SEARCH

SEARCH BY CITATION

Keywords:

  • Biodiversity;
  • braconid;
  • doubletons;
  • singletons;
  • species richness;
  • taxonomic bias

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Abstract.  1. Knowledge of global diversity patterns is important for research into the factors that shape them, and for systematic conservation planning. However, most species inventories are incomplete and biased towards conspicuous, charismatic, geographically widespread, and temperate species. These biases hamper attempts to gain a clear view of underlying diversity patterns, and compromise conservation plans that are based upon what is known.

2. Here we investigate this problem using two methods to estimate species diversity in the parasitic wasp family Braconidae. The first quantifies the effect of taxonomic revisions on species numbers within genera to estimate the present level of underdescription. The second additionally considers the numbers of specimens referred to in descriptions and revisions.

3. Modelling underdescription as a function of region and body size shows that research carried out thus far displays significant geographical and taxonomic biases.

4. Correcting for these biases affects the distribution of inferred undiscovered diversity among braconid subfamilies and among regions, as well as the total diversity estimates for the family.

5. The geographic distribution of levels of underdescription also has implications for latitudinal diversity gradients. Weak or non-existent gradients in some taxa may be caused simply by differences in the number of undescribed species between tropical and temperate regions.

6. Such analyses can enlighten researchers as to where, taxonomically and geographically, research should be directed to economically improve species richness estimates. In the case of braconid wasps the greatest gains are to be made in Africa and southern America, and for the Braconinae and Microgastrinae.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Gaining an accurate knowledge of global biodiversity patterns is a pressing concern for biologists seeking to determine both the actual, and potential, anthropogenic impact upon life on earth (Purvis & Hector, 2000; Mace et al., 2003, 2005; Wilson, 2003). The ability to design effective large-scale conservation strategies intended to address this impact can be greatly increased by a knowledge of what species exist, their abundances and their distributions.

Although there are numerous metrics for biological diversity (Purvis & Hector, 2000; Magurran, 2004), here we concentrate on species richness, the most commonly used biodiversity indicator (see Gaston, 1996; for a review). Quantifying species richness in a particular area, community, or taxonomic group is extremely onerous, and especially so for megadiverse groups. Species richness is usually estimated using one of several sampling methods from the geographical area, community or taxonomic group in question (Gaston, 1996; Purvis & Hector, 2000; Magurran, 2004). Regardless of the method used, the numbers of described species on which these models and estimates are based, are invariably incomplete samples from the total species pool. In addition, and more statistically troubling, these samples are typically biased taxonomically, geographically, or both (Gaston, 1991; Allsopp, 1997; Cabrero-Sanudo & Lobo, 2003; Collen et al., 2004). An extreme example is Erwin’s (1982) famous study of arthropod diversity in a tropical forest canopy which was extrapolated to give a global species diversity estimate of 30 million arthropod species. Sampling such as this cannot be considered random, thus breaking a primary assumption of many commonly used statistical techniques.

The reasons for these biases in the data lie in the fact that the sampling regime is often not designed for the purposes of estimating species richness. Instead, data are commonly extracted from disparate studies with differing aims, which often focus on convenient geographical areas and on large showier taxa.

Taxonomic bias may also be generated as a side-effect of body size because smaller-bodied taxa tend to be described later than large-bodied taxa (Gaston, 1991; Cabrero-Sanudo & Lobo, 2003; Collen et al., 2004). Furthermore, less abundant species have a reduced probability of collection and, therefore, of description (Gaston et al., 1995). Small geographic range size may also reduce the probabilities of discovery and description (Cabrero-Sanudo & Lobo, 2003; Adamowicz & Purvis, 2005). Added to these factors, geographic bias is typically skewed towards temperate regions simply because, historically, most of the world’s taxonomists have been located in temperate regions and these regions have been scrutinised for a longer period of time than tropical regions (Allsopp, 1997; Cabrero-Sanudo & Lobo, 2003; Collen et al., 2004). Other reasons for geographic bias include the ease of travel to particular locations (specimen-level data for terrestrial invertebrates are often clustered around roads and railway stations), and the non-random availability of expertise, museums or technical resources that are required for the discovery, description and curation of specimens.

