Acta Herpetologica 9(2): 203-217, 2014 ISSN 1827-9635 (print) © Firenze University Press ISSN 1827-9643 (online) www.fupress.com/ah DOI: 10.13128/Acta_Herpetol-12875 Factors determining Gekkotan (Reptilia, Sauria) distribution in Tunisia (North Africa) Wided Tlili1,*, Aymen Nefla1, Michel Delaugerre2, Ridha Ouni3, SaÏd Nouira1 1 Unité de recherche “Biodiversité et Biologie des populations”, 05/UR/09-10. Département de Biologie, Faculté des Sciences Mathéma- tique, physique et biologique de Tunis (FST), Campus Universitaire 2092, El Manar Tunis, Tunisia. *Corresponding author. E-mail: wideddada@gmail.com 2 Conservatoire du littoral, Résidence Saint Marc, Rue du Juge Falcone, F20200 Bastia, France 3 Tunisian Wildlife Conservation Society. Faculté des Sciences Mathématique, physique et biologiques de Tunis (FST), Campus Universi- taire 2092, El Manar Tunis, Tunisia Submitted on 2013, 5th May; revised on 2014,2nd June; accepted on 2014, 3rd October Editor: Ernesto Filippi Abstract. Tunisian geckos count nine species (1 is insular relict, 1 is endemic, 2 are ubiquitous and 5 are enfeoffed). We aim to determine factors influencing their distributions. Surveys were founded on environmental divisions. Pres- ence/absence data for 113 grids were analyzed using multivariate tools. 18 environmental variables were revealed and clustered into five factors to model species distributions. Established models were further projected on non-explored areas within Tunisian territory. The distribution of continental geckos follows an indirect bidirectional gradient; the South-northward one is physiologically stressful and the North-southward one is biologically stressful. Five biogeo- graphic regions were established showing concordance with climatic and vegetation regionalization. The distribution of non-anthropophilic species is positively correlated to thermal amplitudes gradient. The distribution of anthropo- philic taxa is positively correlated to agricultural land-use. Oasis, sebkhas and chotts are particular landscapes that disturb both distributions. Predicted areas follow the yielded distribution patterns despite some discrepancy for S. sthenodactylus. The niche characterizing shows that land use and altitude increase the probability of occurrence of H. turcicus and T. mauritanica. Alternatively, they decrease the probability of the presence of T. deserti, T. neglecta, T. tri- politanus and S. petrii. Models could also show that the absence of S. sthenodactylus in northern regions is attributed to high altitudes and cereal land-use. As to T. fascicularis, the displacement of the northern limits of its range is most- ly attributed to an improvement of field investigations. Established model of its distribution shows a restricted area of probable occurrence in central Tunisia confirming its endemism. Keywords. Gekkota, Tunisia, biogeography, clustering, UPGMA, ecological factors, principal component analysis, ecological niche modelling. INTRODUCTION Recent overviews of the distribution of some geckos occurring in Tunisia have been established (Nouira, 1997; Joger, 2003; Rufray et al., 2003; Delaugerre et al., 2011; Tlili et al., 2012 a, b). Reported data have marginally extended the known ranges of the two Phyllodactylidae Tarentola fascicularis and Tarentola deserti (Tlili et al., 2012 a). Most importantly, they have reported new record of Stenodactylus sthenodactylus in sub-humid regions (Tlili et al., 2012 b). These expansions may occur either because of variations in ecological factors or because of traits of the natural history of the species (MacAr- thur, 1972; Holt, 2003). Alternatively, it may be due to anthropogenic impact such as involuntary introduction accompanying human transport. Guisan et al. (1999) and Guisan and Zimmermann (2000) demonstrated that envi- ronmental variables usually replace a combination of dif- 204 W. Tlili et alii ferent resources and do not have direct physiological con- sequence for a species performance. To shed light on the factors that shape species’ ranges, scientists need to classify regional faunas into discrete groups (Guisan and Zimermann, 2000). In the case of Tunisia, this task is complicated by its location in the Mediterranean basin. Besides, it is a transition area between the Palearctic and the Afrotropical ecozones and thereby should be treated differently than the center of these two biogeographical regions (Kreft and Jetz, 2010). Moreover, located astride the Atlas domain and the old African continent, Tunisia might harbor species of many origins (Nouira, 1996). In such cases, several authors sug- gested to start by selecting appropriate environmental variables that might explain the observed distribution of species (e.g., Franklin et al., 2000; Guisan and Zimer- mann, 2000; Hirzel et al., 2002; Holt, 2003). Such studies have been already undertaken for several groups of rep- tiles in Tunisia including Lacertidae (Nouira, 1996) and Scincidae (Kalboussi, 2006). Parallel work for gekkotan fauna is lacking (Tlili et al., 2012 a, b) and questions rela- tive to which factor influence the species distribution and how species will be distributed in time and space remain unanswerable (Heikinheimo et al., 2007). For instance, why did S. sthenodactylus (an arid species) extend its dis- tribution only to a particular sub-humid region (Cape Bon Peninsula) among other similar regions in the extreme north of