«Claudia Gray, Cambridge University, UK Evelien Jongepier, University of Groningen, The Netherlands Abstract The arthropod fauna of Madagascar is ...»
Effect of human disturbance on arthropod diversity at
Kirindy Forest, Western Madagascar
Claudia Gray, Cambridge University, UK
Evelien Jongepier, University of Groningen, The Netherlands
The arthropod fauna of Madagascar is remarkably diverse and includes many endemic genera, yet it
remains poorly documented and under increasing threat from human disturbance. Not only does the
current state of Madagascan arthropod fauna deserve better characterisation, it also provides as yet unrealised potential as a powerful tool in conservation monitoring. We investigate the relationship between diversity of arthropod communities and level of human disturbance in Kirindy Forest, Menabe, Western Madagascar. We demonstrate a decrease in arthropod diversity at the order level with an increase in human disturbance. However, our assessment of individual taxonomic groups shows that the specific effect of human disturbance varies between taxa. Our results have implications for the management of Kirindy Forest reserve and possible restriction of human disturbance within the site. The study also provides strong evidence for the rapid and cost-effective nature of arthropod community assessments, reinforcing the idea that they have an important role to play in contemporary conservation efforts.
The Malagasy invertebrate fauna shows a high level of both species richness and endemicity, making it a conservation priority, and yet much of the fauna remains unknown and undescribed. Of the 1000 described species of ants, for example, 96% are endemic, and this total is thought to represent only a third of the total ant diversity of the country (Fisher, 2003). The Malagasy invertebrate fauna also includes many groups that have retained ‘primitive’ characteristics (such as the Archaeidae group of spiders), which may potentially provide valuable insight into the phylogeny and evolutionary history of insect groups found across the globe (Paulian and Viette, 2003). The unique radiations of many arthropod groups within Madagascar are also likely to provide valuable information on the processes of speciation and convergent evolution. In addition to these issues, arthropods and insects in particular offer great potential as biodiversity indicators. The high diversity and local endemism of ant species, for example, makes this group a good indicator species richness and turnover within a region (Fisher, 2003). They are easy to collect, and in combination with ecological information on the species recorded, can provide useful information about habitat condition for conservation planning. Ants are an important food source for many organisms, as well as exerting an important predation pressure on many other arthropod species.
They also act as ecological engineers, altering the local concentration and cycling of nutrients, and from a wide range of symbiotic interaction with both plants and animals (e.g. Aphaenogaster swammadami is known to disperse the seeds of Commiphora guillaumini in the Menabe region (Bohning-Gaese et al., 1996)). Unfortunately, despite the ecological importance of the invertebrates, it is often excluded from biodiversity inventories and monitoring schemes due to misconceptions about the lack of rapid assessment methods and availability of taxonomic knowledge (Fisher, 2003).
The dry, deciduous forest reserve at Kirindy, Western Madagascar, offers a unique opportunity to use arthropod groups to assess the impact of human land use on biodiversity. In 1978, the Centre de Formation Profesionelle Forestiere (CFPF) established a forestry commission of 1200 ha. Lowimpact logging was carried out on the site until 1990, when the timber production was deemed not to be economically viable and terminated (Sorg et al., 2003). Since the establishment of the concession, a range of research projects on silviculture, exploitation and reforestation techniques were completed. Faunal studies began in the 1980s, and intensified with the establishment of the German Primate Centre (DPZ) in the 1990s. This research group established a grid network of paths at various sites within the reserve in order to facilitate their research. In combination with the old logging roads, these grid systems mean that there is now a range of sites in Kirindy that have received varying levels of human disturbance, from areas of forest left outside of the grids through the small paths to the large logging roads. In addition, variation in logging intensity between different regions within Kirindy forest enables investigation of the effect of this factor on arthropod diversity.
Our study investigated the effect of human disturbance and habitat degradation due to logging, and the continued presence of the roads and path system on the diversity of invertebrate communities.
We tested the hypothesis that species abundance and diversity are negatively impacted by habitat degradation due to human disturbance. In order to do this, we collected data on the abundance of invertebrates, classified to order level, and further into subgroups where appropriate. We also identified ants to morphospecies level, to obtain more detailed data on this particularly abundant, diverse and ecologically important group. The results of the study have implications both for the management of the reserve at Kirindy, and the wider use of invertebrate taxa as ecological monitoring tools.
METHODSIn order to assess arthropod diversity and abundance we used pitfall traps. Each trap consisted of four cups containing water, covered by a weighted lid to prevent damage or disruption. We set traps under 4 different treatment conditions, as follows: big paths (B) included the old logging roads of more than 2m width, with the majority of the path surface lacking leaf litter. Small paths (S) included all paths within the DPZ grid system, approximately 1m in width, and covered with leaf litter. The forest grid (FG) treatment included all sites at the centre of a grid cell (the area of forest between paths in the grid). The forest non-grid (FNG) treatment included all sites in forest outside of the DPZ grid system. We identified two regions that differed in the intensity of logging during the active period of the CFPF commission by consulting data on the percentage of extractable wood removed (Andriambelo, 2005).
