A tale of two Tweefonteins: What physical correlation, geochronology, magnetic polarity stratigraphy, and palynology reveal about the end-Permian terrestrial extinction paradigm in South Africa

Fully continental deposits in the Karoo Basin, South Africa, have been used in the past to record the transition from the latest Permian (Changhsingian) to the earliest Triassic (Induan) ecosystems documenting the fate of the terrestrial Gondwanan biota during the end-Permian extinction event (Ward et al., 2005; Smith and Botha-Brink, 2014; Rubidge et al., 2016; Viglietti et al., 2018; Botha et al., 2020; Botha and Smith, 2020). And, until recently, a 70% estimate in terrestrial biodiversity loss, largely based on the vertebrate record of the Karoo, was interpreted to have occurred across the Daptocephalus to Lystrosaurus declivis Assemblage Zone boundary (AZ; Fig. 1). This biodiversity crisis has been interpreted as coeval with extinction events in the oceans (e.g., Benton and Newell, 2014; Benton, 2018). The Karoo model was extrapolated to other southern (South America: Langer, 2000; Antarctica: Collinson et al., 2006; India: Gupta and Das, 2011; Laos: Battail, 2009) and northern hemisphere continents (Angara: Battail, 1997; and Cathaysia: Tong et al., 2019).

The Karoo Basin preserves an extensive vertebrate record that has been used for more than a century to subdivide a repetitious sandstone-and-siltstone succession several kilometers thick into biostratigraphic assemblage zones (Fig. 2). Vertebrate assemblage zones were assigned either to the Permian or Triassic based on early-twentieth century interpretations of Broom (1906, 1911). These assemblage zones were used by numerous workers (e.g., Keyser and Smith, 1978; Rubidge, 1995) during a time when no numerical age information existed for the rocks in the basin. That condition has changed in the past decade, and the ages of several biozones are now better defined by high-resolution, U-Pb chemical abrasion-isotope dilution-thermal ionization mass spectrometry (CA-ID-TIMS) zircon ages (Rubidge et al., 2013; Day et al., 2015; Gastaldo et al., 2015, 2020a). Several of these ages also are placed into magnetostratigraphic context and correlated with southern hemisphere palynozones established for this critical interval in Earth history (Gastaldo et al., 2015, 2018, 2019a, 2020a). Geochronometric and magnetostratigraphic constraints, in conjunction with the development of lithostratigraphic frameworks at key localities, now allow for a more thorough evaluation of the vertebrate fossil record that has been the basis of a reported turnover event in the basin.

Here, we focus on the lithostratigraphy, geochronology, magnetic polarity stratigraphy (and related rock magnetic properties), and palynology, with previously published vertebrate paleontology (Smith and Botha-Brink, 2014; Gastaldo et al., 2017) and paleobotany (Gastaldo et al., 2017, 2018) of a stratigraphic succession with a lateral, traceable extent of ∼4.5 km that constitutes what has been known in the literature as the locality Tweefontein (see below). We present observations and data from five Tweefontein stratigraphic sections, which have been correlated physically across the area, encompassing >850 m of measured section in strata that previously were inferred to be uppermost Changhsingian to lowermost Induan in age (Fig. 1; Smith and Botha-Brink, 2014; Botha et al., 2020). These Tweefontein sections are physically correlated with measured stratigraphic sections at Old Lootsberg Pass on the Blaauwater Farm (Gastaldo et al., 2018), extending the correlation to 7 km along the strike of the major NW-SE–oriented escarpment ∼40 km southwest of Middelburg (Fig. 3). Both plant and vertebrate fossils confirm that these successions are in the uppermost Daptocephalus and lowermost L. declivis AZs (Gastaldo et al., 2017) as currently defined (Viglietti, 2020; Botha and Smith, 2020). The Tweefontein sections comprise both greenish-gray and reddish-gray siltstone (Li et al., 2017), both laterally and vertically separated, that are dominated by normal magnetic polarity, thus defining a normal polarity magnetozone. We report two new U-Pb CA-ID-TIMS zircon age constraints from these sections that provide maximum depositional ages. Correlation across our stratigraphic framework, in geochronometric and magnetostratigraphic context, indicates that vertebrate taxa used to delimit the pre-extinction Permian Daptocephalus AZ and reportedly post-extinction Lower Triassic L. declivis AZ are coeval and of early Changhsingian age. Both represent vertebrate communities that coexisted in a Glossopteris-dominated landscape.

The Karoo Basin, a foreland basin that developed inboard of the rising Cape Fold Belt (Lindeque et al., 2011; Viglietti et al., 2017), filled with a turbidite to fully continental succession in response to continental deglaciation beginning in the late Carboniferous (Johnson et al., 2006). The Karoo Supergroup contains five lithostratigraphic groups. The basal Dwyka (upper Carboniferous) and Ecca (Lower–middle Permian) groups are generally represented by diamictite and turbidite deposits, respectively, that filled a deep basin that was open to marine influence. As open waters regressed, the basin began to fill with fully continental sediments that are assigned to the Beaufort (middle Permian–Middle Triassic) and Stormberg (Upper Triassic–Lower Jurassic; Catuneanu et al., 2005) groups. The Beaufort Group is subdivided into the lower Adelaide and upper Tarkastad subgroups, of which the Balfour Formation in the former and the Katberg Formation in the latter are considered to span the interval from the Upper Permian to the Lower Triassic (SACS, 1980; Johnson et al., 2006; Fig. 2). The succession culminates in the Drakensberg Group basalts of Early Jurassic age, which are genetically related to the intrusions of the Karoo Large Igneous Province (LIP) dolerite suite.

As is the case for much of the Beaufort Group, Balfour and Katberg Formation rocks in the study area can be described by six basic lithologies that are vertically and laterally arranged in a seemingly monotonous succession (Gastaldo et al., 2018). These are: (1) intraformational conglomerate, (2) fine to very fine feldspathic or lithic wacke, (3) rare medium feldspathic or lithic wacke, (4) coarse to fine siltstone of various hues, (5) silicified siltstone, and (6) devitrified claystone. Intraformational conglomerate consists of calcite-cemented pedogenic glaebules/nodules, disarticulated vertebrate remains (skull and bone or fragments thereof), and mud-clast aggregates as framework grains in either a fine/very fine sand, sandy silt, or coarse silt matrix (Pace et al., 2009; Gastaldo et al., 2013). Intraformational conglomerate was considered to be diagnostic of the Katberg Formation and the L. declivis AZ (Ward et al., 2005; Smith and Botha, 2005; Smith and Botha-Brink, 2014). This lithofacies has been shown to be a component of sandstone bodies in the underlying Elandsberg Member of the Balfour Formation (Fig. 2; Viglietti et al., 2016, 2017; Gastaldo et al., 2018). Fluvial bedload deposits are fine- to very fine-grained, yellowish-gray feldspathic wacke that rarely exhibits a medium grain size, enveloped in coarse-to-fine siltstone. Siltstone varies in color from greenish, olive, or light olive gray (and variants) to reddish gray (and variants; Li et al., 2017; Gastaldo et al., 2019a, 2019b) along with very light gray to white when silicified (porcellanite). Rare light gray to olive-gray devitrified claystone occurs (Gastaldo et al., 2018). Traditionally, siltstone color has been used to distinguish predominantly reddish-gray siltstone (and or mottling with greenish gray) of the Palingkloof Member from the greenish-gray siltstone of the underlying Elandsberg Member in the Balfour Formation (Fig. 2). A perceived increase in the ratio of sandstone to siltstone is used to delineate a transitional contact of up to 100 m between the Balfour and Katberg Formations, which is placed where a marked increase in sandstone abundance can be observed (Johnson, 1976; Groenewald, 1996). The monotonous nature in, and limited variation of, Beaufort Group lithologies required a means by which to subdivide the succession into quasi-chronostratigraphic units. The dearth of macro- and microfossil plants (Gastaldo et al., 2005, 2019a; Barbolini et al., 2018) and the abundance of vertebrate skeletons and skeletal elements (van der Walt et al., 2010, 2015) allowed for coarse age assignments to be placed on the strata following the early interpretations of Broom (1906, 1911). These were later augmented by inter-basinal correlations based on the vertebrate biostratigraphy (e.g., Rubidge, 2005).

Broom (1906) established a six-part biostratigraphic subdivision of Karoo Basin rocks using fossil-vertebrate assemblages by placing specimens of Lystrosaurus into the Triassic and all underlying biozones, including the interval now encompassing the Daptocephalus AZ (Viglietti et al., 2016), into the Late Permian. A recent vertebrate biostratigraphy incorporates geochronological data (Smith et al., 2020; Viglietti, 2020; Botha and Smith, 2020). A Permian (Wuchiapingian) age for the base of the Daptocephalus AZ was reported by Rubidge et al. (2013; 255.24 ± 0.16 Ma; U-Pb ID-TIMS zircon) for the lower part of the Balfour Formation (Fig. 2). Day et al. (2015) presented a middle Permian (Capitanian) age estimate for the base of the underlying Middleton Formation (260.25 ± 0.081 Ma; U-Pb ID-TIMS zircon). Subsequently, the first high-precision U-Pb CA-ID-TIMS ages of zircon grains from two horizons in the upper part of the Balfour Formation were reported by Gastaldo et al. (2015, 2018). A porcellanite bed, in the Elandsberg Member, has a maximum early Changhsingian age (253.48 ± 0.15 Ma; Gastaldo et al., 2015). The second horizon, located higher in the Elandsberg Member or possibly in the lower Palingkloof Member, but in the Daptocephalus AZ, yielded a detrital zircon population of Wuchiapingian age (256.8 ± 0.6 Ma; Gastaldo et al., 2018). This older (maximum depositional) age in strata some 20 m above the porcellanite indicates reworking of an older ashfall deposit in response to landscape degradation (Gastaldo and Demko, 2011) and redeposition into a younger landform. More recently, a late Changhsingian age (252.24 ± 0.11 Ma) was reported from a pristine volcanic ash deposit in the basal part of the L. declivis AZ at Farm Nooitgedacht in the Free State Province (Gastaldo et al., 2020a). This date demonstrates that the L. declivis AZ begins in the latest Permian rather than being solely restricted to the Triassic. Whether a biodiversity crisis is recorded in the Balfour Formation or not depends on the temporal and spatial relationships of taxa considered to be diagnostic of pre- and post-extinction landscapes (Gastaldo et al., 2017, 2019a). Hence, the question remains as to whether the base of the L. declivis AZ, as currently circumscribed (Botha and Smith, 2007, 2020; Fig. 1), is unique in space and time.

Tweefontein is one of seven principal sections in the Karoo Basin around which the model for the terrestrial response to the end-Permian crisis was based. It is used as one of three localities in the Lootsberg Pass area that is reported to contain the Permian–Triassic boundary (Ward et al., 2000, 2005). Tweefontein is positioned between Lootsberg Pass, along the N9 Highway, and Old Lootsberg Pass on the Blaauwater 67 Farm (Fig. 3). Unfortunately, no GPS coordinates or other data regarding the location of either the base or top of any measured section have been provided in the literature (Smith, 1995; Ward et al., 2000, 2005; Viglietti et al., 2018), which would have facilitated duplicative efforts to resolve the reported turnover event. Only a single locality coordinate set is available. The GPS coordinates of Ward et al. (2000, 2005; S31° 49.334′, E(W)024° 48.565′, originally reported as W longitude; WGS84 standard) place the Tweefontein section in an open field with no nearby exposure. Beginning in 2010 and until 2019, we examined all erosional gullies (dongas) below the main escarpment that expose outcrops across a ∼5 km transect from southeast of the reported GPS coordinates northwest to Old Lootsberg Pass. Strata in the area are nearly horizontal, dipping at a low, ∼1° angle to the northwest, and each of our stratigraphic sections intersects one or more traceable sandstone bodies. The bounding surfaces of these sandstone bodies can be physically walked over distances of several kilometers depending on stratigraphic position and cover (Gastaldo et al., 2018), demonstrating the absence of fault displacement along strike. In addition, they allow for an understanding of lateral and coeval lithofacies relationships (Fig. 4).

