Arboreal arthropod biodiversity in woodlands. II. The pattern of recovery of diversity on Melaleuca linariifolia following defaunation
Abstract
This study examines the level of isolation of arthropod faunas present on specimens of the endemic woodland tree Melaleuca linariifolia by investigating the recovery of faunas after defaunation using insecticide. One tree from each of 21 pairs of trees was sprayed at the beginning of the project (early April 1994). After predetermined periods, three test trees were resprayed along with matched control trees. A total of 95 154 arthropods were collected and sorted during the project. The number of species present on the trees recovered within 16 weeks of spraying, with common species recolonizing within a fortnight. The rarer species both of mobile (Diptera) and relatively sedentary (Araneae) taxa reappeared at similar rates. Complete recovery of numbers occurred by week 8 after spraying. Evenness (as Simpson’s D) recovered over the first 2 months; however, both the number of individuals and the evenness continued to diverge from the pattern seen on the control trees until the end of the study. While the rate of movement of individuals and species was such as to provide an apparently complete set of replacement species within several months of perturbation, the structure of the community found on the trees was still seriously disrupted after 1 year. Comparison of the suites of species originally found on the trees with those found in the respray samples and the control samples showed that the set of colonizing species was no more similar to the original fauna of the tree than it was to those on the control tree. The relatively rapid colonization of the trees by a suite of rare species – not necessarily those that were on particular trees before perturbation – indicates that rarity was due neither to inability of the species to colonize the trees nor to the suitability of the trees for these species. Recovery of rare species was to a level similar to that found on the control trees. That the divergences from the controls continued (in number of individuals and in evenness), implies a definite connection between the different faunas of a tree and their partial isolation from fauna communities on other trees. Whatever the forces that maintain suites of species on each tree, it is not the ability of the species to reach and colonize trees, nor certain attributes of a tree, that make it suitable only for a particular subset of the species available. Trees are not isolated entities but neither are they part of a fully integrated community, either chronologically or spacially, and issues of scale are also likely to be important in understanding and estimating the dynamics and factors regulating biodiversity levels.
INTRODUCTION
When the arboreal arthropod fauna of a tree is studied as part of a programme intended to monitor changes in the level of biodiversity, it is valid to ask how the biodiversity on the tree is connected to the diversity on surrounding trees and thus question the representativeness of the sample set. Are the species on an individual host tree simply a partly random subset of the available species (i.e. ‘Are tree communities physically isolated?’), or does each host tree or arthropod community have particular attributes that make it suitable only for a particular subset of the arthropod species available in the area (i.e. ‘Are tree communities biologically isolated?’). The former explanation would mean that tree communities are islands, subject to the vagaries of colonization and extinction ( Janzen 1973; Southwood & Kennedy 1983). The latter would mean that tree communities depend on the genetic attributes of the host or, alternatively, on predator–prey dynamics or competition within the arthropod community. As part of the process of understanding and monitoring the dynamics of biodiversity, we need to know whether the fact that only a proportion of the potential fauna present is found on any given tree (60% in the system studied here; Richardson et al. 1999 ) is the result of limitations in the opportunities for colonization or whether it is due to other factors. Management of the suite of rare species that constitute the vast bulk of species in any arthropod fauna requires a clear understanding of the basis of this rarity, and this becomes of particular concern when it is recognized that species occurring at low density usually also have restricted ranges ( Gaston et al. 1997 ).
This study examines the level and form of isolation of arthropod faunas present on individuals of the endemic woodland tree Melaleuca linariifolia. It focuses on the dynamic response of biodiversity estimates to perturbation by investigating the recovery of faunas after defaunation using insecticide. The recovery rates of evenness and richness indices, over periods varying between 1 week and 1 year after spraying, are calculated, and reflect the rate of recolonization, and therefore the isolation, of the fauna of a tree. The relationships between the recovering fauna and the original fauna of the tree and the fauna found on other trees sampled at the same time are considered. A wide range of species with differing capacities of movement are normally found in arboreal assemblages. As well as the overall effects of defaunation, the patterns of response in less mobile and more mobile groups of organisms (the spiders and Diptera, respectively) are compared.
