The City in the Tropics: Lizard Community Structure and Spatial Resource Use Along an Habitat Gradient of the Dahomey Gap (West Africa)


 Studies on habitat gradients are particularly interesting for the community ecology theory, but almost no investigation has been carried out on community structure of any vertebrate groups along gradients of habitat that were modified by humans in historical or prehistorical times, such as for instance the Dahomey Gap in West Africa. Here, we analyze the community structure of lizards in suburban Lomé (Togo) and in comparison with nearby savannah and forest sites, with a suite of statistical methods. Overall, we recorded 25 lizard species, with a heavy reduction in species richness from forest (18 taxa) to savannah (13) and suburbs (9). 24% of the species occurred in all habitat types, 40% exclusively in forest, and only two were exclusive of suburban habitats. Suburban habitat types were relatively homogeneous in terms of number of observed species (maximum number of taxa per habitat = 6). There were significant interspecific differences in both substratum type preferences and vertical spatial niche by species, but with no evidence of a nonrandom niche partitioning pattern, and hence with a competitively assembled community structure. There was a nonrandom “clustered” distribution of the various species along the available resource categories, thus indicating that species-specific preferences instead of community-driven mechanisms are more likely to explain the observed patterns. We concluded that lizard communities in tropical cities are (i) less species-rich than in the surrounding more natural habitats, (ii) usually clustered into specific habitat/substratum types (often artificial ones), and (iii) not assembled through competitive interactions.


Introduction
Studies on habitat gradients are particularly interesting for the community ecology theory because the environmental continuum may be a useful axis for understanding how attributes of individuals do generate structure at the level of the community (e.g. see Wellborn et al. 1996; Pillsbury and Miller 2008; Khatiwada et al. 2019). It has been shown that community structure across a given gradient may be determined by both physical factors limiting the potential breadth of species distributions (for instance, proximate eco-physiological performances; see Barbault 1991; or habitat characteristics, e.g. Avila et al. 2017) and biotic effects mediated by ecological interactions, such as predation (Nurminen et al. 2018;Freestone et al. 2020) and interspeci c competition (Pianka 1986; Salinas-Ramos et al. 2020; Powell et al. 2021).
Most studies on community structure along tropical habitat gradients have focused so far on habitats that have been modi ed by humans during recent times, for instance burnt versus non-burnt areas (Akani et al. 2018; Costa et al. 2020) or forest-plantation mosaics (Palmeirim et al. 2017; Supsup et al. 2020). However, almost no investigation has been carried out on community structure of any vertebrate groups along habitat gradients that were modi ed by humans in historical or prehistorical times, despite these studies may be important for a better under standing of the community dynamics along habitat gradients. In this regard, the Dahomey Gap in West Africa may offer a excellent scenario where to perform such studies. The Dahomey Gap is a savannah-like vegetation zone that extends all the way to the coast in Benin, Togo, and Ghana, thus separating the forest zone that covers much of the south of the region into two separate forested regions: the Upper Guinean forests and the Lower Guinean forests (Demenou et al. 2018). The Dahomey Gap has been created, according to some studies, by deforestation during the late Holocene, whereas for other researchers it should be a natural transitional vegetation zone between forest and savannah (Demenou et al. 2018). Anyway, now the Dahomey Gap is also characterized by the presence of a myriad of urban centres, with Accra (capital of Ghana), Lomé (capital of Togo) and Cotonou/Porto Novo (Benin) being the largest urban agglomerates.
In this study, we analyze the community composition of sympatric lizards along a habitat gradient, forest-savannah-suburbs (hereby FSS), in south-western Togo, inside the Dahomey Gap of West Africa. We will focus (i) on the taxonomical composition and species richness of lizards within each habitat type, and (ii) on their resource partitioning patterns along the spatial niche axis with regard to the suburban habitat. We selected the suburban habitat for the detailed resource partitioning study because this is the simplest habitat type of the Dahomey Gap in terms of number of sympatric species, thus allowing a easier quantitative assessment and a better understanding of the community dynamics.
More speci cally, in the present paper, we examine the following key questions: 1. Are there any species richness differences among lizard assemblages inhabiting the various habitats in the FSS gradient? This question is relevant because there are empirical evidences that tropical forest communities are more species-rich than savannahs or deforested areas (Estrada et al. 1993 3. Focusing on the ecologically simplest habitat type (in terms of species richness) in the gradient, what are the main assembly rules of the lizard assemblage? That is: is interspeci c competition for the spatial niche an important determinant of the community structure in the simplest habitat type of the FSS gradient? This aspect is relevant because it is easier to detect interspeci c competition mechanisms in simple than in complex arenas, and thus a relatively simple natural scenario may help considerably our understanding on this issue.

