Breadth and specialization in microhabitat selection: the case of the Algerian mouse (Mus spretus) in Central Spain

Rocío Ta r j u e l o 1*, Manuel B. Mo r a l e s 1 & Juan Tr a b a 1

Résumé. — Amplitude et spécialisation de la sélection du microhabitat : le cas de Souris à queue courte (Mus spretus) dans le centre de l’Espagne. — La présente étude analyse la sélection de l’habitat et le patron de distribution spatiale à petite échelle de la Souris à queue courte, une espèce typique des milieux méditerranéens sur laquelle il n’existe que très peu d’informations. L’échelle du microhabitat semble con- venable pour étudier l’influence des changements environnementaux sur le compromis entre disponibilité de nourriture et refuge. Le patron de sélection du microhabitat fut analysé dans trois types de paysages (macrohabitats): oliveraie, culture céréalière et pâturage. Trente pièges Sherman furent installés pendant quatre jours dans chaque macrohabitat (120 nuits-pièges par macrohabitat) et des variables en relation avec la structure verticale et horizontale de la végétation furent mesurées. Des analyses ANOVA furent utilisées pour évaluer les différences entre les points de capture et ceux sans capture en relation avec les gradients environnementaux les plus importants. Ces gradients furent définis à l’aide d’une ACP faite à partir des variables originales. Des modèles linéaires généralisés furent utilisés pour construire des modèles prédictifs de la présence de l’espèce dans chaque macrohabitat en relation avec les variables originales. En outre, le patron de distribution spatiale des points de capture fut analysé à l’aide de la fonction K de Ripley. Les points de capture et ceux sans capture montrèrent des différences significatives par rapport aux deux pre- miers axes de l’ACP, qui s’interprètent respectivement comme des gradients de visibilité et de disponibilité de nourriture. Les modèles prédictifs furent différents pour chaque macrohabitat. Les variables définissant une structure de la végétation plus fermée furent plus importantes dans les aires de céréales, tandis que les variables représentant la disponibilité de nourriture furent plus importantes dans les zones d’oliveraie. Le patron de distribution spatiale à petite échelle fut uniforme, ce qui met en évidence le comportement territo- rial de l’espèce. À des plus grandes échelles le patron tendit à être aléatoire. La souris à queue courte semble présenter une plus grande amplitude de niche par rapport à la structure de la végétation que par rapport aux ressources alimentaires, et elle est capable de modifier son patron de sélection de microhabitat en fonction du macrohabitat dans lequel elle se trouve.

Summary. — This study analyses microhabitat selection and spatial distribution in the Algerian Mouse, a typical small Mediterranean mammal that has been the subject of just a few habitat selection stud- ies. Study on a microhabitat scale seems suitable for determining the influence of environmental changes on the trade-off between foraging and shelter. The microhabitat selection pattern was analysed in three different macrohabitats; a cereal crop, an olive grove and a dehesa (grazing woodland). Thirty Sherman traps were set up on four nights within each macrohabitat (120 trap-nights per macrohabitat) and variables relating to vertical and horizontal vegetation structure were measured. ANOVA was used to evaluate dif- ferences between capture points and those where no captures occurred, with respect to the most important environmental gradients, as defined by PCA axes with the original habitat variables. Predictive models of the presence of the Algerian mouse were constructed using the original variables for each macrohabitat by means of generalized linear models. The point spatial pattern was analysed by Ripley’s K function. PCA axes for vegetation structure and food availability showed significant differences between capture points and no-capture points. Predictive models for each macrohabitat differed in their explicative variables. Vari- ables that defined a more closed vegetation structure were more important in the cereal crop whereas food

1 Groupe d’Écologie Terrestre (TEG), Département Interuniversitaire d’Écologie, Université Autonome de Madrid, Faculté des Sciences (Edifice de Biologie). Rue Darwin 2, 28049 Madrid. * Auteur pour correspondance: Rocío Tarjuelo. Depto. de Ecología, Universidad Autónoma de Madrid. Edificio de Biología. C/ Darwin 2, 28049 Madrid. Espagne. E-mail: [email protected]

Rev. Écol. (Terre Vie), vol. 66, 2011.

­– 145 ­– availability variables had an important role in olive groves. The small-scale spatial pattern was uniform and associated with the territorial behaviour of the species. At a higher spatial scale the pattern tends to random- ness. The Algerian Mouse seems to have a broader niche in relation to vegetation structure than to food resources, and is capable of modifying its microhabitat selection patterns as a function of the macrohabitats in which it occurs.