The outcrop closest to the published Tweefontein coordinates is ∼400 m to the northeast of the reported location of Ward et al. (2000, 2005; Fig. 3). Three correlative sections provide the most exposure in the area (Gastaldo et al., 2017) with the most extensive one being >126 m thick on Blaauwater 65 Farm. Sections were measured in 2012 and 2013 and extended higher onto the escarpment in 2017. Two shorter sections are documented in adjacent gullies and correlated (Fig. 3). All sections were measured and described as previously reported (e.g., Gastaldo et al., 2017, 2019a, 2020a). We have referred to this complex of stratigraphic sections as Tweefontein¹ (Gastaldo et al., 2017).

GPS coordinates of vertebrates collected in these areas, and reported by Smith and Botha-Brink (2014) from their Tweefontein locality, place these specimens on the Lucerne 70 Farm beneath the resistant Katberg Formation sandstones of Lootsberg Pass. The site is >4 km to the southeast of the coordinates published by Ward et al. (2000, 2005). Here, we have measured and correlated two long stratigraphic sections, totaling >240 m, adjacent to where Smith and Botha-Brink's (2014) vertebrate specimens were collected. Herein, we refer to these sections as Tweefontein² (Fig. 3).

To construct our stratigraphic framework for exposures along the northwest to southeast face of the Lootsberg Pass escarpment, we physically traced the upper bounding surface of thick fluvial sandstone bodies with a waypoint taken every 40 m to 80 m (Fig. 4). Gastaldo et al. (2017) presented the results of one such waypointed sandstone body that can be followed along strike for a distance of more than 4 km in, likely, the Elandsberg Member. This basalmost sandstone body is used, herein, as one of several inter-locality correlation datums (Fig. 4; Gastaldo et al., 2017). In 2017, we encountered a well-exposed donga section ∼600 m to the northwest of Tweefontein² where another measured section, totaling ∼205 m in thickness, was completed. This section is referred to, herein, as Tweefontein^1.5 (Fig. 3).

The GPS locations of vertebrate fossils collected in these areas, which were used by Ward et al. (2005), Botha and Smith (2006), and Smith and Botha-Brink (2014), were incorporated into Viglietti et al.'s (2016) database. Each collection site was placed into our measured sections using standard field methods. A Garmin Map62S with barometric altimeter served to assist physical correlation by verifying elevational relationships (see supplemental information in Gastaldo et al., 2019a, for a discussion and test of South African coordinate systems). These data augment the supplemental information published by Smith and Botha-Brink (2014) and Daptocephalus AZ specimens discovered and published by our group (Gastaldo et al., 2015, 2017).

Thin (10 cm) beds of white claystone exposed at both localities (lower Tw¹: S31.81260°, E024.81520°; upper Tw¹: S31.811733°, E024.817450°; Tw² S31.839417°, E024.850317° ± 3 m) were sampled for U-Pb geochronology. Zircon grains were extracted from silicified or devitrified claystone from the upper Tw¹ bed and lower Tw² beds. U-Pb zircon CA-ID-TIMS geochronological methods were employed in the Jack Satterly Geochronology Laboratory, Department of Earth Sciences, at the University of Toronto as previously reported (Gastaldo et al., 2015, 2018, 2020a). Methodological details are provided in the Supplemental Information¹.

Thirty-nine sufficiently competent beds were sampled at Tweefontein¹ and 29 at Tweefontein² including sandstone, siltstone, and concretions in siltstone. Sampling for these beds (each an independent site) was done by drilling seven to 12, and occasionally more, independently oriented cores using a portable field drill with a non-magnetic diamond drill bit (Table SI1; see ^{footnote 1}). In addition, three horizons in hematitic siltstone/mudstone within 1 m of the zircon-bearing bed at Tweefontein² were sampled by extracting small flakes of rock, oriented with respect to the vertical and geographic north, placing them into non-magnetic ceramic boxes packed tightly with either cotton or glass wool, and firmly taping them shut. All beds sampled at the two localities are exposed in the principal gully in which the longest stratigraphic section is described. Our sampling areas are as low as absolutely possible in the two deeply eroded gullies; we avoided sampling any prominent feature susceptible to lightning strikes. The number of independently oriented samples obtained is relatively high for magnetic polarity stratigraphy investigations. This is: (1) because of the implicit need to fully characterize the magnetization in these rocks and (2) to convincingly assess the homogeneity, or lack thereof, of the remanence at each stratigraphic level. This is done because of the potential thermal, or thermochemical, effects of the emplacement of widespread mafic (diabase) intrusions of the Early Jurassic Karoo LIP throughout the Eastern Cape. In this area, most intrusions are of normal polarity (Gastaldo et al., 2018) and, therefore, it is not a straightforward exercise to separate an early acquired, normal polarity remanence in upper Paleozoic and lower Mesozoic strata from a normal polarity Early Jurassic “overprint” secondary magnetization, as recognized by De Kock and Kirschvink (2004). The few initial publications on paleomagnetic results from these strata in the Eastern Cape are not well documented. More recent work in the general study area, as well as in the Free State Province (Gastaldo et al., 2019a, 2020a), has demonstrated evidence of reverse polarity magnetozones, which are defined by well-behaved, stable endpoint magnetizations of relatively high laboratory unblocking temperatures. Evidence of reverse polarity magnetozones is limited, in a stratigraphic context, due to the impact of Late Permian erosion and landscape degradation (Gastaldo et al., 2018).

Core samples were prepared into standard 2.2-cm-high specimens for remanence (progressive demagnetization) and rock magnetic measurements. Rock magnetic measurements include acquisition of isothermal remanence magnetization to saturation (IRMS), measurement of bulk susceptibility as a function of temperature, and anisotropy of magnetic susceptibility (AMS). Methods follow those previously published for these rocks (e.g., Gastaldo et al., 2018, 2020a) and are available in the Supplemental Information (see ^{footnote 1}).

Plant-bearing lithologies at Tweefontein¹ were processed for palynomorphs by Global Geolab Ltd., Alberta, Canada, using standard methods. Palynological residues were mounted in polyester casting resin. The Tweefontein assemblage had a very poor yield and was slightly less mature than the one reported from Blaauwater Farm (Gastaldo et al., 2017). Images were taken with a Nikon DS-Fi1 digital camera mounted on a Leica DM2500 microscope. Sample residues and slides are housed in the Paleobotanical Collections of the University of California Museum of Paleontology, Berkeley, California, USA, curated with the numbers UCMP 398697–398704 (see figure captions).

The rocks exposed in the Tweefontein sections do not differ significantly from those we have reported upon previously to the northwest at Old Lootsberg Pass on Blaauwater Farm (Gastaldo et al., 2014, 2015, 2018). The exception is the presence of white claystone at several levels. The coarsest lithology is a calcite-cemented, intraformational, clast- or matrix-supported conglomerate (Pace et al., 2009) that includes mudchip rip-up clasts, fragmented bone, and pedogenic nodule conglomerate (PNC; Figs. 5A–5B). These lenticular, cross-bedded intervals generally appear as basal channel lag deposits and may be >1 m in thickness. Sandstone bedload deposits, representing barforms of various geometries and thicknesses, consist of fine- to very fine-grained lithic wacke and are organized into trough cross-bedded bedsets that fine upwards to very thin, low angle cross beds (Fig. 6A). Ripple structures or small-scale trough cross beds (Fig. 6B) characterize bed contacts. Paleocurrent data, based on trough cross bed axial directions, indicate a northwest orientation (azimuth = 307°; Fig. 6C). Isolated and scattered euhedral pyrite crystals may occur along bedding contacts (Fig. 6D) in sandstone bodies with the highest proportion commonly encountered in Tweefontein^1.5, Tweefontein², and eastward toward Old Wapadsberg Pass (Gastaldo et al., 2020b). Fining up sequences of sandy siltstone and siltstone, interbedded with thin, decameter-scale lenticular wacke, are common as channel fill successions. Fine-grained rock types range from sandy coarse to fine siltstone, in which fossil floral and faunal elements may be preserved, and devitrified or silicified claystone (Figs. 7B–7D; see Supplemental Information; see ^{footnote 1}).

Fossil occurrences are limited primarily to mudrock intervals. Thin, laminated siltstone beds infrequently preserve megafloral and microfloral elements (Fig. 7A; Gastaldo et al., 2015, 2017, 2018). In contrast, thick siltstone intervals may preserve vertebrate skeletal elements whereas vertebrate remains are rarely encountered in intraformational conglomerate lag deposits (e.g., Gastaldo et al., 2017, 2020c). Ichnofossils are the most common fossil and found in coarse siltstone, regardless of color, and very fine wacke. Katbergia carltonichnus burrows (Fig. 8A; Gastaldo and Rolerson, 2008) are ubiquitous across the escarpment and found throughout all measured sections. These ichnofossils often are associated with concretionary intervals. A unique interval of reddish-gray massive siltstone in Tweefontein^1.5 contains large (50-cm-wide × 30-cm-high) burrows, circular to elliptical in section, filled with greenish-gray siltstone piped down from overlying strata (Fig. 8B). Here, isolated bone fragments are scattered infrequently in both the massive siltstone and burrow fills. Reported vertebrate elements include articulated skeletons, skeletons with isolated limb bones, and isolated skulls (Smith and Botha-Brink, 2014, their supplemental table; Gastaldo et al., 2017).

Three measured sections are physically correlated and form the basis of this locality. The principal section, consisting of 127 m of relatively well exposed rock, is supplemented with two shorter measured sections (Fig. 3), one of which is detailed in Figure 9. A short 10 m section, from which a vertebrate skull was recovered and described (Gastaldo et al., 2017, their Fig. 12) and correlated into the main section, is located in Figure 3. The lowest beds crop out in shallow, <1-m-deep gulleys at the main section's base (S31.82050°, E024.81363°). These consist of fining up sequences of coarse (or sandy coarse) to fine, olive gray, well to poorly cemented siltstone. Several intervals contain pale to moderate to dark brown large carbonate-cemented nodules (Figs. 5C–5D), the exteriors of which are either smooth or exhibit amalgamation features. Septarian concretions, slickensides, and subvertical to subhorizontal burrows are rare. Limited reddish-gray color mottling occasionally is found in olive siltstones, and Katbergia burrows (Gastaldo and Rolerson, 2008) are present (Fig. 9).

An increasing proportion of coarse olive-gray siltstone and very fine, and very thin, pale olive to gray-yellow lithic wacke is encountered in the overlying succession, culminating in a fossiliferous, laminated siltstone (Figs. 7A and 9) in which a Glossopteris (Glossopteridales) and Trizygia (sphenopsid) megaflora and palynological assemblage is preserved (Gastaldo et al., 2017). Here, palynomorph recovery was low, and mixed preservation was exhibited (Fig. 10). Identifiable specimens include Columnisporites (trizygiod sphenophytes) and simple trilete spores (Granulatisporites and Lophotriletes). Granulatisporites is known from very different plant groups, whereas Lophotriletes spores have been described from several herbaceous ferns (see Balme, 1995). Pollen include alete and taeniate bisaccates (cf. Lunatisporites and Protohaploxypinus), similar to those reported from other Lopingian assemblages (Prevec et al., 2009, 2010; Gastaldo et al., 2014, 2015, 2020a; Fig. 10). A centimeter-thick, light gray devitrified claystone caps this interval; no zircon grains were recovered. The siltstone interval is in sharp erosional contact with a basal channel form sandstone body (Fig. 9).

