Hydrobiologia

, Volume 646, Issue 1, pp 49–59

The long-term (1979–2005) effects of the North Atlantic Oscillation on wind-induced wave mixing in Loch Leven (Scotland)

Authors

    • Bryan Spears, Centre for Ecology & Hydrology
  • Ian D. Jones
    • Centre for Ecology and Hydrology, Lancaster Environment Centre
SHALLOW LAKES

DOI: 10.1007/s10750-010-0188-9

Cite this article as:
Spears, B.M. & Jones, I.D. Hydrobiologia (2010) 646: 49. doi:10.1007/s10750-010-0188-9

Abstract

We report on long-term covariation (1979–2005) between indices of the North Atlantic Oscillation (NAO) and wind speed and direction in Loch Leven. The effects of the observed variations in wind speed and direction were combined to produce modelled wave-mixed depths (Zc). Positive correlations were observed between seasonal and annual wind speeds and westerly frequency and indices of the NAO that are in line with general perception: positive NAO was correlated with stronger, more westerly winds and these correlations were strongest in winter and spring. Correlations between NAO and estimates of Zc were strongest in the most westerly exposed site in spring (r2 = 0.701; Zcspring versus spring NAO index). On average, over a 25-year period Zc was deeper in spring and shallower in summer. Major anomalies from the 25-year seasonal means were observed in 1982, 1979, and 1991. Annual average Zc was low in the late 1970s and early 1980s (shallowest average annual Zc of 1.0 m (1984)), high in the late 1980s and early 1990s (deepest average annual Zc of 1.9 m (1990)) and moderate in recent years (up to 2005). This study has major implications for our understanding of potential climate change drivers and the related responses of shallow lake ecosystems, including alterations to littoral habitat quality and benthic–pelagic coupling.

Keywords

North Atlantic OscillationWindWavesLoch LevenHabitat disturbance

Introduction

The North Atlantic Oscillation (NAO) index has been used as a proxy for weather conditions, and is calculated by comparing patterns of sea level atmospheric pressures near the ‘Icelandic low’ (Southwest Iceland (Reykjavik) in the present study) and the ‘Azores high’ (Gibraltar in the present study) of the North Atlantic Ocean (Jones et al., 1997, 2003). The main climatic variations linked with strong positive NAO include higher temperatures and stronger, more westerly winds from late-autumn through early-spring (Slonosky et al., 2000). Past studies have identified NAO as a good (although geographically variable; Gerten & Adrian, 2001) indicator of climatic forcing of processes in European lakes (reviewed by Straile et al., 2003). For example, positive correlations have been observed between indices of the NAO and surface and hypolimnetic lake water temperatures (Livingstone, 1999, 2000; Livingstone & Dokulil, 2001; Dokulil et al., 2006) as well as lake chemistry variables (Monteith et al., 2000; Evans et al., 2001; Weyhenmeyer, 2004). These trends have, in turn, been linked to alterations in plankton dynamics and the onset of the spring clear-water phase (Gerten & Adrian, 2001; George, 2000; Straile et al., 2003).

In shallow lakes, the temperature effects of the NAO may drive ecosystem functioning via the regulation of steady-state change (Scheffer et al., 2001), though the evidence is equivocal (Jeppesen et al., 2003; Van Donk et al., 2003). However, the relationships between NAO and wind speed and direction have received relatively little attention. Further, associations between NAO-driven variation in wave mixing have been suggested (George, 2000), although not investigated in detail. Noges (2004) reported a positive correlation between an index of the Gulf Stream position and wind speed in Võrstjärv, a large shallow lake in Estonia.

Although the effects of temperature are clearly important in regulating the mixing (e.g. through convective processes) of deeper lakes, we focus here on modelling the depth of the wave-mixed layer as an indicator of wind-induced wave mixing in shallow lakes. This may be estimated using simple wave theory which combines the effects of wind direction and wind speed to produce a modelled wave-mixed depth. Variations in the depth of the wave-mixed layer have previously been shown to affect sediment disturbance (Hilton, 1985; Douglas & Rippey, 2000) and internal nutrient loading (Hamilton & Mitchell, 1997; Ogilvie & Hamilton, 1998).

