Recovery of Tawa-dominated forest fragments in the Rotorua Basin, New Zealand, after cessation of livestock grazing


  • Mike B. Dodd,

  • Ian L. Power

  • Mike Dodd and Ian Power are researchers in the Land & Environmental Management section of AgResearch (Ruakura Research Centre, Private Bag 3123, Hamilton 3240, New Zealand; Tel. +64 7 8385912; Fax: +64 7 8385130; Email: This study was part of an indigenous biodiversity restoration project within a research programme on the management of hill country pastoral systems for multiple outcomes.


Summary  Tawa (Beilschmiedia tawa)-dominated forest fragments on farms within the Rotorua Basin were surveyed to quantify the likely recovery processes following exclusion of domestic livestock grazing, using a space-for-time substitution approach. Vegetation structure, plant diversity and soil fertility were measured at 24 sites within 15 forest fragments on six farms, covering a range in time since exclusion from grazing of 1–53 years. The forest fragments were compared with a large area of ungrazed forest in the nearby Lake Okataina Scenic Reserve. As time since exclusion from grazing increased, indigenous plant species diversity increased (up to 30–35 years); ground fern and epiphyte abundance increased (up to 30–35 years); tree seedling and sapling numbers, and litter cover also increased (up to 10–15 years); and overall tree numbers increased, while average tree diameter at breast height and overall tree basal area did not differ significantly. The soil fertility status was highly variable, obscuring clear patterns, although Olsen P status decreased with time since grazing exclusion. Once grazing of forest fragments ceases, significant changes in their diversity, structure and soil characteristics can be expected, which indicate recovery of these plant communities towards the conditions observed in ungrazed forest.


New Zealand's lowland indigenous forest ecosystems have experienced high levels of loss (Park 1983), principally because of the development of pastoral agriculture and production forestry, which now cover 80–90% of low-altitude environments (Leathwick et al. 2003). In most regions, numerous forest fragments remain within the mainly pastoral farming matrix, but they are typically small and isolated, leading to significant impacts of edge effects on their microclimate and vegetation structure (Young & Mitchell 1994). Further pressures on these fragments include historical selective harvesting of emergent conifers, unrestricted grazing by domestic livestock; infestation with non-native weeds and animal pests; and exposure to fertilizer and agrichemical drift. Such multiple and interacting effects are difficult to disentangle, but the net result is that those fragments that are remnants of the original forest cover are degraded in terms of vegetation structure, indigenous species diversity and invasive species (Burns et al. 2000; Norton & Miller 2000; Smale et al. 2005).

In the wider national context, the damage to native forests caused by introduced mammalian pests (e.g. Common Brushtail Possum, Trichosurus vulpecula) is regarded as one of the most pressing conservation problems (Craig et al. 2000). Secondary to this issue in agricultural landscapes is the low level of protection of forest fragments from domestic livestock grazing (Harding 1997). Grazing has strong effects on plant species diversity and abundance. The loss of palatable species through preferential browsing is well documented (Wardle et al. 2001). Livestock also facilitate the ingress of exotic species through transport of propagules and soil disturbance (Tanner 1992). Therefore, significant restoration gains from exclusion of livestock should be expected with fencing and retirement of remnants from grazing animals, although additional management, such as planting and wild animal control may be required to facilitate recovery, particularly where desirable species are locally absent (Spooner et al. 2002). Exclusion of livestock is seen as a priority first step in the restoration of forest fragments, and is strongly emphasized in the relevant practitioner literature (e.g. Porteous 1993; Environment Waikato 2001). However, until recently there has been little documented data on the recovery processes of the vegetation and soils after livestock grazing has ceased (Timmins 2002; Smale et al. 2005).

Building realistic expectations among landowners (who as pastoralists are generally unfamiliar with forest succession ecology) with achievable restoration goals and time frames are required (Norton & Miller 2000). A better understanding of post-exclusion succession would also be useful for evaluating the biodiversity gain in terms of the return on investment in the cost of restoration, as compared with other potential management interventions (e.g. pest and weed control, replanting).

