How Many Species Are There on Earth and in the Ocean? — TEI-tagged view

Camilo MorapersName, Derek P. TittensorpersName, Sina AdlpersName, Alastair G. B. SimpsonpersName, Boris WormpersName (2011date). “How Many Species Are There on Earth and in the Ocean?.” PLoS Biologytitle·j.

Object of study — global and marine speciesterm·dom richness across all domains of life

Research question — Can predictable structure in higher taxonomic ranksterm·dom provide a validated quantitative estimate of the total number of species on Earth and in the ocean?

Contribution — a higher-taxon extrapolation method validated on well-inventoried groups and applied to global and marine eukaryotic speciesterm·dom richness

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

How Many Species Are There on Earth and in the Ocean?

Abstract The diversity of lifeterm·dom is one of the most striking aspects of our planet; hence knowing how many species inhabit Earth is among the most fundamental questions in science. Yet the answer to this question remains enigmatic, as efforts to sample the world's biodiversity to date have been limited and thus have precluded direct quantification of global species richnessterm·dom, and because indirect estimates rely on assumptions that have proven highly controversial. Here we show that the higher taxonomic classificationterm·dom of species (i.e., the assignment of species to phylumterm·dom, classterm·dom, orderterm·dom, familyterm·dom, and genusterm·dom) follows a consistent and predictable pattern from which the total number of species in a taxonomic group can be estimated. This approach was validated against well-known taxa, and when applied to all domains of life, it predicts ∼8.7 million (±1.3 million SE) eukaryotic speciesterm·dom globally, of which ∼2.2 millionterm (±0.18 million SE) are marine. In spite of 250 years of taxonomic classification and over 1.2 million species already catalogued in a central database, our results suggest that some 86% of existing species on Earthterm and 91% of species in the oceanterm still await description. Renewed interest in further exploration and taxonomy is required if this significant gap in our knowledge of life on Earth is to be closed.

Introduction Robert May [1] recently noted that if aliens visited our planet, one of their first questions would be, “How many distinct life forms—species—does your planet have?” He also pointed out that we would be “embarrassed” by the uncertainty in our answer. This narrative illustrates the fundamental nature of knowing how many species there are on Earth, and our limited progress with this research topic thus far [1] – [4] . Unfortunately, limited sampling of the world's biodiversity to date has prevented a direct quantification of the number of species on Earth, while indirect estimates remain uncertain due to the use of controversial approaches (see detailed review of available methods, estimates, and limitations in Table 1 ). Globally, our best approximation to the total number of species is based on the opinion of taxonomic expertsterm·dom, whose estimates range between 3 and 100 million species [1] ; although these estimations likely represent the outer bounds of the total number of species, expert-opinion approachesterm·mth have been questioned due to their limited empirical basis [5] and subjectivity [5] – [6] ( Table 1 ). Other studies have used macroecological patternsterm·mth and biodiversity ratiosterm·mth in novel ways to improve estimates of the total number of species ( Table 1 ), but several of the underlying assumptions in these approaches have been the topic of sometimes heated controversy ( [3] – [17] , Table 1 ); moreover their overall predictions concern only specific groups, such as insects [9] , [18] – [19] , deep sea invertebrates [13] , large organisms [6] – [7] , [10] , animals [7] , fungiterm·dom [20] , or plants [21] . With the exception of a few extensively studied taxa (e.g., birds [22] , fishes [23] ), we are still remarkably uncertain as to how many species exist, highlighting a significant gap in our basic knowledge of life on Earth. Here we present a quantitative method to estimate the global number of species in all domains of life. We report that the number of higher taxaterm·dom, which is much more completely known than the total number of species [24] , is strongly correlated to taxonomic rankterm·dom [25] and that such a pattern allows the extrapolation of the global number of species for any kingdom of life ( Figures 1 and 2 ).

Predicting the global number of species in Animalia from their higher taxonomy. (A–F) The temporal accumulation of taxa (black lines) and the frequency of the multimodel fits to all starting years selected (graded colors). The horizontal dashed lines indicate the consensus asymptotic number of taxaterm, and the horizontal grey area its consensus standard error. (G) Relationship between the consensus asymptotic number of higher taxaterm·dom and the numerical hierarchy of each taxonomic rankterm·dom. Black circles represent the consensus asymptotes, green circles the catalogued number of taxa, and the box at the species level indicates the 95% confidence interval around the predicted number of species (see Materials and Methods ).