Because of their short generation time, arthropods can alter their activities with season, depending on the type of food available ( Schoener 1986). Seasonality is therefore potentially a critical issue in arboreal biodiversity monitoring. However, the effect of the seasonality of arboreal arthropods on biodiversity estimates has not been studied in detail, though limited data are available on arboreal arthropod seasonality in Australia (e.g. Woinarski & Cullen 1984; Bell 1985; Basset 1991; Recher et al. 1996 ). The confounding effect of season on levels of biodiversity and on the similarity of faunas obtained in different seasons is also measured. A knowledge of the effects of season is also necessary to the interpretation of the observed pattern of recovery following defaunation.
METHODS
One member of each of 21 pairs of trees of the species Melaleuca linariifolia sm. was sprayed at the beginning of the project (late March or early April 1994; ‘T’ samples) and the falling arthropods collected. This treatment caused both the defaunation of the trees and provided the pretreatment samples of arthropods from each tree. After a predetermined period, three of these test trees were resprayed (‘R’ samples) along with matched control trees (‘C’ samples) to provide post-treatment and control sample sets. The periods before respraying were 1, 2, 4, 8, 16, 32 and 52 weeks. In order to provide a set of seasonal samples, two additional trees were sprayed in week 24 and a further two trees in week 48. Consequently, samples from two or three untreated trees were available at least every 2 months. Because the aim of the project was to sample foliage-associated arthropods, sampling was arranged so that the effects of any flowering factor (late October and early November) on arthropod abundance was avoided ( Woinarski & Cullen 1984).
This project was conducted within the 3 km2 of open woodland on the southern side of the University of Western Sydney, Hawkesbury campus at Richmond (150°45′44′′E; 33°38′30′′S). The trees in each pair were selected so as to be of similar size and shape, and were growing within 20 m of each other. Pairs were distributed across the study area and several other trees were as close to each test tree as was the control tree. Trees were sprayed using a motorized backpack spraying machine (Solo-Kleimotoren; Sindelfingen, Germany). The spray mixture contained 0.02% (v/v) of the pesticide Permethrin (Ambush; ICI Crop Co., City?, Australia) and 0.05% non-ionic X-77 surfactant mixed with tap water. Spraying was conducted from a double-based stepladder 1.5 m above the ground and 1.5 m away from the canopy. The trees used were all less than 7 m in height, which allowed the entire canopy to be sprayed. Trees were sprayed from four sides over a period of 10–15 min using 5 L of insecticide on each tree. Sampling was carried out on mornings of dry and calm days.
Collectors (diameter 0.9 m; six under the canopy of each tree; total area 3.8 m2) were distributed to collect insects falling from the leafiest parts of the tree. After spraying, most insects fell within the first 20 min but the collectors were left in position for 90 min to maximize the collection. The individuals in a collector were transferred to a jar containing 200–250 mL of 70% ethanol.
Insects were sorted to Order and the remaining arthropods to Class. The animals from each tree were then sorted to morphospecies ( Kitching et al. 1993 ; Kremen et al. 1993 ; Richardson et al. 1999 ). Each species was cross-matched with those from the other sample sets from the pair of trees and indistinguishable combinations recorded. The material studied has been transferred to The Australian Museum, Sydney. Accumulation curves for the six collections taken from a tree ( Richardson et al. 1999 ) show that new rare species are still being collected in the sixth collector, though at a greatly reduced rate. The six collectors covered 70–100% of the crown cover area under a tree.
A set of standard biodiversity estimators was used in this study ( Magurran 1988). Richness was estimated as S (the number of observed species), α (the Index of Diversity, which is also a measure of richness, Magurran 1988), and Chao1 (a non-parametric estimation of the number of species present; Colwell & Coddington 1994). Evenness was estimated as Simpson’s D. The number of individuals (n) was also counted. These three attributes of diversity are independent of one another in the present context (Azarbayjani & Richardson, unpublished observations, 1995; see also Magurran 1988). To compare the distinctiveness of the species arrays found on matching trees, complementarity (in the form of the Marczewski– Steinhaus distance, Colwell & Coddington 1994) was calculated. This measure is to be preferred over the similarity coefficients more commonly used as it satisfies the triangle of inequality (that is, the values are metric and behave as distances that can be subjected to ordination procedures not just classification procedures, Pielou 1984).
