HYDROCLIMATIC VARIATIONS IN SCOTLAND AND NORTHERN IRELAND
SR97(08)D

May 2000

EXECUTIVE SUMMARY

Section 1: Introduction

  1. Since the late 1980s many water resource planners and engineers have become aware of possible changes in the climate of Scotland and Northern Ireland and associated changes in river flow regimes. If substantiated, such changes are potentially of concern to the statutory pollution prevention authorities.
  2. The objective of this report is "to review, within the context of recently recorded climatic variations, the current use and practical significance of those key hydrological variables which relate to both the statutory and advisory duties of the statutory pollution prevention authorities in Scotland and Northern Ireland"
  3. Four tasks are identified: (i) to identify any long-term trends, both national and regional, within the hydroclimatic record; (ii) to undertake a regional (ie basin-specific) examination of low flow records in conjunction with rainfall and evaporation conditions; (iii) to undertake a regional (ie basin-specific) examination of high flows and, in particular, their temporal pattern in the winter and spring seasons; (iv) to compare the overall pattern of recent hydroclimatic variations in Scotland and Northern Ireland with scenarios of future climatic change produced by global circulation models (GCMs).

Section 2: Data bases

  1. A number of pre-existing databases have been identified for the project: Manley's Central England temperature series; Lamb's catalogue of synoptic weather types for the British Isles; Smith's composite precipitation record for Scotland (1757-1992); the all Ireland precipitation record (1862-1995); and the Northern Ireland precipitation record (1931-1995). The following precipitation databases were also compiled for the report: (i) daily and/or monthly precipitation data for 13 sites in Scotland (Braemar, Dumfries, Edinburgh, Fort William, Islay, Inverness, Lilburn (in NE England), Montrose, Plockton, Stranraer, Stornoway, Waulk Glen and Wick; (ii) daily and/or monthly precipitation data for 7 sites in Northern Ireland (Aldergrove, Armagh, Baron's Court, Edenfel, Florence Court, Malin Head (in the Republic of Eire) and Quoile. Precipitation series have been compiled over the long term (1861-1996) and short term (1961-1996) across Scotland. Similar series have been compiled over the long term (1892-1994(5)) and short term (1961 - 1994(5)) for Northern Ireland.
  2. Four river flow databases were compiled and/or updated for the project. Following consultation with SEPA staff in Scotland and DANI staff in Northern Ireland, daily mean flow records were compiled for 71 sites in Scotland and 11 sites in Northern Ireland. The flow records were selected to maximise the natural climatic signal and to minimise the impact of artificial abstractions (surface or groundwater), afforestation or hydro-electric development. Following consultation with SEPA and DoENI staff an existing peaks over threshold (POT) database for high flows at 68 sites in Scotland was updated to 1996 and a new peaks over threshold database compiled for 20 sites in Northern Ireland to 1996. Only stations with continuous records from at least 1975 which also displayed flow regimes not significantly affected by hydro-electric schemes were eligible for inclusion. Where appropriate new rating relationships were established for the extended POT records.

