Environmentally driven extinction and opportunistic origination explain fern diversification patterns

Combining palaeontological and neontological data offers a unique opportunity to investigate the relative roles of biotic and abiotic controls of species diversification, and the importance of origination versus extinction in driving evolutionary dynamics. Ferns comprise a major terrestrial plant radiation with an extensive evolutionary history providing a wealth of modern and fossil data for modelling environmental drivers of diversification. Here we develop a novel Bayesian model to simultaneously estimate correlations between diversification dynamics and multiple environmental trajectories. We estimate the impact of different factors on fern diversification over the past 400 million years by analysing a comprehensive dataset of fossil occurrences and complement these findings by analysing a large molecular phylogeny. We show that origination and extinction rates are governed by fundamentally different processes: originations depend on within-group diversity but are largely unaffected by environmental changes, whereas extinctions are strongly affected by external factors such as climate and geology. Our results indicate that the prime driver of fern diversity dynamics is environmentally driven extinction, with origination being an opportunistic response to diminishing ecospace occupancy.


PaleoDEMs
The paleogeographic digital elevation models 1 were built on the basis of digital topographic and bathymetric datasets of the modern world [2][3][4] . These data were combined into a global dataset with 6minute resolution. The individual grid cells were rotated to their paleopositions using the PALEOMAP Global Plate Tectonic Model 5 , and the modern digital topographic and bathymetric values were corrected and modified using the lithofacies and paleoenvironmental data. These data were constructed using the database of Paleogeographic Atlas Project 6-8 supplemented by additional lithological and paleoenvironmental records for the Permian and Jurassic 9,10 in combination with numerous regional and global paleogeographic atlases. The plate tectonic reconstructions 11 were used to model the expected changes in topography and bathymetry caused by plate tectonic events, such as sea floor spreading, continental rifting, subduction, and continental collision, as well as other isostatic events such as glacial rebound 12 .
Fossil calibration priors for molecular dating

Ferns
Hard minimum age constraint: 359 Ma. Archaeocalamites from Upper Devonian to Lower Permian 13,14 is considered as a member of the horsetail lineage 15 and used in several studies to constrain the first divergence within ferns [16][17][18] . We follow this practice to constrain the minimum age of crown group ferns.
Soft maximum age constraint: 407 Ma. Oldest known Euphyllophytina are from Lower Devonian 19 , and we used this age as the soft maximum age for the ferns.

Equisetum
Hard minimum age constraint: 359 Ma. Archaeocalamites is considered a member of the horsetail lineage 15 and we used their first appearance to constrain the minimum age of the Equisetum stem lineage.
Soft maximum age constraint: 407 Ma. We used the oldest known Euphyllophytina 19 as the soft maximum for the stem lineage.

Tmesipteris
Hard minimum age constraint: 23 Ma. The Oligocene fossil Tmesipteris tasmanica is lacking reproductive structures, but represents the distinctive vegetative characteristics of Tmesipteris 20 . We assigned the constraint for the crown group Tmesipteris because the fossil is associated with extant species having indefinite growth type 20 .
Soft maximum age constraint: 145.5 Ma. Whisk ferns have extremely poor fossil record (Collinson 1996), but are among the first diverging fern lineages 15,17,21 . In the lack of better fossil constraint we applied the beginning of Cretaceous as the soft maximum age for this node.

Botrychium
Hard minimum age constraint: 55.8 Ma. Fossil trophophores, sporophores and spores were assigned to the genus Botrychium with high confidence 22 . The fossil was considered remarkably similar with extant B. virginianum 22 , a species that was resolved together with B. strictum as the sister lineage of all other living Botrychium 21 . A detailed cladistic analysis incorporating morphological characters would be needed to verify whether B. wrightonii represents a crown group Botrychium, or a stem group. We took a conservative approach and allowed the fossil to represent the stem group.
Soft maximum age constraint: 201.6 Ma. Ophioglossales have extremely poor fossil record 22 , but are among the first diverging fern lineages 21 . The split between Botrychium and Ophioglossum have Soft maximum age constraint: 318 Ma. We applied the same soft maximum age constraint for Angiopteris as for Ptisana.

