J Ornithol (2013) 154:481–489 DOI 10.1007/s10336-012-0915-2 ORIGINAL ARTICLE Using self-organizing maps to investigate environmental factors regulating colony size and breeding success of the White Stork (Ciconia ciconia) Abdelkrim Si Bachir • Haroun Chenchouni Nawel Djeddou • Christophe Barbraud • Régis Céréghino • Fréderic Santoul • Received: 15 May 2012 / Revised: 19 October 2012 / Accepted: 12 November 2012 / Published online: 13 December 2012 Ó Dt. Ornithologen-Gesellschaft e.V. 2012 Abstract We studied variations in the size of breeding colonies and in breeding performance of White Storks Ciconia ciconia in 2006–2008 in north-east Algeria. Each colony site was characterized using 12 environmental variables describing the physical environment, land-cover categories, and human activities, and by three demographic parameters: the number of breeding pairs, the number of pairs with chicks, and the number of fledged chicks per pair. Generalized linear mixed models and the self-organizing map algorithm (SOM, neural network) were used to investigate effects of biotic, abiotic, and anthropogenic factors on demographic parameters and on their relationships. Numbers of breeding pairs and of pairs with chicks were affected by the same environmental factors, mainly anthropogenic, which differed from those affecting the number of fledged chicks per pair. Numbers of fledged chicks per pair was not affected by colony size or by the Communicated by P. H. Becker. A. Si Bachir  N. Djeddou Department of Natural and Life Sciences, Faculty of Sciences, University of El Hadj Lakhdar, Batna 05000, Algeria H. Chenchouni Department of Natural and Life Sciences, Faculty of Exact Sciences and Natural and Life Sciences, University of Tebessa, 12002 Tebessa, Algeria C. Barbraud (&) Centre d’Etudes Biologiques de Chizé, CNRS UPR, 1934, 79360 Villiers-en-Bois, France e-mail: barbraud@cebc.cnrs.fr R. Céréghino  F. Santoul Laboratoire d’Ecologie Fonctionnelle et Environnement, EcoLab UMR5245, Université Paul Sabatier, Bâtiment 4R3-b2, 118 Route de Narbonne, 31062 Toulouse cedex 9, France number of nests with chicks. The categorization of the environmental variables into natural and anthropogenic, in connection with demographic parameters, was relevant to detect factors explaining variation in colony size and breeding parameters. The SOM proved a relevant tool to help determine actual dynamics in White Stork colonies, and thus to support effective conservation decisions at a regional scale. Keywords White Stork  Ciconia ciconia  Algeria  Breeding performance  Colony site  Conservation Zusammenfassung Welche Umweltfaktoren regulieren Koloniegröße und Bruterfolg beim Weißstorch Ciconia ciconia?—der Einsatz von Selbstorganisierenden Karten Untersucht wurden Unterschiede in Brutkoloniestärke und Bruterfolg bei Weißstörchen Ciconia ciconia in Nordostalgerien, in den Jahren von 2006–2008. Jede Kolonie wurde anhand von zwölf Umweltvariablen charakterisiert, welche Auskunft über physische Umweltbedingungen, Landbedeckung und menschlichen Einfluss gaben, sowie anhand von drei demografischen Parametern: der Anzahl der Brutpaare, der Anzahl von Paaren mit Küken und der Anzahl flügger Junge pro Paar. Es wurden Generalisierte Lineare Gemischte Modelle und der Selbstorganisierende Karten-Algorithmus (Self-Organising Map, SOM, ein neuronales Netz) angewendet, um die Wirkung biotischer, abiotischer und anthropogener Faktoren auf die demografischen Parameter und die Beziehungen zwischen diesen zu untersuchen. Die Anzahl der Brutpaare und die der Paare mit Küken wurden von denselben (hauptsächlich anthropogenen) Umweltfaktoren beeinflusst. Dagegen wurde die 123 482 Anzahl flügger Küken pro Paar von anderen Faktoren bestimmt: Weder die Koloniegröße noch die Anzahl von Nestern mit Küken hatten hier einen Einfluss. Die Einteilung der Umweltvariablen in natürliche und anthropogen bedingte half in Verbindung mit demografischen Parametern dabei, die Faktoren zu identifizieren, welche die Variation in Koloniegröße und den Brutparametern erklären. Die SOMMethode erwies sich als geeignetes Werkzeug zur Beschreibung der tatsächlichen Dynamik in Weißstorch-Kolonien und stellt somit eine Hilfe bei der Festlegung effektiver Schutzmaßnahmen auf regionaler Ebene dar. Introduction Many studies have investigated biological and socio-economic factors affecting threatened species in order to understand the current biodiversity crisis (Scott et al. 1995). In birds, studies on habitat selection are of growing importance for conservation policy and planning because they deal with quantitative information affecting the dynamics of bird populations (Caughley 1994). Environmental conditions are significant determinants in breeding habitats, and may influence the breeding success of birds and contribute to the variability of