Collectively, these problems mean that methods involving extrapolation can never hope to produce unequivocal estimates. The estimates that are produced typically generate considerable debate in the scientific literature. Some scientists invariably regard them as gross overestimates while other authors, being aware that relatively small organisms may well have been overlooked, regard them as underestimates (Blackburn & Gaston, 1994; Zapata & Robertson, 2007). Nevertheless, given the precision that is required of most estimates and the limited funds and person-hours available to carry out sampling, most workers are initially satisfied with ‘ballpark’ estimates which can later be refined through more directed studies if required.

Here we investigate these problems using the Braconidae, a family of taxonomically poorly known parasitic wasps (Quicke, 1997), with approximately 17 000 recognised species (Yu et al., 2005). The family was used by Dolphin and Quicke (2001) to examine the use of rates of description of new species corrected for taxonomic effort. Using two methods, they estimated that the Braconidae included between 30 873 and 50 886 species. In this paper, we estimate the relationship between the ratio of the number of species before and after taxonomic revisions, and body size and region. We then use these relationships to predict total global braconid diversity from what is currently known about geographic and size distribution in the Braconidae. We show that there are many new braconid species to be discovered and make suggestions about where research (particularly taxonomic revisions) should be concentrated in order to refine these estimates in the most economical way.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

To acquire data on the taxonomic revision work carried out on the Braconidae, we turned to source papers on this group (sources are provided in the Appendix in Supporting Information). We found the majority of studies by searching the literature listed in the Taxapad database (Yu et al., 2005). Our search criteria were that the title of the paper must contain the phrase ‘revision’, but not ‘further revision’, ‘generic revision’ or ‘addition to (the) revision’. In addition, a new genus or species, or new synonymies, must have been established in the work. Slight permutations of the criteria, for example, including sources with the phrases ‘the species of’ or ‘review of’, did not qualitatively affect the results (see Table S1 in Supporting Information). The Taxapad database includes non-English literature, but provides the English translation of the titles for most entries. Nevertheless, we also searched and filtered for French and German translations of these terms, e.g. ‘Les espéces de’, ‘nouvelle révision’ and ‘arten der’.

We wanted to exclude revisionary work that was carried out at a time when many museum and reference collections could be regarded as immature, and when several important groups had not yet received much attention. Therefore, we restricted ourselves to works published after 1965 (i.e. after most museum collections had reached a reasonable level of maturity and many groups had previously been revised). We only included studies based on very limited areas such as single countries if the authors were clearly in a good position to know whether species encountered had been described previously from elsewhere, e.g. through collaboration or examinations of extralimital collections.

For mainland China, the vast majority of taxonomic literature that includes original descriptions is very recent and was published from 1980 onwards (>90%). This contrasts markedly with the figure of 33% for the rest of the world (see Figure S1 Supporting Information). In addition, unlike almost all other parts of the world of a comparable size, very few species have been described or recorded from mainland China during the past 250 years. These two observations lead us to regard the Chinese work as anomalous at the present time and we therefore excluded it from our analyses.

Assignment of studies to regions

In order to gain a clear picture of the geographical distribution of the taxonomic revision studies, we categorised the location of the studies into continental regions based on the Taxonomic Diversity Working Group recommendations (Brummitt, 2001). Where studies dealt with species from more than one such region we subdivided the data accordingly. We extracted the data for the known distribution, diversity, and body size of the Braconidae from Yu et al. (2005), and recorded body size as the longest recorded wing length for each genus.

With these criteria, we included a total of 114 published studies in our analyses (see the Appendix in Supporting Information for sources), and 46 subfamilies were represented. It was possible to split most (109) of the studies into genus-specific records, and by geographic region. Doing so resulted in 172 genus- and region-specific records. There were five studies where it was not possible to split the data up by geographic region, and these were excluded from the analysis.

Of the 172 records available for inclusion, the regional breakdown was as follows: 15 (8.72%) were from Africa, 41 (23.84%) were from Asia-Temperate, 22 (12.79%) were from Asia-Tropical, 11 (6.40%) were from Australia, 39 (22.67%) were from Europe, 30 (17.44%) were from northern America, and 14 (8.14%) were from southern America.