RESULTS
Season
Annual rainfall and temperature show marked seasonality in the study area, with warm wet summers and drier, cooler winters. Mean monthly maximum temperatures during the study ranged between 17.5 and 28.3°C and mean minimum temperature from 1.2 to 15.5°C. Dry conditions extended from 6 months before the experiment started through most months to December 1994 (10 mm month–1). Wetter conditions set in during December 1994 and rainfall reached a peak during March 1995 (150 mm). The study period can be divided into two periods: dry conditions from April 1994 throughout the colder months, followed by a wet warm summer and autumn.
A total of 95 154 individuals was collected and sorted during the project. The results for the T samples (all collected in autumn 1994) and the C trees (collected in the weeks as shown in Table 1) were used to examine the effect of season on the estimates of diversity. It can be seen that all estimates of richness change significantly with time (significance of slopes of all linear regressions for C data of week against richness estimates P < 0.001). Most notably the summer 1994 (week 48) and autumn 1995 (week 52) samples are far richer (80% and 50%, respectively) than the samples from the previous autumn ( Table 1). However, the estimates of α do not change with time if the data are corrected for tree effects (i.e. the T values for trees ultimately used in the 16–52 week respray periods had higher values than the trees used for shorter respray periods ( Table 1). The T/C ratios, however, do not change with time.) The cause of the variation in T values is unknown but presumably some chance sampling effect. The data for the number of individuals show a similar pattern of change to the number of species ( Table 1). There was no consistent effect of season on evenness (D;Table 1). Clearly season affects estimates of diversity, making monitoring for change more difficult.
| Estimator & | Weeks since spraying | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| treatment | 1 | 2 | 4 | 8 | 16 | 24 | 32 | 48 | 52 |
| Number of species (S) | |||||||||
| T | 87.3 ± 9.7 | 99.7 ± 12.7 | 85.7 ± 6.2 | 108.3 ± 6.4 | 87.0 ± 22.1 | 113.0 ± 8.9 | 115.0 ± 16.8 | ||
| C | 69.3 ± 8.8 | 83.0 ± 7.0 | 85.7 ± 9.2 | 118.7 ± 12.6 | 73.0 ± 1.2 | 100 ± 15 | 112.0 ± 32.7 | 195.5 ± 4.5 | 152.3 ± 9.3 |
| R | 46.7 ± 7.2 | 57.3 ± 3.9 | 75.0 ± 14.0 | 104.3 ± 8.8 | 74.0 ± 18.9 | 115.0 ± 9.6 | 154.7 ± 20.4 | ||
| R/C | 0.72 ± 0.05 | 0.71 ± 0.09 | 0.86 ± 0.09 | 0.90 ± 0.10 | 1.01 ± 0.26 | 1.18 ± 0.26 | 1.01 ± 0.11 | ||
| α | |||||||||
| T | 26.0 ± 1.9 | 22.8 ± 1.7 | 28.6 ± 1.2 | 24.4 ± 1.0 | 33.7 ± 4.4 | 33.5 ± 2.7 | 37.1 ± 4.0 | ||
| C | 19.2 ± 0.6 | 26.0 ± 3.4 | 20.8 ± 5.4 | 28.6 ± 1.3 | 25.2 ± 1.0 | 35.0 ± 5.1 | 36.1 ± 4.0 | 48.0 ± 0.2 | 36.9 ± 3.4 |
| R | 20.0 ± 1.6 | 21.8 ± 1.1 | 26.3 ± 3.3 | 24.2 ± 2.5 | 19.7 ± 2.2 | 34.1 ± 4.1 | 31.0 ± 4.0 | ||
