Urbanization effect on the ant diversity and composition in an Arid City

Simple Summary: Urbanization represents a clear threat to biodiversity at the global level, particularly in arid and semi-arid regions. To understand the effect of urbanization on the biological community in the arid region, we studied the ant community along urbanization gradients in Wadi Hanifa in Riyadh, central Saudi Arabia. We found consistency in diversity parameters (abundance, richness, evenness, and Shannon and Simpson diversities) across the urbanization gradient. How-ever, we observed discrete differences in ant community structure across the urbanization gradients. Environmental factors such as vegetation type, soil properties and ground cover proved to be important determinants of ant species composition. Our data supports the use of ants as indicators of urbanization effects and must be considered in order to monitor impacts of urbanization on the biodiversity in arid regions. Abstract: The dramatic increased rates of uncontrolled urbanization in various parts of the World have resulted in loss of native species and overall threats to biodiversity. Over the last few decades Saudi Arabia has witnessed a remarkably rapid population growth and unparalleled levels of urbanization, leading to threats to biodiversity. Ants were pitfall-trapped across an urban-rural gradient to evaluate ant assemblage responses to urbanization in Wadi Hanifa, Riyadh, Central Saudi Arabia. Fifteen sampling sites were selected along three different urbanization gradients, each trav-ersing urban, suburban and rural zones. Within each site 10 traps were distributed and operated for 7 consecutive days, at 3-monthly intervals throughout one year. Vegetation, ground cover, and chemical and physical soil variables at sampling sites were analyzed concurrently. Ant abundance, species richness, evenness, and diversity indices of Shannon and Simpson were calculated for each site using PC-ORD to demonstrate diversity patterns along the urbanization gradients. Ant assemblages were assessed by detrended corresponding analysis (DCA), canonical correspondence analysis (CCA), and analysis of similarity (ANOSIM) using PC-ORD. Indicator species analysis was conducted to define representative species along the urbanization gradient. A total of 42 ant species were identified. The diversity parameters were consistent across the urbanization gradient. How-ever, significant differences were observed in the ant assemblages between rural and urban, suburban and urban, but only marginal between rural and suburban. Eleven ant species were identified as indicator species (IV values between 50.7-80.7%). The ant assemblages were influenced by flora, ground cover, and soil variables.


