The study was conducted on a natural monoclonal poplar stand in the plain of Mitidja (northern Algeria). It is the largest sub-coastal plain of Algeria, with a surface of approximately 140 000 hectares, about 100 km length, and a width varying between 5 and 20 km. On the Northern side, the plain is separated from the Mediterranean sea by the Sahel ride, and on the South, by the Blida Atlas mountain [27].

Two stations have been identified: the Soumàa station, on the foothills of the Blida Atlas is at an altitude of 200 m A.S.L. and south-exposed. It spreads on 4 hectares, and the vegetation is semi-open to open where a shrub layer prevails. The main noticeable tree species are Acacia mimosa, Eucalyptus globulus, Pinus halepensis and Ceratonia siliqua. We also note herbaceous species, as Avena alba, Calendula arvensis, Torilis nodosa, Biscutella didyma, Sinapis alba, Euphorbia helioscopia, Heliotropium europaeum and Borrago officinalis. The Koléa station is located on a surface covering about 6 ha, on the south side of the Sahel at an altitude of 250 m A.S.L. and south-exposed. It has an open vegetation, with an herbaceous layer characterized by the following species: Cyperus rotundus, Centaurea calcitrapa, Dittrichia viscosa, Oenanthe fistulosa, Andryala integrifolia and Sonchus oleraceus.

The climate of both stations is Mediterranean with a continental trend (humid stage with cool winter). The precipitation, mostly observed in winter and spring, is characterized by a great inter-annual and inter-monthly irregularity. The coldest month is January (mean minimum temperature of 4 °C and 5.1 °C, and mean maximum temperature of 8.92 °C and 6.66 °C at Soumàa and Koléa, respectively). The warmest months are July and August with mean minimum temperatures of 26.5 °C and 24.8 °C and mean maximum temperatures of 34.8 °C and 32.3 °C, respectively at Soumàa and Koléa (Fig. 1).

The campaign study (2002–2003) was characterized by an annual rainfall of 806.2 mm at Soumàa and 764.4 mm at Koléa, exceeding by almost 120 mm and 102 mm respectively the means observed in the last 10 years. Moreover, the distribution of these rainfalls was irregular, since more than 79% of rain fell between November 2002 and February 2003. The low rainfall recorded between June and October, together with rising temperatures and radiation, result in a strong evaporative demand [28].

Black poplar is a pioneer species able to settle on fresh environments. The trees are found as spontaneous in alluvial plains, and their growth needs a long favourable period with warm and sunny summers. When these climatic conditions are fulfilled, the budding is precocious and the vegetation growth is sustained until the end of summer [29].

The sampling of entomofauna living on the unit areas defined by the poplar stands began in September 2002 until August 2003, covering a twelve months period. Every 15 days, we carefully examined the aerial compartments of poplar trees (trunk bark, twigs, leaves and galls). Due to a collection orderly and fairly comprehensive in space and time, we have defined for each station 3 transects of 100 m in length (disposition of trees given in Fig. 2), separated by 10 m [30]. In order to get accurate data of community structures, we followed the sampling stratification method described by Atger [31] who showed the good representation of the obtained data. At the level of leaves and twigs of each sampled tree, destructive samplings of 20 cm-long branches bearing about 20 leaves were done in each cardinal directions, at 1.80 m above the ground, so 4 sub-samplings were considered per tree. All the encountered specimens (insects, spiders and galls developed on leaves) were recorded. Fourteen yellow water traps were also placed per station (one 11×13 cm yellow plate per tree at 1.50 m above the ground), and renewed every fortnight for counting. The assignation of each collected species to the different compartments were decided according the specific habitat determined by visual samplings. For example, Tapinoma nigerrimum specimens were added to the leave compartment as this ant is mobile on the trunks but stationary at the level of leaves. We completed these collections by the use of a sweeping net at about 2 m above the ground (5 min as unit of effort per tree, i.e. nearly 4 hours of investigation in all).

