Extreme weather events can have severe consequences for the local environment and population, and are accountable for a disproportionately large amount of climate-related risk. In addition to causing extensive physical damage, these events can have a significant negative impact on society and the economy (Hulme, 2003; McMichael et al., 2006). There is evidence that such weather-related events, at the margins of current local climate distributions, will increase in frequency and intensity in the future (Palmer and Räisänen, 2002; Beniston et al., 2007). It is therefore prudent that action is taken to ensure that social and physical systems have the capacity and capability to absorb and respond to these impacts, such that the structure and function of the systems are maintained and major disruption is avoided.
Future projections of climate change for the United Kingdom (e.g. UKCP09) indicate that the North West region, home to the major conurbations of Manchester and Liverpool, is likely to experience an increasing frequency and intensity of extreme climate events, such as floods, heat waves and storms (Murphy et al., 2009). The impact of these rapid fluctuations in meteorological conditions and shifts in the distribution of local climate variables over time will be felt most acutely in those areas where the human and environmental systems are already marginal (e.g. heavily populated urban areas, coastal zones which suffer frequent flooding). For example, during the 2003 heat wave when the highest temperature ever recorded in the United Kingdom was registered (38.5 °C at Brogdale, Kent) and average daily temperatures in Manchester exceeded 20 °C in August on eight near-consecutive days (compared to the 1971–2000 daily average of 16.1 °C), unhealthy living and working spaces were experienced in buildings situated in the Manchester city region and surrounding suburbs (Wright et al., 2005). In contrast, the heavy snowfall and extreme low temperatures (below − 10 °C during the night) experienced during winter 2009/2010 are reported to have cost the local economy £24 m in 24 h (Manchester Evening News, 2010). Consequently, it is important that the region enhances its preparedness for such events and implements appropriate adaptation strategies.
Temporal analogues offer a valuable resource to aid understanding of how past events have affected the local environment and population. Future climate conditions can then be reported with respect to these historical events in terms of the frequency and severity of recurrence, within the scientific limitations of error and uncertainty associated with climate forecasting. For example, the 2003 European heat wave cited above, which is considered an extreme event (exceeds the 90th percentile) under current climate conditions becomes comparable to median climate conditions for the period 2071–2100 (Beniston, 2004; Beniston and Diaz, 2004). For Greater Manchester the 2003 event is equivalent to median summer temperature conditions under the 2050s high emissions scenario, and in 9 out of 10 years the warmest day in summer will exceed 28 °C for much of the region (Cavan, 2010). As an alternative example, heavy precipitation events which have in the past caused detrimental flooding in Greater Manchester (e.g. flooding in Heywood in 2004 and 2006; Douglas et al., 2010), may occur as frequently as one to four times per year by the 2050s under the high emissions scenario (Cavan, 2010). Such analogues are a useful communication tool for policy makers, stakeholders and the general public because they are based on real events and the likelihood that such events will become increasingly common in the future. They can also provide the type of quantitative information which is required for informed risk evaluation and adaptation planning.
This paper describes the application of a three-stage process to determine quantitative climate thresholds, indicative of extreme events, for the case study area of Greater Manchester, situated in North West England. In the first phase of the process a local climate impacts profiling exercise was carried out to identify the principal extreme weather-related events that have occurred in Greater Manchester in the past. The second phase involved an independent quantitative evaluation of extreme events using historical meteorological records from the UK Met Office. In the final stage of the work the findings from phases 1 and 2 were combined and statistical methods were employed to extract key meteorological thresholds which have in the past resulted in impacts which have affected human health/well-being, have caused damage to the urban infrastructure or have severely disrupted services.
2. Greater Manchester climate impacts profile
Greater Manchester offers an interesting case study for investigation as it is home to some 2.5 million people. With population densities reaching 3779 people km−2 in central areas (Office for National Statistics, 2008), it is one of the most densely populated regions in the United Kingdom. Almost two thirds, 62%, of the region is characterized as urban (793 km2), which is comprised of a range of land use types and built forms (Gill et al., 2008). However, the region also encompasses more rural areas of surrounding farmland with contrasting soil types, and has a significant altitudinal gradient (from 540 m above sea level in the northeast to less than 100 m in the southwest).
