E ø Endangered, V ø Vulnerable, R ø Rare, X ø Presumed extinct, K- Poorly known but suspected as any of the above, i ø Conservation inadequate, a- Conservation adequate

3.3 Vegetation Community Classification

Variations between site groupings revealed few differences between the Bray Curtis and Kulzcynski method. While sites demonstrated a propensity to move within groups they did not shift dramatically to align with different groups within the dendrogram. The Kulzcynski method was chosen as it presented a tighter alignment of wetland communities, although either method could have equally been used.

A large dendrogram displaying the relationships and potential groups for all 2360 sites of the greater Hunter region was produced. Groupings of sites were analysed around the 0.9 level of dissimilarity or 70 groups. Groups emerged which bound homogenous attributes of geology, position in landscape and forest structure. While not perfectly defined, groups described sites occurring on Hawkesbury and Narrabeen Sandstones, Permian Sediments, Basalts, Quaternary Sediments, Sands and Alluvium.

Given the size of the dendrogram (41 pages) it difficult to display the major points of division in the creation of the communities. However, patterns presented in the dendrogram, relevant to the LHCCREMS area, can be described. As an example a large grouping of sites delineated the woodland to dry open forest of the Hunter Valley Floor and Foothills of the Watagan and Port Stephens Ranges. These communities were dominated by Spotted Gum, Corymbia maculata, and various ironbarks, including Eucalyptus fibrosa, E. crebra and E. siderophloia. Further interpretation of this cluster of sites indicated significant variation, demonstrated by sub groups at finer dissimilarity levels. These sub groups reflected distinct changes in ironbark species and understorey characteristics. As an example the spotted gum forests near Cessnock are codominant with E. punctata and E. fibrosa with a shrubby undertorey of Melaleuca nodosa and Davisiea ulicifolia. This contrasts with the Spotted Gum Forest along the foothills of the Watagan Range where codominant tree species are E. siderophloia supported by E. acmenoides and E. umbra. The forest supported by higher rainfall where the mid storey is strongly influenced by Allocasuarina torulosa. At this point these different clusters of sites were marked as vegetation communities.

The process of delineation and interpretation continued for all broad clusters. These clusters mark combinations of sites that describe:

• Spotted Gum-Ironbark Woodlands and Dry Open Forests
• Dry Rainforest and Tall-Open Forest of the Hunter Valley Floor
• Dry Alluvial Woodlands
• Sheltered Narabeen and Hawkesbury Sandstone Forests
• Coastal Wet Sclerophyll and Rainforests
• Coastal Swamp and Wetland Forests and Heaths
• Estuarine Complex of Mangroves and Salt Marsh
• Exposed Coastal Narabeen Sandstone Forests
• Exposed Hawkesbury Sandstone Forests and Woodlands
• Coastal Sandmass Open Forests, Woodlands and Heaths
• Coastal Plains Woodlands on Narabeen and Permian Geologies
• Tertiary sands woodland; Melaleuca scrubs and poorly drained wet heaths.

Fifty-five vegetation communities were derived from these broad groupings relevant to the study area. These communities represent the combination of raw quantitative classification, previous full floristic analyses and field experience. Realignment of sites using physical characteristics such as geology refined groups to improve Community definition in some instances.

Analyses of full floristic data can blur the obvious structural variations that occur in some communities. This is particularly the case for rainforest/wet sclerophyll communities and heath/woodland communities. As Binns (1996) notes the distinction between wet sclerophyll forest and rainforest is a structural one based on the abundance of pyrophytic canopy vegetation. For the purposes of modelling, Map Units 1, 36,28,47, 34, 43 and 3 were mapped as complexes and then structurally separated as sub units using available air photo interpretation. As floristic units they maintain the same combination of species as the parent unit.

Appendix D provides a detailed floristic and structural overview for each Community. A brief summary is also given.