We estimated the spatial (kilometre scale) and temporal (across years and seasons) variability in the modelled wave-mixed layer in Loch Leven, Scotland, over a 25-year period (1979–2005). Long-term data were used to test (1) the seasonal and annual covariation between local wind speed, westerly frequency and indices of the NAO and (2), the effects of the NAO on the modelled depth of the wave-mixed layer in a shallow lake.

Methods

Study site

Loch Leven covers 13.3 km2 and is shallow (mean depth 3.9 m). It has a long, and well-documented history of ecological responses to environmental change, including eutrophication (Bailey-Watts & Kirika, 1999; Spears et al., 2007b), subsequent nutrient control and recovery (Carvalho & Kirika, 2003), alteration to community ecology (Ferguson et al.2008) and temperature variation associated with climate change (Ferguson et al., 2008). The Loch Leven long-term monitoring programme is ongoing (1967—present) and includes data on over 150 variables. The site is one of few with such a data set, making it an excellent case study for the analysis of long-term climate change effects.

Daily average wind speed and direction measurements have been recorded on the western shore of Loch Leven using a cup anemometer and wind vane since 1979. Few data gaps exist, and those that do are mainly in July (owing to holidays) and for less than 2-week periods. There is a gap between 1999 and 2003 due to alterations that were made to the data recording and storage procedures. These gaps have been accounted for in the calculations of means as described below. In this study, we assume that wind speed and direction measured on the shore is characteristic of the entire loch. However, actual wind speed over the loch’s surface is likely to have been slightly higher (Schwab & Morton, 1984).

Six sites (Fig. 1) representing contrasting areas of the loch to wind exposure were used. Data on these sites for sediment and water chemistry have previously been published (Spears et al., 2006, 2007a, b).
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0188-9/MediaObjects/10750_2010_188_Fig1_HTML.gif
Fig. 1

Bathymetric map of Loch Leven showing location of sample sites

Monthly wind speed and westerly frequency

The daily wind speeds and directions were averaged to provide monthly wind speeds and westerly frequencies, the latter being calculated as the percentage of days in the month over which the wind direction lay between 181° and 359°. Where wind variables were not measured on every day in the month (most commonly July), monthly averages were estimated using available data. Annual values and means of wind speed and westerly frequency were calculated for winter (December–February), spring (March–May), summer (June–August), and autumn (September–November).

Modelling the wave-mixed depth

For waves with a wavelength less than half the water depth, wavelength λ (m) and time period T (s), are related by the equation
$$ \lambda = {\frac{g}{2\pi }}T^{2} , $$
(1)
where g is the gravitational acceleration (9.8 m s−2). Waves are often characterised by the significant wave height, i.e. the average of the highest third of all waves, and a corresponding significant wave period Ts (s) and significant wavelength λs (m). Smith & Sinclair (1972) verified for Loch Leven, the U.S. Army (1962) wave prediction equation, showing that the significant wave period was dependent on both the effective fetch (selected using average daily wind direction in this study) F (m) and the wind speed W (m s−1)
$$ T_{\text{s}} = 0.46{\frac{W}{g}}\left[ {{\frac{gF}{{W^{2} }}}} \right]^{0.28} . $$
(2)
Defining the depth of the wave-mixed layer Zc (m) to be half the significant wavelength (Smith & Sinclair, 1972),
$$ Z_{c} = {\frac{{\lambda_{s} }}{2}}, $$
(3)
means Eqs. 13 can be solved to calculate the depth of the wave-mixed layer as a function of fetch and wind speed
$$ Z_{c} = 0.0062F^{0.56} W^{0.88} . $$
(4)
The wind speed used in the original formulation of these equations was that at a height of 8 m. As the Loch Leven wind speed was taken at a height of 2.5 m, we applied a height correction to the wind speed, following the standard logarithmic law
$$ W_{8} = W_{2.5} {\frac{{\ln (8/z_{0} )}}{{\ln (2.5/z_{0} )}}}, $$
(5)
where W8 and W2.5 are the wind speeds at 8 and 2.5 m, respectively, and z0 is the surface roughness length, taken here to be 0.1 m as the appropriate value for grassland (Padisák, 2004). Formally, the wind speed should be the average wind speed over the whole fetch, but here, we follow Douglas & Rippey (2000) who showed that significant results regarding wave-mixed depths could be obtained by shore-based wind speeds. Estimates of the effective fetch across 360° for each site were calculated at 10° intervals and included the effects of all islands in Loch Leven (Håkanson & Jansson, 1983).

