what the deep sea tells us about sampling biases in the fossil record
TRANSCRIPT
Graeme T. LloydDepartment of Palaeontology, Natural History Museum, London, UK
What the deep sea tells us about sampling biasesin the fossil record
Collaborators
Andrew Smith
Jeremy Young
Paul Pearson
Talk Outline
• Introduction• Deep sea record of Coccolithophores and planktic forams
– Deep sea rock and fossil records– Correlations and modelling– Sampling-corrected richness
• Deep sea vs. land-based record of Coccolithophores– Deep sea vs. land rock and fossil records– Correlations and modelling– Sampling-corrected richness: common signal?
• Deep sea coccolithophore species-per-genus patterns– An unusual result!– Potential explanation(s)– Separating signals
• Conclusion
The fossil record is our only empirical record of the history of life
Generic diversity
N M
aps
Land-based rock and fossil records show strong correlation…
…but what about the deep sea?
•Most microfossil groups are highly cosmopolitan…
•…and massively abundant (1000s specimens per gram)
•Many remarkably continuous sections (>10 million years)
•Phylogenies often incorporate ancestors
•Well studied (DSDP/ODP/IODP)
•The best fossil record we have?
Comparing coccolithophore and planktic foraminifera deep sea rock and fossil records
•Questions:
•How does the deep sea rock record change over time?
•How does the deep sea fossil record change over time?
•Are the deep sea rock and fossil records correlated?
•How do the two major calcareous groups compare?
Geotectonic history
•35,416 species occurrences
•16,197 samples
•205 sites
•4,329 names
•19,349 species occurrences
•3,850 samples
•135 sites
•2,462 names
The database
Planktic foramsCoccoliths
Rock record
Planktic foramsCoccoliths
Species record
Planktic foramsCoccoliths
Generic record
Planktic foramsCoccoliths
Species correlation
Planktic foramsCoccoliths
Generic correlation
Planktic foramsCoccoliths
Species detrended
Planktic foramsCoccoliths
Genera detrended
Planktic foramsCoccoliths
Subsampling
Planktic foramsCoccoliths
Modelled versus observed diversity
Planktic foramsCoccoliths
Model-corrected diversity
Planktic foramsCoccoliths
Summary
•How does the deep sea rock record change over time?
•Exponential rise (opening ocean basin)
Summary
•How does the deep sea rock record change over time?
•Exponential rise (opening ocean basin)
•How does the deep sea fossil record change over time?
•Coccolith species ~linear rise
•Coccolith genera ~rapid rise followed by slow fall
•Forams: double sawtooth (K-T divides)
Summary
•How does the deep sea rock record change over time?
•Exponential rise (opening ocean basin)
•How does the deep sea fossil record change over time?
•Coccolith species ~linear rise
•Coccolith genera ~rapid rise followed by slow fall
•Forams: double sawtooth (K-T divides)
•Are the deep sea rock and fossil records correlated?
•Yes, strongly
Summary
•How does the deep sea rock record change over time?
•Exponential rise (opening ocean basin)
•How does the deep sea fossil record change over time?
•Coccolith species ~linear rise
•Coccolith genera ~rapid rise followed by slow fall
•Forams: double sawtooth (K-T divides)
•Are the deep sea rock and fossil records correlated?
•Yes, strongly
•How do the two major calcareous groups compare?
•Forams seem to be less biased than coccos
Comparing sampling bias between the land and the deep sea
Testing sampling bias versus common cause
Deep sea
•Correlations between sampling and diversity are common
•Two main explanations: sampling-bias and common cause
•For coccolithophores we have two records; ideal to test
•Sampling-bias predicts diversity will track sampling
•Common cause predicts shared diversity
•What do the two rock records look like?
•What do the two fossil records look like?
•Are the rock and fossil records correlated?
•Is there evidence for a common palaeobiodiversity?