3.4 The Vegetation Model

3.4.1 Mapping rules
The mapping of the vegetation communities was constructed across the greater Hunter region. While parts of the mapping of the Greater Hunter Region have not been finalised, the mapping for the LHCCREMS study area has been completed and extracted as a stand alone map.

All rule sets developed in the decision tree model relied on parent material as the first predictor. Significant splits were found between communities occurring on the nine class parent material layer, aligning communities on basalts, alluviums, sands, Hawkesbury sandstones, the Narrabeen geologies and coarse and fine sedimentary substrates. New branches of the decision tree grew from each of these nodes. The next level of splitting generally used climatic layers, ruggedness (describing the variation in topography) layers or individual soil landscapes. Soil landscapes were used earlier in the decision tree following splits on parent material where individual soil landscapes were well sampled. Fine scale discrimination of communities used layers such as wetness, topographic position and solar radiation generally at lower levels of statistical significance as fewer sites are available at lower levels in the tree. In some cases expert knowledge forced splits in the data where none were otherwise apparent.

By way of example, Vegetation Map Unit 1 (Coastal Wet Gully Forest) occurs on three different parent materials. The mapping rules used to describe its distribution will differ for each parent material. Its occurrence on Narrabeen sandstone follows derived rules set out below. Annotations are made alongside the rule set to explain:

Rainfall is greater than 1110mm per year; and [Indicating that the Community is restricted to high rainfall areas only]

Wetness is greater than 195; and, [indicating that the Community is restricted to gullies in these high rainfall areas]

Roughness 900 is greater than 45; and [indicating that it prefers deeper more sheltered gully lines within all the gullies of the high rainfall areas]

Temperature is greater than 21.5 C [indicating that proximity to the warm coastal areas is important for the distribution of the Community]

In this way the environmental characteristics of the sites are used to model their distribution. In total over 350 rules were developed to spatially define the distribution of each Community. Several rule sets were explored within the time available, each fine tuning the mapped output. The amalgamations of all rules were used to generate the coverage of the pre 1750 distribution.

The modelling approach worked best in dissected terrain, where data layers are able to delineate fine scale patterns influencing the presence of different communities (typically these are things such as gullies; exposed slopes, ridges etc). On flat terrain, variables are often not at sufficient resolution to always successfully separate communities. The Tomago sand beds are a good example where changes in elevation are very subtle, and drainage features difficult to detect. Sites may indicate the presence of poorer drainage, however such features remained could not be presented as individual units in this environment. In this environment some drainage lines are not mapped on 1:25 000 topographic maps.

Similarly, proximity of the water table to the surface had a strong influence in some areas in the Hunter Valley Floor and Wyong Coastal Plain. Correlations were found with other variables (elevation and proximity to streams) in some instances, however, some areas are likely to remain unmapped.

3.5 Comparison of Derived Communities with other classifications

3.5.1 Within Rems Study Area
No comprehensive effort has been made to correlate the vegetation classification systems contained in this report to all previous classification systems used in the study area. While useful for individual readers to develop an understanding for the communities described in this report; comparisons of this nature may be misleading. Cross-references of this type are based on subjective interpretations of the co-occurrence of similar species between classifications (FEWG, 1997). This is problematic when comparing different classifications describing patterns in canopy species dominance only (SFNSW, 1989) against those using prominent understorey characteristics and geological features (Payne, 1999) to those using numerical classification of all floristic data (Bell; 1997, Binns, 1996; Clarke and Benson, 1986). Difficulty relating field site data to the transcribed forest type vegetation layer used in the modelling process has reinforced these problems.

Nevertheless, many of the communities presented in this work exhibit features which suggest similarity with those described by other researchers (Benson, 1986:Payne, 1996; Payne 1999; Biosis, 1998; SFNSW, 1989; Bell, 1997; Benson and Fallding, 1981; Benson and Clarke; 1986; Floyd, 1990). Others do not demonstrate an immediate equal, and represent either splits or combinations of previous units. The most extensively distributed communities evoke the best similarities between classifications. The following provide examples of these trends.