Nick Brown, Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK (fax + 44 1869 275074; e-mail firstname.lastname@example.org).
1Despite their functional importance, little is known about how and where fungi can be conserved. It is important that we understand the consequences of habitat degradation and fragmentation for fungal assemblages if we are to devise successful conservation strategies.
2We investigated the effects of fragmentation and disturbance on the diversity and landscape distribution of fungi in tropical rain forests in the Kodagu district of the Western Ghats of India. We recorded macrofungi on three occasions over a wet season, in 0·125-ha plots in 10 forest reserve sites, 25 sacred groves and 23 coffee plantations.
3Despite a long history of isolation from continuous forest, sacred groves had the highest sporocarp abundance and the greatest morphotype richness per sample area, while coffee plantations had the lowest. However, coffee plantation samples were more diverse for a given number of sporocarps than a sample of a similar size from other forest types.
4Ordination by non-metric multidimensional scaling suggested that sacred groves had a macrofungal assemblage that was distinct from other forest types. This compositional difference was primarily because of the presence of a group of dead wood and litter decomposing fungi. Coffee plantations and forest reserve sites had very variable but overlapping compositions.
5Neither sacred grove size nor distances between a grove and continuous forest accounted for a significant proportion of the total variation in their macrofungal richness.
6There was no significant correlation between dissimilarity in macrofungal assemblage composition and geographical distance between sample sites. However, we found strong congruence between patterns of dissimilarity in macrofungi and trees between sites.
7Synthesis and applications. These results imply that macrofungal distribution patterns at a landscape scale are determined by habitat requirements rather than dispersal or local population dynamics. This means that habitat degradation is a more serious threat to fungal diversity than fragmentation. Sacred groves, although small, are important for fungus conservation because they provide unique types of habitat.
Forest loss in the humid tropics proceeds by a combination of fragmentation and degradation. Residual patches of natural forest retain relic populations of plants and animals but their species richness has often been found to decline over time because of local extinctions (Fahrig & Merriam 1994). Nevertheless, many forest organisms are dependent on ecological continuity of old-growth conditions in forest patches for their continued survival. This may be because of their limited ability to colonize new habitats and/or requirements for old-growth microhabitats (Norden & Appelqvist 2001). Although much has been learned recently about the dynamics of plant and animal species populations in fragmented landscapes, virtually nothing is known about the diversity and landscape distribution of fungi. Fungi are key functional components of forest ecosystems and the paucity of knowledge on their responses to forest fragmentation and degradation needs to be addressed. In the humid tropics, where the rates of forest loss are highest, so little is known about the ecology of most species of fungi that it is difficult even to make predictions about the likely impacts of forest fragmentation. The effects of disturbance and forest management on fungal diversity have been reported in a number of studies from northern temperate regions (Vogt et al. 1992; Waters et al. 1997; Senn-Irlet & Bieri 1999) but studies from tropical forests are extremely rare (Lodge & Cantrell 1995; Persiani et al. 1998; Allen et al. 2003). In this study we aimed to examine the consequences of tropical forest degradation and fragmentation for fungal diversity in one of the two centres of mega-diversity in India.
One of the principal reasons for the lack of information is the formidable difficulty that fungi present to ecological study (Cannon 1997). Most species are cryptic, rarely or never forming sporocarps, and most species of tropical fungi are undescribed (Hawksworth 2001). Estimates of fungal diversity are therefore based on the identification of morphotypes or other surrogate measures (Cannon 1997). Beattie & Oliver (1994) have argued that this type of approach is the only way that neglected groups such as fungi can be incorporated into biological inventories in the near future. Morphotypes have been used to great effect in biodiversity studies of unknown and highly diverse groups of fungi (Arnold et al. 2000; Balmford, Lyon & Lang 2000). However, inventories based on morphotypes are vulnerable to unpredictable sorting errors and results are difficult to check or compare with other studies (Krell 2004). One means of overcoming some of these problems is to ensure that a thorough and accessible archive of morphotypes is prepared. Detailed morphological descriptions, high-resolution in situ photographs and voucher specimens of both dried sporocarps and spores all provide the means for more precise taxonomy in the future.