The main objective of this study was to quantify the recovery processes in forest fragments with respect to vegetation and soils, using a space-for-time substitution approach. We hypothesized that measurements of a number of vegetation and soil characteristics of the forest fragments would show variation across the scale of time since grazing exclusion, converging to those of a forest reserve, that is native plant diversity would increase, understorey structure would develop, ground cover vegetation and forest litter would increase and soil fertility would decrease.

Methods and materials


Six farms in the Rotorua Basin with a total of 15 forest fragments each of >1 ha in area were selected for this study (Fig. 1). The selection was based on similarity in canopy dominance of Tawa (Beilschmiedia tawa) and availability of information from the farmers in terms of time since grazing exclusion. This led to some unavoidable variation in site characteristics, in that they covered an altitudinal range from 300 to 600 m above sea level and a range in soil types, including recent soils based on Rotomaha mud, recent soils on yellow-brown pumice soils, and yellow-brown pumice soils (Rijkse 1979). Within some of the larger fragments, portions had been fenced and excluded from grazing for varying periods, as reported by the farm owners. Reference to grazing exclusion relates to domestic livestock only – none of the fragments were under management to control other grazers/browsers such as possums and Dama Wallabies (Macropus eugenii) which are known to be abundant in the area. Thus, a total of 24 sites on farms were selected, with another site within the Lake Okataina Scenic Reserve (hereinafter ‘Okataina Reserve’) included for comparison, representing the long-term vegetation condition in the absence of livestock grazing. This reserve also has a recent history of possum and wallaby control via poisoning and hunting (P. Alley, DoC, pers. comm.). A further comparative dataset from the Okataina Reserve was also extracted from the National Vegetation Survey (NVS) database (Manaaki Whenua – Landcare Research 2006).

Figure 1.

Location map of the farms and Okataina Reserve site within the Rotorua basin, North Island, New Zealand.

Botanical nomenclature follows Poole and Adams (1994) and Manaaki Whenua-Landcare Research (2002–2006).


At each site (including the reserve), four plots were randomly located and surveyed using the ‘Quick plot’ techniques in Handford (2000), with some modifications as specified below. All plots were located greater than 50 m inside the forest boundary to avoid sampling edge-affected vegetation (Young & Mitchell 1994). The plots were all 20 m × 4 m in size and the plot edges were at least 10 m apart to avoid resampling the same vegetation.

All vascular plant species occurring in the plots were identified to species or genus and the abundance of the major groups assessed by visually scoring the number of ground ferns, saplings (woody stems >30 cm high and <3 cm diameter), seedlings (woody stems <30 cm high), epiphytes and vines on a scale of 0–3 (0, not present; 1, rare, or present across <5% of the plot; 2, common, or present across 5–30% of the plot; 3, abundant, or present across >30% of the plot). The canopy cover, litter cover and bare ground cover were also visually scored on a percentage basis, using a 5%-increment scale. Individual tree diameters at breast height (d.b.h.) were measured on trees and tree ferns with stems >3 cm d.b.h.

Four soil samples (0–75 mm depth) were taken at regular intervals along the longitudinal centre line of each plot, bulked into one sample for each site and then analysed for pH, Olsen phosphate (Olsen P), potassium (K) and sulphate sulphur (SO4-S) using the methods of Cornforth and Sinclair (1984).

The data extracted from the NVS database included species lists from four 20 m × 20 m Tawa-dominated Recce plots (Allen 1993) surveyed in 1999, and tree density and diameter measurements from these and an additional 17 Tawa-dominated plots recorded at the same time.

Statistical analysis

The plant species richness, visual score and forest structure data from the four plots at each site were averaged for the site. Analysis of these results across time since grazing exclusion was completed using the Bayesian smoothing techniques in Flexi (Upsdell 1994). The individual species presence data (probability of occurrence) were analysed in genstat (GenStat Committee 2006) using a general linear mixed model including terms for farm identity and years since exclusion from grazing. The identification of forest fragments of a similar size across the range of exclosure periods was constrained; hence, potential relationships between fragment size and response variables (species richness, forest structure, etc.) were tested for significance with a linear regression model. The comparison of forest structure attributes (mean d.b.h., tree density and basal area) between the measured sites and the NVS data was analysed with a two-tailed t-test, assuming unequal variance (given differing plot sizes).