Introduction
Urbanization is spreading worldwide, especially within arid and semiarid regions. Many of these arid regions are extremely urbanized, illustrated by countries such as Qatar, United Arab Emirates, Bahrain, Saudi Arabia, and certain parts of United States and Australia. In such regions, urbanization replaces natural systems with hard surfaces and new green spaces such as urban gardens and parks, resulting in the reduction and fragmentation of habitat [1,2]. These spaces may include ponds, parks, and occasionally even biodiversity-friendly rivers and streams. Many of these spaces serve as a refuge for insect species, particularly those that can live and thrive in small areas under containment and human disturbance [3]. The impact of urbanization on arthropods comes from both fragmentation and diminishing of the rural landscape, leading to a mosaic of green remnant patches of various sizes, quality and composition, in a matrix of urban structures. Overall, the transformation of natural landscapes to urbanized landscapes threatens many arthropod species that are characteristic of natural and rural environments and tends to homogenize assemblages [4,5]. Urbanization also unfavorably affects, or at least modifies, ecological processes and interactions, such as decomposition, nutrient cycling, pollination, soil formation, food for vertebrates, and interactions between insects representing various functional groups [6]. The changing of community structure and interactions is a part of biotic homogenization where, in effect, locally distinct communities and processes are substituted by cosmopolitan communities and processes typical of the urban environment [7,8]. Recent studies that have compared ground arthropod communities of agricultural, residential, industrial and natural remnants have revealed great differences in community structure and ecosystem function [9][10][11].
Invertebrates, particularly arthropods, are ideal organisms for investigating urban biodiversity. Their small size, environmental needs and diversity of life histories make them major components of urban faunas. Many studies have revealed that complex arthropod communities can be found within the urban environment [12,13]. Ants display many advantages over the other arthropod taxa as possible subjects for biodiversity assessment and, as such, are one of the most important groups for studying biological diversity across an urban gradient. Worldwide, they represent approximately 14,701 species and have colonized most of the world's terrestrial biomes [14,15]. They impose a strong ecological footprint through their diverse roles as scavengers, predators, granivores, and herbivores [14]. Ants perform major ecological functions such as predation, scavenging, soil turnover, nutrient cycling and pollination, and are also responsible for dispersal of many plant species [16,17]. In addition, ants are active at almost all trophic levels of the food web [18], making them indispensable for the proper functioning of most terrestrial ecosystems [19]. Habitat disturbances and transformations affect ant communities in many ways, either by changing habitat structure, microclimate and availability of resources or by changing the balance of competitive interactions [20]. Ant assemblages are likely to recover after a few years of development, but their abundance is often characterized by the presence of invasive and cosmopolitan species, and composition is generally different to the pre-disturbance ant community [21]. Many studies of urban ant assemblages have noted the presence of exotic ant species, and most recorded a decrease in resident ant biodiversity due to their presence [22,23].
During the last few decades, the population of Saudi Arabia has substantially increased as a result of the oil boom (3.4% per annum) [24]. This has led to dramatic development of large areas of previously relatively wild lands. This is clearly seen in the Sarawat and Hijaz, and the Northern and Central regions [23]. As a consequence, many indigenous species have been locally lost and many that remain are now highly vulnerable. In spite of this, little ecological work has been done in urban settings. In a study of abundance and diversity of darkling beetles (Tenebrionidae) in Huraymala Wadi of the Central Region of Saudi Arabia, it was found that beetles were significantly more diverse in uncultivated sites compared to cultivated ones and species composition was different between these sites [25]. In a sand-fly (Psychodidae) population study covering parts of Wadi Hanifa, Riyadh Province, it was shown that species numbers and density was higher in undisturbed (southern part of the Wadi in regard to Riyadh city) than disturbed sites (Northern one) [26]. In order to increase our understanding of the impacts of urbanization in this extremely arid region, this study aims to assess the environmental impact of urbanization in Wadi Hanifa (WH) (Central Saudi Arabia) along an urbanization gradient (urban, suburban and rural sites) using ants as bioindicators.

Study area
This study was carried out at Wadi Hanifa, a valley in the middle of the Najd Plateau, Riyadh Province, Kingdom of Saudi Arabia (KSA) (Figure 1a). It is a conspicuous natural landmark that runs for 120 km from the northwest to the southeast, cutting through the city of Riyadh. The Valley depth ranges from 10-100 m, and its width ranges of 100-1,000 m; its catchment area is about 4500 km² [27]. A number of towns and villages lie along the valley, including Al Wassyl, Diriyah, Hair, Irqah, Jubaila, and Uyaynah. The valley passes through palm groves and farms.
In the 1980s, the Riyadh Development Authority (RDA) commenced technical studies alongside the development of a strategy for the valley. These studies were aimed at conserving the valley's natural environment, maintaining the valley as a natural drainage course for Riyadh, using of the valley as a recreational area, and also enhancing and upgrading agricultural use [27,28]. In 2001, RDA developed the Wadi Hanifa Comprehensive Development Plan (WHCDP). The WHCDP formed part of a 10-year program of works that split the valley into five zones, with work ongoing over a distance of some 71 km. From the outcomes and achievements, they have planted 30,000 indigenous shade trees, 6,000 date palms, 50,000 shrubs and ground covers, and transplanted 2,000 large native Acacia trees [27]. The rich diversity of flora and fauna [29] that can be found along the vast areas of the valley has proved very attractive and now the people of Riyadh have started using the valley parks and open spaces in large numbers, as evidenced by the almost full capacity crowds on weekends [27].
The monthly climatic conditions are shown in Figure 1. Hot summers and mild winters characterize the climate of WH, with an average annual temperature of 26 °C and average relative humidity of 24.4% ( Figure 2). The lowest value of 7.5°C is during January and the highest value 42.1°C during July. The rains usually start in December and extend through March, with an annual average of 85 mm [27]. During March and April, WH receives more than half amount of these rains, while no precipitation falls from June to September.
The flora of WH includes native Acacia trees (three species: A. gerradii negevensis var. najdensis Zoh, A. gerradii negevensis var. negevensis Zoh, A. ehrenbergiana Hayne, and A. tortilis Hayne); other shrubs and herbaceous plants are also present. In addition, there are algae and several aquatic plants that grow due to the continuous flow of water through the water channel.

