Synoptic weather types in the south of South America and

37

Australian Meteorological and Oceanographic Journal 60 (2010) 37-48

Introduction

The Argentine Pampas Region extends over 600 000 km2

and is one of the most important crop-producing regions in the world (SAGPyA 2008) (Fig. 1(a), shaded rectangle). In

this region, extreme rainfall events often occur and have a significant socio-economic impact. In Argentina, irrigation is very rarely used, and rainfall and its spatio-temporal variability are therefore crucial for successful harvests. Rainfall events depend, among other factors, on the large-scale atmospheric fields. There are two main approaches in synoptic climatology to investigate the links between local-scale environmental features and large-scale circulation

Synoptic weather types in the south of South America and their relationship to daily rainfall in the core crop-producing region in Argentina

M. Laura Bettolli1,2, Olga C. Penalba1 and Walter M. Vargas1,2

1Departamento de Ciencias de la Atmósfera y los Océanos, FCEN,Universidad de Buenos Aires, Argentina

2Consejo Nacional de Investigaciones Científicas y Técnicas, Argentina

(Manuscript received February 2009, revised November 2009)

This paper classifies daily geopotential height anomalies at the 1000 hPa and 500 hPa levels for the period 1979-2001, and examines the capability of these atmospheric patterns to discriminate between the occurrence and non-occurrence of rainfall in the core crop-producing region located in the Argentine Humid Pampas. The resulting optimal classification has seven and five weather-type categories for the 1000 hPa and 500 hPa levels, respectively. The analysis focusses on the season of highest rainfall in the region, which coincides with the growing season of the main summer crops (October-May). These crops are critically dependent on rainfall and its spatial and temporal distribution, since irrigation is rarely employed in the region. The synoptic structures identified in this work can be related to daily rainfall in the region under study. Dry days are significantly favoured by structures with negative geopotential height anomalies at low levels toward the south of the South Atlantic Ocean, inducing an anomalous flow from the southwest over the south of the continent. They are also favoured by positive geopotential height anomalies centred to the east of the continent, inducing stability at low levels. At the 500 hPa level, a positive anomaly in the southwest of the continent centred over the Pacific Ocean, intensifying ridges west of the Andes, favours dry days. On the other hand, rainy days are favoured by patterns with positive geopotential height anomalies at the 1000 hPa level over the south of the continent, enhancing an anomalous flow from the east-southeast in the central region of Argentina and its corresponding moisture advection at low levels. Rainy days are also favoured by the patterns characterised by a cyclonic anomaly to the east of the continent, affecting the whole region. At the 500 hPa level, local rainfall can be related to patterns characterised by negative geopotential height anomalies centred over the Falkland Islands, which induce instability. More regional rainfalls are favoured by structures which weaken the westerlies over Patagonia at the 1000 hPa level and by structures which induce northwest flow over Patagonia and northeast flow over the northeast of the country at the 500 hPa level. The seasonal frequencies of the 1000 and 500 hPa synoptic weather types are consistent with the interannual rainfall variability documented for the region. The influence of the El Niño-Southern Oscillation was also identified in the daily atmospheric circulation anomalies in the south of South America.

Corresponding author address: M. Laura Bettolli, Departamento de Cien-cias de la Atmósfera y los Océanos, Facultad de Ciencias Exactas y Natu-rales, Universidad de Buenos Aires, Argentina.Email: [email protected]

38 Australian Meteorological and Oceanographic Journal 60:1 March 2010

patterns; the environment-to-circulation approach and the circulation-to-environment approach (Yarnal and Draves 1993). The former structures the circulation data based on criteria defined by the environmental variable of interest (e.g. rainfall). Conversely, the latter classifies the circulation data prior to seeking links with the local-scale environment. The second approach is used here to objectively identify synoptic weather types in southern South America and to examine the capability of these atmospheric field patterns to discriminate between the occurrence and non-occurrence of rainfall in the core crop-producing region of Argentina. The environment-circulation approach has been used in a considerable number of studies in Argentina to identify the atmospheric circulation associated with particular surface environmental conditions (e.g. Malaka and Nuñez 1980; Minetti and Sierra 1989; Alessandro and Lichtenstein 1996; Alessandro 1998; Labraga et al. 2002; Müller et al. 2003; Escobar et al. 2004; Scian et al. 2006). The second approach has been addressed by many studies using a wide range of methods in different regions of the world. Recent examples in the literature include Vera and Vigliarolo (2000), Timbal and McAvaney (2001), Gutiérrez et al. (2004, 2005), Pezza and Ambrizzi (2005), Li et al. (2005), Hope et al. (2006) and Lorenzo et al. (2008). However, in Argentina, few studies identify synoptic patterns and then relate these to observed rainfall events. For example, Solman and Menéndez (2003) classified National Center for Atmospheric Research/National Centers for Environmental Prediction (NCAR/NCEP) re-analysis daily averages of geopotential height at the 500 hPa level in winter time from 1966 to 1999, using the k-means clustering method. The authors studied

their duration, persistence, transitions from cluster to cluster and their relationship to the temperature and rainfall in different regions of Argentina. Bischoff and Vargas (2003, 2006) classified daily circulation at the 500 hPa level using Lund’s method on the European Centre for Medium-range Weather Forecasts’s (ECWMF) daily re-analyses and studied its association with extreme climate conditions over the south of South America for a nine-year period (1980-1988). Compagnucci and Salles (1997) characterised the main synoptic patterns over southern South America by means of a principal component analysis (PCA) of observed daily sea-level pressure for the 1972-1983 period without looking for a relationship with surface variables (e.g. rainfall). Similarly, Compagnucci et al. (2001) classified the dominant evolution types of synoptic systems by using an extension of the traditional T-mode PCA methodology called Principal Sequence Pattern Analysis (PSPA) for daily 1000 hPa geo-potential height spatial fields in 1997. The literature on South American synoptic climatology has mostly dealt with individual rainfall events, specific seasons or a limited number of years. In addition, statistical characteristics of the weather types in southern South America and their link to daily rainfall are lacking. Thus, the objective of this paper is to identify daily circulation patterns in southern South America and to associate them with different rainfall conditions in the core crop-producing region in the Argentine Humid Pampas. The present study follows a line of research in applied climatology started by Penalba et al. (2007), whose main goal is the evaluation and quantification of the impact of climate variability on crop production. Classification of the

Fig. 1 (a) Map of the Pampas Region of Argentina (shaded rectangle) and domain chosen for the atmospheric circulation fields (dashed line) and (b) locations and names of the five meteorological stations used in this analysis.

