Water Resources Research

El Niño and La Niña influence on droughts at different timescales in the Iberian Peninsula

Authors


Abstract

[1] This paper seeks to determine the impact of extreme phases of the Southern Oscillation (SO, El Niño/La Niña) on droughts in the Iberian Peninsula. For this purpose, 51 precipitation series (1910–2000) were used. A spatial classification based on monthly precipitation records was made to identify homogeneous regions and to analyze any spatial differences in the influence of these extreme phases. For each region a drought index was calculated (standardized precipitation index) at timescales of 1, 3, 6 and 12 months. El Niño and La Niña years were identified using the Southern Oscillation index (SOI), in line with studies conducted elsewhere. Mean values of the drought index were calculated each month for the four timescales during El Niño and La Niña years as well as for the year following these events. The statistical significance of any anomalies was evaluated by means of the nonparametric Wilcoxon Mann-Whitney test. The results indicate that extreme phases of the SO significantly affect the occurrence of droughts in the Iberian Peninsula. Moreover, spatial and temporal differences, depending on the timescale used, were identified. Large areas of the Iberian Peninsula are affected by significant negative values of SPI during the final months of La Niña years and the initial months of the following year. In contrast, other areas are affected by dry conditions during the first months of El Niño years as well as during the summers and autumns of the following year. The spatial differences in drought conditions during extreme phases of the SO are noticeable, and the drought signal is more consistent for La Niña than it is for El Niño years. Finally, the usefulness of these results for drought prediction and early warning systems is discussed.

1. Introduction

[2] One of the main atmospheric sources of climatic variability in the global climate system is the Southern Oscillation (SO) [Trenberth, 1997; Philander and Fedorov, 2003], with extreme episodes of this atmosphere-ocean coupled mode being known as El Niño and La Niña. El Niño events correspond to phases of the phenomena in which pressure differences across the tropical Pacific are reduced and the sea surface temperature anomalies are positive in the central and eastern tropical Pacific [Philander, 1990]; La Niña events correspond to phases characterized by cold sea surface temperatures and an enhanced sea level pressure gradient across the tropical Pacific running west to east.

[3] Several studies point to the fact that climate is affected greatly by the SO in tropical and subtropical regions [Ropelewski and Halpert, 1986, 1987]. Although the climatic influence of the SO is mainly detected in the intertropical areas, this atmosphere-ocean coupled mode affects the climate of other regions including Africa [Rocha and Simmonds, 2000], Central Asia [Nazemosadat and Cordery, 2000], Turkey [Karabörk et al., 2005], and even extensive areas of the northern hemisphere including North America [Rajagopalan et al., 2000] and Europe [Ropelewski and Halpert, 1987].

[4] The influence of the SO on pressure anomalies and pressure patterns in North Atlantic European areas has been recognized in several studies [e.g., Rogers, 1984; Fraedrich, 1994; Nobre and Shukla, 1996; Huang et al., 1998; Moron and Gouirand, 2003; Pozo-Vázquez et al., 2005a]. Klein et al. [1999] argue that the SO signal is communicated to the tropical Atlantic via a reduction in the surface latent heat flux associated with lighter trade winds [Saravanan and Chang, 1999; Sutton et al., 2000]. Marshall et al. [2001] proposed a mechanism for the feedback effect of the tropical Atlantic on the North Atlantic circulation by rearrangement of Hadley's circulation. Once the El Niño phase influences tropical Atlantic climatology, it is reasonable to expect that this sector plays a key role in carrying the El Niño signal to midlatitudes by means of a feedback mechanism. La Niña signal can also be carried by this mechanism [Enfield and Mayer, 1997]. Nevertheless, other studies ascribe the SO effect to a connection between the southern and northern Pacific Ocean, which can be propagated downstream to the North Atlantic region in the form of Rossby Waves [Trenberth and Hurrell, 1994; Gershunov and Barnett, 1998].

[5] Whatever the mechanism involved, the influence of the SO on European precipitation has been described in a number of studies, mainly done in spring and autumn [Moron and Ward, 1998; Van Oldenborgh et al., 2000; Lloyd-Hughes and Saunders, 2002a; Mariotti et al., 2002], but also during the winter [Fraedrich and Müller, 1992; Pozo-Vázquez et al., 2005b]. Thus extreme phases of the SO (El Niño and La Niña) can have a significant influence on climatic conditions in Europe [Moron and Ward, 1998]. Stockdale et al. [1998] report a significant increase in summer precipitation in areas of southern Europe during El Niño years. Webster and Palmer [1997] report an increase in precipitation in the center and south of Europe during El Niño years, while Pozo-Vázquez et al. [2001, 2005b] have identified positive winter anomalies in North Atlantic surface pressure and droughts in western Europe during La Niña years. In the Iberian Peninsula, Rodó et al. [1997], Rodríguez-Puebla et al. [1998] and Rocha [1999] have found significant relationships between the SO and the frequency and magnitude of precipitation. Rodó et al. [1997] indicated that the time lag between the records of the SO and its impact seen on rainfall series ranges from 3 to 21 months. These authors have also reported that the strongest SO signal is detected primarily in spring and secondarily in autumn. They estimate that the percentage springtime variability attributable to the SO can be as high as 50% in certain areas, mainly in the southeast. Pozo-Vázquez et al. [2005a] have also shown the tendency for positive North Atlantic Oscillation (NAO)-like climate patterns to exist in the North Atlantic region during the winters following autumns of strong El Niño events and thus point to the existence of a potential source of predictability for the North Atlantic climate.