Plant diversity

The total number of native vascular plant species recorded across all forest fragment sites was 55 (see Appendix for species list). This included 23 trees, eight climbers, seven ground ferns, six epiphytic ferns, six herbaceous species, four tree ferns and one epiphytic orchid. No exotic species were found in the plots at any site, although various species of thistle, foxglove and pasture grasses were observed in the forest fragment margins. The greatest number of species identified at any one site was 32, at a site 23 years after grazing exclusion (AGE), and the least number of species recorded at a site was 11 (<1 year AGE). Total species richness within the fragments was not significantly related to fragment size (P = 0.74), but showed a significant increase with time since grazing exclusion (Fig. 2). Species richness appeared to be highest at approximately 35 years AGE, with the limited data being highly variable after that point. Twenty-eight species were recorded in the Okataina Reserve plots, all indigenous. All of those species were also recorded in at least one of the farm fragments. The confidence bands around the fitted curve overlapped the dashed line representing the reserve site at 25–30 years AGE (Fig. 2). With regard to the NVS plots, the mean number of species recorded per 20-m × 20-m plot was 28, with 53 indigenous and seven exotic species recorded across all four plots.

Figure 2.

Vascular plant species richness with increasing time since exclusion from grazing (shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).

The dominant tree within all the fragments was Tawa, which was present in at least one plot at every site (Table 1). The only other ubiquitous species were Mahoe (Melicytus ramiflorus) and the epiphytic Leather-leaf Fern (Pyrrosia eleagnifolia). A number of tree and tree fern species were common components, including Pigeonwood (Hedycarya arborea) and Rewarewa (Knightia excelsa), present at >80% of the sites; and Raurekau (Coprosma grandifolia), Silver Fern (Cyathea dealbata) and Wheki (Dicksonia squarrosa), present at >50% of the sites.

Table 1.  Occurrence of indigenous vascular plant species in the forest fragments
GroupSpecies present at all forest fragment sitesSpecies present at >50% of forest fragment sitesSpecies present only at sites having >10 years of grazing exclusion
  • *

    only recorded once at sites having >20 years of grazing exclusion.

TreesBeilschmiedia tawaCoprosma grandifoliaDacrydium dacrydioides*
Melicytus ramiflorusHedycarya arboreaGriselinia lucida
 Knightia excelsaMyrsine australis
  Pittosporum eugenioides*
  Pittosporum tenuifolium
  Rhopalostylis sapida*
  Schefflera digitata
  Weinmannia racemosa
Tree ferns Cyathea dealbata 
 Dicksonia squarrosa 
Vines Metrosideros diffusa 
 Ripogonum scandens 
 Passiflora tetrandra 
EpiphytesPyrrosia eleagnifoliaAsplenium flaccidum 
 Astelia solandri 
 Blechnum filiforme 
 Earina mucronata 
Herbaceous  Uncinia uncinata
  Microlaena avenacea*
Ground ferns  Asplenium bulbiferum
  Pteridium esculentum

The abundances of vines, epiphytes and ground ferns were highly variable in the more recently grazed plots, but did increase significantly with time since exclusion from grazing in all three groups, appearing to be highest at 30–35 years AGE (e.g. Fig. 3). The predominant vine, present at all but one site, was Supplejack (Ripogonum scandens). White Rata (Metrosideros diffusa) and Passion Vine (Passiflora tetrandra), as well as the epiphytes Hanging Spleenwort (Asplenium flaccidum), Kahakaha (Astelia solandri) Bamboo Orchid (Earina mucronata) and Thread Fern (Blechnum filiforme) were present at more than half of the sites (Table 1).

Figure 3.

Mean visual abundance score of ground ferns, with increasing time since exclusion from grazing (refer to methods for scoring scale; shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).