Leptosporangiate ferns
Hard minimum age constraint: 345 Ma. Tournaisian sporangia of Senftenbergia-type are placed in Tedeleaceae 28 , a group of ferns resolved as stem group leptosporangiates 15 . We assigned the constraint to the leptosporangiate stem lineage.
Soft maximum age constraint: 407 Ma. Oldest known Euphyllophytina are from Lower Devonian 19 , and we used this age as the soft maximum age for leptosporangiate stem lineage.
Soft maximum age constraint: 345 Ma. The hard minimum age of leptosporangiates was used as the soft maximum age for Osmundaceae stem lineage.

Osmunda
Hard minimum age constraint: 201.6 Ma. The Triassic fossil Osmunda claytoniites was described as remarkably similar to modern Osmunda 32 . The fossil shows medially positioned hemidimorphic leaves which are characteristic to Osmunda subgenus Claytosmunda, thus providing a minimum age for the genus 32,33 . We used this fossil to constrain the crown group age.

Stromatopteris
Hard minimum age constraint: 89.3 Ma. Morphology based cladistic analyses 40,41 resolved Boodlepteris as the sister of extant Stromatopteris. Both of these studies suffered from poor taxon sampling and failed to cover the extant diversity within Gleicheniaceae and their topology is not supported by the molecular results. However, Boodlepteris and Stromatopteris share anatomical apomorphies not observed in other members of the family, and we follow a previous study 31 in using this calibration point. We applied the constraint for the Stromatopteris moniliformis-Gleichenella pectinata node.
Soft maximum age constraint: 299 Ma. The soft maximum age of the family was applied to this node.

Lygodiaceae
Hard minimum age constraint: 168 Ma. It has been suggested that mid-Jurassic Stachypteris provides a minimum age for the divergence between Lygodium and Anemia-Schizaea 42 . Similarities in fertile structures suggests affinity with modern Lygodium, despite the highly different frond morphology and habit 42 . We assigned the constraint to the Lygodiaceae stem lineage following a previous study 16 .
Soft maximum age constraint: 299 Ma. Fossil evidence suggests that the second filicalean radiation, into which Lygodiaceae belongs, began in Permian time 43 . Carboniferous-Permian boundary is applied here as the soft maximum age for the stem lineage.

Schizaea
Hard minimum age constraint: 112 Ma. Fossil Schizaeopsis is very similar to living Schizaea in frond morphology, but differs in its spore morphology 42,44,45 . Schizaeopsis has been considered as the sister lineage of Schizaea 42 . The oldest known fossils are from the Lower Cretaceous 45 . We assigned the constraint to the Schizaea stem lineage.
Soft maximum age constraint: 251 Ma. Spores with a possible affinity with Schizaea first appear in the Triassic 42 and consequently we applied the beginning of Triassic as the soft maximum age for the stem lineage.

Anemia
Hard minimum age constraint: 136 Ma. The anatomically preserved fossils were assigned to the genus Anemia, and compared most closely to subgenus Anemiorrhiza 46 . Anemiorrhiza is the sister group to a clade comprising Mohria and the rest of the Anemia 42 . Because the fossil differed from Mohria and presented characters suggesting affinity with Anemiorrhiza 46 , we used it to constrain the basal split of Anemiorrhiza from the remaining clade. Further evidence to support this calibration point is provided by Cretaceous Pelletixia, Ruffordia and Anemia fossils which have been variously resolved within the modern Anemia subgroups in phylogenetic analyses 42,47,48 . This node was constrained as Valanginian also in a previous study 31 . We assigned the constraint to the Anemia stem lineage.
Soft maximum age constraint: 251 Ma. Anemia and Schizaea are sister groups 42 and we apply the same soft maximum age for both of them.

Marsileaceae
Hard minimum age constraint: 140 Ma. Late Jurassic-Early Cretaceous Regnellites was resolved as a member of Marsileaceae in a cladistic analysis of morphological data 49 . This calibration has also been used before 31 . We assigned the constraint to the Marsileaceae stem lineage.
Soft maximum age constraint: 201.6 Ma. Oldest fossils possible related to Marsileaceae are Late Jurassic microspores 50,51 . We therefore applied Triassic-Jurassic boundary as the soft maximum age for the stem lineage.

Marsilea
Hard minimum age constraint: 99.6 Ma. Marsileaceaephyllum mahisensis was nested within Marsilea in cladistic analysis of morphological data 51 . Taxon sampling was poor in that analysis, but the relationships within Marsilea were in accordance with the molecular hypothesis 52 . We assigned the constraint to the Marsilea stem lineage.
Soft maximum age constraint: 145.5 Ma. Based on fossil record it has been estimated that modern Marsileaceae genera started to differentiate at mid-Cretaceous 50 . We used the beginning of Cretaceous as the soft maximum age for the stem lineage.