All the specimens collected were put into jars containing formalin at 33%, and then kept in the laboratory to be identified and quantified. The Diptera were not taken into account due to identification difficulties.

We used the Michaelis–Menten curve as asymptotic function to estimate the richness of each compartment, following the procedure explained in EstimateS 8.0 manual [32].

The correlation between the abundance of species and the phenological variations in Populus nigra stands was analysed by a detrended correspondence analysis (DCA), followed by a hierarchical classification from Euclidean distances calculated on the coordinates of the first 3 axes of the DCA. It helps to consider the differences in composition and assembly of a sample [33]. The analysis was conducted with the software PAST. vers.1.81 [34], on a matrix based on the abundance (direct and indirect samplings) and periods of investigations in the study areas.

To explore the structure of communities living in the different compartments of Populus nigra, rank-abundance data were plotted and compared to the Motomura model: log(N)=aR+b, where N is the total abundance for a species and R the rank [35]. Slope between communities was compared using the procedure described in PAST vers.1.81 [34]. Briefly, an analysis of covariance was conducted considering the slopes as means and squares of standard errors of x as the variances. The probability was calculated with corresponding Barlett's test.

The significance of temporal variations of logtransformed abundances for each community or trophic group was tested by Kruskal–Wallis tests considering all the species belonging to a community or a trophic group.

The adjustment of time variation abundances (log-transformed) to a normal distribution was assessed by calculating mean and variance. The mean was obtained by calculating the barycentre of the data, using the formula: B=Σ(Mi∗ln(1+abundance))/Σ(ln(1+abundance)), where Mi is the number of the ith month, from 1 to 12. The variance was estimated by successive approximation using a least square method. The significance of the lag between temporal distributions of functional groups was assessed by calculating cross-correlation using PAST vers.1.81.

The significance of Jaccard index between 2 communities was assessed by a double resampling method (Permutjac.exe program running on DOS, available on request). Briefly, the Jaccard index between 2 columns is compared to the sorted 1000 ones resulting from resamplings. In each resampling, 100 independent permutations affecting each column are conducted.

A total of 6707 individuals, distributed in 73 species belonging to 35 families were collected in the two study sites. The class of Hexapoda dominated the class of Arachnida with 55 and 18 species respectively. The main orders revealed by this investigation are the Homoptera, Hymenoptera, Coleoptera, and Araneae. Based on the knowledge of their diet and therefore their trophic position, the species fall within primary, secondary, tertiary consumers and trophobionts, in the 4 compartments previously defined (Table 1 and Appendix A). The secondary consumers appeared to harbour the greatest richness with an amount of 35 species. The percentage of primary consumers is relatively similar in the different compartments, often exceeding 80%. Although they belong to different families, they are all specific to Populus genus. Fifty five percent species are opophagous Hemiptera, followed by about 25% of folivorous Lepidoptera, and then folivorous and xylophagous beetles with an abundance of 20%. The Coccinellidae and the Araneae are by far the dominant group among secondary consumers, whereas species belonging to orders Mantidae and Hymenoptera are the most abundant tertiary consumers. These generalist auxiliary species can be found in tree and shrub layers. Finally, trophobionts, especially the families of Formicidae and Apidae, are most of the time confined to Aphididae populations.

In a first time, the table of samplings including the 4 compartments was submitted to a DCA. In the factorial plan defined by Axis 1 × Axis 2, the distribution of species is dispersed (Fig. 3), indicating that the different components of Poplar entomocenosis seem to react differently to seasonal constraints. The ordination plan and the hierarchical classification based on the first 3 axes of DCA (data not shown) show a non-seasonal assemblage, distributed at the centre of the projection, and four well separated seasonal ones, named “early spring”, “spring-summer”, “late summer” and “autumn” on the base of the concerned months. Among the latter assemblages, the autumn one is situated at the right side of axis 1. At the opposite, the summer assemblage is projected toward the negative values of axis 2, whereas early-spring and spring summer ones on the positive values.