The Chao-1 estimator

To further refine our diversity estimates we made use of the Chao-1 estimator (Chao, 2005). Chao-1 is a simple estimator of the absolute number of species in an assemblage based upon the number of rare species present in a sample (those represented by only one or two individuals). For each study that provided sufficient data, we used the numbers of species known from only one individual (singletons) and from only two individuals (doubletons) to estimate ‘uncollected’ species richness using the Chao-1 estimator in a manner analogous to its use in ecological studies (Magurran, 2004), and as also applied by Meier and Dikow (2004) to estimate global species diversity in the robber fly genus Euscelidia (Asilidae, Diptera). Non-parametric estimators such as Chao-l were initially designed for ecological samples, rather than museum specimen data. One of the major assumptions of such estimators is that the data are obtained from random samples. It could be argued that museum data are not random. For example, (i) workers concentrate on particular favourite groups or on rare species; (ii) sampling methods are inconsistent between groups because of different researchers using different methods; and (iii) the extended time span of many museum collections could lead to the overestimation of standing species richness because species turnover is not considered (Petersen & Meier, 2003). On the other hand, it can also be argued that because museum collections involve the activity of many collectors over many years, they are actually less biased than other collections (Petersen & Meier, 2003). Although some researchers have found that such estimators often do not perform particularly well (e.g. Petersen & Meier, 2003; Petersen et al., 2003), others (e.g. Walther & Moore, 2005; Hortal et al., 2006) found that they perform as well, if not better, than other methods.

Modelling the effect of body size and region on the results of taxonomic revision

To investigate the effect of body size and region on the proportional change in species number following taxonomic revision, and the change using the Chao-1 estimator, we used analysis of covariance (ancova) models. Our response variables were proportional change in species number following revision, and proportional change in the Chao-1 estimate following revision: henceforth termed ‘multipliers’. Our explanatory variables were ‘region’ and ‘body size’. Although the use of ratios in regression analysis has been criticised (Atchley et al., 1976), these criticisms mainly revolve around regressions where the denominator of the response variable also appears as an explanatory variable, which is not the case here. In addition, Liermann et al. (2004) showed that, in general, the use of ratios does not adversely affect statistical power or type 1 error rate. We are therefore confident in the use of a ratio as our response variable.

We first loge-transformed both our response and explanatory variables in order to normalise their distributions. We added a constant of 0.1 before transforming in order to cope with zero values. We checked normality by a visual examination of the frequency distributions. We then fitted maximal models that included loge(body size), region, and the interaction between them, and subsequently simplified them by the backward elimination of non-significant terms following Crawley (2007).

Predicting undescribed diversity

For this part of the analysis we asked the question: If all braconid genera were to be revised in the same manner as previous revisions, what would the resulting number of species be? To answer this question we first used the ancova models (described above) to predict the loge(multiplier) for each genus, given a known region and body size. We did this within a Monte Carlo framework: For each genus we had a mean and prediction interval for the loge(multiplier) (derived from the deterministic models) and we used this information to define a normal distribution from which to draw a loge(multiplier) at random. We repeated this 1000 times to produce, after back-transformation, a distribution of likely multipliers for both the basic multiplier and the Chao-1 multiplier. We then applied the multipliers to the current diversity for each genus to estimate predicted diversity for the genus. These values could then be summed to provide an estimate of total diversity. The benefit of this approach is that it is more effective at characterising the uncertainty in the estimates of the underlying model than a deterministic approach, and it allows the robust estimation of error for our estimate of total diversity across the taxonomic groups within Braconidae.

Our analyses were carried out using R version 2.8.0 (2008-10-20) (R Development Core Team, 2008) on a powerpc-apple-darwin8.11.1 platform.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Models of change in diversity following taxonomic revision

Our analysis of the taxonomic revision data indicates that a significant amount of biological diversity remains to be discovered, but that this undiscovered diversity is distributed non-randomly. Across all of the studies included in the analysis, the average loge(multiplier) was 0.63 ± 0.056 (Fig. 1). If this multiplier alone were used to extrapolate braconid diversity (i.e. without correcting for body size and region) we would estimate an increase from the current estimate of 17 392 species to 32 741 [95% confidence limits (CL) = 29 360–36 511] braconid species. We now describe the patterns that are apparent after controlling for body size and region.

image

Figure 1.  Frequency distributions of the loge-transformed ratio of number of recognised species in a genus before and after taxonomic revision. The broken line indicates the population mean.