breeding populations (Lack 1968). The determination of the factors (natural or anthropogenic) affecting breeding success is the main target of many bird conservation studies, since breeding success can often be more easily managed than other demographic parameters (Pedrini and Sergio 2002; Gil-Sánchez et al. 2004; Manning et al. 2004). Strictly or usually colonial bird species exhibit wide variation in colony size, with the smallest and largest colonies within a species often varying by several orders of magnitude. Numerous studies used natural variation in colony size to measure fitness consequences of breeding with different numbers of conspecifics (van Vessem and Draulans 1986; Møller 1987; Brown and Brown 2001). In addition, environmental factors or ecological situations surrounding breeding sites of birds have critical impacts on breeding success in either direct or indirect manners (Burger and Shisler 1980; Cody 1985; Yorio et al. 1995). In its breeding range, the White Stork (Ciconia ciconia) nests either solitarily or colonially (Cramp and Simmons 1977). Although several studies investigated the factors affecting White Stork colony size and fitness parameters, such as breeding success, independently (van Vessem and Draulans 1986; Carrascal et al. 1993; Barbraud et al. 1999; Moritzi et al. 2001; Jovani and Tella 2004; Tryjanowski et al. 2005a, 2005b; Denac 2006), variations in colony size and fitness may be due to different environmental factors, and few studies have aimed at disentangling these factors. In fact, the effect of one environmental factor 123 J Ornithol (2013) 154:481–489 (e.g., ambient temperature) may become evident, sometimes with lagged effects, only when other environmental factors (e.g., food availability) affecting reproduction are taken into account (Tryjanowski and Sparks 2008). Therefore, the aim of this paper was to investigate the main environmental factors affecting the distribution and size of breeding colonies and breeding parameters of the White Stork in an ecotone area located between the arid and semi-arid bioclimatic stages. Most demographic studies on this species have been conducted in temperate areas, mainly in Europe, where environmental conditions and life history traits (e.g., clutch size, colonial or solitary nesting) differ from those encountered in northern Africa. This species is of high conservation interest and its populations have experienced considerable changes during the last decades (Thomsen and Hötker 2006). Our approach is based on artificial neural network analyses that enhance our ability to determine the actual dynamics in colony sizes and breeding performance, and to investigate how these demographic parameters are associated to broad-type natural as well as anthropogenic information. Methods Study area The study was conducted in the Wilaya (department) of Batna, north-east Algeria, in an area of 12,192 km2 located between 6° and 7°E and 35° and 36°N (Fig. 1). The general climate is typically Mediterranean with a continental influence (semi-arid area with cool winters), but ranges from the arid to the humid category according to Emberger’s (1955) classification. This wilaya is characterized by predominance of high montane vegetation forests including tree species such as Holm Oak (Quercus ilex L.), Atlas Cedar (Cedrus atlantica M.), and Aleppo Pine (Pinus halepensis Miller.). Plains are mostly used for cereal crops (mainly Durum Wheat Triticum durum L. and Barley Hordeum vulgare L.) and livestock. Livestock mainly includes extensive sheep grazing and intensive poultry farming. Survey method Colonies of White Storks were censused during 3 years (2006–2008) from early January to early July. We defined a colony of storks as a breeding site with at least two nests built on the same support or on two supports separated by a few meters (\10 m). Based on dates of first hatched eggs (2006: March 10; 2007: March 24; 2008: March 20), colonies censuses were carried out from May 20 to June 8 2006, from June 2 to July 2 2007 and from June 14 to 30 2008, in order to ensure that when the counts were made no chick had already fledged. These dates were dictated J Ornithol (2013) 154:481–489 483 Fig. 1 The wilaya (department) of Batna, Algeria, with locations of the 62 colonies and densities of breeding pairs of White Storks (Ciconia ciconia) per colony (census 2008). Numbers refer to colony codes according to arrival dates of White Storks on one hand and dates of the first hatched chick on the other hand. During the survey period, the color of chicks’ beak and legs was still black, thereby facilitating their detection in the nest without confusion with the parents. According to Schüz (1936), Arnhem (1980), and Whitfield and Walker (1999), at 6 weeks the black feathers on the wings appear, and in the seventh week standing