Complex marine bioturbation ecosystem engineering behaviors persisted in the wake of the end-Permian mass extinction

The end-Permian mass extinction was the most severe mass extinction event of the Phanerozoic and was followed by a several million-year delay in benthic ecosystem recovery. While much work has been done to understand biotic recovery in both the body and trace fossil records of the Early Triassic, almost no focus has previously been given to analyzing patterns in ecosystem engineering complexity as a result of the extinction drivers. Bioturbation is a key ecosystem engineering behavior in marine environments, as it results in changes to resource flows and the physical environment. Thus, the trace fossil record can be used to examine the effect of the end-Permian mass extinction on bioturbating ecosystem engineers. We present a dataset compiled from previously published literature to analyze burrowing ecosystem engineering behaviors through the Permian-Triassic boundary. We report two key observations: first, that there is no loss in bioturbation ecosystem engineering behaviors after the mass extinction, and second, that these persisting behaviors include deep tier, high-impact, complex ecosystem engineering. These findings suggest that while environmental conditions may have limited deeper burrowing, complex ecosystem engineering behaviors were able to persist in the Early Triassic. Furthermore, the persistence of deep tier bioirrigated three-dimensional network burrows implies that benthic biogeochemical cycling could have been maintained at pre-extinction states in some local environments, stimulating ecosystem productivity and promoting biotic recovery in the Early Triassic.

with sharp outlines. The underlying transition layer begins a number of centimeters below the sediment-water interface. The transition layer has a lower water content, as it represents the beginning of compaction, which allows for the preservation of identifiable trace fossils. Bioturbation does not typically occur in the historical layer below the transition layer. The historical layer therefore typically represents bioturbation as it ultimately appears in the stratigraphic record. In marine environments, bioturbation can be divided into various functional groups based on how bioturbators interact with the substrate within these layers 21,22 . Each behavior has its own degree of impact on the sediment chemistry, biogeochemical cycling, and benthic ecosystems 14 . Thus, the persistence or disappearance of these high-impact ecosystem engineering burrowing behaviors across the end-Permian mass extinction is key to understanding how marine ecosystems recovered in the Early Triassic.
Previous research on the trace fossil record across the Permian-Triassic boundary has primarily focused on tiering, ichnofabrics, and burrow size. Induan trace fossils tend to be simple, shallow burrows 12,23 . As early as the early Olenekian (Smithian), trace fossil assemblages exhibit recovery to pre-extinction levels of ichnodiversity, ichnofabric indices, and burrow size 24,25 . However, tiering did not return to pre-extinction depths until the beginning of the Middle Triassic 26 . Herein, we apply two ecosystem engineering framework analyses to previously published Permian-Triassic trace fossil datasets to understand the patterns of ecosystem engineering behaviors across the mass extinction boundary and how these behaviors may have influenced ecosystem recovery in the Early Triassic.

Materials and Methods
ecosystem engineering behavior and trace fossil data collection. The objective of this research is to use the Permian-Triassic trace fossil record to compile and analyze ecosystem engineering behaviors and their impacts on the benthic environment. Trace fossil occurrences were compiled from previously published literature and included the entire Permian through the first stage of the Middle Triassic (Asselian through the Anisian, 299-237 Ma 27 ). Only trace fossils which could be confidently assigned to a tier -surficial, semi-infaunal, shallow, intermediate, or deep 28-30 -were added to the dataset. Tiering is a fundamental part of the ecosystem engineering analyses used in this research, and thus trace fossils without a precise description of penetration depth or tiering cannot be accurately analyzed in terms of ecosystem engineering behavior and the effect it may have had on the benthic environment. Each ichnogenus was counted once per tier because the presence or absence of each ecosystem engineering behavior represents the necessary data to investigate the effect of the extinction on each behavior and the roll that each behavior may have played in the recovery from the extinction. Data were also limited to shallow marine trace fossil occurrences as they generally provide the richest trace fossil record 26 . Due to data filtering based on the lack of precise tiering descriptions, Roadian and Capitanian data are absent. Ultimately, 164 unique ichnogenera were entered into the database given the parameters of the study (confident tiering assignment, one occurrence per tier, one occurrence per time period, and shallow marine). These data were analyzed using two frameworks: the ecosystem engineering occupation cube method 22 and the ecosystem engineering impact (EEI) values method 14,31 . Trace fossils and their categorical assignments for both ecosystem engineering analyses were grouped into stages based on stratigraphic descriptions in the primary literature.