The basal channel system varies in thickness from a maximum of ∼11 m to less than a meter of lithic wacke exposed in correlative sections to the southeast. This sandstone body can be traced laterally along strike for several kilometers (Gastaldo et al., 2018) before it thins to the southeast and is obscured by cover. Clasts of fine- to very fine-grained sandstone are organized into trough cross bed sets ranging in thickness to >1.5 m and consist of millimeter- to centimeter- scale, low angle beds (Fig. 6A). At least four localities to the southeast (Gastaldo et al., 2018, their Figs. 4 and 5; 2020c) expose basal, PNC channel lag deposits that contain pebble- and cobble-sized clasts. The complex is overlain by a ∼15-m-thick interval of fining up siltstone successions with mudstone rip-up clasts in coarser lithologies and a few nodular horizons in olive-gray siltstone. A thin (30 cm), finely laminated, silica-cemented porcellanite and overlying devitrified claystone (lower Tw¹) are correlative with an interval of coarse siltstone in the primary gully that is truncated by an erosional contact of a thin overlying lithic wacke body (Fig. 9).

The major upper sandstone body ranges from a thickness of 8 m in the main section to >11 m to the northwest (Fig. 9). It also is a fine- to very fine-grained lithic wacke, although with a slightly higher feldspar content, organized into trough cross bed sets that attain a thickness of >1 m. Neither basal PNC lag deposits nor mud-pebble conglomerates were observed in these locations. Basal, low-angle trough cross beds, of centimeter- to millimeter-scale, are organized into 5 cm bedsets near the top. Paleocurrent data, based on trough cross bed axial directions, indicate a unimodal flow direction to the northwest (azimuth = 327°; Fig. 6C). These bedload-channel deposits are overlain by interbeds of olive-gray coarse and fine siltstone and thin, very fine wacke. A 10-cm-thick, thinly bedded, zircon-bearing claystone (upper Tw¹), exposed in the shorter section to the northwest, is part of the overlying channel fill complex (Figs. 7B and 9 at asterisk).

Olive-gray siltstone, with few concretionary/nodular horizons, predominates in the overlying stratigraphic section. An isolated skull (AM4757), identified as either Daptocephalus leoniceps or a large Dicynodon sp. (Kammerer et al., 2011; C.F. Kammerer, 2016, personal commun.; Gastaldo et al., 2017, their Fig. 12) was recovered from a nodule-bearing, olive-gray siltstone exposed in the short section to the southeast (S31.8175833°, E024.825233°; Fig. 4). Thereafter, a change in siltstone character occurs in the upper 20 m of section with an increasing proportion of reddish-gray coloration (Fig. 9). As one ascends the slope of the escarpment, exposure becomes limited to resistant sandstone benches with a few meters of underlying fine lithologies, and the remaining intervals are covered by colluvium, talus, and vegetation at higher elevations.

The base of the section is exposed as pavement in this donga (S31.83327°, E024.84056°). The lowest 1.5 m interval is unique and consists of mottled greenish- (5G 6/1) and brownish-gray (5YR 4/1) sandy coarse siltstone in which centimeter-scale mudchips are abundant. This matrix envelopes olistoliths of intraformational conglomerate displaying random orientations. Katbergia burrows are preserved in both the olistoliths and siltstone matrix, and bone fragments are a component of all PNCs. The interval coarsens upwards to lenticular beds of intraformational PNC and bluish-gray (5B 6/1), sandy coarse siltstone. This, in turn, is overlain by a fossiliferous 20 cm interval in which small-diameter unidentifiable plant axes, Katbergia, and slickensides are preserved. Katbergia-burrowed intervals continue upsection in which both grayish-red (10R 4/2), pale olive (10Y 6/2), and mottled olive and grayish-red siltstone are found. Grayish-red siltstone dominates, grading up section to mottled and pale olive lithologies. Katbergia burrows may exceed 25 cm in length along exposed surfaces, and successive lithologies appear to have been independently burrowed, including low-angle, cross bedded lenticular sand bodies. There is a distinct change in sandstone color upsection to yellow-gray (Fig. 11), where euhedral pyrite crystals become common on bedding planes (Fig. 6D).

Sandstone bodies dominate the measured section, each of which exhibits an erosional lower contact with underlying fine-grained lithologies. Intraformational PNC lag deposits are more common in this region of the escarpment and are exposed at the base of nearly all thick sandstone intervals. One particularly thick PNC exhibits sigmoidal cross stratification (Fig. 5A). Sandstones range from thin, lenticular bodies of <0.5 m in thickness to greater than 15 m (Fig. 11) where low-angle trough cross bed sets are organized on the scale of 60 cm or more near the channel base to 15 cm or less near its top. Lenticular beds are en echelon stacked. Due to the prevalence of cover outside of the erosional donga, it is not possible to determine if fining up intervals of thick sandstone to sandy coarse siltstone represent individual channels or if these intervals represent multi-storied bodies. Internal bed thickness of a typical bedset ranges from centimeters overlying an erosional contact to millimeter-scale at the top of the trough fill (Fig. 6B).

In contrast to siltstone intervals in Tweefontein¹, reddish-gray (brownish-gray) siltstone predominates over olive-gray siltstone. Mudchips, ranging from 1 mm to roughly 1 cm in size, are intermixed with these lithologies and also may be found in Katbergia burrow fills. Millimeter- to centimeter-sized, carbonate-cemented nodules are uncommon in rubified intervals. The first reddened mudrock intervals are terminated upsection by greenish-gray to olive-gray siltstone over which lies a sandstone body with a basal erosional contact. Large presumably vertebrate burrows are preserved in a reddish-gray, 4 m interval near the top of the section (Figs. 8B and 11). These elliptical to amorphous-shaped structures are the first to be encountered on any Blaauwater, Lucerne, or Tweefontein Farm exposure (Fig. 3). Burrow casts consist of light bluish gray (5B 7/1) or light olive gray (5Y 6/1) sandy coarse siltstone, in which mudchips and dispersed mica flakes are prominent and surrounded by brownish-gray coarse siltstone. Burrow fills occasionally show brownish-gray mottling or cross-cutting burrow casts of Katbergia and an unidentified 0.5-cm-diameter cylindrical burrow.

Our two measured sections, ∼0.3 km apart, are dominated by the same coarse clastic fraction as in other measured sections, with a higher proportion of sandstone than mudrock intervals (Fig. 11). There is no variation in the characteristics of channel bodies upsection, although there is notable variation in the character of fine-grained sedimentary rocks. The base of our longest section begins in a pavement exposure located in a shallow gully in an open field (S31.84307°, E024.84777°) where light olive-gray and olive-gray siltstone occurs. Reddish-gray siltstone first appears within 10 m of the base and increases in abundance vertically with the presence of both large (decimeter-scale), carbonate-cemented concretions and Katbergia burrows. Thin lenticular or sheet sandstone bodies are intercalated in these basal deposits. A light olive, devitrified claystone overlain by centimeter-scale, fining up cycles of sandy coarse to coarse siltstone and a lenticular, trough-fill claystone crop out in the lower part of the section. Both were sampled for zircons (see below). There is an increasing sand component upsection to the base of the first thick lithic wacke (Fig. 11).

Grain size and sedimentologic features of the basalmost sandstone body are no different here than elsewhere. It attains a thickness of <12 m, and there is no evidence for intraformational pedogenic nodule conglomerate lags in the gully exposures of this unit. The first exposure of PNC is found higher in the stratigraphy. It is exposed in the complementary section to the northwest, where it occurs as a lens enveloped by reddish gray siltstone and is unassociated with any sandstone body (at ∼88 m; Fig. 11). Above it are fining up cycles of coarse to fine light olive-gray and reddish-gray siltstone, in which both Katbergia burrows and vertebrate bone are preserved. These units overlie lenticular and planar sandstone bodies. There is an increasing proportion of sandstone upsection, where a second channel complex is encountered.

The channel complex in the middle of the main section (starting at ∼105 m; Fig. 11) may be multistoried. Meter-scale lenticular bodies overlie an erosional contact with reddish-gray siltstone. These are in erosional contact with overlying PNC and may represent a second phase of channel organization. Exposure in the shorter section to the northwest shows a thick set of fining up successions of light olive siltstone between thick lithic wacke bedsets, which are interpreted as channel fill deposits. These fine-grained rocks are overlain by an erosional contact (at ∼110 m) and another interval of meter-scale wacke bedsets, which may represent a third phase of bedload deposition. Fining up cycles of reddish-gray, coarse- to fine-grained siltstone, in which decimeter-scale carbonate-cemented concretions and Katbergia burrows are preserved, continue and are interrupted by thin beds of lithic wacke.

Resistant sandstone bodies become more prominent at higher elevations of the main section, where fine-grained intervals are covered by talus and vegetation. The measured section is capped by a >15-m-thick channel succession in erosional contact with underlying reddish-gray siltstone. The predominance of reddish-gray siltstone in this part of the escarpment contrasts markedly with the greenish-gray, fine-grained intervals exposed to the northwest at Tweefontein¹ (Figs. 3, 4, and 9).

Zircon grains were recovered from two horizons sampled, one exposed at Tweefontein¹ (Tw¹; S31.81173°, E024.81745°) and the other at Tweefontein² (Tw²; S31.83942°, E024.85032°). The upper claystone at Tweefontein¹ contains abundant zircon, which includes both euhedral elongate crystals, equant multi-faceted grains, and many mechanically rounded, frosted, detrital grains (Fig. 12A). Rounded detrital grains of both rutile and titanite are present in trace amounts. Sharply faceted zircon grains that showed no surface abrasion and, thus, a greater likelihood of derivation directly from a volcanic source were analyzed. Concordant U-Pb results for 14 zircon grains have ²⁰⁶Pb/²³⁸U dates that range from 277.8 ± 0.9 Ma to 254.51 ± 0.33 Ma (Fig. 12A, Table 1). The youngest five results form a cluster (z10–14, Fig. 12A) that has a weighted mean ²⁰⁶Pb/²³⁸U age of 254.73 ± 0.24 Ma (2σ, MSWD = 1.9; Fig. 12A inset; Fig. 9 for stratigraphic context), which is interpreted as a maximum age for the time of deposition.

Abundant zircon grains recovered from a lenticular claystone trough fill at Tweefontein² range from euhedral to subhedral, long prismatic to equant multi-faceted grains (Fig. 12B, inset; Figs. SI1–SI2; see ^{footnote 1}). A selection of the freshest, least rounded grains gave a range of dates typical of a detrital population (Fig. 12B; Table 1). Results of four of the five youngest grains (Z11–14) overlap and have a weighted mean ²⁰⁶Pb/²³⁸U age of 252.432 ± 0.19 Ma (MSWD = 0.5). The youngest date obtained (Z15) plots to the right of the concordia curve, which may indicate minor Pb loss in the grain and, therefore, a ²⁰⁶Pb/²³⁸U that is younger than its crystallization age. Therefore, we consider 252.43 ± 0.19 Ma to represent a maximum age for deposition of the claystone horizon.