is regular and the chicks perform wing beats that prepares the muscles to fly. After the ninth or tenth week, young chicks perform their first flight. Censuses were carried out in all administrative units of the wilaya of Batna with assistance from the local staff of the Forests Conservation Direction, who directed us towards nesting sites of storks. Chick counts were often made during the morning when chicks are fed more frequently and are upright in the nest facilitating their count. The number of chicks was estimated by eye or with binoculars. For each colony, the following parameters were recorded: (1) (2) Site description: name of the site, exact location (longitude, latitude), census dates. The number of breeding pairs (HPa), defined as a male and female holding a nest with or without chicks. (3) (4) The number of pairs with chicks (HPm), defined as a male and female holding a nest with chicks, the presence of chicks indicates the presence of a breeding pair with chicks. The number of fledged chicks per nest (JZG). Fledged chicks were defined as nestlings older than 7 weeks which are still on the nest and about to fledge. At this age, chicks are supposed to survive until fledging since there is no or very slight mortality between the census period and their fledging (Djerdali et al. 2008b). In computation, JZG was entered as the average number of fledged chicks per breeding pair in the colony. Environmental variables For each colony site we collected the following environmental variables: (1) (2) (3) Altitude in meters above sea level measured by an altimeter (±1 m). Annual precipitation (mm). Cumulated precipitation (mm) recorded during November, December, and January (winter rainfall) corresponding to the pre-breeding period. 123 484 (4) (5) (6) (7) (8) (9) (10) (11) (12) J Ornithol (2013) 154:481–489 Cumulated precipitation (mm) recorded during March, April, and May (spring rainfall) corresponding to the early nesting period. Annual mean of minimum temperatures (°C). Annual mean of maximum temperatures (°C). Approximate flight distance between the colony and the closest urban area (in m, ±10 m). Approximate flight distance between the colony and the nearest refuse (rubbish dump) (in km, ±0.1 km). Total useful agricultural surface (UAS in ha). The term ‘useful agricultural surface’ is adopted by official statistics and covers all areas available to all types of agricultural activities (forage, arable or cultivable areas). Irrigated useful agricultural surface (irrigated UAS in hectares). Inhabitant’s density (number of individuals per km2). Type of colony [monospecific for the colonies occupied only by White Storks; plurispecific for the colonies occupied by storks and Cattle Egrets (Ardea ibis)]. Most new colonies of Cattle Egret, which is invasive in Algeria, are built in mixed colonies with or close to the colonies of White Storks (Si Bachir et al. 2011). This has given rise to a multitude of plurispecific colonies of the two species. Environmental variables were extracted in a 1.5-km radius from the center of the colony, which represents the distance covered by the majority of foraging flights (Alonso et al. 1991; Nowakowski 2003). Annual data dealing with the agricultural surfaces (UAS and irrigated) and inhabitant density were provided by the planning and territory management direction of Batna City (DPAT, Direction de Planification et d’Aménagement du Territoire). To this end, data for each municipality in which a colony was established and for each census year were taken into account. Climatic data mainly originated from the meteorological station of Batna. In addition, altitudinal extrapolations of precipitation and temperature were carried out to characterize the climate of some areas (11 colonies out of a total of 61). According to Seltzer (1946), the minimum and maximum temperatures decrease within altitude gradient is respectively 0.3 and 0.7 °C for 100 m, while precipitation increases by 40 mm for each 100 m of altitude. The error in extrapolated values is &10 % (Seltzer 1946). Data analysis and modeling procedure Data from a set of 61 colonies representing a cumulative total of 1,547 nests was analyzed by artificial neural network 123 (ANN) over three consecutive breeding seasons (2006–2008). Specifically, we used the self-organizing map algorithm (SOM, unsupervised neural network) to sort bird colonies according to environmental variables and to highlight relationships between environmental and surveyed breeding parameters (HPa, HPm and JZG). The SOM Toolbox (version 2) for MatlabÒ developed by the Laboratory of Information and Computer Science at the Helsinki University of Technology was used (http://www.cis.hut.fi/ projects/somtoolbox/). With its ability to combine ordination and gradient analysis functions, the SOM is convenient