Geochemical evidence suggests that terminal Ediacaran (ca. 551–539 Ma) oceans experienced expansive anoxia and dynamic redox conditions, which are expected to have impacted animal distribution and behaviors. However, fossil evidence for oxygen-related behaviors of terminal Ediacaran animals is poorly documented. Here, we report a terminal Ediacaran trace fossil that records redox-regulated behaviors. This trace fossil, Yichnus levis new ichnogenus and new ichnospecies, consists of short and uniserially aligned segments of horizontal burrows that are closely associated with microbial mats. Thin-section analysis shows that the trace-making animal moved repeatedly in and out of microbial mats, with mat-burrowing intervals interspersed by epibenthic intermissions. This animal is hypothesized to have been a bilaterian exploring an oxygen oasis in microbial mats. Such intermittent burrowing behavior reflects challenging and dynamic redox conditions in both the water column and microbial mats, highlighting the close relationship between terminal Ediacaran animals and redox dynamics.


Emerging geochemical data indicate an episode of expansive oceanic anoxia in the terminal Ediacaran Period (Evans et al., 2018; Tostevin et al., 2018; Wei et al., 2018; Zhang et al., 2018), with spatially and temporally dynamic redox conditions in shallow oceans (Wood et al., 2015). It has been shown that such dynamic redox conditions influenced the distribution of animals (Tostevin et al., 2016). However, it has not been thoroughly investigated how terminal Ediacaran redox dynamics impacted animal behaviors. Terminal Ediacaran ichnofossils are ideal to address this question because, as a record of animal behaviors (Buatois et al., 2016, 2017; Gehling and Droser, 2018), they are often exceptionally preserved due to low levels of bioturbation (Droser et al., 2002). Here, we describe a new trace fossil that bears on the behavior of terminal Ediacaran bilaterian animals in response to redox dynamics.


The fossil is described as Yichnus levis new ichnogenus and new ichnospecies (see the GSA Data Repository1 for the systematic paleontology; ZooBank urn:lsid:zoobank.org:pub:991964F6-DB19–493A-A91D-BFB72973B6B1). Y. levis is preserved in limestone of the terminal Ediacaran Shibantan Member of the Dengying Formation at Wuhe in the Yangtze Gorges area of South China (Fig. 1; Fig. DR1 in the Data Repository). Specimens were collected at two horizons, ∼20 m and ∼70 m above the base of the Shibantan Member. Y. levis is the only trace fossil thus far found at the lower horizon, but it co-occurs with other as-yet-undescribed ichnotaxa at the upper horizon. Sedimentary features indicate that the Shibantan Member was deposited on a carbonate shelf, between fair-weather and storm wave base (Meyer et al., 2014). Radiometric dates and biostratigraphic data constrain the Shibantan Member to be 551–538 Ma (see the Data Repository).

Yichnus levis consists of disconnected fusiform or spindle-shaped segments that are either isolated (arrowhead in Fig. 2A) or aligned to form a curved uniserial chain (#1–6 in Figs. 2A and 2B). As many as six fusiform segments are preserved in a chain (Fig. 2B), and they are spaced with a gap in between (gap length 10.9–102.2 mm; Table DR1). Gaps are featureless and flush with the bedding surface. Segments are 13.0–84.0 mm in length and 3.4–8.9 mm in maximum width, straight (#1–2 in Fig. 2A) or slightly curved (arrowhead in Fig. 2A), generally smooth (Figs. 2A and 3A), and taper on both ends. Although segments can be curved, the ends of neighboring segments in a chain match perfectly in orientation (#1–6 in Fig. 2B).

The segments are preserved as full reliefs and can be split with the overlying or underlying beds, with the corresponding counterpart preserving a negative mold (Figs. 2A, 2B, and 2E2H). Observations made in longitudinal and transverse thin sections cut perpendicular to the bedding plane suggest (1) they are preserved in close association with organic- and clay-rich calcareous microlaminites interpreted as microbial mats (labeled “m” in Figs. 2E2G and 3C3F); (2) they are filled with micritic and intraclastic sediments, but with a greater amount of cement than in the peloidal and intraclastic sediment in the matrix (Figs. 3C–3F); (3) they are circular to oblate in transverse cross section (Figs. 2H, 3C, and 3D), indicating that the burrows experienced some degrees of postdepositional compaction; and (4) there is no discernible lining or back-filled structures.

To determine whether segments are connected with each other through unexposed intrastratal burrows, one of the specimens was cut across the gap between two segments. No intrastratal structure was found in either the part or the counterpart slab (Figs. 2C and 2D). This was also obvious upon close inspection of the terminal ends of the segment: On both the part and counterpart slabs, the segment appears to direct away from the sediment (Figs. 2A and 2B). In other words, the segment terminates at both ends, rather than going off-plane into either the part or counterpart slab; the termination is also clearly seen in longitudinal thin sections (Figs. 3E and 3F). Thus, segments are not physically connected through intrastratal burrows.