| R/C | 1.05 ± 0.11 | 0.88 ± 0.15 | 1.41 ± 0.38 | 0.84 ± 0.06 | 0.77 ± 0.06 | 0.95 ± 0.06 | 0.87 ± 0.17 | ||
| Chao1 | |||||||||
| T | 138 ± 3.8 | 135.7 ± 8.7 | 169.7 ± 51.9 | 119.0 ± 5.7 | 170.7 ± 36.5 | 163.3 ± 34.6 | 197.3 ± 42.4 | ||
| C | 91.7 ± 7.3 | 137.0 ± 16.3 | 128.7 ± 24.7 | 146.0 ± 10.2 | 127.0 ± 14.3 | 200 ± 15.5 | 156.3 ± 22.9 | 338.0 ± 56 | 242.3 ± 22.2 |
| R | 79.3 ± 9.3 | 99.0 ± 1.2 | 126.7 ± 2.6 | 151.3 ± 18.2 | 119.7 ± 24.5 | 168.0 ± 5.5 | 247.3 ± 40.2 | ||
| R/C | 0.86 ± 0.04 | 0.75 ± 0.10 | 1.04 ± 0.16 | 1.03 ± 0.07 | 0.96 ± 0.24 | 1.11 ± 0.12 | 10.5 ± 0.20 | ||
| Number of individuals (n) | |||||||||
| T | 941 ± 265 | 2241 ± 849 | 566 ± 118 | 1902 ± 249 | 529 ± 106 | 946 ± 86 | 819 ± 237 | ||
| C | 925 ± 362 | 718 ± 180 | 1953 ± 766 | 1929 ± 519 | 427 ± 25 | 827 ± 541 | 1062 ± 681 | 2787 ± 304 | 2496 ± 606 |
| R | 258 ± 78 | 285 ± 35 | 1064 ± 510 | 2138 ± 549 | 1538 ± 998 | 1014 ± 63 | 5021 ± 1632 | ||
| R/C | 0.37 ± 0.13 | 0.48 ± 0.18 | 0.57 ± 0.15 | 1.50 ± 0.73 | 3.82 ± 2.6 | 2.35 ± 1.37 | 2.33 ± 0.97 | ||
| Simpson’s D | |||||||||
| T | 0.93 ± 0.01 | 0.90 ± 0.03 | 0.94 ± 0.02 | 0.91 ± 0.02 | 0.90 ± 0.03 | 0.95 ± 0.01 | 0.91 ± 0.01 | ||
| C | 0.84 ± 0.05 | 0.93 ± 0.01 | 0.78 ± 0.06 | 0.92 ± 0.01 | 0.93 ± 0.01 | 0.84 ± 0.06 | 0.89 ± 0.02 | 0.89 ± 0.03 | 0.86 ± 0.03 |
| R | 0.94 ± 0.01 | 0.94 ± 0.01 | 0.84 ± 0.10 | 0.89 ± 0.02 | 0.80 ± 0.09 | 0.88 ± 0.04 | 0.66 ± 0.10 | ||
| R/C | 1.13 ± 0.08 | 1.01 ± 0.02 | 1.07 ± 0.06 | 0.96 ± 0.02 | 0.90 ± 0.12 | 0.99 ± 0.05 | 0.78 ± 0.14 | ||
- T data are from trees sprayed in week 0, but the data are summarized so as to match the R data for the same set of trees. R/C is the average of dividing R by C for each pair of trees.
Defaunation
Given the effects of season on the amount of biodiversity present, the effects of defaunation are best shown as changes in the values of the various estimators as proportions of the equivalent values for the matching controls. Means and standard errors were calculated from the three proportions that were obtained for each sampling period (R/C; Table 1). It can be seen that the numbers of species obtained on the resprayed trees were substantially below the values obtained for the control trees in the first 2 weeks after spraying (i.e., R/C ≈ 0.7; Table 1, compared to T/C ≈ 1 expected for untreated trees). By week 4, however, the number of species had started to recover and recovery was complete by week 16. A clear effect of defaunation was not demonstrated for α and the non-parametric estimator Chao1 was only affected for the first 2 weeks.
The number of individuals was reduced to 40% on trees 1 week after spraying but recovery of numbers was complete by week 8. The numbers continued to rise, however, leading to an overshoot to more than twice the number of individuals found on the matched control trees at the end of the study.
