WINTER TEMPERATURE COVARIANCES IN THE MIDDLE AND THE LOWER TROPOSPHERE OVER EUROPE AND THE NORTH ATLANTIC OCEAN

Similar documents
SPATIAL AND TEMPORAL DISTRIBUTION OF AIR TEMPERATURE IN ΤΗΕ NORTHERN HEMISPHERE

1. INTRODUCTION. Copyright 2002 Royal Meteorological Society

ON THE RELATION BETWEEN SEA SURFACE AND LOWER TROPOSPHERE TEMPERATURE OVER THE NORTHERN HEMISPHERE

The North Atlantic Oscillation: Climatic Significance and Environmental Impact

A STUDY ON THE INTRA-ANNUAL VARIATION AND THE SPATIAL DISTRIBUTION OF PRECIPITATION AMOUNT AND DURATION OVER GREECE ON A 10 DAY BASIS

STUDY OF FUTURE CLIMATIC VARIATIONS OF A TELECONNECTION PATTERN AFFECTING EASTERN MEDITERRANEAN

THE 850 HPA RELATIVE VORTICITY CENTRES OF ACTION FOR WINTER PRECIPITATION IN THE GREEK AREA

The Arctic Oscillation (AO) or Northern Annular Mode (NAM)

Atmospheric circulation patterns associated with extreme precipitation amounts in Greece

Unusual North Atlantic temperature dipole during the winter of 2006/2007

Chapter outline. Reference 12/13/2016

Definition of Antarctic Oscillation Index

Nonlinear atmospheric teleconnections

June 1989 T. Nitta and S. Yamada 375. Recent Warming of Tropical Sea Surface Temperature and Its. Relationship to the Northern Hemisphere Circulation

10.5 ATMOSPHERIC AND OCEANIC VARIABILITY ASSOCIATED WITH GROWING SEASON DROUGHTS AND PLUVIALS ON THE CANADIAN PRAIRIES

Interannual Variability of the Wintertime Polar Vortex in the Northern Hemisphere Middle Stratosphere1

Extreme precipitation events in NW Greece

First-Order Draft Chapter 3 IPCC WG1 Fourth Assessment Report

Precipitation variability in the Peninsular Spain and its relationship with large scale oceanic and atmospheric variability

SHORT COMMUNICATION EXPLORING THE RELATIONSHIP BETWEEN THE NORTH ATLANTIC OSCILLATION AND RAINFALL PATTERNS IN BARBADOS

Interannual Variability of the South Atlantic High and rainfall in Southeastern South America during summer months

Eurasian Snow Cover Variability and Links with Stratosphere-Troposphere Coupling and Their Potential Use in Seasonal to Decadal Climate Predictions

NOTES AND CORRESPONDENCE. El Niño Southern Oscillation and North Atlantic Oscillation Control of Climate in Puerto Rico

Francina Dominguez*, Praveen Kumar Department of Civil and Environmental Engineering University of Illinois at Urbana-Champaign

Special blog on winter 2016/2017 retrospective can be found here -

THE VARIABILITY OF WINTERTIME PRECIPITATION IN THE NORTHERN COAST OF EGYPT AND ITS RELATIONSHIP WITH THE NORTH ATLANTIC OSCILLATION

June 1993 T. Nitta and J. Yoshimura 367. Trends and Interannual and Interdecadal Variations of. Global Land Surface Air Temperature

Principal Component Analysis of Sea Surface Temperature via Singular Value Decomposition

3. Climate Change. 3.1 Observations 3.2 Theory of Climate Change 3.3 Climate Change Prediction 3.4 The IPCC Process

Climatic study of the surface wind field and extreme winds over the Greek seas

The eastern Mediterranean teleconnection pattern: identification and definition

Global Climate Patterns and Their Impacts on North American Weather

Satellites, Weather and Climate Module??: Polar Vortex

Global Atmospheric Circulation

A STUDY ON EVAPORATION IN IOANNINA, NW GREECE

THE INFLUENCE OF CLIMATE TELECONNECTIONS ON WINTER TEMPERATURES IN WESTERN NEW YORK INTRODUCTION

The role of teleconnections in extreme (high and low) precipitation events: The case of the Mediterranean region

Semiblind Source Separation of Climate Data Detects El Niño as the Component with the Highest Interannual Variability

Analysis Links Pacific Decadal Variability to Drought and Streamflow in United States

Interannual Teleconnection between Ural-Siberian Blocking and the East Asian Winter Monsoon

Trends in Climate Teleconnections and Effects on the Midwest

APPENDIX B PHYSICAL BASELINE STUDY: NORTHEAST BAFFIN BAY 1

DOES EAST EURASIAN SNOW COVER TRIGGER THE NORTHERN ANNULAR MODE?

The Atmospheric Circulation

CHAPTER 1: INTRODUCTION

What kind of stratospheric sudden warming propagates to the troposphere?

Possible Roles of Atlantic Circulations on the Weakening Indian Monsoon Rainfall ENSO Relationship

LONG RANGE FORECASTING OF LOW RAINFALL

Inter-comparison of Historical Sea Surface Temperature Datasets

The high latitude blocking and low arctic oscillation values of December 2009

How far in advance can we forecast cold/heat spells?