We stratified the study site into 4 sampling areas of 10 x 7 grid cells and matched each of these areas to regions of big path and non-grid forest. This allowed us to simultaneously set traps under all 4 treatments, thereby controlling for any variation in trap efficiency due to weather conditions or time of day. Stratifying the study site into 4 sampling areas also controlled for the local variation in ecological conditions (e.g. proximity to the river). Two sampling areas were placed in the more intensively logged forest, and 2 in the less intensively logged forest, enabling us to compare the affect of this variable on arthropod diversity. We set 5 traps under each treatment condition, and left them for 24 hours before collection. The sites of the forest grid (FG) traps, and the small path (S) traps were randomly selected. We excluded the 2 grid cells nearest the big paths to avoid any edge effects. All traps were placed at least 30 m apart (see Fig. 1 below), to ensure that each sampled from different communities, thereby limiting the risk of pseudoreplication. (This distance was justified by preliminary studies showing that the maximum foraging range of an ant was approximately 15 m from its nest entrance.)
Small paths Fig. 1 Schematic representation of grid system in Kirindy forest and different treatment conditions used in our study.
There was limited availability of big path and non-grid forest directly adjacent to each 10 x 7 sampling areas, so we could not randomly select trap sites for these treatment conditions. Instead, we placed traps at 30 m intervals along the big paths and also at 30 m intervals along a transect 30 m into the non-grid forest. We set and collected traps once in each of the 4 sampling areas. For all four areas, we successfully recovered between 3 and 5 replicates for each treatment, as some of the traps were damaged and became non-functional.
Having collected the trap-contents, we counted the total number of individuals within each sample.
We classified the ants, as these were most abundant, according to morphotype. Using a microscope, we identified 26 different morphospecies in total. Where more than approx 150 individuals were present in the trap contents, we sub-sampled the population in order to obtain an estimate of total abundance. All other organisms we identified to order level, or lower taxonomic levels were possible (e.g. Hemiptera were sorted into Homoptera and Heteroptera, and Dictyoptera were sorted into Blattodea, Mantodea and Isoptera). We sieved and washed all samples to ensure that visibility of the trap contents was comparable, and invested more time in the ‘muddiest’ samples in order to ensure that search effort was consistent.
For the analysis of diversity at the order level, we used including the Shannon index (Ingram et al., 2005), as follows: H’ = -ΣpiLog2pi (where p is the proportion of individuals in the ith order). For calculations of arthropod abundance, we did not use the total abundance of ant individuals, as this was greatly affected by the proximity of the trap to an ant nest. Since ants are colonial organisms, we counted each colony as one individual, assuming that all members of each morphospecies belonged to the same colony.
RESULTSArthropod community diversity The Shannon index showed a significant decrease with disturbance level (Fig. 2A.; GLM, F=3.16, df=3,59, p=0.031), where the Shannon index of forest inside grids and non-grid forest tended to be higher than that of the big paths (TukeyHSD, FG-B: p=0.0338, FNG-B: p=0.0734). Neither logging history nor its interaction with disturbance had an effect on the arthropod diversity as given by the Shannon index (GLM, logging history: F=0.00, df=3,58, p=0.944; interaction: F=0.16, df=3,55, p=0.923). In contrast, order richness did not differ significantly between disturbance levels (GLM, F=1.80, df=3,59, p=0.157), logging history (GLM, F=0.14, df=1,58, p=0.708) or their interaction (GLM, F=1.44, df=3,55, p=0.241), although the number of orders sampled tended to decrease with increasing level of disturbance (Fig. 2B). Similarly, the total number of arthropods did not differ between the different levels of disturbance (Fig. 2C; GLM, F=1.79, df=3,55, p=0.159). In addition, neither logging history, nor its interaction with disturbance affected arthropod abundance (GLM, logging history: F=0.53, df=1,55, p=0.471; interaction: F=2.68, df=3,55, p=0.056).
B C Fig. 2 Pitfall trap contents and diversity index for the four levels of disturbance. A) Shannon index of biodiversity (mean ± SE; GLM, F=3.16, df=3,59, p=0.031). B) Number of orders (mean ± SE; GLM, F=1.80, df=3,59, p=0.157). C) Number of individuals. Median, inter quartile range, 95% data range and outliers represented by bar, box, whiskers and stars, respectively (GLM, F=1.79, df=3,59, p=0.159).
Fig. 3 Number of spiders in pitfall traps for the four levels of disturbance and the sites with different logging history (mean; GLM, disturbance: F=4.36, df=3,55, p=0.008, logging history: F=2.90, df=1,55, p=0.094 and interaction: F=3.17, df=3,55, p=0.031).
Abundance of specific taxonomic groups Cockroach abundance was not affected by either disturbance level (GLM, F=1.82, df=3,58, p=0.153), logging history (GLM, F=3.70, df=1,62, p=0.059) or their interaction (GLM, F=0.43, df=3,55, p=0.729).