Temperature and surface lapse rate change: a study of the UK's longest upland instrumental record

Authors


Abstract

A homogenized 76-year daily-mean temperature series is presented for Moor House, an upland peatland meteorological station in northern England, which provides the longest instrumental temperature record in the UK uplands. This paper assesses trends in the temperature record over time compared to the nearby lowland temperature record at Durham. Overall warming for 1991–2006 compared with 1931–1960 and 1961–1990 was approximately the same at the upland site (0.53 °C and 0.73 °C for the two time periods, respectively) as at Durham (0.61 °C and 0.71 °C), maintaining a similar lapse rate. However, winter warming is dominant at Moor House and more pronounced and concentrated over a shorter period than at Durham; summer warming is more prominent at Durham. Therefore, seasonal lapse rates have changed and become more divergent between winter and summer half-years. Rather than changes in the dominance of particular air-masses over the study area accounting for seasonal lapse rate changes, there had been significant steepening of lapse rates over time for some air flow types. Minimum temperatures have increased more than maximum temperatures at Moor House with a consequent significant reduction in diurnal range. At Durham minimum and maximum temperatures have changed in approximate tandem so that there is no significant change in diurnal temperature range. The frequency of air frosts reduced by approximately 23% (from 129 days per year to 99) at Moor House from 1991–2006 compared to 1953–1980, and this may have major implications for the functioning of terrestrial and aquatic systems in the region. These results demonstrate that while overall annual warming in the uplands and lowlands within particular regions may be similar, even where annual lapse rates have been maintained, the seasonal distribution of that warming may be quite different even over short distances and so lowland records are not a reliable indicator of regional upland temperature change. Copyright © 2010 Royal Meteorological Society

1. Introduction

Regional climate models predict a warming, with respect to 1961–1990, of about 3 °C by 2100 for many areas of northern Europe (IPCC, 2007). Although there are many long-term temperature records in the UK such as at Durham (records available since 1850) these are largely confined to lowland areas. A number of authors have suggested that, due to changing surface-based lapse rates, upland areas may not warm in parallel with the lowlands (Pepin, 1995; Diaz and Bradley, 1997). Climate change in the UK uplands is particularly important because the soils are predominately organic and peaty (Holden et al., 2007) and these soils are significant and sensitive carbon stores (Hilbert et al., 2000; Bradley et al., 2005) which may become net sources of carbon under climate change (Worrall et al., 2004). These uplands also contain internationally important ecological habitats and are source areas for water supply (Holden et al., 2007). Where instrumental records of climate exist in the uplands there is therefore a need to ensure their use in understanding climate change (Appenzellera et al., 2008).

The aim of this paper is to compile a homogenized upland temperature series since 1931 for an internationally important upland research site in northern England so that warming trends may be compared with nearby long-term lowland records. This will enable us to test whether regional lowland temperature records can be used to infer local upland temperature change where the records are more scarce. In order to do this, annual, seasonal, monthly and daily mean temperature, and mean maximum and minimum temperatures will be analysed and compared between the sites. The paper assesses whether systematic changes in the lapse rate between the sites have occurred and assesses possible synoptic controls on lapse rate changes.

Moor House is a World Biosphere Reserve in the north Pennine region of northern England. The habitat is internationally important blanket peat moorland. The area is part of the largest contiguous land above 500 m in England. The site has been a focus of decades of scientific research relating to ecology, hydrology, biogeochemistry, geomorphology and climate. Indeed between 1993 and 2006 alone there were 161 ISI journal papers based on work at the site, plus 21 PhD theses, at least 28 MSc theses and 14 book chapters. The site is therefore of major scientific importance. However, while most of the papers and theses based at Moor House report local climate conditions, they are not consistent in the data they present. This is because scientists report data from different periods and using different techniques (Holden, 2007). For example, some report mean annual temperatures based on an average of hourly automatic weather station data since 1991 (e.g. Worrall and Burt, 2005), while others report the mean annual temperatures, or seasonal variability, based on the daily mean of the daily maximum and minimum observations recorded between 1931 and 1980 (e.g. Clark et al., 2005). There is therefore a need to present an analysis of climate data from the site in one published location that scientists can utilize in their future work.