[6] In the Iberian Peninsula, droughts are a climatic risk that have severe consequences for agriculture [Austin et al., 1998; Iglesias et al., 2003], increasing the frequency of fires [Aguado et al., 2003; Pausas, 2004] and significantly reducing water availability for urban and tourist consumption [Morales et al., 2000]. Droughts are frequent, but not spatially uniform in the Iberian Peninsula [Martín-Vide and Barriendos, 1995; Rodrigo et al., 1999], with significant spatial differences in drought patterns being recorded in this region [Vicente-Serrano, 2005]. In general, these spatial differences are determined by atmospheric conditions, the effects of which on drought conditions have been analyzed synoptically [Olcina, 2001]. The general atmospheric patterns in the Northern Hemisphere [Barnston and Livezey, 1987] also affect the magnitude and spatial extent of droughts [Hurrell and Van Loon, 1997; Rodríguez-Puebla et al., 1998]. The influence of the SO on precipitation in the Iberian Peninsula has been described in a number of studies [e.g., Rodó et al., 1997; Rocha, 1999]. However, no detailed analysis has been undertaken of the influence of the extreme phases of the SO on the intensity and spatial extent of droughts in this region.

[7] The objective of this paper is to examine the influence of extreme phases of the SO on drought conditions in the Iberian Peninsula both spatially and temporally. Such research should be useful for drought management and for early warning systems. Similar research has been conducted to improve the forecasting of droughts in a number of regions elsewhere, for example, in Australia by Cordery and McCall [2000], in the United States by Piechota and Dracup [1996], in South Africa by Cook [2001], and in Brazil by Kane [1997].

2. Methods

2.1. Data

[8] In conducting the drought analysis, 51 precipitation series covering the period between 1910 and 2000 were used (Figure 1). Nine of these series were obtained from the Sistema Nacional de Infromação de Recursos Hídricos in Portugal (http://snirh.inag.pt) while the rest were obtained from the Instituto Nacional de Meteorología in Spain. Some series had to be created by combining data from different observatories within the same locality, owing to the frequent changes of location of Spanish observatories during the twentieth century. The data were subjected to a process of quality control, which identified anomalous records using a quartile-based range statistic, in line with that of González-Rouco et al. [2001]. Anomalous records were removed from the original database. To guarantee the final quality of the precipitation series, the homogeneity of each was checked [Lanzante, 1996; Peterson et al., 1998] against an independent reference series, generated by selecting the five series whose difference series correlated best with the difference series of the observatory to be tested [Peterson and Easterling, 1994]. The standard normal homogeneity test (SNHT) was used to identify inhomogeneities in the precipitation series [Alexandersson, 1986; Alexandersson and Moberg, 1997]. For this purpose the ANCLIM program was used [Štìpánek, 2004]. Four nonhomogeneous series were identified and corrected using the difference between the average of the records before and after the date of the inhomogeneity identified, in line with Alexandersson and Moberg [1997]. Finally, temporal gaps were completed using linear regressions upon the respective reference series.

Figure 1.

Spatial distribution of precipitation observatories.

2.2. Classification of the Iberian Peninsula According to Precipitation Patterns

[9] Previous studies describe a highly diverse spatial precipitation pattern in the Iberian Peninsula: several regions with markedly distinct pluviometric regimes [e.g., Fernández Mills, 1995; Rodríguez-Puebla et al., 1998; Esteban Parra et al., 1998; Serrano et al., 1999]. For this reason, prior to analyzing the impact of El Niño and La Niña on drought patterns, a precipitation-based classification of homogeneous areas was obtained to retain the general temporal and spatial patterns of monthly precipitation evolution. This classification was made by means of principal component analysis (PCA) based on the monthly series.

[10] PCA has been widely used to determine the general temporal and spatial patterns of climatic variables. It allows common features to be identified and specific local characteristics to be determined [Richman, 1986; Sneyers et al., 1989; Jollife, 1990]. The application of PCA to climatic series can be performed in S or T modes [Serrano et al., 1999]. The T mode allows the general spatial patterns of climatic variables to be identified. By contrast, the S mode, the mode applied in this study, is used to obtain the general temporal patterns. The areas represented by each mode (component) can be identified by mapping the factorial loadings.

[11] The number of components was selected in accordance with the criteria of an eigenvalue larger than 1, and the components were rotated (Varimax) to redistribute the final explained variance and to obtain more stable, physically robust patterns [Richman, 1986].

[12] The spatial classification of monthly precipitation patterns in the Iberian Peninsula was carried out using the factorial loading values of each component obtained, grouping the observatories by the maximum loading rule. Each observatory was assigned to the component with the highest loading value. This method has been applied in many climatic classifications elsewhere, including those of Karl and Koscielny [1982], Comrie and Glenn [1998], and Bärring [1988] in the United States, Mexico, and Kenya, respectively.

[13] Following Jones and Hulme [1996], once homogeneous areas had been identified, a regional precipitation series for each area was formed using the weighted averages of monthly precipitation records at each observatory. The weight factor was the ratio of the surface area represented by each observatory to the total area of that region based on Thiessen's polygon method. Similarly, a regional series for the whole of the Iberian Peninsula was created adopting the same procedure.

2.3. Calculation of Drought Index at Different Timescales

[14] Many drought indices have been developed during the twentieth century [Heim, 2002], and many studies have used precipitation as the main variable in determining drought conditions [Guttman, 1998; Keyantash and Dracup, 2002]. For this reason, it seems appropriate to use precipitation data only in calculating the drought index [McKee et al., 1995]. McKee et al. [1993] point out that usable water sources include soil moisture, groundwater, snowpack, streamflow, and reservoir storage, and that the time period from the arrival of precipitation until water is available in each usable source differs considerably. Consequently, the timescale over which precipitation deficits accumulate is highly significant. For example, agricultural droughts are usually of a much shorter timescale than hydrological droughts [McKee et al., 1993]. Thus it is important to use a drought index that can be calculated at different timescales so as to identify different drought types, primarily hydrological or agricultural. As far as we are concerned here, it is also important to determine whether the extremes of the SO affect drought conditions, above all agricultural or hydrological droughts, at different timescales in the Iberian Peninsula.

