation (e.g. Shaw, 1962), or meteorological processes (e.g. Collinge et al., 1989; Dore et al., 1992). The purpose here is to focus on one aspect of the rainfall regime in upland areas: the frequency of heavy falls of rain, their contribution to seasonal and annual totals, and changes in heavy rain in relation to atmospheric circulation. This develops the approach of Osborn and colleagues (op. cit.) but with explicit attention being paid to upland areas. It also builds on the work of Malby et al. (2007) who demonstrated a clear change in the seasonality of orographic rainfall in the Lake District of northern England and speculated that this was related to the incidence of higher-intensity events. Wherever possible, the analysis has been extended back beyond the period 1961–2000 in order to examine longer-term changes, taking advantage of some very long daily rainfall records in the region. Changes observed are related to variations in atmospheric circulation, as measured by Lamb Weather Types.

2. Data sources and methods

2.1. Study area

The analysis is based on a transect of rain gauges running approximately east–west across northern England from coast to coast, including the upland regions of the Lake District and the Pennine hills (Figure 1). Because of uncertainty as to whether the western coastal site at Barrow is already influenced by advanced uplift over the Lake District massif (c.f. Hill et al., 1981), we have also included the Armagh gauge from Northern Ireland, giving the added advantage of being able to incorporate another very long record from the mid-nineteenth century. Admittedly, Armagh is 225 km west of Barrow, but it is hard to identify a better ‘upwind’ gauge: those on the Isle of Man would likely be affected by orographic enhancement, and sites further south on the Irish coast are just as far away as Armagh and equally influenced by their location in the east of Ireland.

Figure 1.

The location of the rain gauges included in this study

All data unless specified were taken from the MIDAS land surface dataset available from the British Atmospheric Data Centre (BADC) (MIDAS, 2008). From west to east, the full set of sites included is as follows (with UK Meteorological Office (UKMO) station numbers in brackets where appropriate):

  • Armagh: Daily data are available on the British Atmospheric Data Centre (BADC) website from 1838. We have used data from 1851 to 2008 inclusive. Full details are provided in Butler et al. (1998).

  • Barrow: This is a composite series based mainly on Barrow Sandgate sewage treatment works (#12 740), but infilled after 1960 with Barrow gasworks (#12 739) or Walney Island (#12 741) where gaps exist. For April 1984, since there were no data from the Barrow locality at all, this one month was infilled using data from Stainton Croft (#13 069, 6 km distant). Thus, there is a complete series from 1961 to the end of July 2007. There are 3 years missing in the 1950s, so we do not quote results from that decade. With the exception of 1933, complete data are available for the 1920s to the 1940s inclusive.

  • Ulpha Duddon water works (#12 759): This series runs from 1 January 1935 to 31 July 2007. The only gaps (January 1975; January, February 1983) have not been infilled as there were no available records within a few kilometres.

  • Coniston Holywath (#12 722): Although the record starts in 1910, there are no data for the 1920s and the record does not resume until 1937 but is complete thereafter. Our decadal analysis starts with the 1940s, therefore. The record runs through to 31 July 2007.

  • Ambleside: This is a composite record based from 1932 to 1959 on a complete record from Ambleside (#1077). Records from Rydal Hall (#12 683, 2 km distant) are used to infill the 1960s except for 1964 which is the High Wray (#12 688, 3.6 km distant) record. From 1971, the record is largely that for Brathay (#12 673, 1.5 km distant) although Elterwater (#12 670, 4 km from Brathay) is used for September and October 1988. The only remaining gaps are January 1973 and July 1975. Again, the record runs to 31 July 2007.

  • Holehird (#12 695): This record starts in 1903 and is complete to 1974. Gaps thereafter are infilled using nearby Dubbs Reservoir (#12 616, 1.5 km distant). Only 1992 remains problematic with seven months missing; data completeness for the 1990s is, therefore, only 94%. There are many gaps after November 2002 and so no data are included in our analysis after the 1990s.