Meteorological data were collected at Moor House between January 1931 and May 1980 (Manley, 1980). The pioneering work of Gordon Manley (who also produced the Central England Temperature Series dating from 1659; Manley, 1974) in instrumenting and analysing data from the site is described in Manley (1980), Lamb (1981) and Holden and Adamson (2001). The site was re-instrumented with an automatic weather station in May 1991 which has been running ever since. Data up to the end of 2006 will be presented in this paper. For the missing period of May 1980 to May 1991 a nearby weather station was operating at Widdybank Fell which can be used to help compile the complete record. There have been occasional short periods when data have been lost from the Moor House datalogger at the weather station since 1991. However, by using weather stations with short but overlapping records in the vicinity it has been possible to compile a full daily temperature series for the site from January 1931 to December 2006. This paper presents the compilation of such a series.

2. Homogenizing the temperature record

The Moor House meteorological station (Figure 1) lies at 556 m altitude near the sources of the rivers Tees and South Tyne in the northern Pennines (54.690N, 2.375 W, decimal degrees). This upland area is covered with a blanket peat deposit dominated by Sphagnum, Eriophorum and Calluna vegetation. The station is named after a lone house (Moor House) that used to exist some 7 km distant from the nearest village. Gordon Manley set up the station in 1931. At that time this was the highest point in the British Isles where records had been kept, except for Ben Nevis in Scotland where a weather station had operated from 1883 to 1898. Four temperature datasets are available for use from Moor House. Data from 1931 to 1952 are standardized to 0900 daily readings. Two different techniques had been used within that period (and there is a gap between 1947 and 1952), but Manley (1980) performed an excellent but protracted cross-calibration to produce a continuous and consistent record. The second dataset runs from 1953 to 1980 when Nature Conservancy operated the station. This dataset differs only very slightly because the location of the instrumentation is around 80 m distant from the earlier site that was originally located close to the house itself. Manley (1980) provided correction figures for the earlier data to bring them into line with Nature Conservancy data and so the 1931–1980 data were homogenized and are used in the present study. The third dataset comes from the UK Environmental Change Network (ECN) automatic weather station installed in 1991 as part of a national set of long-term standardized monitoring sites (Sykes and Lane, 1996; Morecroft et al., 1997). This dataset continues up to the present day.

Figure 1.

Location of the two stations used in this study. Shaded areas are those above 400 m within the North Pennines and the Lake District of northern England

2.1. 1980–1991 gap

The Moor House record has missing data between May 1980 and May 1991. However, a nearby weather station at Widdybank Fell operated from 1974 to 1995. Widdybank Fell lies 6.6 km southwest of Moor House at 515 m altitude at 54.653N, 2.321 W (decimal degrees) and is situated within heather moorland. This paper uses the full length of overlapping data available from both Moor House and Widdybank Fell (i.e. June 1974 to April 1980) to produce a calibration that can then be applied to the May 1980 to May 1991 period. The earlier Moor House record then also needs to be calibrated to the ECN instrumentation period (1991 onwards) to produce a homogenized record [Equation (3) below]. Calibration equations for the 1980–1991 period were developed based on both monthly and daily means. For analysis of annual and monthly mean temperatures in this paper the monthly calibration equations were used because the errors were much smaller than for daily temperatures and hence the calibrations were more reliable. However, a daily mean temperature calibration was also required for the 1980–1991 period in order to investigate synoptic controls on changing lapse rates between Moor House and Durham; the daily mean calibration equation (4) was used for daily mean temperature analysis.

The infilling equations (in degrees Celsius) for monthly means were determined as follows at both Widdybank Fell and Moor House (taking the sum of the maximum and minimum temperature recorded for each day and then dividing by two):

equation image(1)
equation image(2)

These equations can then be applied to fill in gaps in the monthly mean temperature record at Moor House. Equation (3) can be obtained by re-arranging Equations (1) and (2) which would determine a small correction factor for any differences between the Manley record and the ECN record at Moor House:

equation image(3)

Daily means were also calculated for Moor House from 1980 to 1991 in order to enable investigation of synoptic controls on lapse rates between Moor House and Durham (see below) over time. The equation for daily mean temperature was:

equation image(4)

Daily means were used for gap filling instead of performing analysis separately by daily maxima and minima. This was because the scatter was quite large for the relationship between Moor House maxima and the Widdybank Fell maxima (r2 = 0.84), and hence in order to ensure the analysis was as robust as possible, we did not homogenize the maxima and minima for the gap period separately.

