Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Environmental correlates of tree alpha‐diversity in New Zealand primary forests

Environmental correlates of tree alpha‐diversity in New Zealand primary forests Leathwick, J. R., Burns. B. R. and Clarkson, B. D. 1998. Environmental correlates of tree alpha-diversity in New Zealand primary forests. - Ecography 21: 235-246. Correlations between environment and tree alpha-diversity in New Zealand's primary forests were examined using an extensive qtiantitaiive dataset (i4 540 plols). Generalised additive models were used to examine relationships between species richness and tetnperature, solar radiation, root-zone moisture deficit, relative humidity, lithology, drainage, and plot size for all trees (112 species), and separately for broadleaved trees (88 species), eomfers (17), and the genus Nothofagus (4). Diversity both for all tree species and for broadleaved trees was predicted to be highest on sites with high temperatures, high solar radiation, and high soil and attnospheric moisture, and on sedimentary atid basaltic substrates. Highest conifer diversity was predicted on sites with intermediate temperatures, low solar radialion, high root-zone and atmospheric moi.sture. and rhyolitic and Quaternary substrates, particularly where drainage was impeded. Highest Nothofagus diversity was predicted for sites combining low temperatures, high solar radiation, high root-zone moisture but low atmospheric moisture, and on granitic substrates. DilTerences in diversity between tbe species groups on different Iithologies are interpreted as rejecting both the elTects of variation in large-scale disturbance histories, and the eftects of confounding environmental factors associated with particular substrates. There were also significant interactions between speeies groups: both broadleaved tree and conifer richness were predicted to be lower on sites where one or more Nothojagus spp. - all of which have marked patchiness in their distribution - are present. Although these results are consistent with the hypothesis that tree diversity' is highest on sites conducive to high productivity, hislory is also indicated as an important delerminant of tree diversity in New Zealand. J- R. Leathwick (leatliwickj( 15 MJ m " - d ' or more, occur in the north of both islands, and values as low as 12 M,l m - d " ' occur in the south of ihe South Island. Annual rainfall varies strongly with topography, and is highest about and to the west of the South Island main divide, where it may reach as high as 12 000 mm {Whitchouse 1988). Much less falls to the east of South Island's tnain divide, with many sites receiving annual totals < 1200 mm, and some declining to as little as 350 mm (Anon. 1983). Highest annual rainfall in the North Island is experieneed in montane areas, mostly in the eentre and west, where it may exceed 6000 mm. Lowland sites generally receive between 1200 and 1600 mrn. but annual totals may fall as low as 800 mm in the southeast. Lowest 9 a.m. relative humidities i<6{)"A>) are experienced in areas to the east of major mountain ranges, mostly in the spring, but remain consistently high through most oi' the year (70- 8O'Vi>) in the west of both islands. Study area and methods New Zealand's environment and forests New Zealand consists of three main islands (Fig. 1) totalling ca 270 000 km- in area, and extending over a latittidinal range of ca 12°. The North Island has the more diverse landforms. The main axial ranges, which are composed mainly of greywackcs dating from the Mesozoic era, run from south to north-east and reach altitudes of 1200-1500 m. Soft younger sediments, mostly formed during marine transgressions in the Tertiary, generally occur at lower altitudes and are tnost extensive in central parts. Rhyolitic volcanoes occur mostly from Lake Taupo to the north-east (Anon. 1972a), and have produced at bast 36 tephra formations over the last 50 000 yr (Froggatt and Lowe 1990). Andcsitic and basaltic volcanoes are more widespread, and include the highest peak in the North Island, Ruapchu (2729 m). The South Island is dominated by the main divide of the Southern Alps, formed mostly from Paleozoie or Mesozoic greywacke and schist, and reaching 3000 m or more in altitude (Anon. 1972b). Mass movement is particularly common in central parts, which arc subject to extreme rates of uplift, i.e., as high as 12 mm yr"' (Whitchouse 1988). More stable landforms, formed mainly on granite and gneiss are extensive both in the south and north-west. Quaternary outwash terraces, many formed during the last glaciation, are extensive at 236 Raukumara Range Urewera Ranges Ahlmanawa Range 40S- South Island Stewart Island 170E 175E Fig. I. Gcogriiphic il-auires of New Zealand ;is mciitiont-d in the text. The locations of the axial mountain ranges of the two main islands are shown by dashed lines. ECOURAI'HY 21;3 (I99S) Primary forests, although subject to widespread clearance since human arrival, still total nearly 6500 knr in area (Newsome 1987). They occur mostly at higher altitudes and along the main axial ranges. Those on warm moist sites are generally dominated by a range of broadleaved evergreen tree species, with local concentrations of conifers, particularly on rhyolitic and Quaternary substrates. In contrast, primary forests on colder sites are often of much more simple floristic composition, and are generally dominated by one or more Nolhoja^us species (Wardle 1991). There are relatively few quantitative accounts of the alpha species diversity of New Zealand's forests {Ogden 1995), and none which presents a national perspective. A number of analyses of alpha-diversity, although from studies of limited geographic extent (e.g., Wilson and Sykes 1988, Wilson et al. 1990. Lee et al. 1991, Burns 1995). indicate that tree diversity generally declines with increasing altitude, conforming to the wider global patterns deseribed above. Two studies with more extensive spatial coverage, but considering the diversity of only selected forest communities, indicate only slight declines in tree diversity with increase in latitude (Wardle 1984, Ogden 1995). Wardle (1984) documents a deeline in tree diversity in Nothofuf-its forests along gradients of decreasing moisture availability and soil fertility. ment. Separate counts were also made of the numbers of species in three groups based on differences in functional morphology and physiognomy, i.e., conifers, Nothofa^^us spp., and broadleaved angiosperm species other than Nothofagus spp. The separation of Nothofagus spp. from the other broadleaved trees has a long history in the New Zealand literature, and is made not only because of their possessitin of ectomycorrhizal rootlets, generally smaller leaves, and 'gregarious' habit (Wardle 1991). but also because of their generally greater tolerance of less favourable climatic conditions than most broadleaved species (Wardle et al. 1983, Leathwiek 1995, Leathwiek unpubl.). Because of ambiguities in the recording of species names in the original data, some closely related groups or pairs of species were amalgamated, i.e., Podoearptts hallii with P. totara, Nestcgis eunninghamii with N. laneeoiata. and Quintinia scrrata and Q. ellipliea. with Q. aetitifolia (Nomenclature tbllows Allan 1961 and Connor and Edgar 1987). Climate estimates for each plot were derived from thin-plate spline surfaces (Hutchinson and Gessler 1994) fitted to monthly data for chmate stations in New Zealand. Monthly estimates for each plot of mean maximum and minimum daily air temperature, precipitation, daily 9 a.m. humidity, and daily solar radiation were converted to summary variables (Table la) tor analysis as follows. Temperature and solar radiation estimates were each converted into mean annual values, i.e., mean annual temperature (T) and mean