Sections 3 and 4: Precipitation - long-term and short-term trends

  1. Temperatures (based on the Central England series) have been steadily increasing over the period 1861-1996. In terms of Lamb's weather types, Westerlies declined steadily from a peak in the 1920s (coincident with an Anticyclonic minimum) to a trough in the late 1970s from which there was some recovery in the 1980s. The Cyclonic weather type occurs with less frequency than the Westerly and Cyclonic types, although at its maximum around 1980 it coincided with a Westerly minimum.
  2. Correlation coefficients have been calculated for the long-term annual precipitation series (1862-1996) across Scotland. Sites in the north and west register significant increases, whilst sites in the east and south register significant decreases over the same period. A similar west-east contrast is found in the seasonal trends with the west reporting significant spring and autumn increases, the north significant winter increases, and the east significant summer decreases. For Scotland as a whole the period 18621992 just registers a significant increase at the 5% level.
  3. For the short-term period (1961-1996) annual precipitation increases at ten sites across Scotland, significantly in the west and north. Seasonally significant winter increases are also recorded. In the summer most sites register a decrease, especially in the east.
  4. Graphical trends for Fort William and Braemar are compared for the long-term period using the IRWsmooth algorithm. Fort William reports an increase of c. 600mm in annual precipitation since the early 1970s, whereas Braemar registers a slight rise followed by a recent decline.
  5. Over the long term period, Braemar and Stornoway report decreases for all four precipitation maxima (1, 2, 7 and 30 day precipitation) whereas Edinburgh and Dumfries reported increases. Over the short-term period all stations, except Dumfries, record increases in precipitation maxima. During the long-term period there is a general decrease in the frequency of 1, 2 and 7 day dry spells. Over the short-term period all stations (except Edinburgh) report strong decreases in the frequency of dry spells.
  6. Annual values of the coefficient of variation over the long term are lower than seasonal ones, with autumn the season with the lowest inter-annular variation. Seasonal coefficients of variation increase over the short term, especially in the winter. Standardised coefficients of variance portray long term changes in annual precipitation for individual stations across Scotland.
  7. Significant correlations exist between weather types and annual precipitation: consistently negative values for the Anticyclone type, positive values for the Westerly type (only in the west) and positive values for the Cyclonic type (only in the east). These results reflect the dominance of contrasting precipitation generators and storm tracks across Scotland.
  8. Over the long term period only Malin Head and Quoile report significant increases in annual precipitation for Northern Ireland, whereas six stations report a significant increase for at least one season. Within the short term period three stations report significant increases in annual precipitation, but only two in seasonal precipitation. Malin Head (at the northwest tip of Ireland) consistently registers the strongest changes over time. Seasonal coefficients of variation are consistently higher than comparable annual values, although in contrast to Scotland spring rather than winter registers the greatest seasonal variation. Differences in values over the long and short term are minimal.
  9. Westerly and Cyclonic weather types generate most of the precipitation across Northern Ireland with the former being of greater significance in the west and the latter dominating the centre and south east. The regional differentiation of precipitation patterns across Northern Ireland is much weaker than for Scotland.
  10. In terms of dominant weather types, the majority of Scotland's precipitation is of Cyclonic origin, although for Stornoway and Dumfries the Westerly weather type dominates, although less so since 1970.
  11. In terms of precipitation maxima, Braemar is the only station to show a significant decrease in 1 and 2 day maxima (1872-1996). Within the period 1961-1996 Stornaway has registered significant increases in 1, 2 and 30 day maxima. The seasonality of Scotland's extreme precipitation peaks in October-November in Stornoway and in August for Edinburgh (reflecting Westerly and Cyclonic dominance respectively). Significant increases in 1 day falls over 25mm, 2 day falls over 50mm and 7 day falls over 100mm are reported for Dumfries (1871-1996) and Stornoway (1961-1996). Edinburgh also registers significant increases in 2 day falls over 50mm and 7 falls over 100mm (1896-1996).
  12. In terms of precipitation minima, Braemar has recorded significant declines in the number of 1 day, 2 day and 7 day periods with no precipitation (1872-1996), a similar pattern also being recorded for 1 day and 2 day periods 1961-1996. Low precipitation events have also been significantly decreasing in Dumfries (1871-1996) and Edinburgh (1961-1996), but increasing in Stornoway (1874-1996).
  13. In terms of precipitation maxima only Malin Head consistently registers a significant increase in 7 day and 30 day maxima for 1941-1995 and 1961-1994. Seasonal precipitation maxima peak in the autumn as follows: Armagh (September), Malin Head (October) and Florence Court (November). Significant increases in 2 day falls over 30mm (Malin Head) and 7 days falls over 45 mm (Edenfel) are reported for 1941-1995 and 1872- 1994 respectively.
  14. Significant decreases in precipitation minima are recorded for most stations in Northern Ireland for 1 day, 2 day and 7 day periods with no precipitation over most time periods. But there is little evidence of changes in the frequency of low precipitation events.