Download figure to PowerPoint

Our final model describing the loge-transformed proportional change in species number after taxonomic revision, which had an R2 of 0.35, showed that region and loge(body size) were both significant factors both as main effects [region: F = 9.68, d.f. = 6, P = 4.6e-09; loge(body size): F = 14.77, d.f.  = l, P = 0.00018] and in an interaction [region × loge(body size) interaction: F = 2.16, d.f. = 6, P = 0.05]. The model indicated that the relationship between loge(body size) and loge(multiplier) was approximately linear in all regions and that larger body sizes were associated with smaller multipliers in all regions (Fig. 2). This means that taxa with smaller body sizes tend to be more underdescribed than those with large body sizes. As body size increases, the multiplier decreases nonlinearly (it is linear on the log-scale) so that differences in body size at the smaller end of the size spectrum have a larger proportional effect on the multiplier than differences at the larger end of the size spectrum. The interaction term [region × loge(body size)] was significant, demonstrating that the effect of body size upon revision varies among regions.

image

Figure 2.  The relationship between loge(body size) and the loge-transformed ratio of the number of recognised species in a genus before and after taxonomic revision. Each point represents a single taxonomic revision at the genus level. The solid lines represent the predictions from an ancova model, while the broken lines represent the 95% confidence limits of the prediction.

Download figure to PowerPoint

Using our deterministic model, we predict that an increase in body size from 2 to 4 mm would result in a change in the absolute (rather than loge-transformed) value of the multiplier from 1.14 to 1.08 in Europe but 4.78 to 2.94 in northern America, and from 2.09 to 2.28 in Africa. Likewise, an increase in body size from 4 to 8 mm would result in a change in the multiplier from 1.08 to 1.02 in Europe but from 4.78 to 2.94 in northern America, and from 2.28 to 2.49 in Africa. To express this in a perhaps more intuitive way, we predict that the probability of discovering a new species of 36.19 mm wasp in the African region is approximately equivalent to the probability of discovering a new 3.89 mm wasp in the Holarctic [because their multipliers are the same (3)].

Refining diversity estimates using the Chao-1 estimator

In order to refine the above diversity estimates we made use of information reported in the taxonomic revision studies on the number of specimens held in museums and other collections to calculate Chao-1 diversity estimates (Chao, 2005). Of the 172 generic or higher revisions used in the regression above, 73 included sufficient data to permit the calculation of such estimates. Our models predicting the loge(Chao-1 multiplier) from loge(body size) and region showed that there was no significant additional effect of either body size or region (region: F = 1.89, d.f. = 8, P = 0.0778; body size: F = 0.113, d.f. = l, P = 0.0721) and that the best model was simply the intercept: a loge(multiplier) of 0.321 ± 0.0335. Thus, the inclusion of the Chao-1 estimate simply increases the diversity estimates of all taxa, regardless of body size and region, by 27.8% [exp(0.321) = 1.28] in addition to the effects of region and body size described above. Note that this is the estimate from the deterministic model and that the Monte Carlo-derived estimates differ slightly.

Predicting undescribed diversity

If we use our models to estimate diversity across the whole of the Braconidae by applying the multipliers derived from the Monte Carlo simulations to all the described genera within the Braconidae the results are instructive. They suggest that if all braconid genera were to be revised along similar lines to the source studies, there would be 31 952 (95% CL = 31 734–32 170) recognised braconid species worldwide, approximately doubling the number described to date.

If, in addition, we apply the multipliers obtained from the Chao-1 model to refine these estimates, the total species richness of the Braconidae would increase to 42 653 (95% CL = 42 342–42 963) species. Interestingly, at the subfamily level, the estimated proportion of undescribed diversity varies considerably (Fig. 3 and Table 1): we predict that most undescribed diversity is in the Braconinae and Microgastrinae.

image

Figure 3.  Current and estimated undescribed diversity for the subfamilies of the Braconidae. The black bars represent current diversity while the white bars indicate total predicted diversity. The error bars represent the 95% confidence interval for the estimate of total diversity.

Download figure to PowerPoint

Table 1.   Braconid subfamilies, the number of genera within them, the numbers of currently described species and the estimated number that would be recognised if all genera were revised based on both regressions of proportions of new species in taxonomic treatments and on regressions based upon the Chao-1 estimate.
SubfamilyGeneraCurrentPredictedMultiplierPredicted using Chao-1Chao-1 multiplier
  1. 95% confidence limits for the predictions are given in parentheses.