Study areas and eld protocol
In order to compare community composition of lizards across a gradient of habitat alteration (FSS), we used data from Segniagbeto et al. (2015) for forest and forest-derived savannah, supplemented by original opportunistic eld observations (years 2013-2020), and compared these data with those originally collected by us in the suburbs of Lomé (see below). The map of southern Togo, showing the various study areas, is given in Fig. 1. For forest, we considered the presence data of lizards for Missahohé, Agome Yo, Kpalimé and Agou, i.e. four hilly, adjacent sites situated nearby the border with Ghana in south-western Togo. For forest-derived savannah,we considered the presence data of lizards at Tsévié, Tgblekope and Agoé, also situated in the south-western Togo.
The eld study at the urban area was carried out in July-August 2013 and in July-August 2014 in a suburban area of Lomé (Baguida), the capital city of Togo, West Africa. Baguida is a residential area situated along the coast, at the western periphery of Lomé, nearby the city port and along the highway connecting Togo with Benin. The climate of this area is tropical, with dry season from October to March, wet season from April to September, but with a short phase of relative rain shortfall during August.
Suburban lizards were searched for along random transects, walked slowly and by two-three researchers, passing through the Baguida landscapes, from about 0900-1200hrs and 2000-2200hrs, for a total of 21 eld days in 2013 and 23 days in 2014. Visual Encounter Survey (VES) methodology was applied, and each lizard was observed at a distance ranging 10-150 cm. During night hours, we used a torch (model Weltool T10) to examine the various substratum types and recording the lizards. We considered in the analyses only the above-ground active individuals, so, we did not consider those lizards (mainly Agama agama) that were observed while sleeping perched on walls, trees or other substrates.
Each observed lizard was identi ed to species and sex; however, H. angulatus and H. mabouia were pooled due to their di cult discrimination in the eld. We recorded the habitat characteristics of each lizard individual along a horizontal and a vertical spatial axes. In order to evaluate the horizontal spatial resource use, we recorded the substratum where each individual was recorded. Lizard sightings were assigned to one of the following six different substratum typesΛ1) cement/asphalt, (2) grass, (3) trees/wood, (4) sand, (5) bushes, and (6) stone-piles. The relative availability of each substratum type was determined by eye by inspecting the relative area occupied by each habitat type in 1000 random spots of 5 x 5 m that were selected along the surveyed area. N order to evaluate the horizontal spatial resource use, we determined the height from the ground of each lizard individual at the sighting time.