  • Appleby Castle (#12 936): This very long record is listed on the UKMO website ( Data are complete from 1891 to 2000 with the exception of September and October 1963 which were infilled from a nearby station (Appleby Bongate #12 937). Data from April 2000 to February 2005 are from Appleby Mill Hill (#30 005). Both stations are within 2 km of the castle.

  • Burnhope Reservoir: This series, supplied by the UK Environment Agency, starts in 1923, but given that most of the data for 1930 are missing, we only include decadal values from the 1930s onwards. The series is complete through to February 2008. Originally, we hoped to include data from the Moor House National Nature Reserve as a complement to Burnhope, Moor House being higher and nearer the Pennine escarpment. However, the use of an automatic weather station at this remote location since 1991 puts some of the heavy winter totals in doubt, so we have not tabulated results for Moor House, although some reference is made to summer rainfall there, since we have greater confidence in those data.

  • Durham (#326): This very long record is complete from 1850; decadal analysis here runs from 1851. Full details are available in Burt and Horton (2007). Since that paper was published, a homogenous daily record for the period December 1871 to July 1879 has been derived (based on Burt, 2009) and the only gaps (July, September, October, December 1854; January, February 1855) infilled using a comparison with Armagh (R2 = 35%, n = 2986, p > 0.0001).

As far as we know, except in the 19th century, all the rain gauges would have been standard UK Meteorological Office (UKMO) design, with an aperture 127 mm diameter 305 mm above the ground (except for Durham since 1999, where data are from a tipping bucket rain gauge (orifice diameter 200 mm; height 340 mm). Given that tall gauges are more susceptible to loss of catch during strong winds, it is fortunate that British gauges are not taller, even if there is the risk of splash (Rodda et al., 1976). Measurements of the collected rain are made at 0900 GMT daily using a glass measuring cylinder. The glass bottle inside the gauge can hold the equivalent of 100 mm rainfall; the overspill bucket (brass) can hold a further 75 mm. We have no evidence that the greater incidence of rainfall and lesser incidence of snowfall in more recent (warmer) years have affected the results presented here. Given the difficulty of measuring snowfall, it is possible that any increase in winter rainfall would be as a result of more rainfall, rather than any change in total precipitation. However, this would not be a factor in changing summer rainfall totals.

2.2. Methodology

Analysis of heavy falls of rain, as indicated by daily totals, follows the analyses of Osborn et al. (2000), Osborn and Hulme (2002), Burt and Horton (2007) and Maraun et al. (2008). In Osborn et al. (2000) a ‘category (10)’ threshold was identified: this is the daily rainfall total above which the top 10% of total rainfall has occurred; here we call this the T10 threshold. This value is derived by ranking all daily rainfall data, cumulating the totals, and identifying the group of highest daily totals that together contribute 10% of the total precipitation over the reference period (1961–1990). Following Osborn et al. (2000), the cumulative rainfall totals are only calculated for rain days; a rain day is defined as a total of 0.25 mm or more and all values below this are excluded; for a very long daily rainfall series, only a very small fraction of total rainfall is thereby excluded. The T10 threshold is calculated at all sites using data for 1961–1990; this follows the convention of the WMO and UKMO whereby climate averages are for the period 1961–1990. In addition, in a similar approach to Karl and Knight (1998), a second threshold has been identified: R0.25, the daily total that is equalled or exceeded on 0.25% of all days (i.e. on average, once every 400 days or approximately once per year). In these cases, all days are included in the analysis, including those with 0.25 mm or less; this mirrors analysis of flow duration curves for river discharge (Ward and Robinson, 1990).