Validating the higher taxon approachterm·mth. We compared the number of species estimated from the higher taxon approachterm·mth implemented here to the known number of species in relatively well-studied taxonomic groups as derived from published sources [37] . We also used estimations from multimodel averagingterm·mth from species accumulation curvesterm·mth for taxa with near-complete inventories. Vertical lines indicate the range of variation in the number of species from different sources. The dotted line indicates the 1∶1 ratio. Note that published species numbers ( y -axis values) are mostly derived from expert approximations for well-known groups; hence there is a possibility that those estimates are subject to biases arising from synonymsterm·dom.

Available methods for estimating the global number of species and their limitations.

Higher taxonomy data have been previously used to quantify species richness within specific areas by relating the number of species to the number of genera or families at well-sampled locations, and then using the resulting regression model to estimate the number of species at other locations for which the number of families or genera are better known than species richness (reviewed by Gaston & Williams [24] ). This method, however, relies on extrapolation of patterns from relatively small areas to estimate the number of species in other locations (i.e., alpha diversityterm·dom). Matching the spatial scale of this method to quantify the Earth's total number of species would require knowing the richness of replicated planets; not an option as far as we know, although May's aliens may disagree. Here we analyze higher taxonomic data using a different approach by assessing patterns across all taxonomic levels of major taxonomic groups. The existence of predictable patterns in the higher taxonomic classificationterm·dom of species allows prediction of the total number of species within taxonomic groups and may help to better constrain our estimates of global species richnessterm·dom.

Results We compiled the full taxonomic classifications of ∼1.2 million currently valid species from several publicly accessible sources (see Materials and Methods ). Among eukaryote “kingdoms,” assessment of the temporal accumulation curves of higher taxaterm·mth (i.e., the cumulative number of species, genera, ordersterm·dom, classesterm·dom, and phyla described over time) indicated that higher taxonomic ranksterm·dom are much more completely described than lower levelsterm·dom, as shown by strongly asymptoting trajectories over time ( [24] , Figure 1A–1F , Figure S1 ). However, this is not the case for prokaryotesterm·dom, where there is little indication of reaching an asymptote at any taxonomic level ( Figure S1 ). For most eukaryotes, in contrast, the rate of discovery of new taxa has slowed along the taxonomic hierarchy, with clear signs of asymptotes for phyla (or “divisions” in botanical nomenclature) on one hand and a steady increase in the number of species on the other ( Figure 1A–1F , Figure S1 ). This prevents direct extrapolation of the number of species from species-accumulation curves [22] , [23] and highlights our current uncertainty regarding estimates of total species richness ( Figure 1F ). However, the increasing completeness of higher taxonomic ranksterm·dom could facilitate the estimation of the total number of species, if the former predicts the latter. We evaluated this hypothesis for all kingdoms of life on Earth.

First, we accounted for undiscovered higher taxaterm·dom by fitting, for each taxonomic level from phylumterm·dom to genusterm·dom, asymptotic regression modelsterm·mth to the temporal accumulation curves of higher taxaterm·mth ( Figure 1A–1E ) and using a formal multimodel averagingterm·mth framework based on Akaike's Information Criterionterm·mth [23] to predict the asymptotic number of taxa of each taxonomic level (dotted horizontal line in Figure 1A –11E; see Materials and Methods for details). Secondly, the predicted number of taxa at each taxonomic rankterm·dom down to genusterm·dom was regressed against the numerical rankterm, and the fitted models used to predict the number of species ( Figure 1G , Materials and Methods ). We applied this approach to 18 taxonomic groups for which the total numbers of species are thought to be relatively well known. We found that this approach yields predictions of species numbers that are consistent with inventory totalsterm·mth for these groups ( Figure 2 ). When applied to all eukaryote kingdoms, our approach predicted ∼7.77 million species of animals, ∼298,000 species of plants, ∼611,000 species of fungiterm·dom, ∼36,400 species of protozoa, and ∼27,500 species of chromists; in total the approach predicted that ∼8.74 million species of eukaryotes exist on Earth ( Table 2 ). Restricting this approach to marine taxa resulted in a prediction of 2.21 million eukaryote species in the world's oceans ( Table 2 ). We also applied the approach to prokaryotesterm·dom; unfortunately, the steady pace of description of taxa at all taxonomic ranksterm·dom precluded the calculation of asymptotes for higher taxaterm·dom ( Figure S1 ). Thus, we used raw numbers of higher taxaterm·dom (rather than asymptotic estimates) for prokaryotesterm·dom, and as such our estimates represent only lower bounds on the diversity in this group. Our approach predicted a lower bound of ∼10,100 species of prokaryotesterm·dom, of which ∼1,320 are marine. It is important to note that for prokaryotesterm·dom, the species conceptterm·dom tolerates a much higher degree of genetic dissimilarity than in most eukaryotes [26] , [27] ; additionally, due to horizontal gene transfers among phylogenetic clades, species take longer to isolate in prokaryotesterm·dom than in eukaryotes, and thus the former species are much older than the latter [26] , [27] ; as a result the number of described species of prokaryotesterm·dom is small (only ∼10,000 species are currently accepted).