[15] Among the drought indices developed over recent decades, the standardized precipitation index (SPI) [McKee et al., 1993] is widely considered the most robust, effective index. It boasts a number of advantages over others [Guttman, 1998; Keyantash and Dracup, 2002]. The SPI allows the duration, magnitude and intensity of droughts to be determined [Hayes et al., 1999], and it can be calculated at different timescales [McKee et al., 1995], thereby allowing the quantification of different drought types: hydrological, agricultural and environmental. The SPI calculated at timescales between 3 and 6 months can be considered an agricultural drought index [Hayes et al., 1999] because it enables vegetation and soil moisture conditions to be monitored [Sims et al., 2002; Ji and Peters, 2003]. The SPI calculated at 12 months is considered a hydrological drought index and is useful for monitoring surface water resources (i.e., river flows [Szalai et al., 2000]). In this study, the SPI were calculated at timescales of 1, 3, 6, and 12 months. The SPI of the month n refers to the last month of the interval considered for its calculation. Timescales larger than 12 months were not considered because their usefulness and spatial coherence is debatable [Vicente-Serrano, 2005].

[16] The SPI has recently been used for drought analysis in various areas of the world, including the United States [Hayes et al., 1999], Turkey [Komuscu, 1999], Europe [Lloyd-Hughes and Saunders, 2002b], South Africa [Roualt and Richard, 2003], Hungary [Domonkos, 2003], Italy [Bonaccorso et al., 2003], East Africa [Ntale and Gan, 2003], Greece [Tsakiris and Vangelis, 2004], Korea [Min et al., 2003], and Spain [Vicente-Serrano et al., 2004].

[17] When calculating the SPI at different timescales, the adjustment of precipitation series to a Pearson III distribution, as opposed to other distributions, has been reported as being appropriate, because of its greater suitability for calculating the index [Guttman, 1999; Vicente-Serrano, 2005] and its enhanced adaptability to precipitation series at different timescales [Vicente-Serrano, 2005]. Then, from the precipitation computed at timescales of 1, 3, 6, and 12 months, the SPI can be calculated using the L moment method [Greenwood et al., 1979; Sankarabrusamanian and Srinivasan, 1999]. The complete procedure using the Pearson III distribution and the L moment method are described in detail by Vicente-Serrano [2005]. The drought index was calculated using the precipitation series from the 51 observatories, the regional series of homogeneous regions obtained by PCA, and the series for the whole of the Iberian Peninsula.

2.4. Selection of the Extreme Phases of the SO (El Niño and La Niña)

[18] Various indices have been proposed to quantify the SO and its extreme phases [Hanley et al., 2003]. One of the most widely used, the SO index, is a measure of the intensity of the coupled atmosphere-ocean mode in the South Pacific and is based on the surface pressure gradient between the western and eastern areas of the South Pacific [Rasmusson and Carpenter, 1982]. The SO index is calculated as the normalized difference between the pressure anomalies recorded at the observatories of Tahiti (east Pacific) and Darwin (west Pacific) [Ropelewski and Jones, 1987].

[19] Selecting the El Niño and La Niña years is highly problematic owing to the high seasonal variation in these extreme phases [Trenberth, 1997]. Various criteria have been used to identify El Niño and La Niña episodes [Fu et al., 1986; Kiladis and Van Loon, 1988; Trenberth, 1997]. In this paper, El Niño years defined by Rasmusson and Carpenter [1983] have been used. This list is based on the 5-month running mean values of the SO index remaining below −0.5 standard deviations for 5 months or longer. Ropelewski and Halpert [1996] include a list of the La Niña episodes, which are defined conversely. The episodes are counted as complete years and only include the first year of multiyear events and ignore successive years. This list of years has been widely used following the lead set by Ropelewski and Halpert [1986, 1987, 1996], where the first harmonic of an idealized 24-month SO cycle was fitted to composites of monthly precipitation anomalies. Fraedrich et al. [1992], Kahya and Dracup [1993], Piechota and Dracup [1996], Vuille [1999], Kahya and Karabörk [2001] and Karabörk and Kahya [2003] have adopted this annual approach when selecting El Niño and La Niña phases, and they have used the list of SO extremes of Ropelewski and Halpert to determine El Niño and La Niña effects in different regions of the world.

[20] According to Ropelewski and Halpert [1996], between 1910 and 2000, El Niño has been recorded for the years 1911, 1914, 1918, 1923, 1925, 1930, 1932, 1939, 1941, 1951, 1953, 1957, 1965, 1969, 1972, 1976, 1982, 1986, and 1991, while La Niña has been recorded for the years 1916, 1924, 1928, 1938, 1950, 1955, 1964, 1970, 1973, 1975, and 1988. In the present article, the most recent events have also been included. Thus the years 1993 and 1997 have been classified as El Niño years, and 1998 and 2000 have been classified as La Niña years.

2.5. Calculation of Differences in Drought Conditions in El Niño and La Niña Years

[21] An empirical methodology, similar to that adopted by Ropelewski and Halpert [1986, 1987, 1996], Piechota and Dracup [1996], and Karabörk and Kahya [2003] (with some modifications), was used to determine the impact of El Niño and La Niña years on drought conditions in the Iberian Peninsula. Average SPI anomalies at different timescales were calculated for El Niño and La Niña years and the year after these events.

[22] Piechota and Dracup [1996] used a bootstrap resampling procedure to determine significant anomalies (i.e., departures from the mean) in the Palmer drought severity index in the United States in relation to El Niño and La Niña years. Here, in order to determine whether the SPI at different timescales represented significant humid or dry conditions during El Niño or La Niña phases, the Wilcoxon Mann-Whitney test was used with this same parameter [Siegel and Castelan, 1988]. The Wilcoxon Mann-Whitney test is based on ranks that do not require normally distributed samples and is slightly less powerful than parametric tests such as the t test [Helsel and Hirsch, 1992]. Although the SPI is a normalized variable for the entire series, the normality of the sample cannot be guaranteed and the distribution may be biased when a sample of years is extracted (i.e., the SPI values during El Niño and La Niña years). For this reason, the use of a nonparametric test is more convenient.