annual solar radiation (S) respectively, to indicate overall thermal and energy conditions. The lowest monthly humidity values for the year (H) were used lo indicate atmospheric moisture deficits, after preliminary analyses had shown its consistently higher correlation with tree diversity than mean annual humidity. Monthly precipitation estimates were combined with estimates of temperature, solar radiation, and soil rooting depth and texture to derive an annual integral of root zone moisture deficit (W) (Leathwiek et al. 1996). Estimates of root zone moisture holding capacity were calculated from data on soil depth and texture derived by overlaying plot coordinates onto a digital copy of the 1:1 000 000 soils map of New Zealand (Anon. 1968). Values for W were then calculated by running a daily water balance model through one year, starting with an assumed saturated soil in late winter, and taking account of daily changes in root zone water storage by adding daily rainfall and subtracting daily evaporation and drainage. Daily water potentials below field capacity were then summed to give an annual integral. Because of their strongly skewed distribution, values of W were log transtbrmed (base 10) betbre further analysis. Data Data for this study were taken from an extensive survey of New Zealand's forests carried out mostly in the 195O's and 196O's, and involving the measurement of over 15 000 plots, each of 0.4 ha (Masters et al. 1957). Plots for which recent fire and/or logging disturbance were noted at the time of survey were omitted, leaving 10 500 plots sampling the majority of North Island's lowland and montane primary forests, and South • Island's lowland primary forests. A further 4000 plots, each of 0.04 ha, measured during more recent (1970"s and I98()'s) protection (brest surveys carried out by the former New Zealand Forest Service, were added to increase the coverage of South Island montane forests (Leathwiek et al. 1996). Beeause of the problems inherent in diversity measures which attempt to combine information on species abundance as well as speeies richness (e.g.. Whittaker 1972), diversity was measured as simply the number of Iree species present, i.e., species richness. For each plot, the number of tree speeies with individuals > 305 mm in diameter was determined non-spermatophyte speeies, i.e.. treeferns. were excluded because of inconsistencies in their measureHCOCiRAPHY The lithology (L) of each plot, a major determinant of edaphic conditions, was detemiined by overlaying ploi coordinates onto a digital copy of the 1:1 O O 000 geoO logical map of New Zealand (Anon. 1972a. b). Geological substrates were grouped for analysis on the basis of their age, weatherability, and composition (Table Ib). An indication of the drainage (D) at each plot was provided by a subjective assessment tnade at the time of original plot measurement using a ihree-step seale. Finally, a variable (PS) was included to indicate the plot size (0.4 or 0.04 ha) for each site. the response variable as in conventional regression, predictions are first formed on a linearised predictor, which is then back-transformed using a link function, which in this case was of logarithmic forrn. Generalised Additive Models differ from GLMs by defining the relationship between response and continuous predictor variables using non-parametric methods (scatter-plot smoothers), rather than parametric combinations of the predictor variables (e.g., X + X-), Tree richness is calculated, then, by first caleulating: linear predictor = a:-l-s,. + . 4-s,, + L,. 4- D., Analysis For all tree speeies. and for each of the three species subgroups, regressions relating species riehness to cnvirontnent were calculated using Generalised Additive Models (GAMs: Hastie and Tibshirani 1990), an exiension of Generalised Linear Models (GLMs; McCullagh and Nelder 1989, Nieholls 1991). The latter offer considerable advantages over conventional regression techniques in their ability to handle dala with error distributions departing from normality, such as the eount data analysed here for which a Poisson error distribution is assumed (Vincent and Haworth 1983). Rather than forming fitted values directly on the scale of where ct is a constant, Si • • • s,, are smooth functions of the continuous climatic variables (Table la), and L,,, D,, and PS; are multi-level coefficients fitted for the three factor variables (Table lb). The richness for any site is then given by richness = e'""--''^ P.'^a,.;ior For each of the four regression models, all six envit"onmcntal predictors were fitted along with the factor variable indicating plot size, and the significance of removing each predictor in turn was tested using changes in residual deviance which are distributed approximately as for a x^-statistic. Non-significant predictors were removed, and the process repeated until only significant Table I. Bitsic statistics Tor cnvironiiiciUal predictors used in the analysis, ti) continuous predictors: b) category predictors, a) Symbol (units) T (=C) S (MJ m- - day W (log,,, MPa day) M (%) b) Symb. Description mean anntial temperature mean annual solar radiation annual integral of root-zone moisture deficit min. monthly mean 9 a.m. humiclit\ Mean (range) 10.1 (3.4- 15.4) 14.1 (11.8-15.4) 0.7 (0 2..^) 74.4 (58 -89) Dcsc. geology Classification G = granite and gneiss, mainly Paleozoic or older, 804 plots: II = hard sedimentary substrates, predom. greywiieke, argillite. and schists. Cretaceous and older. 6206 plots: S^soft sedimentary substrates, predom. calcareous sandstone, siltstone. PliocenePaleocene, 2591 plots: B = more basic volcanlcs. i.e, andesites and basalts, mainly Pliocene and older, but some Quaternary, 1131 plots: R = acid volcanics, i.e.. rhyolites. mainly Quaternary. 928 plots; Q = Quaternary surfaces, sediments deposited by alluvial. glacial, laharic. and aeolian processes, etc., 2880 plots. G = good: M — moderate: P = poor. N = 0 . 4 ha: Q = 0.04 ha. PS drainage plot size F.CO(}RAI'HY l\ predictors remained. Models were then evaluated by plotting both fitted values and residuals against each predictor, and by contouring predictions of tree species richness at various combinations ol' T and H for comparison with actual values. Spatial predictions of tree richness were then calculated for comparison with the raw data, by using the regression equations in conjunction with estimates of the six environmental predictors for points on a 5 km grid across New Zealand. The effect on btHh broadleavcd and conifer diversity of Nolhofagus presence was tested by adding to their regressions, a category predictor indicating whether Nothofagus species were present or absent for each plot. Results Geographic patterns of alpha-diversity A total of 112 tree species occurred in the entire data set, with mean values for total species richness in the 0.4 ha plots ranging from I to 16. and with a grand mean oi 5.4 (Table 2). Predictably, mean values were lower in the 0.04 ha plots, but this difference alsti reflects differences in the range of environmental characteristics sampled by the two plot sizes; the 0.04 ha plots mostly occurred in the South Island, where both temperature and solar radiation are generally lower than in North Island, Plots with high species richness were located mostly in the northern half of North Island, richness gradually declining with progression to higher latitudes and altitudes. Broadleaved trees were the largest subgroup, with 88 species distributed among 30 families as follows; Winteraceae (2). Lauraceae (3). Monimiaceae (2), Violaceae (2), Onagraceae (1). Proteaceae (2), Coriariaceae (1), Pittosporaceae (3), Myrtaceae (6), Elaeocarpaceae (3), Malvaceae (6). Cunoniaceae (3), Escalloniaceae (5 - treated as 3). Papilionaceae (2), Moraceae {!), Corynocarpaceae (1), Icacinaceae (1), Santalaceae (1), Rutaceae (2), Meliaceae (I). Sapindaceae (1), Araliaceae (7), Cornaceae (2), Epacridaceae (4). Myrsinaceae (2), Oleaceae (3 reduced to 2 for analysis), Rubiaceae (8), Asteraceae (9), Myoporaceae (I), and Verbenaceae (1). The maximum broadleaved Table 2. Sutnmary of raw species riehness data. No. positive 0.4 ha plots All species Broadleaved trees Conifers Nothofagus spp. 14 540 10 640 9400 7396 5.4 3.3 2.1 1.7 richness was 14 on 0.4 ha plots, and one or more species occurred on >70% of plots. The spatial distribution of broadleaved richness followed a similar pattern to that for total species richness, i.e., declining with progression away from northern lowland sites to very low levels or complete absence on sites at high latitudes in the south and east. Conifers were the next largest group, totalling 17 species from the families Podocarpaceae (14 - reduced to 13 for analysis), Cupressaceae (2), and Araucariaceae (1). Sites with a high richness of conifers, although concentrated in central North Island and West Coast. South Island, were also scattered over a wide altitudinal range in the north and at lower altitudes in the south, particularly on sites with poor drainage. Four species of Noilwfagus occurred in the dataset; the two subspecies of N. solandri were unable to be accurately differentiated from the original plot data. .Sites with the highest richness in South Island were concentrated in central parts oi' the southern third of the Island, and from the Paparoa Range northwards in the northwest. Sites with high richness in North Island were concentrated on the southern Hanks of Mt. Ruapehu and in the Ahimanawa, Urewera, and Raukumara ranges. The three remaining species occurring in the dataset were all monocotyledonous trees, two from the Agavaceae and one from the Palmae. Correlations between tree species richness and climate Although comparison oi' the overall success of the regression models is hampered by the lack oi' a statistic analogous to the percentage variance explained as used in conventional sums-of-squares regression (McCullagh and Nelder 1989), residual mean deviances were generally close to I. indicating that a substantial amount of the total deviance had been explained (Table 3). Drainage was dropped from the regressions for total species and Nolhofagus richness, but all six environmental predictors were retained in the broadleaved tree and conifer regressions. Changes in residual deviance when dropping each predictor in turn indicate that of the climatic predictors, mean annual temperature (T) has the strongest contribution Mean where present 0.04 ha plots 2.1 1.9 1.4 1.4 0.4 ha plots 1-16 0-14 0-7 0-4 Range over all plots 0.04 ha piois 1 7 0 6 0 4 0-4 EC(KJRAi>HY 21 to the regressions for total species and conifer richness, while minimum monthly mean humidity (H) has the strongest contribution in the regressions for broadleaved tree and Nolhofagus richness. By contrast, mean annual solar radiation (S) and root zone moisture deticit (W) make a much smaller contribution to the results. Examination of the regression functions lor the four climatic predictors (Fig. 2), and predictions formed in relation to the two most strongly correlated of these (T and H. Fig. 3), indicates that total species richness is predicted to reach a maximutn or\ sites combining high mean annual temperature, high solar radiation, low annual root zone moisture deficit, and high miniiiium monthly mean humidity. Within this overall pattern, however, there are marked differences between the three species groups. Predicted pattems of richness for broadleaved trees closely follow those for total species richness, with predicted richness reaching a maximum on warm, high insolation, humid sites with moderate root zone moisture deficit. In contrast, maximum conifer richness is predicted to occur on sites with intermediate temperatures, high humidity, and low insolation and root zone moisture deficit: maximum richness for Nolhofagus is predicted to occur on sites combining cool temperatures, high insolation and low root zone moisture deficit, but low humidity. However, both the latter two groups have a secondary peak of richness at high root zone moisture deficits. and Quaternary substrates {Fig. 4a). Among the species groups, however, broadleaved tree richness is predicted to be highest on hard sedimentary and basaltic substrates, and lowest on granites and Quaternary substrates (Fig. 4b); conifer richness is predicted to be highest on rhyohtes and basalts, and lowest on granites (Fig. 4c); and Nothofagus richness is predicted to be highest on granites, soft sedimentary and Quaternary substrates, and much lower on rhyolites (Fig. 4d). Broadleaved tree richness is predicted to decline as drainage deteriorates (Fig. 5). but that tor conifers is predicted to increase. Does the presence of JSothofafius inHnence the richness of other species groups? Addition to the bi"oad!eavcd tree and conifer regressions of a variable indicating Nothofagus presence/ absence at each plot, resulted in a highly significant reduction in residual deviance in both regressions, but with a more substantial reduction for broadleaved trees than for conifers (Table 4). Responses to other predictors were little changed for either regression model. Combining responses to T and H from the broadleaved tree model while keeping all other environmental factors constant, reveals the substantial decrease in broadleaved tree riehness which occurs in the pi-esence of Nothofagus (Fig. 6a. b). Correlations between tree species richness, and Hthology and drainage Marked variation in species richness is also predicted to occur in relation to lithology (L) but variation is less marked in relation to drainage (D) (Table 3, Figs 4 and 5). The regression coefficients for lithology indicate that total species richness is predicted to be significantly higher on hard and soft sedimentary substrates and basalts, than on granites, rhyolites. Spatial predictions of richness The predicted spatial distribution of tree species richness from our environmental regressions show mixed success (Fig. 7). Predicted patterns of total species richness and broadleaved richness (Fig. 7a, b) correspond well with observed patterns, i.e., with greatest richness predicted for the warm, high insolation sites of northern New Zealand, and declining with progression Table 3 Summary of regressions relating species richness to environmental predictors. For each environmental predictor, table values indicalc ihe increase in deviance resulting from dropping that predictor from the corresponding regression. These values arc distributed approximately as for a /--statistic. For example, for T, S, II, and W, values are signifieant at the 95% and 99% levels of confidence if >7.81 and 11.34 respectively. Model mean devianee 0.739 1.145 0.949 0.828 -T (3 DP) 607.1 1098.9 719,3 767.7 S (3 DF) 184.9 387.5 60,1 378.1 -W (3 DF) 10.4 47..5 87.3 128.1 -H (3 DF) 570.3 1448.9 240.7 945.7 -L (6DF) 162.9 662.1 113.2 1525,.5 -D (3 DF) -PS (2 DF) 783.2 179.2 647.0 117.8 Nolhofagus ECOGRAPJIY 2I:,1 Fig, 2, Non-paramelric regression functions on a logaritlimic seale for species riclincss of all trees, broadleaved Irees, conifers, and :\oflioja,i'iis spp, in relafion to a) mean annual temperature (T): b) mean annual solar radiation (S); e) annual integral of root zone moisture defieit (W); d) minimum monthly mean humidity (H). a) b) Nothotagus 6 8 !0 12 14 13 14 15 Mean annual temperature Mean annual solar radiation C) d) 0,5 Roo1-;one moisture deficit Minimum monthly mean humidity Fig. 3. Contours of tree species richness in relation to mean annual temperature tT) and minimum monthly mean humidity (H) as predicted from environmental regressions: a) all speeies; b) broadleaved trees; c) conifers; d) Notbofagiis spp. The ratige of