Habitat use relates to the manner in which an animal avails itself of both the physical and biological components of its habitat, whereas habitat selection is the active process by which a species chooses between available resources (Johnson, 1980). The latter may thus be seen as the outcome of an evolutionary compromise that maximizes survival and/or lifetime repro- ductive success (Krebs & Davies, 1993). Habitat selection is a multiscalar process that varies between the macrohabitat, the total area in which an organism carries out its life cycle, and the microhabitat, defined as the structural vegetation characters that the organism perceives (Morris, 1987; Traba et al., 2009). It is therefore necessary to determine on what spatial and temporal scales the animals select their habitat in order to develop adequate models that allow their presence and population dynamics to be inferred from defined environmental variables (Oatway & Morris, 2007). Habitat selection patterns in small mammals have often been investigated, the suggestion being that these are predictable in terms of habitat characteristics (Rosenzweig & Winakur, 1969), although there is disagreement regarding the relative importance of macro- and micro- habitat (Corbalán & Ojeda, 2004; Jorgensen, 2004; Corbalán, 2006; Traba et al., 2009). The microhabitat scale proves more effective for measuring environmental variables of parameters that are directly related to the evolutionary compromise between foraging and shelter (Morales & Traba, 2009). It is also suitable for analysing niche breadth, by establishing whether or not particular selection patterns are maintained between different macrohabitats, thus allowing generalist or specialist behaviour to be identified (Traba et al., 2009). Although almost all small mammal habitat selection studies have involved the communities of deserts and temper- ate woodlands (Simonetti, 1989; Morris, 1996; Gonnet & Ojeda, 1998; Dalmagro & Vieira, 2005; Corbalán, 2006), very few studies have addressed habitat selection in the small mam- mals of Mediterranean regions (Alcántara & Tellería, 1991; Díaz et al., 1993; Torre & Díaz, 2004). Several studies have found that rodent communities are structured in accordance with particular spatial patterns (Amarasekare, 1994; Schooley & Wiens, 2001). Such patterns arise from two principal causes: intra- and interspecific interactions and resource availability. The detection and interpretation of such patterns is affected by the scale of study (Levin, 1992). Point data analysis can reveal the structure of communities and populations as a function of their spatial pattern; whether clumped, uniform or random, and also establishes whether such patterns vary with scale. The present study aims to describe the microhabitat selection patterns of the Algerian Mouse (Mus spretus, Lataste, 1883) and to determine its small-scale spatial distribution. Despite being common and widely distributed in the Iberian Peninsula, this species has been the subject of very few ecological studies in Mediterranean ecosystems (Orsini et al., 1982; Gray et al., 1998; Khidas et al., 2002) other than some works on its reproduction and its use of space within reedbeds (Palomo, 1990; Vargas et al., 1991). The analysis of its microhabi- tat selection patterns is based on the premise that the species makes an active selection on that scale. Its degree of specialization is evaluated by analysing its habitat selection within three different environments, considering two alternative hypotheses. One is that habitat selection on the microhabitat scale does not differ between environments, which would mean that the Algerian Mouse has the narrow niche that characterizes specialist species. The other is that such selection does differ between different environments, which would show that the species perceives two levels of habitat scale and has a realized niche typical of generalist species.

­– 146 ­– MATERIALS AND METHODS

Study area The study area was in Garvín district in Cáceres province (Extremadura, central Spain. 39º43’17”N, 5º20’40”W) (Fig. 1). The region has a Mediterranean climate with an annual mean temperature between 13ºC and 17ºC but experiences a sharp winter with minimum temperatures around - 4ºC (Rivas-Martínez, 1987). Mean annual precipitation is around 500 mm. The natural cover is woodland dominated by Iberian Holm Oaks (Quercus ilex subsp. rotundifolia, Miller, 1754). The traditional land use is for grazing which has led to most woods taking the form of ‘dehesas’, grazing woodlands with widely spaced trees in pastures dominated by annual species. The study took place in three different agrarian environments: an area of extensive cereal crops (3.09 ha); an olive grove (Olea europea, Linnaeus, 1753) (5.87 ha); and a Holm Oak dehesa (5 ha) (Fig. 1). These three environments were considered as macrohabitats in the present study given that each was large enough for the mice to carry out their entire life cycles within them, since Algerian mouse shows a mean home range around 350 m2 (Gray et al., 1998).

Data collection Thirty trapping points were set up in each macrohabitat between February and March 2009. Each comprised a Sherman live trap (20 x 6 x 6 cm) with a wooden cover to avoid low temperatures killing individuals that were caught

Figure 1.— The location of Garvín district in Cáceres province, Extremadura, where trapping occurred. The dotted line rectangle indicates the olive grove, the solid line rectangle the cereal crop and the broken line rectangle the dehesa (Ministerio de Medio Ambiente y Medio Rural y Marino, 2009).

­– 147 ­– at night. The trapping points were 30 m apart, forming a grid of 1.8 ha. Each macrohabitat was sampled over four consecutive nights, giving 120 trap-nights per macrohabitat, results of trapping success being assumed to be comparable among macrohabitats (Gurnell, 2006). Traps were baited at dusk with bread soaked in oil and checked the following morning. Captured individuals were identified to species by their general morphology (Blanco, 1998), their capture location within the grid was noted and then they were liberated. Trapping success was calculated as the total number of captures divided by the total number of trap-nights in each macrohabitat (Yahnke, 2006). All necessary permits for these procedures were obtained. In order to determine microhabitat characteristics, data on horizontal and vertical plant cover and structures (Tab. I) were collected within each grid the day following the last trapping session in each macrohabitat. For each macrohabitat data were collected from a minimum of 15 sampling points, including both capture points and points with no captures (no-capture points). In the olive grove, the 8 successful capture points and an equivalent number of randomly selected no-capture points were sampled. In the cereal crop, the 19 successful capture points and the 11 no-capture points were sampled. In the dehesa, where no Algerian mouse was caught, 15 random sampling points were measured to ensure that the microhabitat variability was collected.