ecosystem engineering occupation cubes. The ecosystem engineering occupation cube method 22 assigns 'cube spaces' to ichnogenera according to tiering, sediment interaction 32 , and sediment modification 21,22 . For tiering, literature descriptions were used to classify each trace fossil entry as surficial, semi-infaunal (0-0.5 cm), shallow (0.5-6 cm), intermediate (6-12 cm), or deep (>12 cm) [28][29][30] . Surficial, semi-infaunal, and shallow tier bioturbating organisms commonly occupy the mixed layer, while bioturbators occupying intermediate or deep tiers occupy the transition layer. Each ichnogenus was assigned one of four sediment interaction classifications: intrusion, compression, backfilling, and excavation 32 . Intrusion describes the displacement of the sediment as the animal burrows and sediment closes up behind it; compression describes the movement and compaction around the burrowing animal; backfilling describes the backward passage of sediment either around or through the burrower; and excavation describes the active loosening and transportation of sediment from one point along the burrow path to another 32 . Each ichnogenus was also assigned one of four sediment modification classifications: biodiffusion, gallery biodiffusion, conveyor, and regenerator 21 . Biodiffusion involves the movement of sediment particles over short distances; gallery biodiffusion involves the redistribution of sediment particles from one part of the sediment profile to another; conveying involves transporting sediment particles across and within tiers; and regenerating involves moving sediment up to the surface from below the sediment-water interface 21 . When possible, each ichnogenus was given sediment modification and sediment interaction assignments based on previous descriptions 22 . For ichnogenera which had not been previously described in this framework, the original literature which described the ichnogenera was consulted for the assignments. From these tiering, sediment modification, and sediment interaction categories, each ichnogenus was given an occupied cube which represents a certain ecosystem engineering behavior within a given tiering depth. For each of the time intervals across the Permian-Triassic boundary, the number of occupied cubes is summed to represent the total number of ecosystem engineering behaviors present. ecosystem engineering impact values. The ecosystem engineering impact (EEI) value method 14 rank-order scores trace fossils on the basis of tiering, bioturbation behavior functional group 21 , and bioirrigation potential 14 . We have modified this method to use the same tiering categories used in the ecosystem engineering occupation cube scheme in order to make the two frameworks more comparable. Potential functional group, which describes how the burrowing organism modifies the substrate, scores trace fossils as 1 = epifaunal locomotion, 2 = surficial modification, 3 = biodiffusion, 4 = regeneration, 5 = downward conveying, 6 = upward conveying, or 7 = gallery biodiffusion. Biodiffusion, regeneration, and gallery biodiffusion are the same descriptions as those in the ecosystem engineering occupation cube framework 22 . Epifaunal locomotion describes a surficial animal's movement which does not penetrate the sediment-water interface. Surficial modification describes burrowing which moves particles over short distances only within 2 cm of the sediment-water interface. Conveying is divided into downward conveying, which describes head-down orientation movement, and upward-conveying, which describes movement of an organism with its head at or close to the surface 21 . Bioirrigation potential, which describes the likelihood that the burrow was flushed with sediment, is scored as 1 = improbable, 2 = probable, and 3 = possible. Many ichnogenera can occupy a range in values for each category, so they are given a summated EEI range. The final EEI value range for an ichnogenus is calculated by summing the minimum scores and the maximum scores. Trace fossils which represent simple, low-impact ecosystem engineering have scores lower in range and value, and trace fossils which represent complex, high-impact ecosystem engineering have scores higher in range and value 14 .

Results patterns in burrow tiering. The Permian primarily consists of deep and intermediate tier burrows, with
a smaller component of shallow tier burrows and no reported semi-infaunal and surficial burrows (Fig. 1). The relative abundance of deep tier burrows is stable until it decreases by about a third from the Wordian to the Wuchiapingian and increases again in the Changhsingian. Intermediate tier burrows follow a similar pattern but disappear in the Wuchiapingian and are only a tenth of the trace fossils in the Changhisingian. Abundance of shallow tier burrows is stable from the Asselian to the Kungurian, decreases in the Wordian, increases during the Wuchiapingian, and decreases in the Changhsingian. Surficial and semi-infaunal tier burrows were not reported in the Permian.