In order to understand the synoptic conditions associated with rainfall events and investigate how these may relate to the observed changes in rainfall trends, the daily rainfall data were compared with the objective Lamb Weather Catalogue (Jones et al., 1993) which provides a daily record of synoptic weather conditions and air mass origins over the British Isles. The percentage contribution each Lamb Weather Type (LWT) made to winter and summer rainfall totals were calculated at each station from 1940 to 2005, a period comparable across all stations. Similarly, T10 rainfall events were classified by LWT and the percentage contribution these events made to winter and summer rainfall totals were calculated for each station. Ambleside and Holehird stations were not included in this analysis given the gaps in their records. The comparison of the objective Lamb Weather Catalogue with daily observations is a simple method to assess whether there have been any changes to the rainfall amount provided by a particular weather type. A thorough evaluation of the advantages and limitations of the Lamb typing scheme is given by O'Hare and Sweeney (1993).

In terms of decadal analysis, we have taken decades to run from year 1 to year 0; for example, the years 1981 to 1990 are assigned to the 1980s. This follows the convention of the WMO and UKMO whereby climate averages are for the period 1961–1990. We recognise that choice of decade does affect the resulting totals: this is not a major problem for rainfall totals, less than 5% difference at Durham, for example. However, for numbers of extreme events per decade, where totals are much smaller, the difference can be larger: at Durham, for the T10 index, the difference was up to 38% although less than 15% for 12 decades out of 15. In all cases, the same decade is used at each gauge, so there should not be a problem in comparing totals between sites. Seasons were defined using UKMO convention: winter is December (of the previous year) to February, spring is March to May, summer is June to August, and autumn is September to November. In most cases, as noted above, the datasets are complete; we have only quoted decadal values where records are more than 90% complete; for ‘incomplete’ decades, actual amounts are recalculated pro rata to estimate the value for a complete 10-year period.

Figure 1 shows the location of sites included in this study. Figure 2 shows the relationship between orography and rainfall across the transect. Since gauge altitude is not necessarily a good indication of annual rainfall totals, in particular where the site is located in a deep valley surrounded by high ground, the maximum altitude within 2 km of the rain gauge (MAX2km) has been used to give some indication of the effective height of the surrounding area; using gauge altitude in an upland area may give a false impression the general form of the topography since many gauges are located in valleys (Salter, 1918; Chuan and Lockwood, 1974; Burt, 1980). Average annual rainfall (1961–2000) follows the MAX2km topography across the transect ranging from 652 mm at Durham to 2482 mm at Coniston Holywath (Figure 2). Note, however, that average annual rainfall for the same elevation is lower over the Pennine hills, which lie in the rain shadow of the Lake District. The T10 and R0.25 threshold totals and their relationship to MAX2km topography are also shown on Figure 2 for each gauge. As with average annual rainfall, the T10 and R0.25 threshold totals are closely related to MAX2km topography and are greatest in the Lake District uplands (i.e. Ulpha, Coniston, Ambleside, and Holehird gauges).

Figure 2.

The relationship between the different thresholds, the MAX2km orography, and the annual average precipitation along the study transect

3. Results

3.1. Variations in winter and summer rainfall totals across the transect

As noted above, a number of studies have identified a trend towards increased winter rainfall and decreased summer rainfall in the UK in recent decades; for a recent analysis, see Maraun et al. (2008). This is illustrated in Figure 3a and b, which show rainfall totals from the 1940s to the present relative to the average for the period 1961–1990 (which has been chosen as the baseline comparator c.f. UKMO and IPCC). Strong increases in winter rainfall in the 1980s and 1990s are observed for all the upland gauges (Ulpha, Coniston, Ambleside, and Burnhope Reservoir) and at Appleby. The effect is not seen at the lowland gauges, Barrow and Durham, although a very modest increase is seen at Armagh in the 1990s. It is interesting that the largest ratios in the 1990s are towards the east: 1.3 at Holehird and Burnhope and 1.25 at Appleby. These results confirm and extend the findings of Malby et al. (2007) who noted an increase in rainfall across the region from the 1970s to the 1990s, with greatest increase occurring at elevation and to the east of the Lake District hills. Table I puts the rainfall of recent decades in a historical context and lists the average winter rainfall totals by decade from 1850 where available. These data show the highest average decadal winter rainfall totals for all upland stations and Appleby were recorded in the 1990s, although Appleby and Holehird also show similar totals for the 1910s. The highest average rainfall totals occur prior to the 1940s for Durham and Armagh. Thus, while it is impossible to be sure, the high totals at the end of the 20th century do seem unusually high.