to visualize high-dimensional ecological data in a readily interpretable manner without prior transformation. The SOM algorithm is an unsupervised learning procedure that transforms multi-dimensional input data into a two-dimensional map subject to a topological (neighborhood preserving) constraint (details in Kohonen 2001). The SOM thus plots the similarities of the data by grouping similar data items together using an iterative learning process that was further detailed in Park et al. (2003). The SOM algorithm is specifically relevant for analyzing sets of breeding parameters that vary and covary in a non-linear fashion, and/or that have skewed distributions. Additionally, the SOM algorithm averages the input dataset using weight vectors through the learning process and thus removes noise (Lek and Guégan 2000). Moreover, with the SOM algorithm, there is no effect of outliers on the overall scatter plot distributions, because outliers are displayed in a part of the scatter plot without affecting other parts. A detailed description of the modelling procedure (training, map size selection, number of iterations, map quality measurements) was given in Park et al. (2003). Firstly, we identified relationships between environmental variables and the distribution of colonies. The structure of the SOM for this analysis consisted of two layers of neurons connected by weights (or connection intensities): the input layer was composed of 12 neurons (1 per environmental variable) connected to the 146 samples (44, 43, and 59 colonies in 2006, 2007, and 2008, respectively), and the output layer was composed of 66 neurons (see below) visualized as hexagonal cells organized on an array with 11 rows and 6 columns. With the exception of ‘‘Altitude’’, all variables showed variations over the 3 years. The number of output neurons retained for this study (after testing quantization and topographic errors) fitted the heuristic rule suggested by Vesanto et al. (2000), who reported that the optimal number of map units is close to 5Hn, where n is the number of samples. At the end of the training, each sample is set in a hexagon of the SOM map. Neighboring samples on the grid are expected to represent adjacent clusters of samples. Consequently, samples appearing distant in the modeling space (according to the variables used during the training) represent expected J Ornithol (2013) 154:481–489 485 differences among samples for actual environmental characteristics. Ward’s algorithm was applied to the weight vectors of the neurons in order to divide the SOM units (hexagons) into clusters. Secondly, we aimed to emphasize the relationships between environmental variables and the three breeding parameters recorded for each colony (the number of breeding pairs ‘‘HPa’’, the number of pairs with chicks ‘‘HPm’’, and the number of fledged chicks ‘‘JZG’’). During the aforesaid training, we used a ‘‘mask’’ function to assign a null weight to the breeding parameters, whereas environmental variables were assigned a weight of 1 so that the ordination process was based on the environmental variables only. In addition, setting the mask value to zero for a given component removes the effect of that component on organization (Vesanto et al. 2000; Vesanto and Hollmen 2003; Sirola et al. 2004; Raivio 2006). The values for breeding parameters were thus visualized on the SOM previously trained with environmental variables only. Finally, we tested for factors affecting HPa, HPm, and JZG with generalized linear mixed models using PROC GLIMMIX in SAS (2002). Factors entered in the model were the year for HPa, year and HPa for HPm, and year, HPa, or HPm for JZG. Colony was entered as a random effect. Results Breeding parameters and numbers of nests found each year are shown in Table 1. The number of breeding pairs per colony did not vary significantly between years (F = 0.71, p = 0.49) (Table 2) and was in average 10.6 ± 10.3 (SD). The number of pairs with chicks per colony increased with colony size and varied significantly between years (being higher in 2006 than in other years; Table 2). The mean number of fledged chicks per nest varied between years (being higher in 2006 than in Table 1 Annual variation (±SD) of the breeding parameters of White Storks (Ciconia ciconia) in the area of Batna, Algeria 2006 Breeding pairs per colony (HPa) Pairs with chicks per colony (HPm) Fledged chicks per nest (JZG) 2007 2008 All survey period 10.5 ± 9.6 12.2 ± 11.4 9.7 ± 10.1 10.6 ± 10.3 n = 44 n = 43 n = 58 n = 145 13.4 ± 10.1 10.2 ± 10.5 7.5 ± 8.1 9.2 ± 9.4 n = 15 n = 43 n = 59 n = 117 1.9 ± 0.6 1.6 ± 0.8 1.4 ± 0.7 1.5 ± 0.6 n = 18 n = 42 n = 61 n = 121 Table 2 Testing for factors affecting breeding parameters of White Storks in the area of Batna (Algeria) between 2006 and 2008 Effect F df p Variance SE 0.378 0.133 0.037 