These distinctive specimens of G. cornuta focus attention on other spores within the lower lake that, on examination, also display pervasive malformations. Figure 4 shows the spore Verrucosisporites nitidus, another well-known species (16) that crosses the D-C boundary. Both below (Fig. 4, A to C) and above (Fig. 4, N to P) the boundary, these spores are pale in color and with a well-formed sculpture of verrucae. Within the lower lake, V. nitidus shows a continuous range of morphologies from typical to aberrant forms. However, most specimens that have a well-formed sculpture are brown in color. A minority are pale, smaller in diameter, and with a sculpture that did not fully form, the elements being irregularly spaced and uneven in size. It would appear that spores that developed this pigmentation also acquired a degree of protection during spore wall development. It is more difficult to track malformation in the remainder of the lower lake palynological assemblage, as most of the species in this simple spore assemblage lack spines and have smooth walls. Many of these, in East Greenland and elsewhere, are also dark brown in color and remain so through the entirety of the VI spore zone. However, obvious malformed spores are absent from the upper lake in all three sections that were studied, showing that the cause of the damage has ceased.
Fig. 4 V. nitidus spores from below, in, and above the D-C boundary showing UV-B malformations.
(A) to (C) are from below the boundary (Stensiö Bjerg) and have the characteristic packed verrucae sculpture of V. nitidus. (D) to (K) are from the lower lake (Rebild Bakker), with (H) to (M) more strongly pigmented and more regular sculpture with a wider range of diameter than normal. (D) to (G) are paler in color and smaller in size and have irregularly developed sculpture. (N) to (P) are normal specimens of V. nitidus from the upper lake bed (Rebild Bakker). Sample and slide numbers plus England Finder coordinates are in table S4.
Malformations in spore and pollen walls are well documented in both living (9, 17) and fossil pollen (9, 18) and result from damage to the DNA [typically fusion of opposing base pairs (19)] that occurs before deposition of the protective sporopollenin wall layer. The pollen are then unable to replicate the intricate and regular pattern of the wall sculpture. Such DNA damage can be produced experimentally by exposure to UV-B radiation (17) and atmospheric pollutants (20). Examples from fossil material include the Permo-Triassic (9, 18, 21, 22) and Triassic-Jurassic (23) mass extinctions, where both conifer pollen and spore tetrads show malformation that have generally been interpreted as the direct effect of volcanic gases from large igneous province (LIP) eruptions. However, following experiments with living conifers (9), this has been reinterpreted for the Permo-Triassic mass extinction as the initial destruction of the ozone layer by volcanic gases with the increased UV-B flux and then damaging the pollen and spore DNA. In contrast, that from the Triassic-Jurassic has been attributed (24) to the genotoxic effects of mercury (Hg) generated during LIP eruptions.
Dark-colored spores are also reported from the Permo-Triassic (22) and Triassic-Jurassic (23) mass extinction levels. Although cause and effect remain unproven, spore and pollen walls are naturally pigmented yellow to provide protection (25) against UV-B so any additional opacity increases this protection.
Malformed spores are known (26) from other D-C boundary sections. An important record is from Poland (26) immediately below the D-C boundary (LN spore zone) that includes many malformed specimens of Vallatisporites together with spore tetrads. However, both malformed spores and tetrads are rare in Poland, with only a single sample containing tetrads above 3% that is the background mutation level (18) within normal population of gymnosperm pollen. In addition, there (26) was only a single available VI spore zone sample (and that contained only poorly preserved spores) that is coeval with the malformed spores from Rebild Bakker. These malformations from Poland were attributed (26) to the effects of volcanic eruptions. Another record is from Belgium (27), where malformed spores of equivalent age to the LN spore zone were attributed to cooling/aridity. Again, the Belgian section in common with Poland had no spore recovery in the earliest Carboniferous VI zone that was when the maximum malformation occurred in East Greenland.
Our record of malformed pollen and spores informs us that the kill mechanism acting during the mass extinction was an interval of elevated UV-B related to the loss of the protective ozone layer. The kill mechanism inferred for the Frasnian-Famennian (Late Devonian), end Permian, and Triassic-Jurassic mass extinctions is a collapse of the Earth system linked to known LIPs, which are also evidenced (28, 29) by Hg anomalies. In contrast, compiled records (26, 29, 30) of igneous activity for the D-C boundary show nothing that approached the planetary scale of a LIP eruption.