Monthly average Zc values were calculated from average daily Zc values for each year. Monthly values for the NAO index were obtained (http://www.cru.uea.ac.uk/cru/data/nao.htm).

Data presentation and statistical analyses

All statistical analyses were conducted using MINITAB version 14. All data were tested for normality (Kolomogorov–Smirnov test) and passed the normality criterion (P > 0.05; Townend, 2004). Correlation analyses were used to test the covariation between seasonal and annual values of wind speed, westerly frequency and NAO.

Visual comparisons of effective fetch, annual anomalies (when compared with average values calculated across all years (1979–1999, 2003–2005)) and variation in absolute annual Zc for each site were made. Correlation analyses were conducted to assess (1) linear trends between seasonal and annual indices of the NAO and the corresponding mean values for wind speed and westerly frequency, and (2) linear trends amongst all potential combinations of seasonal and annual values of Zc and NAO, over all years included (i.e. 1979–2005 excluding 2000–2002), for each site.

Two-way analysis of variance (ANOVA) was conducted to assess (1) the variation of average monthly Zc across sites, (2) effects of seasonality on the variation in average monthly Zc, and (3) potential effects of interactions between seasonality and site location on the variation in average monthly Zc.

Estimates of the percentage surface area of the loch in which the modelled wave-mixed layer reached the loch bed (i.e. indicative of sediment disturbance) were made. For each year, this was achieved by first calculating the average of the six annual average wave-mixed depths to obtain an average mixed depth across the loch. This loch average was then converted to surface area of the loch bed shallower than the wave-mixed depth using bathymetric survey data and assuming the average annual loch water level corresponded to the survey datum (106.87 m a.s.l.).

Results

Variation in effective fetch

The effective fetch varied across the sites (Fig. 2). The maximum effective fetch was estimated at 3,500 m in site 6 (wind direction: SE) with the minimum of less than 250 m estimated in site 1 (wind direction: SW). The largest ranges in effective fetch were observed in sites 1, 2, 3 (most sensitive to easterly winds) and 6 (most sensitive to south easterly winds) with sites 4, and 5 showing relatively constant effective fetch estimates across all wind directions.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0188-9/MediaObjects/10750_2010_188_Fig2_HTML.gif
Fig. 2

Estimated effective fetch for each of the six sites across 360°

Covariation between NAO, wind speed and westerly frequency

Both wind speed and the westerly frequency varied positively with NAO (Table 1). These correlations were strongest in winter and spring. A strong positive correlation was observed between wind speed and westerly frequency during winter with weaker, although still significant, correlations observed in annual and spring data. The ranges of annual westerly and easterly frequencies during the study period were 33.0–65.6% and 17.5–27.7%, respectively.
Table 1

Results of correlation analysis between seasonal values of the North Atlantic Oscillation (NAO), average wind speed (m s−1) and westerly frequency (WF) over the study period (α = 0.05)

Comparison

r2

P

NAO versus wind speed

 Annual NAO versus Annual speed

0.470

0.020

 Winter NAO versus Winter speed

0.591

0.002

 Spring NAO versus Spring speed

0.716

0.000

 Summer NAO versus Summer speed

 

NS

 Autumn NAO versus Autumn speed

 

NS

NAO versus WF

 Annual NAO versus Annual WF

0.449

0.028

 Winter NAO versus Winter WF

0.535

0.007

 Spring NAO versus Spring WF

0.576

0.003

 Summer NAO versus Summer WF

0.448

0.028

 Autumn NAO versus Autumn WF

 

NS

Wind speed versus WF

 Annual speed versus Annual WF

0.472

0.020

 Winter speed versus Winter WF

0.654

0.001

 Spring speed versus Spring WF

0.575

0.003

 Summer speed versus Summer WF

 