Intensities of the natural remanent magnetization in the rocks sampled at Tweefontein¹ (Tw¹; N = 39 sites) and Tweefontein² (Tw²; N = 32 sites) typically range from ∼5 mA/m to 0.5 mA/m. Specimens from all samples from both Tweefontein sites studied, to date, yield a first removed well-defined northwest to north-northwest, moderate to steep negative inclination magnetization. At higher laboratory unblocking temperatures or peak alternating field (AF) values, a similar, well-defined component of magnetization usually is isolated, and the remanence is interpreted to be of normal polarity (Fig. 13; Table 2; Fig. SI3 and Table SI1; see ^{footnote 1}). We refer to this remanence as the characteristic remanent magnetization (ChRM) in these rocks. As above, this ChRM component is notably comparable to the present day field direction of the Eastern Cape as well as the ChRM direction of mafic intrusions of the Early Jurassic (ca. 184 Ma) Karoo LIP (e.g., Hargraves et al., 1997; Geissman and Ferre, 2013; Gastaldo et al., 2018, 2019a), which are predominantly of normal polarity in this part of the basin (e.g., Gastaldo et al., 2018). In most sites at both Tweefontein sections, this principal magnetization component is unblocked over a range of laboratory unblocking temperatures to ∼580 °C and randomized in AF demagnetization by 80–100 mT. At the site (= bed) level, the within site (between sample) consistency in magnetization character is typically high (Figs. 14–15) with estimated precision parameters (k; Fisher, 1953) typically greater than 30 (Table 2). For most sites, the maximum laboratory unblocking temperatures of ∼580 °C and median destructive fields of 20–40 mT are interpreted to indicate that magnetite is the principal carrier of the remanence in these rocks. This interpretation is supported by results of IRM acquisition and backfield demagnetization (Fig. SI4; see ^{footnote 1}) and continuous measurements of bulk magnetic susceptibility as a function of heating and cooling (Figs. SI5–SI6; see ^{footnote 1}). However, several sites that are located high in the stratigraphic sequence at Tw² (e.g., sites Tw²_68 and Tw²_69) yield well-defined magnetizations of normal polarity but of laboratory unblocking temperatures that range well above 600 °C (Figs. 13 and 15). These characteristics indicate that hematite is the principal remanence carrier at these horizons. Hematitic mudstones sampled (sites Tw²_2019_1–3) above and below the zircon-bearing claystone at Tweefontein² contain a mixture of both magnetite and hematite, as shown in progressive demagnetization (Fig. 16). This is supported by continuous susceptibility versus temperature measurements. (Fig. SI6; see ^{footnote 1}).

In total, the paleomagnetic data obtained, to date, from both Tweefontein sections show that these sequences are dominated by a ChRM of normal polarity (Fig. 17; Figs. SI7–SI8, see ^{footnote 1}; Table 2). At the site level, the dispersion of individual sample directions is at the acceptable level (estimated α₉₅ values typically less than ∼15°; Table 2; Table SI1). However, in some cases, the dispersion is higher, and this is partially interpreted to be a function of the limited number of samples presently demagnetized and the relatively low natural remanent magnetization (NRM) intensities in some of the beds sampled. The ensemble of estimated site-mean directions from a total of 45 sites (out of 57 reported at present, Table 2) from both Tweefontein sections, which we accept on the basis of having a 95% confidence limit of less than 16.5°, yields a grand-mean direction of Decl. = 331.2°, Incl. = −55.0° (N = 45 sites, α₉₅ = 3.7°, k = 34.0). This grand-mean direction, admittedly uncorrected for possible inclination shallowing (see Fig. SI17 and discussion in caption; see ^{footnote 1}), provides a southern hemisphere pole position of 66.3° S, 95.7° E (K = 17.0, A₉₅ = 5.3 (Table 2). If we can assume that this paleomagnetic result represents a suitably time-averaged record of the geomagnetic field in the latest Permian, then the resulting pole can be compared with paleomagnetic data placed into southern African coordinates (e.g., Torsvik et al., 2012; Fig. 17I). It also can be compared with results from other continents by first placing Southern African coordinates into West African coordinates using a −7.8° rotation about a Euler pole located at 9.3°N and 5.7°E (Torsvik et al., 2012). This results in a southern hemisphere pole location of 57.8°S, 106.2°E (Fig. 17J). We note that if a correction for inclination shallowing was applied (SI7; see ^{footnote 1}), then any resulting pole location will lie closer to the sampling locality in the Eastern Cape Province.

Samples from two sites (Tw¹_19 and Tw¹_20) provide, to varying degrees in progressive thermal demagnetization, subtle evidence for the preservation of a remanence of southeast to south-southeast declination and moderate positive inclination. Admittedly, this was not well-isolated at a high level of confidence (Fig. 18 and Fig. SI9; see ^{footnote 1}). We tentatively interpret these magnetizations to represent the possibility of a reverse polarity magnetozone preserved in the Tweefontein¹ section.

Bulk magnetic susceptibility (MS) values for sites in both of the Tweefontein sections typically range from ∼7 × 10⁻⁴ to 3 × 10⁻⁵ SI volume units and are usually very consistent at the site level (Fig. SI10; see ^{footnote 1}). Anisotropy of magnetic susceptibility (AMS) data from these rocks reveal magnetic fabrics that are consistent with the preservation of primary depositional textures (Fig. SI10) with K_min axes oriented close to vertical. K_max and K_int axes either are relatively well grouped with sub-horizontal orientations (triaxial distribution) or more uniformly distributed within a sub-horizontal plane, in which case the T parameter approaches 1.0. The degree of anisotropy (P) ranges from 1.004 to 1.114 and is typically very consistent at the site level; this is especially true for those sites with a small variation in bulk susceptibility among specimens.

The current terrestrial end-Permian biodiversity crisis and vertebrate-extinction model (e.g., Erwin, 2006; Ward et al., 2005; Smith and Botha-Brink, 2014; Roopnarine and Angielczyk, 2015; Rubidge et al., 2016; Botha et al., 2020; and others, but see: Lucas, 2017, 2018, 2020; Schneider et al., 2019) is based, largely, on lithostratigraphic and vertebrate biostratigraphic patterns reported from the Karoo Basin. The model proposes a succession of five lithofacies, distributed over ∼60 m of stratigraphy, in which an aridification trend is interpreted. Massive, greenish-gray siltstone characterized by meandering fluvial channels (Facies A; Smith and Botha-Brink, 2014) is overlain by an interval of massive maroon and gray mudrock, with some attendant mottling (Facies B), culminating in a thin, interlaminated heterolithic and, reportedly, unique “event bed” (Smith and Ward, 2001) that is interpreted as a basin-wide isochronous playa deposit (Facies C). This horizon is considered to be coeval with the end-Permian marine extinction event of similar duration (Smith and Botha-Brink, 2014) and marks the maximum extent of aridification, vegetational collapse, and vertebrate extinction (Ward et al., 2000, 2005). Predominantly massive maroon siltstone that immediately overlies the heterolithic interval is interpreted to be indicative of harsh, semi-arid conditions in which vertebrate assemblages rapidly recover. Fluvial channels of the overlying Katberg Formation have been interpreted to represent anabranching regimes that formed in response to increased sediment load as a consequence of landscape denudation and the onset of an unpredictable monsoonal climate (Retallack et al., 2003; Smith and Botha, 2005; Smith and Botha-Brink, 2014). This pattern has been extrapolated to other parts of the globe (e.g., Benton and Newell, 2014).

In the Karoo model, taxa of the Daptocephalus (formerly Dicynodon AZ; Viglietti et al., 2016; Viglietti, 2020) AZ are assigned a Permian age and reported from the Elandsberg and lower Palingkloof Members of the Balfour Formation (Figs. 1–2). Olive-gray siltstone is reported to predominate in these pre-extinction landscapes, although an upwards change in siltstone color also is noted (associated with the Palingkloof Member), which includes an increased frequency of mottling and alteration to reddish gray color. Together with sandstone geometry, the presence of PNC, and other features, this change has been associated with climate drying across the basin (Ward et al., 2000, 2005; Smith and Botha-Brink, 2014; Viglietti et al., 2018). A drier climate is interpreted to have pushed ecosystems past physiological thresholds, which resulted in ecological collapse associated with facies C (Ward et al., 2000; Smith and Ward, 2001; Huey and Ward, 2005; MacLeod et al., 2017; but see Retallack et al. [2003] and Gastaldo et al. [2014, 2020b, 2020c] for a different interpretation of climatic conditions during the same interval). This horizon marks the top of the Daptocephalus AZ and the culmination point of extinctions reported over the uppermost ∼20 m of this biozone. The bulk of the vertebrate turnover from the Daptocephalus to L. declivis AZs is reportedly restricted to a narrow stratigraphic interval of ∼20–30 m straddling either side of the biozone boundary (Smith and Botha-Brink, 2014, their Fig. 12; Botha et al., 2020, their Fig. 9). These extinctions equated to the end-Permian event (Phase 2; Smith and Botha-Brink, 2014; but see Gastaldo et al., 2019a, 2020a) reportedly are succeeded by a rapid recuperation of vertebrate taxa (Recovery Phase 1; Botha and Smith, 2006, 2007; Smith and Botha-Brink, 2014) in the overlying L. declivis AZ (Botha and Smith, 2020). This pattern is in sharp contrast to a diversification lag in the marine realm, where the extinction of major invertebrate lineages, ocean anoxia, and a shallow oxygen-minimum zone retarded recovery (Lau et al., 2016). Evidence used to propose an arid setting for the rapid recovery in the lower/basal L. declivis AZ under arid conditions includes: the proposed predominance of loess (Smith and Botha-Brink, 2014; Both and Smith, 2020; Viglietti et al., 2018), vertic structures and iron content reported from paleosols (Smith and Ward, 2001; Smith and Botha-Brink, 2014; Botha et al., 2020), the presence of PNC beds (Smith and Botha-Brink, 2014, Botha et al., 2020), architectural elements associated with low sinuosity fluvial regimes in the upper Palingkloof Member and anabranching regimes for the Katberg Formation (Smith and Ward, 2001; Smith and Botha-Brink, 2014; Viglietti et al., 2018; Botha and Smith, 2020), as well as isotope geochemistry (Rey et al., 2016). Yet, detailed analyses of the transition interval over which the turnover is interpreted to have occurred yields no physical or geochemical evidence of loess deposition (Gastaldo and Neveling, 2016; Gastaldo et al., 2019b) or other evidence for aridity (Li et al., 2017; Gastaldo et al., 2019b, 2020b, 2020c).

Ward et al. (2005) presented magnetostratigraphic information for sections, reportedly sampled across the boundary, at West (Old) Lootsberg and (East) Lootsberg Pass (Fig. 3). The magnetic polarity stratigraphy that they present begins with a normal polarity zone (in Old Lootsberg Pass) and then a short reverse polarity interval, which is succeeded by a normal polarity zone that commences a few meters above their vertebrate-defined boundary. This normal polarity magnetozone is at least 120 m thick and is followed by a reverse magnetozone that begins ∼100 m above the reported biozone boundary and extends higher into the Katberg Formation. To date, we have been unable to locate any physical evidence of conventional (drill hole) sampling in rocks at Old (West) Lootsberg Pass that could be resampled at the identical locations to replicate the reported pattern. In addition, we have been unable to replicate the previously reported polarity pattern at Old (West) Lootsberg Pass (Gastaldo et al., 2015, 2018). We have identified two very thin reverse polarity magnetozones in the section (Fig. 19), one of which is constrained by a U-Pb CA-ID-TIMS maximum depositional Changhsingian age (253.48 ± 0.15 Ma; Gastaldo et al., 2015). Magnetic polarity and geochronometric data from our Tweefontein¹ and Tweefontein² sections, which are in an intermediate geographic position between Old Lootsberg Pass and Lootsberg Pass (Figs. 3 and 4), in conjunction with new and previously published paleontologic data, allow us to test the Karoo Basin paradigm previously used by other workers.