Currently catalogued and predicted total number of species on Earth and in the ocean.

Assessment of Possible Limitations We recognize a number of factors that can influence the interpretation and robustness of the estimates derived from the method described here. These are analyzed below.

Species definitions An important caveat to the interpretation of our results concerns the definition of species. Different taxonomic communitiesterm·dom (e.g., zoologists, botanists, and bacteriologists) use different levels of differentiation to define a species. This implies that the numbers of species for taxa classified according to different conventions are not directly comparable. For example, that prokaryotesterm·dom add only 0.1% to the total number of known species is not so much a statement about the diversity of prokaryotesterm·dom as it is a statement about what a species means in this group. Thus, although estimates of the number of species are internally consistent for kingdoms classified under the same conventions, our aggregated predictions for eukaryotes and prokaryotesterm·dom should be interpreted with that caution in mind.

Changes in higher taxonomy Increases or decreases in the number of higher taxaterm·dom will affect the raw data used in our method and thus its estimates of the total number of species. The number of higher taxaterm·dom can change for several reasons including new discoveries, the lumping or splitting of taxa due to improved phylogenies and switching from phenetic to phylogenetic classifications, and the detection of synonymsterm·dom. A survey of 2,938 taxonomists with expertise across all major domains of life (response rate 19%, see Materials and Methods ) revealed that synonymsterm·dom are a major problem at the species level, but much less so at higher taxonomic levels. The percentage of taxa names currently believed to be synonymsterm·dom ranged from 17.9 (±28.7 SD) for species, to 7.38 (±15.8 SD) for genera, to 5.5 (±34.0 SD) for families, to 3.72 (±45.2 SD) for ordersterm·dom, to 1.15 (±8.37 SD) for classesterm·dom, to 0.99 (±7.74 SD) for phyla. These results suggest that by not using the species-level data, our higher-taxon approach is less sensitive to the problem of synonymsterm·dom. Nevertheless, to assess the extent to which any changes in higher taxonomy will influence our current estimates, we carried out a sensitivity analysisterm·mth in which the number of species was calculated in response to variations in the number of higher taxaterm·dom ( Figure 3A–3E , Figure S2 ). This analysis indicates that our current estimates are remarkably robust to changes in higher taxonomy.