Figure 3.

a, Decadal winter rainfall totals relative to the 1961–1990 baseline. b, Decadal summer rainfall totals relative to the 1961–1990 baseline

Table Ia. Average winter rainfall totals by decade.
1850s181       135
1860s261       144
1870s238       160
1880s189       129
1890s182     213 128
1900s199    531246 107
1910s247    643252 162
1920s229302   596241370151
1930s206288596 548532226400173
1950s215 609754594536154401161
2000s165268553822645 217433142
Table Ib. Average summer rainfall totals by decade.
1850s229       209
1860s197       172
1870s243       193
1880s218       188
1890s253     248 188
1900s223    380209 157
1910s202    354231 181
1920s245271   437266308210
1930s237247453 335380212257185
1950s221 487589429420232283176
2000s221184316431310 194251164

In contrast to winter, a decline in summer rainfall is seen at all gauges, with some recovery in the 2000s from Appleby eastwards (Figure 3b). Historical analysis (Table Ib) shows the highest annual decadal summer rainfall totals occurred prior to 1950, and that the 1990s was the driest decade at all stations (excluding the estimates for Barrow and Ulpha for the 2000s). In combination, the result is a strong increase in the ratio of winter-to-summer rainfall from the 1940s to the 1990s at all sites except Durham (Figure 4), which is also the only station here to have consistently more rainfall in summer than winter. Some of the ratio increases from the 1960s to the 1990s are remarkably large, most notably at Ambleside, Appleby and Burnhope Reservoir; Appleby having switched from having more rain falling in summer to more rain falling in winter after the 1960s. In general, the biggest increases in winter-to-summer ratio occur at the higher gauges, and at Appleby. It is interesting to note that the response at Durham is very different to the other sites. Other than the very wet 1870s (Burt, 2009), Durham summers tended to be much wetter than winters during the latter part of the 19th century whereas in the 20th century, summers have tended to become drier, and winters have become gradually wetter (Burt and Horton, 2007).

Figure 4.

Ratios of winter to summer rainfall from the 1940s

3.2. Changes in the frequency of heavy falls of rain

3.2.1. T10 events

Table II lists the frequency of T10 events by decade for winter (Table IIa) and summer (Table IIb). In general, the upland sites (Ulpha, Coniston, Ambleside, Burnhope Reservoir) have more T10 events in winter compared to summer for the duration of their records. In contrast, there are more T10 events in summer at Durham and Armagh. Although the pattern is not as clear as for seasonal rainfall totals (Table Ia and b), there was a doubling (indeed, tripling at Burnhope Reservoir) in frequency of winter T10 events at Coniston, Ambleside, Holehird, Appleby, and Burnhope Reservoir from the 1970s to the 1990s, when the greatest number of T10 events occurred at all these sites. This accords with the observation of Robson et al. (1998) that the 1970s were comparatively dry compared to the flood-rich 1960s and 1980s. The 1970s to 1990s increase in winter T10 events is not observed at the lowland sites of Armagh, Barrow, and Durham, nor indeed at Ulpha (which is on the western edge of the Lake District upland). This suggests that, as with the change in decadal totals (Section 3.1), the greatest change in rainfall patterns is occurring towards the Pennine hills. It is notable that there were drier than average winters in the 1960s at stations from Barrow to Burnhope Reservoir (Table I), but a greater than average number of T10 events (some of which would have fallen as snow).