In this study we examined the morphotype diversity and landscape distribution of macrofungi (fungal species that form macroscopic sporocarps; Arnolds 1992) in the Western Ghats of India. Macrofungi include most Basidiomycetes (excluding rusts, smuts and yeasts), some Ascomycetes (e.g. Peziza) and Myxomycetes (e.g. Fuligo) (Watling 1995). Although it is a somewhat arbitrarily defined polyphyletic group, it is of potential use for rapid assessment of fungal diversity because of the conspicuous nature of macrofungi (Balmford, Lyon & Lang 2000).
We surveyed macrofungal morphotype diversity in formally protected forest reserves, coffee plantations and sacred groves in the Kodagu district of Karnataka state in the Western Ghats of India. Our aim was to compare the diversity and types of macrofungi found in continuous tracts of natural forest with those found in forest fragments and intensively managed plantations. The formally protected forest reserves in Kodagu are large tracts of natural forest in three wildlife sanctuaries and one national park, which stretch continuously along the western and south-western boundaries of the district. Kodagu has a substantial area of coffee plantation. Coffee plantations are created by clearing the understorey of natural forest whilst retaining a high canopy of native forest trees to provide shade for the coffee crop. Sacred groves in the Western Ghats of India are patches of natural forest that have been continuously protected by the religious beliefs of the local people for more than 1000 years (Sinha 1995; Ramakrishnan 1996; Chandran & Hughes 1997; Chandrashekara & Sankar 1998; Colding & Folke 2001). The Kodagu region is well known for a high density of sacred groves, with an average of one for every 300 ha of land (Kushalappa & Bhagwat 2001). These groves range in size from a fraction of a hectare to a few tens of hectares. Although small, they are important residual patches of semi-natural forest in a landscape that is otherwise intensively cultivated (Nair et al. 1997). While they have a long history of management by local people as an important source of non-timber forest products (NTFP) (Daniels, Chandran & Gadgil 1993), it is believed that most have never been subject to intensive timber harvesting nor cleared for other forms of land use (Chandran 1997).
Materials and methods
The Kodagu (formerly Coorg) district of Karnataka state is found between 11°56′ and 12°52′ N latitude and 75°22′ and 76°11′ E longitude (Pascal & Meher-Homji 1986). The mean temperature of the coldest month in the study area is between 16 and 23 °C; rainfall is between 2000 and 5000 mm year−1, with a 4-month long dry season (Pascal & Meher-Homji 1986).
We sampled macrofungi in 25 sacred groves, 23 coffee plantations and 10 forest reserve sites in a 600-km2 area in south-western Kodagu over a 2-year period. All sites were between 800 and 1000 m a.s.l. We sampled macrofungal sporocarps in a single 250-m long and 5-m wide randomly located transect at each site (following Senn-Irlet & Bieri 1999). Sporocarps were collected from this transect on three different occasions during a single monsoon season at each locality: the beginning (May–June), peak (July) and end (August–September) of the monsoon. Few sporocarps are found at other times of year. A team of at least four people, two on either side of the central line, walked every transect on each occasion, in order to maximize the chance of finding the majority of sporocarps. Some macrofungi produce clusters of sporocarps while others produce only a single one (Vogt et al. 1992). We recorded a cluster as one observation irrespective of the number of sporocarps in that cluster. We also collected sporocarps when new morphotypes were discovered elsewhere in our study forests whilst conducting other survey work. These morphotypes have been described in our data base but have not been included in our comparisons of fungal diversity. We used sporocarp shape, size and colour and morphological features of the spores to establish their identities. We collected at least two sporocarps of each type, and photographed them in situ where possible. We described and recorded the morphological characteristics of each sporocarp before they were cut for spore prints, which were taken on thin, transparent polythene sheets and preserved in polythene bags. Fungal taxonomy is poorly resolved for many parts of India and there are no published field guides. We used the key provided in Jordan's (1995) field guide for identification of European macrofungi to identify most sporocarps to the family level and many to the genus level. A local expert fungal taxonomist helped with the identification of voucher specimens and spore prints. We have created a database that includes a detailed description of each morphotype, the sites at which it was found and photographs, using BRAHMS software. The database has been published on-line at http://herbaria.plants.ox.ac.uk/bol/?kodagu, where it can be viewed and searched. We oven-dried the sporocarps and preserved them as reference specimens in dry paper bags. These specimens have been lodged with the University of Agricultural Sciences College of Forestry, Ponnampet, South Kodagu.
