Despite their laudable efforts to date the channel iron deposits (CID) of the Pilbara of Western Australia (WA), Heim et al., (2006) need to reassess their sampling and the statistical strategy behind their conclusions. Many details they report conflict with data from a major CSIRO-industry research program on CID (Morris et al., 1993; Ramanaidou et al., 2003) and other research (e.g., Stone et al., 2003).
Miocene fluvial goethite/hematite CID of the WA Pilbara range from gravelly mudstones and ooid-rich granular rocks with abundant ferruginized wood fragments, to intraformational pebble, cobble, and occasional boulder conglomerates. Clays and non-ore polymictic basal and marginal conglomerates are also present. The largely pedogenically derived iron-rich sediments occupy numerous meandering paleochannels in a mature surface on Precambrian granitoids, volcanics, metasediments, banded iron formation, and Paleogene valley fill. The porous, consolidated, ooid-rich fine gravels and their intraformational conglomerates from the Yandi and Robe deposits are a >7 billion ton resource of export iron ore.
Heim et al. (2006) make a number of erroneous assertions. To begin, our CID reports do not describe the goethitized cortex of the ooids and pisoids as “vitreous” (as reported by Heim et al. on p. 174). We never used “botryoidal,” nor did we describe matrix as “late-stage cement,” though some post-depositional solution voids in the matrix contain late-stage infill. Yet Heim et al. attribute these and other incorrect statements to Ramanaidou et al. (2003). “Botryoidal” is Greek for “bunches of grapes,” a texture rare in CID ore. Colloform texture, however, is common in late-stage goethite which fills or lines some solution voids in the matrix, and at times appears vitreous. Goethite infill can appear very similar to original matrix in reflected light—see Heim et al.'s Figure A1 (DR1) in the GSA Data Repository 2006032. They wrongly regard this normally minor component as “the ore matrix” (p. 174).
Though shown as “typical Yandi ore,” Figure A1b (DR2006032) is from non-ore, polymictic marginal conglomerate, and Figure A1a, although ore, is from intraformational conglomerate. Original channelfill contains a “large percentage…of Fe-rich clasts” including, the authors claim, “partially weathered, banded iron-formation” (p. 175). This, in fact, is very rare in CID ore, though common in non-ore marginal conglomerates. CID are ferruginous sediments, not “ferruginized” or “Fe-metasomatized ” as claimed (p. 173); however, it is likely some wood/charcoal was goethitized in the channels.
Sampling CID goethite for dating. The use of “botryoidal” by Heim et al. (p. 174) to describe their sampling sources implies masses of pure secondary goethite, which is misleading. So is the 1.5 mm sample spot in Heim et al.'s Figure A1c (DR2006032), since they used “4 mm diameter” drill cores (p. 174), and both are incompatible with the >10 mm fragment of Figures A1e and A1f. Even 1.5 mm cores could include cortex from adjacent ooids as well as original matrix. From crushed cores, Heim et al. selected 0.1–3 mm fragments “devoid of detrital phases,” but said, “None of the samples analyzed in duplicate yields statistically reproducible results” (p. 174). The authors offered three possible explanations, ignoring the most likely. Despite apparently rigorous inspection of duplicates, and without similar data from the analyzed samples, Heim et al. cannot be sure they avoided look-alike primary matrix or ooid cortex. Both are older than the infill goethite they were trying to date, probably by millions of years, and could include even older U/Th in former soil components. Note the wide range in group ages in Heim et al's Figure 1.
The validity of the goethite dating is critically dependent upon what was actually sampled. We suggest Heim et al. included unrecognized original matrix and cortex with the infill goethite.
Statistical forcing of data. Heim et al. used progressively larger age corrections (their Fig. 2), to improve the statistical fit for their conclusions. For example, the 20% correction that they used as a He loss “worst-case scenario” (p. 175), increased the calculated ages of the samples by 30%–32.8%. These ages were used to argue that “ca. 36 Ma” valley fill was ferruginized and cemented from the top down as water levels fell, reaching completion in the Pliocene or later (p. 176). Their extrapolated ~28–36 Ma dates for the start of “ferruginization” assumed an “original channel surface” at 505 m (p. 175), but the overlying Iowa Eastern Member CID is largely absent from their deposit (their Fig. 1). Thus, the original surface was at least 15 m higher, and likely much more, assuming continuous erosion since Heim et al.'s “ca. 36 Ma” date. Extrapolating to this more likely surface could thus support an unrealistic Paleocene or even Cretaceous start for CID sedimentation.
Contemporary deposition of CID matrix. Many features, including partial alluvial matrix and matrix-supported samples with up to 75% matrix, contradict the top-down infill model. Episodic post-consolidation partial leaching of matrix with refilling of some voids is demonstrated by various generations of infill goethite, silica, and oxidized siderite. Scours filled by later CID and well-preserved bedding surfaces confirm lithification occurred soon after sedimentation (Stone et al., 2003). The scours at Yandi are 2–10 m deep, to 20 m wide and over 200 m long, with typically steep, sometimes near-vertical margins. In unconsolidated granule material, such margins would soon slump into the scours to disappear from the CID record. Intraformational conglomerate horizons comprising well-rounded pebble- and cobble-sized clasts of granular CID confirm a contemporary matrix.
Downward younging of CID “infill” goethite. The altered Lower CID basal zone (Heim et al.'s Fig. 1), with its large ochre patches and cavities, is evidence of major leaching due to prolonged basal channel water flow in the past, and the current high water table confirms changing water levels. The presence of unrecognized contemporary CID matrix, altered variably by later episodic solution and by lesser infill events as flow levels fell erratically, as well as by imprecise sampling, is a more valid explanation for the wide-ranging group dates and different younging trends than the top-down infill model of Heim et al.