Table I Microhabitat variables measured in the three macrohabitats studied, the coding employed, the macrohabitat in which each variable was measured and the possible ecological significance of each variable. C = cereal crop, O = olive grove, D = dehesa

Variable Code Macrohabitat Ecological significance Stone cover. SC A measure of horizontal spatial heterogeneity C, O, D and shelter availability Plant litter cover. LC A measure of horizontal spatial heterogeneity C, O, D and food availability Extent of bare ground. BGC Weed cover at height 15cm WC15 Weed cover at height 30cm WC30 Measures of horizontal spatial heterogeneity C, O, D No. of contacts with vegetation at NC15 and detectability by predators height 15cm No. of contacts with vegetation at NC30 height 30cm Cereal cover at height 15cm. CC15

Cereal cover at height 30cm. CC30 Measures of horizontal spatial heterogeneity C Maximum cereal height. MCH and detectability by predators Maximum crop weed height. MWH Woody cover at height 15cm. CSHR15 Measures of horizontal spatial heterogeneity Woody cover at height 30cm. CSHR30 O, D and detectability by predators Woody cover above height 30cm. CTR Distance between trap and nearest TREED Measures of horizontal spatial heterogeneity O, D tree and availability of shelter

At each sampling point (both capture and no-capture), three 1 x 1 m quadrats were set up, one centred on the trap and the others 1.5 m away. Percentage ground cover of bare ground, stones and plant litter, and plant cover at various heights, were measured using two 1m-long rods, divided into 2-cm sections, placed crosswise on the ground (Dalmagro & Vieira, 2005; Silva et al., 2005). The crossed rods were used to delimit 100 2-cm sections, which allowed percentage cover to be assigned to each variable according to the number of sections in which each occurred. Vertical vegetation structure was estimated using a rod of 0.5 cm diameter placed vertically in the centre of each 1 x 1m quadrat. The number of times that vegetation touched the rod at heights of 15 cm and 30 cm were recorded. The distance between the trap and the nearest tree was also measured. All variables were estimated independently of each other and hence their sum could exceed 100% (Aebischer et al., 1993). All the variables measured have been well documented as influencing the presence and abundance of small mammals (Dalmagro & Vieira, 2005; Silva et al., 2005; Tabeni et al., 2007; Traba et al., 2009). Mean values for each sampling point were calculated from the data obtained in the three quadrats.

­– 148 ­– Data analysis A principal components analysis (PCA) was carried out on the correlation matrix of the original microhabitat variables. This procedure allows independent factors with maximum explanatory capacity to be obtained, avoiding the colinearity problems detected during the initial exploratory analysis of the original variables. The use of PCA factors allows sites where mice were captured to be located on gradients of ecological significance at the microhabitat level, that provide information on habitat selection on a small scale (Morris, 1996; Morales et al., 2008; Traba et al., 2009). Differences in microhabitat structure between capture and no capture-points, and among macrohabitats, were tested by means of ANOVAs on the first two axis of the PCA, with a posteriori Tukey test when it was necessary. Predictive models of the presence of Algerian Mouse in each macrohabitat were obtained by means of generalized linear models (GLMs), using the original variables as explanatory variables and mouse presence/absence as the response variable, assuming a binomial distribution. Due to the sensitivity of these analyses to the presence of colinearity between explanatory variables (Quinn, 2000), variables included in the model were selected among those with Pearson correlation coefficients greater than 0.7, as well as including all with coefficients less than 0.7 (Tabachnick & Fidell, 1996). Models were selected according to their Akaike index function (AIC) (Burnham & Anderson, 2003). The Akaike index (or AIC, Akaike information criterion) is a measure of the variance explained by a model, such that for a given model its AIC is a function of its maximized log-likelihood (ℓ) and of the number of estimated parameters (K) (Posada & Buckley, 2004), according to the formula: AIC = -2ℓ + 2K Predictive models for the presence of Algerian mouse finally considered were those included within the two lowest values of the AIC (Burnham & Anderson, 2003). Model robustness was evaluated by calculating the receiver operating characteristic (ROC) curves, which describe the relationship between sensitivity (the proportion of positive assignments) and specificity (the proportion of negative assignments) according to variation in the discrimination threshold (Hanley & McNeil, 1982). The area below the ROC curve is a measure of model robustness and varies between 0.5, giving a random classification of the values predicted by the model, and 1, where there is perfect discrimination. Values between 0.5 and 0.7 indicate imprecise models, those between 0.7 and 0.9 indicate models with useful applications and those with values greater than 0.9 show a high level of precision (Manel et al., 2001). All analyses used transformed variables. Those measured as percentages were transformed using an arcsine function and the variable referring to the distance between a capture point and the nearest tree was transformed logarithmically. To avoid missing values in the statistical analyses this latter variable was assigned a value of 10 000 cm for all observations in the cereal crop macrohabitat, this value being of one order of magnitude higher than the highest value measured in the dehesa (1900 cm). Data treatment and analysis employed the SPSS v.15 and Statistica v.8 statistical packages. In order to evaluate the spatial pattern of capture points of the Algerian Mouse, a Ripley’s K function was used (Wiegand & Moloney, 2004). This function allows second order spatial analysis, taking into account the variance in the distance between points where individuals were present. Unlike other methods, this function also allows the detection of changes in capture patterns as a function of spatial scale. Ripley’s L function is a transformation of the K function that is more often used since it eliminates the effect of scale in the case of independent patterns and stabilizes the variance (Ripley, 1981), which assists the interpretation of the results. In order to establish whether the L function obtained differed significantly from the random null model, confidence intervals were established by means of Monte Carlo permutations of the null model. The spatial pattern obtained is random if the function is delimited by the confidence intervals, it is uniform if above these intervals and it is aggregated if it is below them. This analysis made use of the Passage program (Rosenberg, 2009).