Table IIa. Frequency of T10 events in winter by decade.
1850s3       1
1860s14       3
1870s9       7
1880s0       3
1890s3     8 2
1900s4    119 0
1910s9    185 0
1920s57   139 2
1930s48  8158111
1950s10 11128103142
2000s6741919 7161
Table IIb. Frequency of T10 events in summer by decade.
1850s15       10
1860s11       13
1870s10       5
1880s11       16
1890s20     9 7
1900s14    54 2
1910s12    612 6
1920s713   99 13
1930s1412  39866
1950s11 12948775
2000s1310333 8711

Although the 1990s recorded the most winter T10 events at Coniston, Ambleside, Appleby, and Burnhope Reservoir, this same period saw the fewest number of summer T10 events at these sites and indeed none at all at Ambleside (Table IIb). In fact, the 1990s saw lower than average (as calculated over the duration of the entire record) frequencies of summer T10 events at all stations. This shift from summer to winter rainfall seems to have reversed to some extent in the 2000s: there have been more summer T10 events and a few less T10 winter events, although the number of winter events has remained high at Coniston and Ambleside. Maraun et al. (2008) conclude that the positive trends in heavy winter rainfall reported in earlier papers (Osborn et al., 2000; Osborn and Hulme, 2002) have continued since 1995, whereas the negative summer trends in heavy summer precipitation are more indicative of inter-decadal variability.

The long time series of data available for this study allows the 1990s experience of a large number of winter T10 events and a low number of summer T10 events to be considered in a longer historical context than in most other studies. The 1990s winter T10 total at Appleby (12) is the highest on record, but totals of 8 or 9 were common before the 1940s. Only three T10 events occurred in the 1940s and 1950s at Appleby (Table IIa). Lane (2008) has argued that the middle of the 20th century was ‘flood-poor’ and all stations except Durham record lower than average winter T10 frequencies during the 1940s. In addition Ulpha, Ambleside, and Appleby also record lower than average winter frequencies of T10 events in the 1950s, offering some support to Lane's argument. The longest records of Armagh and Durham show significant variation in the frequency of winter T10 events: other than the 1910s, Armagh recorded fewer than average winter T10 falls from 1880s–1940s; Durham recorded fewer than average winter T10 falls 1890s–1930s. Some more recent decades have recorded higher frequencies of winter T10 events but there is no upward trend. The maximum numbers of winter T10 falls within any decade were recorded prior to 1880 at both stations.

Summer T10 events reveal a more variable pattern (Table IIb). There were fewer than average T10 events during the 1970s (except at Holehird) and 1990s. Only Burnhope and Durham had below-average totals in the 1980s. Generally, the 2000s have shown a swing to wetter times with higher numbers of T10 events at all gauges. The increase in frequency of winter T10 events combined with the reduction in frequency of T10 summer events at upland stations led to a dramatically increased winter : summer ratio in the 1990s (Figure 5). Two stations (Ulpha, Appleby) show a shift in dominance of T10 events from summer to winter from the 1940s to the 1990s. The greatest increases in the winter:summer ratio are observed at the upland sites of Ulpha, Coniston, Ambleside, and Burnhope. Results for the present decade, albeit incomplete, suggest a reduction in the winter:summer ratio of T10 events at most sites; this shift is driven by the increase of summer T10 events in the 2000s. The lowland stations, Armagh, Barrow and Durham, show no clear pattern of change in the ratio of summer to winter T10 rainfall events from the 1940s to the 1990s. At Durham, where heavy falls in summer dominate the annual T10 totals, ratios tended to be lower prior to the 1940s with the exception of the 1890s (0.4) and 1870s (0.9) and have tended to fall again since the 1940s.

Figure 5.

Winter:summer ratios for T10 decadal totals for selected gauges. Note that the 1990s values for Ambleside assumes a summer total of 1 rather than zero

Osborn and Hulme (2002) examined the contribution of T10 events to total rainfall as well as the frequency of T10 events. Table III shows the contribution of T10 events for a selection of upland gauges. There is no consistent pattern by location or decade. Burnhope shows both the lowest contribution (1960s) and the highest (1990s); the latter observation may accord with the large increase in T10 events towards the east, as already described, although the opposite effect is seen at Appleby.