Site selection
The GLOBENET (Global Network for Monitoring Landscape Change) project protocol was followed to assess the impact of urbanization on the biodiversity in WH. Accordingly, 15 sampling sites (Table 1) were selected along an urban-suburban-rural gradient within WH (Figure 1). Urban (WHU) sites were situated in the city of Riyadh, Suburban (WHS) sites were 38 km northwest of WHU, at Al Wassyl town, and Rural (WHN) sites were located in the furthest northwestern part of WH which is located 28 km northwest of WHS, near the town of Uyayna. To prevent pseudoreplication, the distance between sites was set at 300-400 m. The coordinate information for each selected site was recorded by using a GPS unit (Garmin, Montana 650 handheld Global Positioning System).

Sampling methodology
Ants were collected at each selected site using pitfall traps. Traps were 10 cm diameter plastic cups containing 250 mm of 40% propylene glycol as a killing and preserving fluid. Ten traps were set along two lines with 20-50 m between the two lines and 5 m spacing between traps in each line ( Figure 3). This resulted in a total of 150 pitfall traps distributed along the urban-rural gradient. The traps were left open for 7 consecutive days and nights. Sampling was repeated four times during one year (January, April, August, November, 2019). The collected samples were transported to the Museum of Arthropods (KSMA), Department of Plant Protection, College of Food and Agriculture Sciences, King Saud University, Riyadh, Saudi Arabia, for sorting and identification to species level. The number of species, as well as the abundance of each species, was recorded for each trap.

Soil sampling
Within each study site, three cores of soil were collected from the top to 15 cm by soil auger. The collected soil cores were collected along a diagonal line (Figure 3), then mixed together in one plastic bag. This resulted in each plastic bag containing about 1.5-2.0 kg of soil, which was sent for analyses of chemical and physical features at the Soil Laboratory of the Department of Soil Science, College of Food and Agriculture Sciences, King Saud University.

Vegetation measurements
To measure the vegetation cover, 10 1 m 2 quadrats were laid out within each site. The percentages of bare ground, and ground covered by leaf litter, woody debris and plant cover were recorded within each quadrat. Plant species were surveyed once during April, 2019 at each site. The plant specimens were sent for identification to the Herbarium, Department of Botany and Microbiology, College of Science, King Saud University.

Diversity parameters
Ant abundance, richness, evenness and diversity indices (Shannon and Simpson) were calculated using PC-ORD for Windows, version 4.14 [30]. Species richness was taken as the total number of species recorded. The mean of the total count of all individuals for each species collected from each site was used as a measure of abundance.