Bettolli et al.: Synoptic weather types and their relationship to daily rainfall in Argentina 39

atmospheric circulation structures and knowledge of their main properties and links with surface variables are basic elements for diagnosis and forecasting. The advantage of synoptic classification is that it provides a tool that could be introduced as a potential diagnostic and monitoring element to be incorporated in impact models necessary for planning agricultural activities and decision-making.

Data and methodology

DataTwo data-sets were used in this study:(a) Daily average fields of 1000 and 500 hPa geopotential

height from the NCEP Reanalysis 2 (Kanamitsu et al. 2002) during the 1979-2001 period, which covers the satellite period of the 20th century. The choice of 1979 as the start of the study period is motivated by the improved availability and quality of the southern hemisphere data in the NCEP/NCAR re-analyses during the satellite era. In addition, the necessity of stratifying events before and after the Pacific regime shift (Miller et al. 1994; Stephens et al. 2001) that occurred around the mid-1970s is avoided. The domain chosen extends from 15°S to 60°S latitude and from 30°W to 90°W longitude and includes 475 grid-points (2.5° latitude by 2.5° longitude resolution) (Fig. 1(a), dashed rectangle). This domain extends over the Andes Mountains and the South Pacific and South Atlantic Oceans, all of which have a significant influence on the atmospheric circulation over South America. In order to remove the seasonality of daily fields, the daily average field in the period analysed was subtracted from each day. The resulting daily geopotential height anomaly fields were used.

(b) Daily rainfall series in the Argentine Humid Pampas in the period restricted to the period of the re-analysis (1979-2001). These series were provided by the Argentine National Meteorological Service. Although there are more rain gauges in this region, only five stations with sufficiently long records, less than ten per cent of missing data and continuity of records were considered appropriate for the analysis. The stations used were: Paraná Aero (31°47’S 60°29’W), Marcos Juárez Aero (32°42’S 62°9’W), Rosario Aero (32°55’S 60°47’W), Junín Aero (34°33’S 60°55’W) and Nueve de Julio (35°27’S 60°53’W) (Fig. 1(b)). These stations are located in the core region for growing various crops (SAGPyA 2008).

The climatological monthly rainfall in the core agricultural region of Argentina is characterised by Rosario Aero station (Fig. 2). The wet season extends from October to May. These months also correspond to the growing period of the main summer crops in the Argentine Humid Pampas. These crops are critically dependent on rainfall and its temporal distribution, since irrigation is rarely employed in the region. Consequently, the period analysed in both data-sets is only from October 1 to May 31 (243 days) from 1979 to 2001, spanning 22 periods.

MethodologyDefinition of rainfall groups. In order to describe the diff-erent rainfall events in the core crop production region, days were stratified according to rainfall spatial coverage and daily intensity. The daily rainfall groups (R) were given labels of the form Rx,y, where (a) x indicates the intensity of rainfall according to three thresholds: x = W (wet) corres-ponds to daily rainfall greater than 0.1 mm; x = VW (very wet) corresponds to daily rainfall greater than or equal to 10 mm, and x = D (dry) corresponds to days on which there was no rainfall; and (b) y indicates the spatial coverage of rainfall according to the number of stations: y = S (some) corresponds to rainfall at one or two stations; y = M (most) corresponds to rainfall at three or more stations and y = A (all) corresponds to rainfall or no rainfall at all five stations. Taking into account these criteria, different rainfall groups were created: for example, the RWS group corresponds to days with precipitation above 0.1 mm at one or two stations while the RVWA group corresponds to days with precipitation above or equal to 10 mm at the five stations at the same time. Note that only one group involves dry days: the RDA group represents the days without precipitation at the five stations. The 10 mm threshold was used to define intense precipi-tation days since this is the rainfall amount needed to exceed the daily evaporation rate for the time of the year and region considered (Vargas 1979). In addition, of the 31 days (annually averaged) with precipitation above 10 mm, 27 (87 per cent) occur between October and May.

Weather types. There is a wide range of classification techniques represented in the literature. In their review, Yarnal et al. (2001) identified the following: manual typing, correlation-based methods, eigenvector-based methods (such as PCA and Empirical Orthogonal Functions (EOFs)), other multivariate classifications (such as Cluster Analysis and Discriminant Analysis), compositing, indexing and new classification techniques (such as self-organising maps). In the present study, cluster analysis was performed together with PCA to determine the synoptic-scale circulation patterns (as in Romero et al. (1999a); Rodrigues Chaves and Cavalcanti (2001)). PCA was used as a synthesis method and

Fig. 2 Monthly average rainfall at Rosario Aero (mm), 1979-2000.