The segments of Y. levis are interpreted as trace fossils, and specifically endogenic burrows, rather than sedimentary structures or body fossils. Sedimentary structures such as syneresis cracks, tool marks, and some microbially induced sedimentary structures can be spindle in shape, but they tend to be preserved as semireliefs with a U- or V-shaped cross section. In contrast, Y. levis is preserved as full reliefs with an oblate cross section. It is closely associated with microbial mats and filled with well-cemented sediment, identical to documented trace fossil preservation in the Shibantan Member (e.g., Chen et al., 2013, their figures 2C–2F; Meyer et al., 2014, their figures 5–9; Chen et al., 2018, their figure S1). The burrow segments vary widely in length, inconsistent with body fossils, which would have consistent widths and lengths. They are aligned to form uniserial chains, and when multiple chains are found in the same bedding surface, the chains do not have consistent orientation (Fig. 3A), inconsistent with cylindrical body fossils that would be oriented by waves or currents. Importantly, although microbial laminae warp around the segments because of compaction, there are cases where microbial laminae are truncated by the segments (arrowhead in Fig. 3C), unambiguously supporting a trace fossil interpretation. These observations, when considered with the full-relief preservation, clearly identify the segments as endogenic burrows.

The alignment of burrow segments suggests that a burrow chain was made by the same individual animal, not by different individuals operating separately and autonomously. Further, the chain was probably produced by a bilaterian animal capable of directional movement, perhaps within a relatively short amount of time given the coherent course of the chain.

We further interpret that the burrow segments were originally disjointed when they were produced (Fig. DR2). Some continuous burrows can appear disjointed due to biostratinomic or erosional processes. For example, continuous horizontal burrows of Palaeophycus imbricatus can appear discontinuous on rippled bedding surface, with segments exposed in troughs but buried beneath crests (e.g., Jensen, 1997, his figure 46A). Similarly, apparently disjointed burrow segments could result from a continuous intrastratal burrow undulating in and out of the plane of preservation; this model has been proposed to explain the apparently disjointed burrows of Treptichnus (Archer and Maples, 1984; Buatois and Mángano, 1993; Jensen et al., 2000). When these continuous burrows are split, they may appear disjointed (Figs. DR2A–DR2B). However, combined observation of the part and counterpart should still reveal a continuous burrow, whereas our thin-section observation revealed that the burrow segments of Y. levis are genuinely disjointed. Thus, the stratinomic models that could make continuous burrows appear discontinuous can be conclusively ruled out for Y. levis. A third possibility is that the burrow segments of Y. levis were made discontinuous by erosional processes (Figs. DR2C–DR2D). However, erosion would likely partially or entirely remove microbial mats surrounding the burrow, whereas Y. levis is surrounded by intact microbial mats above and below. Also, if a burrow were erosionally exposed before it was filled, it would be cast from above and preserved as semireliefs rather than full reliefs. Finally, to completely remove a segment of horizontal burrow, several millimeters of strata would have been differentially eroded, considering the size of Y. levis. However, there is no evidence for erosion of this magnitude; the bedding surface is flat and smooth, largely defined by microbial laminae undisrupted by erosion. Thus, the most likely scenario is that the burrow segments were originally disjointed when they were made (Fig. DR2E). Such short burrow segments with an opening to the water column can also be more easily filled and cast with sediments than long and continuous burrows.

Finally, the consistent occurrence of Y. levis burrow segments with clay- and organic-rich microbial laminites indicates interactions between the trace maker and microbial mat. Such an association is also seen in other burrows from the Shibantan Member (e.g., Chen et al., 2013, their figures 2C–2I; Meyer et al., 2014, their figures 5–8; Chen et al., 2018, their figure S1A). The observations that the burrow segments occur in millimeter-thick microlaminites (Figs. 2E–2F, 3C, and 3D) and were originally disjointed suggest that the trace maker interacted with live microbial mats at or near the mat-water interface.

Taken together, our observations suggest that the trace maker must have surfed just below and above the mat-water interface, rather than cruising intrastratally and deeply at the interface between a buried/dead microbial mat and overlying layers of intraclastic sediment (as in the case of Curvolithus; Buatois et al., 1998). The trace maker penetrated and burrowed into live microbial mat, produced a short and unlined tunnel within the mat, emerged out of the mat, and then swam or moved epibenthically on the mat, but left no discernible epigenic trails or tracks presumably because of its inability to produce epigenic traces on the relatively thick and firm mat (Evans et al., 2019). This cycle may have been repeated to produce a chain of burrow segments (Animation DR1). Because the burrows are unlined, they are regarded as locomotion traces rather than stabilized dwelling structures. The pattern of movement, with the trace maker periodically moving in and out of the microbial mat (hereafter “in-and-out behavior”), represents relatively complex locomotion behaviors and close ecological interactions between terminal Ediacaran bilaterian animals and microbial mats.