Table III. The contribution of T10 events to total rainfall by site and decade
1890s    8.9 
1900s   9.26.6 
1910s   11.510.1 
1920s   11.89.4 
1930s  8.913.19.410.7

3.2.2. Annual T10 data

All the data presented so far have been decadal totals. Figure 6 presents annual totals of T10 events at Burnhope from 1932 to 2007 in order to illustrate variability at the annual time scale. There was a relative absence of heavy falls in winter in the 1940s and the period 1963–1976, but the 1950s had more heavy falls in winter. Numbers of heavy winter falls increased from 1976 throughout the 1980s and 1990s, with the most T10 falls (5) being recorded in the 1995 and 2000 winters. One summer T10 event per summer was more common in the early part of the record, but from 1975 to 2000 there were many summers without any.

Figure 6.

Annual totals of T10 events at Burnhope (1932–2007)

3.2.3. R0.25 events

As might be expected, R0.25 events are more common in summer at the ‘lowland’ sites: Armagh, Barrow, Appleby, and Durham. Table IVa shows that there were higher than average numbers of R0.25 events in winter at all gauges from Barrow across to Appleby in the 1960s. Numbers were lower in the 1970s for all upland gauges except Burnhope, and then rose in the 1980s and especially in the 1990s: numbers at Ulpha, Coniston, Ambleside, Holehird, and Burnhope were all well above average in the 1990s, and a big contrast with the low 1970s totals. Numbers have stayed high into the 2000s at Coniston, Ambleside, and Burnhope (no data for Holehird) but none have been recorded at Barrow or Ulpha. Generally, numbers of R0.25 events in summer were higher up to and including the 1960s for the upland gauges. As with T10 events, there were few summer R0.25 events in the 1990s at the upland gauges and at Appleby, whereas numbers were much the same at Armagh, Barrow, and Durham. Barrow, Appleby and Durham show some increase in R0.25 events in the 2000s.

Table IVa. Frequency of R0.25 events in winter by decade.
1850s1       1
1860s6       1
1870s3       4
1880s0       0
1890s1     1 0
1900s1    12 0
1910s3    42 0
1920s00   34 1
1930s11  44620
1950s2 4433111
2000s30076 360
Table IVb. Frequency of R0.25 events in summer by decade.
1850s2       6
1860s5       10
1870s4       3
1880s6       9
1890s9     3 4
1900s6    02 0
1910s4    24 3
1920s35   13 6
1930s44  15234
1950s3 5313211
2000s47012 616

3.2.4. Lamb weather type analysis

The daily rainfall observations were compared with the objective Lamb Weather Catalogue. Four weather types, southwesterly (SW), westerly (W), cyclonic (C) and southerly (S) provide more than 50% of total winter rainfall at each station (Table V). SW and W conditions are typified by low pressure to the north, high pressure to the south and a sequence of depressions moving east across the UK (Lamb, 1972); together, SW and W provide around 40% of total rainfall for all sites except Durham. C conditions refer either to the rapid passage of depressions across the county, or a deep depression stagnating over the British Isles. S conditions are associated with a low in the Atlantic that is blocked west of the UK causing depressions to move northerly or northeasterly. In summer, SW, W, C, and S weather types are again major contributors of rainfall along with anticyclonic (A) conditions, which are associated with thunderstorms (Table V).

Table V. The percentage contribution made by the 5 main precipitating air flows to total winter and total summer rainfall, 1961–1990

Figure 7 shows the percentage contribution made by the main precipitating air masses to winter and summer rainfall totals in 1940 and 2005 at each station. These values were calculated from a linear regression through the 65 years of data. In winter (Figure 7a), there are 2 clear trends; firstly, the percentage contribution associated with the C weather type has decreased at all stations except Ulpha, with lowland or more easterly sites, e.g. Barrow, Appleby, Burnhope, or Durham, showing the largest reduction. Secondly, the percentage contribution made by W weather types has increased at all stations, particularly at easterly and/or lowland sites. It is notable that the upland sites of Ulpha and Coniston show the lowest rises in percentage contribution. The increase in rainfall associated with W weather types in winter is reflected in T10 falls, which have increased in percentage contribution. Coniston (4%), Appleby (7%), and Burnhope (4%) stations show the greatest increases.