We also carried out an inventory of trees with stems ≥ 10 cm diameter at breast height in the same 58 sites that were used to sample macrofungi. Trees were also sampled on 5-m wide transects, starting at the same point as our macrofungal sampling and continuing until we had enumerated 100 individuals.
The outlines of all 25 sacred groves, three wildlife sanctuaries and one national park were manually digitized from 42 village land survey maps (scale 1 : 7920) covering our study area. These maps were available locally from Land Revenue departments and depicted the boundaries of each family's landholding as simple numbered polygons. As these maps were originally surveyed in 1901, we verified and updated the position of these boundaries using global positioning system (GPS) surveys. We used MapInfo GIS software (v7·0; MapInfo Corporation 2001) to measure the area of each sacred grove and the distance from each sample site to the nearest large tract of continuous natural forest in a forest reserve (isolation).
Our study compared macrofungal morphotype richness based on equal sampling areas (comparable to species density; sensuHurlbert 1971). However, substantial differences in the number of sporocarps found make it difficult to interpret differences in richness across the three site types. Rarefaction (Hurlbert 1971) allows calculation of the number of species (or in this case morphotypes) that would be expected if all samples were of a standard size. However, this method leads to a great loss of information because the smallest sample (in our case 18) is taken as the standard sample size. We therefore fitted a negative exponential curve to the morphotype accumulation data for all of our plots. Tjørve (2003) recommends the negative exponential curve as the simplest theoretically appropriate model for fitting to species accumulation data compiled from censuses within defined boundaries. The curve was fitted using Matlab's (version 7·0; MathWork Inc. 2004) nlinfit–non-linear regression procedure. Nlinfit estimates the coefficients of a non-linear regression function using least-squares estimation. The residual variation for each plot from its fitted value was calculated and the mean of these residuals for each site type compared with the population mean residual (assuming a sample was drawn without replacement from a finite population).
Compositional differences between samples were investigated using the ordination technique non-metric multidimensional scaling (NMDS). All analyses of composition were conducted using square-root transformed abundance data to reduce the influence of the most common morphotypes. Compositional differences between forest types were tested for statistical significance using a non-parametric analysis of similarities (anosim) with the CAP v3 software package (Seaby, Henderson & Prendergast 2004). The anosim statistic R is a measure of the difference between inter- and intragroup rank dissimilarities. A score of zero indicates that groups have identical composition, while a score of one indicates that all samples within groups are more similar to one another than they are to any samples from different groups (Clarke & Warwick 2001). A similarity percentages procedure (simper; CAP v3) was performed in order to identify which morphotypes were primarily responsible for the compositional difference between sacred groves and the other two forest types.
A Mantel test was used to determine if geographical distance between sites could explain a significant proportion of the variation in assemblage composition. Abundance data were square-root transformed to reduce the influence of dominant species and then the Bray–Curtis distance between all pairs of samples was calculated using CAP v3 (Seaby, Henderson & Prendergast 2004). The matrix of compositional distances was compared with one of Euclidean distance between sites using the simple Mantel test routine in zt software (Bonnet & Van de Peer 2002). The significance of the correlation was assessed by comparing it with a distribution found by randomly re-allocating the order of the elements in one of the matrices 10 000 times. A simple Mantel test was used to look for a correlation between differences in tree and macrofungal composition between sites. Tree abundance data was square-root transformed and the Bray–Curtis distance between all pairs of samples was calculated.
A general linear model (GLM) was used to assess whether (loge) forest patch area and (loge) isolation were significant predictors of macrofungal richness in sacred groves. The analysis was conducted using the GLM routine in Minitab v13 (Minitab Inc. 2000), with both explanatory variables specified as covariates.
Patch size and isolation were used to model the presence/absence of each macrofungal morphotype in turn, using binary logistic regression with a Logit link function (Minitab v13). Only those 28 morphotypes that were present in 10 or more sacred groves were analysed.
We recorded a total of 163 morphotypes. Full descriptions of each morphotype are provided in the on-line database (http://herbaria.plants.ox.ac.uk/bol/?kodagu). There were significant differences in the (loge) abundance of sporocarps in our three site types [F = (2,54) 69·2, P < 0·001, one-way anova, Tukey's pairwise comparisons]. Sacred groves had the highest sporocarp abundance and coffee plantations the lowest (Fig. 1a). There were also significant differences between site types in (loge) morphotype richness (F = 2,54 48·0, P < 0·001, one-way anova, Tukey's pairwise comparisons). Sacred groves had the greatest morphotype richness per sample area (Fig. 1b). We did not use other measures of diversity because they depended on knowledge of the abundance of individuals. Although we were able to quantify the abundance of sporocarps we did not know how these data related to the abundance of fungal individuals.