Data analysis
The following analyses were performed on the ant assemblage composition, using the two ordination techniques, detrended correspondence analysis (DCA) and canonical correspondence analysis (CCA). DCA is an improved eigenvector ordination technique that is based on reciprocal averaging. DCA corrects the two main faults of ordination techniques -arch distortion and violation of the orthogonality criterion [31]. CCA is a multivariate technique that maximizes the correlation between species and sample scores along an assumed gradient [32]. DCA was carried out using the PC-ORD package. Data for all seasons combined were analyzed. Only species that were found at two or more sites were included in the DCA analysis.
Association of both species and sites with the environmental variables was investigated by CCA using the CANOCO statistical software. CCA is a direct gradient ordination technique, where results are simultaneously based on species abundance and environmental variables [33]. The technique selects the linear combination of environmental Similarly, the second and subsequent axes also select the linear combinations of environmental variables that maximize the dispersion of the species scores, but these are subject to the constraint of being uncorrelated with the previous axes. CCA differs from DCA in that the axes are constrained to optimize their relationship with a set of environmental variables. Arrows depict the direction (maximum change) of environmental variables in the ordination, while the length of the arrows shows their degree of influence. The option of species scores as weighted mean sample scores was used in choosing the scaling of CCA ordination scores. The CCA was done in the forward-selection mode of the CANOCO program [33], and the significance of each variable was tested in a sequential fashion using a Monte-Carlo simulation algorithm before it was added to the final model. All variables that were significant at p<0.05 were included in the final model. Only the five most significant flora variables were selected for producing the CCA ordination diagram. Differences in the ant community composition between sites along the three-urbanization gradients were tested by analysis of similarity (ANOSIM, [34]) with a 2-way nested design. Analysis was conducted on the Bray-Curtis similarity matrices [35], with 999 permutations, based on square-root transformed ant abundance data. The ANOSIM analysis was conducted using PRIMER ver. 7.0.17.
Characteristic species (indicator species) were identified for the urbanization gradients using the Indicator Value Method [36]. This method calculated the proportional abundance of a particular species in a particular group, relative to the abundance of that species in all groups. Then, the method calculated the relative abundance of a certain species in a certain group and calculated the proportional frequency of the species in each group. These percentages were regarded as representations of the faithfulness or constancy of presence within a particular group. The two proportions were then multiplied to yield a percentage, which was used as an indicator value for each species in each group. Conversely if either term is low, then the species is considered a poor indicator. Because the component terms are multiplied, both indicator criteria must be high for the overall indicator value to be high. The highest indicator value for a given species across group is saved as a summary of the overall indicator value (IV) of that species and evaluated by the Monte Carlo method, with randomly reassigned SUs (sample units) to groups taking place 1000 times. The probability of a type I error occurring was the proportion of times that the IV from the randomized data set equals or exceeds the IV from the actual data set. The null hypothesis is that IV is no larger than would be expected by chance [37]. The analysis of indicator species was performed using the PC-ORD statistical package. Where possible, natural groupings of sites were visually identified from the clustering of points on the DCA and CCA ordination diagrams.