[23] The sample of SPI values in each of the months of the El Niño/La Niña years were compared with the values of SPI for the months of the normal years and with the opposite sign (i.e., to determine the role of the El Niño years the SPI values during La Niña years were added to the SPI values during normal years and vice versa). The significance level was established at α < 0.05. When significant anomalies were identified, we obtained the significance between El Niño/La Niña years and (1) normal years and (2) the opposite event separately.

[24] The consistency of the response of the monthly SPIs to extreme events of El Niño and La Niña was also studied to determine whether the magnitude of anomalies varies significantly between events. This was addressed by calculating the percentage of consistent signals, defined as the percentage of events having values (SPI values for each particular El Niño/La Niña event) with the same sign as the average (mean SPI values for all El Niño/La Niña events). When 70% of the events were consistent in sign with the mean, a coherence was considered to exist between events, guaranteeing that the influence of SO extremes was not the result of a few major events, but rather the reflection of a fairly stable and quantitatively similar result for all the El Niño/La Niña years.

3. Results

3.1. Classification of Precipitation Patterns in the Iberian Peninsula Based on Monthly Series

[25] Table 1 shows the results of the PCA of the monthly precipitation series. We obtained six components, which explained 76.2% of the total variance. Figure 2 indicates the spatial representativeness of these components. The homogeneity in the temporal variability of precipitation between 1910 and 2000 shows a clear and consistent spatial pattern for the regions of the Iberian Peninsula. Component 1 represents the southwestern areas (southwestern region). Component 2 represents the central areas of the Peninsula (central region). Component 3 is representative of the northern areas (northern region). Component 4 represents the northwestern spaces (northwestern region), and components 5 and 6 represent the northeastern (northeastern region) and southeastern areas (southeastern region), respectively. The patterns are spatially coherent, since there is no overlapping of the areas represented by each component. These patterns, obtained from continuous monthly precipitation series, agree with the results obtained by other authors for annual [Rodríguez-Puebla et al., 1998] and monthly timescales [Serrano et al., 1999]. The spatial differences in precipitation can be attributed to the complexity of the Peninsula's relief and the diversity of flows and atmospheric patterns that affect the Iberian Peninsula [Zorita et al., 1992; Olcina, 2001].

Figure 2.

Spatial distribution of rotated PC loadings.

Table 1. Results of Principal Component Analysisa
ComponentPercent of Variance Explained by the ComponentPercent of Variance Accumulated
  • a

    Varimax rotation.

132.2832.28
214.1846.46
39.6456.10
47.3063.40
56.5069.89
66.2676.16

[26] Figure 3 shows the monthly precipitation classification of the Iberian Peninsula when applying the maximum loading rule. The southwestern region represents the largest surface area and contains the highest number of observatories (22). The central region contains 10 observatories, the northern region contains 7, the southeastern region contains 5, the northeastern region contains 4, and the northwestern region contains 3 with the least surface area.

Figure 3.

Iberian Peninsula classification according to monthly precipitation series using the maximum loading rule.

[27] The spatial representativeness of the regional precipitation series is coherent with the classified patterns. The correlation analysis between the six regional series and the precipitation series of the different observatories indicated that the regional series represented the monthly precipitation evolution of the homogeneous areas obtained from the PCA classification (Figure 4).

Figure 4.

Correlation between the six regional series and monthly precipitation series in the different observatories. R values larger than 0.1 are statistically significant (α < 0.001).

3.2. Drought Evolution in the Different Regions

[28] Figure 5 shows the evolution of the SPI at the 6-month timescale in the six regions, between 1910 and 2000. The periods of negative SPI values indicate drought conditions, becoming more intense in direct proportion to the negativity of the value. Some of the main dry episodes coincide spatially. This was the case of the severe droughts that affected the Iberian Peninsula during the 1940s and 1950s. These droughts can be identified in all regions, with the exception of the southeastern region. In the 1980s and 1990s, major droughts were also recorded in various areas. However, during these years the droughts did not occur simultaneously in all the areas. For example, in the northern region the most severe drought was recorded at the beginning of the 1990s, while the southwestern and the central regions were affected by more severe droughts during the middle years of this decade. Moreover, the southeastern region recorded the most severe droughts in the final years of this decade. Other drought episodes demonstrated more local characteristics. For example, in the 1920s and 1930s, the northeastern (5) and southeastern (6) regions experienced severe dry conditions, while these drought episodes were shorter in the other regions.

Figure 5.

Evolution of SPI series at timescale of 6 months obtained from the six regional precipitation series. 1, southwestern region; 2, central region; 3, northern region; 4, northwestern region; 5, northeastern region; 6, southeastern region.

3.3. Influence of SO Extremes on Drought Conditions Throughout the Iberian Peninsula

[29] Figure 6 shows the mean values of SPI at different timescales during the various months of El Niño (Figure 6a) and La Niña (Figure 6b) years, as well as in the year after these events. The SPI values were obtained from the regional precipitation series for the entire Iberian Peninsula. Thus Figure 6 indicates the general drought response at different timescales when SO extreme phases occur. At a timescale of 1 month, positive averages of SPI during El Niño years in relation to the rest of the years were found between the months of June and November, but these were not statistically significant in August. During El Niño years, significant humid conditions were recorded during the months of September, October and November at the timescale of 3 months but only considering the comparison with the rest of the years (normal and La Niña). However, at longer timescales (12 months), there were no significant positive differences in the SPI averages during the period commencing November in the El Niño year to August of the following year.

