The first published stratigraphic section of the Chiweta Beds was that of Dixey (1926). Total thickness of the Karoo was measured to be 1,219 m with the upper 305 m comprising the Chiweta Beds. Later authors have applied a K1-K7 stratigraphic terminology devised for Tanzanian Karoo deposits, with K6 and K7 equivalent to the Chiweta Beds, measured by Cooper and Habgood (1959) to have a thickness of 244 m. Most bones were said to come from the middle of the unit. Gay and Cruickshank (1999) summarized the Tanzanian K6 and K7 units in their table 1 as Alibashian, Late Permian, Dicynodon Zone, and Anisian, middle Triassic, Cynognathus Zone, respectively. As shown below, these designations are too young as they are applied to the Chiweta Beds.

The Karoo Supergroup of Malawi (Figure 3) exhibits no glacial or marine sediments comparable to the Dwyka Group of South Africa. Coal-bearing strata low in the Karoo section in northern Malawi were correlated with the South African Ecca Group by Cairncross (2001). In southern Malawi, Cairncross (2001) placed the coal shales in the early Tatarian (Late Permian). However, the age of the southern Malawi coal shales is based notably on the presence of the plant species Gangamopteris obovata and Noeggerathiopsis hislopi, which if correct means these beds may well be Early Permian (Habgood 1963). Given the date of the overlying Chikwawa basalt (see below) and the possible ages of plants in the coal shales, the age range of the southern Malawi Karoo could extend from Early Permian to Jurassic, a longer range than that presented by Cairncross (2001, figure 2; Figure 3).

The Age of the Chikwawa Basalt

The distribution of the Drakensberg Volcanics and related rocks that cap the Karoo Supergroup defines the Karoo Large Igneous Province (LIP). Karoo LIP outcrops extend on the east from the Drakensberg Mountains of South Africa and Lesotho, through the Lebombo Mountains of Swaziland and Mozambique. They include the Etjo basalt of Namibia and the Batoka basalts of Zimbabwe. By proximity to the Lebombo volcanics and the Batoka basalts, the older set of extrusive rocks of the Lupata Gorge in Mozambique (Dixey and Smith 1929), and the Chikwawa basalt of southern Malawi appear to be the northern outcrop limit of the Karoo LIP.

The thickness of the igneous rocks capping the Karoo Chikwawa Group of southern Malawi is estimated by Habgood (1963) to be 1067 m, but exposures are poor, faulting complicates the relationship of outcrops, and the age is poorly constrained. Nevertheless, as in Karoo extrusives in other areas, weathered surfaces between flows are rare, suggesting a rapid series of eruptions (Habgood 1963). Based on geochronology and paleomagnetism of the Batoka basalts in northern Zimbabwe, Jones et al. (2001) determined that the duration of emplacement of the Karoo LIP was 5 Myr, ranging between approximately 184 to 179 Ma.

We report here the results of laser incremental-heating 40Ar/39Ar age determinations of a Chikwawa basalt flow (sample LJ-4) and a cross-cutting basaltic dike (LJ-3) from samples collected in 1990. These rocks are nearly identical, fine-grained intergranular plagioclase and pyroxene basalts showing minor alteration of mineral phases in thin section. The flow overlies Karoo Supergroup sediments on the south side of the Nyakamba River in a roadcut for Highway M8 (Figure 1, Figure 4) near the town of Ngabu (162720S, 344345E).

Two aliquots of crushed (0.15 to 0.25 mm) whole rock were analyzed from each of the two samples (Figure 5, see Table 1, Deino et al. 1990, and Sharp et al. 1996, for details of the analytical procedure). Reproducibility of fine details of release patterns in age and geochemical parameters (%40Ar*, Ca/K), and in the derived plateau and integrated ages between aliquots of the same sample, is noteworthy. All spectra released initial fractions (<10% of the cumulative 39Ar release) charged with atmospheric argon and younger than the subsequent plateau, probably reflecting a limited degree of alteration. Nevertheless, broad plateaus occur within the central portion of the release sequence. These plateaus occupy the zone where the Ca/K ratio of the released gasses is less than about 4. In the final 20-30% of the 39Ar release, Ca/K ratios rise sharply and apparent ages fall. This spectral form, characterized by a stair-stepping downward pattern in both the beginning and end of the incremental-heating sequence, is consistent in all four experiments. It can be interpreted as representing recoil implantation of 39Ar released from fine-grained alteration phases into refractory calcic phases. The fine-grained, 39Ar depleted alteration products release their argon early and exhibit artificially old ages and high atmospheric content, while the 39Ar enriched calcic phases retain their argon to the last, exhibiting young ages and high Ca/K ratios. Despite this redistribution of 39Ar locally, older apparent ages in the early release steps are balanced against young ages in the final steps, so that integrated ages are in all cases congruent with the plateau results.