combinations of T and H sampled by the diita is bhown by overlaid points in (a). Values for other environmental predictors were set as follows; S 14.5, W 0,7. L hard sedimentary, PS - 0.4 ha. a) b) Mean annual temperature Mean annual temperature d) Mean annual temperature Mean annual temperature E((KiRAPHY 21:? (I'WS) b) Broadleaved tree c) Conifer Fig, 4, Regression coefficients for litliology classes {L see Table lb} from regressions of species richness versus environment for a) all speeies; b) broadlcaved trees; c) eoniiers; and d) iXotlioJagus spp. Widths of Ihe horizontal bars are proportional to the number of plots for each class, and upper and lower bars indicate Iwo times pointwise standard errors. a) Broadieaved trees b) Conifers Fig, 5, Regression coefficients for drainage classes (I) see Table lb) from regressions of species riehness versus environment for a) broadleaved trees, and b) eonilers. Widths of the horizontal bars are propoitional to the number of plots for each class, and upper and lower bars indicate two times pointwise standard errors. to higher latitudes and altitudes. Predictions of conifer richness also agree well with observed patterns (Fig. 7c), with highest levels predicted for rhyolitie substrates on the central North Island volcanic plateau, and tor mainly Quaternary substrates in the western and southern South Island lowlands. In contrast, predictions of Nothofagus richness show much poorer correspotidence with observed patterns (Fig. 6d). Although high levels are correctly predicted for the northwest of Soulh Island, only low levels are predicted for the LJrcwera and Raukumara ranges, where three Notbofagiis speeies regularly occur together, and moderate levels are predicted through large parts of New Zealand in which no Nothofa^^us spp. are present. Notably, the latter includes both Stewart Island and Soulh Island's 'beeeh gap" (==most of the middle third of the island - Wardle 1963), occurs in climates combining relatively high temperatures, high solar radiation, and high atmospheric and soil moisture. This is consistent wilh results trom similar studies elsewhere which have demonstrated that tree richness is generally highest in conditions favourable to high productivity, assessed using cither derived measures such as available energy or actual evapotranspiration, or the climatic and site factors (e.g.. temperature, solar radiation, and moisture availability) which control plant productivity (Richerson and Lum 198(1. Wright 1983. Currie and Paquin 1987, Adams and Woodward 1989, Currie 1991. Austin et al. 1996). Although the decline we dernonstrate in tree speeies richness with decreasing humidity is a novel result, it is consistent with the general hypothesis already outlined, i.e., that diversity declines on sites whieh are unfavourable to tree growth. The absence of many tree species from sites with lower humidity most likely reflects the severe limitations placed on their growth, and perhaps survival, during extreme events sueh as fohn winds (Leathwiek unpubl.). During these, low humidity is often combined with high ternperatures and solar radiation, and strong winds, all of which Increase evaporative demands on trees (Landsberg 1986). Discussion Correlations between climate and tree richness Results from this study indicate that the maximum total species richness Ibr trees in New Zealand's forests 242 Fig. 6. Contours of broadleavcd tree species richness in relation lo mean annual temperature (T) and minimum monthly mean humidity (H) as predicted from environmental regressions to which a variable Itidicating :\'riiliofagm- presence/absenee has beet! added. The range of combinations of T and H satnplcd by the relevant data are showti by overlaid points in both graphs. Values for other environmental variables are set as in Kig. 3. a) Broadleaved trees Nothofagus absent b) Broadleaved trees Nothofagus present 10 Mean annual letnperature 12 Mean annual temperature The consistently high correlation between mean annual temperatitre and total species richness we demonstrate is also in accord with documented gradual declines along gradients of increasing altitude in both woody species richness {Wardle 1984. Ogden 1995) and tree richness (Wardle 1984) for New Zealand forests. The increase in diversity of all species with inereasing altitude reported from southwestern New Zealand by Wilson ct a!. (1990) appears contradictory. However, this probably reflects their grouping together of differing growth forms for analysis, an approach whieh can obscure important correlations because different mechanisms may control the richness of different groups (Hitston 1994), The correlation we demonstrate between species richness and solar radiation contrasts with the relatively low variation in species richness along latitudinal gradients demonstrated by Wardle (1984) and Ogdcn (1995). particularly given the strong correlation between latitude and solar radiation (Table 5). It may reflect a confounding of their analyses, which used latitude alone, by the east to west variation in solar radiation at any latitude due to interaction between predominantly westerly-borne cloud masses and New Zealand's axial mountain ranges. The pronounced variation between species groups demonstrated here is consistent with their documented environmental preferences, i.e., with gradual replacement of broadleavcd species by Nothofagus with progression towards less favourable growing conditions (e.g., Wardle 1964. Leathwiek 1995, Ogden et al. 1996. Leathwiek unpubl.). and by eonifers on Quaternary or rhyolitic substrates (e.g.. Wardle et al. 1983, Leathwiek and Mitchell 1992, Leathwiek 1995). The less satisfactory spatial prediction oX CCOCiRAI'llY 21:3 Noihofagus richness across New Zealand is consistent with conclusions from other studies that the disjunct distributions of all (bur species are poorly correlated with climate, and probably reflect very slow reoccupation of their potential ranges following displacement by major geomorphie disturbances or glaciation (e.g., Wardie 1984, McGlone 1985, Wardle 1991, Leathwiek ct al. 1996. Leathwiek unpubl.). Disturbance and tree species richness Although a lack of quantitative data prevented analysis of the effects of recent disturbance events ( <ca 200-500 yr), some clues as to the importance of older events are apparent from this analysis, particularly eoncerning the effects of large-scale events. For example, the lower total species riehness predicted for Quaternary and rhyolitic substrates (Fig. 3a), both of which have formed as a result oi geologically recent deposition of new parent material, suggests incomplete occupation of these sites by the range of speeies present on older sedimentary and granitic substrates in comparable climates. Some support for this explanation comes from the distribution of speeies groups across these substrates. For example, the Nothofagus species, which are noted both for their limited dispersal ability (e.g., Wardle 1980. Rogers 1989) and for their slowness to invade nonNothofagiis tbrest communities (Ogden et al. 1996), are predieted to reach their highest speeies riehness on granitie substrates (Fig. 3d). Such substrates oecur in parts of New Zealand which have been more geologically stable since the Oligocene (McGlone 1985), and in some cases, were subject to less severe glaciation (Wardle 1963). Conversely. Nothofu}^us richness is pre243 a) Total tree richness b) Broadleaved tree richness c) Conifer richness d) Nothofagus richness Fig. 7. Predicted lotal tree species richness on l).4 ha plots across New Zealand for a grid of points al a 5 km spacing, and with equivalent environmental data as that used to develop the regressions. No predictions are made for areas SIIOWLI in wliite. ;ts their environments tall outside the range sampled by the dala. dieted