Figure 6.

Mean SPI values at different timescales from regional SPI series in the whole of the Iberian Peninsula as a function of (a) El Niño and (b) La Niña years (boxed) and the year after these events. Black columns indicate significant differences in the SPI values comparing El Niño/La Niña years and the rest of years. Crosses indicate significant differences in the SPI values between El Niño and La Niña years. Circles indicate significant differences in the SPI values between El Niño/La Niña years and the normal years. Arrows indicate months in which the anomalies are consistent among events (more than 70% of the events have the same sign as the mean SPI).

[30] During La Niña years the anomalies in the averages of SPI were more marked than those during El Niño years at all timescales (Figure 6b). At the 3-month timescale, between August of La Niña year and January of the following year, SPI averages were negative. These figures were significant in October, November and December between La Niña and the rest of the years. Moreover, these anomalies were also significant in October and November when El Niño and normal years were considered independently. These anomalies were also consistent because more than 70% of the events presented the same sign as the averages. At the 6-month timescale, the response of SPI to La Niña events was clear, while the dry conditions were intensified during the autumn and the winter of the following year. The differences were also significant in relation to El Niño and normal years in most cases, and the anomalies were consistent between November and January. These significant negative averages were also identified at the 12-month timescale.

[31] Thus stronger influences over the SPI averages were found during La Niña events at different timescales than during its counterpart for the entire Iberian Peninsula. In addition, seasonal anomalies at the 3-, 6-, and 12-month timescales were an evident manifestation of significant dry conditions that appeared during the last quarter of La Niña year and extended to the spring of the following year.

[32] Figure 7 shows the percentage surface area in the Iberian Peninsula in which significant dry conditions were recorded during the El Niño and La Niña years. This was calculated from the SPI series for the whole of the Iberian Peninsula at the 6-month timescale. During El Niño years, significant dry conditions were recorded in area that constituted between 10 and 30% of the Iberian Peninsula between January and May of the El Niño year, both considering differences between El Niño and the rest of the years and also between El Niño and normal years. The surface area with a consistency among events during these months is also important (>20%). A greater surface area, in which average SPI values indicated significant dry conditions, was recorded during the summer and autumn of the year following El Niño, primarily in September. Nevertheless, the percentage of surface is considerably reduced when we analyze the differences between El Niño and the normal years. This indicates that significant dry conditions indicated in Figure 7a are profoundly affected by La Niña behavior. Moreover, the surface area with a consistency between events during these months following the El Niño years is very small.

Figure 7.

Percentage of the Iberian Peninsula, during El Niño/La Niña (boxed), and the following year, with significant negative averages in the 6-month SPI between (a) El Niño/La Niña and the rest of years and (b) El Niño/La Niña and normal years. (c) Percentage of the Iberian Peninsula in which the negative anomalies are consistent among events during El Niño and La Niña years.

[33] During La Niña years a large proportion of the Iberian Peninsula was affected by significant dry conditions when differences between La Niña and the rest of the years are analyzed. More than 20% of the total surface was affected between October of the La Niña year and March of the following year. Moreover, during January of the following year, the SPI averages showed significant dry conditions throughout a high percentage of the area (52%). When the significant dry conditions are analyzed only between La Niña and the normal years the surface area with significant dry anomalies decreases, but the La Niña signal is very consistent, corresponding to large areas (>30%) in which more than 70% of the events have the same sign as the average between January and April of the following year.

3.4. Spatial Differences in the Influence of SO Extremes on Drought Conditions in the Iberian Peninsula

[34] Here the spatial differences presented by the SPI averages at the various timescales during the El Niño and La Niña years are described from the regional series of the six homogeneous regions obtained from PCA. In Figure 8 the average values of SPI according to the different timescales during SO extremes are shown for the southwestern region. During El Niño years there were significant positive average values at the timescale of 1 month in July and September, and negative average values in September of the following year. At the 3-month timescale, these patterns were reinforced, and negative averages were predominant for most of the months in the following year, although only in June were the data statistically significant. At the 6-month timescale, there were negative, although not significant, values during the first months of the El Niño year and positive values were recorded during the final months of this year and during the initial months of the following year. Negative averages were also recorded during the final months of the following year, and these were also significant in September between El Niño and the normal years.

Figure 8.

As in Figure 6, except for southwestern region (1).

[35] The temporal pattern was the opposite during La Niña years and, moreover, there were a higher number of months with significant anomalies. Abnormally dry condition at the 1-month timescale was recorded in October during the event year, and this anomaly was also highlighted when the SPI was analyzed at the 3-month timescale, since between October and December of La Niña years significant negative averages, indicative of dry conditions, were found between El Niño and the rest of the years but also only when considering normal years. At the 6-month timescale, the significant negative anomalies moved to the months of January, February and March of the year following La Niña, indicating significant and consistent dry conditions during the subsequent winter and spring. At the 12-month timescale, positive average values of SPI were obtained during the majority of the months during La Niña years, and negative values during the first months of the following year. However, although the pattern was well defined, the anomalies were not statistically significant.

[36] Figure 9 shows the mean values of SPI during SO extreme years in the central region. In general, the response of the drought index at each of the different timescales was higher than in the southwestern region. A greater number of months were recorded in which the anomalies of the average SPI values (positive and negative) were significant considering the opposite event (El Niño/La Niña) but also the normal years. During El Niño years significant positive averages were recorded in the final months at the timescales of 1, 3 and 6 months. Significant negative averages at the 3-month timescale were recorded in January of the SO extreme year. At the 6-month timescale, there were significant negative average values during the spring of El Niño year. The same pattern was also observed, and more markedly so, at the 12-month timescale, although this is determined mainly by the difference between the SPI values of El Niño and La Niña years, not between El Niño and normal years.