The Triassic marks a shift to a majority of shallow and semi-infaunal burrows. Surficial trace fossils are only present in the Induan and Olenekian. Semi-infaunal tier burrows increase in abundance from the Induan to the Olenekian and decrease in the Anisian (Fig. 1). Shallow tier burrows represent nearly half of the trace fossils in the Induan and continue to make up at least 40% of all trace fossils at each stage throughout the Triassic. Intermediate and deep burrows combined represent less than a fifth of total trace fossils in the Induan and Olenekian and increase to half of all trace fossils in the Anisian. ecosystem engineering impact values. From the Asselian to the Kungurian, ecosystem engineering impact is high in both value and range at EEI = 7-14 (Fig. 2). In the Wordian, EEI values increase in range to EEI = 5-14. EEI values remain high but decrease in range to EEI = 7-14 for the Wuchiapingian and Changhsingian. In the Induan, EEI values increase to EEI = 3-14 and remain consistent into the Olenekian. EEI values decrease in range to EEI = 4-14 in the Anisian.

Discussion
persistence of high-impact ecosystem engineering behaviors. The ecosystem engineering occupation cubes reveal that there are no ecosystem engineering bioturbation behaviors (sediment interaction-sediment modification combination) lost in the aftermath of the end-Permian mass extinction (Fig. 2). We note that some complex trace fossils, such as Zoophycos, do disappear during the Early Triassic, but other trace fossils www.nature.com/scientificreports www.nature.com/scientificreports/ representing the same ecosystem engineering behaviors continue to persist. The three ecosystem engineering behaviors present in the Permian -compression-gallery biodiffusion, backfill conveyor, and compression-biodiffusion -are still present even in the Induan. These results are in agreement with previous observations that, despite the high extinction rates and selectivity associated with the end-Permian mass extinction, there was little change in functional group diversity between pre-and post-extinction benthic ecosystems 33,34 .
Notably, gallery biodiffusion, the highest impact sediment modification functional group 14 , is present even in the deepest tiers in the Induan and Olenekian following the mass extinction event due to the persistence of trace fossils such as Skolithos, Diplocraterion, and Thalassinoides 22 (Supplemental Material, S1). Moreover, the continuation of high EEI values through the Early Triassic reveals that ecosystem engineering remains both complex and high-impact across the Permian-Triassic boundary (Fig. 3). Any effect of the mass extinction on ecosystem engineering is evident only in a decrease in minimum EEI values from the Changhsingian to the Induan (Fig. 3), but this decrease in EEI values without a contraction in range during the Triassic reflects presence of semi-infaunal and surficial trace fossils rather than the loss of high-impact ecosystem engineering ( Figs. 1 and 2). The persistence of all ecosystem engineering behaviors likely reflects either behavioral redundancy in bioturbating ecosystem engineers or that new Early Triassic bioturbators rapidly refilled empty roles of the extinct Permian ecosystem engineers 34 . collapse of bioturbation depth in the early triassic. Although no loss in ecosystem engineering behaviors is evident, the data do reveal a collapse in bioturbation depth associated with the extinction (Fig. 1). It is also evident that there are fewer ichnogenera that represent high-impact ecosystem engineering behaviors present in the deep and intermediate tiers during the Early Triassic than during the Permian (Fig. 3; Supplemental Materials S2, S3). Figure 2, ecosystem engineering occupation cubes for example, only three ichnogenera in the Induan occupy the deep compression-gallery biodiffusion cube space, whereas five ichnogenera occupy the same cube space during the Wordian. The lack of surficial and semi-infaunal trace fossils during the Permian is best explained by the existence of a well-developed mixed layer 35 , while the predominance of shallow tier burrows and loss of deep tiering burrows in the Triassic implies the loss of deep sedimentary mixing in shallow marine environments (Fig. 1). This has been observed in previous research at a variety of temporal and spatial scales 12,26,36,37 .
What caused the loss of deep tier burrows remains unclear. On the one hand, the Early Triassic has been widely associated with widespread marine anoxia 6,[38][39][40] . In modern environments and controlled experiments, some animal burrows become shallower in hypoxic settings 39,41,42 . Thus, the collapse of deep sedimentary bioturbation and lack of recovery by the end of the Early Triassic could be interpreted as the persistence of low-oxygen conditions in shallow marine environments. The persistence of some high-impact ecosystem engineering behaviors (e.g. gallery biodiffusion) in the deep tiers in the Early Triassic, however, most likely suggests that across all localities in the dataset, environmental stress limited burrowing organisms to shallower depths, but some local conditions may have permitted deeper burrowing. This explanation is in agreement with the differences in functional group extinction across the Permian-Triassic boundary reported in different depositional environments and paleolatitudes 33 . Furthermore, deep tier trace fossils tend to occur in the shallowest marine environments, www.nature.com/scientificreports www.nature.com/scientificreports/ where coarse sediment allows for easier oxygen diffusion deeper into the substrate from the sediment-water interface. These shallow facies are not present in trace-fossil bearing Olenekian sections, which likely explains the lower proportion of deep tier trace fossils after the Induan. Finally, we note that the proxies by which marine anoxia are determined may be locally controlled by the intensity of bioturbation, as bioirrigation promotes oxygen penetration into the sediment 16,26 . The geochemical signal of anoxia may, therefore, be amplified by the lack of deeper bioturbation.
