David Lazarus; John Barron; Johan Renaudie; Patrick Diver; Andreas Türke

“ Would extinct species of diatoms be replaced by fully functionally equivalent (e.g. for export productivity) other diatom species, by functionally less efficient taxa, or functionally very different coccolithophores or non-skeletal plankton? Our results, together with evidence that rapid diversity loss is linked to reduced productivity [48] , [49] , suggest that loss of diatom diversity in future oceans, if it occurred, would indeed affect diatom abundance behavior, diatom-carbon export and thus further modify the global ocean pump, with consequences for future ocean regulation of pCO2. ”

“ Many of these implications depend not on diatom diversity itself but diatom export productivity. We distinguish two distinct types of relationship between diversity and export productivity: naturally occurring behavior that has developed over evolutionary time; and perturbed relationships due to rapid extirpation or extinction, in which evolution has not had time to operate.
The relationship between diversity and productivity in living natural systems is controversial, and scale dependent. ”

“ Outliers are an unavoidable aspect of large data compilations, and also exist in the NSB database due to intrinsic aspects of the data such as reworking, taxonomic mis-identifications and age model errors [13] . While trimming fixed percentages of the range ends of age-composited occurrence data (‘pacman trimming’) can remove most outliers [32] this procedure can also create edge effects, particularly for the youngest time bin (s) , and thus complicate comparisons of diversity and environmental history in the late Neogene. ”

“ In addition, we explore how our new diversity history correlates to environmental parameters such as paleoceanographic change, climate, carbon cycle and marine export productivity as derived from sedimentary opal. In these comparisons we do not attempt full analyses of possible causal mechanisms, as these are generally complex and require consideration of many factors, often with detailed time-series analysis and/or modeling, which is beyond the scope of our current study. ”

“ A shift towards higher origination rates for living species, and an increase in the contribution of polar taxa can be seen in the latest Miocene (ca 7-5 Ma) immediately preceding the early Pliocene warm interval (ca 4.5-3.5 Ma) . Nearly 50% of living species originated <5 Ma; ca. 80% (and 50% of genera) in the Neogene (Figs. 9, 10) . This suggests that Neogene, particularly late Miocene-Recent, diversity-environment relationships are most important in estimating future responses to climate change. ”

“ Grasslands
Rabosky and Sorhannus [16] suggested that the relatively low post-Eocene diversity of diatoms in their reconstruction argued against both a grassland expansion-driven increase of silica to the oceans, with consequent diversification of marine diatoms; and against co-evolutionary, inverse relationships between Cenozoic diatom and coccolithophore diversification. Our results do not support their arguments, although there are other reasons to question a grassland-marine diatom hypothesis, as proposed by Falkowski et al. [6] . ”

“ The correlation of warmer climate to lower diversity also suggests that global warming could potentially place a significant fraction of diatom diversity at risk of extinction, particularly as we show that important export productivity species originated only in the last few million years in association with the development of cold polar oceans. ”

“ Our results for Cenozoic marine plankton diatom floras confirm an abiotic control of diversity, but differ from prior results in that: the correlation between diversity and climate is remarkably strong (r of ca 0.6 in detrended data; Spearman's rho >0.9 in the raw data) , and unlike prior results e.g. [24] , [84] , in diatoms colder climate, not warmer, is correlated to higher diversity. ”

“ The relationship to diversity can either be a direct influence of diversity on the relative abundance of diatoms in the plankton (the effect of more diverse evolutionary adaptations on relative nutrient capture) , or indirect via the response of both diatom abundance and diatom diversity to changing amounts of nutrient input into the oceans due to changing Cenozoic climate. ”

“ We argue that we have been able to extract a coherent diversity signal from incomplete, biased occurrence data, as demonstrated by the similarity between diversity estimates made using different methods, and using very different types of source data (raw occurrence data or range estimates from a catalog) . An essential component of our reconstruction method from occurrence data is the use not only of SQS subsampling to address changing geographic structure in the occurrence data, but also the application of an evenness correction factor based on modeling. ”

“ First, we re-analyze the Neptune data. In contrast to prior studies, we explicitly test for sample size bias, and correct subsampled diversity estimates for both changing geographic endemism and changing mean evenness. Second, we use the complementary nature of catalog-derived diversity estimates [13] to confirm the robustness of our results using a new separate catalog (‘BDC’) of diatom species ranges by J. Barron. We also particularly consider the history of living taxa as most relevant to future responses to global warming, as extinct taxa may have had different biologic responses. ”

“ We show that diatoms diversified strongly over the Cenozoic, with major increases near the Eocene/Oligocene boundary, and in the mid Miocene. Diversification occurs primarily by increasing diversity of endemic high latitude floras, and the disjunct development of high productivity regions in Cenozoic oceans is suggested to be a primary reason for diversification of the group. ”

“ Diatom diversity is strongly correlated to evolutionary characteristics of other groups of organisms such as radiolaria; and to the oxygen isotope proxy record of global climate change: diversity not only increases with intervals of cooling, but also significantly decreases during intervals of past warming such as the late Oligocene-early Miocene. Over the last 15 my diatom diversity is strongly correlated to both the global carbon isotope record and to estimated past atmospheric pCO2. ”

“ While the above presents arguments why the correlation of diatom diversity to climate state should be considered in thinking about future climate change, there are limits as well. We argue that while our results provide a useful perspective on eventual system response, a variety of issues, including temporal scaling and other potential casual factors, limit our ability to use our results to make any direct estimate of extinction risk due to anthropogenic change. ”

“ There is considerable a-priori evidence from diatom biology and biogeography to expect diatom diversity to respond to climate, and thus to expect the long-term as well as short term correlations to be at least in part causal, not chance. ”