Statistical analyses
Comparisons of lizard community compositions among the three FSS gradient habitats were made just qualitatively as there were no su cient quantitative data in Segniagbeto et al. (2015) to allow any statistical analysis. Concerning the XXXystemat lizard community, the differences in the frequencies of observed lizards by species and by substratum type, as well as within species (i.e. males versus females versus juveniles) and by substratum type, were assessed by contingency table χ 2 test. Pearson's correlation coe cient was used to analyze the correlation between substratum type availability (independent variable) and number of observed lizards by species (dependent variable). When running the above-mentioned analyses, we considered that, in order to process the data statistically, the su cient number of individuals per species was 30. Thus, we did not analyze the data by contingency tables and correlations for those species with less than 30 sightings. In order to differentiate the various types of substrates in terms of their quantitative lizard community composition, we used a One-Way Analysis of Similarities (ANOSIM). ANOSIM is roughly analogous to an ANOVA in which the univariate response variable is replaced by a dissimilarity matrix, i.e. with distances that were converted to ranks (Clarke 1993). Signi cance was computed by permutation of group membership, with 9,999 replicates, and Bray-Curtis was used as distance measure. Sequential Bonferroni correction was used for post-hoc comparisons of pairwise ANOSIMs between all pairs of groups.
We calculated Chao-1 index in order to predict the theoretical number of species that can be expected to use each substratum type on the basis of the sampling regime. In addition, the following univariate community diversity metrics were calculated for each substratum type: (v) Evenness (e), calculated by Buzan and Gibson's formula (Magurran 1988). For each diversity metric, we also generated upper and lower 95% con dence intervals by performing a bootstrap analysis with 9999 random samples, each with the same total number of individuals as in each original sample generated (Harper 1999).
Niche breadth (B) in resource uses was assessed by Simpson's (1949) index. In order to evaluate the "structure" (sensu Gotelli and Entsminger 2003) of the lizard community in terms of substratum use and height of perches, we rstly calculated the similarity XXXystema species/population groups by Pianka's (1986) symmetric equation (with 0 indicating no similarity and 1 total similarity), and then performed a statistical simulation study. Indeed, to assess whether the various observed overlap values between species/population group pairs occurred by chance, we randomized the original species utilization matrices by shu ing the original values among the resource states by using the randomization algorithms RA2 and RA3 (Lawlor 1980), with 30,000 random Monte Carlo permutations generated by Ecosim software. RA2 tests for structure in the generalist-specialist nature of the resource utilization matrix by conserving guild structure, but destroying observed niche breadth (Gotelli and Graves 1996). RA3 tests for guild structure by conserving niche breadth for each species, but destroying guild structure manifested by the zero structure of the resource utilization matrix (Gotelli and Graves 1996). If the observed overlaps were signi cantly lower than the randomized overlaps, there was evidence of community-wide niche partitioning of the analyzed niche dimensions (substratum type and perch height). If the observed overlaps were signi cantly higher than the randomized overlaps, there was evidence of community-wide convergent use of the niche dimensions (i.e. a "contagion" effect), whereas if observed and randomized overlaps did not differ signi cantly, the lizard community used the niche dimensions by chance, i.e. randomly. The substratum type availability (%) was added to the XXXystematicXXX in the Ecosim software.
The pairwise similarity of each lizard species with the availability of each substratum type was assessed by non-metric multidimensional scaling (nMDS), using cosine as similarity measure between items in the item-item matrix. A Principal Component Analysis (PCA) was used to displaying the various lizard species by substratum type in the multivariate space, retaining only factors with eigenvalues > 1 in the analysis.