NS

 Autumn speed versus Autumn WF

0.662

0.000

Site specific temporal variation in wave-mixed depth

The modelled average (1979–2005) monthly Zc for each site is shown in Fig. 3. Two-way ANOVA results indicated that significant variation in the wave-mixed depth was indicated between sites and across months within each site, but no significant interaction effects were indicated between these two factors (Table 2). Zc was lowest in site 1 followed by sites 2, 3, 6, 4, and 5. In general, Zc was greatest in winter, decreased through spring and summer and increased again through autumn towards winter in all sites (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0188-9/MediaObjects/10750_2010_188_Fig3_HTML.gif
Fig. 3

Average (1979–2005) seasonal variation in the wave-mixed depth at each of the six sites

Table 2

Results of two-way ANOVA testing variation in average monthly wave-mixed depth across the six sites and with time (i.e. 12 months) within each site

Source

DF

SS

MS

F

P

Month

11

1.227

0.112

44.91

0.000

Site

5

6.239

0.125

502.41

0.000

Error

55

0.137

0.003

  

Total

71

7.604

   

R2 = 0.98

     

No interaction between time and sample was observed

Annual average Zc estimates were low during the late 1970s, early 1980s, and early 2000s and deepest during the late 1980s and early 1990s. However, this pattern was not found to be consistent across all sites and seasons. For example, although when considering annual conditions 1982 appeared to be an average year for Zc, closer inspection of the data revealed negative anomalies in winter and spring and a positive anomaly in autumn. Similar seasonal discrepancies were observed in 1979 and 1991.

Long-term seasonal and annual anomalies of Zc in relation to the 1979–2005 average values are shown in Fig. 4a–e. The variation between sites and across the years was most pronounced in winter and least pronounced in summer. Intra-site variation was highest in winter and lowest in summer. In general, similar long-term variation was estimated across all seasons where Zc was lower than average between 1979 and 1985, 1997–1998 and 2003–2005 and higher than average between 1986 and 1991.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0188-9/MediaObjects/10750_2010_188_Fig4_HTML.gif
Fig. 4

Variation in the (a) winter, (b) spring, (c) summer, (d) autumn and (e) annual wave-mixed depths in relation to the 1979–2005 average values. Variation in the absolute annual average wave-mixed depths for each site between 1979 and 2005 are also shown (f)

A number of discrepancies were evident with respect to the consistency of the observed variation across all seasons. For example, in 1982, Zcautumn was higher than the 1979–2005 average. However, little variation was apparent in Zcsummer or Zcspring with a decrease in Zcwinter in the same year. Similar discrepancies are highlighted in 1979 (Zcautumn above average and Zcwinter and Zcsummer below average) and 1991 (Zcwinter below average and all other sites above average).

Long-term variations in average annual Zc are shown in Fig. 4f. In general, the greatest wave-mixed depths and long-term variation (range = 1.34–2.66 m) was at site 5 and the lowest at site 1 (range = 0.61–0.97 m). Zc was low in the late 1970s and early 1980s, high in the late 1980s and early 1990s and moderate in recent years (up to 2005) in all sites.

An estimate of the proportion of the loch bed disturbed by wave mixing, i.e. percentage of the loch bed shallower than the annual average Zc across all sites is shown in Fig. 5.
https://static-content.springer.com/image/art%3A10.1007%2Fs10750-010-0188-9/MediaObjects/10750_2010_188_Fig5_HTML.gif
Fig. 5

Annual average (n = 12; error bars represent standard error of the mean) % surface area of the loch apparently mixed to the sediment surface between 1979 and 2005

Covariation between NAO and wave-mixed depth

Strong positive correlations were observed between long-term variation in at least one of the average seasonal Zc estimates and the corresponding seasonal average NAO index for each site (Table 3). Positive correlations were returned from comparisons of Zcannual and the spring NAO index (sites 2–6), Zcwinter and winter NAO index (sites 4–6 only) and between Zcsummer and the summer NAO index (sites 1 and 2 only). In general, the strongest correlations came from comparisons of Zcspring and the spring NAO index. The strength of the correlations varied across the sites, the strongest being in site 5.
Table 3