To test the LIP hypothesis [and as advocated in (31)], we analyzed Hg through the Stensiö Bjerg, Celsius Bjerg, and Rebild Bakker D-C boundary sections. The results (Fig. 1, tables S1 and S3, and fig. S9) show a number of clear peaks from Stensiö Bjerg. However, when the Hg values are normalized against TOC, the peak coincident with the AOM-rich, high-TOC interval disappears. Hg and TOC analyses (Fig. 1 and fig. S9) from the correlative Rebild Bakker samples (that lack AOM) only reach a maximum of 7 parts per billion (ppb) and 0.3%, respectively. When normalized as Hg/TOC, the values range from 14 to 26, similar to those from Stensiö Bjerg. Therefore, there is no Hg anomaly coincident with the malformed spores, i.e., at exactly the extinction level. Hg is well known (32) to be readily absorbed from lake waters by AOM. Hence, a high content simply reflects high AOM absorbing the available Hg. The remaining Hg/TOC peaks at Stensiö Bjerg (Fig. 1) are coincident with the shifts between carbonate and organic matter–dominated sediments. Hg is also known (32) to become more concentrated in lake waters that lack organic matter. This then becomes available in excess to become absorbed onto any subsequent AOM produced. These, the first records of Hg through a terrestrial D-C boundary section, in common with marine records from China (33), the Czech Republic (33), and Vietnam (31), have yet to demonstrate convincing evidence (28) for a LIP.
A second mechanism to explain an ozone collapse would be the widely repeated hypothesis (34) of reduced atmospheric PO2 (partial pressure of oxygen) across the D-C boundary. This, if correct, would lead to a reduced level of ozone. However, this hypothesis can be refuted as the boundary bed (Fig. 3C and table S5), and the entire Devonian and Carboniferous section in East Greenland and elsewhere (35) contains abundant highly reflective dispersed charcoal with reflectivities over 2% and hence of unequivocal wildfire origin. This, in common with a recently updated atmospheric oxygen modeling curve (36), shows that PO2 was in excess of 16.5% (37).
Regarding the likelihood of DNA damage from local eruptions of volcanic gases, there are no known occurrences of volcanic rocks at this time in east Greenland, the closest in age (38) being some 13 Ma earlier. With regard to airborne pollutants, the evidence (20) from pollen malformation at, for example, modern smelter sites shows that concentrations have to be very high to have any effect. So, this effect can also be discounted as a malformation mechanism because the sedimentary sections (38) through the D-C boundary interval, and throughout the basin, contains no unusual sediments that are indicative of abnormal chemically derived airborne exhalatives.
DISCUSSION
The key observation to understanding the loss of the D-C boundary ozone layer is that the malformed spores occur exactly coincident with the lower lake. In addition, the proportion of these malformed spores and their degree of darkening systematically increases to the midpoint of the lake bed. The pair of lakes present at the D-C boundary was laterally extensive (38) (at least 50,000 km2), with exceptionally high TOC contents, and represents long-lived deep lacustrine systems with a stratified water column. The presence of these lakes is in stark contrast to the normal pattern of dryland fluvial and aeolian sedimentary deposition through the 7 km and 60 Ma of the Devonian to early Carboniferous succession in East Greenland. This prevailing aridity is entirely consistent with its location (39), both in the interior of the Old Red Sandstone Continent (some 1000 km from the sea) and at 15° paleosouth in the southern arid zone. Such large lakes can only form when there is sustained seasonal rainfall sufficient to maintain a perennial standing body of fresh water. The mechanisms that bring the required intensity of seasonal rainfall into such an arid zone continental interior are well understood from many Quaternary (40) and deep-time examples (41, 42). The process is that high summer temperatures in the continental interior cause a reversal of the normal circulation. This reversal draws in moisture-laden air, which then rises and cools to produces an intense seasonal monsoon. In these models, the lakes were thermal highs and, from their normally arid continental interior locations, represent episodes of exceptional and sustained warming. This places the maximum degree of spore malformation at the midpoint of the lower lake coincident with the thermal maximum. This warming is in contrast to the cooling that a LIP causes during the actual eruption, i.e., when the malformation would occur. LIP-related cooling is the direct result (3) of the dual impact of high levels of both ash and SO2 in the atmosphere and further evidence for the absence of planetary scale volcanism at the D-C boundary.
The coincidence of maximum malformation with the climatic maximum now leads us to look for a mechanism for ozone loss driven by increasing temperature. Combined modern observation and modeling data (43) from the continental United States provide such an ozone removal mechanism with an observed increase in mid-latitude summer temperature, leading to an increase in the convective injection of water vapor into the lower stratosphere. This increased high-altitude water content then passes the threshold (43) for an increased production of catalytically active ClO with ensuing increased rate of ozone loss.
