JPE589.fm Page 484 Wednesday, April 11, 2001 10:56 AM
Journal of Applied Ecology 2001 38, 484 – 486
FORUM
Oxford, Journal JAPPL British 20021-8901 38 Measures 589 S.J. Blackwell 001Willott Ecological of UK of Science, Applied sampling Society, Ltd Ecology effort 2001
Graphicraft Limited, Hong Kong 000
Species accumulation curves and the measure of sampling effort S.J. WILLOTT Centre for Biodiversity and Conservation, School of Biology, University of Leeds, Leeds LS2 9JT, UK
Summary 1. Moreno & Halffter (2000) described the problems associated with comparing species richness among communities that have inventories compiled using different methods or with different sampling effort. They used species accumulation curves to standardize samples among sites, to predict the species richness of sites and to estimate the minimum effort required for adequate completeness of inventories. 2. I argue that their measure of sampling effort, number of nights, is inappropriate because it does not distinguish between genuine differences in species richness among sites and differences in trap efficiency. The number of individuals is the best measure of sampling effort to avoid this problem, as illustrated by data on moths collected from a Bornean rainforest. Furthermore, the approach of Moreno & Halffter requires the species accumulation curves to be approaching an asymptote for accurate estimation, and so for practical reasons is probably limited to less diverse taxa. Key-words: biodiversity, moths, species richness, trap efficiency. Journal of Applied Ecology (2001) 38, 484–486
A recent paper by Moreno & Halffter (2000) described the problems associated with comparing species richness among communities that have inventories compiled using different methods or with different sampling effort. They used species accumulation curves to compare the completeness of inventories of bats sampled with mist nets in different vegetation types in Mexico. Different models were applied to these data to predict the total species richness of sites, and the authors proposed that sampling 90% of the predicted species richness is an acceptable level from which to compare communities. I strongly agree that species accumulation curves are an essential component of this kind of work, and they have been used elsewhere to extrapolate species richness (Soberón & Llorente 1993) and, in a somewhat different way, to investigate how the alpha diversity of habitats or sites compares with the gamma diversity of a wider area (DeVries, Murray & Lande 1997; Willott 1999). Also, a critical assessment of the minimum sampling effort required for an adequate inventory is a useful development. However, I question the measure of sampling effort used by Moreno & Halffter, and how applicable their approach is likely to be for very diverse groups. Moreno & Halffter discuss the importance of standardizing sampling effort among sites, favouring © 2001 British Ecological Society
Correspondence: S.J. Willott (fax 0113 233 2835; e-mail
[email protected]).
the number of net metres per hour as their measure, although their subsequent analyses simply use number of nights trapping. I would argue that neither of these measures is the most appropriate. Population censuses are affected by differences in physical structure, vegetation structure or environmental variables among habitats, and these confound comparisons (Sutherland 1996). Unless authors are confident that trapping efficiency or detection probability is the same in each habitat, and present evidence for this (Heydon & Bulloh 1997), I suggest that they use the number of individuals as the measure of sampling effort when constructing species accumulation curves. This will avoid confounding genuine differences in species richness between sites with differences in trap efficiency or observer bias. A further advantage is that it does not require knowledge of the biology of the species. Moreno & Halffter did not sample on moonlit nights because bats are known to be lunar phobic. Had they not known this, and had sampled on all nights, but with more moonlit nights in one habitat relative to the others, then they may have erroneously concluded that this habitat was less species rich. The use of numbers of individuals as the sampling effort would not have produced an underestimate of species richness because of these nights of low activity. Such knowledge of the biology of species is unlikely to be available for many taxa, particularly in the tropics, so appropriate corrections for bias in sampling cannot be made.