General trends
A total of 42 ant species were encountered from 24,510 specimens ( Table 2). Ten of these were known or probable non-native, invasive or cosmopolitan species. A voucher collection is deposited in the Museum. Monomorium niloticum, alone made up 38.34% of the total catch and was the most abundant species in all three sampling gradients ( Table  2). This was followed by Cataglyphis holgerseni (11.10%) and Lepisiota simplex (7.14%) and by the less (or rare species) abundant species; Cataglyphis aurata, Monomorium afrum André, 1884 and Tetramorium syriacum each with values of 0.004%. Nine species were unique to rural sites, one to suburban ones, and eight to urban sites; 16 species (38.09%) were shared among the three gradients (Table 2).  4 The ant diversity parameters were relatively constant across the urbanization gradi-5 ent (P > 0.05; Table 3). Site WHN3 showed the highest value for species richness (23 spe-6 cies) and the mean species richness across all sites was 17.3 ant species. For ant species 7 abundance the highest value (68 individuals) was recorded by the suburban site (WHS1). 8 The mean species abundance across all sites was 38.9 ant species. The ant species diversity 9 (Shannon and Simpson) and evenness exhibited their highest values in the rural site 10 WHN2 and their lowest values in the urban site WHU3 (Table 3). However, one-way anal-11 ysis of variance indicated no significant difference between site types for any of the vari-12 ables 13   Figure 4 shows the result of the DCA ordination analysis for ant species. The output 20 for this procedure also explains the ant species that contributed to the grouping of sites 21 on the ordination diagram. The four urban sites (WHU1, WHU2, WHU4 and WHU5) are 22 located near the right-hand end of the first axis while site WHU3 was placed near the 23 center of the graph. These sites were characterised by species Tapinoma simrothi, 24 Brachyponera sennaarensis, Cardiocondyla minutior, Lioponera longitarsus,  iguum and Cardiocondyla mauritanica. The suburban sites and the rural sites lie at the other 26 end of the first axis with no clear separation between them, although they can be loosely 27 divided into two groups (WHS3, WHS4, WHS5 and WHN1) and (WHN2, WHN3, WHN4, 28 WHN5, WHS1 and WHS2) along axis 2; they were characterised by Tetramorium caespitum, 29 Camponotus thoracicus, Camponotus sericeus, Tetramorium sericeiventre and Cataglyphis livida, 30 all of which were common at these sites. The eigenvalues for the first and second axis were 31 0.31 and 0.147 respectively explained 50 % of the total variation in the species data. 32 The ANOSIM analysis confirmed differences between the three urbanization gradi-33 ent sample groups (R = 0.97, P = 0.001). Pair-wise tests revealed significant differences 34 between rural and urban (R = 0.73, P = 0.008), suburban and urban (R = 0.79, P = 0.008), 35 but only marginal significant differences between rural and suburban (R = 0.25, P = 0.06). 36 37 38 Site variables that we measured in this study (flora, soil pH, soil organic carbon, litter 44 cover, litter depth and log) exhibited important differences between habitat types. 45 The CCA for the ant species and flora is shown in Figure 5. The forward selection 46 procedure resulted in the retention of three significant plant species from the 43-plant 47 species present, Phragmites australis (Cav.) Trin. ex Steud. (Sp32) (P < 0.01), Atriplex num-48 mularia Lindl. (Sp51) (P < 0.01), Pennisetum setaceum (Forssk.) (Sp58) (P < 0.01). The other 49 two-plant species Lactuca serriola L. (Sp.36) and Achillea fragrantissima (Forssk.) (Sp.42) 50 were not significant. This separation of sites was to an extent reflected in the site group-51 ings from the DCA analysis, which distinguished three groups. The five rural sites 52 (WHN1, WHN2, WHN3, WHN4 and WHN5) and the two suburban sites (WHS1 & 53 WHS2) were separated from the rest of the sites, which were in turn divided into two 54 groups: (WHU1, WHU2, WHU4 and WHU5) and (WHS3, WHS4, WHS5 and WHU3). The 55 three significant species were the most important factors in both axes and increased to-56 wards four urban sites (WHU1, WHU2, WHU4 and WHU5) and decreased towards the 57 other two groups. The importance of L. serriola increased towards the first group while 58 the importance of A. fragrantissima increased towards the third group (WHS3, WHS4, 59 WHS5 and WHU3). The eigenvalues for the first and second axis were 0.290 and 0.123 60 respectively.  Figure 6 shows the CCA biplot for the ant species and soil variables. Only two out of 68 the seven variables, soil pH (P < 0.01) and soil organic carbon (SOC) (P <0.01) were signif-69 icant. Soil pH increased towards both the rural and suburban sites, which clustered in one 70 group associated with the abundant species Tetramorium caespitum, Camponotus thoracicus, 71 Camponotus sericeus, Tetramorium sericeiventre and Cataglyphis livida. Soil organic carbon 72 increased towards the urban sites with their characteristic species Brachyponera sennaaren-73 sis, Cardiocondyla mauritanica, Cardiocondyla minutior, and Monomorium exiguum and de-74 creased towards the other group. The two urban sites (WHU3 and WHU4) were concen-75 trated at the middle of the diagram, indicating a weaker association with the soil organic 76 carbon variable. The overlap of soil organic carbon and soil organic matter indicates that 77 they were inter-correlated. The eigenvalues for the first and second axis were 0.279 and 78 0.11 respectively. The relation between the ant species and the five ground cover variables; bare 84 ground%; litter cover %; litter depth "cm"; wild plant cover% and log% (dead wood) is 85 shown in Figure 7. The forward selection procedure resulted in the retention of three sig-86 nificant variables from the five variables, namely: litter cover (Litter C) (F = 4.93, P < 0.01), 87 litter depth (Litter D) (F = 2.38, P < 0.01) and the log variable (Log) (F = 2.85, P < 0.01). The 88 three factors showed a positive correlation with four urban sites (WHU1, WHU2, WHU4 89 and WHU5) along with their characteristic species (Brachyponera sennaarensis, Cardiocon-90 dyla mauritanica, Cardiocondyla minutior, and Monomorium exiguum) and a negative corre-91 lation with the other two groups WHS3, WHS4 and WHU3 (two of the suburban sites and 92 one of the urban sites) and the sites WHN1, WHN2, WHN3, WHN4, WHN5, WHS1, 93 WHS2, and WHS5, which represent the five rural sites and three of the suburban sites. 94 The eigenvalues of the two CCA axes were 0.287 and 0.111.