2.2. Gaps since 1991

There are three short periods where data are missing since 1991 due to failure of the automatic datalogger on site. These are 10 February to 23 March 1993, 4 June to 27 July 2004 and 18–22 October 2006 (all dates inclusive). For these periods daily mean calibration equations were used to homogenize the record. For the first of these periods the correlation with the Widdybank Fell station was used to infill the days with missing data. For the latter two periods the daily mean temperature data have been infilled using a station at Hunt Hall Farm. Hunt Hall has operated since 1 December 1995 and is located at 54.670 north, 2.278 west (decimal degrees), at 370 m altitude some 8 km from the Moor House station and only 3 km from the Widdybank Fell station. The equation for daily mean temperature was:

equation image(5)

It is recognized that the Hunt Hall Farm station is some 184 m lower than Moor House and so this limitation to our approach should be noted. Indeed Pepin et al. (2009) noted important differences between the Hunt Hall Farm and Widdybank Fell records particularly as the Hunt Hall Farm site is more sheltered. However, the periods for which data from Hunt Hall Farm are required for infilling the Moor House record are short in comparison to the whole dataset, and so, accepting an r2 of 0.88, we have included these 59 days of infilling in the homogenized temperature record.

2.3. Durham record

Daily temperature data are available from Durham since 1850. Data from Durham are used for comparison (of temperature changes annually and seasonally) with the upland record from Moor House. Data from Durham are also used to investigate daily near-surface lapse rate (with Moor House) relationships with synoptic air flows. Durham is the closest lowland station with an almost complete record for the same time period as at Moor House and it is also one of the key long-term instrumental records in the UK. Lack of urbanization around the Durham site has meant that external influences are minimal (Burt and Horton, 2007) unlike other important long-term stations such as the Radcliffe Observatory, Oxford. The Durham station is located 53 km northeast of Moor House at 54.768N, 1.585 W (decimal degrees), 860 m southwest of Durham Cathedral, and south of the river Wear (Figure 1). The site is open and well exposed at a height of 102 m above sea level. Temperature data from 1922 to 1933 are presented after having applied a minor correction factor as outlined by Manley (1941) accounting for some instrumental calibration. There are five recent short periods where the automated logger failed at Durham: (a) 1–10 October 1999; (b) 31 October 1999; (c) 26–28 October 1999; (d) 2–25 May 2004; and (e) 16 August–7 September 2005 (all dates inclusive). For all of these periods we used a calibration based on daily mean temperature data from other stations to produce a complete Durham record. For (a), (b) and (c) data from Wycliffe Hall station which is at 120 m altitude some 25 km southwest and which ran from 1 Jan 1989 to 30 November 2000 were used and the equation for daily mean temperature during this period was:

equation image(6)

For (d) and (e) data from Ravensworth station were used to infill the series. Ravensworth station is at 120 m altitude some 35 km southwest and ran from 1 November 2002 to present. The equation which was derived from data from 1 November 2002 to 31 December 2006 was:

equation image(7)

Ravensworth and Wycliffe Hall were the closest stations with complete data for the overlapping periods.

3. Results

Table I provides summary of climate data based on observations at Moor House. The mean annual temperature at Moor House between 1931 and 2006 was 5.31 °C. Figure 2 shows the full time series of mean annual temperatures for Moor House and the lowland Durham site. The relatively cool 1961–1990 period within the context of the warmer 1940s and 1950s can be seen on both records as can the recent warming since the late 1980s. The mean annual warming was approximately equal between Moor House and Durham so that the mean annual surface lapse rate has been maintained (Table II). The warming in the mean annual temperature of the 1991–2006 period compared to the 1931–1960 period was 0.61 °C at Durham and 0.53 °C at Moor House. The equivalent comparisons with the 1961–1990 period (often used as the international standard period for comparison; IPCC, 2007; Jenkins et al., 2007) are a warming of 0.71 °C at Durham and 0.73 °C at Moor House.

