EUROPEAN JOURNAL OF ENTOMOLOGY EUROPEAN JOURNAL OF ENTOMOLOGY

. This study investigated the association of environmental factors with insect richness and nested patterns in the distribution of species of insects on small islands (northeastern Algeria). Insect species on eleven (11) islands were sampled using pitfall traps, sweep netting and a Japanese umbrella. To explain patterns in insect diversity on small islands, a generalized linear model (GLM) was used to determine the association of the overall richness and that of the dominant orders of insects with environmental variables (altitude, surface area, isolation, plant richness and number of habitats). To determine the nestedness pattern on these islands, the online interface NeD was used to measure total matrix nestedness. Of the variables measured, only plant richness was associated with variations in the overall species richness and that of the ﬁ ve dominant orders. The overall number of insects and that of the ﬁ ve dominant insect orders were signiﬁ cantly nested in all null models except the very restrictive FF model. Plant richness accounted for the nested structure of the overall number of insects and that of the ﬁ ve dominant orders (Coleoptera, Hymenoptera, Hemiptera, Diptera and Lepidoptera). Surface area was also important in the nested structure of overall number of insects and that of the order Lepidoptera. The nestedness of the overall number of insects and that of the ﬁ ve dominant orders of insects, however, did not result from passive sampling and selective colonization. The most eﬀ ective way of conserving overall number of insects and that of the ﬁ ve dominant orders of insects on the Algerian islands studied is to protect their plant richness and to maximize the overall number of species and that of Lepidoptera on these islands, conservation should concentrate on the largest islands.


INTRODUCTION
Island biogeography theory and nested species distribution are two important concepts in biogeography and conservation (Lomolino & Smith, 2003). The theory of island biogeography considers that variation in specifi c richness depends on the area and isolation of islands (MacArthur & Wilson, 1967). This theory predicts that species richness decreases with decreasing island area and increasing isolation as these two variables infl uence immigration and the extinction equilibrium (Rosenzweig, 1995). Some authors reject island biogeography theory, as it does not consider the particularities of each species, diversity of ecosystems and their balance, as well as island-specifi c characteristics (Brown & Lomolino, 2000;Lomolino, 2000;Anderson & Wait, 2001;Lomolino & Weiser, 2001).
The major reason for studying island species richness and distribution was to determine the degree of nestedness in an insular environment. Nestedness occurs when assemblages at poor sites are comprised of species that constitute subsets of species that occur at rich sites (Patterson & Atmar, 1986). Nestedness is common in nature, especially, on islands (Patterson & Brown, 1991;Atmar & Patterson, 1993;Worthen, 1996;Wright et al., 1998). The results ob-variables associated with the variation in the overall richness of insects and that of the dominant orders, as well as those associated with the nested structure recorded for the islands studied, could be used for defi ning the management policy for protecting these fragile ecosystems.

Study area
Along the Algerian coast, from Annaba (East of Algeria) to Oran (west of Algeria), there are 196 islands, islets and rocks (Vela, 2016). Of the Islands in the west of Algeria the largest are: Habibas island (60 ha) and Rechgoune (40 ha). Information on insect species diversity is recorded for only eleven (11) islands ( Fig. 1, Table 1). Because of the little data for the rest of the islands, information was collected from published and unpublished fi eld records (Aissat & Moulaï, 2016;Moulaï et al., 2016;Aissat et al., 2023) and Master students' studies (Lachouri et al., 2016;Amrane et al., 2017). The islands included in this study ranged from 7 to 50 m in altitude, and from 0.2 ha to 6 ha in size. Island area, isolation and altitude were obtained from a Google map. Area (m 2 ) was estimated as the planar surface of the islands. Isolation of islands (m) is measured in terms of the shortest distance between an island and the shoreline (Table 1). Information on plant richness for 11 islands was obtained from the following publications (Benhamiche-Hanifi & Moulaï, 2012;Ghermaoui et al., 2016;Lachouri et al., 2016;Amrane et al., 2017).