Figure 7.

The difference in the percentage contribution made by the main precipitating weather types to: a, total winter rainfall; and b, total summer rainfall from 1940 to 2005. The estimated percentage contribution in 1940 and 2005 for each Weather Type and station is given in the table below. See text for details

In summer (Figure 7b), a strong decrease in the rainfall provided by cyclonic weather types from 1940 to 2005 is observed at all stations. There is also a moderate increase in the percentage contribution made by anticyclonic and S weather types at all stations. The percentage contribution made by the T10 rainfall associated with cyclonic weather types has decreased by 2–4% at all stations.

4. Discussion

The results presented here confirm earlier findings (e.g. Maraun et al., 2008) of a widespread shift towards more frequent incidence of and greater contribution from heavy falls of rain during winter, and towards less frequent and lower contribution from heavy rain during summer. The results have been considered in terms of orography and changing atmospheric circulation, and over a longer time period than that generally used in previous analyses (1961–1990).

There has been a significant increase in winter rainfall totals in the uplands in recent decades and this is reflected in the number of T10 and R0.25 heavy falls. In complete contrast, heavy summer rainfall has become much less frequent (at least until the 2000s). Principal components analysis of heavy winter rainfall in the period 1961–2006 (Figure 7; Maraun et al., 2008) indicated a strong downward trend on the Lake District coast, with weaker positive trends in the Lake District and stronger positive trends in the Pennines and towards the east coast. Results in Table IIa for T10 events accord to this pattern to a reasonable extent: there has been little change since the 1960s at the sites on or near the coast, Barrow and Ulpha. In contrast, the number of winter T10 events at Coniston since 1981, and Holehird since 1991 were the highest in their 80-year and 100-year respective records. Appleby and Burnhope Reservoir also showed highest recorded values in the 1990s. In addition, the percentage contribution made by T10 events to the total annual rainfall was highest at Burnhope in the 1990s. These stations are both located east of the Lake District, and this changing pattern of orographic enhancement, with relatively more heavy winter rainfall to the east of the Lake District, may reflect changes in atmospheric circulation, as discussed below.

At the same time, as heavy winter rainfall was becoming more frequent in the uplands, there was a dramatic decline in heavy falls during summer with record low totals for T10 events at all the upland gauges in the 1990s (Table IIb). The 1990s total at Appleby was also the equal lowest on record (with the 1900s). Not surprisingly, the same pattern was evident for R0.25 events (Table IVb); as a result, the winter:summer ratios for R0.25 events exhibit record values in the 1990s for all the upland gauges, including Ulpha, and at Appleby. Given the rarity of T10 and R0.25 events, the results in Tables II and IV lend support to Lane's (2008) argument that the middle part of the 20th century was flood-poor, compared to recent times; in particular, for both T10 and R0.25 events, the 1970s had low winter totals, in stark contrast to record totals in the 1990s.

LWT analysis suggests that the observed increase in winter rainfall (Figure 3a) is driven by an increase in the rainfall provided by W weather types (Figure 7a). This is unsurprising given the recent positive phase of the North Atlantic Oscillation (Hurrell, 1995) and Ferranti et al. (2010) noted an increase in both the frequency and the average wet-day rainfall amount associated with W weather types. The westerly upland stations (Ulpha and Coniston) show the lowest increases in the percentage contribution provided by W weather types from 1940 to 2005. This may be interpreted as follows: in earlier decades, the rainfall associated with W weather types fell predominately in the western Lake District uplands. In more recent decades, W weather types bring a greater amount of rainfall to all stations along the transect. Therefore, the greatest increase in the percentage contribution made by W weather types occurs at stations that were previously relatively dry on W weather type days. This extends the observations of Malby et al. (2007) who noted a greater proportion of winter rainfall is carried over the Lake District hills (from 1970s to 1990s), and also Ferranti et al. (2009) who reported the greatest increase in winter rainfall amount associated with SW weather types (1961–2005) is occurring on Lake District leeward slopes. The easterly propagation of rainfall associated with W weather types is also apparent in T10 falls, which have increased in percentage contribution most notably at Coniston (4%), Appleby (7%), and Burnhope (4%).