Figure 2 shows a negative exponential curve fitted to the morphotype accumulation data for all our plots. The mean residual variation for each site type was calculated and compared with the population mean residual. Z scores and their associated P-values are given in Table 1. Coffee plantations had a significant positive mean deviation from the morphotype accumulation curve, while sacred groves had a significant negative mean deviation. This implied that whilst very few sporocarps were produced in coffee plantations, a collection from a 0·125-ha plot was more diverse than an average sample of a similar number of sporocarps from all the forest types combined. Sacred groves produced large numbers of sporocarps but a collection from a 0·125-ha plot was less diverse than an average collection of similar size from all forest types.
Table 1. A comparison of mean deviations for plots from three forest types from a negative exponential regression through morphotype accumulation data from all forest types combined
The NMDS ordination plot (Fig. 3) shows that, with the exception of a very small number of outliers, assemblages of sporocarp morphotypes in sacred groves across south-western Kodagu were remarkably similar to one another, and distinct from those found in coffee plantations and forest reserves. Table 2 presents the results of the anosim, values for pairwise comparisons between forest types and a global R-value. Sacred groves were shown to be significantly different from other forest types, but the compositional differences between coffee plantations and reserved forests were not significant.
Table 2. Results of an analysis of similarities (anosim) between samples from three forest types, coffee plantations (C, n= 23), forest reserves (R, n= 10) and sacred groves (S, n= 25)
C, R, S
simper calculated the percentage contribution of each morphotype to the total dissimilarity between forest types. A very similar group of morphotypes was responsible for the differences between sacred groves and the other two forest types. The same four morphotypes ranked highest in the comparison of sacred groves with both coffee plantations and forest reserves (Table 3). They all produced abundant sporocarps in sacred groves but were absent or produced very few sporocarps in both coffee plantations and forest reserves. Xylaria sp. (rod like) and Xylaria sp. (club shaped) are ascomycete fungi that were found, preferentially, on dead wood. Similarly, Ganoderma sp. (orange coloured) is a basidiomycete polyporous fungus known to be wood rotting and found on dead wood in our study. Mycena sp. (leaf) is a fleshy fungus in the family Tricholomataceae that was often found growing on leaf litter.
Table 3. The four macrofungal morphotypes that made the greatest percentage contribution to the total dissimilarity between sacred groves and the other two forest types (rank in parentheses)
Xylaria sp. (rod like)
Ganoderma sp. (orange coloured)
Mycena sp. (leaf)
Xylaria sp. (club shaped)
A Mantel test found no significant correlation between geographical distance between sites and differences in macrofungal assemblage composition. However, a highly significant correlation (P = 0·0001, r= 0·281) was found between tree and fungal compositional differences.
Neither (loge) forest patch area nor (loge) isolation accounted for a significant proportion of the total variation in macrofungal richness in sacred groves (loge area, F= 0·22, P= 0·646; loge isolation: F= 0·01, P= 0·938) in our GLM analysis.
Neither patch size nor isolation were significant predictors of the presence/absence of any macrofungal morphotype in a binary logistic regression. No overall model was significant according to the model log-likelihood statistic, at a level adjusted by Bonferroni correction to keep the global level of significance at 0·05.
Many macrofungal species are believed to fruit sporadically, with no consistent pattern of occurrence from year to year (Watling 1995). Furthermore, their sporocarps are ephemeral and, even when produced, may last only a few days before decomposing or being eaten. Many years of intensive surveys may be required to describe the macrofungal communities of a particular area adequately (Tofts & Orton 1998). Although we carried out a meticulous search of a small area at each site on three separate occasions, as our survey of each site was limited to a single year it is likely that we observed only a proportion of the entire fungal assemblage. However, there is no reason for us to suspect that the proportion of the assemblage seen in each site type differed significantly. For this reason we believe that the comparisons that we have made between sites are valid.
There are several possible explanations for low sporocarp abundance in coffee plantations. They have a more open canopy than sacred groves and forest reserves. Consequently they will experience higher light levels, higher temperatures and lower humidity. These environmental differences may have contributed to low sporocarp production. Intensive management of plantations may also have an effect. Fallen dead wood is removed and understorey vegetation is cut back. Coffee bushes are typically fertilized and sprayed with fungicide to control the coffee rust fungus Hemileia vastatrix. These management actions may also reduce fungal abundance.
