CLIMATE RESEARCH Clim Res
Vol. 14: 7–24, 2000
Published January 24
The influence of climate on air and precipitation chemistry over Europe and downscaling applications to future acidic deposition Julie M. Jones1,*, Trevor D. Davies 2 2
1 Institute of Hydrophysics, GKSS, Max-Planck-Strasse, 21502 Geesthacht, Germany School of Environmental Sciences, University of East Anglia, Norwich NR4 7TJ, United Kingdom
ABSTRACT: Atmospheric circulations, and related climate conditions, exert a strong influence on the transport and ultimate deposition of atmospheric pollutants. One potential outcome of climate change is alteration of atmospheric circulation patterns. The first stage in the development of a downscaling methodology to assess the influence of these possible future atmospheric circulation changes on transport and deposition of atmospheric pollutants over Europe is described. Firstly, the main modes of regional-scale circulation over a domain covering the North Atlantic and Europe are determined using principal components analysis of sea level pressure patterns. To determine whether the principal components represent circulation types found in reality, and show realistic relationships with surface temperature and precipitation amount, they are compared to the Lamb Weather Types, Central England Temperature, and England and Wales Rainfall. There are statistically significant relationships between some of the principal components and aqueous and ambient pollutant concentrations at 5 selected stations from the European Monitoring and Evaluation Programme monitoring network. Although, for any one combination of principal component and pollutant variable, the level of explanation can be quite small, notwithstanding some levels of variance explained for individual combinations being up to 69% (45% for the winter season, described in detail here). Because the pollution climate at any one location is a function of many combinations of circulation/pollution relationships, despite the relatively small magnitude of variance explained for individual combinations in this study, the results confirm the utility of atmospheric circulation principal components in deriving downscaling relationships for surface pollution behaviour (via eventual multiple regressions, not reported in this paper), as well as giving further insight into the climatic influences on air and precipitation chemistry over Europe. KEY WORDS: Acidic deposition · Air pollution · Downscaling · Synoptic climatology · Principal components analysis · Climate change
1. INTRODUCTION Acidic deposition is the final stage in a chain of processes linking emission sources and atmospheric transport and transformation. Meteorology exerts a controlling influence in this chain. A number of synoptic climatological approaches have been used in attempts to embrace a range of the pertinent meteorological processes. Examples include the use of Lamb *E-mail:
[email protected] © Inter-Research 2000
Weather Types (LWT) (Davies et al. 1986, 1991, O’Hare & Wilby 1995), statistical clustering of meteorological variables (Sanchez et al. 1990, McGregor & Bamzelis 1995), and the grouping of airmass trajectories (Dorling et al. 1992, Moody et al. 1995). Such studies have demonstrated clear links between atmospheric circulation and the transport and deposition of acidifying compounds. An important implication is that climate change, with possible circulation changes, may lead to changes in the transport and deposition patterns of acidic species (Davies et al. 1986).
8
Clim Res 14: 7–24, 2000
The known links between atmospheric circulation patterns and surface climate variables have been used to produce possible future climate scenarios, on the regional scale, by the synoptic climatological classification of atmospheric circulations projected by general circulation models (GCMs) (e.g. Hewitson & Crane 1992, Goodess & Palutikof 1998). Such methods are now widely used (Hewitson & Crane 1996, Wilby et al. 1998). Extension of such a ‘downscaling’ methodology to the issue of possible future acidic deposition is, therefore, quite logical, given the clear links between present acidic deposition patterns and atmospheric circulation. The objective of this work, as the first step in the production of possible future scenarios of acidic deposition, is to establish whether or not there are distinct relationships between pollutant concentrations and depositions and atmospheric circulation, using the downscaling techniques which have been applied to surface climate variables. If successful, this would then allow the first application of a downscaling technique to future acidic depositions, using the projections of future circulations from GCMs. This would complement the previous, different, approaches to the assessment of the possible influence of future climate change on acid deposition over Europe. Pitovranov (1988) used an analogue method to suggest changes in Grosswetterlagen frequency, and the consequent effects on pollutant transport. Smith (1992) considered the combined effects of possible future emission changes, and possible changes in precipitation regimes, and Alcamo et al. (1995) linked integrated assessment models of climate change and acid deposition. This study uses principal components analysis (PCA) to determine the main modes of spatial variation in atmospheric circulation over the North Atlantic and Europe. Comparisons are then made with an established climatological classification (LWT), and surface observation (Central England Temperature [CET] and England and Wales Rainfall [EWR]), to confirm the physical sense of the principal components (PCs) of the circulation. Relationships are then established between the PCs and observed pollutant concentrations and depositions at 5 stations in the European Monitoring and Evaluation Programme (EMEP) monitoring network (Jones 1997, Jones & Davies 1998, unpubl.). This is an important step since it will determine if the links between the circulation PCs and the pollution are strong enough for application to future scenario studies. Moreover, the physical reality of the relationships may be assessed through comparison with circulation/pollution relationships over the same study region determined by a different approach (e.g. Dorling et al. 1992). The length of the data series used for this work was limited by the availability of the EMEP data. At the
time analysis was begun, only about 10 yr of continuous data with acceptable levels of missing data for the variables considered were available.