0.019 0.013 0.036 0.015 0.032 Number of pairs per colony (v2 =df ¼ 1:00 with a negative binomial distribution and log link) Year 0.71 2, 93.52 0.49 Colony 2 Number of pairs with chicks (v =df ¼ 1:06 with a Poisson distribution and log link) Year 4.09 HPa 576.69 2, 111 0.019 1, 70.45 \0.001 Colony 2 Number of chicks per nest (v =df ¼ 0:52 with a normal distribution and identity link) Year 3.77 2, 84.93 0.027 HPa 0.24 1, 91.07 0.629 Colony 2 Number of chicks per nest (v =df ¼ 0:46 with a normal distribution and identity link) Year 4.45 2, 82.61 0.015 HPm 1.57 1, 88.15 0.213 Colony Test statistics (F and P values) are Type III, and df (numerator, denominator) for fixed effects were estimated using the Satterthwaite approximation. Variance components plus their SE are shown for random effects (Colony). See ‘‘Methods’’ for the definition of variables 123 486 other years), but was not affected by the number of pairs in the colony or the number of pairs with chicks (Table 2). The majority of Stork colonies were distributed in the north-eastern part of the wilaya, particularly on the plains of El Madher, Malel, Belezma and Ain Touta. These areas host the most populated colonies, while in the southern part of the country, the colonies are very sparse and little populated (Fig. 1). After training the SOM with the 12 environmental variables, colonies were separated into five subsets (clusters A–E) according to the quantitative structure of the colonies assemblages (Fig. 2a). Cluster A included colonies built at low altitudes, subjected to low precipitation and high temperature (both maximum and minimum), low inhabitant density, moderate irrigation, and located far from urban areas but close to refuse. These colonies corresponded roughly to those located in the arid and sparsely populated region. Cluster B and C included colonies located in areas with low inhabitant density, relatively close to urban areas and refuse tips, and intermediate altitude, temperature and precipitation. Compared to cluster C, cluster B included colonies located in areas with large irrigated agricultural areas. These areas corresponded to the regions where intense farming uses irrigation. Cluster D included colonies established at high altitude with high precipitation (annual, winter and spring) and low temperature, located close to urban areas, and with high useful agricultural surface. These characteristics are typical of mountainous areas with small villages surrounded by agricultural lands. Cluster E included colonies situated in areas with low useful agricultural surface, far from refuse, but close to the cities with high inhabitant density. These colonies were located in, or near, the largest urban areas and were also occupied by Cattle Egrets (Fig. 2b). When explaining breeding parameters, the ordinate on the SOM revealed a similar gradient for both number of breeding pairs per colony and number of pairs with chicks (Fig. 2c). Compared with the distribution of environmental variables on the SOM, this indicated that high numbers of breeding pairs per colony, as well as high numbers of pairs with chicks, were in the areas located close to urban areas with high population density, far from refuse, and these colonies were also occupied by Cattle Egrets (Fig. 2c). However, the gradient for the mean number of fledged chicks per nest showed a different pattern (Fig. 2c). The mean number of chicks fledged per nest was higher in relatively small monospecific colonies at high altitude, with high precipitation (annual, winter and spring), relatively close to urban areas and refuse, and with low population density (Fig. 2c). 123 J Ornithol (2013) 154:481–489 Discussion Our results suggest that the largest colonies were established close to urban areas and refuse, and consequently in the most human-populated zones. Most of these colonies were also shared with the Cattle Egret. The smallest colonies were situated at high altitudes where precipitation was most important. Although the SOM suggested the classification of breeding colonies into five groups, these could be gathered into two super-sets based on the nature of the environmental factors characterizing the habitats: (1) White Storks breeding close to urban areas where they exploit food resources mainly on refuse (Djerdali et al. 2008a, 2008b) and are not demanding in terms of weather, and (2) pairs breeding in colonies situated in suburban or natural areas where environmental conditions foster the abundance of prey species (mainly Orthoptera and terrestrial Coleoptera). In this case, the Storks feed both in natural habitats and in irrigated farmland (Boukhemza et al. 2006). By simultaneously investigating the effects of the same environmental variables on colony size and breeding parameters, the trained models revealed that the environmental