Although there are comparatively few sporocarps found in coffee plantations, the few that are found form a relatively diverse macrofungal morphotype assemblage. Resource abundance, host diversity and habitat heterogeneity have all been linked to fungal diversity in tropical forests (Lodge 1997). However, in this case we believe that high diversity is a consequence of strong spatial autocorrelation in plot-based inventories of sporocarps (Schmit, Murphy & Mueller 1999). Some fungal colonies may be very large (Ferguson et al. 2003) and produce sporocarps across a sizeable patch of forest. Where sporocarps are abundant a species accumulation curve may rise comparatively slowly when based on an inventory of sporocarps in a plot that falls entirely within such a patch.
We found a distinct assemblage of fungal morphotypes in sacred groves. This assemblage differed from those found in coffee plantations and forest reserves because of the presence of species that decompose dead wood and litter. Dead wood is removed from coffee plantations as part of normal maintenance procedures and is therefore artificially scarce in these types of site. We hypothesize that dead wood is more abundant in sacred groves than in reserved forest. Tree falls and standing dead trees have been shown to be more common in forest fragments and near forest edges (Laurance et al. 1997; Peltonen 1999). Xylariales are wood-decomposing saprotrophs and weak parasites with many tropical species (Rogers, Callan & Samuels 1987). Members of this group may live as endophytes within living trees, and initiate wood decay when a branch dies so that decay starts within the canopy (Rayner & Boddy 1988). They would therefore be expected to be more abundant in trees with dead and damaged branches and in the litter of fallen dead branches and twigs that they shed. Species of Mycena and Marasmius are typically found as wood and leaf litter saprotrophs (Alexopoulos, Mims & Blackwell 1996). Sacred groves may also have other unique habitat features associated with their small size and strong edge effects, which foster a distinct macrofungal assemblage.
We found no correlation between the similarity of macrofungal assemblages and their geographical proximity. This implies that dispersal is not a significant constraint on macrofungal species in this environment and that assemblage composition is determined more by habitat requirements or internal population dynamics. The congruence of similarity patterns between our sites in their tree flora and macrofungal assemblages supports the idea that habitat requirements are the most important determinant of species composition. Further support comes from the absence of any relationship between the size and isolation of sacred groves and their macrofungal richness. Isolation and patch size did not explain the presence/absence of any macrofungal morphotype. These results imply that macrofungal distribution patterns at a landscape scale are determined by habitat requirements rather than dispersal or local population dynamics.
There are two important implications of this work for conservation policy. (i) Despite a long history of isolation from continuous forest, sacred groves had the highest sporocarp abundance and the greatest morphotype richness per sample area. The large areas of degraded and modified forest found in and around coffee plantations had the lowest sporocarp abundance and morphotype richness. We conclude that habitat degradation is a more serious threat to fungal diversity in the Western Ghats than fragmentation. Quality rather than quantity of habitat appears to be the key to successful fungal conservation.
(ii) Our study has shown that although small, sacred groves are important for fungus conservation because they provide unique types of habitat that sustain a distinct fungal assemblage. The strong cross-taxon congruence in similarity patterns between trees and macrofungi revealed by our study implies that future conservation efforts for fungi should be directed towards the protection of a network of complementary habitats rather than large, undisturbed areas.
Although there is a growing awareness amongst conservation managers of the importance of mutualisms involving fungi for the survival of many other organisms, and an appreciation of their functional importance, fungi are rarely considered when developing conservation priorities. Our research suggests that optimum strategies for conservation of fungal diversity may not necessarily coincide with those for plants and animals.
This project was funded by a research grant from the Conservation, Food and Health Foundation, Boston, MA, USA. S. A. Bhagwat's doctoral research was supported by the Rhodes Trust, the Radhakrishnan Memorial Bequest, Linacre College and the University of Oxford Graduate Studies Committee. We thank M. D. Ashfaq, K. T. Boraiah, H. R. Kamal Kumar, K. M. Nanaya, C. Shivanad and B. S. Tambat for field assistance, and C. G. Kushalappa for facilitating the fieldwork. Prasad Lamrood assisted in the identification of fungi and Caroline Bampfylde with data analysis.