Assessment of factors affecting the higher taxon approachterm·mth. (A–E) To test the effects of changes in higher taxonomy, we performed a sensitivity analysisterm·mth in which the number of species was calculated after altering the number of higher taxaterm·dom. We used Animalia as a test case. For each taxonomic level, we added or removed a random proportion of taxa from 10% to 100% of the current number of taxa and recalculated the number of species using our method. The test was repeated 1,000 times and the average and 95% confidence limits of the simulations are shown as points and dark areas, respectively. Light gray lines and boxes indicate the currently estimated number of species and its 95% prediction interval, respectively. Our current estimation of the number of species appear robust to changes in higher taxonomy as in most cases changes in higher taxonomy led to estimations that remained within the current estimated number of species. The results for changes in all possible combinations of taxonomic levels are shown in Figure S2 . (F–J) The yearly ratio of new higher taxaterm·dom in Animalia (black points and red line) and the yearly number of new species (grey line); this reflects the fraction of newly described species that also represent new higher taxaterm·dom. The contrasting patterns in the description of new species and new higher taxaterm·dom suggest that taxonomic effortterm·dom is probably not driving observed flattening of accumulation curves in higher taxonomic levels as there is at least sufficient effort to maintain a constant description of new species. (K–O) Sensitivity analysisterm·mth on the completeness of taxonomic inventories. To assess the extent to which incomplete inventoriesterm·dom affect the predicted consensus asymptotic values obtained from the temporal accumulation of taxa, we performed a sensitivity analysisterm·mth in which the consensus asymptotic number of taxaterm was calculated from curves at different levels of completeness. We used the accumulation curves at the genusterm·dom level for major groups of vertebrates, given the relative completeness of these data (i.e., reaching an asymptote). Vertical lines indicate the consensus standard error. (P–T) Frequency distribution of the number of subordinate taxaterm·dom at different taxonomic levels. For display purposes we present only the data for Animalia; lines and test statistics are from a regression model fitted with a power function.

Changes in taxonomic effortterm·dom Taxonomic effortterm·dom can be a strong determinant of species discovery ratesterm·dom [21] . Hence the estimated asymptotes from the temporal accumulation curves of higher taxaterm·mth (dotted horizontal line in Figure 1A–1E ) might be driven by a decline in taxonomic effortterm·dom. We presume, however, that this is not a major factor: while the discovery rate of higher taxaterm·dom is declining (black dots and red lines in Figure 3F–3J ), the rate of description of new species remains relatively constant (grey lines in Figure 3F–3J ). This suggests that the asymptotic trends among higher taxonomic levels do not result from a lack of taxonomic effortterm·dom as there has been at least sufficient effort to describe new species at a constant rate. Secondly, although a majority (79.4%) of experts that we polled in our taxonomic survey felt that the number of taxonomic expertsterm·dom is decreasing, it was pointed out that other factors are counteracting this trend. These included, among others, more amateur taxonomists and phylogeneticists, new sampling methods and molecular identification tools, increased international collaboration, better access to information, and access to new areas of exploration. Taken together these factors have resulted in a constant rate of description of new species, as evident in our Figure 1 , Figure 3F–3J , and Figure S1 and suggest that the observed flattening of the discovery curves of higher taxaterm·dom is unlikely to be driven by a lack of taxonomic effortterm·dom.

Completeness of taxonomic inventories To account for yet-to-be-discovered higher taxaterm·dom, our approach fitted asymptotic regression modelsterm·mth to the temporal accumulation curve of higher taxaterm·dom. A critical question is how the completeness of such curves will affect the asymptotic prediction. To address this, we performed a sensitivity analysisterm·mth in which the asymptotic number of taxa was calculated for accumulation curves with different levels of completeness. The results of this test indicated that the asymptotic regression modelsterm·mth used here would underestimate the number of predicted taxa when very incomplete inventoriesterm·dom are used ( Figure 3K–3O ). This underestimation in the number of higher taxaterm·dom would lower our prediction of the number of species through our higher taxon approachterm·mth, which suggests that our species estimates are conservativeterm·dom, particularly for poorly sampled taxa. We reason that underestimation due to this effect is severe for prokaryotesterm·dom due to the ongoing discovery of higher taxaterm·dom ( Figure S1 ) but is likely to be modest in most eukaryote groups because the rate of discovery of higher taxaterm·dom is rapidly declining ( Figure 1A – 3E , Figure S1 , Figure 3F–3J ).

Since higher taxonomic levels are described more completely ( Figure 1A–1E ), the resulting error from incomplete inventoriesterm·dom should decrease while rising in the taxonomic hierarchy. Recalculating the number of species while omitting all data from genera yielded new estimates that were mostly within the intervals of our original estimates ( Figure S3 ). However, Chromistaterm·dom (on Earth and in the ocean) and Fungiterm·dom (in the ocean) were exceptions, having inflated predictions without the genera data ( Figure S3 ). This inflation in the predicted number of species without genera data highlights the high incompleteness of at least the genera data in those three cases. In fact, AdlpersName et al.'s [28] survey of expert opinions reported that the number of described species of chromists could be in the orderterm·dom of 140,000, which is nearly 10 times the number of species currently catalogued in the databases used here ( Table 1 ). These results suggest that our estimates for Chromistaterm·dom and Fungiterm·dom (in the ocean) need to be considered with caution due to the incomplete nature of their data.