Tunisia (Tlili et al., 2012b)? Besides, why did some juxtaposed areas show variation concern- ing their biodiversity despite their proximal locations and similar abiotic characteristics (Tlili et al., 2012a)? Given the above, this work has several aims: 1) to elaborate a map of biogeographical regions based on gek- kotan species richness; 2) to extract the environmental variables constraining their distributions in Tunisia, 3) to identify areas of probable occurrence and 4) to assess their ecological niche (habitat). MATERIAL AND METHODS Study area Located in the southern shore of the Mediterranean Sea, Tunisia is separated from Europe by the Channel of Sic- ily (140 km). Tunisian landscape is naturally regionalized into units with characteristic climate, landforms, soil and vegeta- tion (Fakhfakh and Laclavère, 1979). The landscape is marked by a general low relief crossed by the Dorsal Mountain Chain (Fig. 1A). Within the Cap Bon and Medjerda lowland regions all the original woodlands and forests have been cleared for agriculture. Khroumirie and Mogod regions are areas in the north-west comprising mountainous Mediterranean forest and maquis. The central steppe region marks the transition zone between the Dorsal Mountain and the desert and harbors many salt lakes including ‘chott El Jrid’. Jeffara and Dhahara regions are areas of subdesert, desert and Sahara landscapes where the stony ergs and the large sand-dunes of the Great Eastern Erg occur. The climate of Tunisia is mainly Mediterranean divided into 5 bioclimatic stages: Humid, Sub-humid, Semi-arid, Arid and Saharan (Emberger, 1950; Tlili et al., 2012b). Sampling and mapping Preliminary sampling took place in spring and summer seasons from 1996 to 2012 as described in Tlili et al. (2012 b). However, concentration of occurrence data was showed around locations of anthropophilic geckos’ populations and differenc- es in sampling efforts have been revealed from north (highly inventoried) to south (less inventoried). Thus, on the basis of distributions maps (Tlili et al., 2012 a, b), 0.3°×0.3° equal-area grids were chosen as a sampling unit to re-sample an equal number of replicates per environmental combination (Fig. 1B). Teams of work followed climatic gradient where each bioclimat- ic stage was surveyed by two observers in spring and summer seasons of 2011 and 2012 (Graham and Hijmans, 2006; Kreft and Jetz, 2010). Observers alternated between day and night attempting to seek for one specimen of each species for each grid (Guisan and Zimmermann, 2000). Maps were composed using QUANTUM-GIS software (Sillero and Tarroso, 2010). Species distributions and regionalization Absence/presence data were converted into a species per site incidence matrix in order to calculate species richness (S.R.) and grids similarities. A matrix of similarity coefficients between grids was established using Jaccard’s index (Jaccard, 1908) suitable for binary data (Real and Vargas, 1996; Kreft and Jetz, 2010): ȷ = c/N (c = number of attributes present in both operational taxonomic units OTUs, N = total number of attrib- utes). The statistical significance of obtained pairs of OTUs was tested using the statistical table of probability values (Real, 1999). Herein, the probabilities associated with Jaccard’s index depend only on the total number of attributes (N) present in either of the two OTUs being compared (N = p + q - c). The 113 grids were clustered according to their species richness. A matrix of dissimilarities was established from Jac- card’s distance (D (j, k) = 1 - ȷ). Obtained distances range from 0 to 1 (0 when both units have the same attributes, 1 when they share no attributes). The Unweighted Pair-Group Average (UPGMA (Sneath and Sokal, 1973)) was used as a linkage rule. The retained number of clusters was based on the number of bioclimatic stages (Guidi, et al., 2009). Biogeographical patterns Environmental variables were automatically selected by following environmental gradients detected from the clusters’ structure (Guisan and Zimmerman, 2000; Hirzel et al., 2002). 205Gekkotan distribution in Tunisia The 113 cells were characterized with variables relating to: 1) relief (the mean value), 2) Climate (Emberger coefficient, Ther- mal amplitudes (°C) and precipitation levels (mm)), 3) lithology (dominant type), 4) soil (dominant type), 5) water plan (pres- ence of sebkhas, Chotts or lakes), 6) vegetation cover (wood- land, degraded forest, steppe, subdesert, and desert), 7) Agricul- ture (cereal, olive, palm) and 7) urbanism (habitations, indus- trial, touristic and road densities). Spatial data were provided by the “Office Tunisien de la Cartographie” as printed maps and/ or numerical data. Taking into account the strong correlations between most of these variables (Appendix 1), a linear transfor- mation of correlated variables into uncorrelated variables was made via a Principal Component Analysis PCA (Kaplunovsky, 2005; Lawley et al., 2011). The extraction of principal compo- nents amounts to a variance maximizing (VARIMAX) rotation of the original variable space (Harrell et al., 1996; Guisan and Zimmermann, 2000; Clark et al., 2003; Kaplunovsky, 2005; Bri- to et al., 2011; Lawley et al., 2011). To confirm the orthogonal character of obtained factors, clusters of items and rotated axes were identified; then, correlations between those (oblique) fac- tors were computed, and that correlation