Figure 9.

As in Figure 6, except for central region (2).

[37] The average anomalies were more significant during La Niña years. At the 1-month timescale, between January and November of La Niña years, the mean negative values of SPI were recorded, and although these anomalies were not statistically significant, they explain why, at timescales of 3, 6, and 12 months, significant negative averages were recorded during the winter and spring of the following year. These values were more intense than those observed in the southwestern region. The consistency of these anomalies between events was high and also significant differences were recorded between El Niño and La Niña years and between La Niña and normal years at the timescales of 6 and 12 months.

[38] Figure 10 shows the mean values of SPI at different timescales during SO extremes in the northern region, where no significant anomalies were recorded in relation to El Niño and La Niña events. Figure 11 shows the average values of SPI during SO extreme events in the northwestern region. Corresponding to El Niño years, there were no significant anomalies and the behavior was similar to that in the northern region. However, the influence on drought conditions was clearer during La Niña years, and, although in this area significant negative values were not recorded at the 3-month timescale, as in southwestern and central regions, the negative averages were statistically significant at timescales of 6 and 12 months. However, these anomalies were displaced in time in relation to the southwestern and central regions, and also the number of months with consistent anomalies was lower than in the central region and determined mainly by the differences between La Niña and El Niño years, not between La Niña and normal years.

Figure 10.

As in Figure 6, except for the northern region (3).

Figure 11.

As in Figure 6, except for northwestern region (4).

[39] In the northeastern region, there was a major discordance between the average values of SPI for El Niño and La Niña years (Figure 12). There were significant negative averages of SPI in the months of March, April and May during the year following El Niño at the 3-month timescale. The temporal pattern is different to that recorded in any of the other regions where negative SPI averages were not so marked at this timescale. However, during La Niña years more significant negative and consistent averages were recorded for several months at different timescales: Negative averages of SPI at the 1-month timescale during the majority of the months of La Niña years, negative averages of SPI between July and November of La Niña year at the 6-month timescale, and between September and February of the following year at the 12-month timescale. During the majority of these months the differences in the SPI values of El Niño and La Niña years and between La Niña and normal years were also significant.

Figure 12.

As in Figure 6, except for northeastern region (5).

[40] Finally, the southeastern region recorded the highest number of statistically significant and consistent negative or positive SPI averages during El Niño and La Niña years (Figure 13). Between December of El Niño year and November of the following year, the SPI values at the 1-month timescale were negative, particularly during the months of March, May and August. This caused significant average drought conditions to be recorded during several months at the 3-, 6-, and 12-month timescales. During La Niña years, the behavior of the drought indices was very similar to observations in areas 2 and 5, with significant negative averages being recorded in several months. Further, the mean anomalies were very high, indicative of drought conditions, while the differences between El Niño and La Niña years and between La Niña and normal years were significant in most months and the anomalies were consistent among events.

Figure 13.

As in Figure 6, except for southeastern region (6).

[41] Therefore it can be concluded that the areas in which the drought indices were most affected by SO events were the eastern sectors of the Iberian Peninsula. Moreover, La Niña years produced more significant negative averages of SPI, which also affected wider areas than was the case in El Niño years. However, it was also observed that the significant negative SPI averages can occur in different months depending on the region. This indicates that although the SO extremes affected wide areas, these are not stable in time, and the negative anomalies can be produced in different months within the Iberian Peninsula.

[42] The areas in which significant negative averages were observed changed during the various months of SO extremes. To illustrate this situation, maps of the average SPI values at the 6-month timescale, obtained from the SPI series of the different observatories, are shown between the months of May and December of the year following that of the El Niño episode (this is the period when a high percentage of the surface area of the Iberian Peninsula was affected by significant negative SPI values, Figure 14), and also between September of La Niña year and April of the following year (Figure 15). Figures 14 and 15 include areas with significant differences between the corresponding extreme event of the SO and the rest of the years (solid line), areas with significant differences between El Niño/La Niña and normal years (dashed line) and the observatories in which a consistency among events has been identified.

Figure 14.

Mean values of SPI during different months of the year following the El Niño event (pluses). SPI at the timescale of 6 months. Solid lines isolate areas with significant negative SPI values comparing El Niño and the rest of years. Dashed lines isolate areas with significant negative SPI values between El Niño and normal years. Solid circles indicate observatories in which anomalies are consistent among El Niño events.

Figure 15.

Mean values of SPI during different months of La Niña year (open circles) and during different months of the year following the La Niña event (pluses). SPI at the timescale of 6 months. Solid lines isolate areas with significant negative SPI values comparing La Niña and the rest of years. Dashed lines isolate areas with significant negative SPI values between La Niña and normal years. Solid circles indicate observatories in which anomalies are consistent among La Niña events.

[43] Between May and August of El Niño years, the negative averages of SPI values, indicative of significant dry conditions between El Niño and the rest of the years, were recorded in the east of the Iberian Peninsula and in the Mediterranean coastal lands. The area is smaller when comparing the El Niño and normal years, but the spatial pattern is coherent with previous findings. In September there was an increase in the number of areas with significant negative anomalies in the central region (between El Niño and the rest of the years but also between El Niño and normal years), though this phenomenon was most marked in the southwest. In October, the areas with significant negative averages of SPI were those located in the interior and on the Mediterranean coast of the southern Iberian Peninsula, while the areas with significant negative averages the previous months (Mediterranean coastlands) did not show significant negative averages of SPI. In November and December, only some isolated observatories in the interior presented significant negative averages of SPI. The consistency among events is only recorded in a few observatories, mainly in August and September of the year following El Niño.