factors positively affecting the number of breeding pairs and the number of pairs with chicks are not necessarily beneficial for the number of fledged chicks per nest. Artificial neural networks suggested that factors explaining the establishment of large colonies were mainly anthropogenic variables (e.g., distance to urban areas, distance to refuse, population density). However, the trained models showed that the pairs breeding at altitude with high precipitation rates and near to slightly anthropogenic habitats were mainly those that fledged more chicks. Our results differ from those obtained on other colonial species, where breeding success has been found to be positively related to colony size (Young 1994; Barbosa et al. 1997; Brunton 1999). Given the large body size of White Storks and the very small number of its nest predators, colony size probably does not play a role against predation. On the other hand, reproductive success in White Storks is known to be affected by climatic conditions (Sasvári and Hegyi 2001; Jovani and Tella 2004), food availability and accessibility (Alonso et al. 1991; Pinowski et al. 1991), proximity of food abundant feeding sites (Barbraud et al. 1999; Tortosa et al. 2002; Nowakowski 2003), and the presence of livestock where the birds have better access to food resources (Tryjanowski et al. 2005c). The additional presence of food emanating from household refuse also enhances breeding success (Tortosa et al. 2002; Aguirre 2006; Djerdali et al. 2008b). In our study, the higher number of fledged chicks per nest in colonies situated in areas with relatively high precipitation may reflect an indirect effect of precipitation on food abundance, as J Ornithol (2013) 154:481–489 487 b Altitude (m) 1510 Minimum temperatures ( °C) 9.28 UAS(ha) 237 a B2 B44 C52 C57 C58 A2 C2 C51 C53 C62 A44 C44 C3 A B38 A23 B23 C23 A35 B35 B10 C10 C37 C35 C46 C60 C61 A1 A30 B1 B30 C1 B16 C16 A42 B20 C42 A6 A33 B6 B33 C6 C47 C33 C38 B65 C59 A5 A36 B5 B36 C5 C36 C B17 C45 A37 C8 B37 A10 C50 B C43 C17 C24 A17 A19 A8 A43 A24 B19 A9 B9 C15 B15 C9 A14 C14 A16 A12 A39 B39 C39 A26 B12 C12 C26 7.66 124 547 6.04 10.5 Annual precipitations (mm) Maximum temperatures ( °C) Distance/Refuses (Km) 624 25.1 97.4 A40 B40 C48 C49 B31 C54 A21 C56 A31 C40 B21 C21 C31 436 21.6 49.6 248 18.2 1.79 Distance / urban area (m) Inhabitants density / Km² 1280 1490 B8 B24 B13 A13 C13 A27 C41 1030 Precipitations DJF(mm) 164 B18 B27 D A20 A22 C20 A32 A34 B22 B32 B34 C22 C32 A3 B3 A25 B25 C25 A11 A29 B11 B29 A38 C11 C29 E A4 B4 C45 A18 A41 B14 B41 C18 C27 A15 B43 C55 111 834 666 58.8 174 49.4 Precipitations MAM(mm) 237 A28 B28 C28 Type of colony 14800 2.14 190 9550 1.57 143 4300 1 A7 B7 C7 C HPm HPa Irrigated UAS(ha) JZG 21.4 18.1 1.96 13.4 11.2 1.54 5.47 4.29 1.13 Fig. 2 a Distribution and clustering of colonies on the self-organizing map (SOM) according to the 12 environmental variables. Codes within each hexagon (e.g., A1, B37 and C62) correspond to colonies (sampling units). Colonies are numbered from A1 to A44 for the 44 colonies listed in 2006; the letter B precedes the number of the 43 colonies of 2007 and the letter C for the 59 colonies of 2008. The geographical location and the average colony size of these 62 colonies are reported in Fig 1. b Gradient analysis of the value for each habitat variable on the trained SOM represented by a shaded scale (light low value, dark high value). Each small map represents one parameter that follows similar patterns can be compared to (or superimposed on) the map representing the distribution of colonies presented in panel (a), thus showing the distribution patterns of the various environmental variables (in shade and gray) within each sub-area of the SOM. c The three demographic parameters that characterize the colonies in shades of gray (HPa number of breeding pairs, HPm number of pairs with chicks, JZG number of fledged chicks per nest). The mean value for each variable was calculated in each output neuron of the SOM previously trained with environmental variables data suggested by other studies on this species (Denac 2006; Nevoux et al. 2008). In North Africa, White Storks benefit directly from the use of agricultural habitats, since the creation of irrigated habitats enables multiple opportunistic species such as Storks and Cattle Egrets to benefit from it (Boukhemza et al. 2006; Si Bachir et al. 2011), contrary to what is observed in Europe (Senra and Alés 1992; Carrascal et al. 1993). 