We take the weighted means of the plateau ages of the replicates as the reference age, and obtain 179.4 0.8 for LJ-3 and 168.6 0.8 for LJ-4. These determinations are clearly distinct and present the apparently inverted age relationship of an older age for the crosscutting basalt than the lava flow it lies within. Either alteration has affected the argon systematics more than is apparent from the spectra and the lava flow is relatively too young, or the dike contains a component of excess argon and is relatively too old.

In either case, these ages are relatively young, but broadly compatible with the emplacement history of the Karoo LIP (184 to 179 Ma, Duncan et al. 1997; Jones et al. 2001). Considering the geographic proximity of the Chikwawa basalt relative to recognized Karoo LIP rocks, the stratigraphic position of the Chikwawa basalt atop Karoo Supergroup sediments, and the age of the lava flow and dike at Ngabu, the Chikwawa basalt is taken to be the northern outcrop extent of the Karoo LIP. The age of the Chikwawa basalt is Jurassic, and in the absence of a major unconformity beneath this flow, the Karoo strata of southern Malawi are much younger than the Triassic age assigned by Cairncross (2001, figure 2; Figure 3).

Northern Malawi

The Chiweta Beds of northern Malawi are fluvial in origin. Bones within overbank deposits are often found encased in pedogenic carbonate, similar to the condition of those from the Beaufort Group of South Africa (Smith 1990, 1993). The most detailed stratigraphic work conducted on the Karoo Supergroup of the Mount Chombe-Chiweta area is that by Yemane et al. (1989), Yemane and Kelts (1990), and Yemane (1993). Approximately 300 m of section were measured, all underlying the therapsid-bearing Chiweta Beds, which were excluded from their studies. Yemane et al. (1989) interpreted the environment of deposition to be lacustrine, and while these lacustrine sediments do not preserve bone, they contain pollen (Yemane 1994). Of 20 identified pollen taxa, 15 are unknown after the Permian-Triassic boundary. Eleven of those taxa have records prior to the Late Permian, and four range throughout the Late Permian. Three pollen taxa (Taeniaesporites noviaulensis, Lycopodiacidites pelagius, and Rimaesporites aquilonalis) range only from after the beginning of the Tatarian (late Capitanian or younger fide Gradstein and Ogg 2004) and into the Triassic. Therefore, an independent maximum age limit for the overlying Chiweta Beds of <263 Ma (base of the Tatarian, timescale of Gradstein and Ogg 2004) is set from palynological evidence.

Age of the Chiweta Beds Based on Therapsid Biostratigraphy. Dixey (1926) recognized two main bone producing horizons in the Chiweta Beds, although he acknowledged that bone could be found throughout. The main bone producing horizons correspond to his units 5 and 7. Unit 5 is the Lower Bone Bed (or B1), and unit 7 is the Upper Bone Bed (B2) of Haughton (1926). Our collections were made from the Lower Bone Bed (B1, seen in the foreground in Figure 2) with the possible exception of an articulated but unprepared and unidentified skeleton lacking the skull, which is likely from the Upper Bone Bed (B2).

The upper age limit of the Chiweta Beds can be refined by comparison of the B1 therapsid assemblage with that of the Beaufort Group, the vertebrate biostratigraphy of which is reviewed in Rubidge (1995). Of particular importance is the genus Oudenodon, which we show to occur at Chiweta (Figure 6, Figure 7). King diagnosed the Tribe Oudenodontini (which contained only Oudenodon) as:

Medium-sized to large dicynodonts (skull length ranging from 100 mm to over 300 mm). Teeth lacking in both upper and lower jaws. Postorbitals well separated on skull roof by parietals. Septomaxilla recessed within external naris, lachrymal in some species extends forward above maxilla to posterior margin of naris. Nasal forms boss over naris. Maxilla carries weak caniniform process, with sharp edged posterior crest. Palatal portion of palatine divided into inflated posterior area, and a smooth anterior part that meets the premaxilla. Vomers form short septum in anterior part of interpterygoidal fossa. Ectopterygoid large with palatal exposure, pterygoid does not contact maxilla… (King 1988, p. 85).

The characters observed in the dicynodont skulls designated Malawi Department of Antiquities (Mal) 108 and 129 (Figure 6, Figure 7) from the Chiweta Beds indicate that both belong to the genus Oudenodon. Mal 108 is lacking its posterior portion. Mal 129 is postdepositionally dorsoventrally compressed. Measurements (Table 2) demonstrate that the two specimens are similar in size and represent a single species, but Mal 129 appears more robust.

Oudenodon groups with Tropidostoma, Rhachiocephalus, and their most recent common ancestor within a clade (Cryptodontinae) within the Dicynodontidae (sensu King 1988, 1990; Angielczyk 2001; compare Angielczyk and Kurkin 2003a, 2003b). Members of this clade are diagnosed by having a postcaniniform crest. A less inclusive clade within the Cryptodontinae, diagnosed by parietals exposed in a depression between the postorbitals and a long interpterygoid vacuity, includes Tropidostoma and Oudenodon, but not Rhachiocephalus (Angielczyk 2001). Of these taxa, teeth are present in Tropidostoma but lacking in Oudenodon.