Acampsohelconinae3101183.65 (174.51–192.80)1.82242.29 (230.39–254.19)2.40
Agathidinae4910152094.34 (2057.56–2131.13)2.062809.63 (2755.20–2864.06)2.77
Alysiinae10119122377.83 (2357.45–2398.22)1.243177.30 (3146.23–3208.36)1.66
Amicrocentrinae2531.13 (28.15–34.11)6.2341.25 (37.28–45.22)8.25
Aphidiinae53558749.35 (743.69–755.01)1.341003.99 (995.63–1012.35)1.80
Apozyginae113.23 (3.05–3.41)3.234.36 (4.10–4.61)4.36
Betylobraconinae71628.38 (27.89–28.87)1.7738.21 (37.46–38.96)2.39
Blacinae12196314.72 (307.12–322.32)1.61423.23 (412.00–434.46)2.16
Brachistinae9380492.50 (485.26–499.73)1.30656.34 (645.66–667.03)1.73
Braconinae18127887398.70 (7261.03–7536.38)2.659857.59 (9665.10–10050.09)3.54
Cardiochilinae16206358.98 (350.47–367.49)1.74475.80 (462.82–488.78)2.31
Cenocoeliinae876131.17 (126.86–135.48)1.73175.42 (169.19–181.65)2.31
Charmontinae2811.53 (11.19–11.88)1.4415.56 (15.04–16.07)1.94
Cheloninae1812231716.41 (1684.82–1747.99)1.402282.07 (2236.91–2327.23)1.87
Dirrhopinae156.21 (6.03–6.39)1.248.32 (8.05–8.59)1.66
Doryctinae16513032839.18 (2811.63–2866.74)2.183787.44 (3746.35–3828.53)2.91
Ecnomiinae2910.32 (9.98–10.67)1.1513.50 (13.02–13.99)1.50
Euphorinae5010811680.93 (1662.29–1699.56)1.552256.10 (2227.93–2284.27)2.09
Exothecinae790119.41 (116.82–121.99)1.33159.66 (155.78–163.54)1.77
Gnamptodontinae460113.17 (110.99–115.35)1.89151.05 (147.76–154.34)2.52
Helconinae31212353.32 (348.42–358.22)1.67471.41 (464.39–478.42)2.22
Histeromerinae258.33 (8.07–8.60)1.6711.11 (10.72–11.49)2.22
Homolobinae358124.01 (118.12–129.90)2.14166.53 (156.62–176.45)2.87
Hormiinae15131247.43 (244.14–250.72)1.89331.19 (326.10–336.28)2.53
Hydrangeocolinae31957.18 (55.04–59.31)3.0176.13 (73.02–79.25)4.01
Ichneutinae1083133.13 (131.23–135.03)1.60177.79 (174.90–180.68)2.14
Khoikhoiinae2833.77 (32.36–35.19)4.2244.98 (42.96–47.01)5.62
Lysiterminae1389158.90 (156.64–161.17)1.79212.33 (208.79–215.88)2.39
Macrocentrinae8218308.00 (298.32–317.68)1.41411.00 (397.71–424.30)1.89
Masoninae1410.72 (10.36–11.07)2.6814.44 (13.88–14.99)3.61
Maxfischeriinae110.99 (0.93–1.05)0.991.31 (1.23–1.40)1.31
Mendesellinae21439.86 (38.31–41.41)2.8553.06 (50.79–55.32)3.79
Mesostoinae3915.49 (14.96–16.03)1.7220.73 (19.96–21.49)2.30
Meteorideinae21728.58 (27.80–29.36)1.6838.00 (36.85–39.14)2.24
Microgastrinae4620374024.73 (3894.92–4154.54)1.985369.83 (5181.74–5557.92)2.64
Microtypinae3912.67 (12.39–12.94)1.4116.88 (16.49–17.28)1.88
Miracinae23355.94 (54.73–57.15)1.7074.80 (72.98–76.62)2.27
Opiinae3318063469.04 (3397.10–3540.98)1.924639.82 (4531.87–4747.77)2.57
Orgilinae12345538.74 (526.66–550.82)1.56715.93 (699.35–732.51)2.08
Pambolinae1372141.01 (138.14–143.88)1.96188.41 (184.26–192.55)2.62
Pselaphaninae113.14 (2.95–3.32)3.144.22 (3.95–4.50)4.22
Rhysipolinae95393.93 (91.89–95.97)1.77125.53 (122.59–128.48)2.37
Rhyssalinae115783.72 (82.32–85.12)1.47111.83 (109.73–113.92)1.96
Rogadinae526801256.03 (1236.54–1275.52)1.851673.00 (1645.15–1700.85)2.46
Sigalphinae63676.46 (74.16–78.75)2.12102.36 (99.12–105.61)2.84
Trachypetinae383.48 (3.25–3.71)0.444.65 (4.30–5.01)0.58
Xiphozelinae21512.10 (11.52–12.69)0.8116.31 (15.45–17.18)1.09