A number of species were recorded only in sites that had been ungrazed for at least 10 years (Table 1), including Puka (Griselinia lucida), Mapou (Myrsine australis), Kohuhu (Pittosporum tenuifolium) and Pate (Schefflera digitata); also the ferns Hen and Chicken Fern (Asplenium bulbiferum) and Bracken Fern (Pteridium esculentum); and the sedge Uncinia uncinata. Another four species were only recorded (once) at sites that had been ungrazed for at least 20 years, including Kahikatea (Dacrydium dacrydioides), Tarata (Pittosporum eugenioides), Nikau Palm (Rhopalostylis sapida) and Bush Rice Grass (Microlaena avenacea).

Ten species showed significant (P < 0.05) increases in their probability of occurrence with increasing time since exclusion of grazing: Hen and Chicken Fern, Raurekau, White Rata, Wheki, Titoki (Alectryon excelsus), Shining Spleenwort (Asplenium oblongifolium), Sickle Spleenwort (Asplenium polyodon), Lance Fern (Blechnum chambersii), Hound's Tongue Fern (Microsorum pustulatum) and Fragrant Fern (Microsorum scandens). Filmy ferns (Hymenophyllum, not identified to species) also showed a significant increase. No species showed significant decreases in their probability of occurrence with increasing time since exclusion from grazing.

Forest structure

Tree canopy cover was similar across all the sites (including the reserve), at 65–75% cover, with no significant trend over time since grazing exclusion. Average tree d.b.h. appeared to decrease slightly with increased time since grazing exclusion (Fig. 4a), although the high level of variability in sites of <10 years AGE meant that this pattern was not significant. Average tree density (stems per hectare) increased significantly with increasing time since grazing exclusion (Fig. 4b). The net result of these patterns was little overall change in tree basal area (55–60 m2 per ha) with time since grazing exclusion, slightly greater than that in the Okataina Reserve (45 m2 per ha).

Figure 4.

Differences in (a) mean tree diameter at breast height (d.b.h.) and (b) tree stem density, with increasing time since exclusion from grazing (shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).

The mean tree d.b.h. across the 21 NVS plots was 12.6 ± 0.7 cm, significantly lower (P < 0.01) than that of the forest fragments and Okataina Reserve stand (Fig. 4a). The mean tree density of the NVS plots was 2870 ± 270 stems per ha, significantly higher (P < 0.01) than that of the forest fragments and Okataina Reserve stand (Fig. 4b). The mean basal area of the NVS plots was 63 ± 3 m2 per ha, not significantly different from the forest fragments and Okataina Reserve stand (P = 0.15).

The abundance of seedlings increased relatively rapidly with time since grazing exclusion (Fig. 5a). Seedling abundance in the fragments was highest at approximately 15–20 years AGE, but was similar to the Okataina Reserve for periods longer than ~10 years AGE. A very similar pattern was observed for saplings (Fig. 5b). Sapling abundance was highest at 15–20 years AGE, but was similar to the Okataina reserve for periods longer than ~10 years AGE. The majority of saplings observed were either Mahoe or Tawa.

Figure 5.

Mean visual abundance score of (a) seedlings and (b) saplings, with increasing time since exclusion from grazing (see methods for scoring scale; shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).

Bare ground cover decreased significantly with time since exclusion from grazing, from 23% down to a level similar to the reserve (1–2%) at 10–15 years AGE (Fig. 6). Percent litter cover showed an inverse pattern, increasing significantly with time since grazing exclusion, from 50% to a level similar to the Okataina Reserve (81%) at 10–15 years AGE.

Figure 6.

Percent cover of bare ground with increasing time since exclusion from grazing (shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).

None of the forest structure variables were significantly related to fragment size (P > 0.05) based on the linear regression analysis.

Soil fertility

The soil analysis results showed no significant relationship between any soil fertility characteristics and fragment size and no significant difference in soil pH, K, and SO4-S levels with time since exclusion from grazing. Soil Olsen P was also highly variable in plots that had been grazed within 20 years of measurement, but there was a statistically significant decrease in Olsen P with increasing time since exclusion from grazing (Fig. 7).