A positive feedback mechanism could also be at play in the Late Devonian and early Carboniferous, as there were extensive shelf seas with high volumes of carbon burial (2, 44) and hence high organic productivity. Methyl halogens are produced naturally by a wide range of organisms (45); hence, any increased organic productivity ultimately increased supply. Progressive ozone loss and increased influx of damaging UV-B would have led to the observed collapse of the Devonian forest environment together with its structured community. The resultant higher rates of runoff would then produce a one-time nutrient flush into the shelf seas and hence increased production of methyl halogens. The process only slowed when temperatures fell from the peak of the lacustrine cycle, allowing the ozone layer to reestablish. A time duration can be estimated from the thickness (0.8 m) of the exceptionally high TOC interval in the lake deposit at Stensiö Bjerg that represents the most persistent stratification and warmest conditions. It represents about 50% of the lacustrine cycle, so if driven by precession (42, 46), a duration of about 9 ka is likely, although it must be noted that many of the spores had already disappeared earlier in this lake cycle with the malformations being a characteristic of the survivors.
The Stensiö Bjerg sediments in the interval immediately below the two boundary lakes (Fig. 5) show a strong paleoenvironmental contrast, as the rocks are bright red in color, with indicators for aridity that include pervasive shrinkage cracks, calcrete aridisols, and a single 50-cm-thick stage 3 calcrete that represents an interval of sustained aridity (47). Palynomorphs place this interval as a correlative (10) to the LN spore zone that in marine shelf sections in western Europe (Fig. 5; Stockum II, Germany) includes, at its base, the Hangenberg Black Shale. This widely recognized black shale was at the level at which most of the latest Devonian marine extinctions were concentrated and now reinterpreted (48) as a low paleolatitude response to the forced lowering of sea level during the final glacial cycle of the latest Devonian glaciation. The total time occupied by the latest Devonian glaciation was somewhat longer [~3 Ma; (49)] and had within it separate orbitally forced glacial to interglacial cycles. Palynological correlations also directly tie this arid interval from east Greenland to sections with glacial diamictites (Fig. 5) both at high paleolatitudes, e.g., Chaguaya, Bolivia (60°S) (50, 51), and in eastern North America (52) (estimated at 30°S to 45°S). The multiple occurrences (52, 53) of glacial sediment in eastern North America demonstrate the severity of the terminal Devonian glacial cycle that was exceptional for reaching into low paleolatitudes and closer to the equator than any comparable Quaternary glaciation. The evidence from eastern North America includes giant dropstones (53), so the ice centers were not simply at altitude but glaciers were reaching to sea level. Therefore, what we have identified in East Greenland is the far-field expression of both the final Devonian glacial cycle as an episode of sustained cool aridity and its subsequent collapse that was an equally severe episode of exceptional warming. This is evidence that the climate system around the D-C boundary had a reduced resilience in that it was more easily perturbed to extremes.
Fig. 5 The D-C boundary correlated between low paleolatitude terrestrial (Stensiö Bjerg, Greenland) and marine shelf [Stockum II, Germany; (11)] sections with high paleolatitude (Chaguaya, Bolivia).
The LN* to VI spore boundary is the correlation tie and the interval of terrestrial extinctions. The terrestrial arid interval on Stensiö Bjerg correlates to the interval at Stockum II between the Hangenberg Black Shale and the LN* spore zone and equivalent to the glaciation. The Chaguaya composite section (note the reduced scale) includes a channel-filling sandstone of glacigenic origin that terminates immediately below the D-C boundary.
Our discovery provides a model for understanding the extinctions and subsequent recovery of the terrestrial biota at the D-C boundary. The initial phase of extinctions in the marine realm that occurred during deposition of the Hangenberg Black Shale (2) was the response to a glacio-eustatic sea-level fall. These extinctions would have been driven by this pronounced low latitude cooling, its disruptive effects, and the reduction in the area of the shelf seas. The second phase of extinctions, largely occurring in the terrestrial environment, was at a time of warming and caused by loss of the ozone layer. There was a collapse in the terrestrial ecosystem with the global extinction (54) of a number of hitherto very successful and widely distributed plant groups. This habitat destruction and disruption was accompanied by mass extinctions and evolutionary bottlenecks in both fish (4, 55) and tetrapods (56) that reset the trajectory of vertebrate evolution. There was also a second phase of extinctions (2) in the trilobites, ammonites, conodonts, chitinozoans, and acritarchs at this time. It is notable that many nektonic marine groups such as the conodonts (57) exhibited a “bloom” of morphotypes within separate lineages and took time to reestablish as stable clades. It can be hypothesized that this was the result of UV-B penetrating (58) into shallow water with a continuing high rate of mutations that would equally have applied to the tetrapods (59) and caused enhanced evolutionary rates during this turning point in their evolution.
The recognition that a known extinction kill mechanism, the loss of the ozone layer, occurred not only during emplacement of a LIP but at times of high global temperature identifies a new mechanism for mass extinctions. Recognition of the significance of bolide impacts (60) and LIPs (3) as kill mechanisms has transformed our understanding of the mass extinction process. However, unlike a LIP or a bolide impact, higher temperatures are a certainty in the immediate future with implications for a similar collapse of the ozone layer.