A comparison of our stratigraphic sections (Figs. 9 and 11) with those previously presented for Tweefontein (Ward et al., 2000, their Fig. 1) reveals some similarities along with major differences in their overall character. Single-storied sandstone bodies are reported to be enveloped in olive-colored siltstone of the uppermost Daptocephalus AZ, where occasional reddish-gray mottling may be found (Smith, 1995; Ward et al., 2000, 2005) that increases in abundance upsection (Botha and Smith, 2006; Viglietti, 2020). Single-storied channel deposits incised into olive-gray siltstone with occasional mottling are consistent with our observations at Tweefontein¹ but not at Tweefontein², where basal erosional surfaces of channel architectures are in contact with reddish-gray siltstone (Fig. 19). At Tweefontein¹, greenish-gray siltstone predominates in the first ∼100 m of stratigraphy rather than being restricted to the lowest 25 m of section, as illustrated by Ward et al. (2000), which is more consistent with the pattern at Tweefontein². Key sedimentary successions interpreted to represent the post-extinction facies of the L. declivis AZ (Ward et al., 2000) consist of multistoried sandstone bodies that are frequently bounded at the base by isolated PNC lag deposits. This change in fluvial architecture reportedly is preceded by a change in siltstone color to reddish gray ∼20–30 m above the base of the Lootsberg and Tweefontein sections of Ward et al. (2000). In contrast, we find no evidence in any Tweefontein section for a change in fluvial architecture and, previously, PNC lenticular lag deposits were reported in association with single-story channels situated tens of meters below the base of the Katberg Formation (Gastaldo et al., 2018; also see Viglietti et al., 2017). Rather, we find evidence in support of similar channel architectures continuing into the L. declivis AZ as it is currently defined (Figs. 9, 11, and 19). All sandstone channels display: (1) basal, intraformational PNC lag deposits; (2) similar maximum thicknesses; (3) trough cross bed sets of similar thickness and stacking patterns; (4) the same set of primary structures; (5) identical grain size distributions and mineralogical features; and (6) similar architectural organization in exposures along strike. The reported differences in previous channel architecture interpretations may be due, in part, to the location at which each section was measured or how the composite stratigraphic section was constructed. We note that there appears to be no evidence in the Sydney Basin, Australia, to support a change in fluvial style across the Permian–Triassic boundary sections there (Fielding et al., 2019). Also, given that the basal Lystrosaurus AZ, as currently defined, is apparently latest Permian in age rather than Early Triassic (Gastaldo et al., 2020a), this particular feature ascribed to the global continental extinction model (Benton and Newell, 2014) should be reconsidered.

Color differences in mudrock deposits continue to be used as one of a suite of diagnostic criteria to identify lithofacies in the Daptocephalus and the L. declivis AZs (Smith and Botha-Brink, 2014; Viglietti et al., 2018; Botha et al., 2020). Kitching (1968) described mudrock color as generally greenish-gray at the base of his Lystrosaurus AZ with a change to overlying bright reddish maroon and purple mudrock upsection. The Palingkloof Member, characterized by the predominant reddish color exhibited by the mudstones, was subsequently established for the uppermost interval of the Balfour Formation (Johnson, 1976; S.A.C.S., 1980). Smith (1995, p. 87) placed the last occurrence of Dicynodon (= Daptocephalus) coincident with the first rubified mudrocks of the Palingkloof Member. Subsequently, Ward et al. (2000, p. 1741) described the interfluvial silt-and-mudstone facies of the L. declivis AZ as being predominantly maroon in color when compared to the olive colors of the Permian (then equated with the Daptocephalus AZ). The position of the biostratigraphic boundary was reassigned to the lower strata of the Palingkloof Member (Smith and Botha, 2005; Botha et al., 2020; Viglietti, 2020). But, in general, siltstones of the Daptocephalus AZ have been considered to be dominated by color variants of olive-gray hues, albeit with “patchy rubification of the mudrocks” at the very top of the biozone (Smith and Botha-Brink, 2014, p.103). In contrast, the fine-grained rocks of the L. declivis AZ are reported to be dominated by variants of maroon, grayish-red, or brownish-gray color.

Spatial variation in color and mottling of these end-member hues has been documented vertically and laterally throughout both assemblage zone intervals (Gastaldo et al., 2017, 2018, 2019b). The Tweefontein transect along the escarpment demonstrates this pattern on a wider spatial scale. Color variation and mottling are a function of the localized effects of diagenetic processes associated with changes in the position of the water table (Li et al., 2017). Coeval siltstones in the Tweefontein¹, Tweefontein^1.5, and Tweefontein² sections show this gradient over a distance of more than 4.5 km. If we expand the stratigraphic framework to include our measured sections at Old Lootsberg Pass on Blaauwater Farm (Fig. 19; Gastaldo et al., 2015, 2018), the lateral (7 km distance) and vertical (250 m) gradient pattern becomes more pronounced. The greater part of the stratigraphy in the northwestern part of the escarpment is dominated by olive-gray siltstone variants. In contrast, siltstones to the southeast are dominated by reddish-gray siltstone variants at stratigraphically correlative horizons (Figs. 19–20). A parallel pattern is observed in the upsection increase of channel fill sandstone complexes, which is used to delineate the base of the Katberg Formation (Johnson et al., 2006; Groenewald, 1996). The highest sandstone-to-siltstone ratios commonly are encountered where roadcuts (preferentially exploited by many researchers) expose resistant sandstone ledges. Laterally, though, these channel deposits pinch out and are replaced by siltstone-dominated lithologies, which supports observations by Groenewald (1996, p. 20) that the Katberg Formation exposures at Lootsberg Pass exhibit a high siltstone content. This clearly demonstrates the gradational and spatially variable nature of the lithostratigraphic boundaries in the study interval.

Beds of whitish to light gray silicified or devitrified claystone, from which euhedral zircon grains are recovered, are exposed in the stratigraphic sections at both Tweefontein localities and elsewhere (Gastaldo et al., 2015, 2018, 2020a; unpublished data). Zircon grains recovered from a reworked, devitrified claystone at Tweefontein¹ (Figs. 9 and 19) yield a maximum depositional age of 254.73 ± 0.24 Ma (Fig. 12A). This Wuchiapingian (259.1–254.14 Ma) age conforms to a reworked tuff (youngest zircon at 256.8 ± 0.6 Ma; Gastaldo et al., 2018) that was recovered from the same relative stratigraphic interval reported ∼1.3 km to the northwest in the Old Lootsberg Pass section (Fig. 19). These zircon dates are interpreted as maxima for the time of deposition of each bed and assume that all dated grains in the Tweefontein¹ sample are detrital or re-worked to some extent. These two horizons act as one additional tie-point in the stratigraphic framework. Another tie point in the stratigraphic framework is in the main Tweefontein² section. Here, we report on a maximum depositional age based on results from four of the youngest zircon crystals, at 252.43 ± 0.19 Ma (Changhsingian; Fig. 12B), from a tuffaceous deposit. This horizon is at a similar relative position as the horizon yielding a Changhsingian age at Old Lootsberg Pass (Gastaldo et al., 2015; Fig. 19). These circumstances further support the importance of acquiring dates for multiple beds in a given section to demonstrate adherence to stratigraphic superposition (Bowring and Schmitz, 2003; Eberth and Kamo, 2019; Rasmussen et al., 2021). We caution against the use of maximum depositional dates from single zircon grains as the true time of deposition, especially when they are used to estimate the age of the Permian–Triassic boundary (e.g., Marchetti et al., 2019; Botha et al., 2020). Herein, we use such dates only to guide and approximate our stratigraphic correlation.

Our interpreted magnetic polarity stratigraphy for the Tweefontein area (summarized in Figs. 9 and 11) differs from that reported by Ward et al. (2005), which is reportedly based on data from two nearby Lootsberg Pass sections as noted above. Demagnetization data from most beds (specific sites) at the Tweefontein¹ section show them to be dominated by normal polarity. Only two beds (Tw¹_19 and Tw¹_20) exhibit a semi-consistent hint, albeit very poorly defined, of the presence of a south-directed magnetization of moderate positive inclination. This inferred possible reverse polarity remanence is a small percentage of the NRM and, again, not well defined (Fig. 18; Fig. SI9; see ^{footnote 1}), in particular when compared to data from horizons reported from Old (West) Lootsberg Pass (Gastaldo et al., 2018). To date, all demagnetization data from the Tweefontein² section have yielded exclusively normal polarity results, including those horizons immediately below and above the claystone horizon that has yielded U-Pb zircon age data (Fig. 16). Results from other localities sampled (e.g., a resampling of Old Lootsberg Pass: Gastaldo et al., 2015, 2018; Bethulie: Gastaldo et al., 2019a; Nooitgedacht: Gastaldo et al., 2020a) are interpreted to indicate that the normal polarity ChRM typically persists up to ∼580 °C in thermal demagnetization and, for some horizons where hematite carries a substantial fraction of the remanence, the ChRM persists up to ∼680 °C. The principal component of the NRM in these rocks is likely a composite of a lower laboratory unblocking temperature, secondary magnetization associated with Early Jurassic Karoo magmatism (and possibly even younger events), and an earlier acquired remanence. At some Old Lootsberg Pass sites, as well as several sites at the Bethel Farm (Free State Province) locality, this normal polarity overprint magnetization is demonstrated to be fully unblocked by laboratory unblocking temperatures of ∼450 °C or less, and a south-directed, moderate positive inclination remanence (interpreted as reverse polarity) is sufficiently well-isolated at higher temperatures (Gastaldo et al., 2015, 2018). We interpret the reverse polarity magnetizations as early acquired, likely primary magnetizations, and show that they can be isolated at temperatures above ∼450 °C depending on the character of the rocks containing this remanence. This observation is consistent with the results reported and interpreted by Tohver et al. (2015) and Lanci et al. (2013) from lower Beaufort Group strata in the westernmost Cape Province, which were sampled in an area that is largely removed from any major exposures of Early Jurassic Karoo LIP intrusions. Thus, based on the available demagnetization data, we interpret that the Tweefontein¹ section possibly contains a thin reverse-polarity magnetozone, sandwiched between underlying and overlying normal polarity magnetozones, but its presence is poorly defined. Such a reverse polarity magnetozone may exist within the Tweefontein² section, but it either has yet to be identified in laboratory analysis, may simply not have been sampled, or may be missing from the sedimentary record as a consequence of erosion. Regardless, successions in both sections represent a time interval dominated by normal polarity chrons. This differs from our data at Old Lootsberg Pass, more than 1.5 km to the northwest, where we have presented a refined magnetostratigraphy for the area.

Two very short siltstone intervals yielding results that are interpreted to define magnetozones of reverse polarity have been identified at Old Lootsberg Pass (Gastaldo et al., 2015, 2018; Fig. 19). Both intervals are less than a few meters thick and directly underlie an erosional contact with a thick sandstone channel body, in which intraformational PNC lag deposits occur. This lithostratigraphic relationship indicates the scale of missing section represented by a significant diastem of possibly 1 Ma duration based on the physical relationships of our age constraints (Fig. 20). The diastem is in close association with landscape degradation where fluvial and interfluvial deposits were eroded in response to a change in fluvial gradient (Gastaldo and Demko, 2011). During the latter, fine-grained detritus was transported farther into the basin, whereas the coarsest fraction (mudclast aggregates [Gastaldo et al., 2013], pedogenic nodules [Pace et al., 2009; Gastaldo et al., 2020b], and bone [Gastaldo et al., 2017]) were retained locally as a function of fluvial competence rather than capacity. Hence, without the development of a stratigraphic framework for the area, geochronometric constraints, and an even more rigorous sampling program for magnetic polarity stratigraphic information in that context, one could arrive at two very different interpretations of how much time is represented in the succession.