101
Eleven ant species (26.2% of the total) with significance (P < 0.05) were identified as 102 indicators for the urbanization gradient (Table 4). Six species were identified for urban 103 sites, five for rural sites plus two suburban sites WHS1 and WHS2, and no indicator spe-104 cies were found in the suburban sites WHS3-5 (Table 4).

111
Urbanization is intimately associated with habitat modifications, and it is thought 112 that community attributes interact in a number of ways with it [20,[38][39][40][41]. The fact that ant 113 diversity does not vary but species composition does vary across the urbanization gradi-114 ent is one of the main findings of this study. Overall, 42 different species of ants across the three urbanization gradients were 118 trapped. Monomorium niloticum was the most abundant followed by Cataglyphis holgerseni, 119 Lepisiota simplex, Tapinoma simrothi, Cataglyphis livida, and Monomorium abeillei. These spe-120 cies are considered native generalists [42][43][44]. 121 We found that ant diversity parameters (species richness, abundance, evenness, and 122 Shannon and Simpson's indices) did not vary significantly along the urbanization gradi-123 ent. Thus, these findings at variance with both the increasing disturbance hypothesis [45] 124 or the intermediate disturbance hypothesis [46]. However, as in our investigation, several 125 other studies have also failed to support these hypotheses [9,[47][48][49]. 126 The lack of significant difference in ant diversity parameters might be attributed to 127 the difference in reactions of different species to urbanization. Urban habitat is highly sus-128 ceptible to being inhabited by both native and exotic species [50][51][52][53]. Some native ants are 129 able to coexist with invasive ants as a result of many factors. One is the fact that certain 130 invasive species are not yet numerically dominant [54]. This is the case with Monomorium 131 exiguum (30 individuals in urban sites out of a total 24,510); this invasive is at the border 132 of its abiotic tolerance [55]. Secondly, both native species such as Camponotus sericeus, C. 133 thoracicus, Cataglyphis aurata, C. holgerseni, C. minima, Monomorium abeillei,M. niloticum,M. 134 bicolor, M. venustum, and invasive or cosmopolitan species such as Brachyponera sennaaren-135 sis, Cardiocondyla mauritanica, C. minutior, Lioponera longitarsus, Paratrechina longicornis, 136 Pheidole indica, Solenopsis abdita, Tetramorium caespitum, Tetramorium simillimum, and 137 Trichomyrmex mayri occupy different niches (resource partition) and can use different food 138 resources [56,57], or use they can forage at different time (e.g. C. thoracicus, and C. hol-139 gerseni [56,58] or they may possess potent chemical defenses (Brachyponera sennaarensis) 140 [21]. 141