implications for ecosystem recovery. The persistence of high-impact ecosystem engineering behaviors and loss of deep bioturbation across the end-Permian mass extinction both likely affected benthic biogeochemical cycling during the Early Triassic. Bioturbation exerts a control on the rate of nutrient cycling in marine ecosystems 43,44 . Globally, a cessation of deep tier bioturbation in general may have increased nutrient burial, as nutrients that reach the seafloor would not have been as likely to be reworked by conveyors and regenerators and resuspended back into the water column. Additionally, a shallowing of the mixed and transition layers would have caused the redox profile discontinuity (RPD) to migrate upwards towards the sediment-water interface 16,42 , which may have affected redox-sensitive nutrient flux and benthic microbial processes 45 . However, in some local environments, deep high-impact ecosystem engineering behaviors would have maintained benthic nutrient flux and the RPD depth by transporting nutrients throughout the sediment and promoting nutrient flow to microbial communities 46 . More specifically, gallery biodiffusion (the highest impact bioturbation functional group 14 ), vastly increases the surface area for solute exchange and active biogeochemical reaction sites [47][48][49] . This means that these complex bioturbation behaviors are a critical aspect of benthic nitrogen cycling, for example, as nitrification-denitrification processes require an oxic-anoxic interface 50 . Increased bioirrigation through gallery network burrows thus can result in increased areas of stimulated nitrification 51 and decrease the flux of inorganic nitrogen to the water column 52 . Bioirrigation may also influence benthic phosphate cycling, as phosphate adsorbs to positively charged clay particles in oxic conditions 49 . Increased bioturbation intensity by animals which create bioirrigated gallery networks has also been found to cause increased abundance of microbial communities which are unique to the burrow, either due to increased oxygen 45 or organic matter in the burrow wall 53 . Thus, as these high-impact ecosystem engineering behaviors and burrow size returned to pre-extinction levels within the www.nature.com/scientificreports www.nature.com/scientificreports/ intermediate and deep tiers, progressively more complex bioturbation would have allowed benthic biogeochemical cycling to return to its pre-extinction state.
Precisely how these bioturbation-induced changes in biogeochemical cycling may have influenced biotic recovery of benthic macrofauna is not entirely resolved, and research investigating the links between these benthic processes remains limited. However, a number of studies of modern benthic ecology have linked habitat quality and ecosystem functioning to oxygenated and bioturbated substrates 52,54,55 . An increased supply of nutrients to microbes due to complex bioturbation behaviors would have stimulated marine productivity, which may have maintained or increased the diversity of life 27,40,55,56 in the wake of the end-Permian mass extinction. However, it is still implausible that ecosystems would have fully recovered until larger, deep tier bioturbators regained pre-extinction levels on a global scale 33 . Although, there is still potential for high-impact ecosystem engineering behaviors, most notably gallery biodiffusion, to maintain habitable environments during biotic crises and mass extinctions, particularly in environments where deep tier burrows persisted in the Early Triassic.

conclusions
The data presented here show two key observations: (1) there is no loss in ecosystem engineering bioturbation behavior after the end-Permian mass extinction, and (2) the highest-impact ecosystem engineering behaviors (deep tier gallery biodiffusion) persisted in the Early Triassic. There is a pronounced shift from deep and intermediate tier burrows in the Permian to shallow and semi-infaunal tier burrows in the Early Triassic, likely related to significant loss of mixed layer development and shallowing of the transition layer. However, ecosystem engineering complexity does not seem to be majorly affected, as highly effective sediment mixing behaviors were still present throughout the entire Early Triassic. These two key observations suggest that although shallow marine environments may have still been inhospitable during the Early Triassic, generally causing bioturbating organisms to be limited to the shallow tier, some local environmental conditions were favorable enough to bioturbators to reach the deep sediment tier with rather complex bioturbation behaviors. The persistence of complex engineering behaviors across the most devastating mass extinction in Earth history suggests that there was still potential for these organisms to maintain habitable benthic environments by increasing nutrient flow throughout the sediment profile and to the surface.