the cluster analysis was carried out in the subspace given by the leading unrotated principal components (PCs). As shown by Gong and Richman (1995), this combination of methods provides the most separable cluster system. For this data-set in particular, the PCA results were discussed by Bettolli et al. (2007). The authors performed the method in the T-mode using the correlation matrix as input (Jolliffe 1986; Richman 1986), with the daily fields as variables and the grid-points as observations. The first six PCs account for 80.7 per cent and 83.5 per cent of the variance at the 1000 and 500 hPa levels, respectively. Orthogonal rotations failed to produce an effective redistribution of the variance (see Bettolli et al. (2007) for more detail). Therefore, the information contained in the subspace of the unrotated six principal components was used in this study at both levels as input for the cluster analysis. The clustering algorithm used in this study is the ‘k-means’ method, which is a partitioning method that classifies all days into a predefined number of clusters (MacQueen 1967). The method minimises the variability within each cluster and maximises the variability between clusters. In this approach, the number of partitions is pre-specified, so the choice of the optimal number of clusters was established by the pseudo-F statistic (Calinski and Harabasz 1974). Milligan and Cooper (1985) conducted a comprehensive evaluation of more than 30 methods for determining the number of clusters in a data-set. They found the pseudo-F statistic to be one of the more effective procedures. It is defined as

...1

...2

pseudo-F =

Z =

= 0.0234FFDI ...7

...8

A(n – k)

W(k – 1)

p1– p2

fp1 = Pr(Rx,y | WTi) =

n1 n2

p1(1 – p1) + p2(1 – p2)

∂FFDI

ΔFFDI = 0.0234FFDIΔV

∂V

Rx,y i

ni

fp2 = Rx,y

n2

where A and W are the among and within-cluster sum of squared distances respectively (among centres of clusters and from each object to the centre of its cluster), n is the number of objects to be classified (daily fields) and k is the number of clusters. Generally, the pseudo-F statistic shows high values for small k (when A is high and W is small), decreasing as k increases. Maximum local peaks indicate an appropriate number of clusters (Romero et al. 1999b). In order to establish the number of clusters suggested by the method, a progressive clustering was performed, with k ranging from 2 to 30. In the clustering final solution, the synoptic weather types (WT) are indicated by their group number and atmospheric level. For example, WT1L denotes the first low-level (1000 hPa) synoptic weather type while WT2M denotes the second mid-level (500 hPa) synoptic weather type.

Comparison between probabilities. The probability of occurrence of a given rainfall event conditioned to a specific WT was compared to the climatological probability of its occurrence. The Z-statistic was used to quantify the difference of probabilities (Infante Gil and Zárate de Lara 1984). Here,

...1

...2

pseudo-F =

Z =

= 0.0234FFDI ...7

...8

A(n – k)

W(k – 1)

p1– p2


n1 n2

p1(1 – p1) + p2(1 – p2)

∂FFDI


∂V

Rx,y i

ni

fp2 = Rx,y

n2

where p1 is the conditional probability of occurrence of a daily precipitation event (Rx,y) conditioned to a given WT. This is

...1

...2

pseudo-F =

Z =

= 0.0234FFDI ...7

...8

A(n – k)

W(k – 1)

p1– p2


n1 n2

p1(1 – p1) + p2(1 – p2)

∂FFDI


∂V

Rx,y i

ni

fp2 = Rx,y

n2

where ni is the number of occurrences of the specific weather type WTi, i being its pattern number (ranging from 1 to 7 at the 1000 hPa level and from 1 to 5 at the 500 hPa level), and fRx,yi

is the number of occurrences of WTi days that fulfill the Rx,y precipitation condition. Lastly, n1 = ni . The conditional probability p1 is compared with the climatological probability p2 of the Rx,y precipitation event, where

...1

...2

pseudo-F =

Z =

= 0.0234FFDI ...7

...8

A(n – k)

W(k – 1)

p1– p2


n1 n2

p1(1 – p1) + p2(1 – p2)

∂FFDI


∂V

Rx,y i

ni

fp2 = Rx,y

n2

Here, fRx,y is the number of occurrences of the Rx,y

precipitation event in the total sample of size n2. The Z-statistic follows a standard normal distribution (i.e. Z ~ N(0,1)) if n1 and n2 are large enough. If n1 and n2 are small, Z follows a Student’s t distribution with (n1 + n2 – 2) degrees of freedom (Infante Gil and Zárate de Lara 1984). A two-tailed test of significance was used. Thus, if the Z-statistic value is positive (negative) and is significant, the specific circulation pattern has (does not have) a significant contribution to the rainfall event.

Results

Characterisation of rainfall groupsInitially, the most important characteristics of the rainfall groups in the core crop-producing region were studied. Table 1 shows the percentages of days corresponding to each rainfall group. The percentage of simultaneous dry days is 51.8 per cent (Table 1, RDA), indicating that slightly more than half the period shows dry conditions. These days do not wholly represent the lack of rainfall in the region; however, they are useful when analysing the synoptic situations associated with no rainfall at the regional scale. The extreme situation of no rainfall in the region will be determined by the frequency and sequentiality of these synoptic situations. The percentage of rainy days at three or more stations RWM (25.2 per cent) is slightly higher than that of local rainfall RWS (23 per cent), indicating that the probability of one

Table 1. Percentages of days corresponding to each rainfall group Rx,y.

Rainfall Percentage Rainfall Percentagegroup of days group of days

RWS 23.0% RVWS 15.9%RWM 25.2% RVWM 8.3%RWA 9.5% RVWA 1.8%RDA 51.8%


local rainy day is similar to that of a generalised rainy day. However, when analysing rainfall intensity, the percentage of days with intense rainfall in three or more stations RVWM (8.3 per cent) represents half of the percentage of days with intense local rainfall RVWS (15.9 per cent). This indicates that the probability of one day with intense local rainfall is double the probability of one day with intense regional rainfall. In the RVWA group, only 1.8 per cent of the total days satisfy the conditions of intensity and spatial coverage set forth (Table 1). When making a detailed analysis of the rainfall, it was found that most days satisfying this condition show rainfall amounts above 20 mm at at least four of the five stations considered (results not shown). In isolation, it is possible to find rainfall totals above 80 mm, showing the complexity of the problem when it is addressed at the synoptic scale. Thus, the low probability of occurrence of these days and the high levels of precipitation determine the condition of an extreme event.