to be dramatically lower on rhyolitic substrates, which have been subject to a number of large-scale eruptions over the last 50 000 yr (Froggatt and Lowe 1990), despite the demonstrated climatie suitability of these sites (Leathwiek and Mitchell 1992. Leathwiek 1995). Similarly, the higher richness of the predominantly bird-dispersed eonifers on rhyolites (Fig. 3c) probably reflects in part the strong ability of many of these speeies to colonise new mineral soils (MeKelvey 1963). However, their higher diversity on this substrate may also be favoured by the predisposition to fire of the generally flat rhyolite plateau landscapes, because of a lack of topographic barriers to fire spread. Although natural fires are rare in New Zealand, firing has been extensive since Polynesian occupation 900-1000 yr BP (McGlone 1983). and some fire modifieation of the supposedly primary forests that grow there, particularly under lower rainfall regimes, cannot be ruled out. Conversely, the low richness predicted for eonifers on gran244 ites (Fig. 3c) probably reflects ihat the localised disturbance events through mass movement, which they exploit elsewhere (Ogden and Stewart 1995). arc less eommon on these substrates (Rcif and Allen 1988. Allen et al. 1991). However, the confounding effects of more subtle environmental differences between these substrates must also be considered alongside likely disturbance effects. For example, the richness of conifers on rhyolites may relkct not only their good dispersal abihty, but also a degree of exclusion of shade-tolerant but more eold-sensitivc broadleaved competitors from these prcdominiiiitly Hat and frost-prone sites (Leathwiek and Mitchell 1992). Similarly, on Quaternary substrates the lower richness of broadleaved trees and Nothofagus spp. and the higher richness of eonifers probably iti part also reflects their relative responses to edaphie differences, and particularly the poor drainage eonditions whieh develop as a consequence of" intense leaching and the relative absence of periodic rejuvenation F.COCiRM'llY :i:3 (IWK) through local tectonic activity (Wardle ct a!. 1983, Rcif and Allen 1988). Conclusions Both environmental and disturbance paradigms are important for understanding the distribution of tree species richness in New Zealand's primary forests. In common with studies from other countries, tree richness at a broad scale is correlated with - and probably in some manner controlled by - the environmental factors of temperature, solar radiation, and moisture availability in both the atmosphere and soil, which in turn control productivity. However, at large to intermediate spatial scales the efteets of cataclysmic but infrequent disturbances resulting from the dynamic nature of New Zealand's landscape, become more apparent. The effects of recent ( < ca 300 yr) disturbance could not be examined in this study, but are undoubtedly also important, particularly at smaller spatial scales. Although more precise partitioning of the relative effects of environment and this variatitjn in disturbance regime is desirable from a theoretical perspeetive, its analysis is problematic with these data given the difficulty in accurately quantifying recent disturbance histories for large numbers of sampie sites. Aekiunvledgmeiits - A number of people have coiUribiited lo ihe development of llic ideas presented in this paper, and in piirlicular M. P. Austin, ,1. Dale, and G. M. Rogers. The paper benefited greatly from ii review in i(s draft stages by W. Lee. Access Io part of the data used in the analysis was facilitated by L. R, Burrows and G, M. .1. Hull. The project was funded by the New Zealand Foundiitlon for Researeli. Science, and 7"eclinolojiy under Contract no. C09405. The influence of Nothofagus Less direct imprints of historical factors are evident in the negative relationship which we have demonstrated particularly between broadleaved richness and the presence of the patchily distributed Nothofagus spp. (see above). This suggests that on sites within the clitiiatic range of Nothofagus. bi'oadleaved tree richness is controlled not only by envirohment but also by whether Nothofagus species have arrived on site. Where (he lalter occur, the richness of other broadieaved trees can be expected to be significantly reduced, perhaps owing to the superior ability of (he ectomycorrhizal Nothofagus to capture nutrients (Wardle 1984. Leathwick unpubl.). Such ;i reduction in riehness consequent on the presenee ol competitive but poorly dispersed species is fundamental to the dynamic equilibrium model of community structure proposed by Huston (1979, 1994). and has been observed in both tropical (Hart et a!. 1989) and temperate settings (e.g.. Loucks 1970). In contrast, richness of conifers appears to be less aOeeted by the presence of Notbofagus. perhaps a consequence of their greater stature enabling them to coexist in at least some Nothofagus Ibrests as emergents. or of their greater tolerance of the infertile soils which ol'ten develop under a Nothofagus eanopy. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecography Wiley

Environmental correlates of tree alpha‐diversity in New Zealand primary forests

Leathwick, J. R.; Burns, B. R.; Clarkson, B. D.
Ecography , Volume 21 (3) – Jun 1, 1998
Download PDF
13 pages

Loading next page...
 
/lp/wiley/environmental-correlates-of-tree-alpha-diversity-in-new-zealand-NMT5rbJwG0

References (54)

Publisher
Wiley
Copyright
Copyright © 1998 Wiley Subscription Services, Inc., A Wiley Company
ISSN
0906-7590
eISSN
1600-0587
DOI
10.1111/j.1600-0587.1998.tb00561.x
Publisher site
See Article on Publisher Site

Abstract

Leathwick, J. R., Burns. B. R. and Clarkson, B. D. 1998. Environmental correlates of tree alpha-diversity in New Zealand primary forests. - Ecography 21: 235-246. Correlations between environment and tree alpha-diversity in New Zealand's primary forests were examined using an extensive qtiantitaiive dataset (i4 540 plols). Generalised additive models were used to examine relationships between species richness and tetnperature, solar radiation, root-zone moisture deficit, relative humidity, lithology, drainage, and plot size for all trees (112 species), and separately for broadleaved trees (88 species), eomfers (17), and the genus Nothofagus (4). Diversity both for all tree species and for broadleaved trees was predicted to be highest on sites with high temperatures, high solar radiation, and high soil and attnospheric moisture, and on sedimentary atid basaltic substrates. Highest conifer diversity was predicted on sites with intermediate temperatures, low solar radialion, high root-zone and atmospheric moi.sture. and rhyolitic and Quaternary substrates, particularly where drainage was impeded. Highest Nothofagus diversity was predicted for sites combining low temperatures, high solar radiation, high root-zone moisture but low atmospheric moisture, and on granitic substrates. DilTerences in diversity between tbe species groups on different Iithologies are interpreted as rejecting both the elTects of variation in large-scale disturbance histories, and the eftects of confounding environmental factors associated with particular substrates. There were also significant interactions between speeies groups: both broadleaved tree and conifer richness were predicted to be lower on sites where one or more Nothojagus spp. - all of which have marked patchiness in their distribution - are present. Although these results are consistent with the hypothesis that tree diversity' is highest on sites conducive to high productivity, hislory is also indicated as an important delerminant of tree diversity in New Zealand. J- R. Leathwick (leatliwickj( 15 MJ m " - d ' or more, occur in the north of both islands, and values as low as 12 M,l m - d " ' occur in the south of ihe