Strata sampled at Tweefontein² yield normal polarity magnetizations that persist well above 450 °C in demagnetization and indicate, at least based on our sampling strategy and currently available data, the presence of one long, and likely uninterrupted, normal polarity magnetozone. Applying the traditional model for the end-Permian crisis to this terrestrial pattern, in the absence of geochronometric constraints, would easily allow for an interpretation that the stratigraphic record was continuous (e.g., Ward et al., 2005; Smith and Botha-Brink, 2014; Botha et al., 2020). In addition, any paleontological change could be equated to the marine extinction, which, based on magnetic polarity data from the Meishan section in SW China, is thought to have occurred within a normal polarity chron of ca. 700 ka duration (Li and Wang, 1989; Yin et al., 2001, Zhao et al., 2007; Szurlies, 2013). In contrast, the identification of two reverse polarity magnetozones of unknown duration, as a consequence of landscape degradation, situated between three normal polarity intervals in the Old Lootsberg Pass section and the possible presence of a short reverse polarity interval at Tweefontein¹, provides a plausible means to correlate these sections to proposed global magnetic polarity time scales for this interval of time (Hounslow and Muttoni, 2010; Ogg, 2012; Henderson et al., 2012; Szurlies, 2013; Hounslow and Balabanov, 2016). We note, though, that the magnetic polarity time scales referenced here differ in their placement of the Permian–Triassic boundary in the chron record and in their approach to develop each polarity time scale. Hounslow and Muttoni (2010) and Hounslow and Balabanov (2016) place the Permian–Triassic boundary at the very base of a ca. 700 ka duration chron of entirely or almost entirely normal polarity. These two estimates of the magnetic polarity time scale represent a composite of a number of magnetic polarity stratigraphic sections from numerous localities, some of which have more robust age control than others. In contrast, Ogg (2012), Henderson et al. (2012), and Szurlies (2013) place the Permian–Induan boundary, and the span of marine extinction events, within a comparable ca. 700 ka duration normal polarity chron. The magnetic polarity time scale of Szurlies (2013) is based on data from the largely non-marine Germanic Basin and has complete stratigraphic superposition. The period-level time scales of Ogg (2012) and Henderson et al. (2012) are also based on composites from a number of sources. Gastaldo et al. (2018), in conjunction with temporal constraints from a U-Pb CA-ID-TIMS age from a porcellanite (253.48 ± 0.15 Ma, 2σ error; MSWD = 0.47), as discussed above, placed deposition of these Karoo Basin rocks in the earliest quarter of the Changhsingian stage (254.14–251.902 Ma), a time interval dominated by normal polarity. New U-Pb CA-ID-TIMS zircon dates recovered from other beds in the stratigraphic framework (Figs. 19–20) help to support a maximum age limit for these rocks.

Two palynological biozone schemes are proposed for the Karoo Basin and surrounding basins over the past 25 years. Aitken (1998) described 10 zones based on palynological assemblages from the northeastern part of the Karoo Basin, specifically from the Vryheid, Volksrust (Ecca Group), and Normandien Formations (Beaufort Group). Aitkin's upper five biozones (VI–X) were largely based on samples from the borehole PA106, Lindley, Free State Province. These zones were characterized by concurrent range zones of two taxa and were not always defined by the presence of certain taxa with a last appearance datum (LAD) or first appearance datum (FAD). Rather, zones were characterized by overlap and relative abundance of the particular taxa. Aitken (1998) counted his samples quantitatively and, although the changes in relative abundance of major pollen-and-spore categories show a clear change in dominance from striate to trilete and disacciatriletes, two thick intervals in the core were unproductive (between biozones VII–VIII and IX–X). Unfortunately, the three sources of information in Aitken's (1998) thesis—biozone taxa, range charts, and zone presence accompanying the taxon description—sometimes produce contradictory information. Range charts indicate a termination of five out of seven Protohaploxypinus and Striatopodocarpites (Glossopteris-associated) taxa going from biozone VII to VIII but, at the same time, the number of taxa recorded in the higher biozones also declines. For example, there are declines of 38, 18, and nine taxa, respectively, for biozones VIII, IX, and X, while more than 10 samples were counted per biozone. This suggests that the preservation in these rocks is much poorer than that in the richer lower biozones.

The most recently published palynological biozone scheme for the Karoo Basin is by Barbolini et al. (2018). They combined data from previous studies from the western, southern, and northeastern part of the Karoo Basin with new data from roadcuts and outcrops. These authors describe 11 biozones that cover the same stratigraphic interval as Aitken's. Barbolini et al. (2018) established a latest Permian palynozonation using indicator taxa to separate the Daptocephalus AZ of the Palingkloof Member from underlying Elandsberg and Barberskrans (now Ripplemead) members. This uppermost palynozone (then considered to be latest Permian) was based on only two productive sample sites (Barbolini, 2014, table 3.1) and data from Aitken's (1998) PA106 core. These data were used to establish the Dictyophyllidites mortonii Interval Zone (Barbolini et al., 2018). The authors indicate that both localities originate from the transition from the Daptocephalus to the L. declivis AZs and equate this turnover to the end-Permian extinction event. Gastaldo et al. (2019a, their supplemental information) detail efforts to recover palynomorphs from these sediments, only one of which was found to be productive.

The one productive palynological assemblage comes from a thin heterolithic interval on Donald 207 (Fairydale) Farm (Gastaldo et al., 2019a; GPS coordinates are S30° 24.416ˊ, E026° 14.261ˊ in Barbolini, 2014, p. 103, table 3.1). Barbolini et al. (2018) indicate that the locality represents the transition across the Permian–Triassic boundary. All vertebrates reported from Donald 207 (Fairydale) Farm are assigned to the L. declivis AZ by Smith and Botha-Brink (2014; their supplemental table) and are not coincident with the reported vertebrate boundary on the adjacent Bethel Farm locality, which is >50 m lower in the section. There was no question that the palynomorph assemblage reported by Gastaldo et al. (2019a) originates from high in the L. declivis AZ and did not originate either in the uppermost Daptocephalus AZ or in the basal part of the overlying assemblage zone (Gastaldo et al., 2020a). Hence, we have not incorporated this palynostratigraphic nomenclature into our interpretations. Both studies confirm what we have experienced over the years. First, it is difficult to find productive samples from strata that are presumed to be of latest Permian age; as a result, it is not feasible to produce a high-resolution record of the Late Permian floral turnover. Second, few of the sections in the Karoo Basin provide numerical age data. Because of these limitations, we continue to rely on correlation of our palynological assemblages to the revised Australian palynozonation using overall species associations and the relative abundance of major pollen-and-spore groups.

The Australian Basins are one of the few regions in Gondwana where Lopingian to Early Triassic age palynostratigraphic biozones have been calibrated against both marine invertebrate zones and U–Pb zircon dates. These pollen-and-spore biozones are mostly based on first, or consistent, occurrence of indicator taxa in the western and eastern Australian basins (Metcalfe et al., 2015; Laurie et al., 2016). Palynological records in both regions show a floral transition from diverse assemblages characterized by multitaeniate bisaccate pollen to those dominated by algal remains with low amounts of other multitaeniate bisaccate and alete non-taeniate bisaccate pollen as well as cavate spores. This suggests the decline in diversity and abundance of glossopterids that produced multitaeniate bisaccates Protohaploxypinus and Striatopodocarpites was stepwise, while non-glossopterid taeniate forms (e.g., Lueckisporites, Lunatisporites noviaulensis and Protohaploxypinus microcorpus) and both simple and cavate spores, respectively, represent increased dominance of other gymnosperm taxa, ferns, and lycopods. More recently, and based on well-dated material from the Sydney Basin, the start of a major compositional floral change was demonstrated to have occurred at the transition between the Australian Dulhuntyispora parvithola and the Playfordiaspora crenulata zones (Fielding et al., 2019; Vajda et al., 2020; Mays et al., 2020). These data contrast with a previous interpretation that indicated a floral turnover from the P. crenulata zone to the overlying P. microcorpus zone (or, if the indicator taxon P. crenulata is missing, in the basal part of the P. microcorpus zone). Hence, the palynological assemblage recovered from Tweefontein¹ (Fig. 10) and the assemblage reported from Old Lootsberg Pass (Gastaldo et al., 2017) now are assigned to the Dulhuntyispora parvithola zone. And a consequence of these new findings is that the palynological assemblages that we described from the Wapadsberg section (Prevec et al., 2010), also in the Eastern Cape Province and ∼13 km to the southeast, would now more appropriately be correlated to the D. parvithola zone of latest Permian age. That assemblage originally was correlated to the basal part of the Australian P. microcorpus zone.

The ability to physically trace bounding surfaces of major sandstone bodies across the escarpment from Old Lootsberg Pass to Lootsberg Pass, along the valley floor exposures of the Elandsberg Member and to the flanks of the mountainside where the Katberg Formation crops out (Fig. 4; Gastaldo et al., 2018), provides the opportunity to correlate the Tweefontein stratigraphic sections and place the reported vertebrate records into chronostratigraphic, magnetostratigraphic, and palynostratigraphic context (Fig. 19). The stratigraphic position of vertebrates, based on GPS coordinates of collection sites documented by Smith and Botha-Brink (2014) and incorporated by Viglietti et al. (2016) into their Daptocephalus AZ, are reported to be of high resolution when compared to collections made prior to 1976. These older collections appear in catalogues only as farm names or, at best, as estimated coordinates, at which specimens were recovered and identified as “farm centroids” (van der Walt et al., 2010, 2015). Such records have been omitted in recent analyses due to the inherent vagaries (Viglietti et al., 2016). Biostratigraphic patterns and conclusions about extinction and diversification drawn by workers since rely on more accurate locality data with GPS coordinates that allow placement of vertebrate fossils into measured stratigraphic sections (Viglietti et al., 2016). It is on this data set that a three-phased extinction pattern, based on a narrow interval displaying a sharp turnover from the Daptocephalus to the L. declivis AZ faunas (Smith and Botha-Brink, 2014, their Fig. 12; Botha et al., 2020, their Fig. 9), was proposed. It is this pattern that has been widely accepted as the end-Permian event across the vertebrate-biozone boundary (e.g., Ward et al., 2005; Smith and Botha-Brink, 2014; Botha et al., 2020; Smith et al., 2020). And, although ∼86% of the database on which Smith and Botha-Brink (2014) developed their extinction model was collected only on three adjacent farms—Bethel, Heldemoed, and Donald 207 (Fairydale; Gastaldo et al., 2019a)—and the Tussen-Die-Riviere game reserve in the Free State Province, the distribution of taxa assigned to either the Daptocephalus or L. declivis AZs along the escarpment in the Eastern Cape Province can be used to further test the extinction model (see Gastaldo et al., 2019a, for a similar test on the Free State Province records).

