minimum of 40% Gal/una cover formed the major choice criterion. At each site a 1 0 metre by 1 0 metre area was chosen and pegged. The species present, within the pegged areas, at each site were listed, and the percentage cover of the 3 or 4 most prominent cover species was estimated by averaging values of three independent observers.
Exploratory data analysis was with 'Minitab' and comparative analysis with Systat (Systat, inc. 1 989). The data were used to form a presence/absence matrix on which ordinations, and cluster analysis were performed. A cluster analysis (average linkage, Euclidean distance) shows the level of similarity between paired sites to confirm the pairing of sites initially made by eye in the field. Principal component analysis (PCA) was applied to a reduced matrix involving just the uninvaded and Gal/una only sites to better represent the distinctive characters of each native vegetation type (such as successional state). The ordination displays the sites onto which a vector plot of the component loadings is super-imposed. A table (Table 2.4) is supplied listing species that comprise each vector.
Resou rce m o d ifications
Biomass
Biomass samples were collected from the five Gal/una invaded sites, the Gal/una only site, and three other areas called: 'Pukeonake' (the base of Pukeonake on the south side of the road, Fig. 2.2), 'Mangatepopo-a', and 'Mangatepopo-b' (approximately 1 00 metres south of the Pukeonake site, and themselves separated by 50 metres in a southerly direction). All these sites are within the original planting area of Gal/una and should represent the biomass of Gal/una in its dominant state. Sampling was done in August (1 993) at the end of winter when the standing crop was most likely at its annual minimum. All Gal/una material above ground was removed from a 50 cm by 50 cm square with projected vertical sides. Four samples were taken haphazardly from each site. The percentage cover of Gal/una in a 1 Ox1 0 m area at each site was noted. The samples were dried in a Cuddon vacuum oven (-600C, -1 5 atm) at HortResearch, Palmerston North for 1 4 days with test samples being removed on days 8, 1 0 and 1 3 to determine if weights were stable.
The data were log transformed and analysed using ANOVA testing for differences between sites, on which a Bonferroni means comparison was done, this being a robust multiple range test (Systat manual). The Bonferroni results are presented as line diagrams joining means which are not statistically different (at a 5% confidence level).
Mineral content
To assess the mineral content of above ground parts of a Gal/una plant, three individual plants from the Mangatepopo area were removed, again from the area of oldest and most mature Gal/una. The plants were separated into new foliage (this year's growth), old foliage (> 1 year), and woody tissue, and
then f reeze dried. Six haphazardly c hosen samples of -3 g of eac h of the three categories were p laced into c rucibles and weighed. The samples were then incinerated at 500 °C for 1 2 hours, and reweighed. The ash remaining was considered the mineral portion, the burnt off portion the carbon and nitrogen compounds.
N i t r o g e n
Soi ls and plant material were taken for analysis of n itrogen content in w inter ( 1 7/8/1 992) from the tussock/Cal/una
,
Dracophylfum IGaffuna , manukaiGalluna , flax/Cal/una , and Gal/una only sites (Table 2.2). Gal/una samples were also collected f ro m the same sites once in summer (6/1 /1 993) in o rder to establish seasonal differences between nitrogen levels in Gal/una. F rom each site the top 1 0 c m of foliage (ca. 1 0 g rams wet weight) of five Gal/una , or five native p lants were taken, i.e. five replicates per s ite per plant type. The samples were f reeze dried (72 hours), g round to powder (,Glen Creston' seedmill) , and analysed by the Kjeldhal method (an acid extraction method) by the Soil Science Department, Massey Univers ity. Soil samples were collected, oven dried at 600C for four days, g round, and analysed for nitrogen by the above method.
Plant a rchitecture
T he measurement of plant architecture is very new, and there is, as yet, I believe, no
comprehensive method. Commonly used is the point-intersection method , or some derivation (Brown, 1 99 1 ) but this best works in g rasslands (but see Dickinson and Mark 1 992). I have surveyed features of bushes considered relevant to insect use , and then constructed an index of architectural complexity.