Insect sampling
The inventory of the insect fauna took place in spring and summer. Insects were collected for 40 days (20 days for each year Island biogeography theory and nestedness have been used to investigate the patterns in insect species richness on islands in the north of the Mediterranean Sea (Dennis et al., 2008;Fattorirni, 2010Fattorirni, , 2011, but not in the south of the Mediterranean Sea, other than for Algerian island breeding birds (Aissat, 2020).
The fauna on islands is the least known of all Algerian ecosystems. Thus, recently there have been a few studies on the faunal diversity on some islands off the coast of Algeria (Aissat & Moulaï, 2016;Aissat, 2020;Aissat et al., 2023). Many of these islands are subjected to several disturbing factors such as the presence of large colonies of gulls and the fl ow of visitors throughout the year, especially in summer.
Understanding the functioning of these island ecosystems and defi ning the factors associated with the diversity of insects is important for determining the measures necessary for protecting their entomofauna and the conservation of their fragile habitats.
This study aimed to highlight the association of environmental variables with insect diversity on islands, and determine if the pattern in nestedness in the overall distribution of insects and that of the dominant insect orders (Coleoptera, Hymenoptera, Hemiptera, Diptera and Lepidoptera) is similar on small islands. In addition, how nestedness among these assemblages is associated with environmental variables was determined. Identifying the environmental sampled) on 11 islands during April and July of the year 2016-2017. Sites sample per island was related to their surface area and habitat heterogeneity. Islands with large surfaces (Grand Cavallo island: 6 ha, Petit Cavallo island: 4 ha; Collo island: 2.8 ha) were surveyed for 5 days and those with small surfaces (not exceeding 1.2 ha) were surveyed for 3 days (Table 1). It was not possible to sample the islands in the diff erent regions at the same time due to logistical reasons and the distances between the islands. Surveys were conducted during daytime between 8:00 and 16:00 h. In total, 98 sites were sampled, which included all the habitats in the areas studied (high mattoral, herbaceous strata, rocky zones and sand dunes) ( Table 1). Most of the sites were sampled using three sampling methods. For ground dwelling insects pitfall traps were used, which consisted of plastic cups, 10 cm in diameter and 15 cm deep, which were fi lled to one third of their capacity with a mixture of seawater (80%) and ethanol (10%) and laundry powder (10%). Forty (40) pitfall traps were used on the largest island and 20 on the smallest island, with a minimum distance of 5 m between traps. The contents of pots were recorded according to the time of sampling. The insects in the herbaceous layer were collected by sweep netting for 10 min. For each island there was a minimum of twenty 10-min samples and sampling continued until no new species were recorded. Insects on shrubs and trees were collected by using a Japanese umbrella, which consisted of a square piece of canvas of light colour (120 × 120 cm) stretched on a folding wooden frame. The umbrella was held with one hand under the foliage of shrubs and branches of trees which were rigorously shaken by the other hand. Each sample consisted of 20 min of collecting between 08:00 and 16:00. Active collection took place within a 200 m radius of the pitfall traps. Insects caught sweep netting were preserved in 70% ethanol. For each type of sampling, collections were considered complete when no new species were recorded for some time. Collected insects were identifi ed using the reference collection of the laboratory of Applied Zoology and Ecophysiology of the University of Bejaia (Algeria), and available literature: Portevin (1924);Perrier (1935Perrier ( , 1961Perrier ( , 1963Perrier ( , 1964Bernard (1968);Plateaux-Quénu (1972);Delvare & Aberlenc (1989); Auber (1999).

Data analysis
Sampling eff ort Cook (1995) suggests that before carrying out a nestedness analysis, it is fi rst necessary to verify that the list of species is as complete as possible. To estimate the true species richness for each island a non-parametric chao1 estimator was used. Past software was used to generate the estimate of the true species richness.

Collinearity and relationships of the environmental variables
To verify collinearity between independent variables a multiple regression analysis was used. Myers (1990) suggests removing the variables with the highest variance infl ation factor (VIF) values and redoing the analyses until all remaining variables obtain a VIF value below the designed threshold (i.e. < 10). The level of collinearity between the environmental variables was quantifi ed with the variance infl ation factor calculated using Exelstat software. For this analysis all the variables were log-transformed.
The eff ects of island area, island isolation, altitude, plant richness and number of habitats on overall number of insects and that of species in the dominant orders were analysed using a generalized linear model (GLM). All analyses were done using Excelstat software.