JPE589.fm Page 485 Wednesday, April 11, 2001 10:56 AM
485 Measures of sampling effort
Table 1. Measures of sampling effort, species richness and diversity for moths collected in tropical rain forest near the Danum Valley Field Centre, Borneo. Abbreviations: C, canopy sites; L, logged forest sites; P, primary forest understorey sites; S, total number of species; N, total number of individuals; alpha, diversity index (alpha of the log-series); D 1000, number of nights of sampling required to collect 1000 individuals; S1000, number of species in a random sample of 1000 individuals; S8D, number of species after 8 nights of sampling; N 8D, number of individuals after 8 nights of sampling; alpha 8D, diversity after 8 nights of sampling Site
S
N
alpha*
D1000
S1000*
S8D
N8D
alpha8D*
C1 C2 L1 L2 P1 P2 P3 P4
391 389 392 368 459 344 375 352
1175 1138 1323 1061 1486 1115 1058 1128
205 ± 19 210 ± 20 189 ± 16 200 ± 20 226 ± 19 170 ± 16 207 ± 20 176 ± 17
17 7 14 9 13 10 20 13
356 ± 9 363 ± 17 338 ± 10 355 ± 5 367 ± 12 323 ± 6 362 ± 6 329 ± 7
178 389 233 342 250 311 213 219
417 1138 569 948 561 934 438 521
118 ± 19 209 ± 20 147 ± 20 192 ± 20 173 ± 24 163 ± 17 164 ± 26 142 ± 20
*± 95% confidence interval.
© 2001 British Ecological Society, Journal of Applied Ecology, 38, 484– 486
The problems arising from the choice of measure of sampling effort may not be severe if the results using the different units are strongly correlated. However, they may not be, as illustrated using a data set of moths sampled from eight sites in tropical forest around the Danum Valley Field Centre in Sabah, Malaysian Borneo. A full description of the site and methods may be found in Willott (1999). For each site, I present different measures of species richness, sampling effort and diversity. Species richness is measured as S, the total number of species in the sample; S1000, number of species in a random sample of 1000 individuals; S8D, number of species after 8 nights of sampling. Sampling effort is measured as N, the total number of individuals in the sample; D1000, number of nights of sampling required to collect 1000 individuals; N8D, number of individuals after 8 nights of sampling. The diversity index used is alpha of the log-series, where alpha is the diversity of the total sample, and alpha8D is the diversity after 8 nights of sampling. Table 1 shows the different measures of sampling effort, species richness and diversity among sites. Data were collected until there was a minimum of 1000 individuals collected from each site, but between 7 and 20 nights per site were required to achieve this. Comparing two sites as an example, site P3 required twice as many nights to collect 1000 individuals compared with site P2. After a comparable number of nights, species richness of site P2 (S8D) was greater than that of P3, but this was a result of the number of individuals sampled: with equal numbers of individuals (S1000), site P3 had the greater species richness. As noted by Moreno & Halffter, there are problems comparing species inventories of sites where there is no indication of how complete the inventory is, or when different methods have been used. To this list can be added the problem that the use of different measures of sampling effort may alter the conclusions. Using all sites of the moth data set, there were no significant correlations among estimates of species richness and diversity using equivalent numbers of individuals as the measure of sampling effort (S1000, alpha) and
Table 2. Correlation coefficients (above) and P-values (below) among measures of species richness and diversity with different measures of sampling effort (n = 8). Data and abbreviations from Table 1
S8D alpha8D
S1000
alpha
0·084 (0·843) 0·382 (0·349)
0·017 (0·968) 0·319 (0·441)
equivalent numbers of nights of sampling (S8D, alpha8D) (Table 2). The choice of units of ‘equal’ sampling effort strongly affects conclusions about the relative richness and diversity of sites. I applied the methods of Moreno & Halffter to the moth data from sites P2 and P3. Data were randomized using the EstimateS software (Colwell 1999), with number of individuals or number of nights as the sampling unit. The species accumulation curves are presented in Fig. 1. Using the number of individuals (Fig. 1a), site P3 was the most species rich for equal sampling effort, whereas using the number of nights (Fig. 1b), site P2 was the most species rich. The linear dependence (LDM) and Clench models were fitted to the randomized data to estimate the lower and upper limits of the asymptotic total species richness. The Clench model predicted greater species richness than the LDM model in each case (Table 3), but the relative magnitude of total species richness of sites predicted by the models was reversed. Using number of individuals, the LDM and Clench models predicted that site P3 had approximately 23% or 29% more species, respectively, whereas using number of nights, site P2 had 24% or 14% more species. The choice of units of ‘equal’ sampling effort may strongly affect predictions about the relative total species richness of sites. Moreno & Halffter conclude that their methods may be applied to any other biological group. This is true in principle, but there is a major caveat: they are
JPE589.fm Page 486 Wednesday, April 11, 2001 10:56 AM
486 S.J. Willott
400
richness when only a relatively small proportion of the assemblage has been sampled produces large errors and the result is strongly dependent on the estimator used (Colwell & Coddington 1994). It is in the hyperdiverse groups of arthropods, nematodes, fungi and microorganisms where these problems of sampling and estimation are most acute and to which the methods of Moreno & Halffter are unlikely to be applicable.