“ These studies have generally concluded that global warming, by reducing global latitudinal wind stress, will lead to more highly stratified, oligotrophic oceans, reduced abundances of diatoms, and possibly reduce the effectiveness of the ocean carbon pump. Studies of extinction risk in marine biota by contrast have concentrated on benthos or nekton: there are essentially no studies of marine plankton extinction risk. ”

“ As biotas usually consist mostly of rare species, a sample's diversity is usually much less than the total diversity of the sampled biota; thus sampled diversity increases with sample size. If sample size varies in a data series, sampled diversities, and also range-through diversity from catalogs compiled from such samples, can be biased [17] . Standardized subsampled Neptune data is needed to correct for the strong variation of sample sizes in Neptune with time. Subsampling though yields biased results if either taxa ranked abundance distributions are not constant between samples ”

“ While diversity estimates compiled from raw data can have major data size biases, subsampling can also easily produce misleading results when underlying assumptions are violated (e.g., homogeneity of geographic structure and constancy of evenness of occurrence of species in samples) , producing in such cases artificially ‘flattened’ diversity vs time curves. ”

“ Although on short time scales (weeks to seasonal) , or in local environments diatom export productivity is associated with blooms dominated by very few species, over broader scales diversity and productivity appear to be correlated. In the modern Atlantic ocean living diatom diversity is strongly linked to diatom abundance/export of carbon (diversity:log biomass correlation r = .864, from data in [45] , Fig. S1) ; productivity is significantly correlated to diversity also in freshwater plankton [46] . ”

“ These include uneven completeness of diversity recording by different authors, non-uniform data coverage by time, geographic region and author, differing amounts of fixed-list biostratigraphic vs diversity survey data, age model errors, reworking, and others. Diversity data is drawn for example from a much broader number of deep-sea sections, with on average poorer quality age models, than the much smaller number of sites with high-quality age models typically used for high-resolution paleoceanography, and compilation of stable isotope curves. ”

“ When the average number of species in a list is short compared to the total diversity in a bin, which is true of NSB data [13] the effectiveness of list methods is reduced (and collapses as list size approaches 1 taxon to simple rarefaction) ; modeling subsampling by lists is more complex and requires information on taxon clustering in lists (e.g. samples) , which would make modeling the effect of evenness on subsampling (below) much more difficult ”

“ Errors are important if they create outliers that extend ranges of species in time, causing incorrect diversity estimates. The NSB version of Neptune (for ‘Neptune Sandbox Berlin’) used in this study is a new implementation of the database hosted at the Museum für Naturkunde in Berlin. It was forked off the Chronos version of the Neptune database in early/mid 2010. Analyses of NSB content [32] , [33] have specifically checked, and not found data errors like those reported by Lloyd et al. [31] , suggesting that these, if present in Chronos Neptune, postdate the NSB fork. ”

“ This analysis shows that our ‘untrimmed’ subsampled NSB diversity curve is not strongly affected by outliers, and if anything is a conservative estimate of the degree of dynamics of diversification over the Neogene.
Subsampled diatom diversity (sample size = 100) from the NSB database, using different levels of Pacman trimming of data from the ends of species ranges (7) to remove outliers due to possible data errors in NSB.Blue solid line - no data trimming; dashed red - 15% of the youngest and 10% of the oldest occurrences for each species removed ”

“ Due to long standing policies on leg staffing and sample access, the large majority of deep-sea drilling legs have only a single primary diatom paper reporting occurrence data. There is still a strong correlation between number of legs/bin and diversity/bin (Fig. 5c) , suggesting that using numbers of papers as a proxy for number of samples does not mask a sample size-diversity correlation.
Neptune (NSB) diatom diversity vs data density.a) Diversity (red) vs number occurrences (black) . ”

“ Data outliers, due to age model errors, reworking and mis-identifications, may individually contribute only a small amount to inflated diversity estimates but cumulatively, particularly in range through calculations, may have also increased diversity by significant amounts. ”

“ Despite obtaining a coherent diversity signal, our results may still be inaccurate if there are systematic biases in the data which would produce coherent biases in our diversity estimates. We specifically consider changes in preservation over time which might bias our results. ”

“ In order to do this, we need much more information on the proximate controls on individual species distributions, both for now extinct species, and for modern species, whose ecology also is often only poorly known. ”

“ In Bopp et al. [91] , this moderate degree of global change in pump efficiency is in part due to increased pump activity in polar regions, which partially compensates for an a broader drop in pump functioning in lower latitudes. ”

“ Results of simple rarefaction subsampling two virtual computer populations of equal diversity (100 taxa) and total sample size but with averaged Neogene or Paleogene relative abundance distributions as seen in NSB diatom occurrence data.a) for subsample sizes between 0–1,000; b) subsample sizes 0–10,000. Blue dots - Paleogene subsampled diversity; red - Neogene subsampled diversity; green squares - percentage underestimation of Neogene diversity relative to Paleogene diversity. Each point is an average of 50 subsampling trials. Age scale: [38] . Outliers and other errors in occurrence data ”

“ Larger bins are not desirable, decreasing our ability to compare more rapid (often significantly less than 1 my) changes in climate state to diversity. In smaller e.g. regional, more homogeneous microfossil diversity data sets Renaudie and Lazarus [33] explored the effect of bin size on diversity studies and find no significant effect between 0.5 and 1 my bins. Their study also points out that the incompleteness and/or biostratigraphic bias of data available in databases such as NSB also limits effective resolution of time-series change in primary biodiversity signals to at best ca .5 my. ”