Regnellidium
Hard minimum age constraint: 83.5 Ma. Fossils from the Late Cretaceous were assigned to Regnellidium based on preserved sporocarp, megaspore and microspore characters 50 . This calibration point has been used also in a previous study 31 . We applied this constraint to the Regnellidium diphyllum-Pilularia globulifera node.
Soft maximum age constraint: 145.5 Ma. Based of fossil record it has been estimated that modern Marsileaceae genera started to differentiate at mid-Cretaceous 50 and we assigned the beginning of Cretaceous as the soft maximum age for this node.
We applied this constraint for the Azolla stem lineage.
Soft maximum age constraint: 201.6 Ma. The soft maximum age of Marsileaceae (sister family of Salviniaceae, in which Azolla belongs) was applied as the maximum age for Azolla stem lineage.

Azolla section Azolla
Hard minimum age constraint: 13.8 Ma. The minimum age of this section was constrained following a previous study 54 . The age constraint was applied to the section including the stem.
Soft maximum age constraint: 33.9 Ma. All modern Azolla share a collared megaspore, a character that is observed in the fossil record first time in the Oligocene 54 . The age constraint was applied to the stem lineage.

Loxsomataceae
Hard minimum age constraint: 112 Ma. Relationship of the Aptian Loxsomopteris anasilla rhizome with Loxsomataceae was suggested by similarity of the hairs covering the rhizome 55 . We assigned the constraint to the Loxsomataceae stem lineage.
Soft maximum age constraint: 251 Ma. Triassic fossils have often been associated with tree ferns 56 , although it is possible that they represent extinct stem lineages of tree ferns or tree ferns and Polypodiales 17 . Molecular clock suggested that the divergence between Polypodiales and tree ferns occured in Late Triassic 16 . We apply the beginning of Triassic as the soft maximum age for the tree ferns.

Lophosoria
Hard minimum age constraint: 112 Ma. Lophosoria cupulata from the Aptian of Antarctica is placed in the extant genus 57 and was used here to constrain the split between Lophosoria and Dicksonia.
Soft maximum age constraint: 251 Ma. We apply the beginning of Triassic as the soft maximum age for the tree ferns.

Cyatheaceae
Hard minimum age constraint: 125 Ma. Permineralized fertiled structures have diagnostic characters of modern Cyathea, including abaxially borne sori with globose indusia, annulate sporangia with multiseriate stalks and vertical annuli that are not interrupted by the stalk, and 64 trilete spores 58 . The Barremian Cyathea cranhamii represents the oldest unequivocal member of modern Cyatheaceae 58 and is used here as the minimum age constraint, although the permineralized stem genus Cyathocaulis appears to be slightly older (Hauterivian) 59 and was resolved as a member of Cyatheaceae in a cladistic analysis 60 . We assigned the constraint to the Cyatheaceae stem lineage.
Soft maximum age constraint: 251 Ma. We apply the beginning of Triassic as the soft maximum age for the tree ferns.

Lindsaeaceae
Hard minimum age constraint: 99.6 Ma. Cretaceous fossil roots display sclerenchymatous outer cortex, parenchymatous inner cortex, and an innermost layer of six large cells 61 . This root anatomy is only known in the family Lindsaeaceae 61 . We assigned the constraint to the Lindsaeaceae stem lineage.
Soft maximum age constraint: 201.6 Ma. Some Jurassic fossils have been associated with Lindsaeaceae, although their affinity is highly uncertain 61,62 . Oldest fossils possibly associated with Polypodiales are also Jurassic 17,56,62 and we used the beginning of the Jurassic as the soft maximum age for the stem lineage.

Pteridaceae
Hard minimum age constraint: 93.5 Ma. The shape of the pinnae and the presence of pseudoindusia in a Cenomanian fossil 63 closely resemble pteridoid ferns, and this fossil has been used to constrain the origin of pteridoid lineage 16,17,31 . We assigned the constraint to the Pteridaceae stem lineage.
Soft maximum age constraint: 201.6 Ma. Oldest known putative members of Polypodiales are Jurassic 17,56,62 and therefore, we used the beginning of Jurassic as the soft maximum age constrain for the stem lineage.