Further analysis of the genus-level predictions revealed a significant negative association between loge(body size) for the genus and the estimated proportion of diversity that remains undescribed [GLM: delta AIC for removal of loge(body size) term = −208.04, d.f. = 2241 P = 1.3e-47]. This means that as the body size of the group increases, the proportion of diversity that remains undescribed decreases.

We can also use the above predictions to examine potential changes in our understanding of the geographical distribution of the group (Fig. 4). Current diversity is predominantly located in the temperate region of Asia (20.98%), with least diversity in Australasia (4.13%). However, our models predict that most diversity actually lies in Africa (40.12%), with least diversity in Australasia (2.19%).

image

Figure 4.  Current and estimated undescribed diversity for each region. The black bars represent current diversity while the white bars indicate total predicted diversity. The error bars represent the 95% confidence interval for the estimate of total predicted diversity.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We show that there is still considerable undescribed diversity within the Braconidae. There are about 17 000 recognised braconid species (Yu et al., 2005), and we now estimate that there is a total diversity of 42 653 (95% CL = 42 342–42 963) species and thus there may be as many as 25 911 species still to be described. These figures are similar to the total richness estimates of 30 873 and 50 886 species given by Dolphin and Quicke (2001) which were calculated using two different methods.

Furthermore, we demonstrate that the described diversity of the Braconidae is uneven both taxonomically and geographically: the multipliers vary systematically with taxon body size and region. Unevenness of sampling methods is a common feature of studies of biological diversity (Walther & Moore, 2005) and a failure to correct for this kind of bias would undoubtedly result in flawed estimates of biodiversity. In this case, a failure to control for taxonomic (body size) and geographic (region) bias results in a change in our species richness estimate: omitting body size reduces the estimate, while omitting region increases the estimate. However, the important point is not that the gross estimate is changed, it is that the estimate of where undescribed diversity lies in terms of geography and taxonomic group is markedly altered.

Perhaps more importantly, however, we show that the undescribed diversity is likely to be distributed in a predictable way, with, proportionally, most gains to be made in tropical regions and for smaller-bodied taxa. It has been noted that the observed latitudinal gradient of species richness is dependent on body size, with a smaller effect size for smaller organisms (Hillebrand & Azovsky, 2001). Hillebrand and Azovsky maintain that the lack of a gradient (or weak gradients) for small taxa is due to decreasing influence of regional and long-term processes on community structure and species richness associated with decreasing body size. Our results imply that there may be an alternative explanation: there may be a lack of an observed gradient simply because there are more undescribed species in the tropics than there are in temperate regions.

These results indicate that, as is common in studies aiming to predict biological diversity, there is significant bias in the research that has been carried out into this particular group. However, it is clear that the correction of such bias may be possible with the inclusion of ancillary data such as body size in models aiming to quantify undescribed diversity. In principle, the approach used here could be used to make finer-scale corrections and predictions. This approach would then allow diversity correction at a scale that would be relevant for systematic conservation planning. For example, given appropriate data from the revisions or specimens, it would be possible to correct for a range of factors including habitat type and accessibility in addition to geographic corrections at a finer scale than the regions used here.