Subjectivity in the Linnaean system of classificationterm·dom Different ideas about the correct classification of species into a taxonomic hierarchy may distort the shape of the relationships we describe here. However, an assessment of the taxonomic hierarchy shows a consistent pattern; we found that at any taxonomic rankterm·dom, the diversity of subordinate taxaterm·dom is concentrated within a few groups with a long tail of low-diversity groups ( Figure 3P–3T ). Although we cannot refute the possibility of arbitrary decisions in the classification of some taxa, the consistent patterns in Figure 3P–3T imply that these decisions do not obscure the robust underlying relationship between taxonomic levels. The mechanism for the exponential relationships between nested taxonomic levels is uncertain, but in the case of taxa classified phylogenetically, it may reflect patterns of diversification likely characterized by radiationsterm·dom within a few clades and little cladogenesisterm·dom in most others [29] . We would like to caution that the database we used here for protistan eukaryotes (mostly in Protozoa and Chromistaterm·dom in this work) combines elements of various classification schemes from different ages—in fact the very division of these organisms into “Protozoa” and “Chromistaterm·dom” kingdoms is non-phylogenetic and not widely followed among protistologists [28] . It would be valuable to revisit the species estimates for protistan eukaryotes once their global catalogue can be organized into a valid and stable higher taxonomy (and their catalogue of described species is more complete—see above).

Discussion Knowing the total number of species has been a question of great interest motivated in part by our collective curiosity about the diversity of lifeterm·dom on Earth and in part by the need to provide a reference point for current and future losses of biodiversity. Unfortunately, incomplete sampling of the world's biodiversity combined with a lack of robust extrapolation approaches has yielded highly uncertain and controversial estimates of how many species there are on Earth. In this paper, we describe a new approach whose validation against existing inventories and explicit statistical nature adds greater robustness to the estimation of the number of species of given taxa. In general, the approach was reasonably robust to various caveats, and we hope that future improvements in data quality will further diminish problems with synonymsterm·dom and incompleteness of data, and lead to even better (and likely higher) estimates of global species richnessterm·dom.

Our current estimate of ∼8.7 million speciesterm narrows the range of 3 to 100 million species suggested by taxonomic expertsterm·dom [1] and it suggests that after 250 years of taxonomic classification only a small fraction of species on Earth (∼14%) and in the ocean (∼9%) have been indexed in a central database ( Table 2 ). Closing this knowledge gap may still take a lot longer. Considering current rates of description of eukaryote species in the last 20 years (i.e., 6,200 species per year; ±811 SD; Figure 3F–3J ), the average number of new species described per taxonomist's career (i.e., 24.8 species, [30] ) and the estimated average cost to describe animal species (i.e., US$48,500 per species [30] ) and assuming that these values remain constant and are general among taxonomic groups, describing Earth's remaining species may take as long as 1,200 years and would require 303,000 taxonomists at an approximated cost of US$364 billion. With extinction rates now exceeding natural background rates by a factor of 100 to 1,000 [31] , our results also suggest that this slow advance in the description of species will lead to species becoming extinct before we know they even existed. High rates of biodiversity loss provide an urgent incentive to increase our knowledge of Earth's remaining species.

Previous studies have indicated that current catalogues of species are biased towards conspicuous species with large geographical ranges, body sizes, and abundances [4] , [32] . This suggests that the bulk of species that remain to be discovered are likely to be small-ranged and perhaps concentrated in hotspots and less explored areas such as the deep sea and soil; although their small body-size and cryptic nature suggest that many could be found literally in our own “backyards” (after Hawksworth and Rossman [33] ). Though remarkable efforts and progress have been made, a further closing of this knowledge gap will require a renewed interest in exploration and taxonomy by both researchers and funding agencies, and a continuing effort to catalogue existing biodiversity data in publicly available databases.