Figure 2.

Mean annual temperatures for Moor House (1931–2006) and Durham (1851–2006) with five-year moving averages

Table I. Summary of climate data for Moor House for periods for which they exist
Variable1931–20061953–1980 1991–2006Specific date
  • The cells marked with

  • *

    have been left empty because of the gap during the 1980–1991 period. Other empty cells are where no data were collected.

  • #

    For both 1953–1980 and 1991–2006.

Temperature ( °C)
Mean daily maximum*8.29 8.78 
(standard deviation) (5.78) (5.64) 
Mean daily minimum*1.96 2.87 
(standard deviation) (5.09) (4.78) 
Mean5.315.13 5.81 
Absolute minimum−18.5   31 Jan 1972
Absolute maximum27.6   9 Aug 2003
Mean air frosts per year*129 99 
(standard deviation) (19) (22) 
Mean days when temperature was both above and below zero per year*98 81 
Mean humidity   88% 
(standard deviation)   (5%) 
Precipitation (mm)
Mean annual total  2012#  
(standard deviation)  (470)#  
Maximum in any year  3372# 2000
Minimum in any year  1434# 1973
Mean ppt days per year  244#  
(standard deviation)  (15)#  
Mean number of days snow lying 69 41 
Maximum consecutive days snow lying  65# 1st February–5th April 1969
Wind
Mean wind speed (m s−1)   4.3 
Mean wind direction (°N)   221 
Sunshine and visibility
Mean annual sunshine (h)   1170 
Annual fog days (visibility < 1 km at 0900 h) 52   
Table II. Mean yearly and half-yearly temperatures and lapse rates for measurement periods at Moor House and Durham
Measurement periodMoor House ( °C)Durham ( °C)Mean lapse rate ( °C/100 m)
 Oct–MarApr–SeptYearOct–MarApr–SeptYearOct–MarApr–SeptYear
1931–20061.928.655.315.2812.278.790.740.800.77
1931–19601.758.755.285.1612.268.700.750.770.75
1961–19901.728.405.085.1112.048.600.750.800.78
1991–20062.628.965.815.8212.749.310.710.830.77

There has been a much larger warming in the winter months of January and February than other months of the year. Those months at Moor House which are significantly warmer (at p < 0.05, using a Mann–Whitney U-test) between 1991 and 2006 compared to 1931–1960 and 1961–1990 are shown in Figure 3. Similar data are shown for comparison for the lowland Durham record. At Moor House the warming is only significant (using p < 0.05 throughout) over both the 1931–1960 and 1961–1990 periods for January and February (Figure 3). At Durham, March is also included in this category. Figure 3 shows that warming between the summer months of May–September is significant and greater for May, July and August at Durham compared to Moor House (where only August warming is significant). Conversely, winter warming is greater at the upland Moor House site. The result of these differences is that seasonal lapse rates have changed. Table II shows that for the 1991–2006 period compared with earlier periods the lapse rates have become shallower (less steep) for the winter half-year between the two sites, while the lapse rates have become steeper for the summer half-year (Table II). Figure 4 shows the trend for changes in October–March and April–September surface lapse rates between the two stations for the whole time period. The trend is significant for October–March at p = 0.023 while it is significant at p < 0.001 for April–September.

Figure 3.

Mean monthly temperature increase at (a) Moor House and (b) Durham for the period 1991–2006 compared with 1931–1960 and 1961–1990. Fully shaded bars indicate months where temperature difference is statistically significant. Striped bars indicate those months where there is no significant difference to the 1991–2006 period

Figure 4.

Surface lapse rates ( °C per 100 m) between Moor House (556 m) and Durham (102 m) for half-year periods between 1931 and 2006. Straight lines indicate best-fit trend lines which are significant at p < 0.023 for October–March and p < 0.001 for April–September time series

Daily temperature data were used to compute daily lapse rates between the two stations. Each day was also assigned a synoptic air flow using the Lamb classification (Lamb, 1972) which is now compiled by the Met Office Hadley Centre. The daily Lamb classifications for the study period were obtained from the British Atmospheric Data Centre.

