488 
PACIFIC SCIENCE, Vol. XX, October 1966 
the upper part of the massive layer of the sandy 
(clay) loam, regardless of whether or not the 
surface shows terracettes. The density of the 
"corona” around the base of the tussocks is less 
than that in texture-contrast soil. Rather many 
roots penetrate deeper, into the layer with 
weathered rock fragments. Deepest roots are 
observed at 1.3 m. 
Consequently, rooting patterns except for 
being somewhat more diffuse in the regosols are 
not very different in various kinds of soil, and 
only a firm, massive, sandy loam layer as en- 
countered in certain texture-contrast soil pro- 
files seems to be unfavourable for deeper root 
development. 
Ecological Significance of the Results 
The investigation revealed that correlations 
between vegetation and soil conditions seem to 
have only a localised validity. For instance, the 
distribution pattern of Eucalyptus alba encoun- 
tered on transect 2, viz. its restriction to texture- 
contrast soils, is not consistent with that on 
transect 1, where E. alba occurs over the whole 
transect. Also, a hypothesis based on observa- 
tions at transect 1, that the boundary between 
the Themeda-Capillipedium and the Themeda- 
Sehima ground cover groups coincides with the 
boundary between texture-contrast soil and 
regosol, would appear untenable when applied 
to transect 2. On the other hand, black soils 
carry only the Themeda-Heteropogon group, 
which however also extends over other soils. 
Some conditions find ready expression in the 
vegetation (e.g., steep, rocky slopes are covered 
with a Sehima-Heteropogen ground cover, and 
disturbed soils with a T heme da-Sorghum 
ground cover), but generally the determining 
factors remain unknown. The presence of an 
algal cover, great local variations in grass cover, 
and differences in certain soil characteristics 
mentioned in the discussion on the ground 
cover groups, suggest a great influence of soil 
moisture, whilst terracettes and local concen- 
trations of grit and gravel point to slope wash 
as an important factor. On the other hand, dif- 
ferences in moisture regime were hardly, if at 
all, expressed by rooting patterns as the root 
systems of the grasses as well as those of E. 
papuana appeared to be similar for different 
soils, whilst E. alba showed a considerable 
variation within one soil type. 
Observations over a period of years, also 
during the wet season, and on a regional basis 
could lead to more conclusive results, but an 
investigation into the influences of burning will 
probably reveal fire to be the overriding factor 
of the environment. 
COMPARISON WITH AUSTRALIAN VEGETATION 
Savannahs comparable in physiognomy and 
to a certain extent also in floristic composition 
occur in the higher rainfall areas of tropical 
and subtropical Australia, but relevant litera- 
ture is scarce and is concerned mainly with the 
characteristics of the ground cover for use as 
natural pasture. 
At Katherine, N.T., Arndt and Norman 
(1959) studied a savannah vegetation of a 
richer floristic composition but which included 
most of the species encountered at Port 
Moresby. Predominant in the ground cover are 
Sorghum plumosum, Theme da australis, and 
Chrysopogon fallax. The dry matter yield for 
unburnt areas of a lightly grazed pasture was 
about 110 g/m 2 at the end of the dry season 
(October), and rose to about 290 g/m 2 at the 
end of the wet (April-May). For areas burnt 
at the end of the dry season the dry matter 
yield at the end of the wet season was only 
about 140 g/m 2 . 
Shaw and Bisset (1955) have given figures 
for subtropical Queensland, which can be as 
low as 60 g/m 2 for dry season yield and as 
high as 425 g/m 2 for yields at the end of the 
wet season. 
The figures obtained for the Themeda aus- 
tralis-Eucalyptus savannah in Papua are much 
higher than those mentioned above. Oven-dry- 
ing of the samples instead of air-drying would 
have resulted in figures about 10 to 15% lower, 
which gives a figure of, say, 530 g/m 2 , which 
still is much higher than the figures for Kather- 
ine and Queensland. Because our observations 
are from only one season, it is impossible to say 
how near they are to the average. On the other 
hand calculations by Fitzpatrick (1965) about 
the periods of useful pasture growth demon- 
strate that the growing season at Jackson’s 