SEASONAL ENVIRONMENTAL CONDITIONS RELATED TO HURRICANE ACTIVITY IN THE NORTHEAST PACIFIC BASIN

Impacts of Climate Change on Autumn North Atlantic Wave Climate

Lecture 8: Natural Climate Variability

PRMS WHITE PAPER 2014 NORTH ATLANTIC HURRICANE SEASON OUTLOOK. June RMS Event Response

Changes in Southern Hemisphere rainfall, circulation and weather systems

2013 ATLANTIC HURRICANE SEASON OUTLOOK. June RMS Cat Response

Impacts of Long-term Climate Cycles on Alberta. A Summary. by Suzan Lapp and Stefan Kienzle

El Niño / Southern Oscillation

CPTEC and NCEP Model Forecast Drift and South America during the Southern Hemisphere Summer

P3.6 THE INFLUENCE OF PNA AND NAO PATTERNS ON TEMPERATURE ANOMALIES IN THE MIDWEST DURING FOUR RECENT El NINO EVENTS: A STATISTICAL STUDY

A CLASSIFICATION OF AMBIENT CLIMATIC CONDITIONS DURING EXTREME SURGE EVENTS OFF WESTERN EUROPE

Global Temperature Report: December 2018

ENSO, AO, and climate in Japan. 15 November 2016 Yoshinori Oikawa, Tokyo Climate Center, Japan Meteorological Agency

J1.7 SOIL MOISTURE ATMOSPHERE INTERACTIONS DURING THE 2003 EUROPEAN SUMMER HEATWAVE

THE INFLUENCE OF EUROPEAN CLIMATE VARIABILITY MECHANISM ON AIR TEMPERATURES IN ROMANIA. Nicoleta Ionac 1, Monica Matei 2

Inter ENSO variability and its influence over the South American monsoon system

TROPICAL-EXTRATROPICAL INTERACTIONS

Climate Outlook for December 2015 May 2016

Climate Change 2007: The Physical Science Basis

Stratospheric Influences on MSU-Derived Tropospheric Temperature. Trends: A Direct Error Analysis

ON THE KEY REGIONS OF 500 hpa GEOPOTENTIAL HEIGHTS OVER NORTHERN HEMISPHERE IN WINTER

Monitoring and Prediction of Climate Extremes

An ENSO-Neutral Winter

WIND TRENDS IN THE HIGHLANDS AND ISLANDS OF SCOTLAND AND THEIR RELATION TO THE NORTH ATLANTIC OSCILLATION. European Way, Southampton, SO14 3ZH, UK

Anthropogenic warming of central England temperature

identify anomalous wintertime temperatures in the U.S.

Assessment of the Impact of El Niño-Southern Oscillation (ENSO) Events on Rainfall Amount in South-Western Nigeria

Verification of the Seasonal Forecast for the 2005/06 Winter

March was 3rd warmest month in satellite record

Our climate system is based on the location of hot and cold air mass regions and the atmospheric circulation created by trade winds and westerlies.

Characteristics of Storm Tracks in JMA s Seasonal Forecast Model

Diagnosis of systematic forecast errors dependent on flow pattern

Name: Climate Date: EI Niño Conditions

Observed Climate Variability and Change: Evidence and Issues Related to Uncertainty

The nonlinear association between ENSO and the Euro-Atlantic winter sea level pressure

2. DYNAMICS OF CLIMATIC AND GEOPHYSICAL INDICES

Climatic changes in the troposphere, stratosphere and lower mesosphere in

Problems with EOF (unrotated)

Relationship between atmospheric circulation indices and climate variability in Estonia

The Formation of Precipitation Anomaly Patterns during the Developing and Decaying Phases of ENSO

Atmospheric patterns for heavy rain events in the Balearic Islands

Decadal Changes in the Atmospheric Circulation and Associated Surface Climate Variations in the Northern Hemisphere Winter

Extremely cold weather events caused by arctic air mass and its synoptic situation in Finland from the year 1950 onwards

The feature of atmospheric circulation in the extremely warm winter 2006/2007

International Journal of Climatology - For peer review only

WINTER NIGHTTIME TEMPERATURE INVERSIONS AND THEIR RELATIONSHIP WITH THE SYNOPTIC-SCALE ATMOSPHERIC CIRCULATION

Proceedings, International Snow Science Workshop, Banff, 2014

Name: Date: Hour: Comparing the Effects of El Nino & La Nina on the Midwest (E4.2c)

Predictability and prediction of the North Atlantic Oscillation

Transcription:

INTERNATIONAL JOURNAL OF CLIMATOLOGY Int. J. Climatol. 21: 679 696 (2001) DOI: 10.1002/joc.651 WINTER TEMPERATURE COVARIANCES IN THE MIDDLE AND THE LOWER TROPOSPHERE OVER EUROPE AND THE NORTH ATLANTIC OCEAN C.J. LOLIS and A. BARTZOKAS* Laboratory of Meteorology, Department of Physics, Uni ersity of Ioannina, Ioannina, Greece Recei ed 10 May 2000 Re ised 18 January 2001 Accepted 26 January 2001 ABSTRACT In this work, the variability and covariability of winter temperatures in the middle and the lower troposphere are studied over Europe and the North Atlantic Ocean. Temperature variations of the middle and the lower troposphere are examined in terms of (a) 500 700 hpa and 700 l000 hpa thickness and (b) air temperature on the isobaric surfaces of 500 hpa and 700 hpa. At first, factor analysis (FA) defined areas with characteristic temperature variability in each layer (and on each isobaric surface) and then, canonical correlation analysis (CCA) revealed areas in both layers (and on both isobaric surfaces) presenting common temperature variations. A temperature see-saw between N Europe and W Asia was revealed for both layers and isobaric surfaces implying that temperature changes in these areas are vertically spread. Another well-defined area, appearing in both analyses, is the area of the Labrador Sea and S Greenland. This region is also teleconnected to other regions, though not very clearly in every height. These temperature patterns are mainly attributed to the Eurasian (EU) and North Atlantic Oscillation (NAO) circulation patterns, which are responsible for large air mass exchanges in the area, being vertically extended in the middle and the lower troposphere. Copyright 2001 Royal Meteorological Society. KEY WORDS: air temperature variability; canonical correlation analysis; Europe and North Atlantic Ocean; factor analysis; middle and lower troposphere 1. INTRODUCTION The study of air temperature variability near the Earth s surface presents certain difficulties because of the complex land sea distribution and the corresponding complicated atmospheric circulation. This is especially valid for the European area, which, being surrounded by regions of origin of different air masses such as the Sahara desert, the continental region of Siberia and the North Atlantic Ocean, is affected in various ways. Moreover, North Atlantic Oscillation (NAO), being one of the dominant global teleconnection patterns, plays a major role on the European climate (Lamp and Peppler, 1987; Hurrell and van Loon, 1997; Kapala et al., 1998). The variability of air temperature over the North Atlantic and Europe has been investigated by many researchers in the past (see e.g. van Loon and Rogers, 1978; Maheras, 1989; Weber, 1990; Bartzokas and Metaxas, 1993; Deser and Blackmon, 1993). Most of these papers deal either with surface or upper air temperature (thickness or temperature at specific pressure levels) without examining different levels or layers simultaneously. During the last decade, several publications have appeared using satellite data (microwave sounding units, MSU), which have a better coverage of the globe avoiding the poor monitoring of the remote regions (Spencer and Christy, 1990; Spencer et al., 1990; Christy and McNider, 1994). Furthermore, MSU data provide height-averaged values reducing undesirable effects of the surface such as urbanization. Since temperature time series * Correspondence to: Laboratory of Meteorology, Department of Physics, University of Ioannina, 45110 Ioannina, Greece; e-mail: abartzok@cc.uoi.gr Copyright 2001 Royal Meteorological Society

680 C.J. LOLIS AND A. BARTZOKAS generated by these data do not confirm to the global warming theory, a constructive debate on the explanation of this difference has taken place during the last decade (Hansen et al., 1995; Christy et al., 1998; Hurrell and Trenberth, 1998). In the present work, a contribution to the investigation of the vertical homogeneity in the troposphere, concerning temperature variability, is attempted by examining whether the temperature variations are vertically propagated. Specifically, it is investigated whether a temperature trend detected in the lower troposphere also appears in the middle troposphere and vice versa. If this is not the case for the whole area but it is valid for some sub-regions only, we define these sub-regions and we examine their temperature variability. This is achieved by comparing temperature variability in different layers and on different isobaric surfaces. In particular, the synchronous fluctuations of (a) 500 700 hpa and 700 1000 hpa thickness as an equivalent to the mean temperature of the layer and (b) 500 hpa and 700 hpa air temperatures are studied in order to reveal the similarities as well as the differences in temperature variability between the middle and the lower troposphere. 2. DATA AND METHODS USED The database for this work consists of mean monthly values at grid points spaced by 5 in latitude and 10 in longitude of the following parameters: (a) 500 700 hpa and (b) 700 1000 hpa thickness for the period 1949 1992 and (c) 500 hpa and (d) 700 hpa air temperatures for the period 1963 1992. The 500 700 hpa and 700 1000 hpa thickness values were estimated by using the 500 hpa and 700 hpa heights and the 1000 500 hpa thickness data. Our study area is defined by the 70 W and 70 E meridians and the 25 and 80 N parallels. It comprises Europe, the North Atlantic Ocean, N Africa, W Asia and the northeastern coasts of North America. We used 154 out of the 156 grid points, omitting two grid points in SW Asia because of too many missing data (Figure 1). We then constructed the time series of the above four parameters for January, a typical winter month. A few missing values were estimated from the neighbouring grid points by using linear regression analysis. The statistical methods used are factor analysis (FA) and canonical correlation analysis (CCA). The methods are now briefly described. Figure 1. The 154 grid points used