Lapse rates are significantly controlled by synoptic airflow (Figure 5) with shallower lapse rates associated with southerly airflows and steeper lapse rates for northerly flows. There were no significant changes over the study period, when considered as a whole, in the frequency of any of the 27 classes of airflow. When data from the winter six months are separated from the summer six months, the only significant changes in air flow frequency, out of all 54 possible air flow categories (27 summer and 27 winter), were a reduction in straight westerly (W) air flows in summer (p = 0.023) and increased southwesterly (SW) airflows in winter (p = 0.013). These changes are not sufficient to explain the significant changes in lapse rates seen between the stations over time. However, comparing Figure 5(b) and (c) suggests that changes in lapse rates within particular synoptic conditions may be important. Significant changes with time in daily lapse rates were found for cyclonic westerly (CW) flows (c = 0.151, p = 0.001), cyclonic (C) flows (c = 0.105, p < 0.001), cyclonic northerly (CN) flows (c = 0.014, p = 0.005), westerly (W) flows (c = 0.062, p = 0.006) and unclassified (U) airflows (c = 0.180, p = 0.008). All of these relationships were positive suggesting steepening lapse rates over time under these synoptic conditions. Synoptic conditions have different effects on lapse rates depending on whether it is day (affecting the maximums) or night (affecting the minimums) and these patterns are discussed by Pepin (2001) and are not repeated here.

Figure 5.

Box plots of lapse rates for each synoptic class (Lamb) for (a) the whole study period; (b) 1961–1965 and (c) 2001–2005. A = anticyclonic, C = cyclonic, N = northerly (straight), NE = northeasterly (straight), CNE = cyclonic northeasterly

While it is possible for the mean annual temperature to increase due to both rising minimum and maximum daily temperatures, it appears that increases in minimum temperature are greater with a consequent reduction in the diurnal range at Moor House (Figure 6(a)). Minimum temperatures at Moor House during February, for example, are 2.2 °C warmer between 1991 and 2006 than between 1953 and 1980. Data are only presented for maximum and minimum temperature from 1953–1980 compared to 1991–2006 in Figure 6 to ensure robust reliability of comparison of maxima and minima, which is particularly important for analysis of air frost frequency. This is because the homogenization of temperature data for the 1980–1991 gap period at Moor House was much stronger when daily means were used rather than performing the analysis separately on minima and maxima. At Moor House, maximum temperatures are significantly greater during the more recent period for only three months of the year, while minimum temperatures are significantly greater for eight months of the year. The patterns of change in mean maximum and minimum temperatures at Durham are rather different than at Moor House (Figure 6(b)). Maximum temperatures are significantly greater during the 1991–2006 period for seven months and minimum temperatures for eight months of the year at Durham. The impact on diurnal temperature range at Moor House is that it is significantly reduced from October to February inclusive (Figure 7(a)), while the pattern is quite different for the same time periods at Durham where there are no significant differences in diurnal range except in February where the range has increased (Figure 7(b)). The median annual diurnal range is significantly lower at Moor House (p = 0.004) in the 1991–2006 period (5.8 °C) compared to the 1953–1980 period (6.3 °C), while there was no significant difference in the annual diurnal range at Durham between these two periods.

Figure 6.

Monthly mean difference in daily maxima and minima between 1953 and 1980 compared with 1991 and 2006; (a) Moor House, (b) Durham

Figure 7.