matrix of oblique fac- tors is further factor-analyzed to yield a set of five orthogonal factors that divide the variability in the items into that due to common variance (secondary factors), and unique variance due to the clusters of similar items in the analysis (primary factors) (Appendix 2). Species distribution modeling Modeling of the distribution of the eight gekkotan spe- cies in Tunisia was assessed by means of Maxent 3.3.3 (Phillips et al., 2006, Philips and Dudik, 2008). The model for each of the eight gekkotan species expresses the suitability of each grid cell as a function of its environmental variables by mean of maxi- mum entropy method. We considered variables extracted by factor analysis to produce “features” which constrain the proba- bility distribution. For every cell occurring within Tunisian bor- ders, Maxent produced a value of probability for the presence of a gecko species, ranging from 0 to 1. If p(x) is the raw output for environmental conditions x, the corresponding logistic value is c p(x) / (1 + c p(x)) for a particular value of c (namely, the exponential of the entropy of the raw distribution). For easier interpretation, cumulative probability values were transformed into three main distribution classes: 1) the ‘out of range’ area, 2) the ‘suboptimal’ area, and 3) the optimal area (Anadón et al., 2012). The importance of each environmental factor in explain- ing the observed distribution of geckos was evaluated by the Fig. 1. Map of Tunisia. Geographic location and natural regions (A); sampling units numbers (B) 206 W. Tlili et alii percent of contribution (PC) and Permutation importance. Response curves obtained from Maxent were also used to dis- cuss the niche of Tunisian Gekkota. RESULTS Taxonomic status and species richness Tunisian geckos belong to three gekkotan families: Gekkonidae, Phyllodactylidae and Sphaerodactylidae (Table 1). Figure 2 illustrates the spatial variation of the species richness to latitudinal, climatic and geomor- phological gradients. Beyond the latitude 36°, only two anthropophilic species T. mauritanica and H. turcicus occur (22.2% of the total gecko-diversity). Between lati- tude 35° and 36°, S. sthenodactylus occurs as well as the two anthropophilic species; it occupies 56.6% of the 113 grid cells. Between latitude 34° and 35°, five gekkotan species were found: T. mauritanica, T. fascicularis, H. tur- cicus, T. tripolitanus and S. sthenodactylus. Southward the latitude 34°, only species highly adapted to drastic Saha- ran conditions occurred: S. sthenodactylus, S. petrii, T. deserti, T. neglecta and T. tripolitanus. Considering bioclimatic stages (Fig. 2A), species richness ranges from two to five species. Each stage con- tains only two gekkotan families. Only Euleptes europaea (relict of northern islands) occupies humid regions. Giv- en its insular statute, it was not included in the follow- ing parts of analysis. The two anthropophilic species are widely distributed and occupy all bioclimatic stages. Six of the nine gekkotan species occur in arid environments. According to lithology and vegetation (Fig. 2B), we note that the presence of geckos is heavily dependent on natural shelters availability such as rocky crevices, stones, sandy borrow or tree barks. Most importantly, we noticed that the aggregation of the non-anthropophilic geckos (in central and southern regions) is related to local vari- ations of vegetal landscape. T. fascicularis is usually found in open steppe type in which S. sthenodactylus occupies either the bark of trees or small shrubs. T. tripolitanus is associated with burrows under rocks. In ergs, S. stheno- dactylus and S. petrii are found exclusively at the base of the rare and dispersed plants. Table 1. Taxonomic status of Tunisian Gekkota. S.I. species incidences. Family Genus Species Global distribution S.I. Statutes Sphaerodactylidae Euleptes europaea (Gené, 1839) Mainly restricted to Western Mediterranean island. Some small continental isolated populations have been reported in south of France and west of Italy (Salvidio and Delaugerre, 2003; Renet et al., 2008; Delaugerre et al., 2011; Salvidio et al., 2010). --- Insular relict Phyllodactylidae Tarentola mauritanica (Linnaeus, 1758) It ranges from the Iberian Peninsula to Italy in the north of Mediterranean sea; and from Morocco to the Nile Delta in the south (Kluge, 2001) 59.3% Ubiquitous neglecta Strauch, 1895 It is known with certainty in Algeria North-western Libya, Southern Tunisia and Chad (Schleich et al., 1996). 11.5% Desert habitat deserti Boulenger, 1891 It is present in Morocco, Algeria and Tunisia (Schleich et al., 1996, Harris et al., 2006). 44.25% Desert habitat fascicularis (Daudin, 1802) This species was recently elevated to the specific rank by Joger and Bshaenia (2010). 13.3% Endemic to central Tunisia Gekkonidae Hemidactylus turcicus (Linnaeus, 1758) A recent phylogeny assigned populations of North Africa to the Arid clade. This taxon has a large circum- Mediterranean distribution. 