Keyser and Smith (1978, p. 30) designated the succession at Lootsberg Pass as the stratotype for their Lystrosaurus AZ (now L. declivis AZ) which, in turn, is the basis for the Lootsbergian land-vertebrate faunachron (Lucas, 2010). Thirty-one vertebrates are reported from the general area (Old Lootsberg Pass [N = 14], Tweefontein [N = 8], and Lootsberg Pass [N = 9]; Smith and Botha-Brink, 2014, their supplemental data; Fig. 19). Five specimens were collected from pavement exposures and assigned (based on current criteria) to the Elandsberg and Palingkloof Members (Balfour Formation) in the low relief fields <0.9 km from our primary measured section at Tweefontein¹. Five specimens were collected from exposures in, or in very close proximity to, our measured sections at Tweefontein² assigned to the upper Palingkloof Member and overlying Katberg Formation. All these vertebrate records have been correlated physically into our stratigraphic framework. One specimen, RS17 (Tweefontein¹; Fig. 19), was used by Viglietti et al. (2016, their supplemental data) as part of the data set on which the Upper Daptocephalus AZ was established. The FAD of two taxa, Lystrosaurus murrayi and L. declivis, are used to define the base of the overlying Lystrosaurus AZ and are considered to be recovery faunal elements (Smith and Botha, 2005; Botha and Smith, 2007, 2020; Smith and Botha-Brink, 2014; Botha et al., 2020). Therefore, we consider the stratigraphic positions of these 10 specimens relative to each other as accurate in the following analysis, with some acknowledged variance (±5 m stratigraphically) due to potential problems associated with GPS replication (see Gastaldo et al., 2019a, their supplemental information for a thorough discussion of South African coordinate system usage).

When the reported sites of Daptocephalus and L. declivis AZ taxa are placed into our stratigraphic framework, discordant and inexplicable relationships appear (Fig. 19). For example, the remains at Tweefontein² of a Moschorhinus (SAM-K-K10698), a therocephalian therapsid that is one of only four genera considered to span the biozone boundary (Smith et al., 2020), previously was assigned by Smith and Botha-Brink (2014) to the L. declivis AZ (Fig. 20). Yet, it overlaps specimens of Dicynodon lacerticeps (=Daptocephalus leoniceps RS16) and Lystrosaurus maccaigi (RS81), and both taxa are diagnostic of the upper Daptocephalus AZ (Smith and Botha-Brink, 2014; Viglietti et al., 2016; Botha et al., 2020) near Tweefontein¹ (AM4757; Gastaldo et al., 2017). The Daptocephalus AZ fauna extends even higher at Tweefontein¹, where a skull cf. Dicynodon leontocephalus (Gastaldo et al., 2017) occurs at nearly the same stratigraphic horizon as L. declivis (SAM-K-10583) and L. murrayi (RS90) from Tweefontein² where both taxa are diagnostic of the L. declivis AZ (Botha and Smith, 2007; Botha and Smith, 2020). Extending the correlation northwest to Old Lootsberg Pass (Gastaldo et al., 2017, 2018) does not change the relationships between these vertebrate taxa; if anything, the degree of biozone overlap increases. We reiterate that there is no evidence of faulting along the escarpment that could have resulted in displaced strata (Fig. 4).

Together with the degree of biostratigraphic overlap, which has also been reported at Bethel (Gastaldo et al., 2019a) and Wapadsberg Pass (Neveling et al., 2016a; Gastaldo et al., 2020b), these spatial relationships of vertebrate taxa along the Lootsberg Pass escarpment represent strong evidence of faunal co-occurrence. The only difference between these collection sites is siltstone color; olive gray dominates the sections at Tweefontein¹ (colors commonly associated with the Daptocephalus AZ; e.g., Smith and Botha-Brink, 2014) and reddish gray dominates the lithostratigraphy at Tweefontein² (colors associated with the L. declivis AZ). These relationships are depicted in a simplified panel diagram (Fig. 20), which illustrates that diagnostic taxa of the upper Daptocephalus AZ taxa (Viglietti, 2020) were coeval with diagnostic taxa of the L. declivis AZ (Botha and Smith, 2020). Both assemblage zones are of early Changhsingian age and both assemblages existed in a glossopterid landscape (see above). Our data indicate that the proposed assemblage zone boundary at Lootsberg Pass (Fig. 4, white arrow; Retallack et al., 2003), considered by many to represent the terrestrial expression of the Permian–Triassic boundary and used as a correlation datum across the basin (e.g., Ward et al., 2005; Smith and Botha-Brink, 2014; Botha et al., 2020), lies significantly below the position reported by Ward et al. (2000, 2005), as field checked and documented by Gastaldo et al. (2009, 2018), at Old Lootsberg Pass. The stratigraphic position of this reported “event bed” (Smith and Ward, 2001; Retallack et al., 2003) is early Changhsingian and, applying this criterion in association with the reported vertebrate biostratigraphic boundary at each classic locality, demonstrates that the vertebrate-defined PTB occurs in different magnetozones in different stratigraphic positions in the D. parvithola palynozone (Fig. 21).

In summary, when vertebrate collections in the area of Lootsberg Pass are placed into a lithostratigraphic framework constrained by chrono- and magnetostratigraphy (Figs. 19–20), there appears to be no criterion on which to distinctly separate either the Daptocephalus (Viglietti, 2020) or L. declivis (Botha and Smith, 2020) AZ. Vertebrates that have been assigned to a latest “Permian” age at Tweefontein¹ come from a stratigraphic section dominated by olive-gray mudrock whereas those taxa collected at Tweefontein² localities, assigned an earliest “Triassic” age, come from equivalent stratigraphic horizons dominated by reddish-gray siltstone and sandstone. The overlap between assemblage zone taxa is amplified when considering their occurrence in Old Lootsberg Pass. There is no definitive level at which one assemblage replaces another and results in a biostratigraphic pastiche. It is not our intention, here, to redefine either the vertebrate assemblage zone or the biozone boundary (Fig. 21), which would necessarily require a thorough re-evaluation of the vertebrate fossil record, with accurate, high-precision collection sites placed into a stratigraphic framework developed at each collection locality and correlated using a multidisciplinary approach (see below). We note that the recent published biostratigraphic nomenclature (Botha and Smith, 2020) necessitates a lower position for the biozone contact than previously was assumed (Smith and Botha-Brink, 2014; Gastaldo et al., 2015, 2017). These findings also highlight the risks and limitations of using lithostratigraphic data to constrain a biostratigraphic horizon (Smith and Botha, 2005; Smith and Botha-Brink, 2014; Botha et al., 2020) in a basinal setting that is characterized by gradational lithostratigraphic boundaries.

To date, two different approaches have been employed to recognize the boundary between the Daptocephalus and L. declivis AZs as it is currently defined (Rubidge, 1995; Botha and Smith, 2007, 2020; Smith and Botha-Brink, 2014; Botha et al., 2020; Viglietti, 2020). These approaches control how stratigraphic sections are interpreted and related in space and time. The first is based solely on a lithofacies approach. The second utilizes a multidisciplinary approach combining lithostratigraphic frameworks, geochronology, magnetostratigraphy, and palynostratigraphy.

In the lithofacies approach, the stratigraphy spanning the biozone boundary is presumed to exhibit a characteristic facies sequence (i.e., facies A–E; Smith and Botha-Brink, 2014; Botha et al., 2020) that reportedly corresponds with biostratigraphic trends. A key component of this model is the presence of a “unique” heterolithic interval that has been used as the criterion to separate the underlying Daptocephalus AZ from the overlying L. declivis AZ (Smith and Ward, 2001; Retallack et al., 2003; Ward et al., 2005; Smith and Botha-Brink, 2014; Botha et al., 2020). This interlaminated bed of olive-gray and reddish-gray sandstone/siltstone, of a few meters in thickness, is reportedly isochronous. It is interpreted as a playa-lake deposit and used for correlation purposes across the Karoo Basin (Ward et al., 2000; Smith and Botha-Brink, 2014). In contrast, physical and geochemical evidence indicates that the bed is neither a playa-lake deposit nor does it occur at a single stratigraphic horizon in the Bethel Farm section where it is typified and to which other Karoo sections are correlated (Neveling et al., 2016b; Gastaldo et al., 2009, 2019b). The heterolithic facies is neither unique nor does it occur as a singleton at one specific stratigraphic horizon in the basin. Although these facts are acknowledged by Ward et al. (2012), the “event bed” concept continues to be applied as a correlation datum from sections in the Free State Province to those in the Eastern Cape Province, a distance of >200 km (Smith and Botha-Brink, 2014; Botha et al., 2020). And, given that this “unique” lithofacies at Lootsberg Pass (e.g., Retallack et al., 2003, their Fig. 3A) is located at an elevation below our stratigraphic framework (Fig. 4, white arrow) and is likely early Changhsingian (Fig. 19) rather than latest Permian, its usefulness as a correlation datum should be abandoned.

Botha et al. (2020) presented results of LA-ICP-MS analyses on suites of detrital zircon grains from two measured sections on Farm Nooitgedacht in the Free State Province. Here, four major Paleozoic modal ages, three of which are Cambro-Ordovician and one that is late Paleozoic to earliest Mesozoic, were identified. The largest modal peak appears to be latest Permian with all dates reported with uncertainties of several million years. A U-Pb TIMS age of 251.7 ± 0.3 Ma, based on five detrital zircons, also is reported (Botha et al., 2020, their Table 2), but no U-Pb data are presented either in the text or supplemental data. Therefore, we only can note that their earliest Triassic detrital zircon maximum age for the upper Daptocephalus AZ potentially postdates the end-Permian event in the oceans (251.941–250.880 Ma; Burgess et al., 2014) and may indicate little effect on the terrestrial biota if true. In contrast, a latest Changhsingian U-Pb age of 252.24 ± 0.15 (2σ) Ma, based on 13 overlapping zircon results obtained by CA-ID-TIMS methods for the base of the L. declivis AZ, as currently recognized (see above), has been documented from a pristine volcanic ash deposit at the same Nooitgedacht locality in lithostratigraphic and magnetostratigraphic context (Gastaldo et al., 2020a). This discrepancy between age estimates from a pristine volcanic ash deposit and a detrital sedimentary rock and, hence, a maximum depositional age, reinforces the problem identified by numerous workers with using the latter approach to constrain time (e.g., Ibañez-Mejia et al., 2018; Andersen et al., 2019; Rasmussen et al., 2021). Our age constraint at Nooitgedacht, in conjunction with data presented herein, provides a means for correlation in the Karoo Basin using the second approach (Fig. 21).

Our maximum depositional age of 253.48 Ma from a porcellanite at Old Lootsberg Pass in the Daptocephalus AZ is located in a normal polarity magnetozone, the base of which is undefined (the section continues downwards for at least 50 m, all of which is of normal polarity, into the subsurface). This normal polarity magnetozone is overlain by a reverse polarity magnetozone of unknowable original thickness and, thus, duration (see above for reasons; Gastaldo et al., 2015, 2018). This geochronometric and magnetostratigraphic relationship is consistent with the global geomagnetic polarity time scales proposed by Ogg (2012), Henderson et al. (2012), Szurlies (2013), and Hounslow and Balabanov (2016; Fig. 21). The LAD of Daptocephalus AZ taxa could be extended (based on the occurrence of cranial material assignable to either a Dicynodon, Daptocephalus, or L. maccaigi skull) into an intraformational conglomerate lag deposit at Old Lootsberg Pass (datum sandstone; Fig. 19). This position overlies the higher, short reverse polarity chron on Blaauwater Farm (Gastaldo et al., 2017). We note, however, that Botha et al. (2020, their Table 1) recently extended upwards the stratigraphic range of L. maccaigi, a diagnostic taxon of the Daptocephalus AZ (Viglietti, 2020), overlapping the basal elements of the L. declivis AZ at Nooitgedacht. This upsection shift of the taxon's range removes one criterion that is used to constrain the purportedly sharp terrestrial extinction of Smith and Botha-Brink (2014). Glossopteris megafossils and palynologic assemblages assigned to the D. parvithola biozone of middle to Late Permian age are preserved in coeval deposits at Old Lootsberg Pass. Here, megafloral elements extend into the overlying normal polarity chron that has been considered to be in the basal part of the L. declivis AZ, as currently defined (Gastaldo et al., 2015, 2017, 2020c). A similar magnetostratigraphic and vertebrate relationship has been reported from Bethel Farm, which is some 200 km away to the north-northeast in the Free State Province.