Haughton (1926) reported two species of Dicynodon distinguished by size in the material sent him by Dixey from Chiweta. Both are tuskless. The illustrations in Haughton (1926) of the smaller species, designated by him Dicynodon sp. A, show the parietals exposed between the postorbitals, indicating that it should be referred to Oudenodon. The larger species was represented by a fragmentary skull, a humerus, and parts of other postcranial bones, which Haughton designated Dicynodon sp. cf. D. grandis. The referral of those specimens to D. grandis, a species named by Haughton (1917), was based on size alone. Keyser (1975), in his review of tuskless anomodonts, referred D. grandis to the genus Oudenodon, consolidating the genus and synonymizing a large number of species into three: O. baini, the genotype, most abundant, and widespread species; O. grandis, a rare but large species whose only diagnostic character other than size is having the parietal foramen lie in a depression on the dorsal surface of the skull; and O. luangwaensis, known from Zambia and defined on the basis of “…the great width of the zygomatic arches, giving the skulls their very characteristic heart shape” (Keyser 1975, p. 57).

Measurements of Mal 108 and Mal 129 (Table 2) fall within the range of Oudenodon baini presented by Keyser (1975, p. 39-40) indicating that those specimens do not pertain to O. grandis. In addition, a low boss lies near the parietal foramen of Mal 108, which if a reliable character, also indicates that Mal 108 is not O. grandis. Mal 129 lacks the great width and therefore the “very characteristic heart shape” of the skull of O. luangwaensis, even though this character in Mal 129 is accentuated by dorsoventral flattening. Haughton’s Dicynodon sp. A is smaller than either Mal 108 or Mal 129, but still within the range of O. baini. We therefore identify Mal 108, 129, and Dicynodon sp. A of Haughton (1926) as Oudenodon baini.

Haughton (1926) also reported teeth and jaw fragments of Endothiodon cf. E. bathystoma. He questioned the level from which they were obtained but considered them most likely to come from the Lower Bone Bed (B1), and therefore less likely to have come from the Upper Bone Bed (B2). We have not recorded Endothiodon in the collections we made from B1. Regardless, Endothiodon is a distinctive genus (Cox 1964) and its occurrence in the Chiweta Beds has biostratigraphic significance. Haughton (1926, p. 82), in evaluating the age of the Chiweta fauna, states, “…had [the assemblage] been found in South Africa, [it] would unhesitatingly have been assigned to the base of the Cistecephalus zone or top of the Endothiodon zone of the Lower Beaufort Beds.”

Numerous revisions to the biostratigraphy of the Beaufort Group have been made since Haughton’s work, but his observations are on the mark. In the latest comprehensive revision, Endothiodon is a component of the upper Pristerognathus Assemblage Zone (Smith and Keyser 1995a), the Tropidostoma Assemblage Zone (Smith and Keyser 1995b), and at least the lower portion of the Cistecephalus Assemblage Zone (Smith and Keyser, 1995c). Oudenodon is found throughout the Cistecephalus and the overlying Dicynodon assemblage zones (Kitching 1995), becoming extinct near the Permian-Triassic boundary. Gorgonops is known from the Tropidostoma and most of the Cistecephalus Assemblage Zones (Smith and Keyser 1995b). Thus, the overlapping ranges of Endothiodon and Oudenodon, consistent with the presence of Gorgonops? dixeyi and Aelurognathus, argue for a lower Cistecephalus Assemblage Zone placement for the Chiweta Beds. The youngest likely age falls just below the midpoint of the Tatarian Stage (Hancox and Rubidge 2001), or following Gradstein and Ogg (2004), about at the middle of the Wuchiapingian Stage, or approximately 256 to 258 Ma. We take that range as a best estimate for the age of the fauna.

In conjunction with his phylogenetic analyses of dicynodonts, Angielczyk (2001, 2002); Angielczk and Kurkin (2003a, 2003b) fitted preferred cladograms to the stratigraphic record of the taxa, thereby demonstrating the extent of ghost lineages associated with most parsimonious trees. The most significant implication for this study affects the overlap between Endothiodon and Oudenodon, a co-occurrence that limits the correlation of the Chiweta Beds with the Cistecephalus Assemblage Zone, but eliminates a correlation with the uppermost portion of that same zone. The downward ghost extension of Oudenodon, it might be argued, could affect the correlation of the Chiweta Beds with the Karoo Basin of South Africa by allowing a correlation with the underlying Tropidostoma Assemblage Zone. However, the biostratigraphic ranges of the robust suite of pollen taxa (Yemane 1994) limit the maximum age of the therapsid assemblage and mitigate concern over downward extension of the Oudenodon range.