RESULTS

Small mammal captures There were 67 captures in total across all the macrohabitats (Tab. II). Two species were trapped: the Algerian Mouse and the Wood Mouse (Apodemus sylvaticus, Linnaeus, 1758). More Algerian mice were captured in the cereal crop (n = 49) than in the olive grove (n = 17), and none were captured in the dehesa. Only one Wood mouse was caught, in the dehesa, and so the presence of that species was not taken into account in the subsequent analyses. In accord- ance with these results, trapping success was greater in the cereal crop (Tab. II).

Habitat selection patterns The first two PCA factors accounted for 72.96% of the variance (Tab. III). The first fac- tor was positively associated with the number of contacts with vegetation at heights of 15 cm (NC15) and 30 cm (NC30), with cereal cover at 15 cm (CC15 and 30 cm (CC30), with maxi- mum cereal height (MCH), with crop weed cover at 15 cm (WC15) and 30 cm (WC30), with maximum crop weed height (MWH) and with the distance to the nearest tree (TREED). This first factor can be taken as a gradient of small-scale vegetation structure. The second factor

­– 149 ­– Table II Captures of Algerian Mouse during sampling sessions in the three macrohabitats. The number of trap/nights (T/N), the number of different traps that caught mice and trapping success in each macrohabitat

T/N Captures Nº of successful traps Trapping success (%) Dehesa 120 0 1 0.83 Olive grove 120 17 8 14.16 Cereal 120 49 19 40.83

Table III Results of the two Principal Component Analysis factors indicating the independent variables of ecological significance that summarize microhabitat variation and the explained variance of each. Asterisks indicate values greater than 0.7. See Table I for variable abbreviations

Variables Factor 1 Factor 2 SC -0.251 -0.511 LC 0.121 0.886* BGC 0.716 -0.428 WC15 0.775* 0.179 WC30 0.775* 0.235 NC15 0.912* 0.055 NC30 0.852* 0.114 CC15 0.948* 0.011 CC30 0.859* 0.081 MCH 0.979* 0.024 MWH 0.971* 0.035 CTR -0.433 0.617* TREED 0.884* -0.044 % Variance 59.71 13.25

was positively associated with cover of woody species above 30 cm (CTR) and with plant litter cover (LC). This second factor can be interpreted as a gradient of food availability, given that dead plant material favours abundant food resources for the Algerian Mouse such as insects, and trees supply both green matter and seeds. Microhabitat characteristics, estimated by the PCA factors differed significantly among macrohabitats (F = 199.698; P < 0.001) (Fig. 2). The cereal crop was significantly different from the olive grove (a posteriori Tukey test; P < 0.001) and the dehesa (P < 0.001), whereas the olive grove and dehesa did not differ significantly from each other (P = 1.000), showing that the two latter macrohabitats had similar small-scale vegetation structure. These three mac- rohabitats represent the two extremes of the gradient of small-scale vegetation structure. The macrohabitats also differed from each other in the second PCA factor (F = 5.870; P = 0.005) but here there were significant differences only between the olive grove and the dehesa (P < 0.003), whereas the cereal crop did not differ significantly from the other two macrohabitats. The second factor provides a gradient of food availability that ranges from the highest values found in the dehesa to the lowest ones found in the olive grove. Capture and no-capture points showed significant differences in both the first (F = 5.067; P < 0.029) and the second PCA factor (F = 7.681; P = 0.008), the dehesa being excluded since