An examination of the results for evenness show that Simpson’s D is also affected by the treatment. In the first week, D was higher on the sprayed trees, that is the numbers of species were more evenly distributed on the resprayed trees than on the matching control trees. This result is due to the fact that most species present were in similarly low numbers. Recovery was apparently complete by week 8 after spraying. Evenness, however, continued to change, with the faunas on the treated trees slowly becoming more unbalanced with time. The observed effects on n and D in the longer recovery periods were due to the presence of very large numbers of a few species 8 months to 1 year after spraying. It is noteworthy that although the total number of species and individuals originally present were recovered within 2 months, the effects of the perturbation were still evident in community structure 1 year after the event.
The analysis of complementarity is summarized in Table 2. In weeks 2 and 4 not all pairs of trees were cross-matched, so less data are available. Comparison of the trees sampled in week 0 (T) with the control trees (C) sprayed at a later date showed a steadily increasing complementarity (i.e. increasing distinctiveness) with time until week 16 (early spring). Even when the matched trees were sprayed in the same season but 1 year later (week 52), there was no decrease in complementarity. It can be concluded that the suite of species constituting the fauna of the site was steadily changing with time during the study and that this shift was not due to a seasonal cycle in the suite of species.
| Complementarity | Weeks since spraying | ||||||
|---|---|---|---|---|---|---|---|
| between: | 1 | 2 | 4 | 8 | 16 | 32 | 52 |
| C & T | 61 ± 1 | 63 ± 0 | 65 | 70 ± 1 | 76 ± 1 | 74 ± 3 | 77 ± 3 |
| R & T | 69 ± 2 | 68 ± 1 | 66 | 67 ± 1 | 76 ± 3 | 75 ± 3 | 77 ± 0 |
| C & R | 63 ± 2 | 71 ± 4 | 58 | 59 ± 1 | 65 ± 1 | 69 ± 4 | 71 ± 2 |
| n | 3 | 2 | 1 | 3 | 3 | 3 | 3 |
- Values shown are means ± SE. Note that complementarity was not calculated for all three pairs of trees on all occasions.
The comparison of the week 0 trees (T) with the same trees sprayed at a later date (R) showed a higher complementarity than that of the equivalent T and C pairs for weeks 1 and 2. The number of species on the sprayed trees was still low during this period (see above) and, consequently, the number of species in common was low as a proportion of the total number of species on the two trees. By week 4 the recovery of the species richness was 85% complete and similar T/R complementarity values were obtained to those found for the matching T/C pairs for the rest of the study. Thus the recovery of the faunas on the trees was not to the same species as were found before spraying but to a set of species no more similar to the original fauna than the set found on the matching control trees.
Examination of the complementarities between the R and matching C trees showed a different pattern. R and C faunas were more similar until week 16 than the matching R and T pairs, but they were as similar as those of the C and T comparisons. The reverse would have been expected; in the first few weeks, the low number of species on the R trees would be expected to lead to results similar to the R and T comparison. The only anomaly in this pattern is week 2, where one tree had an unusually high complementarity value, and this result should be discounted. Why the recolonization process should include a high proportion of the species found on the control trees in the early stages of recolonization is unclear and the validity of the result needs to be examined with further data sets. It is possible that because pairs of test and control trees were physically close to one another, the control trees provided the major source of colonizers. Other trees, however, were just as close. In the later stages of the study, C and R complementarity values were lower than those observed for the C and T and the R and T comparisons. This was due to the previously discussed changes over time in the suite of species present, that is the sample sets collected at the same time but from different trees were more similar than sets collected at different times from the same tree.