A database of weather related events, which have negatively impacted on human health, human well-being and on infrastructure and housing or have caused disruption to services in Greater Manchester for the period 1961–2009 was constructed through consultation of a variety of archival sources. This formed part of a broader Local Climates Impact Profile which was carried out for Manchester City Council. The key sources of data used are listed in Table I.
Table I. Archive sources consulted to construct the database of past meteorological events in Greater Manchester, 1961–2009
Source of data
Association of British Insurers
Supplementary information, table VII: major weather incidents in the United Kingdom
Association of Greater Manchester Authorities (AGMA)
Local authorities asked to collate data from their civil contingencies, education, highways, social services, drainage, media and archive departments
Strategic flood risk assessment for Greater Manchester
Catchment management plans
Greater Manchester Fire and Rescue Service
Flooding call-out records
Local media reports (Manchester Evening News, Bolton Evening News, Ashton Reporter, Post and Chronicle, Wigan Observer, Wigan Evening Post)
Available from Manchester Central Library Archives and local studies unit
North West Public Health Observatory
Emergency hospital admissions database
For the 30 year period 1961–1990, 94 weather-related events were identified across Greater Manchester. Between 1971 and 2000, 112 weather-related events were experienced. These are events which manifest as flooding, heat stress, storm damage and tree falls, lightning damage and wild fires, weather related transport damage and delays, ground instability (land slides and subsidence), poor air quality due to fog and smog, drought or extreme cold (hypothermia and ice damage). The rise in the number of events for the two 30 year periods does not necessarily indicate that extreme events are becoming more frequent, but may simply be the result of growing media interest in climate change related impacts (Boykoff and Rajan, 2007).
The proportional split of different types of extreme event shows some variation over the period 1961–2009 (Figure 1). Summertime events associated with high temperatures or reduced precipitation amounts do not feature at all prior to 1971 but now account for ∼10% of extreme events. This is reflective of the significant increasing temperature trend experienced by the North West between 1961 and 2006 (Jenkins et al., 2008). In particular, maximum summer temperatures have risen by 1.63 °C during this period, and this has been coupled with a 13.2% decrease in summer precipitation (Jenkins et al., 2008). While the percentage of reported wind events has fluctuated slightly, the actual number of events between 1961 and 2000 has not changed (five per decade in 1961–1970; six per decade from 1971 onwards). This is in contrast to the suggested UK-wide increase in windstorms made by Jenkins et al. (2008). However, as Jenkins et al. (2008) state, trends in storminess are difficult to identify due to the low number of events annually. In addition, their analysis at the UK aggregate-level may not be representative of storm trends in the Northwest. Furthermore, this contradiction may be due to improved building standards and arboreal services and the reduction in pedestrians on the streets in suburban areas leading to a decline in the severity of the impacts on the local population and/or infrastructure, and consequently a lack of archive evidence of such events.
Flooding is the dominant climate impact within the region across all decades, and is found to be attributable for 37–55% of decadal extreme events. There is some suggestion from the database that flood events are becoming more frequent (Figure 2). This would correspond to changes in precipitation patterns exhibited over the same period, exemplified by both winter precipitation totals and the frequency and contribution of intense rainfall events increasing (Osborn and Hulme, 2002). However, the time series is not sufficiently long enough to draw any robust conclusions. Indeed, the majority of pluvial flood events actually occur during the summer months when overall precipitation totals have declined. This may indicate the significance of brief but very intense episodes of rainfall during the summer and/or changes to land use over time, such as urban infill trends, which increase surface runoff and consequently pressure on drainage systems.
Mapping the location of flood event occurrence through time (Figure 3) suggests that flooding in the north of the region (Irwell catchment) has become more prevalent over the past decade. To the south of the region (Mersey catchment) the number of events has remained comparatively high throughout the period under investigation, but there has been a shift from riverine to pluvial events. In particular, riverine flooding on the Mersey has decreased considerably since the opening of the Sale and Didsbury temporary flood storage basins in the 1970s (Environment Agency, 2009). The decrease in the severity of recent flood events, with respect to the impact on the health and well-being of the local population, infrastructural damage and delivery of services as suggested by the various archive records (Table I), is perhaps a reflection of the success of flood prevention measures throughout Greater Manchester in recent years.