­– 150 ­– 2,5

2

1,5

1

0,5 2

or 0 act

F -0,5

-1

-1,5

-2

-2,5 -1,5-1- 0,500,511,52

Factor 1 Figure 2. — Centroids and standard deviations of the two PCA factors for each of the three macrohabitats studied, capture points and total control points, and capture points and control points in the olive grove and cereal field respectively. Control points are in white, capture points in black. Squares indicate total points, circles points in the olive grove and triangles points in the cereal crop. The first factor represents small-scale vegetation structure and the second may be interpreted as a gradient of food availability. Ellipses show the location of corresponding points in the different macrohabitats: solid line = dehesa, dotted line = olive grove, broken line = cereal crop.

no Algerian mice were caught there. Figure 2 shows that Algerian mouse capture sites gave higher values for both factors, indicating that they selected areas with more closed vegetation and greater food abundance. However, the range of microhabitat selection as a function of food resources is narrower than that of vegetation structure on a small scale. Capture points differed significantly from no-capture points in both the first and second PCA factors in the cereal crop (F = 20.275; P < 0.001; F = 9.849; P = 0.004), but not in the olive grove for either factor (F = 0.071; P = 0.793; F = 2.396; P = 0.144).

Predictive models The predictive models for the presence of Algerian Mouse with the originally selected microhabitat variables encompass both those associated with vegetation structure and with availability of food resources. All the selected models for the olive grove included the bare ground variable, which had a negative effect on Algerian Mouse presence, and the stones cover variable figured in two of the three selected models (Tab. IV). The plant litter and woody veg- etation above 30 cm variables also figured. The precision of the three models selected for the olive grove, as evaluated from the area below the ROC curve, was high, showing elevated pre- dictive capacity. The models selected for the cereal crop included the variables for the number of contacts with vegetation at heights of 15 cm and 30 cm (Tab. V) and were also robust, giving areas below the ROC curve above 0.7.

­– 151 ­– Table IV Selected predictive models for Algerian mouse presence in the olive grove, ranked according to Akaike index (AIC) values. Degrees of freedom (d.f.), significance level of model and the area under the ROC curve (AUC) are shown

Model Variables and coefficients d.f. AIC P AUC 1 0.08 - 12.02BGC + 28.34SC + 7.84LC 3 20.65 0.02296 0.813 2 2.54 - 14.56BGC + 21.55SC 2 20.76 0.02448 0.813 3 2.44 - 10.26BGC + 4.32CTR 2 21.31 0.03228 0.813

Ta b l e V Selected predictive models for Algerian Mouse presence in the cereal crop, ranked according to Akaike index (AIC) values. Degrees of freedom (d.f.), significance level of model and the area under the ROC curve (AUC) are shown

Model Variables and coefficients d.f. AIC P AUC 1 -3.82 + 1.38NC15 + 1.35NC30 2 30.12 0.00047 0.804 2 -9.98 + 0.14BGC + 1.62 NC15 + 1.37 NC30 3 30.97 0.00091 0.785 3 -3.85 + 1.23NC15 + 1.08 NC30 + 0.18WC30 3 31.52 0.00118 0.758 4 -1.93 + 1.51NC30 + 0.05 WC15 2 31.93 0.00117 0.739

Spatial distribution pattern The analysis of the pattern of capture points showed a uniform spatial pattern both within the cereal crop and the olive grove (Fig. 3), which may be due to small-scale territorial behav- iour by the Algerian Mouse. A tendency towards random spacing as the spatial scale increases was detected.

DISCUSSION

Habitat selection patterns In spite of the low sampling effort, the present study seems to demonstrate active micro- habitat selection by the Algerian Mouse in the Mediterranean environments where it abounds. Differences between microhabitats are taken to be perceived by organisms when some of those available are used more frequently than others (Simonetti, 1989). In this study, the microhabitat differences between capture points and no-capture points indicate that the mouse is capable of perceiving and selecting between different microhabitat types. Furthermore, Algerian Mouse seems to be able to detect not just small-scale differences but also those at the macrohabitat scale, being a generalist species that is capable of altering its microhabitat selection pattern as a function of the macrohabitat in which it is found. Within the cereal crop, the search for areas with greater herbaceous cover is imposed by the need to hide from possible predators. Various studies have found that small mammals prefer areas with more enclosed low level vegetation that reduces detectability by predators, as happens within the shrubby vegetation at field edges (Benton et al., 2003; Hole et al., 2005). In the olive grove mouse captures occurred in areas that offer high protection, being close to trees and small shrubs where the vegetation structure is more enclosed. Tree cover offers dif- ferent forms of shelter from those available in treeless crops, since the mice can have their runs within the trunks as well as near the trees (pers. obs.). This selection pattern seems to follow the need to reduce the risks of detection and predation but the predator avoidance strategies chosen are shaped by the microhabitats present in each macrohabitat, as seen with other rodent species (Korpimäki et al., 1996). In accordance with the hypotheses Algerian Mouse should be regarded as a generalist for vegetation structure.

­– 152 ­– Figure 3. — Analysis of Ripley’s L function for the spatial pattern of points within the cereal crop and olive grove. The solid line indicates the observed function and the broken lines indicate 95% confidence intervals. The spatial pattern is random when the function falls within the confidence intervals, it is uniform if it occurs above them and it is clumped if occurs below them.