Results of the correlation analysis conducted between average seasonal values of wave-mixed depth and NAO for each year (n = 25) over the study period

Site and variables in correlation

r2

P

Site 1

 Zcspring versus spring NAO index

0.426

0.038

 Zcsummer versus spring NAO index

0.410

0.047

Site 2

 Zcannual versus spring NAO index

0.470

0.021

 Zcspring versus spring NAO index

0.562

0.004

 Zcsummer versus spring NAO index

0.404

0.050

Site 3

 Zcannual versus spring NAO index

0.546

0.006

 Zcspring versus spring NAO index

0.628

0.001

Site 4

 Zcannual versus annual NAO index

0.413

0.045

 Zcannual versus spring NAO index

0.588

0.003

 Zcwinter versus winter NAO index

0.519

0.009

 Zcspring versus spring NAO index

0.672

0.000

Site 5

 Zcannual versus annual NAO index

0.440

0.030

 Zcannual versus spring NAO index

0.594

0.002

 Zcwinter versus winter NAO index

0.575

0.003

 Zcspring versus spring NAO index

0.701

0.000

Site 6

 Zcannual versus spring NAO index

0.510

0.011

 Zcwinter versus winter NAO index

0.497

0.013

 Zcspring versus spring NAO index

0.619

0.001

Zcannual = average annual wave-mixed depth estimate; Zcwinter = average winter wave-mixed depth estimate; Zcspring = average spring wave-mixed depth estimate; Zcsummer = average summer wave-mixed depth estimate; Zcautumn = average autumn wave-mixed depth estimate

Discussion

Untangling spatial and temporal variability in Zc

Spatial variations

In Loch Leven, where the prevailing wind direction is generally from the west, the sites that were most affected by wind mixing, as indicated from modelled estimates of wave mixing, were those at the furthest distance directly downwind. With respect to the effective fetch estimates, sites 4 and 5 were most sensitive to variations in wind speed regardless of direction, whereas variations in all other sites were dependent on variations in wind direction as well as wind speed. For example, sites 1–3 were most sensitive to variations in easterly winds and site 6 to variations in southerly winds.

Spatial variation in sensitivity to changes in wind speed and direction was highlighted in the results of the Zc comparisons (Table 2; Figs. 3, 4). As expected Zc was greatest in site 5, the most exposed site, and least in site 1, the most sheltered site. This is in agreement with past surveys of the loch substrata in which sandy beaches were found on the westerly shore, but not on the east shore, indicating higher intensity sediment redistribution from the former (Calvert, 1974).

Average seasonal variations

Seasonal variations in Zc were observed in all sites (Fig. 3). As expected, this variation was highest in the most exposed sites (i.e. sites 4 and 5) with the deepest estimates of Zc corresponding to winter and the shallowest estimates of Zc corresponding with summer in all sites.

Long-term seasonal and annual variations in Zc

The highest anomalies (compared with 25-year averages) of Zc occurred in winter and spring in all sites, with the lowest in summer. Similarly, the range of anomalous Zc estimates across the sites was greatest in winter and spring and lowest in summer. This would suggest that the period, within which wave mixing is most important, is winter through spring in Loch Leven. The largest positive anomaly occurred in winter at site 5 (+1.5 m; 1990) with the largest negative anomaly occurring in the same site in spring (−1.1 m; 1984). This variation of ±2 m from the long-term mean Zc estimate may have implications for the ecology of Loch Leven. Any change in the wind conditions to favour deepening of summer wave-mixed layers may result in significant alterations to lake processes. However, variations in the NAO are unlikely to alter summer conditions to the same extent as winter and spring conditions (Jones et al., 2003).

Identifying the drivers of change

Both wind speed and the westerly frequency increased as the NAO increased, with the strongest correlations between all variables in winter and spring. In addition, correlations between NAO indices and wind speed were stronger than correlations between NAO indices and westerly frequency. This is in agreement with other studies, in which the effects of the NAO are reported to be strongest in winter and spring (George et al., 2000; Scheffer et al., 2001). The sensitivity of any lake to variations in wind speed and direction will be governed by lake and catchment morphometry. In Loch Leven, where the wind direction is predominantly westerly, the most significant NAO driver at a whole lake scale is likely to be wind speed.