[44] In September and October of La Niña years (Figure 15), only certain areas of the southeastern and northeastern regions presented significant negative SPI averages, but in November a high percentage of the eastern half of the Iberian Peninsula presented significant negative averages between La Niña and the rest of the years and also between La Niña and the normal years. On the other hand, between December of La Niña years and February of the following year, a significant change was recorded in the spatial patterns. Some areas of the interior of the Iberian Peninsula and its western regions showed significant negative SPI averages, while in the Mediterranean coastlands the averages were close to 0. During March and April, the significant values were limited to certain areas of the west and even positive SPI values were recorded in the Mediterranean areas. During the period of La Niña influence, there was a displacement of the significant negative SPI values from the east to the west, the interior areas being affected between December and February. Only the northern areas of the Iberian Peninsula were not affected by significant negative averages of SPI, as we also concluded above in the analysis of the regional series. In comparison with the El Niño years, the consistency among events is recorded in a large number of observatories, which indicates that La Niña effects have a coherent behavior. During La Niña years droughts were identified in large areas of the Iberian Peninsula with a high degree of regularity and were more marked than during El Niño. Therefore the negative SPI anomalies during El Niño years are more dominated by a few major events than during La Niña years.

[45] We also observed these spatial changes at the timescales of 3 and 12 months (data not shown). Thus the drought response to the SO extremes in the Iberian Peninsula is particularly complex in terms of time and space.

4. Discussion and Conclusions

[46] The influence of extreme phases of the SO on droughts at different timescales was evaluated by the SPI in the Iberian Peninsula. In general, the results indicate that drought conditions vary significantly in time and space according to the impact of these extreme phases.

[47] Kiladis and Diaz [1989] described the uniform reversal in the sign of anomalies at the global scale during El Niño and La Niña events, respectively. This effect has also been identified in the Iberian Peninsula in terms of drought indices. During the final months of La Niña year, and during the initial months of the following year, wide areas of the Iberian Peninsula show significant negative SPI averages at a range of timescales (3, 6 and 12 months). The significant negative values are also found to vary during the year depending on the timescale. Although it is initially recorded at the 3-month timescale, the influence is more marked at the timescales of 6 and 12 months. However, the influence of La Niña years varies spatially. In general, more significant negative values seem to be recorded during longer periods in the northeastern, southeastern and central regions, while in the northern region no significant influences were found. Nevertheless, the displacement of significant negative averages from east to west during the year was marked and none of the months in which negative anomalies were recorded in these areas coincided. Moreover, the consistency analysis showed that the majority of significant anomalies during La Niña years were not the result of a few events, and that they are stable during most years. During El Niño years, the negative anomalies were lower than they were during La Niña years, but there were also significant negative averages of SPI between May and December of the following year and there were spatial changes during the various months.

[48] The anomalies observed in the average SPI values are statistically consistent and have a great spatial and temporal coherence. In general, although the identification of the impact of SO extremes on drought conditions in the European continent is new, several studies have shown the marked impact of SO extremes on European precipitation. Stockdale et al. [1998] have described an increase in precipitation in southern Europe during the summers of El Niño. Fraedrich and Müller [1992], Landsea et al. [1994] and Fraedrich [1994] have shown that during El Niño years, negative anomalies occur in sea level pressure from Ireland to the Black Sea. This appears to be related to the increase in cyclonic weather types in western Europe [Wilby, 1993; Fraedrich, 1994]. This atmospheric dynamic could explain the general humid conditions found in the Iberian Peninsula during El Niño years. These results agree, in general, with the drought index response reported in this paper for the Iberian Peninsula. Nevertheless, there are major spatial differences. In the northern region, an influence of the SO extremes on drought indices was not identified, whereas in the northeastern, southeastern and central regions such an influence was very clear.

[49] On the subject of the significant negative SPI averages found during the summer and autumn of the year following that of El Niño, Laita and Grimalt [1997] reported an increase in the frequency of anticyclonic days in the western Mediterranean areas during the spring following an El Niño year. The low precipitation associated with this weather type could reinforce drought conditions during subsequent months. Fraedrich [1990, 1994] has reported more anticyclonic days over western and central Europe during the winters of La Niña years. Pozo-Vázquez et al. [2005b] have also shown, for the European area during La Niña events, a statistically significant winter precipitation anomaly pattern resembling that associated with the positive phase of the NAO. This could also explain the general dry conditions found in the Iberian Peninsula at different timescales during La Niña.

[50] The present paper also shows that El Niño effects have a temporal delay, because the main negative and significant averages in SPI values are recorded during the autumn and winter of the year following that of El Niño. This result also agrees with those reported elsewhere, which show that the SO influence on European climate presents a delay of several months [Fraedrich, 1994; Van Oldenborgh et al., 2000; Laita and Grimalt, 1997]. In this paper, we have shown that the droughts at the timescales of 6- or 12-months in the southwestern, central and northwestern regions during La Niña years were caused by the precipitation decrease in summer and more importantly by the decrease in autumn. Nevertheless, in northeastern and southeastern regions, the significant dry conditions at the timescales of 6 and 12 months during La Niña years do not affect the spring but mainly the autumn and winter months. The significant dry anomalies cannot be explained only by anomalies in the precipitation in autumn, but also by negative 1-month SPI values in the majority of months during La Niña year. This indicates the complexity of the effect of SO extremes on droughts in the Iberian Peninsula because these events can affect different areas during different seasons and respond to anomalies of precipitation in different months.

[51] Results have also shown that the central timing of signal seasons tends to shift further by a couple of months as the timescale goes from the shorter to the longer. This might be resulted from SPI calculation procedure because longer timescales generates more smooth fluctuations and, in turn, a larger sequence of anomalies with same sign. It is also important to note that longer seasonal averages as a signal season are more plausible to be representative of slowly progressing atmospheric events (i.e., El Niño). In this study as the timescale increases the timing of the signal season moves to the mature or decaying phase of the event, reflecting more realistic findings for remote locations from Pacific Ocean.