Active microhabitat selection with respect to food supply also occurs but, unlike with veg- etation structure, the pattern is the same in both macrohabitats. In the cereal crop dead organic matter derived from cereal and crop weed fragments and seeds that were uniformly distributed throughout the study area. However, in the olive grove the food comprises dead plant matter, fruits and seeds originating from trees and small shrubs and hence is concentrated in small patches. Although the Algerian Mouse has been considered to be a generalist (Blanco, 1998), these results show a food resource selection strategy more associated with specialist species. No Algerian mice were caught in the dehesa even though its small-scale vegetation struc- ture is very similar to the olive grove. Its absence from this macrohabitat may be due to a lack of its food requirements, given that it has a narrower niche in this respect. There was never- theless a degree of overlap in the food availability gradient between the olive grove and the dehesa. The absence of the Algerian Mouse may thus be due to other biotic factors, like niche segregation for some type of available resource sharing with other small mammals as occurs with other rodents (Dalmagro & Vieira, 2005; Tabeni et al., 2007). Various studies have shown that the density of Algerian Mouse populations is inversely proportional to those of the Wood Mouse (de Alba et al., 2001; Khidas et al., 2002; Torre & Díaz, 2004; Pons & Pausas, 2007), although no solid evidence of the presence of the latter species was found in this case.

Predictive models In accordance with the microhabitat selection pattern described above, the selected mod- els predicted Algerian Mouse presence as a function of variables relating to the protection offered by vegetation structure and food availability, both type of variables being related to

­– 153 ­– survival chances and reproductive success (Corbalán et al., 2006). The models nonetheless differed among macrohabitats. The presence of the mouse in the olive grove is defined as much by the availability of shelter as of food resources. In this macrohabitat, the mouse avoids areas with bare ground where the risk of predation is higher. Its strategy focuses on choosing places covered by stones, near trees and patches of shrubby vegetation to conceal it. Furthermore, mouse chooses areas with greater leaf litter cover that will supply food resources. However, food resources may not play such a decisive role in the cereal due to a greater amount and more evenly distribution of crops weeds and cereal plants. Thus, mouse presence could be predict- able in terms of vegetation structure that offers shelter, as shown by variables that recorded the number of contacts.

Spatial distribution pattern The results showed differences in mouse captures between the cereal crop and olive grove, taking into account the same trapping effort in both macrohabitats. This may be related to differences in the number of high quality microhabitats that each macrohabitat offers, that will increase reproductive success and individual survival (Morris, 2003). The higher number of mouse captures in the crop may be due to greater food availability than in the olive grove, in the form of large quantities of green matter and various arthropods. The uniform spacing pattern found in both macrohabitats may reflect territorial behaviour between individuals for the access to high quality microhabitats. Nevertheless, the observed mouse spatial distribution seems to be scale-dependent. There is a tendency towards random distribution with increased distance between sample points, when territoriality ceases to act, since the intensity of biotic interactions may vary with spatial scale (Diggle, 2003). The spacing of other territorial species also shows this mixed pattern, with uniform small-scale distributions that become clumped or random when seen on a larger scale (Schooley & Wiens, 2001; Cornulier & Bretagnolle, 2006). These results show that the Algerian Mouse seems to be capable of perceiving its environ- ment both on a small scale (that of vegetation structure) and on a macrohabitat scale, modify- ing its selection pattern as a function of the macrohabitat in which it finds itself. Its behaviour and morphology give rise to a selection pattern that allows it to adjust the compromise between obtaining shelter and food according to environmental characteristics. This species shows a wider niche breadth for vegetation structure than for food resources and adjusts the microhabi- tat selection strategy in each macrohabitat, showing it to be a fine-grained opportunistic species (sensu Morris, 1984).

ACKNOWLEDGMENTS

We are grateful to A. Mostajo, R. Mostajo and D. Tarjuelo for their help with fieldwork and to F. García González for helpful advice on identification. We are also grateful to the Dirección General del Medio Natural de la Junta de Extremadura for granting the scientific trapping permits, and to the owners of the properties on which sampling took place. A. Planillo, P. Delgado and C. Pérez provided valuable advice on various aspects of the study. The work was included as a REMEDINAL investigation (S-2009/AMB1783) cofinanced by the CAM and the European Social Fund.

References

Ae b i s c h e r , N., Ro b e rt s o n , P. & Ke n wa r d , R. (1993). — Compositional analysis of habitat use from animal radio- tracking data. Ecology, 74: 1313-1325. Al c á nta r a , M. & Te l l e r í a , J.L. (1991). — Habitat selection of the Wood mouse (Apodemus sylvaticus) in cereal steppes of Central Spain. Z. Säugetierkunde, 56: 347-351. Am a r a s e k a r e , P. (1994). — Spatial population structure in the banner-tailed kangaroo rat, Dipodomys spectabilis. Oecologia, 100: 166-176. Be nt o n , G., Vi c k e r y , J.A. & Wi l s o n , J.D. (2003). — Farmland biodiversity: is habitat heterogeneity the key? TREE, 18: 182-188. Bl a n c o , J. C. (1998). — Mamíferos de España I. Editorial Planeta, Barcelona.

