Table 4.

Summary of anthracological and micromorphological results

	Type A	Type B	Type C
	Flat combustion features	Dug combustion features	Palimpsest
Combustion feature description	Four different microfacies were identified in type A combustion features (Fig. 7, see Appendix 1.0), with from bottom to top: (1) Dark brown layer (concave shapes visible in stratigraphy) composed of geogenic sands, with a high proportion of microcharcoal and carbonised vegetal particles (Fig. 7c). (2) Orange-pinkish layer composed of geogenic sands and ash particles (Fig. 7b). (3) Layer with charcoal chunks embedded within the geogenic sands (Fig. 7a). (4) Whitish layer (Fig. 7c) mostly composed of ash particles (pseudomorphs of plant calcium oxalate) and phytoliths that are often still in anatomic connection (articulated).	Type B combustion features are characterised by the digging of a pit or depression where the fire was lit and frequently reused. These dug features sometimes cut through the flat combustion features immediately below, disturbing the orientation of particles (Fig. 8b). Type B combustion features contain large fragments of charcoal embedded within a fine matrix of geogenic sands and/or random oriented ash particles (Fig. 8a).	Type C features are accumulations composed mainly of by-products of combustion (predominantly ash and charcoal), along with a mix of non-burnt vegetal parts and a minor proportion of geogenic sands (Fig. 9). The high proportion of combustion residues gives the Holocene deposits its grey colour (Figs. 2 and 3). Under the microscope, the microfacies show a very bright calcitic crystallic b-fabric as a result of the high proportion of ash particles (pseudomorphs of plant tissues) (Fig. 9b, d). These particles are found both in anatomic connection (Fig. 9c, d) and disturbed (Fig. 9a, b) which indicates some mixing in this level.
Hypothesised function	Small, open-air, single-use hearths which were lit directly onto the ground surface and used for heating, lighting, or cooking purposes during short term visits to the site.	Earth or ground ovens which were constructed by digging a pit or depression into the substrate. A fire was lit within the depression and subsequently buried during a hypothesised underground cooking process.	A palimpsest of numerous type A and B features coupled with an increase of combustion residues resulting from maintenance activities such as reworking (cleaning rake-out of hearths but also human and animal trampling and turbation, see Fig. 4).
Economy of wood	The charcoal within the type A combustion features is mostly completely combusted with the exception of the F3-A and F9-A assemblages. F9-A is the only anthracological sample from Riwi Cave which was not dominated by bloodwood/eucalypt; the dominance of Erythrophleum sp., coupled with the low diversity of F9-A taxa, is indicative of a single firing episode. Taxa were collected predominantly from the valley floor as per the associated scattered Pleistocene contexts.	The taxonomic diversity of the type B dug features is similar to the associated scattered contexts and might indicate re-use of the combustion features. The dominance of bloodwood savanna taxa indicates that firewood was collected predominantly from the valley floor, with a minor presence of dry rainforest taxa, which are associated with the limestone outcrop.	The savanna taxa of the valley floor continue to be well represented alongside a higher proportion of dry rainforest taxa, which indicates a potentially more extensive use of the surrounding landscape and its various ecological niches.
Inference of hunter gatherer strategies	Occupation was intermittent from 47 to 34 ka, with enough time between site visits to allow for these flat, open-air, single-use hearths to be covered by natural, rapid aeolian deposition.	More frequent and/or longer occupation at the site during this time (c. 34 ka).	Frequent occupation of longer duration during the Holocene sequence producing a palimpsest of archaeological deposition, which forms this type C secondary context.