After flowing sluggishly along on the surface of the table-land, the Mbutu-mbutu River arrives suddenly at the edge of a line of cliffs of volcanic agglomerate, that here form the southern border of the plateau, and with a volume 30 to 40 feet across, it plunges down into the ravine 150 feet below. As shown in the view from the gorge below, there is a break in the middle of the descent. These falls, however, are not easily accessible. They are best approached by proceeding from Wainunu to Ndawathumi and thence up the gorge of the Mbutu-mbutu River.

The surface of the plateau of Na Savu is densely wooded. In places it is marshy, and here thrives the Giant Sedge (Scirpodendron costatum). The Makita tree (Parinarium laurinum) also flourishes in the wet districts; and in the drier localities occur the Ndakua (Dammara vitiensis) and the Ndamanu (Calophyllum-burmanni) together with a palm of the genus Veitchia. Here on this level watershed between the basins of the Wainunu and Sarawanga rivers, the sluggish streams flow aimlessly along in but slightly eroded channels; and it is not always possible to determine the side of the island to which they ultimately direct their course. In their beds are pebbles and irregularly formed concretions of an impure reddish flint which I have described on page [354]. On the north and south sides the table-land is much excavated by the tributaries of the Sarawanga and Wainunu rivers. On the west where it meets the foot of the Seatura slope portions of columns of basaltic rocks appear on the surface, and deep gorges are worn by the large streams descending from the mountain. On the east towards Nuku-ni-tambua and Tambu-lotu, the surface is also much cut up. The preservation of this table-land in a region, where the denuding agencies are very active in their operations all around it, is to be attributed to its being a level watershed, where the head-waters of the Wainunu and Sarawanga rivers in part take their rise but have little or no eroding power.

It is not easy to obtain a good general view of the district of the falls on account of the dense forest-growth. When making the traverse from Tambu-lotu to Ndawa-thumi, it is observed that there is here a singular hollow, about half a mile in length, which receives the falls at the western end. The river crosses this hollow and is at once received into the gorge below, but there is no stream to explain the origin of the cavity. On its north side the cliffs of agglomerate rise to a height of 150 to 200 feet from their base, but on the south the sides are much lower. Here there seem to be the remains of the crater of the ancient vent from which all the tuffs and agglomerates of the district were derived. We must look for their origin in the vicinity, and the only evidence of a crateral cavity is this streamless hollow extending east from the falls of Na Savu.

With reference to the basic tuffs and agglomerates of this plateau it may be observed that they cover the massive basic rocks and are probably not over 100 or 150 feet in maximum thickness. They are well exposed where the streams cut into the borders of the plateau. The tuffs are sometimes bedded and slightly inclined, and they may be fine or coarse grained. They are more or less palagonitised hyalomelane-tuffs, being composed mainly of fragments of a basic glass, often finely vesicular and even fibrillar, the vacuoles being filled with different materials, whilst the palagonitisation is well advanced. Sometimes they have a brecciated appearance, and in that case when the alteration of the basic glass is very extensive we find angular fragments, 1 to 2 inches across, of a greenish palagonite imbedded in a pale matrix of palagonitic debris, the whole rock having a soapy feel and a steatitic appearance. This is well shown on the sides of the stream-course at Ndawathumi which lies at the border of the table-land. These tuffs effervesce but slightly with an acid.

The basic agglomerate is displayed in the face of the falls and in the gorges. The blocks are as a rule composed of semi-vitreous basaltic andesites of varying type, showing no olivine and containing a fair amount of smoky glass in the groundmass. At times they are scoriaceous and display amygdules of calcite or a zeolite. In places the rock shows large phenocrysts of plagioclase and a semi-ophitic groundmass, when it is referred to the porphyritic group of genus 9 of the augite-class. In a few of the scoriaceous blocks the augite of the groundmass is for the most part prismatic and rarely granular (genus 5).

The massive rocks underlying the agglomerates in the vicinity of Na Savu are aphanitic augite-andesites, differing in important characters from the rocks of the agglomerates. They probably represent ancient lava flows of the Na Savu vent. They are compact (sp. gr. 2·72-2·76), and display a groundmass formed of a felt of felspar-lathes, averaging ·05 or ·06 mm. only in length, and in flow-arrangement. That occurring just below the falls is almost aphanitic, but is referred to genus 13, species A, sub-species a, of the augite-andesites. The rock from the gorge below is of the same character, but on account of its opaque plagioclase phenocrysts it is referred to genus 14, and is described on p. [279].

In one place on the plateau a tuff-agglomerate is penetrated by veins, a few inches thick, formed apparently of a finely brecciated tuff of basic glass fragments in a palagonitic matrix. It is, however, pointed out on p. [340] that they were originally veins of basaltic glass which have been subjected to crushing, and that the palagonite has since been produced.

In concluding this description of the table-land of Na Savu, it may be inferred that the source of its basic tuffs and agglomerates is to be found in the same locality; and probably the original vent is now represented by the hollow extending eastward from the falls. With the exception of a large block of silicified coral found in the vicinity of Ndawathumi and of the impure flints of the surface of the plateau, which are described on pages [354], &c., no direct testimony of its submarine origin offered itself to me. The palagonitic characters of the tuffs afford, however, indirect evidence in this connection; and indeed the occurrence of submarine tuffs and limestones in the vicinity of Tembenindio on its lower northern slopes (see page [131]), and the existence at elevations of several hundred feet above the sea of fossiliferous tuffs and clays in the Wainunu and Ndrandramea districts to the eastward, afford strong presumptive evidence that the tuffs and agglomerates of the table-land were deposited under the sea, and I may add in a period subsequent to that of the formation of the great basaltic flows of Seatura and Wainunu.

CHAPTER VI
DESCRIPTION OF THE GEOLOGICAL AND GENERAL PHYSICAL FEATURES
(continued)

The Basaltic Plateau of Wainunu.—This table-land extends for a distance of seven miles from the base of the Ndrandramea mountains in the heart of the island, where it is elevated 1,100 to 1,200 feet above the sea, to the valley immediately north of the hill of Ulu-i-ndali, where within a short distance of its termination it still retains a height of 700 to 800 feet. Limited on the west by the valley of the Wainunu River and on the east by that of the Yanawai River, its breadth varies usually between four or five miles. It is best seen in profile when viewed from the south-west on the western shores of Wainunu Bay, between Korolevu and Nasawana, when it presents itself to the eye as a table-land, descending with a very gradual slope from the interior towards the coast. From such a point of view the two great basaltic slopes of Seatura and Wainunu may be seen together, the former descending eastward to the Wainunu valley at an angle of 3 or 4 degrees, the latter descending at right angles to it to the southward with a similar small gradient of 2 or 3 degrees.