­– 154 ­– Bu r n h a m , K.P. & An d e r s o n , D.R. (2003). — Model selection and inference: A practical information-theoretic approach. 2nd ed. Springer-Verlag, New York. Co r b a l á n , V. (2006). — Microhabitat selection by murid rodents in the Monte Desert of Argentina. J. Arid. Environ., 65: 102-110. Co r b a l á n , V. & Oj e d a , R. (2004). — Spatial and temporal organization of small mammal communities in the Monte Desert, Argentina. Mammalia, 68: 5-14. Co r b a l á n , V., Ta b e n i , S. & Oj e d a , R.A. (2006). — Assessment of habitat quality for four small mammal species of the Monte Desert, Argentina. Mamm. Biol., 71: 227-237. Co r n u l i e r , T. & Br e ta gn o l l e , V. (2006). — Assessing the influence of environmental heterogeneity on bird spacing patterns: a case study with two raptors. Ecography, 29: 240-250. Da l m a g r o , A.D. & Vieira, E.M. (2005). — Patterns of habitat utilization of small rodents in an area of Araucaria forest in Southern Brazil. Austral Ecol., 30: 353-362. De Al b a , J.M., Ca r b o n e l l , R., Al o n s o , C.L., Ga r c í a , F.J., Dí a z , M., Sa nt o s , T. & Te l l e r í a , J.L. (2001). — Distri- bución invernal de los micromamíferos en bosques fragmentados de llanura del centro de España. Galemys, 13: 63-78. Dí a z , M., Go n z á l e z , E., Mu ñ o z -Pu l i d o , R. & Na v e s o , M.A. (1993). — Effects of food abundance and habitat struc- ture on seed-eating rodents in Spain wintering in man-made habitats. Z. Säugetierkunde, 58: 302-311. nd Di gg l e , P.J. (2003). — Statistical analysis of spatial point patterns. 2 ed. Arnold, London. Go nn e t , J.M. & Oj e d a , R.A. (1998). — Habitat use by small mammals in the arid Andean foothills of the Monte Desert of Mendoza, Argentina. J. Arid Environ., 38: 349-357. Gr a y , S.J., Hu r s t , J.L., St i d w o rt h y , R., Sm i t h , J., Pr e s t o n , R. & Ma c d o u g a l l , R. (1998). — Microhabitat and spatial dispersion of the grassland mouse (Mus spretus Lataste). J. Zool., 246: 299-308. Gu r n e l l , J. (2006). — Live trapping small mammals: a practical guide. The Mammal Society, London. Ha n l e y , J.A., & McNe i l , B.J., (1982). — Maximum attainable discrimination and the utilization of radiologic exami- nations. J. Chron. Dis., 35: 601-611. Ho l e , D.G., Pe r k i n s , A.J., Wi l s o n , J.D., Al e x a n d e r , I.H., Gr i c e , P.V. & Eva n s , A.D. (2005). — Does organic farm- ing benefit biodiversity?Biol. Conserv., 122: 113-130. Jo h n s o n , D.H. (1980). — The comparison of usage and availability measurements for evaluating resource preference. Ecology, 61: 65-71. Jo r g e n s e n , E.E. (2004). — Small mammal use of microhabitat reviewed. J. Mammal., 85: 531-539. Kh i d a s , K., Kh a mm e s , N., Kh e l l o u f i , S., Le k , S. & Au l a gn i e r , S. (2002). — Abundance of the wood mouse Apo- demus sylvaticus and the Algerian mouse Mus spretus (Rodentia, Muridae) in different habitats of Northern Algeria. Mamm. Biol., 67: 34-41. Ko r p i m ä k i , E., Ko i v u n e n , V. & Ha kk a r a i n e n , H. (1996). — Microhabitat use and behavior of voles under weasel and raptor predation risk: predator facilitation? Behavioral Ecology, 7: 30-34. Kr e b s , J.R. & Da v i e s , N.B. (1993). — An introduction to behavioural ecology. Blackwell Scientific, Oxford. Le v i n , S.A. (1992). — The problem of pattern and scale in ecology. Ecology, 73: 1943-1967. Ma n e l , S., Wi l l i a m s , C.H. & Or m e r o d , S.J. (2001). — Evaluating presence-absence models in ecology: the need to account for prevalence. J. Appl. Ecol., 38: 921-931. Ministerio d e m e d i o a m b i e nt e , y m e d i o r u r a l y m a r i n o (2009). — Visor SIGPAC. http://sigpac.mapa.es/fega/ visor/ Acceso el 20/5/2009. Mo r a l e s , M.B. & Tr a b a , J. (2009). — Compromisos adaptativos en la selección de hábitat de aves esteparias. In press in: H. Dopazo & A. Navarro (eds). Adaptación y evolución. 150 años después del Origen de las Especies, SESBE, Madrid. Mo r a l e s , M.B., Tr a b a , J., Ca r r i l e s , E., De l g a d o , M.P. & Ga r c í a d e l a Mo r e n a , E.L. (2008). — Sexual dif- ferences in microhabitat selection of breeding little bustards Tetrax tetrax: ecological segregation based on vegetation structure. Acta Oecol., 34: 345-353. Mo r r i s , D.W. (1984). — Patterns and scale of habitat use in two temperate-zone of small mammal faunas. Can. J. Zool., 62: 1540-1547. Mo r r i s , D.W. (1987). — Ecological scale and habitat use. Ecology, 68: 362-369. Mo r r i s , D.W. (1996). — Coexistence of specialist and generalist rodents via habitat selection. Ecology, 77: 2352- 2364. Mo r r i s , D.W. (2003). — Toward an ecological synthesis: a case for habitat selection. Oecologia, 136: 1-13. Oa t w a y , M.L. & Mo r r i s , D.W. (2007). — Do animals select habitat at small or large scales? An experiment with meadow voles (Microtus pennsylvanicus). Can. J. Zool., 85: 479-487. Or s i n i , P., Ca s s a i ng , J., Du p l a nt i e r , J.-M. & Cr o s e t , H. (1982). — Premières données sur l’écologie des popula- tions naturelles de souris Mus spretus Lataste et Mus musculus domesticus Rutty, dans le midi de la France. Rev. Ecol. (Terre Vie), 36: 321-336. Pa l o m o , J.L. (1990). — Características de los desplazamientos del ratón moruno, Mus spretus (Lataste, 1883), en cultivos de cañas de azúcar de la provincia de Málaga. Ecología, 4: 185-189. Po n s , J. & Pa u s a s , J.G. (2007). — Rodent acorn selection in a Mediterranean oak landscape. Ecol. Res., 22: 535-541.