Synoptic weather typesThe values of the pseudo-F statistic (Eqn 1) as a function of the number of groups or clusters in the k-means method are shown in Fig. 3. The results suggest a seven-cluster solution for the 1000 hPa level. At the 500 hPa level, however, the number of clusters cannot be clearly defined by means of this criterion. Local peaks of the pseudo-F statistic can be observed for four and six clusters. Consequently, five clusters (the average of these) were adopted for this level. These solutions determine well separated and homoge-neous groups, since the highest spatial correlations between the centres of the clusters were 0.38 and 0.27 at the 1000 and 500 hPa levels, respectively, showing a good separation between them (according to Mo and Ghil 1988). Climatological values of the standard deviation of geopotential height (over the 1979-2001 period) are more than 80 m at the 1000 hPa level, and more than 120 m at the 500 hPa level over the south of the region studied. For the solutions found, the highest standard deviations are lower than 60 m and 90 m respectively (results not shown). Therefore, the clusters found at each pressure level may also be considered homogeneous from this point of view. The synoptic weather types are shown in Figs 4 and 5. At the 1000 hPa level, WT1L is characterised by a negative geopotential anomaly centred at 60°S 45ºW, inducing an anomalous flow from the southwest over the south of the continent and adjacent ocean, along with an intensification of the westerlies over the South Atlantic Ocean. The centre of negative minimum anomalies is aligned in the northwest-southeast direction and can be associated with the trough axis or cold front affecting the region. WT2L is characterised by a dipolar structure between both oceans: a centre of positive anomalies over the South Atlantic Ocean and a centre of negative anomalies over the South Pacific Ocean, both at approximately 55°S, intensifying the flow from the north over the southeast of the continent. WT3L shows a positive anomaly of geopotential height centred at the southern

tip of the continent, favouring an anomalous flow from the east-southeast in the central region of Argentina and its corresponding moisture advection at low levels. WT4L shows an anticyclonic anomaly with centre to the east of the continent. This pattern can be associated with an anomalous flow induction from the north all over the region, which intensifies the warm and humid advection from subtropical zones. In WT5L, a centre of positive anomalies dominates the circulation over the South Pacific Ocean. It can be related to the intensification of migratory anticyclones from the west, generating cold air advection in the south of South America. This pattern is accompanied by a negative anomaly to the east, over the South Atlantic. WT6L is characterised by a cyclonic geopotential height anomaly centred at 45°S 55°W. This negative anomaly pattern affects almost the whole country, inducing an anomalous flow from the south over the region. Finally, WT7L shows negative anomalies at the southwest extreme of the continent, intensifying the westerlies south of 40°S. A short-wave disturbance can be observed in the centre of Argentina. At the 500 hPa level (Fig. 5), WT1M is characterised by a positive anomaly at the extreme south of the continent. This anomaly pattern weakens westerlies at subtropical and middle latitudes. In WT2M, it is possible to observe two opposed centres of geopotential height anomalies located over the oceans, inducing anomalous southerly flow. The

Fig. 3 Pseudo-F statistics as a function of the number of clusters for the 1000 and 500 hPa levels.


WT3M presents the opposed dipolar structure, inducing an anomalous flow from the north. The WT4M shows a cyclonic anomaly centred over the north of the Falkland Islands and affecting the whole country. This configuration causes an anomalous flow from the south in Patagonia (south of the country) and the centre of Argentina. Finally, WT5M shows negative anomalies to the south of the continent favouring flow from the west over Patagonia; at the same time, a positive anomaly over the Atlantic Ocean induces an anomalous flow from the northeast at lower latitudes.

Fig. 4 1000 hPa Synoptic Weather Types (WT) and percent-age of days corresponding to each group. The dashed (solid) lines represent negative (positive) geopoten-tial height anomalies. The contour interval is 20 m.

Fig. 5 As in Fig. 4, except for 500 hPa.


At both levels, the percentage of days corresponding to each WT was calculated (numbers in Figs 4 and 5). The most remarkable result is the low level of variation in the number of days in each cluster, the main exceptions being WT4L (10.9 per cent) and WT6L (11.3 per cent) at the 1000 hPa level, and WT4M (15.9 per cent) at the 500 hPa level, with the fewest number of days. Some of the WTs found here are similar to those found by Solman and Menéndez (2003) and Bischoff and Vargas (2003) at the 500 hPa level. However, the main differences are in the domain extension, and the months and years of study. In this sense, these results provide a complementary approach to the classification of atmospheric circulation patterns in the south of South America.

Synoptic weather types and daily rainfallThe purpose of this section is to evaluate how much rainfall information for the core crop-producing region is contained in the circulation structures at the regional scale. In order to make this analysis, the Z-statistic (Eqn 2) was calculated. Its values and significance are shown in Table 2.

Dry days. At the 1000 hPa level, WT1L and WT4L have the highest contribution to dry days, showing positive and significant values of the Z-statistic for the RDA condition (first column in Table 2). The configuration of anomalies of these WTs induces stability at lower levels in the region of study. The WT2L pattern also has a significant contribution to dry days, though not as many as the first and fourth types. At the 500 hPa level, the Z-statistic values indicate that the WT2M pattern may be responsible for an increase in the frequency of dry days. This WT shows a positive anomaly in the southwest of the continent centred over the Pacific Ocean, intensifying ridges west of the Andes, inducing an anomalous advection of anticyclonic vorticity over the Humid Pampas, which favours stability at the middle and low levels of the atmosphere.