South Island. Annual rainfall varies strongly with topography, and is highest about and to the west of the South Island main divide, where it may reach as high as 12 000 mm {Whitchouse 1988). Much less falls to the east of South Island's tnain divide, with many sites receiving annual totals < 1200 mm, and some declining to as little as 350 mm (Anon. 1983). Highest annual rainfall in the North Island is experieneed in montane areas, mostly in the eentre and west, where it may exceed 6000 mm. Lowland sites generally receive between 1200 and 1600 mrn. but annual totals may fall as low as 800 mm in the southeast. Lowest 9 a.m. relative humidities i<6{)"A>) are experienced in areas to the east of major mountain ranges, mostly in the spring, but remain consistently high through most oi' the year (70- 8O'Vi>) in the west of both islands. Study area and methods New Zealand's environment and forests New Zealand consists of three main islands (Fig. 1) totalling ca 270 000 km- in area, and extending over a latittidinal range of ca 12°. The North Island has the more diverse landforms. The main axial ranges, which are composed mainly of greywackcs dating from the Mesozoic era, run from south to north-east and reach altitudes of 1200-1500 m. Soft younger sediments, mostly formed during marine transgressions in the Tertiary, generally occur at lower altitudes and are tnost extensive in central parts. Rhyolitic volcanoes occur mostly from Lake Taupo to the north-east (Anon. 1972a), and have produced at bast 36 tephra formations over the last 50 000 yr (Froggatt and Lowe 1990). Andcsitic and basaltic volcanoes are more widespread, and include the highest peak in the North Island, Ruapchu (2729 m). The South Island is dominated by the main divide of the Southern Alps, formed mostly from Paleozoie or Mesozoic greywacke and schist, and reaching 3000 m or more in altitude (Anon. 1972b). Mass movement is particularly common in central parts, which arc subject to extreme rates of uplift, i.e., as high as 12 mm yr"' (Whitchouse 1988). More stable landforms, formed mainly on granite and gneiss are extensive both in the south and north-west. Quaternary outwash terraces, many formed during the last glaciation, are extensive at 236 Raukumara Range Urewera Ranges Ahlmanawa Range 40S- South Island Stewart Island 170E 175E Fig. I. Gcogriiphic il-auires of New Zealand ;is mciitiont-d in the text. The locations of the axial mountain ranges of the two main islands are shown by dashed lines. ECOURAI'HY 21;3 (I99S) Primary forests, although subject to widespread clearance since human arrival, still total nearly 6500 knr in area (Newsome 1987). They occur mostly at higher altitudes and along the main axial ranges. Those on warm moist sites are generally dominated by a range of broadleaved evergreen tree species, with local concentrations of conifers, particularly on rhyolitic and Quaternary substrates. In contrast, primary forests on colder sites are often of much more simple floristic composition, and are generally dominated by one or more Nolhoja^us species (Wardle 1991). There are relatively few quantitative accounts of the alpha species diversity of New Zealand's forests {Ogden 1995), and none which presents a national perspective. A number of analyses of alpha-diversity, although from studies of limited geographic extent (e.g., Wilson and Sykes 1988, Wilson et al. 1990. Lee et al. 1991, Burns 1995). indicate that tree diversity generally declines with increasing altitude, conforming to the wider global patterns deseribed above. Two studies with more extensive spatial coverage, but considering the diversity of only selected forest communities, indicate only slight declines in tree diversity with increase in latitude (Wardle 1984, Ogden 1995). Wardle (1984) documents a deeline in tree diversity in Nothofuf-its forests along gradients of decreasing moisture availability and soil fertility. ment. Separate counts were also made of the numbers of species in three groups based on differences in functional morphology and physiognomy, i.e., conifers, Nothofa^^us spp., and broadleaved angiosperm species other than Nothofagus spp. The separation of Nothofagus spp. from the other broadleaved trees has a long history in the New Zealand literature, and is made not only because of their possessitin of ectomycorrhizal rootlets, generally smaller leaves, and 'gregarious' habit (Wardle 1991). but also because of their generally greater tolerance of less favourable climatic conditions than most broadleaved species (Wardle et al. 1983, Leathwiek 1995, Leathwiek unpubl.). Because of ambiguities in the recording of species names in the original data, some closely related groups or pairs of species were amalgamated, i.e., Podoearptts hallii with P. totara, Nestcgis eunninghamii with N. laneeoiata. and Quintinia scrrata and Q. ellipliea. with Q. aetitifolia (Nomenclature tbllows Allan 1961 and Connor and Edgar 1987). Climate estimates for each plot were derived from thin-plate spline surfaces (Hutchinson and Gessler 1994) fitted to monthly data for chmate stations in New Zealand. Monthly estimates for each plot of mean maximum and minimum daily air temperature, precipitation, daily 9 a.m. humidity, and daily solar radiation were converted to summary variables (Table la) tor analysis as follows. Temperature and solar radiation estimates were each converted into mean annual values, i.e., mean annual temperature (T) and mean annual solar radiation (S) respectively, to indicate overall thermal and energy conditions. The lowest monthly humidity values for the year (H) were used lo indicate atmospheric moisture deficits, after preliminary analyses had shown its consistently higher correlation with tree diversity than mean annual humidity. Monthly precipitation estimates were combined with estimates of temperature, solar radiation, and soil rooting depth and texture to derive an annual integral of root zone moisture deficit (W) (Leathwiek et al. 1996). Estimates of root zone moisture holding capacity were calculated from data on soil depth and texture derived by overlaying plot coordinates onto a digital copy of the 1:1 000 000 soils map of New Zealand (Anon. 1968). Values for W were then calculated by running a daily water balance model through one year, starting with an assumed saturated soil in late winter, and taking account of daily changes in root zone water storage by adding daily rainfall and subtracting daily evaporation and drainage. Daily water potentials below field capacity were then summed to give an annual integral. Because of their strongly skewed distribution, values of W were log transtbrmed (base 10) betbre further analysis. Data Data for this study were taken from an extensive survey of New Zealand's forests carried out mostly in the 195O's and 196O's, and involving the measurement of over 15 000 plots, each of 0.4 ha (Masters et al. 1957). Plots for which recent fire and/or logging disturbance were noted at the time of survey were omitted, leaving 10 500 plots sampling the majority of North Island's lowland and montane primary forests, and South • Island's lowland primary forests. A further 4000 plots, each of 0.04 ha, measured during more recent (1970"s and I98()'s) protection (brest surveys carried out by the former New Zealand Forest Service, were added to increase the coverage of South Island montane forests (Leathwiek et al. 1996). Beeause of the problems inherent in diversity measures which attempt to combine information on species abundance as well as speeies richness (e.g.. Whittaker 1972), diversity was measured as simply the number of Iree species present, i.e., species richness. For each plot, the number of tree speeies with individuals > 305 mm in diameter was determined non-spermatophyte speeies, i.e.. treeferns. were excluded because of inconsistencies in their measureHCOCiRAPHY The lithology (L) of each plot, a major determinant of edaphic conditions, was detemiined by overlaying ploi coordinates onto a digital copy of the 1:1 O O 000 geoO logical map of New Zealand (Anon. 1972a. b). Geological substrates were grouped for analysis on the basis of their age, weatherability, and composition (Table Ib). An indication of the drainage (D) at each plot was provided by a subjective assessment tnade at the time of original plot measurement using a ihree-step seale. Finally, a variable (PS) was included to indicate the plot size (0.4 or 0.04 ha) for each site. the response variable as in conventional regression, predictions are first formed on a linearised predictor, which is then back-transformed using a link function, which in this case was of logarithmic forrn. Generalised Additive Models differ from GLMs by defining the relationship between response and continuous predictor variables using non-parametric methods (scatter-plot smoothers), rather than parametric combinations of the predictor variables (e.g., X + X-), Tree richness is calculated, then, by first caleulating: linear predictor = a:-l-s,. + . 