The same ordination plan shows a superposition of projections stations of Koléa and Soumàa between May and November, a period during which we observe the highest diversity and densities, reflecting the overall composition consistency of both stations (Table 2). This uniformity has led us to consider the mean abundances of all species for each group over the 2 stations. The estimated richness of each compartment during the May–November period (Table 3) is very close to the corresponding observed numbers of species, providing a good confidence in our samplings.

For each compartment, the adjustment of rank-abundance (log-transformed abundances) diagrams to the geometric series of Motomura was assessed by calculating Pearson coefficients. Unexpectedly, two communities are identified in the leaves compartment, significantly differing by their respective slopes (P=3.81×10−36^∗∗∗). In contrast, the other three compartments contain only one community, as proved by homogeneous rank-abundance profiles. It appears that the 5 communities (Leaves 1, Leaves 2, Twigs, Galls and Trunk, named respectively L1C, L2C, TwC, GC and TrC in the following) are well modelled by the series of Motomura (Table 4).

The slope comparisons establish significant differences between the structures of the different communities, except for L1C compared to GC, and marginally for GC and TrC (Table 4 and Fig. 4). The richest and most balanced community is L2C, comprising 37 species, while the poorest one is L1C with only 5 (Chaitophorus leucomelas, C. populi-albae, Monosteira unicostata, Tapinoma nigerrimum and Crematogaster scutellaris). In L2C, the abundance ranks of primary consumers are lower than those of tertiary consumers, indicative of a significantly higher average abundance (Kruskal–Wallis test, P=0.043). In the same community, the trophobionts have marginally significant lower abundance ranks than tertiary consumers (P=0.073). As for TwC, no significant difference was recorded in the average abundance ranks between trophic groups.

The calculation of Jaccard similarity indices (Table 5) shows that the faunal compositions differ highly significantly between communities. Indeed, only three species are shared by 2 compartments: Sphodromantis viridis (tertiary consumer) is common to TwC and L2C, Oenopia doublieri (secondary consumer) to TwC and GC and Crematogaster scutellaris (trophobiont) to L2C and GC.

The position of abundance–time curves, their forms and their limits illustrate how species choose the compartment in which they specialize; this adjustment ensures succession meaning and the speed of their dynamics. In each community, we have separated the different functional groups in order to better visualize their temporal variation, using a synthetic curve (average abundances of different species).

The L1C comprises three opophagous Hemiptera (Chaitophorus leucomelas, C. populi-alba and Monosteira unicostata) and two trophobiont ants (Tapinoma nigerrimum and Crematogaster scutellaris). Their density are maintained at a high level throughout the season, and show no significant temporal variation (Kruskal–Wallis test, P=0.425).

In L2C, the primary consumers are organized into two groups, significantly different by their temporal abundance variations (Kruskal–Wallis, P=0.0368). The early group, peaking in April (Fig. 5a), includes mostly opophagous species leaving on galls (Pemphigus versicarius, P. bursarius and P. protospirae). The following group culminates in July and comprises mainly folivorous species (Coleoptera Chrysomelidae; Lepidoptera Lyonetiidae, Lymantridae, Lithocolletidae, Stigmellidae and Notodontidae). The secondary consumers also show significant temporal variations, but with a unimodal distribution (P=5.14×10−6), peaking in June (Fig. 5b), where 3 groups follow each other with a significant time lag (Table 6): Araneae, Anthocoridae–Coccinellidae and Hymenoptera (cross-correlation tests, P=5.15×10−4 and 2.24×10−2 respectively). Tertiary consumers (mostly Mantidae) and trophobionts do not show significant variations over time (Kruskal–Wallis test, P=0.884 and P=0.425 respectively).

At the level of TwC, the primary and secondary consumers are recorded throughout the year, without significant temporal variation (Kruskal–Wallis test, P=0.120 and P=0.595 respectively). However, within each type consumers in TwC, 2 groups can be identified and well separated, the first in spring-summer and the other in autumn (Table 6). In contrast, GC contains mainly secondary consumers where no subgroup can be defined, and no temporal variation observed (Kruskal–Wallis, P=0.844).