4 The Methodology of Bioindication
4.2 Indicator values
group of invertebrates, i.e. ground beetles, and common methodology in a similar landscape, i.e. urban mosaics (Niemelä et al., 2000b). The outcome of the programme to date has shown some success, with generality in the effect of urbanization on the composition of ground beetle assemblages found across several cities and continents (Ishitani et al., 2003). Approaches, such as GLOBENET, present underexploited, yet potentially powerful opportun- ities for developing general bioindicators. However, because few such case studies exist, it is not possible to estimate under which circumstances or how commonly generality is likely to be found. None the less, the criterion of generality is not a necessary requirement for the successful local application of a bioindication system.
open, mixed woodland. Detector species may be more useful indicators of the direction of change than highly specific (characteristic) species restricted to a single state (Fig. 7.6). This is because the abundances (and thus the fidelity) of characteristic species may decline rapidly under changing environmental conditions to the point where they are regarded as vulnerable (Fig. 7.6). These species will become increasingly difficult to sample (Fig. 7.6), and may disap- pear rapidly with no further value for monitoring thereafter. Characteristic indicator species also provide no information on the direction of ecological change (although changes in their abundance may remain useful for moni- toring within the habitat to which they are specific), because they are highly specific and thus restricted to a single ecological state (Fig. 7.6). By contrast, species with moderate specificity levels (detector species, Fig. 7.6), may thus be more useful for monitoring change. Bioindication using insects in aquatic and soil systems makes use of species such as these that have a range of pref- erences for different environmental states (e.g. van Straalen and Verhoef, 1997; Mouillot et al., 2002), but this distinction has less commonly been made in above-ground terrestrial bioindication.
The IndVal method has several advantages over other indicator measures used for ecological bioindication (McGeoch and Chown, 1998). For example, the IndVal is calculated independently for each species, and there is complete flexibility with regard to the state (site, sample or habitat) categorization on which the IndVal measures are based (McGeoch and Chown, 1998). However, although habitat specificity is a comparatively inflexible species-specific trait (Southwood, 1988; Blackburn and Gaston, 2005), the abundances of spe- cies (and thus their fidelity) are likely to vary as a consequence of stochas- tic, seasonal, as well as disturbance factors. Insect abundances may also be higher under disturbed than undisturbed condition, as shown for dung bee- tles in coffee plantations versus cloud forest in Mexico (Pineda et al., 2005).
Abundance will thus not have a straightforward relationship with the EP of interest, resulting in potential problems with the interpretation of the IndVal (Hiddink, 2005). The sensitivity of the IndVal to such changes will thus ultim- ately determine its usefulness for bioindication.
Medium
Medium
High
High
Tramp Rural
Indicator Detector species
Indicator Characteristic
species Vulnerable
Fidelity (occupancy) Low
Low
Specificity
Fig. 7.6. Species characterized by a combination of their degree of environmental specifi city and fi delity, and classifi ed on this basis as either indicators (characteristic or detector species), tramp, rural or vulnerable species. (Redrawn with permission from McGeoch et al., 2002.)
Indeed, a comprehensive understanding of the behaviour and properties of indicator measures and indices in bioindication is critical (e.g. Chovanec and Waringer, 2001; Allegro and Sciaky, 2003; Garcia-Criado et al., 2005). This must include an understanding of the formal relationship between index components, such as the fidelity and specificity components of the IndVal index. Failure to examine the properties of indicator measures or indices prior to their application can result in the misinterpretation of outcome values, the failure to recognize complex changes in the relationships between sys- tem components on which the aggregate measure is based, compounding of biases as a consequence of uncertainty or high variability in the constituent components and loss of information (Cousins, 1991; Gaston, 1996; Eiswerth and Haney, 2001; Niemi and McDonald, 2004; Loh et al., 2005). Using a dung beetle assemblage, McGeoch et al. (2002) showed that species with signifi- cant, high IndVal tended to remain so when tested in different locations and at a different time (Fig. 7.7a). Although the fidelity component of IndVal is sensitive to species abundance fluctuations (Fig. 7.7b), the fidelity value
0 20 40 60 80 100
Indicator value t1 0
10 20 30 40 50 60 70 80 90
Change in Indicator value [t1–t2]
(a)
0 1 2 3 4
Abundance (log) 0.0
0.2 0.4 0.6 0.8 1.0
Fidelity
(b)
Fig. 7.7. Properties of the Indicator Value (IndVal) index of Dufrêne and Legendre (1997), stylized from McGeoch et al. (2002). (a) Relationship between the percentage indicator values of an assemblage of dung beetle species in one season and the change in this percentage 2 years later. (b) Relationship between the fi delity component (that lies between 0.0 and 1.0) of the IndVal and the logarithmically transformed abundance of species in the assemblage.
is calculated from relative, rather than absolute differences in the frequency of occurrence of a species across habitats. As a result, if the abundance of a species changes in a similar direction across environmental states of interest this may not affect a change in its fidelity value. Furthermore, the logistic nature of the relationship between fidelity and abundance (as well as the fact that abundance is logarithmic in the relationship) means that a substantial abundance change (over 1 order of magnitude) may not result in any change in fidelity (Fig. 7.7b). Properties such as these make the IndVal method a par- ticularly effective tool for ecological bioindication. More widespread appli- cation of the method is, however, necessary to establish the generality of its properties.