Comparative effects on sedentary and mobile groups in the fauna
Recovery of species and the number of individuals of the Diptera (more mobile) and Araneae (less mobile) are summarized in Table 3. It can be seen that the number of species has recovered in each group by week 4. The number of individuals of Araneae had recovered by week 8 while the number of individuals of Diptera had recovered by week 1. There was no difference (at least under the conditions used in this study, e.g. time of year) in colonization rates by species in the two groups. The low number of species involved makes analysis of evenness uninformative. Examination of the primary data (not presented) showed that the difference between the patterns of recovery in the two taxa was due to the behaviour of common dipteran species. The common dipteran species all returned within 1 week in numbers equivalent to those seen in the control trees. However, the rare species were slower to return. For the Araneae, the rare species were also slow to return while the common species returned more rapidly. However, even when returning they appeared at much lower numbers than in the control trees. In summary, rare species returned slowly irrespective of the mobility of the group, while the recovery of common species was faster than for rare species although slower in the less mobile taxon. Nevertheless, recovery of numbers was complete within a few weeks in both groups.
| Estimator & | Weeks since spraying | ||||||
|---|---|---|---|---|---|---|---|
| treatment | 1 | 2 | 4 | 8 | 16 | 32 | 52 |
| Species | |||||||
| Diptera | 0.71 ± 0.30 | 0.68 ± 0.16 | 1.12 ± 0.19 | 0.70 ± 0.26 | 2.13 ± 0.65 | 1.53 ± 0.55 | 1.04 ± 0.22 |
| Araneae | 0.34 ± 0.08 | 0.59 ± 0.04 | 1.02 ± 0.30 | 0.85 ± 0.21 | 0.82 ± 0.13 | 1.20 ± 0.20 | 0.94 ± 0.25 |
| Individuals | |||||||
| Diptera | 1.66 ± 0.84 | 3.65 ± 2.82 | 0.97 ± 0.27 | 1.51 ± 1.15 | 5.56 ± 3.50 | 3.21 ± 2.66 | 3.42 ± 2.78 |
| Araneae | 0.48 ± 0.12 | 0.35 ± 0.02 | 0.71 ± 0.17 | 1.02 ± 0.45 | 1.44 ± 0.49 | 1.26 ± 0.42 | 1.24 ± 0.72 |
- Values shown are the mean ± SE for the ratios of resprayed to matched control trees. Typically there were 10–20 species of each taxon, 10–20 spider individuals and over 100 dipteran individuals in each control group.
DISCUSSION
Under a regime of continuous change due to natural and anthropogenic factors, ecosystem structures undergo continuous adjustment and have a dynamic form ( Loreau et al. 1995 ). In this context, both the level and distribution of biodiversity should be considered as dynamic, not static, attributes. As a result, it is unhelpful to think of biodiversity in typological terms, though biodiversity estimates continue to be reported and used as if they had some absolute meaning. It is clear from the present study, for example, that estimates of diversity are dependent on environmental conditions and the history of the community before sampling. The observed changes in the number of species and individuals between the autumn samples during the 2 years can probably be ascribed to the long dry period that affected the study area before the study began and continued until the end of spring 1994 (week 32). Clearly, weather affects estimates of diversity, making monitoring for human-induced change that more difficult.
In the present study, the number of species present on the trees recovered within 16 weeks of spraying, with the common species recolonizing within 1 or 2 weeks. This observation in part may have been due to the presence of a few chance survivors of the spraying. Surprisingly, the rarer species of both mobile (Diptera) and relatively sedentary species (Araneae) reappeared at similar rates. The recovery of species richness took a similar period to that observed in two previous defaunation studies (mangrove islands, Simberloff 1978; Spartina islands, Rey 1981). The recovery of the total number of dipteran individuals was also speedy. This was due to the rapid recovery of the numbers of individuals of common species that make up most of the dipteran population. Evenness also recovers over the first 2 months, with the distribution on sprayed trees being more even than that on the control trees in the first weeks. Interestingly, both the number of individuals and the imbalance between the numbers of individuals in each species continued to increase on the treated trees, but not the control trees, throughout the study. While the rate of movement of individuals and species is such as to provide an apparently complete set of replacement species within several months of perturbation, the structure of the community found on the treated trees remained seriously disrupted when compared with the matching control trees, as reflected in the changes in D and n present after 1 year. The perturbation, however, is more subtle than a simple change in the number of species present. How long it would take the faunal community living on a tree to recover from an extreme perturbation is unknown. Monitoring evenness may well be a suitable way of detecting environmental insults to ecological communities.