Materials and Methods Databases Calculations of the number of species on Earth were based on the classification of currently valid species from the Catalogue of Lifeterm ( www.sp2000.org , [34] ) and the estimations for species in the ocean were based on The World's Register of Marine Speciesterm ( www.marinespecies.org , [35] ). The latter database is largely contained within the former. These databases were screened for inconsistencies in the higher taxonomy including homonyms and the classification of taxa into multiple clades (e.g., ensuring that all diatom taxa were assigned to “Chromistaterm·dom” and not to “plants”). The Earth's prokaryotesterm·dom were analyzed independently using the most recent classification available in the List of Prokaryotic Names with Standing in Nomenclatureterm database ( http://www.bacterio.cict.fr ). Additional information on the year of description of taxa was obtained from the Global Names Indexterm database ( http://www.globalnames.org ). We only used data to 2006 to prevent artificial flattening of accumulation curves due to recent discoveries and descriptions not yet being entered into databases.

Statistical Analysis To account for higher taxaterm·dom yet to be discovered, we used the following approach. First, for each taxonomic rankterm·dom from phylumterm·dom to genusterm·dom, we fitted six asymptotic parametric regression models (i.e., negative exponentialterm·mth, asymptotic, Michaelis-Mententerm·mth, rational, Chapman-Richardsterm·mth, and modified Weibullterm·mth [23] ) to the temporal accumulation curve of higher taxaterm·dom ( Figure 1A–1E ) and used multimodel averagingterm·mth based on the small-sample size corrected version of Akaike's Information Criteria (AIC c ) to predict the asymptotic number of taxa (dotted horizontal line in Figure 1A–1E ) [23] . Ideally data should be modeled using only the decelerating part of the accumulation curve [22] – [23] , however, frequently there was no obvious breakpoint at which accumulation curves switched from an increasing to a decelerating rate of discovery ( Figure 1A–1E ). Therefore, we fitted models to data starting at all possible years from 1758 onwards (data before 1758 were added as an intercept to prevent a spike due to Linnaeus) and selected the model predictions if at least 10 years of data were available and if five of the six asymptotic models converged to the subset data. Then, the estimated multimodel asymptotes and standard errors for each selected year were used to estimate a consensus asymptote and its standard error. In this approach, the multimodel asymptotes for all cut-off years selected and their standard errors are weighted proportionally to their standard error, thus ensuring that the uncertainty both within and among predictions were incorporated [36] .

To estimate the number of species in a taxonomic group from its higher taxonomy, we used Least Squares Regression modelsterm·mth to relate the consensus asymptotic number of higher taxaterm·dom against their numerical rankterm, and then used the resulting regression model to extrapolate to the species level ( Figure 1G ). Since data are not strictly independent across hierarchically organized taxa, we also used models based on Generalized Least Squaresterm·mth assuming autocorrelated regression errors. Both types of models were run with and without the inverse of the consensus estimate variances as weights to account for differences in certainty in the asymptotic number of higher taxaterm·dom. We evaluated the fit of exponential, power, and hyperexponential functionsterm·mth to the data and obtained a prediction of the number of species by multimodel averagingterm·mth based on AIC c of the best type of function. The hyperexponential functionterm·mth was chosen for kingdoms whereas the exponential functionterm·mth for the smaller groups was used in the validation analysis (see comparison of fits in Figure S4 ).

Survey of Taxonomists We contacted 4,771 taxonomy experts with electronic mail addresses as listed in the World Taxonomist Databaseterm ( www.eti.uva.nl/tools/wtd.php ); 1,833 were faulty e-mails, hence about 2,938 experts received our request, of which 548 responded to our survey (response rate of 18.7%). Respondents were asked to identify their taxon of expertise, and to comment on what percentage of currently valid namesterm·dom could be synonymsterm·dom at taxonomic levels from species to kingdom. We also polled taxonomists about whether the taxonomic effortterm·dom (measured as numbers of professional taxonomists) in their area of expertise in recent times was increasing, decreasing, or stable.

Supporting Information Completeness of the higher taxonomy of kingdoms of life on Earth.

(DOC) Click here for additional data file.

Sensitivity analysisterm·mth due to changes in higher taxonomy.

(DOC) Click here for additional data file.

Assessing the effects of data incompleteness.

(DOC) Click here for additional data file.

Comparison of the fits of the hyperexponential, exponential, and power functions to the relationship between the number of higher taxaterm·dom and their numerical rankterm.

(DOC) Click here for additional data file.