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 681 (a) FA describes a set of p correlated variables X 1, X 2,...,X p in terms of a smaller number of uncorrelated indices or factors and hence elucidates the relationship between the original p variables. Thus, each of the p initial variables can be expressed as a linear function of m (m p) uncorrelated factors, i.e. X i = i1 F 1 + i2 F 2 + + im F m, where F 1, F 2,...,F m are the factors and i1, i2,..., im are the factor loadings which express the correlation between the factors and the initial variables. The sum m j=1 2 ij is referred to as the communality of X i and it is the part of its variance that is related to the m factors. The values of each factor are called factor scores and they are usually presented in standardized form, having zero mean and unit variance (Jolliffe, 1986; Manly, 1986). The number m of the retained factors has to be decided, by using various rules (Horel, 1981; Overland and Preisendorfer, 1982; Jolliffe, 1986; Rogers, 1990) and considering the physical interpretation of the results (Bartzokas and Metaxas, 1993). A widely used process is the rotation of the axes, which creates new factors with different variances, but it keeps the cumulative variance of the m factors unaffected. Rotation, by maximizing some factor loadings and minimizing some others, succeeds in a better separation among the initial variables and thus in a better interpretation of the factors. There are various types of rotation. Varimax rotation is generally accepted as the most accurate orthogonal rotation. It maximizes the sum of the variances of the square factor loadings, keeping the factors uncorrelated (Richman, 1986). FA is a method similar to principal component analysis (PCA), which has been also used extensively in climatology. Both methods are concerned with the reduction of the initial p variables to a smaller number, m, of new variables but they are not identical. As it is shown above, FA estimates how the variables are related to the factors and how much variance in each variable separately is explained by them. On the other hand, PCA finds m linear functions accounting for maximal amounts of variance in the original variables, expressing the m new variables in terms of all the initial ones, i.e. PC j = j1 X 1 + j2 X 2 + + jp X p, where PC 1,PC 2,...,PC m are the principal components and j1, j2,..., jp are the component loadings (Manly, 1986; Jolliffe, 1993). (b) CCA investigates the relation between two sets of variables X 1, X 2,...,X p and Y 1, Y 2,...,Y q.at first, considering the linear combinations W 1 = 11 X 1 + 12 X 2 + + 1p X p, V 1 =b 11 Y 1 +b 12 Y 2 + +b 1q Y q. CCA calculates 11, 12,..., 1p and b 11, b 12,...,b 1q such that the correlation coefficient between W 1 and V 1, C 1 =cor(w 1, V 1 ), is at a maximum. W 1 and V 1 are called canonical variates and C 1 is called canonical correlation. In the next step, another set of canonical variates is identified: W 2 = 21 X 1 + 22 X 2 + + 2p X p, V 2 =b 21 Y 1 +b 22 Y 2 + +b 2q Y q such that C 2 =cor(w 2, V 2 )=maximum and W 2 and V 2 are uncorrelated with W 1 and V 1. That is, the two sets of canonical variates are uncorrelated. This procedure is continued up to the mth set of canonical variates, where m=min(p, q). Thus, m sets of canonical variates (W 1, V 1 ), (W 2, V 2 ),..., (W m, V m ) are created in a way that: (i) the corresponding canonical correlations C 1, C 2,...,C m are at a maximum and (ii) cor(v j, V k )=cor(w j, W k )=cor(w j, V k )=0, j k (Anderson, 1984; Sharma, 1995). In practice, the number of the canonical pairs used never equals the total number of pairs. Similarly to FA, only the statistically significant pairs are used, indicated by various statistical criteria (Sharma, 1995). We note that while in FA an increase of the number of retained factors may affect the results, in CCA, a selection of more canonical pairs does not affect the previous ones as the method proceeds with steps. In this work, we applied these two methods in order to succeed in interrelating the temperature variability of the two layers (and the two isobaric surfaces). In other words, we looked for cases of vertical homogeneity of the troposphere or teleconnective interaction between the two layers, as it concerns their temperature variability. At first, FA (S-mode) (Richman, 1986) was applied separately for each layer (and each isobaric surface), in order to reduce the number of the initial variables (grid points) and define sub-areas within each layer (and each isobaric surface) consisting of grid points with a common temperature variability. Many researchers have followed this process, i.e. the reduction of the dimensionality of the initial data sets before the application of CCA (Zorita et al., 1992; Knappenberger and Michaels, 1993; Corte-Real et al., 1995; Diaz et al., 1998; Xoplaki et al., 2000). This is considered