2. PRINCIPAL COMPONENTS ANALYSIS OF ATMOSPHERIC CIRCULATION The United Kingdom Meteorological Office (UKMO) northern hemisphere 5° latitude by 10° longitude gridded mean sea level pressure dataset was used (Jones 1987). Analysis was undertaken on a daily basis for the 4 standard meteorological seasons, for the region 40° E–40° W, 30–80° N, and for the period 1982 to 1991. This period corresponded to the time of greatest EMEP data availability for this study. Orthogonally rotated PCA was applied to the data, using the varimax rotation (von Storch & Zwiers 1999). Opinions differ on whether or not to use rotation (Richman 1986, Jolliffe 1987). Because interpretation of patterns is regarded as more straightforward with rotation (Yarnal 1993), it was used in this study. Components with eigenvalues greater than 1 were retained for rotation (i.e. only components which explained more variance than the original variables), a criterion used by Horel (1981) and Villalba et al. (1997). This led to between 14 and 16 components in each season being retained for rotation. The retention of a relatively large number was desirable, since a characteristic of acidic deposition is that it may be episodic (i.e. a large proportion of annual deposition at a site may occur under relatively infrequent meteorological conditions; Smith & Hunt 1978). In extreme cases, in some locations, up to one-third of the total annual deposition of acidic species may occur in 1 day (Davies et al. 1992). Thus, although much deposition may be expected under frequently occurring conditions, inclusion of the more uncommon synoptic types increases the probability that a greater proportion of deposition may be accounted for by the components. Thus components which explain only a few percent of the total variance have been retained. Jolliffe (1982) suggests that PCs with smaller eigenvalues can be important in PC regression. The rationale that PCA identifies the most frequent modes of variance in a dataset naturally means that perhaps very infrequent conditions important to deposition at the stations may not be captured by the classification. There has been some discussion as to which PC retention rules to use. North et al. (1982) suggest that if the sampling error of a particular eigenvalue is equal or larger to the spacing between it and the neighbouring eigenvalue then the sampling error of the empirical orthogonal function (EOF) is comparable to the size of the neighbouring EOF. Von Storch (1995) suggests that this approach is more objective from a
Jones & Davies: Climatic influence on air and precipitation chemistry
statistical point of view. Kidson (1988) uses a combination of this method and the ‘scree test’. The latter involves plotting the eigenvalue against PC number. A break in slope is then identified, after which the eigenvalues of the following components, which represent uncorrelated noise, should decay exponentially. This methodology has the advantage of being easy to understand, and is relatively conservative. It involves however a subjective judgement about the location of the break in the plotted curve. The results in the following sections, of relationships between the components and CET, EWR, the LWT and with the EMEP station data, support the physical reality of the components retained. The component loadings are shown in Figs. 1 (winter) & 2 (summer). The maps show the contribution of the variable at each grid-point to each PC. The values in parentheses are the percentage variance explained by each component. In the winter PC3, explaining 9.8% of the variance, for example, the highest variance
9
occurs in the grid points over the southern UK. It can be seen that the majority of the patterns are common between these 2 seasons, as in spring and autumn (not shown), although there are some differences in the ordering. For example, the PC6 map for winter (Fig. 1) is clearly the equivalent to that of PC3 in summer (Fig. 2). For each time-step of the analysis a score for each PC is obtained, which represents the strength of a PC weight on a particular day. Interpretation of the PC patterns can be aided by the construction of composite circulation plots. Figs. 3 (winter) & 4 (summer) show the mean pressure pattern of those days with component scores of more than twice the standard deviation (positive and negative) from the mean value. A greater number of composites are shown for winter as these are referred to later. These composites reveal that high positive scores correspond to high pressure at the centre of maximum variance, and large negative scores correspond to low pressure. Taking winter PC3 as an
Fig. 1. Winter principal component loadings, 1982 to 1991. Values in parentheses are the percentage variance explained
10
Clim Res 14: 7–24, 2000
Fig. 2. Summer principal component loadings, 1982 to 1991. Values in parentheses are the percentage variance explained
example again, the composite map for high positive score days shows an anticyclone centred on the region of maximum variance (Fig. 3). Similarly, the composite for the days of large negative scores reveals low pressure in this region. For brevity, hereafter, ‘high scores’ refer to high positive scores, and ‘low scores’ to large negative scores. As previously indicated, the majority of patterns are common to all 4 seasons. The first component in all seasons represents variation in the strength of the zonal circulation. High pressure over Greenland and low pressure to the north of Scandinavia on high score days give a weakened zonal circulation (see PC1 in Figs. 3 & 4). Conversely, low scores indicate a strength-
ening of flow with low pressure to the north of the domain, over Greenland, and high pressure to the south. In summer, this low pressure area is further north, reflecting the northward movement of the northern hemisphere storm tracks, with a ridge of high pressure extending from the southwest of the domain into northwest Europe (compare Fig. 4 with Fig. 3). Another common pattern is a centre in the northeast of the domain (winter PC6, Fig. 1; summer PC3, Fig. 2). This component represents changes in the location of the north Atlantic storm track. On high score days of winter PC6 (Fig. 3), pressure is high over the Barents Sea and depressions are deflected southwards over northern Europe. In the opposite phase, pressure is low in this region as depressions travel to the north of high pressure over Europe. Further discussion, and interpretation, of the PC patterns, in climatological terms, will be restricted to a comparison with the LWT (below). This is because the purpose of this paper is to establish links with pollutant
Jones & Davies: Climatic influence on air and precipitation chemistry
11
Fig. 3 (Above and on the following 2 pages). Winter mean sea level pressure composites for days with component scores > 2.0 (high) and 2.0 (high) and