While the process of collating this evidence base has been rigorous and has drawn on a broad range of resources, there are several external factors which may limit the completeness of the extreme events database. Namely, other important sources of information on the impacts of weather-related events in Greater Manchester may have been overlooked, and those that were consulted may be incomplete. For example, the press is inconsistent in recording weather related events. As well as depending on the whims of journalists, the recording of such events will also be influenced by the amount of other newsworthy events at the time. Furthermore, it is not always possible to extrapolate the precise location of the event from media reports. For example, an event is invariably recorded as having taken place ‘in Salford’, ‘in the Irwell Valley’, ‘across north Manchester’. In addition, some of the Local Authority officers consulted experienced difficulty in obtaining records of past weather related events from within their respective departments. The recording of past weather related impacts by Local Authorities is haphazard, a weakness that should ideally be addressed if climate change adaptation is to be a priority for the future.
3. Quantifying extreme events in Greater Manchester
Ringway (Manchester International Airport), offers the only suitably long, uninterrupted record of observed meteorological data in Greater Manchester. However, there is a denser network of rainfall monitoring stations, due to the high spatial variability of precipitation (Figure 4(a)). Daily meteorological data for Ringway, along with precipitation data for 14 stations with continuous or near continuous rainfall records, were extracted from the British Atmospheric Data Centre (BADC) for the period 1961–2009. Extreme events for a range of climate variables were classified using the Ringway dataset according to a set of objective criteria (Table II). These indicators are based upon previously defined indices (Frich et al., 2002; Alexander et al., 2006) but are necessarily more stringent than those that have been applied at a global scale due to the spatially localized nature of this study. The percentiles and other criteria used to classify extreme events were set at a level when impacts which affect human health/well-being, cause damage to the urban infrastructure or severely disrupt services is highly likely to occur.
Table II. Indicators used to identify extreme weather events from historical Met Office records
Gale days (mean wind speed reaches or exceeds 34 km for a period of 10 min)
3.1. Temperature extremes
Extremely high temperature events are quantified as those days where maximum daily temperature (Tmax) was greater than or equal to 29.2 °C. The frequency of these events has risen from 0.2 per annum during 1961–1970 to greater than 1 per annum over the past two decades (Figure 5). Although the majority of the high temperature days experienced during the latter decades occurred during the summers of 1995 and 2003, the data exhibit an upward trend in the frequency of hot days across all decades, which is consistent with findings from previous studies (Jenkins et al., 2008). Over a quarter of the interdecadal variation in high temperature extremes is attributable to the increasing tendency toward extremes over time. In general, the days when extreme maximum temperatures were recorded coincided with days when Tmin was also considered unusually high. However, on over a quarter of occasions when Tmax was high but not classified as extreme, Tmin was classed as extreme. In relation to human health and well-being it is often elevated night time temperatures which are considered more serious than the daytime maxima (Kalkstein, 1991; Hajat et al., 2002), and this must therefore be taken into account when defining critical meteorological thresholds for the local population.
The time series of extreme low temperature events mirrors that of the high temperature events, displaying an overall decreasing trend across the five decades (Figure 5). As with high temperature extremes, over 25% of the changing interdecadal frequency of low temperature extremes is accounted for by the time series trend. However, the dataset is skewed somewhat by the unusually cold decade from 1961 to 1970. Indeed, quantifying an extremely low temperature event for the climatological periods 1961–1990 and 1971–2000 gives a 0.8 °C discrepancy (daily Tmax < − 2.4 °C and daily Tmax < − 1.6 °C, respectively).
3.2. Precipitation extremes
Extreme precipitation events are classed as days with total precipitation exceeding or equal to 24.6 mm, based on the Ringway dataset. However, due to the significant NE-SW precipitation gradient which exists across Greater Manchester (Figure 4(b)), there is a 14 mm difference in the defining threshold for extreme precipitation at meteorological stations situated across the region. Ringway has the lowest threshold and Broadhead Noddle (Figure 4(a)) the highest (38.6 mm). Moreover, extreme events at these two stations coincide on less than a third of occasions (31% of the time), emphasizing the spatially variant nature of intense precipitation episodes. Consequently, some form of distance function is required in order to establish the most appropriate precipitation risk metric for a particular location, based on the nearest and most geographically comparable meteorological station.