Monthly mean difference in diurnal range between 1953 and 1980 compared with 1991 and 2006; (a) Moor House, (b) Durham. Fully shaded bars indicate months where temperature difference is statistically significant at p < 0.05. Striped bars indicate those months where there is no significant difference to the 1991–2006 period

Given that mean temperatures for several winter months are close to 0 °C at Moor House the large increase in temperature during the winter (when compared to insignificant summer changes) may be of major biological and geomorphological significance in peatlands (Tallis, 1973). Air frosts have been recorded during all months of the year at Moor House but they have declined in all months of the year except October (Figure 8). There has been a decrease in the annual number of days with air frost from 129 during the 1953–1980 period to 99 per year between 1991 and 2006. The mean number of days per year when the air temperature was both above and below freezing (and hence when freeze-thaw processes might be more likely) was 98 during 1953–1980 and 81 days for the 1991–2006 period. Changes in winter freezing patterns have also been associated with large reductions in snow cover at Moor House. Snow cover at the site is synoptically controlled and a typical winter season will see several complete accumulation and melt cycles. Comparison of the two snow recording periods (1953–1980 and 1994–2006) indicates a reduction in mean snow-lie days from 69 to 41 (Table I). Manley (1971) quoted a mean of 72 days with snow lying between 1949 and 1970 at Moor House with 106 days on Cross Fell at 893 m altitude at the study site.

Figure 8.

Mean frequency of frost days per month at Moor House (a) and Durham (b) comparing the 1991–2006 period with 1953–1980

4. Discussion

Data from Moor House can be used to produce the longest UK upland instrumental temperature record. In the context of global studies, a 76-year instrumental record may not seem all that long. Chapter 3 of the IPCC (2007) report focusses on surface temperature observations and yet it makes no mention of observed differences in upland and lowland records. However, there are very few descriptions of differences in upland and lowland temperature change within regions based on long-term instrumental records; most analysis relies on short term datasets. Hence, much earlier work has been inconclusive as to whether upland sites are warming faster or slower than nearby lowland sites. For example, inconsistent changes in surface lapse rates have been found in the instrumental records of the European Alps (Beniston et al., 1994; Beniston and Rebetez, 1996) and the Rocky mountains (Pepin, 2000). Diaz and Bradley (1997) examined 116 global stations and found that some high-elevation sites showed enhanced warming although there were few long-term local upland–lowland records. Liu and Chen (2000) reported steepening surface lapse rates in the Tibetan Plateau while Vuille and Bradley (2000) reported more gentle lapse rates (i.e. lower elevations warming faster than higher ones) in the Andes. Pepin and Lundquist (2008) examined over 1000 upland stations and found that there has been no simplistic elevational increase in warming rates but that observed temperature trends are most rapid near the annual 0 °C isotherm due to snow-ice feedback.

If we had only analysed lapse rates based on annual differences between Moor House and Durham then we would have come to the conclusion that there were no significant changes. However, the seasonal contrast in lapse rates has significantly strengthened. Pepin (2001) showed that the Durham-Widdybank Fell temperature contrast is a fair representation of the broader regional upland/lowland climatic gradient. Therefore, using nearby Moor House with its homogenized record allows a much longer-term and reliable investigation of regional lapse rate changes. While in the 1990s some had suggested that temperature changes at Moor House were not significant within the context of the warming felt across the lowlands of Britain (Garnett et al., 1997), it is clear that the recent temperature changes recorded at the upland Moor House site are indeed significant and during some seasons they are in fact more prominent than in the nearby lowland Durham record. While the overall pattern of warming is similar from year to year between Moor House and the nearby lowland Durham site (e.g. Figure 2), the distribution and magnitude of that warming in terms of seasonal and diurnal patterns are quite different. This difference comes out more strongly than that shown in the recent UK-wide analysis of climate trends (Jenkins et al., 2007), highlighting that the changes in climate that have been observed and which may occur in the future might vary greatly over just short distances. The changes in the climate observed are very different between the upland and lowland sites examined in the study, despite the distance between the sites being only 53 km. Therefore, the observed impacts are likely to be different not simply because the upland site is cooler, but because the distribution of changes observed at the upland site is quite different.

Surface-based lapse rates are strongly influenced by synoptic climatology (Barry et al., 1981). Changes in the prevalence of air-mass flow, or convective activity, over the British Isles may therefore play a role in the differences found in seasonal lapse rates during our study. However, out of 27 synoptic classes we only found one significant change in the winter half-year (increased straight southwesterly airflows) and one in the summer half-year (reduction in straight westerly air flows). These do not explain the significant changes in summer and winter lapse rates that we found. Importantly, however, a significant steepening of lapse rates within a number of synoptic air flow types (westerly, cyclonic westerly, cyclonic northerly and cyclonic) was found.

