Measuring disturbance in tropical forests: a critique of the use of species–abundance models and indicator measures in general


Recent papers have considered the merits or otherwise of using species–abundance models to detect disturbance in tropical forests (Hill et al. 1995; Hamer et al. 1997; Nummelin 1998; Hill & Hamer 1998). The purpose of this paper is to argue: (1) that there are simpler, more reliable ways of assessing forest disturbance; and (2) that the use of species–abundance models to assess whether a forest is disturbed or not ignores the different types, degrees and scales of disturbance that exist, each of which potentially have a different impact on forest biodiversity.

Tropical forest biodiversity is undeniably threatened. Although estimates on the number of species in tropical forests vary greatly, there is a widespread assumption that at least half the world's species are concentrated in only 7% of the terrestrial land mass (e.g. Stork 1988; Myers 1990; Bibby et al. 1992; World Conservation Monitoring Centre 1992). As Hill et al. (1995), Hamer et al. (1997) and many others point out, tropical forest species are threatened by logging and forest clearance for a range of purposes, from the establishment of small farms as part of the practice of shifting cultivation to the large scale clearance of forest for pasture and plantation crops of exotic timber tree species, rubber, oil palm, and many other crops. Shifting cultivation and logging have been practiced in tropical forests for thousands of years, but the area partially or wholly cleared for these purposes has increased dramatically during the twentieth century (Phillips 1997). Thus, many authors rightly argue that there is an urgent need for more research on the impact of forest disturbance on biodiversity (e.g. Phillips 1997; Watt et al. 1997). In fact Hill & Hamer (1998) justify the development of methods of assessing whether or not a forest is disturbed as a precursor to the measurement of the impact of forest disturbance on tropical forest communities.

Hill et al. (1995) first suggested that species–abundance models could be used to indicate forest disturbance after finding that butterfly data from an unlogged area of forest on Buru, Indonesia, fitted a log-normal distribution, whereas the data for an area of logged forest did not. Nummelin (1998) tested the applicability of species–abundance models as indicators of disturbance by studying ladybirds (Coccinellidae), tortoise beetles (Cassidinae), dung beetles and plants in forest near Kibale, Uganda. Nummelin found that the species–abundance distributions fitted a log-normal distribution irrespective of whether they came from logged or unlogged forest. In a more recent paper on Indonesian butterflies (and therefore not available to Nummelin at the time of preparing his paper), Hamer et al. (1997) again showed that species–abundance data collected in Sumba, Indonesia, were affected by forest disturbance, although in this case only the butterflies in the most disturbed of three disturbed forest sites sampled showed a log-normal distribution. Hill & Hamer (1998) respond to Nummelin by listing a number of requirements that should be met for assessing the usefulness of species–abundance models as indicators of forest disturbance.

The purpose of this paper is not primarily to prolong the discussion on the usefulness of species–abundance models as indicators of disturbance. However, it is worth noting that the restrictions in the use of species–abundance models described by Hill & Hamer imply that this technique cannot be regarded as an easy ‘instantaneous indicator of forest disturbance’ (Hamer et al. 1997), but at best must be applied cautiously. In addition, although Hill et al. (1995) found that the presence of a log-normal distribution distinguished disturbed from undisturbed forest, data from only one of the three disturbed forest sites sampled by Hamer et al. (1997) showed a log-normal distribution.

A more important point, however, is that it is surely unnecessary to sample butterflies to assess whether or not a forest has been disturbed. It is, of course, possible to detect extreme disturbance in the form of complete deforestation, both from the ground and from satellite (e.g. INPE 1997). Less severe disturbance in the form of selective logging may also be assessed through the presence of logging roads and trails; the soil compaction caused by the removal of logs resulting in poor growth of vegetation over long periods of time (e.g. Pinard, Howlett & Davidson 1996; Guariguata & Dupuy 1997; Whitman, Brokaw & Hagan 1997). For example, Nummelin (1998) noted that ‘15–20 years after logging, logged areas were still easily recognizable due to the large gaps in the canopy’ and Hill et al. (1995) sampled logged forest ‘adjacent to a logging road’. Even more simply, information on disturbance history can also be obtained from local people (directly or indirectly), logging companies and government forestry departments. Presumably, this was how Hill et al. (1995), Nummelin (1998) and Hamer et al. (1997) selected their sites.

These direct ways of assessing forest disturbance – visual evidence or local records of logging history – have the merits of being simpler, more reliable, free of assumptions about the impact of disturbance and faster than using species–abundance models.

The second problem with the use of species–abundance models to assess forest disturbance is that, even if they could be used to detect accurately whether or not a forest had been disturbed in some way, they provide no quantitative measure of the degree of forest disturbance: forests are either classified as being disturbed or not. However, forest disturbance takes many forms, some of them natural, such as lightning and hurricane damage (e.g. Walker et al. 1996; Middleton et al. 1997), and many others of an anthropogenic nature (e.g. Phillips 1997; Whitman, Brokaw & Hagan 1997). It is surely more important to measure the impact on biodiversity of different forms and degrees of disturbance (and related aspects such as the time since disturbance and the relationships between the effects of disturbance on different groups of species) than to assess whether or not disturbance, of any form, has occurred.

There is increasing evidence that disturbance has an impact on forest biodiversity. For example, Belshaw & Bolton (1993) examined the impact of replacement of natural forest with cocoa plantations on litter ants and Chey, Holloway & Speight (1997) compared the diversity of moths in intact forest and plantations in Sabah, Malaysia. There have been a few studies on the effects of selective logging on biodiversity [e.g. Holloway, Kirk-Spriggs & Chey (1992) and see below] and even fewer studies on the effect of ‘disturbance gradients’ on biodiversity (Eggleton et al. 1995, 1996; Jullien & Thiollay 1996; Watt et al. 1997; Lawton et al. 1998). These, and other, studies suggest that different types and scales of disturbance have different effects and that different groups of species, even different groups of insects, respond to the same form of disturbance in different ways.

This paper should not be taken as an overall criticism of the work done by Hill et al. (1995) and Hamer et al. (1997). Indeed, their work is extremely valuable in demonstrating how disturbance in the form of logging affects butterflies. In addition to describing the species–abundance patterns of butterfly communities in different sites, they described the diversity and species composition of butterflies in those sites. They used a variety of measures for diversity (in its widest sense): both species richness and diversity indices. Views on diversity indices and the value of indices in general can be found in Magurran (1988) and Pielou (1995), respectively. Although different ecologists prefer different measures of diversity, Hill, Hamer and co-workers usefully present a wide range. Most usefully of all, perhaps, they present an analysis of the geographic range of the species found in different forest sites. This work is unique in several ways (particularly in examining the effect of logging on the taxonomic distinctiveness of the insect community) and much more of its kind is needed.

In conclusion, species–abundance models are not a useful method for determining whether or not a forest has been disturbed. Forests are disturbed in many ways and research should not focus on trying to find an ‘ecological test’ to detect whether disturbance of any form has occurred. Rather, research should focus on quantifying the impact of the many forms of disturbance on biodiversity as a basis for minimising the potentially serious threat of land use change on biodiversity.


I thank Gerry Lawson, Roger Leakey and Oliver Phillips for their comments on earlier drafts of this paper.

Received 28 February 1998