123 488 One may thus hypothesise that the increase in White Stork populations in North Africa has been partly due to an increase in the surface areas of irrigated agriculture (Anonym 2006), although other factors such as wintering climatic conditions may also have played a role (Kanyamibwa et al. 1990; Barbraud et al. 1999; Schaub et al. 2005; Saether et al. 2006). In the region of Batna, the increase in local breeding populations of the White Stork is in agreement with the results of the sixth international census (2004–2005) which highlighted an increase of about 39 % since 1994–1995 over the entire breeding area (NABU 2006). Our study thus supports the idea that identifying the variables which significantly affect the breeding parameters of White Storks should be a research priority for conservationists and environmental policy makers. Acknowledgments We warmly thank Mr. Abdellatif Gasmi, the head office for Forest Conservation of the wilaya of Batna, the staff having participated to this survey, and all people, in particular Mr. Athmane Briki, who kindly helped us in the field. We declare that the experiments of this work comply with the current laws of Algeria. We thank two anonymous reviewers for constructive comments. References Aguirre JI (2006) Factores que afectan a la supervivencia juvenil de la Cigüeña blanca Ciconia ciconia. Doctoral thesis,Complutense University of Madrid Alonso JC, Alonso JA, Carrascal LM (1991) Habitat selection by foraging white storks, Ciconia ciconia, during the breeding season. Can J Zool 69:1957–1962 Anonym (2006) Statistiques environnementales dans les pays méditerranéens. Compendium 2005, Commission Européenne. Office des publications officielles des Communautés Européennes, ISBN 92-79-01539-7 Arnhem R (1980) Nos oiseaux (XX). La Cigogne blanche Ciconia ciconia. Homme Oiseau 18:76–77 Barbosa A, Moreno J, Potti J, Merino S (1997) Breeding group size, nest position and breeding success in the chinstrap penguin. Polar Biol 18:410–414 Barbraud C, Barbraud J-C, Barbraud M (1999) Population dynamics of the white stork Ciconia ciconia in western France. Ibis 141:469–479 Boukhemza M, Boukhemza-Zemmouri N, Voisin JF, Baziz B (2006) Ecologie trophique de la Cigogne blanche (Ciconia ciconia) et du Héron garde-bœufs (Bubulcus ibis) en Kabylie (Algérie). Ecol Mediterr 32:15–28 Brown CR, Brown MB (2001) Avian coloniality: progress and problems. Curr Ornithol 16:1–82 Brunton D (1999) ‘‘Optimal’’ colony size for least inter colony study opposing selective predators. Condor 101:607–615 Burger J, Shisler J (1980) Colony and nest site selection in laughing gulls in response to tidal flooding. Condor 82:251–258 Carrascal LM, Bautista LM, Lázaro E (1993) Geographical variation in the density of the white stork Ciconia ciconia in Spain: influence of habitat structure and climate. Biol Conserv 65:83– 87 Caughley G (1994) Directions in conservation biology. J Anim Ecol 63:215–244 123 J Ornithol (2013) 154:481–489 Cody ML (1985) Habitat selection in birds. Academic Press, New York Cramp S, Simmons KEL (1977) Birds of the western Palearctic, Vol. 1. Oxford University Press, Oxford Denac D (2006) Resource-dependent weather effect in the reproduction of the white stork Ciconia ciconia. Ardea 94:233–240 Djerdali S, Tortosa FS, Doumandji S (2008a) Do white stork (Ciconia ciconia) parents exert control over food distribution when feeding is indirect? Ethol Ecol Evol 20:361–374 Djerdali S, Tortosa FS, Hillstrom L, Doumandji S (2008b) Food supply and external cues limit the clutch size and hatchability in the White Stork Ciconia ciconia. Acta Ornithol 43:145–150 Emberger L (1955) Une classification biogéographique des climats. Rec Trav Lab Bot Geol Zool Fac Sci Montpellier (Ser Bot) 7:3– 43 Gil-Sánchez JM, Moleón M, Otero M, Bautista J (2004) A nine year study of successful breeding in a Bonelli’s eagle population in southeast Spain: a basis for conservation. Biol Conserv 118:685– 694 Jovani R, Tella JL (2004) Age-related environmental sensitivity and weather mediated nestling mortality in white storks Ciconia ciconia. Ecography 27:611–618 Kanyamibwa S, Schierer A, Pradel R, Lebreton J (1990) Changes in adult annual survival rates in a western European population of the white stork Ciconia ciconia. Ibis 132:27–35 Kohonen T (2001) Self-organisation maps, 3rd edn. Springer, Berlin Lack D (1968) Ecological adaptations for breeding in birds. Methuen, London, pp 124–133 Lek S, Guégan JF (2000) Artificial neuronal networks: application to ecology and evolution. Springer, Berlin Manning AD, Lindenmayer DB, Barry SC (2004) The conservation implication of bird reproduction in the agricultural ‘‘matrix’’: a case study of the vulnerable superb parrot of south-eastern Australia. Biol Conserv 120:363–374 Møller AP (1987) Breeding birds in habitat patches: random distribution of species and individuals. J Biogeogr 14:225–236 Moritzi M, Maumary L, Schmid D, Steiner I, Vallotton L, Spaar R, Biber O (2001) Time budget, habitat use and breeding success of white storks Ciconia ciconia under variable foraging conditions during the breeding season in Switzerland. Ardea 89:457–470 NABU (2006) Preliminary results of the 6th