[52] Moreover, a similar drought response to opposite SO extreme phases has not been identified. There are indications that negative anomalies are more intense during La Niña years. Related to this result, the evidence suggests that the influence of El Niño and La Niña phases on the atmospheric circulation at the global scale is not the same. Smith and Ropelewski [1997] indicate that, in several regions in which the SO influence has been identified, the impact is greater for one extreme (El Niño or La Niña) than it is for the other and, in general, these authors indicate that the influence on precipitation at the global scale is higher during La Niña than it is during El Niño. Pozo-Vázquez et al. [2001] identified a relatively strong anomalous pressure pattern over the North Atlantic during La Niña winters, with an anomalous high-pressure band over the central Atlantic that resembles the positive phase of the NAO, with an intensification of the Azores high and drier conditions in southern Europe [Hurrell and van Loon, 1997]. This may explain the greater influence on drought conditions in the Iberian Peninsula, as well as the spatial differences, since in the northern region significant negative averages of SPI are not recorded, and in this area the precipitation is not affected by NAO [Rodríguez-Puebla et al., 1998; Martin-Vide and Fernández, 2001]. Pozo-Vázquez et al. [2001] have also indicated that the influence of El Niño phases on atmospheric circulation in the North Atlantic areas is more diffuse than it is during La Niña periods. These findings also agree with the results shown here, because the influence of El Niño phases on drought conditions in the Iberian Peninsula is not as great: smaller and a greater degree of dispersion in the affected areas and a wider temporal lag. This might occur because different types of El Niño phase exist and these could produce different responses in extratropical atmospheric circulation [Pozo-Vázquez et al., 2001]. Fu et al. [1986] identified three types of El Niño phenomena according to the spatial distribution of the sea surface temperature anomalies in the South Pacific area. This points to the need for future research to explore the use of separately categorized El Niño events so as to determine their independent impact on European climate, as Kahya and Dracup [1994] did in their exploration of type I El Niño events in the southwest of the United States.

[53] The usefulness of the results reported in this paper for drought prediction and management could establish certain doubts regarding the fact that during the twentieth century a certain temporal uncertainty has been identified in the climatic response to the SO [Trenberth and Hoar, 1997]. In Europe, changes in the climatic response to the SO have been identified: in atmospheric pressure anomalies [Hamilton, 1988; Gouirand and Moron, 2003], and in precipitation [Lloyd-Hughes and Saunders, 2002a; Knippertz et al., 2003]. Between 1950 and 2000, the relationship between the SO and southern European precipitation was more robust than in previous decades [Rodó et al., 1997; Moron and Ward, 1998]. This might be due to the higher frequency of extreme SO phases recorded during the final decades of the twentieth century [Trenberth, 1997]. Nevertheless, although this temporal uncertainty has been observed, the influence of the SO on the climate of the Iberian Peninsula is not a recent phenomenon and the effects have been observed on wider timescales. Muñoz et al. [2002] described a connection between the sedimentary phases of the Pliocene in the Iberian Peninsula and the SO cycles. The historical influence of the SO on the Mediterranean areas has also been recorded in other studies. Piervitali and Colacino [2001] have shown from the study of historical documents in Italy that during the last five centuries many drought events have been linked to SO activity. De Putter et al. [1998] have also provided evidence of a historical influence of the SO on Nile discharges.

[54] Finally, it should be stressed that the development of drought monitoring and early warning systems is crucial for agricultural and hydrological management [Svoboda et al., 2002]. Here drought indices are an important tool [Heim, 2002] and drought prediction is a fundamental objective [Gordon, 1993]. Forecasts of the SO extremes are routinely provided and distributed today, but the limits of El Niño predictability are still the subject of debate [Philander and Fedorov, 2003; Palmer et al., 2004], presenting considerable difficulties for prediction [Latif et al., 1998]. However, major efforts and advances are being made in this field. Tang et al. [2004] has used two hybrid coupled models to examine the predictability of El Niño, obtaining a significant improvement in the predictive skills of the El Niño sea surface temperature anomaly. Similarly, Chen et al. [2004] have presented a retrospective forecast of the interannual climate fluctuations in tropical areas of the Pacific Ocean for the period 1857 to 2003, using a coupled ocean-atmosphere model, which successfully predicts all prominent El Niño events within that period at a lead time of up to 2 years. This implies that the evolution of El Niño is controlled to a larger degree by self-sustaining internal dynamics than by stochastic forcing. These advances indicate that predicting El Niño is possible. Hence knowing the impact of these events on atmospheric circulation in the European continent is crucial, as is knowing the impact on and spatial differences in climate extremes such as droughts, which in the Mediterranean regions have severe negative effects.

[55] The study reported here should be of great use in the development of early warning systems and in mitigating the effects of drought in an area that is especially prone to such phenomena. Nevertheless, and in spite of these advances, we must be cautious. For drought prediction in the Iberian Peninsula it may also be necessary to take into account the influence of other atmospheric circulation patterns or even the effect of random weather, which can be important and make it difficult to use ENSO as a single drought predictor. For this reason, further research is needed to improve the development of drought early warning systems by means of other atmospheric indicators.

Acknowledgments

[56] The author wants to acknowledge financial support from the following projects: BSO2002-02743 and REN2003-07453 (Financed by Ministerio de Educación, Cultura y Deporte Spain and FEDER) and “Programa de grupos de investigación consolidados” (grupo Clima, Cambio Global y Sistemas Naturales, BOA 147 of 18-12-2002), financed by Aragón Government. Research of the author was supported by postdoctoral fellowships by the Ministerio de Educación, Cultura y Deporte, Spain. The author also wants to express his gratitude to the National Institute of Meteorology (INM) for the readiness of the data used in this work, and also he would like to thank to Santiago Beguería, Juan I. López-Moreno, Ercan Kahya, and the two anonymous reviewers for their helpful comments.

Ancillary