Figure 7.

Soil Olsen P (µg/L) with increasing time since exclusion from grazing (shaded area represents 83% confidence interval of the trendline; dashed line represents the ungrazed Okataina Reserve).


Of the numerous studies into the impacts of introduced mammalian browsers on New Zealand indigenous forest, most have focused on possums, deer and goats, with only a few directed at domestic livestock, most of them being observational studies (Timmins 2002). The more detailed opportunistic study of existing exclosures by Timmins (2002) in south-westland showed significant, but quite variable, impacts of cattle disturbance on vegetation species composition, structure and regeneration that lasted for decades. In an associated 10-year study of newly established exclosures, Buxton et al. (2001) also noted the complex responses of the forest communities to cessation of grazing, but consistently observed increases in woody plant cover in the browse layer (0.3–2.0 m). Miller (2006) observed few significant differences in sapling and seedling populations between grazed and ungrazed (for 10–20 years) podocarp-dominated forest patches in south-westland. Another recent study of Kahikatea fragments in the Waikato measured recovery of species and population structures following fencing and identified a 20-year time frame as a turning point (Smale et al. 2005). Although grazing and exclusion impacts appear to vary with forest type and location, similar effects on the ‘browse layer’ are commonly observed.

This study confirmed the that most striking effects across the continuum of length of time since grazing exclusion were variation in forest structure in terms of herbaceous cover, seedling abundance and sapling abundance (Figs 3 and 5). Since these are features of the lower forest tiers, much of this effect can be directly attributable to removal of browsing and other disturbance factors associated with cattle (e.g. uprooting small trees, Jane 1983). Seedling and sapling numbers increased rapidly with time since exclusion of grazing, being similar to the levels observed in the Okataina Reserve from 10 years AGE and apparently stabilizing around this level in the longer term. Smale et al. (2005) similarly observed the development of sapling populations in sites within 15 years since grazing exclusion, and this result concurs with the increase in shrub density in forest plots observed by Buxton et al. (2001). Most of the saplings observed in the forest fragments this study were Mahoe and Tawa, also the predominant canopy species. Thus, the contribution of sapling regeneration to species richness is minimal, but the contribution to the maintenance of forest structure is important.

The density of the tree population also increased with time since grazing exclusion, from a mean of ~1200 stems per ha at <1 year AGE to ~2000 stems per ha at 50 years AGE (Fig. 4b). This pattern can be expected to result from cohorts of saplings that appear to establish in the 10-year period following release from grazing, moving through into the mature tree population. Given the recognized relationship between tree density and tree size in forest stands, modified for stem diameter rather than tree biomass data (White 1985), a concomitant decline in mean tree d.b.h. with time since grazing exclusion might be expected. A slight but non-significant decrease in mean d.b.h. was observed, along with no change in forest basal area across the time variate. A similar pattern was also observed in Kahikatea fragments in the Waikato (Smale et al. 2005), although the basal areas observed in that study were much greater (more than 90 m2 per ha cf. 55–60 m2 per ha in the present study). The stand dynamics of these recovering stands appear to show movement down the theoretical self-thinning line (Fig. 8). The original grazing pressure, involving the destruction of young stem cohorts and reduction in stem density, had apparently driven the stands up the line relative to the ungrazed reserve, as indicated by the NVS data. Thus, grazing may have had similar effects to thinning in the short term, allowing improved growth in the larger established tree cohorts, but which would ultimately result in reduced basal area in the long term, where tree mortality opened up gaps that were not filled by new cohorts. The process of recovery after grazing exclusion appears to be reversing this process, although long-term monitoring studies are required to confirm this hypothesis.

Figure 8.

Tree density and size relationships for Tawa-dominated forest sites in the Rotorua Basin (AGE, after grazing exclusion). The visually imposed self-thinning line is represented by y = –0.67x + 3.5 after White (1985). The hypothesized dynamics of forest fragments in response to grazing and exclusion are indicated by the dashed arrows.