General community characteristics
Pooling all the study areas, there were 25 lizard species in total, belonging to ve different families (Table 1). Comparing the species composition along the FSS gradient, there was a heavy reduction in the species richness, from 18 taxa in the forest to 12 in the savannah up to 9 in the suburbs ( Table 1). Six species (24% of the total) were found in all habitat types along the gradient, ten (40%) were found exclusively in forest, four (16%) only in savannah, two were found in both forest and savannah (8%), in both forest and suburbs, and in only suburban habitats. This exclusively suburban species were the geckos Ptyodactylus ragazzi, that was recently introduced by merchants of charcoal and rewood transported from the northern regions of Togo (our unpublished observations), and Tarentola ephippiata. These latter species are currently common in some areas of Lomé, for example in the Be-Kpotà sector of the main city centre. There were signi cant differences in the frequency of species by ecological niche (arboreal/climbers; arboreal-terrestrial; terrestrial; subterranean) across the FSS gradient (contingency table: χ 2 = 23.4, df = 6, P < 0.001), although arboreal-terrestrial and arboreal/climbers were dominant in all habitat types (Fig. 2). The forest lizard community had however a higher breadth of utilized niches compared to the other two habitats (Fig. 2).
Community structure in the suburban area In the suburban site, during our surveys we observed a total of 611 individuals belonging to 6 species of lizards. Three of the observed species were Scincidae Species richness was adequately sampled within each habitat type, as revealed by the plateau phase reached by the respective saturation curves (Online Supplementary Figure S1). Habitat types were relatively homogeneous in terms of number of observed species, with the highest number (n = 6) being observed in bushes (Table 2). Chao-1 estimated that the maximum number expectable per habitat type is 7 (Table 3). In terms of diversity metrics (calculated from the number of individuals by lizard species sighted on each substratum), the various substrates showed relatively similar values, with bushes and trees/wood being used by a slightly higher diversity of species and with a slightly higher evenness than the other substrates (Table 3). However, there were no statistically signi cant differences among substrates for any of the diversity metrics using ANOSIM (mean rank within substratum types = 29.46; mean rank between substratum types = 37.6; R = 0.05, at least P > 0.155 in all metrics). Suburban lizards: horizontal niche resource use The synopsis of the observations by species and by substratum type, and in relation to the relative availability of each substratum, is given in Table 2 Overall, cement (29.7% of the sightings) and grass (29.8%) were the substrates where lizards (all pooled species) were more frequently observed, followed by trees (12.7%), bushes (10.3%), sand (7.8%) and stonepiles (0.4%).
Substrates were used in relation to their relative availability by A. agama, but not by T. quinquetaeniata and H. angulatus (Table 4), showing that the former species was more a substratum type generalist than the two latter species.  (Fig. 3). In this analysis, Component 1 explained 77.02% of the whole variance and Component 2 explained 22.35%. The three species were also clearly spaced apart in the multidimensional space, with A. agama being very close to the habitat availability centroid, in a PCA diagram (Fig. 4). In both analyses, Hemidactylus sp. centroid was spaced far away from that of the other two species within the multidimensional space.

Suburban lizards: vertical niche resource use
The raw data on the distribution of lizard sightings by vertical height from the ground are given in Table 5. As for A. agama, males had much higher niche breadth (B = 4.46) than females (B = 2.41) and juveniles (B = 2.03). This difference was due to the fact that males (1) used more frequently perches at > 101cm height (26.4% of observed individuals) than females (7.4%) or juveniles (2.9%), and (2) they were observed more frequently at 10-50 cm height, whereas females and juveniles mostly on ground (Fig. 5), Thus, males usually perched higher than females and juveniles and overall, the differences between the three groups of individuals were statistically signi cant (χ 2 test, P < 0.001). On the other hand, in T. quinquetaeniata the three groups of individuals did not differ signi cantly (χ 2 test, P < 0.001), with all of them being found essentially on ground (Fig. 6). However, also in this species the males exhibited a wider niche breadth (B = 3.26) than females (B = 1.82) and juveniles (B = 1.56).  and P(observed > = expected) = 0.024. In this case, both RA3 and RA2 gave consistent results, showing that there was a non-random preference for same height category by the whole lizard community.