56.6% Ubiquistous Stenodactylus petrii Anderson, 1896 It is found in Africa, the Middle east and the southwest of Asia 23% Desert habitat sthenodactylus (Lichtenstein, 1823) It has been reported in Africa, the Middle East and Saudi Arabia (Schleich et al., 1996; Padial, 2006) 56.6% Sandy areas Tropiocolotes tripolitanus Peters, 1880 It ranges in North Africa, Middle East and parts of Asia. Populations of the Middle East have been transferred to the species T. somalicus. Those of Asia have been moved to Asiocolotes genera. Individuals from North Africa are now part of the species T. tripolitanus which is located in Tunisia, Libya and Egypt (Loveridge, 1974; Schleich et al., 1996) 41.6% Desert habitat (Hamada) 207Gekkotan distribution in Tunisia Fig. 2. Distribution of Gekkota species richness according to: bioclimatic stages (A), geomorphology (B), lithology (C) and vegetal cover. Maps were provided by the Office of the Ministry of Agriculture (Forestry Management Department) and further processed with QUAN- TUM-GIS software (Sillero and Tarroso, 2010). 208 W. Tlili et alii Grids clustering and regionalization Species richness varies from two to five species per grid cell. Calculated values of ȷ indices between the grids were all significant at the probability level 0.05. The den- drogram of Fig. 3A split at ȷ value of 0.6 to form two groups. The first group (yellow shades) assemblages grids occurring on Calci-magnesian and isohumic soils, under precipitation levels higher than 150 mm and with ther- mal amplitude lower than 20°C. This sub-branch con- tains two clusters: 1) Northern regions with the Sahel region (grids 49-56-57-61); and 2) Central region with Golf de Gabes region (grids 68-73-78-61). The second group (brown shades) gathers together the grids occur- ring on hydromorphic and poorly developed soils, under precipitation levels lower than 150 mm and with thermal amplitude higher than 20°C. It contains four clusters: 1) Chott El Jrid region (grids 65-66-70-73-74); 2) Tyaret region (grids 109-110-111-112-113); 3) Jbil region (grids 83-84-85-89-91-92-93-102-103-107-108) and 4) a cluster with no defined geographical area (grids 65, 66, 71 112, 113). A combination of corresponding colors was used to delineate regions according to cluster memberships (Fig. 3B). Fig. 3. Dendrogram (A) and map (B) resulting from UPGMA hierarchical clustering of grid cell assemblages of Gekkota based on Jaccard dissimilarity matrix at the species level. The six major biogeographical divisions are highlighted in the dendrogram with large colored rec- tangles. 209Gekkotan distribution in Tunisia Environmental factors 18 environmental variables were identified based on the hierarchical tree (Fig. 3A) and shared distribution patterns (Fig. 3). Most of these variables are highly cor- related producing great redundancy. A linear transforma- tion of correlated variables into uncorrelated variables via PCA retained five factors with Eigenvalues ≥ 1.0 (accord- ing to the Kaiser criterion). The total variance explained by all variables was 71.33% (Appendix 2). The orthogonal character of obtained factors was confirmed and clusters of items and rotated axes were identified (Table 2). Five orthogonal factors were yielded containing 11 environ- mental layers to develop gekkotan distribution models (Table 3). Distribution model The 11 environmental layers produced “features”, which constrain the distribution of computed prob- abilities. In our case, the set of features depend on the number of presence records for the species. The receiver operating characteristic (ROC) curves show that models for all Phyllodactylidae (Appendix 3) and most of Gek- konidae (Appendix 4) will well perform in predicting occurrences compared to a random selection of points. In fact, AUC values are all higher than 0.7 except for S. sthe- nodactylus which is 0.578 (Table 4). Also, table 4 shows threshold models that permitted to identify suitable cells for each species. The analysis of variables’ contributions Table 2. Factor Loadings with VARIMAX as a rotation method. Underlined loadings are >0.6; Bold values are negative correlations. S.R. species richness. Factor 1 Factor 2 Factor 3 Factor 4 Factor 5 Rain-thermal -0.895183 0.006325 -0.048915 0.078812 0.127187 Precipitations 0.746916 0.350094 0.080001 0.105304 0.116798 Lithology 0.108549 -0.353286 0.17636 -0.65443 0.16875 Soil 0.745688 -0.035934 -0.069964 -0.020273 -0.053747 Accident 0.447775 -0.419113 0.322996 -0.214502 0.421146 Vegetation 0.874349 0.07004 0.136885 -0.045448 -0.140143 Cereal 0.884682 0.019462 0.048258 0.036604 -0.107117 Olive 0.26776 0.424258 -0.215968 0.250303 0.130972 Almond 0.27634 -0.052531 -0.188399 0.430077 -0.303976 Palm -0.248537 0.132684 -0.045195 0.10083 0.839285 Water plan 0.045374 -0.040519 -0.03302 0.893364 0.145629 Coast 0.140236 0.887385 0.028427 -0.052517 0.032802 Altitude 0.581316 -0.466223 0.147382 -0.33244 0.023178 Urban 0.48289 0.551476 0.204528 0.190292 -0.02548 Tourism 0.131498 0.753782 0.056591 0.080994 0.147267 Species richness 0.049573 0.009893 -0.961705 0.101292 0.012076 Anthropophlic S.R. 0.858163 0.215614 -0.200664 0.148998 0.019184 Natural S.R. -0.665359 -0.184663 -0.650443 -0.027808 -0.009454 Explained variance 5.67628 2.589464 1.711235 1.74057 1.129378 Proportion .Total 0.315349 0.143859 0.095069 0.096698 0.062743 Table 3. Extended Factor Loading Matrix: Correlations of variable clusters (oblique factors) with primary factors. Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Secondary 1 0.242825 0.552793 -0.336319 0.545912 0.059906 Secondary 2 0.590997 0.220702 0.375875 -0.132976 -0.353716 Primary 1 0.769258 0.000000 0.000000 0.000000 0.000000 Primary 2 0.000000 0.803561 0.000000 0.000000 0.000000 Primary 3 0.000000 0.000000 0.863486 0.000000 0.000000 Primary 4 0.000000 0.000000 0.000000 0.827223 0.000000 Primary 5 0.000000 0.000000 0.000000 0.000000 0.933432 210 W. Tlili et alii to each model (Table 5) shows that three to four variables are involved in controlling distributions of ubiquitous species (H. turcicus, S. sthenodactylus, T. mauritanica, T. deserti and T. tripolitanus). Only one variable contribute with more than 70% in modeling distributions of scarce and rare species (T. neglecta and S. petrii). Response curves (Fig. 4) show how the logistic prediction changes as each environmental variable is varied, keeping all other environmental variables at their average sample value. For instance, S. sthenodactylus does not show specific environmental requirement; however the probability of it occurrence decreases with increased cereal land use. The probability of occurrence of T. neglecta increases with thermal amplitudes. In fact, areas with thermal ampli- tudes higher than 19 are particularly favorable for T. fas- cicularis, T. deserti, T. tripolitanus and S. petrii. T. fascicu- laris showed very narrow environmental requirements; it occurs in area with 200 mm of precipitation, between 20 and 22 of thermal amplitudes, 200 to 400 m asl, less than 100 of population density and 200 quintal of cereal yield- ed per hectare. Variable jackknife showed that the envi- ronmental variable that decreases the gain the most when Table 4. AUC (Area under curves), Cumulative threshold (C.T.), Logistic threshold (L.T.) and corresponding Fractional predicted Area (F.P.A.) for the MTP (Minimum training presences), MTSS (Maximum training sensitivity plus specificity) and EED (Equaled entropy of thresholded and original distributions). T.m. Tarentola mauritanica, T.n. T. neglecta, T.d. T. deserti, T.f T. fascicularis, H.t. Hemidactylus tur- cicus, S.p. Stenodactylus petrii, S.s. S. sthenodactylus, T.t. Tropiocolotes tripolitanus. H.t. S.p. S.s. T.d. T.f. T.m. T.n. T.t. A.U.C. 0.743 0.853 0.578 0.768 0.831 0.788 0.896 0.723 MTP C.T. 3.197 5.835 1.902 4.437 9.081 3.108 3.161 4.648 L.T. 0.166 0.277 0.280 0.223 0.215 0.226 0.338 0.247 F.P.A. 0.623 0.362 0.931 0.499 0.442 0.524 0.290 0.643 MTSS C.T. 6.55 13.186 12.861 4.434 17.655 8.489 17.017 12.569 L.T. 0.28 0.394 0.452 0.223 0.366 0.72 0.390 0.411 F.P.A. 0.587 0.303 0.779 0.499 0.326 0.463 0.216 0.512 EED C.T. 3.197 2.595 1.121 2.331 50.819 2.429 3.161 3.185 L.T. 0.166 0.235 0.236 0.152 0.141 0.221 0.338 0.208 F.P.A. 0.653 0.401 0.954 0.548 0.524 0.537 0.290 0.679 Table 5. Analysis of variable contributions T.m. Tarentola mauritanica, T.n. T. neglecta, T.d. T. deserti, T.f. T. fascicularis, H.t. Hemidactylus turcicus, S.p. Stenodactylus petrii, S.s. S. sthenodactylus, T.t. Tropiocolotes tripolitanus. Variable Percent contribution Permutation importance H.t. S.p. S.s. T.d. T.f. T.m. T.n. T.t. H.t. S.p. S.s. T.d. T.f. T.m. T.n. T.t. Thermal amplitudes 43.6 0 10.4 0.2 0.1 13.3 3.8 1.2 70.4 0 14.3 0.3 4.3 75.4 8.1 0 Cereal 20.9 75.1 37.4 71.9 0.4 55.9 2.5 0.1 12.3 41.8 17.5 32.8 25.5 15.4 0 4.3 Precipitations 13.6 1.3 1.8 0 0 0.6 9.2 0 0.8 14.3 2.4 0 0 2 76.2 0 Urbanism 10.1 0 0.3 4.7 32.1 23.1 0 39.5 13.8 0 11.8 9.9 2.5 5.6 0 39.6 Emberger coefficient 5 0.3 17 0.6 20.1 0.3 78.2 0.2 0 1.8 9 23.3 0 0 0 6.6 Vegetation 3.8 0.3 12.9 17.6 22.4 6.7 0 49.5 0 0 9 15.9 35 0 0 41.3 Soil 1.7 4.4 8.6 3.1 12.5 0 1 2.6 0 3.9 16.7 8.5 18.1 0 0.6 1.6 Coast 0.6 0 0 0 1.7 0 0 1.4 1.7 0 0 0 13.5 0 0 4.1 Palm 0.5 0 0.1 0 1.4 0.1 0 0.1 0.9 0 0 0 0.2 0.9 0 0 Lithology 0.1 1.1 3.2 0.6 1.3 0 5.03 1.6 0 4.4 4.5 0 1 0.7 14.4 0 Waterplan 0 0 0.2 0 0 0 0 3.8 0 0 0 0 0 0 0 1.1 Tourism 0 0 2.6 0.8 1.9 0 0 0 0 0 9.6 8.1 0 0 0 1.2 Olive 0 0.4 0 0 0 0 0 0 0 1.6 0 0.1 0 0 0 0 Altitude 0 1.7 1.6 0 2.6 0.1 0 0 0 3.9 1.7 0 0 0 0 0 Accident 0 3.2 0 0 0 0 0 0 0 3.9 0 0 0 0 0 0 Almond 0 0 3.8 0.4 0.5 0 0 0 0   3.5 1.2 0 0 0 0 211Gekkotan distribution in Tunisia Fig. 4. Response curves explaining gekkotan niches. Climatic items (A), relief (B) and land-use-items (C). 212 W. Tlili et alii Fig. 5. Predicted probabilities of presence of Gekkonidae within Tunisian territory and habitat quality classes. H. turcicus (A), S. sthenodac- tylus (B), S. petrii (C) and T. tripolitanus (D). 213Gekkotan distribution in Tunisia Fig. 6. Predicted probabilities of presence of Phyllodactylidae within Tunisian territory and habitat quality classes. T. mauritanica (A), T. fascicularis (B), T. neglecta (C) and T. deserti (D). 214 W. Tlili et alii it is omitted is thermal amplitudes for H. turcicus, cereal for Stenodactylus genus, soil for T. fascicularis, thermal amplitudes for T. mauritanica, precipitations for T. deser- ti, and vegetation for T. tripolitanus. Figures 5 and 6 show the projection of the model to unexplored grids covering Tunisian territory. DISCUSSION Tunisian geckos count nine species showing expan- sions of their ranges (Tlili et al., 2012 a, b). They also pre- sented distinct biogeographical affinities mostly concord- ant with the different habitat selection patterns (Fig. 3). This variation does not concern climatic affinities because 77.8% of Tunisian geckos are adapted to arid climate (unlike Tunisian Lacertidae) (Nouira, 1996; Nouira and Blanc, 2004) and Scincidae (Kalboussi, 2006). However, within this climatic frame, species of the same genus are latitudinally replaced to form several guilds according to vegetal and soil characteristics. The so far recognized distribution of gekkotan fauna shows that species