Most air-masses moving over the British Isles are westerly from the Atlantic and so sea surface temperatures in the North Atlantic are likely to be important, potentially influencing lapse rates within westerly or cyclonic air-masses producing a broad regional-scale signal. Polyakov et al. (2010) have found a general warming trend in the North Atlantic since the 1920s although this was often accentuated by multi-decadal variability. Based on a mean daily temperature record, as westerly air-masses move inland from a warmer than usual sea, they will more often become cooled from beneath inland thereby resulting in an overall increased relative stability inland and shallower daily mean lapse rates. This compares to a cooler sea which would increase instability inland and strengthen mean daily lapse rates. The role of upwind sea surface temperatures would explain why changes in mean daily lapse rate are seen mainly in the westerly synoptic types.

Pepin (2001) showed that increased (warm) sea surface temperature anomalies to the west of the UK correlated with weaker lapse rates by night, and to a lesser extent during the day, although most strongly in different parts of the North Atlantic. We did not examine daytime and night-time daily lapse rates in our data as we only analysed mean daily temperature to keep the analysis as robust as possible (due to the infill calibrations for maxima and minima for the 1980–1991 period being a lot weaker than those for mean daily temperature). However, Pepin's (2001) findings may well be manifest at a seasonal scale in our study as the winter half-year has longer night-time hours and we have found a significant weakening of lapse rates in our study for the October–March half-year. Conversely, we found a significant strengthening of lapse rates in the summer half-year which may be associated with daytime lapse rate strengthening. This suggests that our observations, which cover a much longer time period than that used by Pepin (2001), are broadly consistent with those in that earlier study.

Pepin (2001) reported that there was a significant change between 1968 and 1995 in the latitudinal gradient in sea surface temperatures in the North Atlantic immediately to the west of the British Isles with more frequent periods of lower temperatures in the north and higher temperatures in the south. Steepening of latitudinal sea surface temperature gradients were found to be consistent with the trend towards steeper lapse rates by day and weaker ones at night. Whether these changes in North Atlantic sea surface temperature gradients are really a driver of findings manifest in our data through steeper lapse rates in summer and weaker lapse rates in winter, however, requires further investigation.

Our analysis of minima and maxima only used data from time periods when data were complete (i.e. no calibrated data). This was to ensure that the analysis was robust since maxima and minima calibrations were found to be the weakest between Widdybank Fell and Moor House while the mean daily and monthly temperature regressions were very strong. Using the intact periods of the record we found that minimum temperatures have increased more than maximum temperatures at Moor House with consequent significant reduction in the diurnal range. At the same time, minimum and maximum temperatures have changed in approximate tandem at Durham and there is therefore no significant change in diurnal temperature range at the lowland site. The winter warming and large increase in winter daily minimum temperatures has been very marked at Moor House when comparing 1991–2006 with the earlier 1953–1980 period when there is comparative maximum/minimum data. The consequent reduction in the frequency of air frosts represents a 23% decline which is close to the UKCIP08 (2008) 25% frost reduction prediction for northern England by 2065 (compared to 1961–1990). This may be significant for erosional and biological activity in the upland peatlands of the UK with impacts for plant communities and soil organisms which in turn impact stream water quality (e.g. Cole et al., 2002). Peatlands are very sensitive to climate change, and minor changes in temperature or hydrology can have marked impacts on their function and carbon cycling (Holden, 2005). Thus, the spatial and temporal nature of temperature change in upland peatland areas is of considerable importance.

The significant reductions in frost and snow cover (Table I) observed at Moor House may provide an important mechanism that has contributed to decreased winter lapse rates. Pepin and Lundquist (2008) suggested that there were faster warming rates at lower temperatures, with particularly enhanced warming at temperatures near freezing. This feedback should not be consistent with elevation but theoretically should be greatest around 0 °C. They suggested that this process was related to reduced albedo through decreases in snow and ice cover at altitudes around 0 °C. It is notable that mean winter temperatures at Moor House are close to zero (1.0 °C in December, 0.0 °C in January and −0.2 °C in February). So it is likely that the Moor House–Durham region fits into this category and this provides one additional causal mechanism for the observed decreases in winter lapse rates (i.e. faster rates of warming at the upland site compared to the lowland site).