­– 155 ­– Po s a d a , D. & Bu c k l e y , T.R. (2004). — Model selection and model averaging in phylogenetics: advantages of Akaike information criterion and bayesian approaches over likelihood ratio tests. Systematic Biol., 53: 793-808. Qu i nn , G.P. (2000). — Experimental design and data analysis for biologists. Cambridge University Press, Cambridge. Ri p l e y , B.D. (1981). — Spatial statistics. Wiley, New York. Ri v a s -Ma rt í n e z , S. (1987). — Memoria del mapa de series de vegetación de España 1:400.000. ICONA, Madrid. Ro s e n b e r g , M.S. (2009). — PASSaGE. Pattern Analysis, Spatial Statistics and Geographic Exegesis. Version 2 (β 2.0.9.24). Ro s e n z w e i g , M.L. & Wi n a k u r , J. (1969). — Population ecology of desert rodent communities: habitats and environ- mental complexity. Ecology, 50: 558-572. Sc h o o l e y , R.L. & Wi e n s , J.A. (2001). — Dispersion of kangaroo rat mounds at multiple scales in New Mexico, USA. Landscape Ecol., 16: 267-277. Si l v a , M., Ha rt l i ng , L. & Opp s , S.B. (2005). — Small mammals in agricultural landscapes of Prince Edward Island (Canada): Effects of habitat characteristics at three different spatial scales. Biol. Conserv., 126: 556-568. Si m o n e tt i , J.A. (1989). — Microhabitat use by small mammals in Central Chile. Oikos, 56: 309-318. Ta b a c h n i c k , B. & Fi d e l l , L. (1996). — Using multivariate statistics. Harper Collins, New York. Ta b e n i , S., Ma s t r a nt o n i o , L. & Oj e d a , R.A. (2007). — Linking small desert mammal distribution to habitat struc- ture in a protected and grazed landscape of the Monte, Argentina. Acta Oecol., 31: 259-269. To r r e , I. & Dí a z , M. (2004). — Small mammal abundance in Mediterranean post-fire habitats: a role for predators? Acta Oecol., 25: 137-143. Tr a b a , J., Ac e b e s , P., Ca mp o s , V. & Gi a nn o n i , S.M. (2009). — Habitat selection by two sympatric rodent species in the Monte desert, Argentina. First data for Eligmodontia moreni and Octomis mimax. J. Arid Environ., 74: 179-185. Va r g a s , J.M., Pa l o m o , J.L. & Pa l mqv i s t , P. (1991). — Reproduction of the Algerian mouse (Mus spretus Lataste, 1883) in the South of the Iberian Peninsula. Bonn. Zool. Beitr., 42: 1-10. Wi e g a n d , T. & Mo l o n e y , K.A. (2004). — Rings, circles, and null-models for point pattern analysis in ecology. Oikos, 104: 209-229. Ya h nk e , C.J. (2006). — Habitat use and natural history of small mammals in the central Paraguayan Chaco. Mastozo- ología Neotropical, 13:103-116.

­– 156 ­–