Section 5: Annual and seasonal mean flows, low flows

  1. In terms of annual mean flows (1970-1996) all stations except three in Scotland report an increase: significant at the 5% level at 12 sites. The majority of stations exhibiting significant increases are in the south and west with none in the northeast. Each record which registered a significant increase was scaled and a trend generated using the IRWsmooth algorithm. A strikingly consistent temporal signature emerges in which the mean flows rise from 70-80% of the mean in 1970 to a peak of 105-110% in the late 1980s falling back to 100-110% by 1996. The 26 stations which did not register a significant increase in annual mean flows report a different signature with a rapid rise from 70-90% of the mean in 1970 to 105-120% in the early 1980s falling back to 85-105% by 1996. In terms of seasonal flows, 17 of the 38 stations show significant increases in the spring and autumn and 15 show significant increases in mean winter flows. It is noteworthy that all the stations draining to the Solway Firth and within the southern part of the Clyde comprise 12 of the stations reporting increases in the autumn and spring and 11 reporting increases in the winter. Twenty stations (more than half) report decreases in mean summer flows, but none of these is significant.
  2. In terms of low flow extremes in Scotland, twenty-one of the 38 gauging stations report a decrease in Q95 over the period 1970-1996, but only one of these (the White Cart Water) is significant at the 5% level. When the low flows for each station are scaled and a trend derived using IRWsmooth, most records show values below the mean until 1980, rising above the mean in the years 1985-90 but thereafter registering a strong decline to 1996 when virtually all stations are at 60- 100% of the mean.
  3. Trends in the magnitude of 1, 2, 7 and 30 day lows are also analysed. Significant decreases in low flows for all four durations are reported on the R Dee (Woodend) 1950-1996. Similar significant decreases in low flows are also recorded on the R. Dee (Woodend), the Dean Water (Cookston), the R Almond (Almondbank) and the R. North
  4. Esk (Dalmore Weir) 1960-1996. By 1970-1996 only the R Clyde (Blairston) and the White Cart Water (Hawkhead) continue to report significant 7 day low flows.
  5. Trends in mean annual flows across Northern Ireland from the late 1970s/mid 1980s to 1995/6 are more muted than in Scotland with only the Bush (Seneirl) registering significant increases in spring, summer and winter flows (1980-1996) and the Crosh (Owenkillew) significant increases in autumn and winter (1981-1996). Otherwise spring, autumn and winter generally report increased flows and the summer decreases. In terms of low flow extremes only the Faughan (Drumahoe) registers a significant decrease in Q95 (1977-1995) although nine of the remaining stations do report deceases in Q95. The only significant increase in the magnitude of 1, 2 and 7 day minimum flows
  6. is also at the Faughan (Drumahoe).

Section 6: High flows (peak over threshold data) - Scotland

  1. In terms of high flow extremes in Scotland, changes in both the magnitude and frequency of POT events have been analysed for the period 1970-96. POT magnitude is expressed as the mean flow once the threshold has been exceeded. Only 4 stations register significant changes in magnitude: two increasing and two decreasing. By contrast 15 of the 44 stations report significant increases in frequency (typically from about two to four events per year), the majority being in the central and southern part of Scotland.
  2. The seasonality of flooding across Scotland is reported in polar plots with the direction of the arrow depicting the average day of POT events, the length of the arrow the strength of the seasonal signal and the thickness of the arrow the average size of floods. During the period 1970-1996 the mean day of flood is in December with stations in the south east showing a particularly strong seasonality. This pattern changes as one moves to the north and east (January now includes the mean day of flooding) and changes the period of analysis. November and December mean days of flooding in 1967-1976 have typically shifted later by 1-2 months by 1987-1996.
  3. Different weather types dominate flood generation across Scotland (1970-1996). In the east the Cyclonic weather type is dominant, whereas in the south west of Scotland the South Westerly type is now extremely important. In central and western areas the Westerly type generates most floods. These patterns change for some sites when only
  4. the 20 largest floods are analysed. Thus the largest floods on the Findhorn (Shenacie) are of Cyclonic origin although most of the floods at this station are generated by Westerly weather types.
  5. Flood growth curves for Scottish stations (using a power law model) are reported for all floods, for floods divided according to their season and according to the generating weather type. For most stations winter and autumn floods for a given return period are larger than summer and spring floods. Ratios of seasonal to annual floods are reported for all stations enabling the season with typically larger floods to be identified. Flood growth curves for each station partitioned according to the dominant flood generator (Westerly, Cyclonic and South Westerly) as appropriate are reported. In most cases the growth curve for the dominan1: flood generator is steeper than that for other weather types and the total data set.
  6. Where possible, the impact of length of record on growth curves is also reported. In the case of the R Dee (Woodend), the growth curve for annual floods with return periods >10 years for 1934-1996 is steeper than that for 1970-1996. Similar seasonal contrasts occur depending on which record length is used. Where return periods are needed for a specific station both seasonal and weather type partitioned curves should be consulted along with that base on the total data set. Ideally, growth curves should be based upon the longest record available, although where regional comparisons are to be made, standardisation of records (eg 1970- 1996) is necessary.