Rainy days. The analysis of the atmospheric circulation associated with rainy days shows that results are more variable and depend on the rainfall conditions considered. At the 1000 hPa level, the lowest values of the Z-statistic for dry days (RDA, first column in Table 2) are those for the WT3L and WT6L patterns, with these having positive Z values for the rainy days groups (remaining columns in Table 2, although with one exception – WT3L/RVWS). WT3L induces east-southeast anomalous flow and consequently increases moisture advection over the region considered (Fig. 4). WT6L shows negative anomalies in the geopotential heights with a centre to the east of the continent over the South Atlantic Ocean, probably associated with cold front passages or cyclogenesis (Fig. 4). In accordance with this, the WT4M pattern at the 500 hPa level appears to be associated with a lower frequency of dry days (RDA) and a higher frequency of rainy days at the local level (RWS). WT5M favours the frequency of days with daily precipitation above 10 mm at subregional-regional scale (RVWM in Table 2, column 6). WT3M, which represents cyclonic anomalies over the Pacific Ocean and anticyclonic anomalies over the Atlantic Ocean, does not appear to be significantly associated with increased or decreased frequencies for any of the rain day groups, nor does it show consistency of Z-statistic sign. This suggests that the structure could be related to both dry and rainy days of any group. This would depend on the position of the trough over the South Pacific Ocean, which would condition the anomalous cyclonic vorticity advection, and on the persistence of the east-northeast flow at the surface, which would provide the moisture needed for precipitation.

Regional heavy rainy days. No significant values for the Z-statistic were found for the 1000 hPa level (Table 2, last column) for the most extreme regional rainfall events (RVWA

group), the highest (but not significant) values coming from the WT3L and WT7L weather types.

Table 2. Z-statistics of the comparison between the conditional probability of occurrence of a day with a certain rainfall condition in each circulation cluster at 1000 and 500 hPa and the climatological probability of that day. Significant values at 95 per cent (*) and 90 per cent (**).

1000 hPa RDA RWS RWM RWA RVWS RVWM RVWA

WT1 3.78* −0.82 −3.67* −3.24* −1.61 −1.68** −0.09WT2 1.74** −1.32 −0.65 0.12 −9.05* 0.57 −0.09WT3 −3.97* 1.24 3.25* 2.95* −0.77 2.62* 0.96WT4 3.41* −2.49* -1.59 −1.87** −1.18 −2.05* −0.60WT5 −2.76* 2.48* 0.77 −1.76** 2.03* −0.64 −1.36WT6 −3.98* 1.99* 2.52* 1.04 3.02* 1.01 0.47WT7 0.52 0.46 −0.94 −0.50 0.50 1.01 1.21

500 hPaWT1 −1.01 0.59 0.71 0.74 −1.20 0.49 0.96WT2 1.96** −0.82 −1.36 −2.63* −0.60 −1.38 −1.48WT3 0.69 −0.20 −0.46 0.60 −0.48 0.48 0.51WT4 −2.59* 1.86** 1.22 −0.43 1.24 −1.40 −0.46WT5 −0.37 0.51 0.07 −0.38 0.99 2.61* 0.90


At the 500 hPa level, WT1M and WT5M are the weather types with the strongest associations with the RVWA group (Table 2, last column), although as at the 1000 hPa level neither association is statistically significant, while WT2M contributes least to heavy rainy days, which is consistent with its association with the dry days group (RDA).

Vertical structureIn order to analyse the vertical structure of the atmosphere in relation to rainfall in the core crop-producing region, the distribution of the joint 1000 and 500 hPa frequencies were included in the study. The most frequently occurring combinations of WTs at the two levels are WT3L/WT1M, WT5L/WT2M, WT6L/WT4M and WT7L/WT5M (results not shown). The WT1L and WT4L patterns show different combinations with WTs at the 500 hPa level, indicating that these surface patterns have more variable vertical structures. The WT2L/WT2M, WT3L/WT5M and WT5L/WT3M combinations were found to be very rare. The joint frequency of the circulation patterns at both levels was analysed for the most extreme conditions: heavy rainy days (RVWA) and dry days (RDA) (Table 3). The conditional probability of occurrence of an RVWA or RDA day, for a given combination of WTs, was compared with the climatological probability of occurrence of that rain day group, using the Z-statistic. For the RVWA group, there are no statistically significant increases in the probability of occurrence. However, the weather-type combinations WT3L/WT1M and WT7L/WT5M patterns have the highest frequencies of joint occurrence with positive Z-statistic, consistent with previous findings. Only the combination WT5L/WT2M shows a significant departure from the climatological frequency, resulting in fewer wet days. This result indicates that this last combination of spatial structures does not contribute significantly to regional heavy rainy days. This vertical structure responds to an anticyclonic disturbance, located at the southwest of the continent, associated with a ridge at upper levels. There are numerous structure combinations for which no RVWA events occurred (as indicated by the empty cells in Table 3).

For the RDA group, WT2M combined with the WT1L and WT7L patterns significantly favours dry days. In the first combination, WT1L/WT2M, the vertical structure as a whole favours dry conditions. In the second (and rarer) combination WT7L/WT2M, it is the ridge at middle levels that provides the mechanisms needed to inhibit rainfall in the region. Another combination of structures significantly favouring dry days is WT4L/WT3M corresponding to a positive anomaly over the east of the continent and adjacent ocean, inducing subsidence and warm advection across the layer. The combinations that result in statistically significant reductions in the numbers of dry days (i.e. those that have significant negative values of the Z-statistic) include WT3L/WT1M (positive anomalies to the south of the continent, anomalous flow from east-southeast) and WT6L/WT5M (surface cyclonic disturbance associated with an upper-level trough). WT3M, combined with WT3L and WT6L, also reduces the frequency of dry conditions. In the first case, WT3L provides the necessary moisture, favouring the advection from the east-southeast over the region; in the second case, WT6L provides the cyclonic disturbance needed for flow convergence at the lower levels of the atmosphere. Likewise, WT4M does not contribute to dry conditions when combined with WT3L, WT5L and WT6L. These results agree with the results from two pressure levels taken separately, so it is the vertical structure that does not favour dry conditions.

