The first half of tetrapod evolution, sampling proxies, and fossil record quality

The first half of tetrapod evolution witnessed substantial diversification of the clade and several major turnovers and mass extinctions. In the time since their origin, more than 380 Myr ago, to the beginning of the Middle Jurassic 175 Myr ago, tetrapods apparently diversified fitfully, reaching their highest level in the Middle Permian, and showing major diversity declines in the late Moscovian, Early Permian, Wordian, lower Wuchiapingian, end-Permian, lower Anisian, lower Ladinian, Late Triassic (lower Norian to upper Rhaetian), end-Triassic, and Early Jurassic (upper Sinemurian, lower Pliensbachian). Of these diversity drops, only the end-Permian and end-Triassic correspond to recognised mass extinctions, and the late Moscovian and early Norian drops to other previously identified environmental crises. The remainder could be real extinction or turnover events, or partially artefacts of biased sampling. There are strong correlations between formation counts and tetrapod palaeodiversity, suggesting a sampling component in the raw data, but the covariation is not uniform through the whole time span, being poor from Devonian to Middle Permian, and better from Late Permian to Early Jurassic. There is limited evidence for covariation between the tetrapod palaeodiversity time series and other putative sampling metrics, such as specimen completeness, numbers of publications, map areas, gap-bounded sedimentary units, rock volumes, formations, and fossil collections. Modelling by multiple correlations shows that formation count is generally the best explanatory model, either on its own, or combined with other ‘sampling’ time series. However, it is not clear that formation count is independent of the palaeodiversity time series, because rises and falls in both signals could reflect variations in original diversity or in preservation or in sampling.

► Early tetrapods underwent several extinctions and diversifications
► The diversity curve reflects variable preservation, sampling, and diversity signal.
► Diversity correlates with formation counts, but not from Devonian to Middle Permian
► Map areas, author counts, and sample quality do not covary with palaeodiversity.
► ‘Sampling proxies’ may be partly redundant with diversity data