distri- bution follows a bidirectional gradient. The South-North direction is physically stressful considering the increas- ing humidity and development of vegetal cover; these two parameters are constraining the species occurrences and only ubiquitous ones are present. The North-South direction is biologically stressful taking into account the increasing species richness (MacArthur, 1972; Real et al., 1997; Guisan and Zimmermann, 2000). Assumptions yielded from field observations were confirmed and shared distribution patterns were deter- mined (McLaughlin, 1992; Holt and Keitt, 2005; Heikin- heimo et al., 2007; Escalante, 2009; Kreft and Jetz, 2010). Thereby, a biogeographical regionalization of Tunisia according to gekkotan species richness could be estab- lished yielding five regions (Fig. 3). Region I harbors Anthropophilic geckos widely distributed generally occu- pying habitats proximal to the sea (coastal line) or to an important water body (lagoon, lac …). The greatest parts of these lands are widely exploited for agricultural or industrial purposes. The south-western part of this region harbors Mediterranean taxa specific to North Afri- ca; it represents the connecting point between Tunisia and southern neighboring countries and contains a large traffic network. Region II represents some culminant points in central Tunisia (Dorsal Mountain) and har- bors endemic geckos in addition to anthropophilic ones. Region III harbors Saharo-sindian geckos which occupy area of arid steppe lands characterized by the continen- tal influence and drastic conditions of life. As to region IV, it harbors Saharan geckos that occupy the Grand Erg Oriental. Finally region V harbors desert species that pre- fers rocky lands and small shrubs habitats. The gecko-based regionalization observed in south- ern Tunisia coincides with that based on Lacertidae fauna (Nouira, 1996; 2004). For instance, Saharan and Saharo-sindian Lacertidae were reported in southern Tunisia; their differentiation was attributed to edaphic gradient and climatic conditions (Nouira and Blanc, 2003; 2004). While Acathodactylus scutellatus inhabits bordering regions of the Sahara, A. dumerilli occupies sandy habitats of barchans regions and A. longipes inhab- its sandy Saharan biotopes of the grand erg oriental. The gecko-based regionalization is also concordant with our first assumptions yielded from field work. Nevertheless, we note the presence of a sixth cluster grouping spa- tially disjointed grids representing western and south- ern regions. Also, faunas of Northern regions are com- posed of two widespread cosmopolitan species which provide little useful biogeographical information. The question was then which of the environmental variables evidenced by clustering analysis is a more accurate rep- resentation of the distribution of geckos (Lawley et al., 2011; Strand, 2011). It is known that for ectotherms, including lizards, climate has been proposed as a key factor of their dis- tribution (e.g. Adolph and Porter, 1993; Doughty and Shine, 1995; Arad et al., 1997; Kaspari and Valone, 2002; Carretero, 2008; Salvidio and Oneto, 2008). Tempera- ture and rainfall intervene indirectly by co-limiting pri- mary production and physiologically limiting access to that production (Kaspari et al., 2000). Edaphic and relief characteristics take part in burrows selection (Zaady and Bouskila, 2002). Results provided herein suggest retain- ing five oblique factors to explain gekkotan distributions: 1) Environmental items, 2) coastal items, 3) biotic items, 4) sebkhas and chotts landscape items and 5) oasis land- scape items. It also shows that gekkotan species richness is positively correlated to environmental items (climate and natural land-use). It is negatively correlated to coast- al items (urban and touristic land-use) and geomorphol- ogy (altitudes and geological accidents). A strong rela- tionship has been revealed between reptiles’ occurrences and the orographic structures in Tunisia (Nouira, 1996). For instance, the Dorsal Mountain constitutes a barrier to the expansion of Saharan species toward the Palearctic domain and vice versa. So far, this is in accordance with our observations concerning the distribution of the spe- cies S. sthenodactylus being limited to the south-eastern limit of the Dorsal. The absence of the species in north- ern regions remains not explained by climate require- ments alone; it could be related to geographical barriers and to their history of colonization and settlement. In 215Gekkotan distribution in Tunisia fact, S. sthenodactylus, considered as Saharan, becomes related to widely Mediterranean species and expand its ecological niche to sub-humid regions. A better under- standing of the niche of the species is shown in figure 4 where the probability of occurrence of the species decreases with high precipitations, low thermal ampli- tudes, altitudes higher than 250m, but most importantly, cereal land-use. Maps of the Fig. 2 show that these are the same characteristics of the extreme north of Tuni- sia which explains the absence of the species within this area. Map B of Fig. 5 shows that the suboptimal area of occurrence of S. sthenodactylus lays beyond the Dorsal Mountain and in the Cap Bon peninsula. However, area of very low probabilities of occurrence did appear in the map despite the certain presence of the species within this area. For the other