Temporal variabilityThe time series of the seasonal WT frequencies are examined in order to explore their main characteristics (Figs 6 and 7). This analysis is performed even though the analysed period is relatively short to infer temporal variabilities. These time series were analysed for linear trends at the 95 per cent confidence level, considering the correlation coefficient as an indicator of a possible trend (numbers in Figs 6 and 7). At the 1000 hPa level, the seasonal frequencies of WT7L show a gradual and significant increase over time. WT7L and WT1L, which both represent enhanced westerlies, show the strongest interannual variability. A very different behaviour is found for WT4L, which is characterised by an anticyclonic anomaly inducing an anomalous flow from the north. It has

1000 hPa RVWA RDA

WT1 WT2 WT3 WT4 WT5 WT6 WT7 WT1 WT2 WT3 WT4 WT5 WT6 WT7

WT1 (+)1 (−)2 (+)15 (−)2 (+)4 WT1 (+)8 (+)67 (−)239 (+)131 (−) 60 (+)6 (+)6WT2 (−)6 (+)1 (−)2 (−)5 (+)1 WT2 (+)231 (+)2 (+)26 (+)80 (−)272 (+)14 (+)7WT3 (−)12 (+)2 (+)2 (+)9 WT3 (+)9 (+)396 (−)30 (+)64 (−)1 (−)16 (−)121WT4 (−)2 (+)2 (−)2 (−)7 (+)1 WT4 (+)98 (−)7 (−)32 (+)8 (−) 63 (−)179 (+)16WT5 (+)6 (+)2 (+)3 (+)3 (+)12 WT5 (+)140 (−)14 (−)3 (+)64 (−)5 (−)49 (+)310

5

00 h

Pa

Table 3. Absolute frequencies of days for the RVWA and RDA events according to the combination of patterns at the 1000 and 500 hPa levels. Bold: relative frequencies that are significantly different from the climatological relative frequency at the 95 per cent confidence level. In parenthesis: sign of the Z-statistic.


little trend and little interannual variability. The maximum frequency of occurrence of this WT appears in the 1994/95 season, coincident with the below-normal precipitation during 1994 and the drought in 1995 which affected a vast region of Argentina (Alessandro and Lichtenstein 1996).

Seasonal frequencies of WT5L, which is associated with an anticyclone inducing enhanced cold air advection situations, present a progressive decrease over time. This temporal behaviour is coincident with the increase of summer minimum temperature values observed in recent decades over the studied area (Easterling et al. 1997; Rusticucci and Barrucand 2004). At the 500 hPa level, strong interannual variability is observed for WT3M and WT5M, while the opposite, with a significant decrease in the frequencies, is found for WT4M (Fig. 7). WT5M shows a positive trend in the seasonal frequencies, but one which is not significant. However, this characteristic is consistent with the temporal behaviour of its corresponding pattern at the 1000 hPa level, WT7L. These findings are in agreement with the changes observed both in the rainfall totals and in the frequency of rainy days in the last decades (see for example, Minetti et al. ( 2003) and Penalba and Vargas (2004)). Several studies have documented that a large part of the interannual variability of the southern hemisphere circulation is associated with the El Niño–Southern Oscillation (ENSO) phenomenon (e.g. Kiladis and Mo 1998). In this section, the relationship between ENSO and WT interannual variations is investigated. During the strongest El Niño events of the study period (1982/83, 1991/92 and 1997/98), WT7L and WT1L show pronounced low frequencies, implying a reduced zonal flow component. During La Niña events, however, high values of seasonal frequency are observed (e.g. the maximum frequency in WT7L in the 1988/89 La Niña event). During El Niño events, the atmospheric circulation is more perturbed than during other years, and the transient component and the meridional transport are therefore enhanced. According to van Loon (1979), the mechanisms for heat transport are dominated by the transient component of the circulation in the southern hemisphere. In this way, the increase in the frequencies of WT6L also represents an increase of the poleward heat transport. These results are consistent with the results observed at the 500 hPa level, evidencing a weakening of zonal flow and an enhancement of meridional flow. WT5M shows minimum frequency during El Niño events, while WT4M shows higher frequency. According to several authors, ENSO is responsible for most of the interannual rainfall variability in the region occurring mainly during the summer season (e.g. Grimm et al. 2000; Boulanger et al. 2005). In this sense, the WTs’ temporal variability is in agreement with the observed precipitation variability. Although longer series are necessary to achieve more rigorous results, these findings provide insight into the possible mechanisms that may relate the daily scale circulation with the variability observed in the rainfall regimes of the region, especially in the warm period.

Conclusions

This paper classifies daily geopotential height anomalies at the 1000 hPa and 500 hPa levels in order to identify the most frequent circulation patterns in the south of South America. In addition, it examines the capability of these atmospheric

Fig. 6 Time series of the seasonal frequencies of 1000 hPa WTs. The years on the horizontal axis refer to the start year of the season. For example, 1979 refers to the season Oct 1979-May 1980. The correlation coef-ficients of the linear trend are indicated. Bold denotes significant values at the 95 per cent confidence level.