4-s,, + L,. 4- D., Analysis For all tree speeies. and for each of the three species subgroups, regressions relating species riehness to cnvirontnent were calculated using Generalised Additive Models (GAMs: Hastie and Tibshirani 1990), an exiension of Generalised Linear Models (GLMs; McCullagh and Nelder 1989, Nieholls 1991). The latter offer considerable advantages over conventional regression techniques in their ability to handle dala with error distributions departing from normality, such as the eount data analysed here for which a Poisson error distribution is assumed (Vincent and Haworth 1983). Rather than forming fitted values directly on the scale of where ct is a constant, Si • • • s,, are smooth functions of the continuous climatic variables (Table la), and L,,, D,, and PS; are multi-level coefficients fitted for the three factor variables (Table lb). The richness for any site is then given by richness = e'""--''^ P.'^a,.;ior For each of the four regression models, all six envit"onmcntal predictors were fitted along with the factor variable indicating plot size, and the significance of removing each predictor in turn was tested using changes in residual deviance which are distributed approximately as for a x^-statistic. Non-significant predictors were removed, and the process repeated until only significant Table I. Bitsic statistics Tor cnvironiiiciUal predictors used in the analysis, ti) continuous predictors: b) category predictors, a) Symbol (units) T (=C) S (MJ m- - day W (log,,, MPa day) M (%) b) Symb. Description mean anntial temperature mean annual solar radiation annual integral of root-zone moisture deficit min. monthly mean 9 a.m. humiclit\ Mean (range) 10.1 (3.4- 15.4) 14.1 (11.8-15.4) 0.7 (0 2..^) 74.4 (58 -89) Dcsc. geology Classification G = granite and gneiss, mainly Paleozoic or older, 804 plots: II = hard sedimentary substrates, predom. greywiieke, argillite. and schists. Cretaceous and older. 6206 plots: S^soft sedimentary substrates, predom. calcareous sandstone, siltstone. PliocenePaleocene, 2591 plots: B = more basic volcanlcs. i.e, andesites and basalts, mainly Pliocene and older, but some Quaternary, 1131 plots: R = acid volcanics, i.e.. rhyolites. mainly Quaternary. 928 plots; Q = Quaternary surfaces, sediments deposited by alluvial. glacial, laharic. and aeolian processes, etc., 2880 plots. G = good: M — moderate: P = poor. N = 0 . 4 ha: Q = 0.04 ha. PS drainage plot size F.CO(}RAI'HY l\ predictors remained. Models were then evaluated by plotting both fitted values and residuals against each predictor, and by contouring predictions of tree species richness at various combinations ol' T and H for comparison with actual values. Spatial predictions of tree richness were then calculated for comparison with the raw data, by using the regression equations in conjunction with estimates of the six environmental predictors for points on a 5 km grid across New Zealand. The effect on btHh broadleavcd and conifer diversity of Nolhofagus presence was tested by adding to their regressions, a category predictor indicating whether Nothofagus species were present or absent for each plot. Results Geographic patterns of alpha-diversity A total of 112 tree species occurred in the entire data set, with mean values for total species richness in the 0.4 ha plots ranging from I to 16. and with a grand mean oi 5.4 (Table 2). Predictably, mean values were lower in the 0.04 ha plots, but this difference alsti reflects differences in the range of environmental characteristics sampled by the two plot sizes; the 0.04 ha plots mostly occurred in the South Island, where both temperature and solar radiation are generally lower than in North Island, Plots with high species richness were located mostly in the northern half of North Island, richness gradually declining with progression to higher latitudes and altitudes. Broadleaved trees were the largest subgroup, with 88 species distributed among 30 families as follows; Winteraceae (2). Lauraceae (3). Monimiaceae (2), Violaceae (2), Onagraceae (1). Proteaceae (2), Coriariaceae (1), Pittosporaceae (3), Myrtaceae (6), Elaeocarpaceae (3), Malvaceae (6). Cunoniaceae (3), Escalloniaceae (5 - treated as 3). Papilionaceae (2), Moraceae {!), Corynocarpaceae (1), Icacinaceae (1), Santalaceae (1), Rutaceae (2), Meliaceae (I). Sapindaceae (1), Araliaceae (7), Cornaceae (2), Epacridaceae (4). Myrsinaceae (2), Oleaceae (3 reduced to 2 for analysis), Rubiaceae (8), Asteraceae (9), Myoporaceae (I), and Verbenaceae (1). The maximum broadleaved Table 2. Sutnmary of raw species riehness data. No. positive 0.4 ha plots All species Broadleaved trees Conifers Nothofagus spp. 14 540 10 640 9400 7396 5.4 3.3 2.1 1.7 richness was 14 on 0.4 ha plots, and one or more species occurred on >70% of plots. The spatial distribution of broadleaved richness followed a similar pattern to that for total species richness, i.e., declining with progression away from northern lowland sites to very low levels or complete absence on sites at high latitudes in the south and east. Conifers were the next largest group, totalling 17 species from the families Podocarpaceae (14 - reduced to 13 for analysis), Cupressaceae (2), and Araucariaceae (1). Sites with a high richness of conifers, although concentrated in central North Island and West Coast. South Island, were also scattered over a wide altitudinal range in the north and at lower altitudes in the south, particularly on sites with poor drainage. Four species of Noilwfagus occurred in the dataset; the two subspecies of N. solandri were unable to be accurately differentiated from the original plot data. .Sites with the highest richness in South Island were concentrated in central parts oi' the southern third of the Island, and from the Paparoa Range northwards in the northwest. Sites with high richness in North Island were concentrated on the southern Hanks of Mt. Ruapehu and in the Ahimanawa, Urewera, and Raukumara ranges. The three remaining species occurring in the dataset were all monocotyledonous trees, two from the Agavaceae and one from the Palmae. Correlations between tree species richness and climate Although comparison oi' the overall success of the regression models is hampered by the lack oi' a statistic analogous to the percentage variance explained as used in conventional sums-of-squares regression (McCullagh and Nelder 1989), residual mean deviances were generally close to I. indicating that a substantial amount of the total deviance had been explained (Table 3). Drainage was dropped from the regressions for total species and Nolhofagus richness, but all six environmental predictors were retained in the broadleaved