682 C.J. LOLIS AND A. BARTZOKAS necessary to avoid quasi-degeneracy of the autocovariance matrices of the data sets. Furthermore, this process filters the data by eliminating the noise (Barnett and Preisendorfer, 1987; Bretherton et al., 1992). CCA was then applied on the two sets of the resultant factor scores time series. The basic advantage of the above process is the fact that it takes into account both sets simultaneously, investigating the best interrelations between them. This would have not been the case if individual correlation coefficients had been calculated among the scores time series. Finally, in order to achieve an interpretation of the resultant canonical variates, each of them was correlated with all the original time series of the corresponding layer (and isobaric surface) and then, for each canonical pair, the correlation coefficients were plotted on two maps. By comparing the two maps of each canonical pair, we are able to elucidate the similarities as well as the differences in temperature variability between the two layers (and the two isobaric surfaces). 3. RESULTS AND DISCUSSION 3.1. 500 700 hpa and 700 1000 hpa thickness The application of FA on 500 700 hpa thickness and on 700 1000 hpa thickness, revealed that 11 and 12 factors respectively describe well the variability in the study area, explaining about 87% of the total variance in each case. In both layers, the geographical sub-regions formed cover the area of interest satisfactorily, since the grid points not classified in any factor are very few (Figure 2). For the upper layer, 500 700 hpa (Figure 2(a)), the 11 factors correspond to well defined areas, i.e. at least four grid points of each factor present loadings higher than 0.60, while a teleconnection appears between the central North Atlantic Ocean and NW Asia (factor 3). The main features of the long term temperature changes in these sub-regions, as they are revealed by the factor scores time series (Figure 3), will be described briefly in the following, since the main purpose of this paper is the interrelation between the two sets of variables. In the Labrador Sea and S Greenland (factor 2), a cooling has appeared since 1981. The teleconnected areas of the North Atlantic and NW Asia (factor 3) present a temperature minimum around 1970, a maximum in the early 1980s and a cooling since then. The central Atlantic (factor 4) presents a relatively warm period (1949 1972) followed by a cool period (1973 1992). A warming trend appears in NW Europe (factor 5) since 1984, while in NE Europe (factor 6), there is a cooling from 1973 towards the minimum of 1987. The east Mediterranean and NE Africa area (factor 7) presents a general positive trend since 1967. Finally, in the Arctic Ocean (factor 11), the most interesting features are the warm decade of 1950s and the pronounced minimum around 1963. For the lower layer 700 l000 hpa (Figure 2(b)) the groupings are in general similar with those of the upper layer, but there are some differences: The teleconnection between the north Atlantic Ocean and NW Asia does not appear here, while a weak teleconnection (factor 4) emerges between the middle Atlantic Ocean and W Arabia. Furthermore, a new weak grouping of two grid points only (factor 12) arises in NW Africa. The main features of the corresponding long term temperature changes are the following (Figure 4). In SW Asia (factor 1), there is a broad maximum during the early 1960s, which appeared weaker in the upper layer 500 700 hpa. The Labrador Sea (factor 3) presents a temperature minimum during the 1970s, while the opposite (maximum) is valid for central Atlantic (factor 4) during the same period. The cooling detected for 500 700 hpa layer over the Labrador Sea during the last decade, is also detected for the 700 l000 hpa layer and this finding is in agreement with the results of Shabbar et al. (1997) for surface temperature variability over the same region. It seems also that in the continental region of N Greenland, the 700 1000 hpa layer (factor 6) is warmer during the last years in relation to the long term mean, while this does not appear for the 500 700 hpa layer (factor 9). This means that, a recent warming is detected in this region, being confined in the lowest tropospheric layer, close to the Earth s surface. The same is valid to a lesser degree for the continental area of SW Asia (factor 1). East Mediterranean and NE Africa (factor 9), present a broad maximum around 1970, followed by a cooling (Bartzokas and Metaxas, 1991; Maheras and Kutiel, 1999).

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 683 Figure 2. Sub-regions formed by FA for (a) 500 700 hpa thickness and (b) 700 1000 hpa thickness (the 0.60 loadings isopleths have been drawn)

684 C.J. LOLIS AND A. BARTZOKAS Figure 3. Factor scores time series for 500 700 hpa thickness (standardized), smoothed (bold line) using 5-year moving averages with binomial coefficient weights

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 685 Figure 4. As in Figure 3 but for 700 1000 hpa thickness