Fig. 7 As in Fig. 6, except for 500 hPa.


field patterns to discriminate between the occurrence and non-occurrence of rainfall in the core crop-producing region located in the Argentine Humid Pampas. The analysis was made during the season of highest rainfall in the region, coinciding with the growing season of the main summer crops (October-May). These crops are critically dependent on rainfall and its spatial and temporal distribution, since irrigation is rarely employed in the region. The synoptic structures identified in this work can be related to daily rainfall in the region of study. The classification scheme is effective not only in discriminating dry and rainy days, but also in differentiating between a range of rainfall intensities and spatial distributions. In some cases, the vertical structure, through the combination of structures at low and middle levels of the atmosphere, contributes to each rainfall condition. In other cases, the structures at a certain level play a more important role in defining the rainfall or no rainfall event. In this sense, the patterns found here represent a first approach to defining and differentiating structures at a larger scale as they organise the lower scale processes responsible for generating precipitation. This analysis, then, could be considered as the first step in a hierarchical classification of structures and variables through which the occurrence of precipitation or no precipitation in the region studied can be explained. The temporal variability of the weather type (WT) frequencies was also analysed. This analysis determined the main characteristics of the atmospheric circulation in the south of South America during the period analysed. The seasonal frequencies of the 1000 and 500 hPa synoptic WTs are consistent with the interannual rainfall variability documented for the region, providing insight into the daily scale circulation anomalies relating to rainfall anomalies. The influence of the ENSO global forcing was identified in the daily atmospheric circulation in the south of South America. Its effect in inducing circulation anomalies is manifested in the changes in the WT frequencies that are related to the transient component and the meridional transport. From this point of view, the structures found at the two pressure levels are representative of the daily atmospheric circulation that can be related to the ENSO-rainfall connection in the Pampas Region. The results showed which atmospheric structures do or do not favour dry and rainy days in the region. These findings provide an approach to the study of the relationships between atmospheric circulation in the south of South America and rainfall events in the Humid Pampas Region. The structures found can be additional objective tools that could help to improve short and middle term forecasts. In the context of impact studies, they represent an important element for the monitoring of the crop growing season and for planning strategies. They can also be useful as a downscaling tool in order to prospect possible climate evolution, especially taking into account the projected global changes for the 21st century.

Acknowledgments

The authors wish to thank the Associate Editor and the anonymous reviewers for their valuable suggestions that improved the manuscript. This work was supported by the following grants from Universidad de Buenos Aires; X170, X228 and X605, BID 1728/OC-AR-PICT 38273 and CLARIS LPB Project (European Community’s Seventh Framework Programme under Grant Agreement No. 212492).

ReferencesAlessandro, A.P. and Lichtenstein, E. 1996. Situación Sinóptica Asociada a

la sequía del invierno de 1995. Revista Geofísica, 45, 125-43.Alessandro, A. 1998. Aspectos sinópticos de las intensas lluvias del verano

1982-1983. Meteorologica, 23, 3-14.Bettolli, M.L., Vargas, W.M. and Penalba, O.C. 2007. Aspectos

metodológicos del uso del análisis de componentes principales en campos de anomalías de altura geopotencial en el sur de Sudamérica. Revista Brasileira de Meteorologia, 22, 273-85.

Bischoff, S.A. and Vargas, W.M. 2003. The 500 and 1000 hPa weather type circulations and their relationship with some extreme climatic conditions over southern South America. Int. J. Climatol., 23, 541-56.

Bischoff, S.A. and Vargas, W.M. 2006. Climatic properties of the daily 500 hPa circulation ECWMF reanalysis data over southern South America during 1980-1988. Meteorologica, 23, 3-12.

Boulanger, J.P., Leloup, J., Penalba, O., Rusticucci, M., Lafon, F. and Vargas, W. 2005. Observed precipitation in the Paraná-Plata hydrological basin: long-term trends, extreme conditions and ENSO teleconnections. Clim. Dyn., 24, 393–413.

Calinski, R.B. and Harabasz, J. 1974. A dendrite method for cluster analysis. Commun. Stat., 3, 1-27.

Compagnucci, R.H. and Salles, M.A. 1997. Surface pressure patterns during the year over southern South America. Int. J. Climatol., 17, 635-53

Compagnucci, R.H., Araneo, D. and Canziani, P.O. 2001. Principal sequence pattern analysis: a new approach to classifying the evolution of atmospheric systems. Int. J. Climatol., 21, 197-217.

Easterling, D., Horton, B., Jones, P., Petterson, T., Karl, T., Parker, D., Salinger, M. J., Rasuvayev, V., Plummer, N., Jamason, P. and Folland, C. 1997. Maximum and minimum temperature trends for the globe. Science, 277, 364-7.

Escobar, G. Compagnucci, R. and Bischoff, S. 2004. Sequence patterns of 1000 hPa and 500 hPa geopotential height fields associated with cold surges over Central Argentina. Atmósfera, 7, 69-89.

Gong, X. and Richman, M.B. 1995. On the application of cluster analysis to growing season precipitation data in North America east of the Rockies. Jnl Climate, 8, 897-931.

Grimm, A.M., Barros, V.R. and Doyle, M.E. 2000. Climate variability in southern South America associated with El Niño and La Niña Events. Jnl Climate, 13, 35–58.

Gutiérrez, J.M., Cofiño, A.S., Cano, R. and Rodríguez, M.A. 2004. Clustering methods for statistical downscaling in short-range weather forecasts. Mon. Weath. Rev., 132, 2169-83.

Gutiérrez, J.M., Cano, R., Cofiño, A.S, and Sordo, C. 2005. Analysis and downscaling multi-model seasonal forecasts in Peru using self-organizing maps. Tellus, 57A, 435–47.

Hope, P.K., Drosdowsky, W. and Nicholls, N. 2006. Shifts in the synoptic systems influencing south west Western Australia. Clim. Dyn., 26, 751-64.

Infante Gil, S. and Zárate de Lara, G. 1984. Métodos Estadísticos. Trillas México, 643 pp.

Jolliffe, I.T. 1986. Principal Component Analysis. Springer-Verlag, 271 pp.Kanamitsu, M., Ebisuzaki, W., Woollen, J., Yang, S.-K, Hnilo, J. J., Fiorino,

M. and Potter, G.L. 2002. NCEP-DOE AMIP-II Reanalysis (R-2). Bull. Am. Met. Soc., 83, 1631-43.

Kiladis, G.N. and Mo, K.C. 1998. Interannual and intraseasonal variability in the Southern Hemisphere. Meteorology of the Southern Hemisphere. Met. Monogr., No. 49, Amer. Met. Soc., 307–36.