We argue that scientists studying biodiversity should investigate these biases with the aim of correcting for them statistically and coordinating research to correct for bias through directed research in understudied areas. For example, here we predict that tropical wasps from taxonomic groups with a small body size are relatively underdescribed. This presents a strong case for an increase in research in this area, which should result in greatly improved biodiversity estimates for the group as a whole. Unfortunately, this is likely to be hampered at present because many museum collections, upon which such revisions are largely based, are biased towards the relatively larger and showier taxa. Although, in the case of parasitic wasps and similar groups, modern collecting procedures, such as the use of malaise traps, yield large numbers of the smaller-bodied species. However, these still require processing (mounting and sorting), which is time-consuming, and study, which is difficult. Furthermore, such collection efforts are highly unevenly distributed. Parasitic wasps in Costa Rica, for example, have been extensively studied (>150 malaise trap-years; Gaston (1996)) and an extensive parasitoid rearing programme exists (Smith et al., 2008), but similar programmes are still rare in the tropics, and those that do exist are not consistent in their methodology from place-to-place. We believe that a priority ought to be to attempt taxonomic sampling on a small but representative group of genera that should be selected randomly (rather than by selecting, for example, showier or larger-bodied taxa). In addition, the sampling ought to involve substantial new collecting according to a structured design permitting statistical analysis while minimising bias. The bulk of world multicellular diversity is composed of insects and yet its distribution remains poorly understood in comparison with vertebrate diversity (Stork, 2007), partly because of these sampling issues. There is still debate around the determinants of insect diversity and surrounding the potential drivers, and even the existence of, insect diversity gradients (Dyer et al., 2007; Novotny et al., 2007). Our methods provide another tool that will help biodiversity researchers gain a deeper understanding of insect diversity and distribution.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This work was supported by grant NE/C519583 from the Natural Environment Research Council. Bruno Walther and anonymous referees provided useful comments on an earlier draft of this work.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Adamowicz, S. & Purvis, A. (2005) How many branchiopod crustacean species are there? Quantifying the components of underestimation. Global Ecology and Biogeography, 14, 455468.
  • Allsopp, P. (1997) Probability of describing an Australian scarab beetle: influence of body size and distribution. Journal of Biogeography, 24, 717724.
  • Atchley, W., Gaskins, C. & Anderson, D. (1976) Statistical properties of ratios. 1. Empirical results. Systematic Zoology, 25, 137148.
  • Blackburn, T. & Gaston, K. (1994) Animal body-size distributions change as more species are described. Proceedings of the Royal Society of London Series B-Biological Sciences, 257, 293297.
  • Brummitt, R. (2001) World Geographical Scheme for Recording Plant Distributions. Hunt Institute for Botanical Documentation, Carnegie Mellon University, Pittsburgh, Pennsylvania, U.S.A.
  • Cabrero-Sanudo, F. & Lobo, J. (2003) Estimating the number of species not yet described and their characteristics: the case of Western Palaearctic dung beetle species (Coleoptera, Scarabaeoidea). Biodiversity and Conservation, 12, 147166.
  • Chao, A. (2005) Species richness estimation. In Encyclopedia of Statistical Sciences (ed. by N.Balakrishnan, C.Read and B.Vidakovic). pp. 79097916, Wiley, New York, U.S.A.
  • Collen, B., Purvis, A. & Gittleman, J. (2004) Biological correlates of description date in carnivores and primates. Global Ecology and Biogeography, 13, 459467.
  • Crawley, M. (2007) The R Book. John Wiley and Sons, Chichester, U.K.
  • Dolphin, K. & Quicke, D. (2001) Estimating the global species richness of an incompletely described taxon: an example using parasitoid wasps (Hymenoptera : Braconidae). Biological Journal of the Linnean Society, 73, 279286.
  • Dyer, L.A., Singer, M.S., Lill, J.T., Stireman, J.O., Gentry, G.L., Marquis, R.J., Ricklefs, R.E., Greeney, H.F., Wagner, D.L., Morais, H.C., Diniz, I.R., Kursar, T.A. & Coley, P.D. (2007) Host specificity of Lepidoptera in tropical and temperate forests. Nature, 448, 696699.