Discussion
General community characteristics along the habitat gradient Although not methodological consistent across habitat types, our study revealed that along the FSS gradient (i) there was a clearly higher species richness in the forest than in the savannah and suburbs, and that (ii) the three communities were quite taxonomically distinct. Thus, there was a "forest-speci c" assemblage of species (including e.g. Varanus ornatus) and lowly diverse assemblages in savannahs (including e.g. Varanus exanthematicus) (Fig. 7) and suburbs, mostly characterized by ecological generalists with wide habitat tolerances. Both these diversity patterns were not unexpected (McKinney 2008).
Concerning pattern ( Concerning pattern (ii), the fact that we observed habitat-speci c assemblages of lizards at our study area mirrors observations from Brazilian Cerrado lizard communities (Nogueira et al. 2009). In the Brazilian Cerrado, in fact, habitat-specialists with patchy distributions in the forest-savannah-plantation mosaics create habitat-structured lizard assemblages with faunal overlap between forest and savannah being limited and forested versus open areas acting as mutual barriers to lizard distribution (Nogueira et al. 2009). In our study case, the taxonomical composition of the savannah assemblage was similar as that of the suburban habitat (although in the savannah we also observed Agama sankaranica and a few other species), thus suggesting that there is no clear barrier acting against lizard distribution between savannah and suburban areas in southern Togo.
The species richness reduction from forest to urban areas may also be interpreted as an effect of habitat loss on forest-specialists (Turner 1996 Suburban lizards: analysis of community structure and niche characteristics Our data revealed interspeci c differences in both substratum type preferences and vertical spatial niche, but with no evidence of a mathematical structure consistent with a nonrandom niche partitioning pattern, and hence with a competitively assembled community structure. Instead, we detected a nonrandom "clustered" distribution of the various species along the available resource categories. Therefore, the main conclusion of our study is that species-speci c preferences instead of community-driven mechanisms are more likely to explain the observed patterns at the study area.
Thus, we found no support for the hypothesis that lizards, as they can increase net energy gain from careful thermoregulation and hence selection of appropriate basking substrates, can compete for, and exclude potential competitors from, optimum basking sites (e.g. Huey and Slatkin 1979). When interspeci c competition cannot be advocated, the proximate external ecophysiological constraints are often considered the drivers of species-speci c differences in niche characteristics (Barbault 1988; Barbault and Stearns 1991

General considerations on lizard communities in tropical cities
It is well known that urbanization, involving the conversion of natural habitats into human-modi ed ecosystems, usually reduce both the diversity and the abundance of indigenous animal communities, up to leading to the extinction of some species or even facilitating the establishment of non-indigenous communities in cities and towns (Hamer and McDonnell 2010). So, the general observations that we made (i.e. a reduced species richness in the suburban habitat of the FSS gradient) is not surprising. However, as mentioned above, our study also suggests that the lizard species used the available resources in a clustered way within the suburban habitat in Lomé. Whereas many studies analyzed reptile assemblages in secondary forests and altered habitats (e.g., Luya et al. 2008), the data available in the literature with regard to communities of lizards (or other reptiles) in tropical cities are relatively few and based on different methodologies from ours, and so comparisons with our study are not straightforward. However, in a snake community inhabiting a Brazilian city, Franca and Franca (2019) observed a clustered distribution for two species and dispersed distributions for eight species, and in a Indian city, Janiawati et al.
(2016) demonstrated that the various reptilian species were clustered around water sources and vegetation cover patches. In fact, in urban fragments embedded in an "arti cial matrix" hostile to the movement of individuals, most species tend to be concentrated into suitable microhabitats (and using clustered types of substratum) as extinction rates (stochastic or anthropogenic) are not compensated by colonization rates (Laurance et al. 2006a(Laurance et al. , 2006bSalomão et al. 2019).Thus, we may conclude that lizard communities in tropical urban areas are (i) less species-rich than in the surrounding more natural habitats, (ii) usually clustered into speci c habitat/substratum types (often arti cial ones), and (iii) not assembled through competitive interactions. Further studies should verify the validity of these conclusions.

Declarations
Funding -This study was indirectly supported by funds provided from Mohamed Bin Zayed Species Conservation Fund and Turtle Conservation Fund.

Con icts of interest -none
Ethics approval -not applicable Consent for publication -All authors read and approved the nal manuscript, and accepted to be co-authors in the published paper.
Availability of data and materials -Data will be available by the rst author (LL) on request Figure 1 Map of southern Togo, showing the various study areas. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 3
Non-metric multidimensional scaling (with COSINE as similarity measure) representing suburban lizard species in relation to their substratum use and availability in the eld. In this graphic, cement and grass are placed along the Component 1 (positive values) and trees, bushes and stone piles along    Percentage of suburban Trachylepis quinquetaeniata individuals observed at the various height categories from the soil. Sample sizes: males, n = 44; females, n = 64; juveniles, n = 14.