Labraga, J.C., Scian, B. and Frumento, O. 2002. Anomalies in the atmospheric circulation associated with the rainfall excess or deficit in the Pampas Region in Argentina. J. Geophys. Res., 107, 1-15.

Li, F., Chambers, L.E. and Nicholls, N. 2005. Relationships between rainfall in the southwest of Western Australia and near-global patterns of sea-surface temperature and mean sea-level pressure variability. Aust. Met. Mag., 54, 23–33.

Lorenzo, M.N., Taboada, J.J. and Gimeno, L. 2008. Links between circulation weather types and teleconnection patterns and their influence on precipitation patterns in Galicia (NW Spain). Int. J. Climatol., 28, 1493-505.

MacQueen, J. 1967. Some methods for classification and analysis of multivariate observations. Fifth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley CA, University of California Press, 281–97.

Malaka, I. and Nuñez, S. 1980. Aspectos sinópticos de la sequía que afectó a la República Argentina en 1962. Geoacta, 10, 1–21.

Miller, A.J., Cayan, D.R., Barnett, T.P., Graham, N.E. and Oberhuber, J.M. 1994. The 1976-77 climate shift of the Pacific Ocean. Oceanography, 7, 21-6.

Milligan, G.W. and Cooper, M.C. 1985. An examination of procedures for determining the number of clusters in a data set. Psychometrika, 50, 159-79.

Minetti, J.L. and Sierra, E.M. 1989. The influence of general circulation patterns on humid and dry years in the Cuyo Andean region of Argentina. Int. J. Climatol., 9, 55-68.

Minetti, J.L., Vargas, W.M., Poblete, A.G., Acuña, L.R. and Casagrande, G. 2003. Non-linear trends and low frequency oscillations in annual precipitation over Argentina and Chile, 1931-1999. Atmósfera, 16, 119-35.

Mo, K. and Ghil, M. 1988. Cluster analysis of multiple planetary flow regimes. J. Geophys. Res., 93D, 10927-52.

Müller, G.V., Compagnucci, R.H., Nuñez, M.N. and Salles, M.A. 2003. Surface circulation associated with frost in the wet Pampas. Int. J. Climatol., 23, 943-61.

Penalba, O.C. and Vargas, W.M. 2004. Interdecadal and Interannual variations of annual and extreme precipitation over central-northeastern Argentina. Int. J. Climatol., 24, 1565-80.

Penalba, O.C., Bettolli M.L. and Vargas, W.M. 2007. The impact of climate variability on soybean yields in Argentina. Multivariate regression. Meteorol. App., 14, 3-14.

Pezza, A.B. and Ambrizzi, T. 2005. Dynamical conditions and synoptic tracks associated with different types of cold surge over tropical South America. Int. J. Climatol, 25, 215-41.

Richman, M.B. 1986. Rotation of Principal Components. Int. J. Climatol., 6, 293-335.

Romero, R., Sumner, G. Ramis, C. and Genovés, A. 1999a. A classification of the atmospheric circulation patterns producing significant daily rainfall in the Spanish Mediterranean area. Int. J. Climatol., 19, 765–85.

Romero, R., Ramis, C., Guijarro, J.A. and Sumner, G. 1999b. Daily rainfall affinity areas in Mediterranean Spain. Int. J. Climatol., 19, 557-78.

Rodrigues Chaves, R. and Cavalcanti, I.F.A. 2001. Atmospheric circulation features associated with rainfall variability over southern Northeast Brazil. Mon. Weath. Rev., 129, 2614-26.

Rusticucci, M.M. and Barrucand, M.G. 2004. Observed trends and changes in temperature extremes over Argentina. Jnl Climate, 17, 4099-107.

SAGPyA. 2008. Estimaciones Agrícolas. Ministerio de Producción, Secretaría de Agricultura, Ganadería, Pesca y Alimentos de la Nación Argentina. http://www.sagpya.gov.ar/SAGPyA/index.php.

Scian, B., Labraga, J.C., Reimers, W. and Frumento, O. 2006. Characteristics of large-scale atmospheric circulation related to extreme monthly rainfall anomalies in the Pampas Region, Argentina, under non-ENSO conditions. Theor. Appl. Climatol., 85, 89–106.

Solman, S.A. and Menéndez, C.G. 2003. Weather regimes in the South American sector and neighbouring oceans during winter. Clim. Dyn., 21, 91-104.

Stephens, C., Levitus, S., Antonov, J. and Boyer, T. 2001. The Pacific regime shift. Geophys. Res. Lett., 28, 3721–4.

Timbal, B. and McAvaney, B.J. 2001. An analogue-based method to downscale surface air temperature: application for Australia. Clim. Dyn., 17, 947-63.

van Loon, H. 1979. The association between latitudinal temperature gradient and eddy transport. Part 1: Transport of sensible heat in winter. Mon. Weath. Rev., 107, 525–35.

Vargas, W.M. 1979. Atlas de excesos y déficits de humedad en la región húmeda y semiárida argentina. Ministerio de Economía, Instituto Nacional de Ciencias Técnicas y Hídricas, 140 pp.

Vera, C.S. and Vigliarolo, P.K. 2000. Diagnostic study of cold-air outbreaks over South America. Mon. Weath. Rev., 128, 3-24.

Yarnal, B. and Draves, J.D. 1993. A synoptic climatology of stream flow and acidity. Clim. Res., 2, 193-202.

Yarnal, B., Comrie, A. C., Frakes, B. and Brown, D.P. 2001. Developments and prospects in synoptic climatology. Int. J. Climatol., 21, 1923–50.

Documents

Synoptic weather types in the south of South America and