Acrostichum-Ceratopteris -clade
Hard minimum age constraint: 65.5 Ma. Macrofossils from Maastrichtian were assigned to Acrostichum 64 , but more conservative views consider this fossil as putative stem group member 17,31 .
This more conservative approach is also followed here.
Soft maximum age constraint: 201.6 Ma. The soft maximum age of Pteridaceae was applied to the stem lineage.

Ceratopteris
Hard minimum age constraint: 37.2 Ma. Diagnostic Ceratopteris spores are known from the Bartonian 47 . This calibration has also been used in previous studies 17,31 [69][70][71] . Molecular studies have also suggested Cretaceous origin for the main eupolypod groups 16,17,31 . We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups.

Drynaria mollis
Hard minimum age constraint: 2.6 Ma. The morphology-based and total-evidence analyses resolved Pliocene D. callispora as sister to the extant species D. mollis, and these two formed sister lineage to D. sinica 68 . Although the morphological results somewhat differed from the molecular results (e.g. Drynaria monophyletic vs. paraphyletic), the nodes critical to place D. callispora in the phylogeny remained the same and we constrained the D. mollis-D. sinica split to have occured before 2.6 Ma.
Soft maximum age constraint: 65.5 Ma. We applied the beginning of Paleogene as the soft maximum age for this phylogenetically recent split.

Cyclosoroids
Hard minimum age constraint: 33.9 Ma. Eocene fossils can be assigned as members of cyclosoroids 72 , which form a clade within Thelypteridaceae 73 . This calibration has also been used in previous studies 17,31 . We assigned the constraint to the cyclosoroid stem lineage.
Soft maximum age constraint: 145.5 Ma. We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups (see above).

Onoclea sensibilis
Hard minimum age constraint: 55.8 Ma. A large number of relatively complete Paleocene fossils conform to the living species in all recognizable features 74 . This calibration has been accepted in previous studies 17,31 , although in these more sparsely sampled analyses the calibration was assigned at deeper phylogenetic level. We assigned the constraint to stem lineage of the clade of two Onoclea sensibilis varieties.
Soft maximum age constraint: 145.5 Ma. We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups (see above).

Stenochlaena
Hard minimum age constraint: 23 Ma. Distinctive spores of Stenochlaenidites are essentially identical with the spores of modern Stenochlaena and first appear in the late Oligocene 72 . The constraint was assigned to the Stenochlaena stem lineage.
Soft maximum age constraint: 145.5 Ma. We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups (see above).

Woodwardia
Hard minimum age constraint: 55.8 Ma. Several fossil Woodwardia species are known from the Paleocene onwards 72 , including W. bureiensis from Wuyun Formation of China 75 . This calibration has been used previously as well 31 . The constraint was assigned to the Woodwardia stem lineage.
Soft maximum age constraint: 145.5 Ma. We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups (see above).

Woodwardia virginica
Hard minimum age constraint: 15.4 Ma. Combination of vegetative pinnules, rhizome and stipe anatomy, and fertile pinnules with sori and sporangia from the Middle Miocene were described to be similar with the extant Woodwardia virginica 76 . The fossil bed was dated 15.6 ± 0.2 Ma by Ar/Ar dating technique 76,77 . We used the lower estimate as the hard minimum age to constrain the crown group Woodwardia.
Soft maximum age constraint: 145.5 Ma. We apply the soft maximum age of the genus as the soft maximum age for this early diverging species.

Athyriaceae
Hard minimum age constraint: 37.2 Ma. The Middle Eocene Makopteris princetonensis is the oldest fossil assigned to the family 78 and has been used to calibrate Athyriaceae in other studies as well 17,31 .
The constraint was assigned to the Athyriaceae stem lineage.
Soft maximum age constraint: 145.5 Ma. We apply the beginning of Cretaceous as the soft maximum age for the eupolypod groups (see above).      Shrinkage weights greater than 0.5 (highlighted in bold) indicate significant evidence for correlation (positive or negative depending on the respective G or H value). The curves were numbered as follows: 0 fern diversity; 1 selenium concentration in marine sediments; 2 extent of wet tropical biome; 3 extent of cool temperate biome; 4 magmatic activity; 5 atmospheric CO2; 6 gymnosperm diversity; 7 angiosperm diversity; 8 extent of warm temperate biome; 9 extent of polar biome; 10 continental fragmentation; 11 extent of arid biome; 12 atmospheric O2; 13 eustatic sea level; 14 diversity of lycophytes etc. (free sporing vascular plants excluding ferns); 15 global mean temperature; 16 mountain area.