The presence of a reverse polarity magnetozone of undetermined thickness on the Bethel Farm, which is considered to be the golden spike for the end-Permian vertebrate record (Smith and Botha-Brink, 2014), was first reported by Neveling et al. (2016a). Subsequently, further analyses refined the stratigraphic extent of the reverse polarity magnetozone and indicated that the magnetozone could be traced across three measured sections that encompass more than 20 m of stratigraphy in the interval reported to transition from the currently defined Daptocephalus to L. declivis AZ (Fig. 21; Gastaldo et al., 2019a). Leaves of Glossopteris are preserved in these transitional rocks (Gastaldo et al., 2005), above which a normal polarity chron characterizes the basal part of the L. declivis AZ at this locality (Gastaldo et al., 2019a, 2019b). One palynologic assemblage assigned to the Protohaploxypinus microcorpus zone of latest Permian age, which postdates the disappearance of Glossopteris in Australia (Mays et al., 2020; Vajda et al., 2020), is preserved higher in the L. declivis AZ on the Donald 207 (Fairydale) Farm (Gastaldo et al., 2019a, 2019b). To date, we have been unable to locate any volcanic deposit or reworked tuffite in the sections in this area. However, a latest Permian U-Pb date from a thin volcanic ash deposit in magnetostratigraphic and palynological context that lies ∼33 km to the northwest on Farm Nooit-gedacht forms the basis for further correlation.

The stratigraphic interval yielding a latest Permian date of 252.24 ± 0.15 Ma at Nooitgedacht lies in a normal polarity magnetozone (Gastaldo et al., 2020a). Botha et al. (2020) report the presence of Glossopteris and sphenopsid megafloral elements, similar to those reported by Prevec et al. (2010) and Gastaldo et al. (2014, 2017), in rocks ∼17–20 m below the pristine volcanic ash deposit in the upper Daptocephalus AZ. We identify a palynoflora assignable to the D. parvithola biozone from a heterolithic unit immediately overlying the thin ash bed. Both the ash and palynological assemblage occur ∼16 m above the biozone contact as defined in previous work at Farm Nooitgedacht (Botha-Brink et al., 2014; for reasons of uncertainty about the biozone position, see Gastaldo et al., 2020a). This relationship demonstrates the continued presence of glossopterid vegetation in the overlying normal polarity magnetozone and extending into the basal L. declivis AZ at Nooitgedacht, similar to what has been reported on Blaauwater Farm (Gastaldo et al., 2015, 2017, 2018). A palynological assemblage assigned to the Playfordiapora crenulata biozone occurs ∼10 m higher at Nooitgedacht above an erosional contact with an intraformational, conglomerate-bearing sandstone body (Fig. 21). This palynological pattern, constrained by geochronology and magnetostratigraphy, parallels that reported in Australia, albeit the turnover in vegetation may be slightly younger in South Africa.

In summary, vertebrate taxa currently assigned to either the Daptocephalus or L. declivis AZs were coeval on early Changhsingian landscapes possibly as early as 252.43 Ma (Figs. 19–21). There is no evidence that the L. declivis AZ replaced the upper Daptocephalus AZ stratigraphically. Second, these landscapes were vegetated, at times, by glossopterid-dominated forests (Gastaldo et al., 2014, 2019a) and, during cooler and drier times when calcic Vertisols developed (Gastaldo et al., 2020b), other gymnosperm groups previously considered characteristic of the Early Triassic were more common (Gastaldo et al., 2018). Third, these coeval taxa persisted for a time interval of likely 100 k.y. or more before the stratigraphic disappearance of Daptocephalus and other genera currently considered as restricted to the Daptocephalus AZ (e.g., Smith and Botha-Brink, 2014; Viglietti et al., 2016; Botha et al., 2020). If the absence of Daptocephalus AZ taxa higher in the stratigraphy is not due to either sampling bias (Marshall, 2005) or misinterpretation of their original collection sites (see Gastaldo et al., 2019a; Gastaldo and Neveling, 2020), the timing of their last occurrence is somewhere high in the D. parvithola or possibly P. crenulata palynozone. Nonetheless, there is no evidence for the evolutionary first appearance of the diagnostic taxa identified as part of a terrestrial “recovery” (Botha and Smith, 2006) higher in the section. Rather, at least several “recovery” taxa were present in the early Changhsingian and persisted into the latest Permian, and the narrow lithofacies interval over which there was a reported vertebrate turnover occurs at several stratigraphic horizons in different magnetozones across the basin. Lastly, the proposal that a single, isochronous lithofacies can be used to correlate vertebrate assemblages across the Karoo Basin is demonstrated to be unrealistic when evaluated in a multidisciplinary context (Fig. 21).

A stratigraphic framework correlated across the escarpment from Old Lootsberg Pass to New Lootsberg Pass, South Africa, demonstrates that lateral facies relationships account for, and help elucidate, interpretations about the latest Permian terrestrial extinction paradigm in the Karoo Basin. Greenish-gray mudrock coloration, previously used as a criterion for recognizing Upper Permian deposits, and reddish-gray mudrock color, previously used as a criterion to recognize lowermost Triassic deposits, are laterally equivalent over at least 100 m of vertical stratigraphic section and do not represent a time-transgressive phenomenon. Rather, color variation is a function of early diagenetic processes associated with individual landscape changes in the water table over space and time. When vertebrate fossils used by previous workers to circumscribe two vertebrate assemblage zones—the upper Daptocephalus and Lystrosaurus declivis—are placed into their respective collection sites in stratigraphic context along the escarpment, taxa diagnostic of each are shown to be coeval rather than stratigraphically replacive. Hence, the coexistence of vertebrate taxa previously considered diagnostic of pre-extinction and post-extinction ecosystems in glossopterid-dominated landscapes precludes the interpretation of a sudden extinction and biodiversity crisis as a consequence of the loss of this vegetation. And the interpretation of a rapid recovery of a Lystrosaurus-dominated community appearing above the biozone boundary, which occurs in different magnetozones in different parts of the basin, now is questionable. These observations are placed into geochronometric, magnetostratigraphic, and palynologic contexts, all of which indicate that these assemblages are early to late Changhsingian.

We present two new U-Pb zircon maximum depositional ages for devitrified claystone deposits with maximum depositional ages, which augment our previously published age constraints for the area. One horizon exhibiting a Wuchiapingian age (254.73 ± 0.24 Ma) conforms in stratigraphic position to a similar age published for another detrital deposit (Gastaldo et al., 2018), both of which are interpreted as reworked. These two detrital ages in reworked channel deposits serve as one tie point in our stratigraphic framework. The second horizon, which is lower in the section and in a stratigraphic position similar to our previously published silicified siltstone, yields a Changhsingian age (252.43 ± 0.19 Ma). This horizon serves as a second chronometric tie point in the stratigraphic framework and is located in an area where all vertebrate fossils have been assigned a Triassic age in the L. declivis AZ. Hence, we conclude that diagnostic taxa assigned to the Daptocephalus AZ and L. declivis AZ in the Lootsberg Pass area, which demonstrate coeval stratigraphic relationships, were contemporaries beginning sometime in the early Changhsingian. This temporal relationship likely continued for at least one hundred thousand years if not longer. Palynologic and macropaleobotanic evidence indicates that these vertebrate communities occupied glossopterid-dominated landscapes during wet intervals and, likely, landscapes dominated by other Permian gymnosperm groups during cooler and drier intervals. But, again, there appears to be little evidence in the Lootsberg Pass area for a stratigraphic replacement or turnover of vertebrate assemblages over time.

Results of magnetic polarity stratigraphic investigations, at the current level of sampling, demonstrate the presence of normal polarity magnetozones in the eastern half of the escarpment. In contrast, two thin magnetozones of reverse polarity have been documented in the western part of the area, both of which underlie a basal channel erosional contact indicative of landscape degradation and loss of a part of the stratigraphic record. Hence, whether one interprets a single long, normal polarity magnetozone interval of more than 100 m or several oscillations in magnetic polarity over the same stratigraphic thickness, either interpretation is a function of both sampling and the inherent processes that operated in the area and resulted in the available rock record. In combination with constraining U-Pb CA-ID-TIMS zircon ages, it appears that a time interval of up to 1 Ma may be missing in the eastern part of the escarpment within the documented single long, normal polarity magnetozone, which would suggest a record of inconsistent sediment preservation at Old Lootsberg Pass. Without a multidisciplinary approach, in which high-resolution stratigraphic frameworks are combined with geochronologic, palynostratigraphic, and magnetostratigraphic data, the generalities inherent in composite or coarse stratigraphic trends may result in over-interpretation and the recognition of pseudo-events.

We propose a correlation of classic Upper Permian localities between the Eastern Cape Province (Old Lootsberg Pass, Tweefontein, Lootsberg Pass) and the Free State Province (Bethel, Heldenmoed, Donald 207 [Fairydale], Nooitgedacht) using this multidisciplinary approach. The vertical extent of our measured sections and lithostratigraphic framework of the Eastern Cape Province encompass the early to late Changhsingian, where three normal polarity chrons are separated by at least two reverse polarity chrons. Stratigraphic frameworks and measured sections in the Free State Province are correlated with the youngest reverse and normal polarity chrons at Old Lootsberg Pass and are constrained to the latest Changhsingian with a U-Pb CA-ID-TIMS zircon age from a pristine volcanic ash in the basal part of the L. declivis AZ at the Farm Nooitgedacht section. The current correlation model differs markedly from one based solely on a lithofacies approach and should aid in our future understanding and interpretation of the end-Permian crisis and preceding events.

The authors appreciate the hospitality of J. and L. Kingwill, Blaauwater Farm, and J.P. and H. Steynberg, Ganora Farm; field assistance by S. Makubalo and V. Nxumalo, Council for Geoscience; and T. Chizinski ‘14, M. Langwenya ‘14, K. Spencer ‘14, J. Li ‘16, K. Lipshultz ‘16, T. Sasajima ‘16, S. Sinkler ‘18, and K. Kus ‘18, Department of Geology, Colby College; laboratory assistance with magnetic susceptibility/anisotropy of magnetic susceptibility measurements by B.J. Lycka; laboratory assistance with thermal demagnetization measurements during the COVID-19 pandemic by S. Akatakpo; assistance with data reduction by Z. Haque and assistance with scanning electron microscope and petrographic microscope inspection by Z. Haque and L. Roberts, University of Texas at Dallas; curatorial assistance by Diane Erwin at the University of California Museum of Paleontology; and R.H.M. Smith for the vertebrate database on which the end-Permian extinction model was based. We appreciate comments by the editor, associate editor, and two anonymous reviewers that strengthened the final manuscript. Student participation was supported by the Selover Family Student Research Endowment and Barrett T. Dixon Geology Research and Internship Fund for undergraduate experiences in the Department of Geology, Colby College. R.A. Gastaldo's research was supported by the Council for Geoscience; National Science Foundation EAR 0749895, 0934077, 1123570, and 1624302; a Fulbright Award to R.A. Gastaldo at the Geology Department, Rhodes University; and faculty start-up funding to J.W. Geissman from the University of Texas at Dallas.