(a)
350 300 250 200 150 100
P2 P3
50 0 Species
0
400
200
400 600 800 1000 Individuals
Acknowledgements
(b)
350 300 250 200 150 100 50 0
0
2
4
6 Nights
8
10
Fig. 1. Randomized species accumulation curves for sites P2 and P3 generated using (a) number of individuals and (b) number of nights as the measure of sampling effort.
Table 3. Total species richness of sites P2 and P3 predicted by fitting the linear dependence model (LDM) and Clench model to data randomized using either the number of individuals or nights as the measure of sampling effort Model Site
LDM
Clench
P2 P3
396 489
592 761
P2 P3
483 390
739 649
Individuals
Nights
© 2001 British Ecological Society, Journal of Applied Ecology, 38, 484–486
only likely to yield meaningful results where the species accumulation curves have reached, or are clearly approaching, an asymptote. This may be practical in assemblages of mammals (the gamma diversity of the bats across all the sites of Moreno & Halffter was 20 species), and even in relatively less diverse assemblages of insects in the tropics (e.g. 130 species of butterflies; DeVries, Murray & Lande 1997). It is not likely for very diverse assemblages where there may be thousands of species in an area and an asymptote is not reached even after extensive sampling (e.g. moths: Willott 1999; beetles: Didham et al. 1998). Extrapolating total species
Data collection was funded by a Royal Society Post Doctoral Fellowship. Permission to work in Sabah was granted by the Economic Planning Unit of the Malaysian Government and by the State Secretary, Internal Affairs and Research, Chief Ministers Department, and to work at Danum by Yayasan Sabah (Forestry Division) and the Danum Valley Management Committee. I thank Jamal Majid, Stephen Sutton, Sheila Wright and Darline Lim for help with sampling, Gaden Robinson for a programme to calculate alpha, and Stephen Hartley for assistance with computing. This is project DV85, paper number A/316 of the Royal Society’s South-east Asian Rain Forest Research Programme.
References Colwell, R.K. (1999) EstimateS 5: Statistical Estimation of Species Richness and Shared Species from Samples, Version 5·0·1 User’s Guide. http://viceroy.eeb.uconn.edu/estimates [accessed on 27/03/00]. University of Connecticut, Storrs, CT. Colwell, R.K. & Coddington, J.A. (1994) Estimating terrestrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society of London, Series B, 345, 101–118. DeVries, P.J., Murray, D. & Lande, R. (1997) Species diversity in vertical, horizontal, and temporal dimensions of a fruitfeeding butterfly community in an Ecuadorian rainforest. Biological Journal of the Linnean Society, 62, 343–364. Didham, R.K., Hammond, P.M., Lawton, J.H., Eggleton, P. & Stork, N.E. (1998) Beetle species responses to tropical forest fragmentation. Ecological Monographs, 68, 295–323. Heydon, M.J. & Bulloh, P. (1997) Mousedeer densities in a tropical rainforest: the impact of selective logging. Journal of Applied Ecology, 34, 484 –496. Moreno, C.E. & Halffter, G. (2000) Assessing the completeness of bat biodiversity inventories using species accumulation curves. Journal of Applied Ecology, 37, 149–158. Soberón, J. & Llorente, J. (1993) The use of species accumulation functions for the prediction of species richness. Conservation Biology, 7, 480–488. Sutherland, W.J. (1996) The twenty commonest censusing sins. Ecological Census Techniques: A Handbook (ed. W.J. Sutherland), pp. 317–318. Cambridge University Press, Cambridge, UK. Willott, S.J. (1999) The effects of selective logging on the distribution of moths in a Bornean rain forest. Philosophical Transactions of the Royal Society of London, Series B, 354, 1783 –1790. Received 3 April 2000; revision received 13 August 2000