Species richness increased significantly in response to time since grazing exclusion, although at a much lower rate than observed in Kahikatea fragments in the Waikato (Smale et al. 2005). Overall mean richness in the forest fragments after 25–30 years was similar to that recorded in the equivalent-sized plots surveyed in the Okataina Reserve, and there were no unique species in those plots. This indicates good potential for the recovery of forest fragments in terms of plant species richness, relative to a conservation reserve as a benchmark. However, numerous species were recorded in the NVS plots (covering a greater area of 1600 m2 in total) that were not found in any of the forest fragments. This may be due to the additional animal pest control occurring in the reserve (for wallabies and possums) but it is not clear from this data comparison whether this is anything more than a sample area effect. The lack of exotic species in the forest fragments was surprising, especially since they were recorded in the NVS plots. Without prior data on their presence in the fragments while being grazed, it is not clear whether the recovering forest has successfully excluded them or whether this is also an artefact of the smaller areas sampled.

The recovery of native species richness in forest fragments after cessation of grazing could be due to either removal of direct grazing pressure on specific understorey species or the development of forest structure creating suitable environmental conditions for establishment. Certainly, there was evidence of the re-emergence (or immigration) of a number of understorey species, particularly the ground ferns, that is Hen and Chicken Fern, Lance Fern, Shining Spleenwort and Wheki. The two non-ferns that also increased in their probability of occurrence (Titoki and Raurekau) were largely observed as small seedlings and saplings in the ‘older’ sites. With regard to a number of the species that did not appear at sites with less than 10 years of grazing exclusion (i.e. Mapou, Pate, Puka and Bush Rice Grass), other studies have noted the susceptibility of these to browsing by cattle, deer and goats (Jane 1983; Wardle 1984; Sweetapple & Burns 2002; Timmins 2002).

With regard to environmental conditions favouring establishment, the epiphytic ferns Sickle Spleenwort and Microsorus spp., along with filmy ferns, also increased in their probability of occurrence with time since exclusion from grazing. Given that humidity is known to be an important factor in germination success of epiphytes (Dawson 1988), the development of forest structure following exclusion from grazing may be improving microclimatic conditions and indirectly benefiting these species. However, since the measurements did not distinguish between the locations of epiphytes (e.g. canopy vs. fallen logs), this may also be simply a direct grazing effect.

Another important aspect of the environment for plant establishment for both epiphytic vines and trees is the condition of the forest floor. Although livestock effects on the condition of the forest floor appear to be poorly studied, stock are observed to disturb the litter and organic layers of soil and cause compaction of the mineral layers (Schneider et al. 1978), which likely creates a hostile environment for seed germination. In terms of the effects of livestock removal observed in this study, litter cover increased while exposed roots and bare ground cover decreased relatively quickly (i.e. in the first 10 years). Removal of this source of disturbance, in terms of hoof and browsing damage, may have slowed the physical breakdown of litter, allowing it to recover to levels similar to ungrazed sites. However, recovery of forest soil structure is likely to take much longer than the time frames encompassed by this study.

The soil fertility status overall showed a high degree of variation between sites which meant that clear patterns over the time scale of this study were difficult to detect. In these Rotorua volcanic soils with low natural P levels, soil Olsen P represents a good indicator of the impact of the application of phosphate fertilisers and animal transfer through dung deposition. Elevated phosphate levels under grazed forest fragments (compared with ungrazed forest soils) have been shown elsewhere and attributed to direct fertilizer application and animal transfer (Stevenson 2004). In this study, sites up to 20 years since grazing exclusion had variable levels of Olsen P, some >70 units. Persistently high levels may well be the result of high original inputs from fertilizer drift and animal transfer combined with low loss rates once livestock are removed.

Recent review and policy documents (e.g. DoC 2000; PCE 2001) have emphasized concerns about the poor state of indigenous biodiversity in lowland agricultural landscapes, but at the same time recognized the great potential for recovery and restoration in this context. There are a number of pathways that restoration practitioners could take: conducting animal pest control across the broader landscape; focusing on the condition of remnant forests through livestock exclusion and legal protection; and increasing the extent and connectedness of remnant forests through land retirement and supplementary planting. This study has attempted to examine the impact of exclusion of livestock in isolation, recognizing that it is one of the management activities currently being strongly advocated by local government agencies and non-government organizations (NGO) (e.g. Porteous 1993). It is apparent that significant recovery of indigenous plant diversity and forest structure can be expected to occur over timescales of greater than 10 years following grazing exclusion, providing some guidance as to realistic expectations for forest restoration.