T he survey was on ten haphazardly c hosen plants of each of eight plant types in November 1 994. Tussock plants (Ghionochloa rubra), the colonising form of manuka « 1 metre tall), and the isolated bush, dwarf type Gal/una (short Gal/una < 50 cm) common on the eastern side of the Park were measured on the Desert Road approximately 2 k m south of the Waihohonu track. Dracophyllum,
Celmisia, the dense tall
(
> 1 metre) form of manuka, and the tal l rangy, dense, interlinked form of Gaffuna(tall Ga{{una), found predominantly north of Hauhaungatahi towards the northern Park boundary and west of the mountain axis were measured fro m the end of the Mangatepopo road. F lax ( Phormium tenaXj was
measured from around E rua (Fig. 2.2).
In the f ield, 1 0 p lants of the type required were haphazardly cho sen and surveyed. T he measurement of features was divided into 2 sections: 'objective' measures; and 'subjective' measures , as described below. The 'objective' measures consisted of 1 4 measures g rouped into 4 categories: plant structu re, ste m and branch attributes, leaf attributes , and bus h dimensions (height plus bush radius). The 'subjective' measures had two categories, base complexity, and lichen cover. My own 'F iddler' score is based on the premise that the measures of the survey miss nuances an observer may notice about plant structu re in relation to invertebrate use.
Architectural index
Objective measures (ct. appendix 1 a for an example score sheet): 1 . Structure, a whole plant aspect.
3 7
From observation it was determined if a plant was separate, touching, or interlinked with neighbours of the same species, and, if touching, the number of individuals it was touching was recorded. A plant's own branch inter-linkage, i.e. branches crossing within an individual plant, was scored through estimation by eye and consisted of the categories - no branches crossed, 1 -1 0%, 1 0-50%, 50-75%, 75%+ crossed (scored as 0-4). General branch orientation, mostly vertical, mostly horizontal, or some of both was recorded (1 -3).
2. Stem .
The number of main stems was counted. Then the number of branches from a haphazardly selected main stem in the categories 1 0, 20, 30, 40, as divisions of branching o rder, were counted, and the angles of the branches from their immediate superior recorded.
3. Height and radius.
Heig ht to highest point, and radius at the roundest point were measured and recorded for each bush.
4. Leaf.
Leaf attributes involved where they were found, i.e. on ultimate, penultimate or on many branches (1 -3) ; their spacing in relation to each other, i.e. single or packed as a unit ( 1 -2); and then the spacing
relationship of these units on the branch, e.g. a few units spaced, to lots of units spaced, to lots clustered (1 -6).
Subjective measures: 1 . Litter.
Litter and debris, live or dead, at the base of a plant was scored from 1 (meaning no litter) to 1 0 and described the amount of litter.
2. Lichen.
The amount of lichen hanging in the foliage and between stems was scored from 1 to 1 0.
To form the index the raw values of a category were transformed to a percentage of the largest score for that category. This reduced the weightings of larger categories, given the different scales used. Then all the category scores for each plant were summed to g ive a pre-index value and the mean value for each plant type attained. These values were again divided by the highest mean-index score, giving a range of 0 to 1 for the index.
'Fiddler' score
My own score (1- 1 0) was purely subjective, but i magines that the survey method does not account for the intuitive measure that an observer has. For example Dracophyl/um has many thin leaves arranged horizontally and packed densely. It looks complex and sheltering; however the landing sites are m in imal (due to its horizontal nature) and the foliage is 'harsh'. T hus it is only 'useful' for crawling insects suc h as beetles, spiders, and wood scale. The score ranks the p lant species based u pon knowledge of the invertebrates likely to use each plant and their probable requirements (see Appendix 1 a for more detail).
Analysis
Both indices allowed a ranking of the plants, in terms of structure, from most 'useful' to
invertebrates to least. A PCA o rdination analysis was used to s eparate the plant species by their indices, and g roupings circ led and identified by eye with s hading. A tabl e of components illustrates w hat features are prominent in which plant, w hile multifactorial ANOVA analysis and Bonferroni range tests were used to determine statistical ly s ignificant differences between plants index components. Resource
arrangement was illustrated for all the plant community types explored through pyramid diagrams of base, stem, structure, and leaf , depicted by size of bar. The size of a bar of a pyramid was calculated by
multiply ing the index component score by the percent cover of the plant species involved in that com munity .