Nestedness analysis
Nestedness analysis was carried out on the overall number of insects and that of each of the fi ve dominant insect orders, Coleoptera, Hymenoptera, Diptera, Lepidoptera and Hemiptera. The hypothesis tested was that overall number of insects and that of the dominant orders are nested within each other based on the overlap and decreasing fi ll (NODF) values, which range from 0 to 100, with large numbers indicating increased nestedness (Almeida-Neto et al., 2008;Ulrich et al., 2009).
Nestedness data for overall number of insects and that of the dominant orders was organized in binary presence / absence matrix, where each row was a species (n Overall species = 216; n Coleoptera = 68; n Hymenoptera = 48; n Hemiptera = 48; n Diptera = 36; n Lepidopptera = 16) and each column an island (n = 11). The online interface "NeD" was used to calculate the nestedness value (https://ecosoft.alwaysdata. net/) developed by Strona et al. (2014). Five null models were run to test the signifi cance level of nestedness: EE: equiprobable rows and column totals null model; CE: proportional rows and column totals null model; EF: equiprobable rows totals and equiprobable column totals null model; FE: Fixed row totals and equiprobable column totals null model; FF: Fixed column and row totals. The FF null model is highly restrictive and EE the least restrictive (Ulrich & Gotelli, 2012;Strona & Fattorini, 2014;Matthews et al., 2015). All of these null models have strengths and weaknesses (Ulrich & Gotelli, 2012;Strona & Fattorini, 2014). The random placement model was used to test the passive sampling hypothesis. According to this model, the number of species S(α) in a given relative area α, α = AK / ∑ k k-1 AK, and the overall abundance n 1 , n 2 ,...ns of the S species in collection C (Coleman, 1981;Coleman et al., 1982): The random distribution hypothesis should be rejected if more than one-third of the points lie outside one standard deviation of the expected curve (Coleman, 1981;Coleman et al., 1982).
To test the presence of nestedness in the overall number of insects and that of the fi ve dominant orders, the order in which the islands were sorted using NeD software were correlated with that based on environmental variables. Partial Spearman's rank correlation was used to determine the associations between island area, island isolation, plant richness and number of habitats, in order to test: (1) the selective extinction of species hypothesis, (2) diff erential colonization of habitat hypothesis and (3) habitat nestedness hypothesis. All analyses were done using IBM SPSS software (IBM Corp., Armonk, NY, USA).
The number of species collected on each island varied between 15 and 140 species. Survey completeness for the 11 islands ranged between 69.87% and 92.59%. According to the estimated species richness, for 7 of the 11 islets more than 80% of the species were collected and for the rest the mean values were higher than 74.38% (Table 1). This indicates that in general the islands studied were well sampled.

Relationships of environmental variables and nestedness analysis
The VIF values obtained in the fi rst analyses were acceptable (VIF < 10) ( Table 2). Thus surface area, island isolation, plant richness, altitude and the number of habitats were used in the subsequent analysis.
GLM test indicated plant richness was the only variable signifi cantly associated with variation in overall number of insects and that of the dominant orders (Coleoptera, Hymenoptera, Hemiptera, Diptera and Lepidoptera) ( Table 3).
The overall number of insects and that of the fi ve dominant insect orders were signifi cantly nested according to all null models except the very restrictive FF model (Table 4).
The nested structure of the overall number of insects and that of the fi ve (5) dominant orders was not attributable to passive sampling (Fig. 2). For the overall number of insects and that of the fi ve dominant orders, more than one third of the observed data points lay outside ± 1 SD of the expected species-area curves computed using the random placement models (Fig. 2).
There was a strong correlation between island rank in terms of nestedness of overall numbers of insects and that of Lepidoptera with that in terms of surface area and plant richness. The island nested ranks for four (4) orders of insects (Coloeptera, Hymenoptera, Hemiptera and Diptera) were also strongly correlated with island rank in terms of plant richness (Table 5).

DISCUSSION
This study revealed the environmental variables associated with the variation in the overall number of insects and that of the fi ve dominant orders recorded on Algerian islands and the occurrence and causes of nestedness.

Relationships between species richness and environmental variables
The overall number of insects and species richness of the fi ve dominant orders was strongly correlated with plant There was no general relationship between altitude and overall insect species richness. The low altitude of most islands (< 30 m), however, resulted in them being fl ooded periodically, which has resulted in the formation of temporary ponds on these islands. This may favour the transfer of marine resources that could infl uence the structure of their insect communities. Several studies report that small islands generally benefi t from the transfer of resources from aquatic environments (Anderson & Wait, 2001;Jonsson et al., 2009). Barrett et al. (2003) even suggest that the transfer of so called aquatic subsidies can aff ect the species richness of an island.
In Algeria, the Yellow-legged gull (Larus michahellis) is a disturbing factor, especially in coastal ecosystems, because they can be very abundant and their decaying waste has a strong eff ect in terms of soil acidifi cation, which in turn favours the establishment of invasive species of plants (Vidal, 1998). On the islands studied, the eff ect of gull colonies on insect species richness is refl ected clearly in the presence of a few families like the Calliphoridae, Muscidae and Sarcophagidae, which are capable of exploiting the decaying corpses of gulls.