  • Erwin, T. (1982) Tropical forests: their richness in Coleoptera and other arthropod species. Coleopterists Bulletin, 36, 7475.
  • Gaston, K. (1991) Body size and probability of description – the beetle fauna of Britain. Ecological Entomology, 16, 505508.
  • Gaston, K. (1996) Species richness, measure and measurement. In Biodiversity (ed. by K.Gaston). pp. 77113, Blackwell Science, Oxford, U.K.
  • Gaston, K., Blackburn, T. & Loder, N. (1995) Which species are described first – the case of North American butterflies. Biodiversity and Conservation, 4, 119127.
  • Hillebrand, H. & Azovsky, A. (2001) Body size determines the strength of the latitudinal diversity gradient. Ecography, 24, 251256.
  • Hortal, J., Borges, P. & Gaspar, C. (2006) Evaluating the performance of species richness estimators: sensitivity to sample grain size. Journal of Animal Ecology, 75, 274287.
  • Liermann, M., Steel, A., Rosing, M. & Guttorp, P. (2004) Random denominators and the analysis of ratio data. Environmental and Ecological Statistics, 11, 5571.
  • Mace, G., Gittleman, J. & Purvis, A. (2003) Preserving the tree of life. Science, 300, 17071709.
  • Mace, G., Masundire, H., Baillie, J., Ricketts, T., Brooks, T., Hoffmann, M., Balmford, A., Purvis, A., Reyers, B., Wang, J., Revenga, C., Kennedy, E., Naeem, S., Alkemade, R., All-nutt, T., Bakarr, M., Bond, W., Chanson, J., Cox, N., Fonseca, G., Hilton-Taylor, C., Loucks, C., Rodrigues, A., Sechrest, W., Stattersfield, A., Van Rensburg, B. & Whiteman, C. (2005) Biodiversity. In Millennium Ecosystem Assessment: Current State and Trends: Findings of the Condition and Trends Working Group. Ecosystems and Human Well-being (ed. by R.M.Hassan, R.Scholes and N.Ash). pp. 5398, Island Press, Washington, DC, U.S.A.
  • Magurran, A.E. (2004) Measuring biological diversity. Blackwell Publishing, Maiden, Massachusetts, U.S.A.
  • Meier, R. & Dikow, T. (2004) Significance of specimen databases from taxonomic revisions for estimating and mapping the global species diversity of invertebrates and repatriating reliable specimen data. Conservation Biology, 18, 478488.
  • Novotny, V., Miller, S.E., Hulcr, J., Drew, R.A.I., Basset, Y., Janda, M., Setliff, G.P., Darrow, K., Stewart, A.J.A., Auga, J., Isua, B., Molem, K., Manumbor, M., Tamtiai, E., Mogia, M. & Weiblen, G.D. (2007) Low beta diversity of herbivorous insects in tropical forests. Nature, 448, 692695.
  • Petersen, F. & Meier, R. (2003) Testing species-richness estimation methods on single-sample collection data using the danish diptera. Biodiversity and Conservation, 12, 667686.
  • Petersen, F., Meier, R. & Nykjaer, M. (2003) Testing species richness estimation methods using museum label data on the danish asilidae. Biodiversity and Conservation, 12, 687701.
  • Purvis, A. & Hector, A. (2000) Getting the measure of biodiversity. Nature, 405, 212219.
  • Quicke, D.L.J. (1997) Parasitic wasps, 1st edn. Chapman & Hall, London, U.K.
  • R Development Core Team. (2008) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
  • Smith, M.A., Rodriguez, J.J., Whitfield, J.B., Deans, A.R., Janzen, D.H., Hallwachs, W. & Hebert, P.D.N. (2008) Extreme diversity of tropical parasitoid wasps exposed by iterative integration of natural history, dna barcoding, morphology, and collections. Proceedings of the National Academy of Sciences of the United States of America, 105, 1235912364.
  • Stork, N.E. (2007) Biodiversity – world of insects. Nature, 448, 657658.
  • Walther, B. & Moore, J. (2005) The concepts of bias, precision and accuracy, and their use in testing the performance of species richness estimators, with a literature review of estimator performance. Ecography, 28, 815829.
  • Wilson, E. (2003) The encyclopedia of life. Trends in Ecology & Evolution, 18, 7780.
  • Yu, D., Van Achterberg, K. & Horstmann, K. (2005) World Ichneumonoidea 2004. Taxonomy, Biology, Morphology and Distribution. Taxapad, Vancouver, Canada.
  • Zapata, F.A. & Robertson, D.R. (2007) How many species of shore fishes are there in the tropical eastern Pacific? Journal of Biogeography, 34, 3851.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
ICAD_057_sm_FigureS1.pdf48KSupporting info item
ICAD_057_sm_TableS1.pdf41KSupporting info item
ICAD_057_sm_Appendix1.txt21KSupporting info item

Please note: Neither the Editors nor Wiley Blackwell are responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.