international white stork census. International Ornithological Congress, Hamburg Nevoux M, Barbraud J-C, Barbraud C (2008) Breeding experience and demographic response to environmental variability in the white stork. Condor 110:55–62 Nowakowski JJ (2003) Habitat structure and breeding parameters of the White Stork Ciconia ciconia in the Kolno Upland (NE Poland). Acta Ornithol 38:39–46 Park YS, Céréghino R, Campin A, Lek S (2003) Applications of artificial neuronal networks for patterning and predicting aquatic insect species richness in running waters. Ecol Model 160:265–280 Pedrini P, Sergio F (2002) Regional conservation priorities for a large predator: golden eagles (Aquila chrysaetos) in the Alpine range. Biol Conserv 103:163–172 Pinowski J, Pinowska B, de Graaf R, Visser J, Dziurdzik B (1991) Influence of feeding habitat on prey capture rate and diet composition of White Stork Ciconia ciconia (L.). Stud Nat 37:59–85 Raivio K (2006) Analysis of soft handover measurements in 3D network. In: Proceedings of the 9th ACM International Symposium on Modeling Analysis and Simulation of Wireless and Mobile Systems. (Terromolinos, Spain, October 02-06, 2006). ACM, New York, pp 330–337 Saether BE, Grøtan V, Tryjanowski P, Barbraud C, Engen S, Fulin M (2006) Climate and spatio-temporal variation in the population J Ornithol (2013) 154:481–489 dynamics of a long distance migrant, the white stork. J Anim Ecol 75:80–90 SAS (2002) Statistical Analysis System 9.1 for Windows. SAS Institute, Cary Sasvári L, Hegyi Z (2001) Condition-dependent parental effort and reproductive performance in the white stork Ciconia ciconia. Ardea 89:281–291 Schaub M, Kania W, Köppen U (2005) Variaiton in primary production during winter induces synchrony in survival rates in migratory white storks Ciconia ciconia. J Anim Ecol 74:656–666 Schüz E (1936) The white stork as a subject of research. BirdBanding 7:99–107 Scott JM, Tear TH, Mills LS (1995) Socioeconomics and the recovery of endangered species: biological assessment in a political world. Conserv Biol 9:214–216 Seltzer P (1946) Le climat de l’Algérie. Carbonel, Algiers Senra A, Alés EE (1992) The decline of the white stork Ciconia ciconia population of western Andalusia between 1976 and 1988: causes and proposals for conservation. Biol Conserv 61:51–57 Si Bachir A, Ferrah F, Barbraud C, Céréghino R, Santoul F (2011) The recent expansion of an avian invasive species (the Cattle Egret Ardea ibis) in Algeria. J Arid Environ 75:1232–1236 Sirola M, Lampi G, Parviainen J (2004) Using self-organizing map in a computerized decision support system. In: Pal N, Kasabov N, Mudi R, Pal S, Parui S (eds) Neural information processing. Springer, Berlin, pp 136–141 Thomsen K, Hötker H (2006) The sixth International white stork census: 2004–2005. In: Boere GC, Galbraith CA, Stroud DA (eds) Waterbirds around the world. The Stationery Office, Edinburgh, pp 493–495 Tortosa FS, Caballero JM, Reyes-López J (2002) Effect of rubbish dumps on breeding success in the white stork in southern Spain. Waterbirds 25:39–43 489 Tryjanowski P, Sparks TH (2008) The relationships between phonological traits and brood size of the white stork Ciconia ciconia in western Poland. Acta Oecol 33:203–206 Tryjanowski P, Sparks T, Profus P (2005a) Uphill shifts in the distribution of the white stork Ciconia ciconia in southern Poland: the importance of nest quality. Divers Distrib 11:219– 223 Tryjanowski P, Sparks TH, Jakubiec Z, Jerzak L, Kosicki JZ, Kuzniak S, Profus P, Ptaszyk J, Wuczynski A (2005b) The relationship between population means and variance in reproductive success differs between local populations of white stork (Ciconia ciconia). Popul Ecol 47:119–125 Tryjanowski P, Jerzak L, Radkiewicz J (2005c) Effect of water level and livestock on the productivity and numbers of breeding White Storks. Waterbirds 28:378–382 van Vessem J, Draulans D (1986) On the adaptive significance of colonial breeding in the Grey Heron Ardea cinerea: inter- and intracolonial variability in breeding success. Ornis Scand 17: 356–362 Vesanto J, Hollmen J (2003) An automated report generation tool for the data understanding phase. In: Abraham A, Jain L (eds) Innovations in intelligent systems: design, management and applications, studies in fuzziness and soft computing. Chapter 5. Springer, Berlin Vesanto J, Himberg J, Alhoniemi E, Parhankangas J (2000) SOM toolbox for matlab 5. Technical report A57, Neural Networks Research Centre, Helsinki. University of Technology, Helsinki Whitfield Ph, Walker R. (1999) Le grand livre des animaux. Lavoisier, Paris Yorio P, Bertellotti M, Quintana F (1995) Preference for covered nest sites and breeding success in Kelp Gulls (Larus dominicanus). Mar Ornithol 23:121–128 Young E (1994) Skua and penguin: predator and prey. Cambridge University Press, Cambridge 123