142
Many studies have shown a decrease in ant species richness along gradients of rural 143 to urban forest [50], sometimes associated with reduced size and senescence of habitat 144 fragments surviving in urban areas [59], ranging from parks at urban edges to inner-city 145 squares [52]. By contrast, others have found no decline in ant richness with increases in 146 urban extension or with decreasing size of natural habitat fragments in urban areas 147 [9,47,49]. However, almost all studies have shown clear changes in ant species composi-148 tion in urban habitats compared with nearby natural areas. Although our results have not 149 shown a change in ant richness along the urbanization gradient, our results have demon-150 strated that the species composition does vary between habitat types. The urban habitat 151 has an ant composition that differs between rural and suburban habitats, as evidenced by 152 the DCA results. The lack of relationship between species richness and disturbance might 153 be related to the occurrence of generalist species (e.g. Cataglyphis holgerseni, C. livida, Mon-154 omorium abeillei, M. niloticum) which offests the loss of some disturbance-sensitive species 155 (e.g. Messor meridionalis,Messor picturatus,Crematogaster chiarinii, bauti. These can also be replaced by other more generalist species such as Brachyponera 157 sennaarensis, M. exiguum, Tapinoma simrothi or opportunist species [42,[60][61][62]. In such a 158 way, as Hoffmann [63] has emphasized, the disturbance has caused changes in species 159 composition, but not to species richness. Urban communities can be a subset of the re-160 gional species pool, often biased towards generalists (63.3% of the total species) that are 161 better adjusted to a stressful environment [64,65], or they may be novel to the region, com-162 prising many non-native species [53]. 163 While we recorded few non-native species (23.8%, 10 species) along the entire gradi-164 ent, the urban sites in this study harbored many generalist (63.3%) and open-land species 165 (e.g. Camponotus sericeus, Cataglyphis holgerseni, Cataglyphis livida, Lepisiota simplex, 166 Monomorium niloticum), which is in line with other studies (e.g. [64,65]). Urban sites were 167 inhabited by ant species that were practically absent from the rural and suburban sites. 168 The unique aspects of the ant composition in urban sites can be partially explained by the 169 occurrence of species pre-adapted to urban environments, such as Cardiocondyla mauritan-170 ica and C. minutior [66], invasive generalized foragers such as Monomorium exiguum [44],0 171 native generalized scavengers such as M. afrum and M. bicolor; and an invasive arboreal 172 twig-nesting Lioponera longitarsus [67]. 173 The lack of variation in overall species abundance or richness or diversity indices 174 across the gradients should not be considered as unresponsiveness of ant communities, 175 but rather as a weakness of such variables as indicators of environmental impact. By con-176 trast species composition changed in all studies where it has been used (e.g. [68][69][70]). 177