The relatively rapid colonization of the trees by a suite of rare species, but not necessarily those that were on the particular trees before the perturbation, means that rarity is due neither to the inability of the species to colonize the trees nor to the suitability of the trees for these species. Rare species recovered to similar numbers to those found on the control trees and included only a relatively small proportion of the rare species available in the study area. There also was little difference in the recovery of the biodiversity represented by highly mobile (Diptera) and less mobile (Araneae) groups.
It is clear that care needs to be taken in the selection of estimators to be used when monitoring for changes in biodiversity. Numbers of species and individuals are grossly affected by environmental influences, as can be seen in the between-year autumn data. However, α and D are little affected and provide more robust estimates of richness and evenness. As a measure of richness, α is unaffected by the perturbation and may be a better, more robust measure of richness than the number of species. On the other hand, the measure is insensitive to quite massive perturbations in the fauna of a tree and the value of the estimator in detecting changes in a fauna as part of a monitoring programme would need to be examined further; the model may simply be insensitive. Modelling shows that Chao1 gives excellent estimates of the number of species over a wide range of species sample sizes. At low sample sizes, however, it underestimates the number of species present ( Colwell & Coddington 1994). What constitutes too small a sample size remains to be determined, though in the system modelled by Colwell & Coddington (1994), 25 species were sufficient for an accurate estimate. It is clear that, with the sample sizes used in the present study (100 species), the number of species known to be present in the area was grossly underestimated (for example, at least 300 species were collected during the T sampling process, though only 138 were predicted by Chao1, Table 1).
Analysis using complementarity shows that there was increasing divergence in the faunas on the trees with increasing time between the first and second sampling occasions until week 16, presumably reflecting a turnover in the species composition on the study site ( Table 2). Whether this was simply due to seasonal changes during the study period is unknown. However, the failure to show any return to similar suites of species in the autumn of the second year to those seen in the autumn of the first year demonstrated that the effect was not simply the result of a seasonal succession cycle of species in the fauna. The issue of the effect of sample size on estimates of complementarity raised by Colwell & Coddington (1994) was considered in the previous part in this series ( Richardson et al. 1999 ), where it was shown to be unaffected by variation in sample size of the magnitude observed in the present study.
There is a clear demonstration in this study that individual trees of the same species support distinct sets of arthropod species. Despite this, the colonizing capacity of the available suite of species is such as to ensure that many more species than those found could colonize a tree over a period of a few months (and so the effect is not due to physical isolation). At the same time it is difficult to believe that the presence of a single individual of a small rare species somewhere on a relatively enormous tree would provide sufficient competition to stop colonization by a single individual of another small rare species. However, the continuing abnormalities in the abundance of individuals and the distribution of relative frequencies of individuals 12 months after spraying implies definite connections between the different segments of the fauna of a tree. Given that this only occurred on the treated trees, there is also partial isolation of the arthropod community on a tree from those only a few metres away on other trees. The assemblage recolonizing a tree, however, was no more similar to the original assemblage than it was to that on the control tree. Consequently, the differences in the suites of species on these trees was not due to biological differences between the host trees.
Whatever the force then that maintains suites of species on each tree, it is not the ability of the species to reach and colonize trees; that is, trees are not physically isolated. Biological isolation is present but it is not the genetic attributes of each tree that make it suitable for only a particular subset of the available species. What underlies the observed biological isolation (e.g. community assembly rules, predator or parasite/prey relationships) is unknown. It is clear, however, that, for whatever reason, there are simply not enough individuals available of rare species to inhabit all suitable trees, and regulation of their numbers must occur at some other scale. Trees are not isolated islands but neither are their faunas part of a widespread, fully integrated community, either chronologically or spacially, and as a consequence issues of scale are likely to be important in understanding the dynamics and the factors regulating biodiversity levels and their estimation and monitoring. Clearly there is an urgent need to examine further the dynamics of biodiversity and its measurement in an experimental fashion.
Acknowledgements
We would like to acknowledge the Ministry of Science and Higher Education of the Republic of Iran for support during this project (F.F.A).
Thanks to the staff of the School of Biological Sciences, University of East Anglia, UK, for their hospitality to B.J.R. during the preparation of this manuscript.