Our interpretation of Y. levis suggests that the trace maker had biotic interactions with live microbial mats, rather than mining dead organic matter in buried microbial mats. This distinction is important because the former is animal-mat interaction but the latter is animal-sediment interaction. In ecological terms, live mats offer food and a dynamic redox microenvironment related to the diurnal cycle of photosynthesis, whereas dead mats offer only food and probably a hostile redox microenvironment (Gingras et al., 2011).

The nature of the animal-mat interaction represented by Y. levis is uncertain. Considering that the Shibantan Member was deposited in the photic zone (Meyer et al., 2014), the microbial mat was probably constructed by O2-producing cyanobacteria. We thus hypothesize that the trace maker was mining the microbial mat primarily for O2 (Meyer et al., 2014) and perhaps secondarily for food (e.g., Gehling and Droser, 2018), if the lack of back-filled structures and burrow lining was a taphonomic artifact.

We further hypothesize that the in-and-out behavior as recorded by Y. levis reflects the dynamic redox conditions in the water column (Wood et al., 2015) and in the microbial mats (Gingras et al., 2011). Ediacaran atmospheric pO2 levels are uncertain, but they were likely a fraction of present atmospheric level (PAL; Sperling et al., 2015; Lu et al., 2018), so that [O2] of fully saturated Ediacaran seawater was comparable to that of oxygen minimum zones (OMZs) in modern oceans (Sperling et al., 2013). Such redox conditions can support many animals (Mills et al., 2014), but it would be a challenge for animals engaged in energetically demanding activities such as carnivorous predation and burrowing (Sperling et al., 2013). Furthermore, with >20% of ocean floor bathed in anoxic water (Tostevin et al., 2018; Zhang et al., 2018), a percentage that is two orders of magnitude greater than in the modern ocean (0.1%; Helly and Levin, 2004), significant portions of terminal Ediacaran continental shelves were likely affected by anoxia. As in the modern ocean, anoxic seawater likely periodically flooded terminal Ediacaran outer shelves that intersected with the OMZ. Indeed, anoxia and dynamic redox conditions have been documented in terminal Ediacaran shelf deposits (Wood et al., 2015), and they have been shown to limit animal distribution (Darroch et al., 2015; Wood et al., 2015; Tostevin et al., 2016). Thus, terminal Ediacaran animals were likely compelled to explore O2 oasis in light of such redox dynamics in the water column.

Cyanobacterial microbial mats could provide, if only temporally, an O2 oasis for these animals. However, only small and mobile animals would have been able and compelled to take advantage of this O2 oasis because of their high body-mass-specific metabolic rate, as well as the strong diurnal redox dynamics and millimeter thickness of microbial mats (Fig. 4). Empirical measurements show that dissolved [O2] in modern cyanobacterial mats can reach 4–8 PAL during the day, but it drops to essentially 0 PAL during the night, when a sulfidic environment develops (Canfield and des Marais, 1993; Wieland and Kühl, 2006; Gingras et al., 2011). These remarkable redox dynamics are also confirmed by the observation of [O2] bubbles in cyanobacterial mats during the day (Bosak et al., 2010), which indicate [O2] > 5 PAL, and by diffusion-reaction models parameterized with O2 production and consumption rates as measured in modern microbial mats (Revsbech et al., 1986). Thus, although cyanobacterial mats can provide sufficient O2 to support millimeter-sized mobile animals during the day, they are a challenging microenvironment for any animals that need even a moderate amount of O2 throughout the light-dark cycle. The combined redox challenges and dynamics in the water column and microbial mats may have driven the in-and-out behavior as recorded by Y. levis.


Building upon previous observations of the close association between terminal Ediacaran burrows and microbial mats (Meyer et al., 2014), we show that uniserially aligned burrows of Yichnus levis were produced by a bilaterian animal that developed an in-and-out behavior, interacted with live microbial mats, and repeatedly burrowed into and emerged out of live microbial mats. We hypothesize that the in-and-out behavior was an evolutionary innovation driven by the dynamic redox conditions in both the water column and the microbial mat. Dynamic redox conditions were a challenge for early animals (Boag et al., 2018), but this challenge could be mitigated by mobile bilaterians with the capability to explore dynamic and localized O2 oases, which together with the heterogenic distribution of food (Budd and Jensen, 2017), may have been a stimulus for the evolution of animal mobility. This hypothesis can be further tested through a comprehensive analysis of terminal Ediacaran trace fossils exhibiting potential signs of in-and-out behavior (Jensen et al., 2000; Jensen and Runnegar, 2005; Meyer et al., 2014) to demonstrate the global scale of this innovation.


This research was funded by the National Science Foundation (EAR-1528553), the National Geographic Society (9564–14), and the Chinese Academy of Sciences (XDB18000000, QYZDJ-SSW-DQC009, and XDB26000000). We thank Xiang Chen for field assistance, and three anonymous reviewers for constructive comments.

1GSA Data Repository item 2019367, stratigraphic setting, methods, and systematic paleontology, is available online at http://www.geosociety.org/datarepository/2019/, or on request from editing@geosociety.org.
Gold Open Access: This paper is published under the terms of the CC-BY license.