There is an absence of a temporal trend in the frequency of extreme precipitation events across the precipitation monitoring network. All stations display an apparent oscillatory pattern, switching between decades with a high number of extreme rainfall events (>2.2 per annum during 1961–1970, 1981–1990, 2001–2009) and those with a comparatively low number (<1.5 per annum during 1971–1980, 1991–2000). Osborn and Hulme (2002) found a seasonal disparity in trends of heavy rainfall events across the United Kingdom, with heavy winter precipitation events increasing in frequency whilst the number of intense summer rainfall events decreased. To check whether this seasonal difference was being averaged over the annual scale data, the data were also analyzed for summer and winter separately. However, there was no evidence of the existence of a trend in either the winter or the summer datasets. The lack of a trend in the meteorological data, at odds with the archive evidence (Figure 2), is likely to be the result of the daily resolution of the dataset. The daily total precipitation is too coarse to capture the short duration (maximum 3–4 h), high intensity rainfall episodes that lead to excessive runoff, and cause pluvial flooding.
Heavy snowfall events (daily snowfall ≥ 6 cm) have dropped off markedly since 1970 and almost three quarters (R2 = 0.74) of the interdecadal variation in the frequency of extreme snow days is explained by the temporal trend. Given the lack of trend in the precipitation dataset it is likely that this decreasing trend is due to the significant winter warming which has occurred over the past 50 years. It is also worth noting that the winter North Atlantic Oscillation index has been in a positive phase since 1970 (Jenkins et al., 2008). This is associated with warmer and wetter conditions in northern Europe, which may provide some explanation for the decline in heavy snowfall events since this time. There is also evidence that snowfall events are becoming less extreme. Prior to the heavy snowfall that occurred during winter 2009/2010 there were no occurrences of daily snowfall exceeding 10 cm since 1984, in spite of events of this magnitude being not uncommon during the 1960s, 1970s and early 1980s. Although the frequency and intensity of extreme snowfall events are decreasing, the archive evidence suggests that the disruptive consequences in the urban environment may be increasing due to the changing provision of mitigation tools, such as snow ploughs and localized supplies of salt and grit. It is therefore necessary to consider the non stationarity of the social, political and physical environment when considering what constitutes an extreme event in the past and how this can be translated into the future.
Drought is a complex process to objectively quantify as it is a function of several physical, biological and socio-economic factors, and unlike other extreme events such as heat waves or floods, the effects are the result of a longer term deviation from average climatological conditions. For example, the hosepipe ban that came into effect in July 2010 in northwest England was the result of an extensive period of drier than average conditions, with much of the region receiving less than 70% of the 1971–2000 average spring rainfall amount (Met Office, 2010). Here, low precipitation as an indicator for drought conditions is based on the consecutive number of dry days (daily precipitation < 1 mm) rather than a percentile, following from previous work on extreme climate indicators (Frich et al., 2002). This is necessary to capture the aggregation of low rainfall days over time which may lead to drought, and because precipitation exhibits a strong negatively skewed distribution deeming a percentile approach inappropriate. The two periods with the highest number of consecutive dry days occurred in September/October 1986 (CDD = 30) and in March 1993 (CDD = 29). There is no evidence that drought incidence has increased over time, however, the number of summer droughts has increased consistently decade on decade. This supports previous findings that summer precipitation totals are decreasing (Osborn and Hulme, 2002; Jenkins et al., 2008). This is potentially a more serious threat for the local population and environment as it coincides with the time of year when soil moisture conditions may already be stressed and it is more likely that the drought conditions will be coupled with high temperatures.
3.3. Storminess and wind extremes
The UK Met Office identifies extreme wind events as ‘Gale Days’ in their climate records. These are defined as a day when mean wind speed reaches or exceeds 34 km (force 8 on the Beaufort scale) over a period of at least several minutes (10 min in the case of a station equipped with an anemograph). As with extreme snowfall events, wind events have reduced rapidly since 1970 and exhibit an overall decreasing trend with time (Figure 5). However, the magnitude of events shows no overall trend, with incidents of windspeeds exceeding 60 knots occurring in all five decades between 1961 and 2009.
4. Meteorological thresholds for adaptive planning
In order to establish meteorological thresholds for Greater Manchester, which are indicative of extreme events, the two sources of temporal analogue data discussed above, are combined. A comparison is made between the dates of extreme events as classified by the objective analysis of the Met Office data and the dates of weather related events in Greater Manchester which have resulted in impacts on human health/well-being, infrastructure and/or service provision, extracted from the historical archive search. This allowed for the construction of contingency tables (as an example Table III). These can then be used to calculate verification statistics which offer a useful gauge to establish the robustness of the quantitative thresholds as a warning mechanism for extreme weather event occurrence (Thornes and Stephenson, 2001).