686 C.J. LOLIS AND A. BARTZOKAS We then applied CCA to the two sets of factor scores in order to search for the possible existence of connections between the 11 uncorrelated time series of the upper layer and the 12 uncorrelated time series of the lowest layer. In other words, we looked for cases of temperature covariability between the two layers. CCA resulted into 8 (out of 11) canonical pairs with statistically significant canonical correlation at 95% confidence level. In order to obtain a physical interpretation of the results, we considered the correlation coefficient of the canonical variates with the original thickness time series at each point. From the eight canonical variate pairs, five were found not to be correlated with any original time series with a correlation coefficient at least 0.6 and therefore were rejected. Hence in the following, the first three canonical pairs will be presented. The first canonical pair (W 1, V 1 ) explains 37% of the total variance and corresponds in both layers to a see-saw between N Europe and W Asia (Figure 5(a) and (b)). This finding is in agreement with a similar see-saw detected, at a lesser degree, for 1000 500 hpa thickness by Bartzokas and Metaxas (1993). In these two areas, the temperature variability of the lower layer is similar to that of the upper layer, which means that this see-saw is vertically extended. It can be attributed to Eurasian (EU) and Pacific North American (PNA) patterns (Wallace and Gutzler, 1981; Barnstom and Livezey, 1987) since both of them are related to a temperature see-saw between the above areas (Gutzler et al., 1988; Hurrell and van Loon, 1997). The two canonical variates W 1 and V 1 are almost identical because of the very high canonical correlation (greater than 0.98). According to the scores of the canonical variates, north Europe is characterized by a temperature maximum during the early 1970s, a minimum during the middle 1980s and a warming trend during the last years. The opposite is valid for west Asia, where minimum values dominated in early 1970s, maximum values in the middle 1980s and a decreasing trend is recorded during the last years. The second canonical pair (W 2, V 2 ) (24% of the total variance) with a correlation coefficient of 0.98, corresponds to a teleconnection between the Labrador Sea and N Africa mainly in the 700 1000 hpa layer (Figure 6(a) and (b)). This teleconnection appears very weak in the upper layer 500 700 hpa, where only the Labrador Sea area presents high (above 0.6) correlation coefficients. Thus, according to the second canonical pair there is a well-established vertical homogeneity of the troposphere over the Labrador Sea, while this is not clear for N Africa. According to W 2, V 2 scores, the Labrador Sea presents a temperature maximum about 1980 and a generally decreasing trend since then, while the same is valid mainly for 1000 700 hpa layer over N Africa. There is a similarity between the scores time series and the temporal variation of the Baffin Island West Atlantic (BWA) index. This index reflects the variability of the western structure of the NAO and has been found to explain temperature variability better in the above area than the structure characterized by the NAO index (Shabbar et al., 1997). The negative trend of the BWA index and the corresponding positive trend of the NAO index since about 1980 are in accordance with the W 2, V 2 time series and the temperature decrease over the Labrador Sea area (World Meteorological Organization, 1998). The third canonical pair (W 3, V 3 ) (13% of total variance) with a correlation coefficient of 0.96, corresponds to a vertically extended see-saw between the polar area and E Europe (Figure 7(a) and (b)). A similar temperature see-saw has been detected for 850 hpa temperature field, but no statistically significant relation with any of the already known teleconnection patterns was found (Gutzler et al., 1988). The main feature of the two time series of the scores is the temperature minimum experienced in the mid 1980s in E Europe, corresponding to a maximum over the Pole. 3.2. 500 hpa and 700 hpa temperature The same procedure was followed for 500 and 700 hpa temperatures. FA resulted in nine and eight factors respectively, explaining 86% and 85% of the total variance. The sub-regions formed are presented in Figure 8. The main features of these maps are three teleconnections which appear on both surfaces. The polar region is teleconnected with W Asia (see-saw), the Labrador Sea is teleconnected with a small area in the west Atlantic (see-saw) and NW Asia is teleconnected with an area around the central Atlantic. A fourth teleconnection between NE and low latitudes Atlantic appears confined on the 700 hpa surface

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 687 Figure 5. The first canonical pattern and the time series of the canonical variates. (a) for 500 700 hpa thickness and (b) for 700 1000 hpa thickness. Isopleths of correlation between the canonical variates and the thickness time series have been drawn. The canonical variates time series have been smoothed (bold line) using 5-year moving averages with binomial coefficient weights

688 C.J. LOLIS AND A. BARTZOKAS Figure 6. As in Figure 5 but for the second canonical pattern

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 689 Figure 7. As in Figure 5 but for the third canonical pattern

690 C.J. LOLIS AND A. BARTZOKAS Figure 8. As in Figure 2 but for (a) 500 hpa temperature and (b) 700 hpa temperature

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 691 only. The long term temperature variability in these areas presented by factor scores time series is not shown, as this is outside the scope of this work. By applying CCA to the two sets of factor scores time series and following the previous analysis process, we found out that three of the resultant canonical pairs are physically interpretable with canonical correlations exceeding 0.98. The first canonical pair (W 1, V 1 ) (42% of total variance) corresponds to a see-saw between NW Europe and a broad area with two centres located over N Africa and E Black Sea (Figure 9(a) and (b)). This see-saw has a clear vertical extent and exists on both isobaric surfaces. Its existence can be attributed to blocking effects. When dynamic high pressure systems prevail over central Europe positive temperature anomalies arise over NW Europe and negative temperature anomalies arise over N Africa and SW Asia (Metaxas et al., 1993). According to the corresponding time series of the canonical variates, there is a temperature increase in NW Europe during the last 8 years of our study period. The opposite is valid for the double centre area of N Africa and SW Asia, where a cooling has appeared since about 1984. This trend is in agreement with the finding of Beniston et al. (1994) that these blockings have been intensified during the last years. The second pair (W 2, V 2 ) (29% of total variance) corresponds to a see-saw phenomenon between N Europe and W Asia (Figure 10(a) and (b)). It looks similar to the first canonical pair of the previous analysis, but there is a slight northward displacement of the two high correlation centres. According to the W 2 and V 2 time series, there is a temperature maximum in N Europe during the early 1970s, a minimum during the early 1980s and an increasing trend since then. The opposite is valid for the teleconnected area of W Asia. The third pair (W 3, V 3 ) (14% of total variance) corresponds to a vertically extended temperature see-saw between an area comprising the Labrador Sea and Greenland and an area located over NW Asia (Figure 11(a) and (b)). This see-saw is slightly displaced relative to the well-known temperature see-saw between Greenland and N Europe, which has been attributed to inter-annual variations in the intensity of the Icelandic low (van Loon and Rogers, 1978). 4. COMPARISONS The results of the two analyses for thickness and isobaric surfaces temperature present impressive similarities but also certain differences. As mentioned above, the most striking similarity is the one between the patterns of the first canonical pair of thickness and the second canonical pair of isobaric surfaces. This similarity is further enhanced if the fluctuations of the time series of their canonical variates are compared. The maximum in the early 1970s and the minimum in the mid 1980s are common to both curves (Figures 5 and 10), as they ought to be. In order to present a quantitative proof of the relationship between these two canonical pairs and to search for relationships between the others, we calculated the correlation coefficients between their corresponding time series for the common period 1963 1992. As the correlation coefficients between the canonical variates of the same pair (W i, V i ) are very high (above 0.96), we calculated only the correlation coefficients between the W-time series (500 700 hpa thickness and 500 hpa temperature). As expected, it was found that W 1 of 500 700 hpa thickness and W 2 of 500 hpa temperature are well correlated (r=0.83, statistically significant at p=0.01), indicating that the variables of the two pairs (W 1 V 1 and W 2 V 2 respectively) co-vary to a high degree. The correlations between 500 hpa temperature W 2 and 500 700 hpa thickness W 2 and W 3 were found to be very low (0.08 and 0.35 respectively), not being statistically significant at the 0.01 level. Then, the correlation coefficients between W 3 of 500 hpa temperature and W 1, W 2, W 3 of 500 700 hpa thickness were estimated and it was found that W 3 of temperature at 500 hpa corresponds to W 2 of 500 700 hpa thickness (r=0.81, statistically significant at p=0.01). In Figures 6 and 11, the teleconnections found may at first not lead to a conclusion of resemblance of the two patterns. However, a careful inspection reveals that from NW to SE, three belts of alternate signals are encountered. Hence, although in some cases the teleconnections and the see-saws are weak, they do exist in all patterns. But even if these teleconnections are not considered as the