Environmental variables 178
In rural-suburban-urban gradients, variables including habitat, landscape, and com-179 petitive interactions are the main groups of factors that affect ant community composition 180 [20,23,59]. Habitat fragmentation [71,72], the degree of urban development [50,73,74], soil 181 properties and vegetation structure [23,75,76], and the role of exotic ant species 182 [22,23,77,78] are important determinants of ant community responses to urbanization. Site 183 environmental variables, such as flora, ground cover, and soil properties measured in our 184 study may explain much of the variation in the ant composition. 185 In accordance with our results, earlier studies have indicated that relatively high an-186 thropogenic disturbance favours the occurrence of various types of plants in urban areas 187 [79,80], often as a result of the introduction of exotic plant species, such as Atriplex num-188 mularia (Chenopodiaceae), Pennisetum setaceum (Gramineae), and Phragmites australis (Po-189 aceae) [81,82]. As shown by the CCA analysis, the difference in plant composition could 190 possibly explain the difference in ant community on both the rural and the suburban than 191 those of the urban areas [83]. 192 Based on our CCA results, several soil elements (soil pH, SOC and SOM) explained 193 the part of the variation in the ant community. These soil attributes could affect ants di-194 rectly via their nesting activities or indirectly via their effect on plants [84]. In general soil 195 organic matter (SOM) often increases in perennial vegetation [85]. Moreover, a lower soil 196 pH in urban sites is related to a higher amount of organic matter and available nutrients 197 [86], which enhances the suitability of nest sites [87]. All the urban ant species revealed 198 some association with soil factors. The abundance of Brachyponera sennaarensis, Cardiocon-199 dyla mauritanica, Cardiocondyla minutior, and Monomorium exiguum abundance were posi-200 tively related to high concentrations of SOC, SOM and lower pH. In the rural and subur-201 ban sites, abundance of Tetramorium caespitum, Camponotus thoracicus, Camponotus sericeus, 202 Tetramorium sericeiventre and Cataglyphis livida were negatively correlated with high SOC 203 and SOM and positively with increased pH. 204 The conversion from native flora to perennial vegetation and the irrigation in un-205 managed urban sites leads to the increase in litter cover and depth and log percentage, 206 which all showed a significant effect on the variation among the different studied sites in 207 the CCA result. Logs are an important nesting resource for both leaf-litter and arboreal 208 ant species [88,89]. Many ant species that forage in the leaf litter use logs for protection 209 and foraging and to enlarge their colonies [90][91][92]. In accordance with several studies [93-210 95], we found a positive correlation between the ant species, Brachyponera sennaarensis, 211 Cardiocondyla mauritanica, Cardiocondyla minutior, and Monomorium exiguum in urban sites 212 and the presence of deep litter cover and logs (dead wood). Discrete differences in ant community structure were seen across the sites. Urban 215 sites with the highest degree of disturbance differed in their composition from the other 216 rural and suburban sites. Eleven indicator species were identified for different habitats, 217 suggesting that these predominant species are key to shaping ant community composition 218 and indicating that the variation within each habitat is driven by shifts in the highly 219 abundant species. These indicator species showed different ecological behavior within 220 each habitat [90,91]. Six predominant urban ant species have been identified as urban bi-221 oindicators that can easily be found in high abundance and can adjust their nesting habits 222 to different human environments. Brachyponera sennaarensis is a general scavenger tramp 223 species that nests in humid urban and disturbed sites next to human settlements, espe-224 cially where waste ground and rubbish dumps exist [43,60,96]. Tapinoma simrothi is a na-225 tive species, a generalized forager and nests among roots of Gramineae plants in wild 226 habitats of the Arabian deserts, where it attends mealybugs [97]. It is observed attending 227 aphids and protecting them from predators [98]. Cardiocondyla minutior is a tramp species 228 that nests directly in humid soil of disturbed sites and date palm plantations of the Ara-229 bian Peninsula and in the Socotra Archipelago where soil is rich in waste of domestic live-230 stock [99]. Little is known about the biology of this species but the majority of Cardiocon-231 dyla species are known to inhabit anthropogenically or naturally disturbed sites [66]. The 232 predatory ant, Lioponera longitarsus; is another tramp species, that build nests directly in 233 ground with a single, small entrance hole [100]. However, Lioponera longitarsus is known 234 to nest in hollow twigs [67] and this nesting habit facilitates species dispersal when or-235 ganic material is moved around by humans. Monomorium exiguum is a native species, and 236 is a generalized forager that inhabits the leaf litter and topsoil layer in public gardens, date 237 palm plantations, and urban sites near to human settlements where there is plenty of or-238 ganic matter [99,101]. 239 Five indicator native species showed an affinity towards both rural and suburban 240 environments, namely Cataglyphis livida, Camponotus sericeus, C. thoracicus, Tetramorium se-241 riceiventre, and T. caespitum. These are generalized scavengers, that build nests directly in 242 the ground under stones and other objects next to Acacia and Calotropis procera (Apocyna-243 ceae) trees of the Arabian Peninsula [102]. 244

245
Our results indicate that the ant diversity parameters do not vary appreciably along 246 the three-urbanization gradient (urban, suburban and rural). By contrast, discrete differ-247 ences in ant community structure exist across the urbanization gradient. Urban sites with 248 the highest degree of disturbance differed in their composition from the other rural and 249 suburban sites. Environmental factors such as vegetation type (native and exotic species), 250 soil properties (soil pH, SOC and SOM) and ground cover (litter cover litter depth and 251 wood debris) proved to be important determining factors of ant species composition. As 252 with similar investigations, this study demonstrates that ants are good indicators of ur-253 banization effects and remain successful and dominant in all habitats.