Table III. Contingency table to compare the results of the archive search with the meteorological threshold for extreme high temperature events derived from Met Office data
Classified as extreme by Met Office data
Not classified as extreme by Met Office data
Extreme event identified from the archive search
a = 19
b = 2
No extreme event identified from the archive search
c = 8
d = 4479
n = 4508
The type of information that can be extracted by analyzing the data in this way includes the percentage of times the two methods independently recognize an event as being extreme:
However, due to the intrinsic rarity of extreme events, this measure is likely to be heavily skewed by the large amounts of days when an extreme event doesn't occur. Perhaps more important parameters, particularly from a planning for adaptation perspective, are the risks of Type 1 and Type 2 errors. Type 1 errors (denoted as b in Table III) are defined as days when an extreme event has been recognized to occur from the archive evidence, resulting in a negative impact on the local population, infrastructure and/or service delivery, but no corresponding extreme event has been identified from the historical Met Office data. Consequently, if the meteorological thresholds were being used in conjunction with weather forecasts as an alert for an extreme event occurrence, the days with a Type 1 error would be those days when it is deemed safe to take no action, assuming an accurate meteorological forecast, when actually the weather conditions do cause harm to the local population or environment. These are days when some form of preparedness would be beneficial but the conditions appear to decision makers to be of low risk.
Type 2 errors, in contrast, occur when the meteorological conditions suggest that there is likely to be an extreme event which will impact on the local area, when actually the day passes without significant consequence for local inhabitants. Type 2 errors are considered less serious than Type 1 errors. In this instance a Type 2 error would imply that unnecessary action is taken by, for example, the local authorities. There may be a cost associated with this, however this is likely to be less costly than damage to infrastructure, disruption to services and in the worst case, loss of lives, which may arise due to the result of inaction during a Type 1 error day.
Other measures of quantifying the reliability, accuracy and skill of climate metrics are discussed by Thornes and Stephenson (2001). These include calculating bias (B) as an indicator for reliability:
using the Miss (M) and False Alarm Rates (F) to establish accuracy:
and using skill scores such as, the Odds Ratio Skill Score (ORSS):
4.1. Temperature extremes
The threshold of 29.2 °C, extracted from the analysis of the Ringway Met Office data, as an indicator of an extreme high temperature event produces favourable results. As a metric of negative impact upon the local population, environment or service provision the threshold was correct on 99.78% of days. However, there were two occasions when a Type 1 error occurred and eight when a Type 2 error occurred (Table III). The number of Type 2 errors can be reduced, and even eliminated, by making the threshold more stringent. For example, every day with a maximum temperature greater than 30.9 °C coincided with an archive report associated with the negative consequences of higher temperatures. However, increasing the threshold will not help to reduce Type 1 errors, which are arguably the more severe outcome.
As a tool for communication and adaptive planning it may be appropriate to provide the threshold warning information along with some indication of the likelihood of climate risk. For example, a traffic light warning system could be used to denote the level of preparedness required, given a temperature forecast of a particular magnitude. To illustrate, for the extreme high temperature scenario a red light would be used for days where maximum temperature is predicted to exceed 30.9 °C, signifying that it is very likely that the local population and/or environment will suffer negative impacts and therefore decision makers ought to take action. For maximum daily temperatures between 29.2 and 30.9 °C an amber warning would be applied, indicating that there is a likelihood of negative impact and that stakeholders should be prepared to take action. For temperatures below this threshold a green light would indicate that the thermal environment falls within the safe boundaries and, therefore, no action is required. This type of information allows for more informed decision-making and could be integrated into a more detailed cost-benefit analysis (Thornes and Stephenson, 2001).
The few ‘heat’ events extracted from the archive sources which were not picked up by the Met Office data were often indirect consequences of the high temperature/heat wave conditions, for example poor air quality, lightning and water shortages. These are all the result of a complex interplay between a number of meteorological variables and, therefore, may not lend themselves to identification by assessing a single meteorological variable in isolation.