Patterns in nestedness
The nested structure recorded was not related to passive sampling. Although as a general rule the data collected should be examined for passive sampling before testing hypotheses, in the case of island insects this is rarely done probably because of the lack of or the diffi culty of obtaining data on the abundance of species (Xu et al., 2017). In the present study, passive sampling plays no role in the nested structure of the overall number of insects and that of the fi ve dominant orders.
Simberloff (1976) states that the nested structure of assemblages of insects on islands, particularly depends on the capacity of dispersion of diff erent taxa. In the case the Table 4. Comparative analyses of nestedness for overall number of insects and that of fi ve dominant orders, on the 11 Algerian islands studied. Nestedness metrics and related parameters are provided for overall number of insects and that of the fi ve dominant orders. Pvalues were based on 1,000 Monte Carlo simulations: EE -equiprobable rows and column totals null model; CE -proportional rows and column totals null model; EF -equiprobable rows totals and equiprobable column totals null model; FE -fi xed row totals and equiprobable column totals null model; FF -fi xed column and row totals null model SD, standard deviation; * signifi cant nestedness (P < 0.05), ** (P < 0.001), *** (P < 0.0001). islands studied, the nestedness did not appear to result from selective colonization, as it was not correlated with isolation. Many species of insects can actively disperse (Murakami & Hirao, 2010), often over great distances (Carlquist, 1965;Holzapfel & Harrell, 1968;Lee et al., 2008), which may weaken correlations between isolation and species nestedness. In this study, particular insect orders appear to be good island colonizers and are present even on islands far from the mainland on which they are abundant.

Matrix
Consequently, the absence of particular patterns of colonization may be attributable to their ability to disperse. It is necessary to consider the role of human activity in the general structure of some taxa on the islands studied. Although the islands studied are not inhabited, they are connected to the mainland as fi sherman use their boats to ferry visitors to the islands. Thus species of continental origin with limited powers of dispersal, like ants, could to be unintentionally introduced on to small islands (Blard et al., 2003). That the analysis of nestedness was done for the overall number of insects and that of the fi ve dominant orders, provides an opportunity of determining how they are structured among islands and with which environmental variables they are associated. The overall number of insects and that of the fi ve dominant orders were all signifi cantly nested, but diff erently associated with the environmental variables. Plant richness was the single best variable infl uencing the nestedness of the overall number of insects and that of fi ve dominant orders, because almost all of these insects assemblages are phytophagous, e.g. Chrysomelidae, Coccinellidae, Mordellidae, Apidae, Colletidae, Miridae, Pentatomidae, etc. Thus, the plant richness on the islands might determine the distribution of the overall numbers of Coleoptera, Hymenoptera, Hemiptera, Lepidoptera and Diptera. The association of plant richness with the nestedness pattern may account for the eff ect of environmental variables on the nested structure of insect distribution. Island surface area, however, was associated with the nested structure of overall numbers of insects and Lepidoptera. These results indicate island surface area is important in determining the overall distribution of insects and species of Lepidoptera. The importance of island surface area on the pattern of nestedness in overall number of insects and species of Lepidoptera indicate the role of selective extinction of species in structuring the overall number insects and species of Lepidoptera on the islands studied. Similarly, Murakami & Hirao (2010) indicate the relative importance of spatial factors (island isolation) and environmental factors (vegetation height and island area) in determining the nested distributions of insect species on small islands in the Bahamas.
In addition, the nested distribution may be due to other factors. Coleoptera are the fi rst to colonize environments (Jeff ries, 1994), thus older environments are more likely to harbour more rare species (Fattorini, 2011). Unfortunately, there is no information on the age of the islands to confi rm this. Furthermore, in terms of colonization and dispersal ability, aquatic Coleoptera quickly colonize new environments (Eyre et al., 1992). For Diptera, decomposing carcasses in gull colonies on all islands probably contribute to the diversity and distribution of insects. In this context, at least at the scale of an island, Morrison (2005) emphasizes the importance of the decomposition of marine algae in the Bahamas islands for the distribution Diptera.

CONCLUSION
This study revealed the association between plant richness and the variation in insect species richness and that of fi ve dominant insect orders (Coleoptera, Hymenoptera, Hemiptera, Lepidoptera and Diptera). The mechanisms determining nestedness diff ered for the overall number of insects and that of fi ve dominant orders of insects. These results have implications for the conservation of insect diversity on the Algerian islands studied. First, for the overall number of insects and that of the fi ve dominant orders of insects, plant richness is the best predictor. Accordingly, the plant richness on these islands should at least be pro-tected from the eff ect of colonies of Yellow-legged gulls and human activity. Second, as surface area is signifi cantly associated with nestedness the larger islands are the most important in terms of conservation. (Algeria) is thanked for its support and two anonymous reviewers for providing constructive feedback that helped improve the manuscript. Moulaï R., Chelli A., Hamadi K., Amrane H., Ihdene A., Lachouri T. and Mouloudj L. are thanked for providing unpublished data.