gekkotan species, predicted areas fol- low the general distribution patterns previously yielded. Besides, established models defined suitable habitats for each species and confirmed the latitudinal gradient of distribution. Land use and altitude increase the prob- ability of occurrence of H. turcicus and T. mauritanica. Alternatively, they decrease the probability of the pres- ence of T. deserti, T. neglecta, T. tripolitanus and S. petrii. H. turcicus occurs within area with high cereal land use; this is in accordance with field observations that revealed that the species inhabits preferentially hay fields. As to T. fascicularis, it showed a little displace- ment of the northern limit of its range towards semi- arid regions. Being a newly described species this dis- placement must be the result of better field explorations, however, modeling of its niche predicted areas of prob- able occurrence within the central arid steppe. High probabilities of occurrence of the species are restricted to domains characterized by less than 200 mm of pre- cipitations, 22 to 21 thermal amplitudes, 200 m of alti- tudes and very low land use (Fig. 4). ACKNOWLEDGMENTS We thank the Office of the Ministry of Agriculture (Forestry Management Department) who provided nec- essary authorization for animal collecting (N° 902 and 1294). We also thank the officials from the meteorologi- cal stations and the Tunisia Wildlife Conservation Society TWCS who provided information and documents neces- sary to our work. Special thanks are addressed to Marc Cheylan, Claudia Corti and José D. Anadón for their helpful advices. Our gratitude is addressed to Doghri Moadha, Mrabet Kais, Doghri Ali, Khemiri Rajeh, and Doghri Aymen for material help. SUPPLEMENTARY MATERIAL Supplementary material associated with this article can be found at: Manuscript number 12875. Appendix 1, Appendix 2, Appendix 3 A, Appendix 3 B. REFERENCES Adolph, S.C., Porter, W.P. (1993): Temperature, activity and lizard life histories. Am. Nat. 142: 273-295. Anadón, J.D., Giménez, A., Graciá, E., Pérez, I., Ferrán- dez, M., Fahd, S., Fritz, U. (2012): Distribution of Tes- tudo graeca in the western Mediterranean according to climatic factors. Amphibia-Reptilia 33: 285-296. Arad, Z., Schwarzbaum, A., Werner, Y. (1997): Tempera- ture selection and thermoregulation in the Moorish gecko, Tarentola mauritanica. Amphibia-Reptilia 69: 269-282. Brito, J.C., Acosta, A.L., Álvares, F., Cuzin, F. (2009): Biogeography and conservation of taxa from remote regions: An application of ecological-niche based models and GIS to North-African Canids. Biol. Con- serv. 142: 3020-3029. Carretero, M.A. (2008): Preferred temperatures of Taren- tola mauritanica in spring. Acta Herpetol. 3: 57-64. Clark, W.C., Kuhl, J.P., Keohan, M.L., Knotkova, H., Winer, R.T., Griswold, G.A. (2003): Factor analy- sis validates the cluster structure of the dendrogram underlying the multidimensional affect and pain sur- vey (MAPS) and challenges the a priori classification of the descriptors in the McGill pain questionnaire (MPQ). Pain 106: 357-363. Delaugerre, M., Ouni, R., Nouira, S. (2011): Is the Euro- pean Leaf-toed gecko Euleptes europaea also an Afri- can? Its occurrence on the Western Mediterranean landbrige islets and its extinction rate. Herpetol. Notes 4: 127-137. Doughty, P., Shine, R. (1995): Life in two dimensions: natural history of the southern leaf-tailed gecko, Phyl- lurus platurus. Herpetologica 51:193-201. Emberger, L. (1950): Rapport sur les régions arides et semi-arides de l’Afrique du Nord. Institut national de la zone aride. UNESCO, Paris. Escalante, T. (2009): Un ensayo sobre regionalizacion bio- geografica. Rev. Mex. Bio. 80: 551-560. Fakhfakh, M., Laclavère, G. (1979): Les Atlas de Jeune Afrique, Tunisie. Éditions Jeune Afrique. Paris. Franklin, J., McCullough, P., Gray, C. (2000): Terrain var- iables for predictive mapping of vegetation communi- 216 W. Tlili et alii ties in Southern California. In: Terrain Analysis: Prin- cipals and Applications. Wilson, J., Gallant, J., Eds, John Wiley and Sons, New York. Graham, C.H., Hijmans, R.J. (2006): A comparison of methods for mapping species ranges and species rich- ness. Global Ecol. Biogeogr. 15: 578-587. Guidi, L., Ibanez, F., Calcagno, V., Beaugrand, G. (2009): A new procedure to optimize the selection of groups in a classification tree: Applications for ecological data. Ecol. Model. 220: 451-461. Guisan, A., Weiss, S.B., Weiss, A.D. (1999): GLM versus CCA spatial modeling of plant species distributions. Plant. Ecol. 143: 107-122. Guisan, A., Zimmermann, N.E. (2000): Predictive habitat distribution models in ecology. Ecol. Model. 135: 147- 186. Harrell, F.E., Lee, K.L., Mark, D.B. (1996): Multivariable prognostic models: Issues in developing models, eval- uating assumptions and adequacy, and measuring and reducing errors. Stat. Med. 15: 361-387. Harris, D.J., Batista, V., Lymberakis, P., Carretero, M.A. (2004): Complex estimates of evolutionary relation- ships in Tarentola mauritanica (Reptilia: Gekkonidae) derived from mitochondrial DNA sequences. Mol. Phylogenet. Evol. 30: 855-859. Heikinheimo, H., Fortelius, M., Eronen, J., Mannila, H. (2007): Biogeography of European land mammals shows environmentally distinct and spatially coherent clusters. J. Biogeogr. 