Section 7: High flows (peak over threshold data) - N. Ireland

  1. Trends in the magnitude and frequency of floods in Northern Ireland (based on POT series) are reported. Two sites (Blackwater at Maydown Bridge and the Braid at Ballee) report a significant increase in magnitude, and two sites (Six Mile Water at Antrim) and the Claudy at Glenone Bridge) report significant decreases in magnitude for 1975-1996. There is a general increase in flood frequency with three sites (Blackwater at Maydown Bridge, Braid at Ballee and the Six Mile Water at Antrim) recording significant increases. All but four of the stations show increased flood frequency.
  2. The pattern of seasonality in Northern Ireland mirrors that for Scotland with the mean day of flood typically being in December, although the strength of seasonality is not always well-developed. Between 1975- 1985 and 1986- 1996 a shift towards flooding occurring later in the winter has been detected. There is also a regional pattern with the mean day of flood in the south being a month or so later than in the north.
  3. The Cyclonic weather type provide the dominant flood generator, with the Westerly type next in frequency and of much less significance, although increasing towards the west of the province. Flood growth curves are partitioned according to season and dominant weather type. The ratio of the magnitude of seasonal floods to annual floods increases for higher return periods and more dominant seasons. In this case the dominant season is autumn rather than winter (as in Scotland). The Cyclonic type is the dominant flood generator and typically generates larger floods (for a given return period) than the Westerly or, where available, South Westerly types.

Section 8: Climate change scenarios

  1. There is mounting evidence that the global climate is changing in response to human activity. Although global warming is the most obvious manifestation, regional increases/decreases in precipitation also provide evidence of climate change. If sustained over the next few decades, it is clear that climate change will have a significant impact on the regulation and management of water resources in many parts of the world.
  2. Two main strategies have evolved in assessing the hydrological effects of climate change. The first is to identify past hydroclimatic variability over appropriate timescales by examining changes in precipitation and runoff. The second involves the development of climatic scenarios coupled with Global Circulation Models and appropriate downscaling of outputs of temperature and precipitation at the regional or catchment scale and over a range of time periods.
  3. In general, current climate change predictions for Scotland and Northern Ireland by the 2080s are as follows: 1.2 C to 2.6 C warmer; annual precipitation increased by 5-20%; increases in both intense daily precipitation and severe winter gales plus higher evapotranspiration losses. In more detail the mean increases in annual precipitation for the 2020s to the 2080s for Scotland and the Scottish Borders based on the Medium-high scenario for emissions and the HadCM2 model are 6-16% and 6-14% respectively. The seasonal distribution of these increases is concentrated in the autumn (SON) and winter (DJF) with increases as high as 24% and 29% for the autumn months by 2080. By contrast predicted summer (JJA) changes in precipitation vary from -5% to +5%. Evapotranspiration losses are likely to increase by 2-3% (Scotland) and 3-7% (Scottish Borders) with the highest seasonal increases (12-20%) over the Scottish Borders by the 2080s. It is important to note that whilst predicted temperatures changes can be assigned to human activity, the situation is much less clear cut for precipitation. Many of the predicted changes in precipitation are within limits of natural variability.
  4. Changes in annual runoff across Scotland and Northern Ireland by the 2050s (based on an earlier climate change experiment: IS92a) are likely to be of the order of +5%-25%. Under this scenario annual runoff for four representative Scottish catchments (R Don, R Almond, Lyne Water and R Nith) is predicted to increase by 8.9-11.6%. Applying a water balance model across the whole of the UK and reporting the results for a 0.5 x 0.5 grid, the predicted change in annual runoff across the whole of Scotland for the 2050s is 5-15%, but locally reaches 25%. For Northern Ireland the comparable value is +5-15%. Turning to the simulated flow duration curves for the 2050s, only the R Don, R Almond and R Nith register an increase in Q95 of 5% or less. By contrast Q5 values (the flow exceeded 5% of the time) report increases of 10% (R Don), 16% (R Almond), 24% (Lyne Water) and 11% (R Nith), again typically much higher than for catchments in England. Thus far there have been no attempts to predict changes in runoff based on the HadCM2 model outputs.
  5. In terms of regulation and water resource planning these predicted changes in runoff should be regarded as first order approximations. They take no account of natural climatic variability and could generate different absolute values if other scenarios and climatic models were used.

Copies of this report are available from the Foundation: Final Report 50.00 - Appendices (5 Volumes) 35.00 per volume, less 20% to FWR members.