While its location is not at a particularly high altitude by world standards, the site is itself of international importance through its designation as a World Biosphere Reserve. The climate data for Moor House presented in this paper will be of use to the scientists working at the site over the coming years. It is recommended that workers at the site use data presented in this paper as the standard climatological data when describing site conditions in their manuscripts (e.g. taken from Table I or Table II).

5. Conclusion

If this paper had assessed mean annual temperatures alone then we would conclude that warming at the upland and lowland stations has occurred in tandem. However, our analysis shows that the trends in temperature vary greatly between the two stations on a monthly and seasonal basis. The differences between the changes observed at Moor House compared to Durham include: (1) winter warming is dominant at Moor House and more pronounced than at Durham; (2) summer warming is more prominent at Durham than Moor House; (3) minimum temperatures have increased more than maximum temperatures at Moor House with consequent significant reduction in the diurnal range; (4) minimum and maximum temperatures have changed in approximate tandem at Durham where there is no significant change in diurnal temperature range.

While there were no significant changes in the surface lapse rates for the mean annual temperature, there were significant differences in lapse rates within the winter and summer half-years. Indeed the difference between the lapse rate in winter and that in summer has significantly increased. It is therefore concluded that we cannot use nearby lowland records to infer the nature of temperature change in the uplands and that more work is required to understand controls on regional lapse rates and their changes within different regions. Most climate models suggest enhanced warming at higher altitudes (e.g. Chen et al., 2003). However, our research suggests that caution is needed and that changes in lapse rates may be seasonal such that the lapse rate contrasts between seasons are enhanced while in some regions no changes in the overall annual lapse rate will be observed. Thus, if results from models are downscaled based on mean annual temperature changes then this may mask more important and significant seasonal lapse rate changes.

Future temperature change in upland areas will be spatially variable. This is because the response of local air-mass types is important. We found significant increases in lapse rates for four (out of 27) air-mass types during our study period. Therefore, future changes in near-surface lapse rates under climate change may be different between regions due to the differential local importance (and changes) of air-mass characteristics within regions and the different types of air-masses that dominate regions. Thus, we are unlikely to observe consistent changes in near-surface lapse rates across the globe (Pepin and Lundquist, 2008) and patterns are likely to vary regionally. There were significant changes (steepening) of lapse rates over time for a few westerly/cyclonic air flow types in our study region.

Additionally, it is suggested that decreased near-surface winter lapse rates may be related to reduced snow cover and albedo at upland sites. Theoretically, where snow and ice cover are common, temperature increases should be more enhanced closer to 0 °C. The mean winter temperature at Moor House for December, January and February is 0.2 °C. Hence, local albedo changes need to be taken into account when assessing whether lowland records can be used to infer nearby upland temperature change.

This paper has highlighted the importance of lapse rate studies and the need to understand the controls on lapse rates within local regions (e.g. synoptic controls and the impact of changing synoptic controls on lapse rates, and controls related to changing snow cover) because upland temperature changes cannot be inferred from lowland records alone. This has been shown to be the case even when the near-surface lapse rate of mean annual temperature is static. These local controls may mean that downscaling of climate model results to infer near-surface lapse rate changes need to be mediated by an improved regional understanding of near-surface lapse rate controls. These regional and local processes also mean that we are unlikely to observe a consistent pattern of near-surface lapse rate change across the world and re-affirm the need to monitor temperature at more high-altitude sites.

Acknowledgements

The work was funded by a Philip Leverhulme Prize awarded to Joseph Holden. We are grateful to the Environmental Change Network and the British Atmospheric Data Centre for data supplied for this project. We are grateful to those stations that provide data to the British Atmospheric Data Centre for scientific use and as such we acknowledge the original sources of those data, including the University of Durham. The authors are grateful to John Adamson (retired member of CEH Lancaster) for the additional information supplied to aid this project. Thanks are also offered to the referees who helped improve the paper.

Ancillary