Land
205 sites, 16,197 samples,36,416 occurrence records
462 sections, 5,563+ samples,22,745 occurrence records
Deep sea Land
The database
Deep sea Land
Rock records
Time (Ma)
Number of cores recovering rock of given age
Number of localities with published nannofossil taxonomic lists
Time (Ma)
Time (Ma) Time (Ma)
Number of species
Raw species diversity
Number of species
Deep sea Land
Species richness
Deep sea Land
Log (Nsites)
Log (species richness) Log (species richness)
Log (Nsites)
Species richness versus rock record (1): raw data
Deep sea Land
Log (species richness) Log (species richness)
Log (Nsites) Log (Nsites)
Species richness versus rock record (2): first differences
Deep sea Land
Orange = empirical patternWhite = diversity at equal subsampling
Time (Ma)
109 samples per bin
Time (Ma)
Species diversity (max) Species diversity (max)
106 samples per bin
Estimating true diversity: 1, subsampling
Time (Ma) Time (Ma)
Species richness Species richness
True richness modelled as invariant (observed richness = sampling)
Deep sea Land
Estimating true diversity: 2, modelling
Yellow = empirical patternBlue-green = model prediction assuming diversity is
invariant and shaped by rock abundance
Deep sea Land
Time (Ma)
Residuals from modelled richness
Time (Ma)
Estimating true diversity: 2, modelling
Time (Ma)
Deep sea Land
Species
Time (Ma)
Species
Time (Ma)
Estimating true diversity: 3, alpha diversity
Mean number of species recorded per site
• The recorded history of coccolithophorid diversity over last 150 Ma changes dramatically according to whether data is drawn from land-based records or deep-sea records
• Coccolithophorid diversity correlates strongly to the shape of the rock record it is recovered from
• Subsampling, modeling and estimates of mean alpha diversity all point to a third, much more uniform diversity irrespective of which record is used
Deep sea Land
Summary
Species per genus patterns
• Used since the earliest diversity curves…
• …and continue to be (e.g. Alroy et al. 2008)• Originally pragmatic (less data required)• Then argued that species are inadequate
• But, adequacy of higher taxa to represent species-level patterns is essentially untested
Higher taxa as species proxies
Taxonomic level affects pattern
• Only explicit test of species-to-higher taxon ratio
• Compared families to number of named species in Zoo. Record (Raup 1976)
• Pattern of change differs
• Families become more speciose
Flessa and Jablonski 1985
• Species are standardised (synonyms)
• Species are assigned to genera• Species are often widespread• Species are long-ranging
• Species are comparatively stable taxonomically
• Questions:
• How does the species-to-genus ratio change over time?• How does the sampling change over time?• How does the number of taxonomists change over time?• Do neither, either or both sampling and taxonomists shape
the signal?
Our database is superior
Species per genus
Number of sites (sampling)
Number of authors (taxonomists)
Long-term correlation (raw)
N sites (rho = 0.95)N authors (rho = 0.93)
Short-term correlation (sampling)
Species per genusN sites
Rho = 0.43
Short-term correlation (taxonomists)
Species per genusN authors
Rho = 0.44
Correlations
• Both number of sites and number of authors significantly correlate with species-per-genus
• Fit 3 models:– spg ~ N sites– spg ~ N taxonomists– spg ~ N sites + N taxonomists
• Which is the best explanatory model?– Akaike weights = N Sites (marginally more than a
combined model)– Variance partitioning = a combined model
• So is it sites or combined?
Subsampling (rarefaction by occurrences for sites)
Subsampling (rarefaction by occurrences for papers)
Summary
• How does the species-to-genus ratio change over time?– In a two-step ‘punk eek’ way
Summary
• How does the species-to-genus ratio change over time?– In a two-step ‘punk eek’ way
• How does the sampling change over time?
– The same
Summary
• How does the species-to-genus ratio change over time?– In a two-step ‘punk eek’ way
• How does the sampling change over time?
– The same
• How does the number of taxonomists change over time?– The same
Summary
• How does the species-to-genus ratio change over time?– In a two-step ‘punk eek’ way
• How does the sampling change over time?
– The same
• How does the number of taxonomists change over time?– The same
• Do neither, either or both sampling and taxonomists shape the signal?– Both contribute to the pattern
Summary
• How does the species-to-genus ratio change over time?– In a two-step ‘punk eek’ way
• How does the sampling change over time?
– The same
• How does the number of taxonomists change over time?– The same
• Do neither, either or both sampling and taxonomists shape the signal?– Both contribute to the pattern
Genera are not an accurate proxy for species
Conclusion: what does the deep sea tells us about sampling biases in the fossil record?
• The deep sea record shows the same correlation with sampling as land-based studies
• This argues in favour of the sampling-bias interpretation and not the common cause
• The deep sea record is more biased than the land-based• The deep sea coccolithophore record is more biased than
the deep sea planktic foram record
• Once sampling has been accounted for there is convergence on a single palaeobiodiversity estimate (at least for Coccolithophores)
• Taxonomic structure (species-per-genus) for deep sea Coccolithophores is biased by both sampling and the number of taxonomists