Low temperature extremes are not adequately captured by using a low temperature threshold alone. The impacts of low temperatures that prove hazardous are associated with ice and snow, and therefore a suitable climate warning system would require additional information about precipitation and atmospheric stability in combination with temperature data.
4.2. Precipitation extremes
The precipitation threshold of daily precipitation ≥ 24.6 mm derived from the Ringway data is an unreliable indicator of an extreme event. The number of Type 1 and Type 2 errors are 28 and 47, respectively, and the Miss Rate is greater than 0.5. The number of Type 2 errors can be reduced considerably by raising the threshold to 38 mm (0.999 percentile). Now, all but three of the days exceeding this threshold correspond to reports of local flooding, with two out of these three occurring in 2003, which may be the result of recent improvements to infrastructure and drainage to withstand extreme events. The large amount of Type 1 errors will still exist, however, as days with precipitation totals much lower than this are often reported as flood events. This is a consequence of the high spatial variability of precipitation, which means that events in the north of the region (Bolton, Wigan, Rochdale) could not be identified using the Ringway data. Using the Broadhead Noddle data, with the higher threshold calculated for this station, in conjunction with the Ringway data, reduces the number of Type 1 errors by 50%. Specifically, by incorporating this additional constraint the thresholds are now capable of identifying riverine flooding events, which have been extracted from the archive searches, associated with the River Irwell and the River Douglas (Figure 3). Pluvial flooding events, however, remain difficult to identify due to the temporal resolution (daily total) of the precipitation data being insufficient (Section 3.2). Other data sources, such as the hourly rain gauge network maintained by the Environment Agency, may in future provide a more useful means of forecasting flood risk. However, because this dataset was not available for the full extent of the period under investigation here (1961–2009) and because the data may not be freely available to decision-makers, this dataset was not used in the current analysis.
Days with snowfall greater than or equal to 6 cm were frequently reported as having caused some disruption to the region, but the Miss Rate using this criterion still exceeded 0.5. The majority of these unreported events occurred between 1961 and 1970, and this may have been because events of this magnitude were not considered particularly unusual at this time. By discounting the ‘Misses’ during this period, the Miss Rate improves to 0.19, with the remaining unreported incidents identified as isolated events, where heavy snowfall (6–8 cm) occurred on an odd day during an otherwise climatologically average period. There is a slight bias using this threshold (0.87), which indicates underforecasting of extreme events and therefore some risk of a Type 1 error occurring. This bias can be reduced by using the consecutive number of snow days as an additional criterion for extreme event identification.
Very few drought events were extracted from the archive searches. All of these were matched by events in the Met Office database using the 15 consecutive dry days criterion (Table II). On the other hand, there is a considerable number of occasions when this criterion is exceeded yet no drought is identified in the archives. This can be attributed to drought being a more complex event to capture from a single meteorological variable, compared to flooding and heat wave events. Similar to the temperature traffic light warning system described above, it may be that planning for drought should simply be on a ‘be prepared’ basis. Alternatively, further analysis may reveal drought to be definable through a specific combination of two or more variables.
4.3. Storminess and wind extremes
The more stringent threshold (than that indicated in Table II) of days with a maximum gust speed of greater than or equal to 60 knots, as identified from the Ringway data, were all found to correspond to reports in the archive sources of negative impacts to the local population and/or environment. A number of days with lower maximum gust speeds were also reported (Type 1 errors). These were all recorded as gale days and still experienced high wind speeds but in these instances it may be that the wind speed was high over a longer period of time which caused more damage than can be inferred from the maximum wind speed alone. In addition, some of the storm and gale events which were unidentified by the Ringway data analysis were located in the north of the region and therefore may have been subject to differing conditions than those recorded at Ringway due to localized and topographical (particularly altitude) differences.
In light of these Type 1 errors, it may seem sensible to lower the threshold which corresponds to extreme wind events. However, this would then cause the number of Type 2 errors to rise. This trade off between Type 1 and Type 2 errors is a weakness of using the objective quantitative metrics but may in part be overcome by using the sliding scale of risk indicated by the traffic light system, a more detailed cost-benefit analysis of past events and/or the integration of other meteorological parameters. This latter suggestion is borne out by a high proportion of Type 1 events which occurred on days subject to high wind speed conditions (>34 knots) together with heavy rain or snowfall.