692 C.J. LOLIS AND A. BARTZOKAS Figure 9. As in Figure 5 but for (a) 500 hpa temperature and (b) 700 hpa temperature

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 693 Figure 10. As in Figure 5 but for the second canonical pattern and for (a) 500 hpa temperature and (b) 700 hpa temperature

694 C.J. LOLIS AND A. BARTZOKAS Figure 11. As in Figure 5 but for the third canonical pattern and for (a) 500 hpa temperature and (b) 700 hpa temperature

TROPOSPHERE WINTER TEMPERATURE COVARIANCES 695 main feature on every map, the large area over Greenland and the Labrador Sea is definitely an area where common temperature fluctuations are clearly marked in every layer and every height of the middle and lower troposphere. Thus, in this area (Figures 6 and 11), temperature has presented maxima in mid 1960s, late 1970s and mid 1980s at every height. Finally, the two variates left, W 1 of 500 hpa temperature and W 3 of 500 700hPa thickness do not seem to correspond to each other (r=0.12). The different data sets used may be responsible for this disagreement. We remind the reader that: (a) we have used different parameters (thickness and isobaric surface temperature); (b) the atmospheric height of the two data sets is not the same, since for the second analysis, the upper boundaries of the first analysis layers were used, i.e. the second analysis is referred to greater atmospheric heights; and (c) we have used different study periods (1949 1992 for thickness and 1963 1992 for isobaric surface temperatures). 5. CONCLUSIONS The winter temperature variability and covariability of the middle and the lower troposphere over Europe and the North Atlantic Ocean was studied by using FA and CCA. By applying FA, we succeeded in describing the temperature variability separately in each of the layers 500 700 hpa and 700 1000 hpa and on each of the isobaric surfaces 500 hpa and 700 hpa. By applying CCA, we managed to detect the main areas of temperature covariance in the lower and the middle troposphere. Two main patterns were found to be common on both analyses: a strong temperature see-saw between N Europe and W Asia and a weaker one with its main pole over the Labrador Sea and S Greenland. Other patterns, not appearing on either thickness or isobaric surface temperature fields, were also statistically significant, but they may be affected by the different data sets used in the two analyses. The results confirm the influence of NAO and EU in the formation of temperature characteristics of their neighbouring areas. REFERENCES Anderson TW. 1984. An Introduction to Multi ariate Statistical Analysis. Wiley: New York. Barnett T, Preisendorfer R. 1987. Origins and levels of monthly and seasonal forecasts skill for the United States surface air temperatures determined by canonical correlation analysis. Monthly Weather Re iew 115: 1825 1850. Barnstom AG, Livezey RE. 1987. Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Monthly Weather Re iew 115: 1083 1126. Bartzokas A, Metaxas DA. 1991. Climatic fluctuation of temperature and air circulation in the Mediterranean. In Proceedings of the European School of Climatology and Natural Hazards Course, Arles/Rhone, France, Duplessy JC, Pons A, Famtechi R (eds). Commission of the European Communities: Brussels; 279 297. Bartzokas A, Metaxas DA. 1993. Covariability and climatic changes of the lower-troposphere temperatures over the Northern Hemisphere. Il Nuo o Cimento 16: 359 373. Beniston M, Rebetez M, Giorgi F, Merinucci MR. 1994. An analysis of regional climate change in Switzerland. Theoretical and Applied Climatology 49: 135 159. Bretherton CS, Smith S, Wallace J. 1992. An intercomparison of methods for finding coupled patterns in climate data. Journal of Climate 5: 541 560. Christy JR, McNider RT. 1994. Satellite greenhouse signal. Nature 367: 325. Christy JR, Spencer RW, Lobl ES. 1998. Analysis of the merging procedure for the MSU daily temperature time series. Journal of Climate 11: 2016 2041. Corte-Real J, Zhang X, Wang X. 1995. Large-scale circulation regimes and surface climatic anomalies over the Mediterranean. International Journal of Climatology 15: 1135 1150. Deser C, Blackmon ML. 1993. Surface climate variations over the North Atlantic Ocean during winter: 1900 1989. Journal of Climate 6: 1743 1753. Diaz AF, Studzinski CD, Mechoso CR. 1998. Relationships between precipitation anomalies in Uruguay and southern Brazil and sea surface temperature in the Pacific and Atlantic Oceans. Journal of Climate 11: 251 271. Gutzler DS, Rosen RD, Saistein DA, Peixoto JP. 1988. Patterns of interannual variability in the Northern Hemisphere wintertime 850 mb temperature field. Journal of Climate 1: 949 964. Hansen J, Wilson H, Sato M, Ruedy R, Shah K, Hansen E. 1995. Satellite and surface temperature data at odds? Climate Change 30: 103 117. Horel JD. 1981. A rotated principal component analysis of the interannual variability of the northern hemisphere 500 mpa height field. Monthly Weather Re iew 109: 2080 2092. Hurrell JW, Trenberth KE. 1998. Difficulties in obtaining reliable temperature trends: reconciling the surface and satellite microwave sounding unit records. Journal of Climate 11: 945 967.