34: 1053-1064. Hirzel, A.H., Hausser, J., Chessel, D., Perrin, N. (2002): Ecological-niche factor analysis: how to compute hab- itat-suitability maps without absence data? Ecology 83: 2027-2036. Holt, R.D. (2003): On the evolutionary ecology of species’ ranges. Evol. Ecol. Res. 5: 159-178. Holt, R.D., Keitt, T.H. (2005): Species borders: a unifying theme in ecology. Oikos 108: 3-6. Jaccard, P. (1908): Nouvelles recherches sur la distribu- tion florale. Bull. Soc. Vaud. Sci. Nat. 44: 223-270. Joger, U. (2003): Reptiles and Amphibians of Southern Tunisia. Kaupia 12: 71-81. Joger, U., Bshaenia, I. (2010): A new Tarentola subspe- cies (Reptilia: Gekkonidae) endemic to Tunisia. Bonn. Zool. Bull. 57: 267-274. Kalboussi, M. (2006): Biosystématique, Biogéographie et Ecologie des Scincidae (Reptilia) de la Tunisie. Unpublished doctoral dissertation. Faculté des Sci- ences de Tunis. Kaplunovsky, A.S. (2005): Factor analysis in environmen- tal studies. HAIT J. Sci. Eng. B 2: 54-94. Kaspari, M., Donnell, S., Kercher, J.R. (2000): Energy, density, and constraints to species richness: studies of ant assemblages along a productivity gradient. Am. Nat. 155: 280-293. Kaspari, M., Valone, T.J. (2002): On ectotherm abun- dance in seasonal environment-studies of a desert ant assemblage. Ecology 83: 2991-2996. Kluge, A.G. (2001): Gekkotan lizards taxonomy. Hama- dryad 26: 1-209. Kreft, H., Jetz, W. (2010): A framework for delineating biogeographical regions based on species distribu- tions. J. Biogeogr. 37: 2029-2053. Lawley, E.F., Lewis, M.M., Ostendorf, B. (2011): Envi- ronmental zonation across the Australian arid region based on long-term vegetation dynamics. J. Arid Environ. 75: 576-585. Loveridge, A. (1947): Revision of the African lizards of the family Gekkonidae. Bull. Mus.Comp. Zool. 98: 1-50. MacArthur, R.H. (1972): Geographical ecology: patterns in the distribution of species. Princeton University Press, Princeton. McLaughlin, S.P. (1992): Are floristic areas hierarchically arranged? J. Biogeogr. 19: 21-32. Nouira, S. (1996): Systématique, Ecologie et Biogéog- raphie évolutive des Lacertidae (Reptilia, Sauria). Importance dans l’herpétofaune tunisienne. Unpub- lished doctoral dissertation. Faculté des Sciences de Tunis. Nouira, S. (1997): Biodiversité de l’herpétofaune tunisi- enne: II. Les Gekkonidae (Reptilia, Sauria). Bull. Soc. Sci. Nat. Tunis. 26: 66-74. Nouira, S., Blanc, C.P. (2003): Distribution spatiale des Lacertidae (Sauria, Reptilia) en Tunisie; caractéris- tiques des biotopes et rôle des facteurs écologiques. Ecol. Mediterr. 29: 71-86. Nouira, S., Blanc, C.P. (2004): Organisation spatiale et modalités de mise en place du peuplement des Lacer- tidé (Sauria, Reptilia) en Tunisie. Bull. Soc. Herp. Fr. 110: 5.34. Phillips, S.J., Anderson, R.P., Schapire, R.E. (2006). Maxi- mum entropy modeling of species geographic distri- butions. Ecol. Model. 190: 231-259. Phillips, S.J., Dudik, M. (2008): Modeling of species dis- tributions with Maxent: new extensions and a com- prehensive evaluation. Ecography 31: 161-175. Real, R. (1999): Tables of significant values of Jaccard’s index of similarity. Miscellània Zoològica 22: 29-40. Real, R., Vargas, J.M. (1996): The probabilistic basis of Jaccard’s Index of similarity. System. Biol. 45: 380-385. Renet, J., Gerriet, O., Jardin, M., Magne, D. (2008): Les populations de Phyllodactyle d’Europe Euleptes euro- paea Gené, 1839 Reptilia, Sauria, Gekkonidae dans les Alpes Maritimes: premiers éléments sur leur répar- 217Gekkotan distribution in Tunisia tition et leur écologie. Faune de Provence  (CEEP) 24-25: 117-126. Rufray, V., Duguet, R., Durand, C., Fradet, V. (2003): Découverte d’une troisième station continentale de Phyllodactyle d’Europe Euleptes europaea en France et mise au point sur le nouveau statut taxonomique de l’espèce. Faune de Provence (CEEP) 21: 13-14. Salvidio, S., Delaugerre, M. (2003): Population dynamic of the european leaf-toed gecko (Euleptes europaea) in NW Italy: implications for conservation. Herpetol. J. 13: 81-88. Salvidio, S., Lanza, B., Delaugerre, M. (2010): Euleptes europaea (Gené, 1839). In: Fauna d’Italia. Reptilia, pp 258-270. Corti, C., Capula, M., Luiselli, L, Razzetti, E., Sindaco, R., Eds, Edizione Calderini de Il Sole 24 Ore, Bologna. Salvidio, S., Oneto, F. (2008): Density regulation in the Mediterranean leaf-toed gecko Euleptes europaea. Ecol. Res. 23: 1051-1055. Schleich, H.H., Kästle, W., Kabisch, K. (1996): Amphibi- ans and Reptiles of North Africa, Ed., Koeltz, Koenig- stein. Sillero, N., Tarroso, P. (2010): Free GIS for herpetologists: free data sources on Internet and comparison analy- sis of proprietary and free/open source software. Acta Herpetol. 5: 63-85. Sneath, P.H.A., Sokal, R.R. (1973): Numerical Taxonomy. The principles and practices of numerical classifica- tion. Freeman, San Francisco. Strand, G.H. (2011): Uncertainty in classification and delineation of landscapes: a probabilistic approach to landscape modeling. Environ. Modell. Softw. 26: 1150-1157. Tlili, W., Delaugerre, M., Ouni, R., Nouira, S. (2012): Dis- tributional review of the genus Tarentola (Reptilia, Sauria) in Tunisia (North Africa). Herpetol. Notes 5: 485-492. Tlili, W., Ouni, R., & Nouira, S. (2012): New distribution records of the genus Stenodactylus (Reptilia, Sauria) in Tunisia (North Africa). Herpetol. Notes 5: 413-418. Zaady, E., Bouskila, A. (2002): Lizards burrows associa- tion with successional stages of biological soil crusts in an arid sandy region. J. Arid Environ. 50: 235-246.