This paper has examined the variation of extreme event occurrence over the past five decades. Extreme events have been identified using two methodologies. The first approach made use of historical archive information in order to build up a picture of how extreme events have impacted upon the population, environment, infrastructure and services of Greater Manchester. The second approach involved the quantitative analysis of past meteorological data from the UK Met Office. The outcome of the latter approach was, to some degree, particularly for some variables, influenced by the altitude and location of the Met Office station used (Ringway).
Temperature extremes, with respect to the 1961–1990 baseline period, are found to have increased for high temperatures but decreased for low temperature events in both datasets. In line with these trends, extreme snowfall events are also found to have declined over the past five decades. These findings are consistent with other studies, and are projected to reflect longer term patterns of meteorological change for Greater Manchester in the future (Jenkins et al., 2008; Murphy et al., 2009). There was some disagreement between the two data sources in relation to temporal trends in precipitation extremes. While both approaches suggested summer drought occurrence to be increasing in frequency, only the archive database revealed an increasing trend in flood events from intense precipitation episodes, no trend was evident in the meteorological data. This may, however, be the result of the daily precipitation total data being too coarse to effectively resolve the short duration, high intensity rainfall episodes responsible for pluvial flooding. Storms, and in particular wind extremes, are difficult to capture. There is some evidence of a decreasing trend in the frequency of high wind speed events: however, the magnitude of the maximum gust speed associated with extremes has remained stable.
An attempt was made to define quantifiable thresholds indicative of extreme events by combining the qualitative archive evidence with the quantitative records of meteorological variables. Such thresholds offer a useful means of evaluating the risk to the local population and/or environment for local stakeholders and decision makers when presented with weather forecast information. They are designed as a guidance measure to be used in conjunction with other relevant information in order to support informed decision making as oppose to being used to illicit an ‘Action vs No Action’ response. This latter response, as demonstrated in Section 4, could potentially prove hazardous.
The thresholds had varying degrees of success. For extremes which are the function of a single meteorological variable (e.g. heat waves, pluvial flooding, heavy snowfall) the thresholds proved to be reliable and skillful. However, caution should be used when applying them to highly spatially variant parameters, such as precipitation. For these variables it may be necessary to ensure that the thresholds are based on a meteorological station a minimum distance away or which is geographically similar, or derived from spatially interpolated climate datasets (e.g. Perry and Hollis, 2005). Extreme events which are the result of a more complex interaction between variables (e.g. drought, freezing conditions) were less well captured by applying the thresholds associated with a single variable in isolation. This shortcoming of the threshold approach could be overcome through further analysis and the integration of multiple meteorological parameters into a single extreme metric. Another weakness of the current approach is the trade off between Type 1 and Type 2 errors which is required when assigning the quantitative thresholds. The use of a sliding scale of climate risk (traffic light approach) when communicating the thresholds, and a more comprehensive cost-benefit analysis approach may assist with this issue.
In general, the threshold approach for the identification of extreme events outlined here offers a valuable tool for the region's decision-makers, stakeholders and population. The quantitative and objective means of identifying an extreme event allows for informed decision making and preparedness, and will inevitably assist in reducing costs and minimizing risk to the local population and environment. When used in conjunction with projections of future climate change (e.g. UKCP09), the thresholds will help to determine the probabilities of extremes occurring in the future and identify the elements or areas most at risk. This information can subsequently be used to prioritize and target climate adaptation strategies, with the important caveat that extraneous factors are nonstationary (Wilby et al., 2009). For example, improvements to adaptive capacity (e.g. changes to surface cover properties, policy interventions) may mean that society and the environment become increasingly resilient to extreme climate events, and therefore the thresholds may have to be re-evaluated and adjusted accordingly.
Claire Smith gratefully acknowledges the support of Manchester Geographical Society for their contribution to this work. Nigel Lawson was funded by the EcoCities project. Thanks also to all of the people who assisted with the collation of data on past extreme events: Dave Hodcroft (Bury MBC), Jonathan Mayo (Bolton MBC), Georgina Brownridge (Oldham MBC), Barry Simons/Jonathan Kershaw (Rochdale MBC), Andy Williams (Stockport MBC), Christina Sexton (Tameside MBC), James Noakes (Wigan MBC), Paul Needham/John Thompson (Environment Agency) and Ken Brown (GM Fire and Rescue Service). The Met Office data were accessed via the British Atmospheric Data Centre.