696 C.J. LOLIS AND A. BARTZOKAS Hurrell JW, van Loon H. 1997. Decadal variations in climate associated with the North Atlantic Oscillation. Climate Change 36: 301 326. Jolliffe IT. 1986. Principal Component Analysis. Springer-Verlag: New York. Jolliffe IT. 1993. Principal Component Analysis: A beginners guide II. Pitfalls, myths and extensions. Weather 48: 246 253. Kapala A, Mächel H, Flohn H. 1998. Behaviour of the centres of action above the Atlantic since 1881. Part II: associations with regional climate anomalies. International Journal of Climatology 18: 23 36. Knappenberger PC, Michaels PJ. 1993. Cyclone tracks and wintertime climate in the mid-atlantic region of the USA. International Journal of Climatology 13: 1 24. Lamp PJ, Peppler RA. 1987. North Atlantic oscillation: concept and an application. Bulletin of the American Meteorological Society 68: 1218 1225. Maheras P. 1989. Principal component analysis of western Mediterranean air temperature variations 1866 1985. Theoretical and Applied Climatology 39: 137 145. Maheras P, Kutiel H. 1999. Spatial and temporal variations in the temperature regime in the Mediterranean and their relationship with circulation during the last century. International Journal of Climatology 19: 745 764. Manly BFJ. 1986. Multi ariate Statistical Methods: A Primer. Chapman & Hall: London. Metaxas DA, Bartzokas A, Repapis CC, Dalezios NR. 1993. Atmospheric circulation anomalies in dry and wet winters in Greece. Meteorologische Zeitschrift 2: 127 131. Overland JE, Preisendorfer RW. 1982. A significance test for principal components applied to a cyclone climatology. Monthly Weather Re iew 110: 1 4. Richman MB. 1986. Rotation of principal components. Journal of Climatology 6: 293 335. Rogers JC. 1990. Patterns of low-frequency monthly sea level pressure variability (1899 1986) and associated wave cyclone frequencies. Journal of Climate 3: 1364 1379. Shabbar A, Higuchi K, Skinner W, Knox JL. 1997. The association between the BWA index and winter surface temperature variability over Eastern Canada and West Greenland. International Journal of Climatology 17: 1195 1210. Sharma S. 1995. Applied Multi ariate Techniques. Wiley: New York. Spencer RW, Christy JR. 1990. Precise monitoring of global temperature trends from satellites. Science 247: 1559 1562. Spencer RW, Christy JR, Grody NC. 1990. Global atmospheric temperature monitoring with satellite microwave measurements: method and results 1979 84. Journal of Climate 3: 1111 1128. van Loon H, Rogers JC. 1978. The seesaw in winter temperatures between Greenland and Northern Europe. Part I: general description. Monthly Weather Re iew 106: 296 310. Wallace JM, Gutzler DS. 1981. Teleconnections in the Geopotential Height Field during the Northern Hemisphere winter. Monthly Weather Re iew 109: 784 812. Weber GR. 1990. Tropospheric temperature anomalies in the northern hemisphere 1977 1986. International Journal of Climatology 10: 3 19. World Meteorological Organization. 1998. The Global Climate System Review, December 1993 May 1996. In World Climate and Monitoring Programme, Nichols JM (ed.). WMO-No. 856. Xoplaki E, Luterbacher J, Burkard R, Patrikas I, Maheras P. 2000. Connection between the large scale 500 hpa geopotential height fields and rainfall over Greece during winter time. Climate Research 14: 129 146. Zorita E, Viacheslav K, Von Storch H. 1992. The atmospheric circulation and sea surface temperature in the North Atlantic area in winter: their interaction and relevance for Iberian precipitation. Journal of Climate 5: 1097 1108.