Exclusion of livestock from forest fragments on farms in the Rotorua Basin has led to recovery in a range of forest structure and diversity characteristics over a 30–35-year time frame (i.e. species diversity, tree d.b.h., epiphyte and ground fern abundance). However, the abundance of seedlings and saplings, and ground cover of litter were observed to recover over a shorter time frame of 10–15 years. Other characteristics showing recovery included increases in a number of palatable fern and shrub species, epiphytic species, and a decline in soil phosphorus status.


We thank Tony Carr, John Ford, David Reeves, David Royal, the Ngati Rangiteaorere Te Ngae Farm Trust and Ngati Whakaue Tribal Lands Inc. for their cooperation in allowing us to survey the forest fragments on their properties; the Department of Conservation staff in Rotorua and Colin Stace at Environment BOP for assistance with locating suitable sites; Martin Hawke for assistance in the field; Martin Upsdell and Catherine Cameron of AgResearch for statistical analyses; and Bruce Burns of Landcare Research for comments on the manuscript. The valuable suggestions of two journal reviewers improved the manuscript. We acknowledge the use of data from the National Vegetation Survey, originally collected by Department of Conservation staff. This research was funded by the Foundation for Research, Science and Technology (FRST), contract C10X0223.

Table Appendix.  Species list for Tawa-dominated forest fragments in the Rotorua Basin
TreesAlectryon excelsusTitoki
Aristotelia serrataWineberry
Beilschmiedia taraireTaraire
Beilschmiedia tawaTawa
Carpodetus serratusPutaputaweta
Coprosma grandifoliaRaurekau
Dacrycarpus dacrydioidesKahikatea
Dacrydium cupressinumRimu
Geniostoma rupestreHangehange
Hedycarya arboreaPigeonwood
Knightia excelsaRewarewa
Laurelia novae-zelandiaePukatea
Litsea calicarisMangeao
Lophomyrtus bullataRamarama
Macropiper excelsumKawakawa
Melicytus ramiflorusMahoe
Myrsine australisMapou
Pittosporum eugenioidesTarata
Pittosporum tenuifoliumKohuhu
Pseudopanax arboreusFive-finger
Rhopalostylis sapidaNikau
Schefflera digitataPate
Weinmannia racemosaKamahi
ClimbersGriselinia lucidaPuka
Metrosideros diffusaWhite Rata
Metrosideros fulgensScarlet Rata
Metrosideros perforataClinging Rata
Muehlenbeckia australis 
Parsonsia heterophyllaNew Zealand Jasmine
Passiflora tetrandraPassion Vine
Ripogonum scandensSupplejack
Ground fernsAsplenium bulbiferumHen and Chicken Fern
Asplenium oblongifoliumShining Spleenwort
Blechnum chambersiiLance Fern
Blechnum fluviatileRay Water Fern
Diplazium australe 
Pneumatopteris pennigeraFeather Fern
Pteridium esculentumBracken
EpiphytesAsplenium flaccidumHanging Spleenwort
Asplenium polyodonSickle Spleenwort
Blechnum filiformeThreadfern
Hymenophyllum spp.Filmy ferns
Lygodium articulatumMangimangi
Pyrrosia eleagnifoliaLeather-leaf Fern
HerbaceousAstelia solandriKahakaha
Collospermum hastatum 
Microlaena avenaceaBush Rice Grass
Microsorum pustulatusHound's Tongue Fern
Microsorum scandensFragment Fern
Uncinia uncinataHook Grass
Tree fernsCyathea dealbataSilver Fern
Cyathea medullarisBlack Mamaku
Cyathea smithiiPonga
Dicksonia squarroseWheki
OrchidEarina mucronataBamboo Orchid