Due to the location of the anemometer on the western shore of the loch, the measured wind speed is likely to be lower than the fetch-averaged wind, when winds are from the west and greater than the fetch-averaged wind, when winds are from the east. Results shown here will, therefore, tend to underestimate the links between wave-mixed depth and the NAO. This study does not account for spatial variability in wind speed and direction across the loch and, as the NAO is a global scale indicator of westerly winds, does not necessarily reflect changes in any other wind direction. In general, positive values of NAO are associated with stronger more westerly winds at Loch Leven. This is likely to lead to variations in the response at a site specific level where the strongest response will be apparent in the most westerly exposed sites.

Potential implications for shallow lake structure

The general seasonal and long-term trends reported here between the NAO and Zc are similar to those reported for other European lakes with positive NAO effects (high temperature and wind) during the early 1990s and negative NAO effects (low temperatures and winds) in the mid-1980s (Blenckner et al., 2007; Dokulil et al., 2006). It is possible that shallow lakes may be more sensitive to changes in wind speed and direction than deeper lakes and that this may, at least in large shallow lakes with high fetch, supersede the temperature effects previously reported in the literature.

Spatial variations in habitat quality may be sensitive to long-term variations in wind speed or direction. On average, the area of the loch bed that was shallower than Zc was between 18% (1984) and 35% (1990; Fig. 5). This is likely to result in variation in littoral habitat quantity and quality. These results are in agreement with a previous study conducted at Loch Leven, in which the maximum wave-mixed depth was estimated at about 35% (Beaufort wind force 8) of the total loch volume (Smith, 1974). In comparison, Nagid et al. (2001) estimated that up to 70% of the sediments in Lake Newman (27 km2, 1.5 m average depth) could be impacted by wave mixing.

Wind-induced wave mixing is an important driver of shallow lake structure and function. Sediment disturbance as a result of wave mixing alters water clarity as a result of increased sediment suspension (Kristensen et al., 1992), the recruitment of settled/benthic microbes (e.g. bacteria and algae; Carrick et al., 1993; Cotner et al., 2000; Schallenberg & Burns, 2004; Verspagen et al., 2004) to the water-column, the supply of dissolved oxygen to the lake bed (Belanger, 1981), entrainment of nutrients from the lake bed to the water-column (Søndergaard et al., 1992) and the regulation of benthic habitat quality and community structure and biomass (macrophytes: Doyle, 2001; Riis & Hawes, 2003; invertebrates: Brodersen, 1995; James et al., 1998). Variations in these processes may significantly alter the ecological status of a lake, and it is important to quantify such variation, and the drivers thereof, over long time scales (George et al., 2006).

The importance of the extent of the littoral zone for overall lake ecosystem functioning is well-known, and has important implications for water quality management (Smith et al., 1987; Carvalho & Anderson, 2001; Moss, 2008). The results presented highlight the effects of NAO-related drivers on the likely disturbance regime in the littoral habitats of Loch Leven. Variability in the magnitude of wave mixing is likely to significantly alter the ecological structure and ecosystem functions performed within the littoral zone, especially in shallow lakes, in which the littoral zone is the dominant habitat.

Acknowledgements

Thanks go to two anonymous reviewers, Prof. Brian Moss, Dr. Matthew O’Hare and Dr. Laurence Carvalho for comments leading to the improvement of this manuscript. Thanks also go to William Wilson and the staff at the Loch Leven Estates Ltd. for their continued and long-term support of limnological research at Loch Leven and for the provision of the meteorological data used in this study. We would also like to thank the Climate Research Unit, University of East Anglia for the open provision of NAO data. This study is a contribution to the European Union Sixth Framework Programme integrated Project Euro-limpacs (GOCE-CT-2003-505540) and was part funded by the UK Natural Environment Research Council.

Copyright information

© Springer Science+Business Media B.V. 2010