An increase in the contribution made by the W weather type in winter is also linked to the decreased importance of rainfall associated with the C weather type. The percentage rainfall contribution made by the C weather type has decreased in both winter and summer, and Ferranti et al. (2010) reported a decrease in both the frequency and the average wet-day rainfall amount provided by C weather types. This suggests the observed decrease in summer rainfall (Figure 3) is driven by a reduction in the frequency and rainfall amount provided by C weather types (Figure 7). This is also reflected in the decrease in T10 rainfall provided by C weather types.

In relation to the summer floods of 2007 in central England, Lane (2008) argues that these should not have been unexpected when viewed in the longer term. He notes that relatively few rainfall records are available prior to 1960. He argues that the period since 1960 is relatively flood-poor compared to the last 150 years as a whole which is why the late 1990s and early 2000s were perceived as exceptionally flood rich, but in fact are not so exceptional in the context of much longer records. The results presented here lend support to Lane's argument, confirming the lack of heavy winter rainfall in the 1970s and a declining number of heavy summer falls through the second half of the 20th century. Robson et al. (1998) also observed that the 1970s were comparatively lacking in floods compared to the flood-rich 1960s and 1980s. The results presented here underline the value of long-term monitoring and the maintenance of records from key historic sites (Robson et al., 1998; Burt, 1994; Burt et al., 2008). Long records for upland sites are rare, so the agencies responsible for climate monitoring must be careful to maintain a few key records in order to ensure continuity. An example of bad practice in this regard was the cessation of daily rainfall measurement at Appleby Castle in 2000, notwithstanding that the record was continuous, excepting one brief gap, from 1891—indeed the station is, ironically, one of the UKMO's ‘historic sites’ on its website. Here, it has been necessary to assemble composite records for key locations in the Lake District using nearby gauges; whilst acceptable, there must always be some doubt about this method and it is always preferable to have records from a single site.

The need for long records from upland gauges must be emphasised, because this is where widespread floods are most likely to be generated in the UK (since most large catchments have their headwaters in upland areas) and, as this study shows, upland areas seem particularly sensitive to recent changes in climate. Very long records like those at Durham and Armagh are very rare and so invaluable, but as the analysis presented here has shown, they are not good sites for assessing the likely response of heavy rainfall in upland areas.

5. Conclusions

There were large increases in winter rainfall in the 1980s and 1990s at all the upland gauges and at Appleby. The effect was not seen at the lowland gauges Barrow and Durham, although a very modest increase was seen at Armagh. In contrast, total summer rainfall has declined in recent decades.

For the upland gauges, the changes in total rainfall in recent decades are reflected in the frequency of heavy falls of rain, as defined using two threshold values. In winter, as the threshold value increases, so the contrast between the 1990s and earlier decades becomes even more marked.

The 1990s saw record numbers of heavy falls in winter but an almost complete absence of heavy summer rainfall in the uplands, in marked contrast to lowland gauges.

The increase in winter rainfall amount is related to an increase in the rainfall provided W weather types. The decrease in summer rainfall is related to a decrease in the rainfall provided C weather types.

The contrast between the 1990s and immediately preceding decades lends support to the argument that the high frequency of flooding in recent years was unexpected because evidence from earlier decades (especially before 1961) had been overlooked. The results presented underline the value of long-term monitoring and the maintenance of climatic records from key historic sites.


Most of the data presented in this paper were downloaded from the British Atmospheric Data Centre (BADC). We also acquired data from the Met Office's historic station data website:

Rachel Merrix (Environment Agency) provided the data for Burnhope Reservoir. We thank Mark Bailey and colleagues at the Armagh Observatory for providing additional data through to the end of 2008.