tree and conifer regressions. Changes in residual deviance when dropping each predictor in turn indicate that of the climatic predictors, mean annual temperature (T) has the strongest contribution Mean where present 0.04 ha plots 2.1 1.9 1.4 1.4 0.4 ha plots 1-16 0-14 0-7 0-4 Range over all plots 0.04 ha piois 1 7 0 6 0 4 0-4 EC(KJRAi>HY 21 to the regressions for total species and conifer richness, while minimum monthly mean humidity (H) has the strongest contribution in the regressions for broadleaved tree and Nolhofagus richness. By contrast, mean annual solar radiation (S) and root zone moisture deticit (W) make a much smaller contribution to the results. Examination of the regression functions lor the four climatic predictors (Fig. 2), and predictions formed in relation to the two most strongly correlated of these (T and H. Fig. 3), indicates that total species richness is predicted to reach a maximutn or\ sites combining high mean annual temperature, high solar radiation, low annual root zone moisture deficit, and high miniiiium monthly mean humidity. Within this overall pattern, however, there are marked differences between the three species groups. Predicted pattems of richness for broadleaved trees closely follow those for total species richness, with predicted richness reaching a maximum on warm, high insolation, humid sites with moderate root zone moisture deficit. In contrast, maximum conifer richness is predicted to occur on sites with intermediate temperatures, high humidity, and low insolation and root zone moisture deficit: maximum richness for Nolhofagus is predicted to occur on sites combining cool temperatures, high insolation and low root zone moisture deficit, but low humidity. However, both the latter two groups have a secondary peak of richness at high root zone moisture deficits. and Quaternary substrates {Fig. 4a). Among the species groups, however, broadleaved tree richness is predicted to be highest on hard sedimentary and basaltic substrates, and lowest on granites and Quaternary substrates (Fig. 4b); conifer richness is predicted to be highest on rhyohtes and basalts, and lowest on granites (Fig. 4c); and Nothofagus richness is predicted to be highest on granites, soft sedimentary and Quaternary substrates, and much lower on rhyolites (Fig. 4d). Broadleaved tree richness is predicted to decline as drainage deteriorates (Fig. 5). but that tor conifers is predicted to increase. Does the presence of JSothofafius inHnence the richness of other species groups? Addition to the bi"oad!eavcd tree and conifer regressions of a variable indicating Nothofagus presence/ absence at each plot, resulted in a highly significant reduction in residual deviance in both regressions, but with a more substantial reduction for broadleaved trees than for conifers (Table 4). Responses to other predictors were little changed for either regression model. Combining responses to T and H from the broadleaved tree model while keeping all other environmental factors constant, reveals the substantial decrease in broadleaved tree riehness which occurs in the pi-esence of Nothofagus (Fig. 6a. b). Correlations between tree species richness, and Hthology and drainage Marked variation in species richness is also predicted to occur in relation to lithology (L) but variation is less marked in relation to drainage (D) (Table 3, Figs 4 and 5). The regression coefficients for lithology indicate that total species richness is predicted to be significantly higher on hard and soft sedimentary substrates and basalts, than on granites, rhyolites. Spatial predictions of richness The predicted spatial distribution of tree species richness from our environmental regressions show mixed success (Fig. 7). Predicted patterns of total species richness and broadleaved richness (Fig. 7a, b) correspond well with observed patterns, i.e., with greatest richness predicted for the warm, high insolation sites of northern New Zealand, and declining with progression Table 3 Summary of regressions relating species richness to environmental predictors. For each environmental predictor, table values indicalc ihe increase in deviance resulting from dropping that predictor from the corresponding regression. These values arc distributed approximately as for a /--statistic. For example, for T, S, II, and W, values are signifieant at the 95% and 99% levels of confidence if >7.81 and 11.34 respectively. Model mean devianee 0.739 1.145 0.949 0.828 -T (3 DP) 607.1 1098.9 719,3 767.7 S (3 DF) 184.9 387.5 60,1 378.1 -W (3 DF) 10.4 47..5 87.3 128.1 -H (3 DF) 570.3 1448.9 240.7 945.7 -L (6DF) 162.9 662.1 113.2 1525,.5 -D (3 DF) -PS (2 DF) 783.2 179.2 647.0 117.8 Nolhofagus ECOGRAPJIY 2I:,1 Fig, 2, Non-paramelric regression functions on a logaritlimic seale for species riclincss of all trees, broadleaved Irees, conifers, and :\oflioja,i'iis spp, in relafion to a) mean annual temperature (T): b) mean annual solar radiation (S); e) annual integral of root zone moisture defieit (W); d) minimum monthly mean humidity (H). a) b) Nothotagus 6 8 !0 12 14 13 14 15 Mean annual temperature Mean annual solar radiation C) d) 0,5 Roo1-;one moisture deficit Minimum monthly mean humidity Fig. 3. Contours of tree species richness in relation to mean annual temperature tT) and minimum monthly mean humidity (H) as predicted from environmental regressions: a) all speeies; b) broadleaved trees; c) conifers; d) Notbofagiis spp. The ratige of combinations of T and H sampled by the diita is bhown by overlaid points in (a). Values for other environmental predictors were set as follows; S 14.5, W 0,7. L hard sedimentary, PS - 0.4 ha. a) b) Mean annual temperature Mean annual temperature d) Mean annual temperature Mean annual temperature E((KiRAPHY 21:? (I'WS) b) Broadleaved tree c) Conifer Fig, 4, Regression coefficients for litliology classes {L see Table lb} from regressions of species richness versus environment for a) all speeies; b) broadlcaved trees; c) eoniiers; and d) iXotlioJagus spp. Widths of Ihe horizontal bars are proportional to the number of plots for each class, and upper and lower bars indicate Iwo times pointwise standard errors. a) Broadieaved trees b) Conifers Fig, 5, Regression coefficients for drainage classes (I) see Table lb) from regressions of species riehness versus environment for a) broadleaved trees, and b) eonilers. Widths of the horizontal bars are propoitional to the number of plots for each class, and upper and lower bars indicate two times pointwise standard errors. to higher latitudes and altitudes. Predictions of conifer richness also agree well with observed patterns (Fig. 7c), with highest levels predicted for rhyolitie substrates on the central North Island volcanic plateau, and tor mainly Quaternary substrates in the western and southern South Island lowlands. In contrast, predictions of Nothofagus richness show much poorer correspotidence with observed patterns (Fig. 6d). Although high levels are correctly predicted for the northwest of Soulh Island, only low levels are predicted for the LJrcwera and Raukumara ranges, where three Notbofagiis speeies regularly occur together, and moderate levels are predicted through large parts of New Zealand in which no Nothofa^^us spp. are present. Notably, the latter includes both Stewart Island and Soulh Island's 'beeeh gap" (==most of the middle third of the island - Wardle 1963), occurs in climates combining relatively high temperatures, high solar radiation, and high atmospheric and soil moisture. This is consistent wilh results trom similar studies elsewhere which have demonstrated that tree richness is generally highest in conditions favourable to high productivity, assessed using cither derived measures such as available energy or actual evapotranspiration, or the climatic and site factors (e.g.. temperature, solar radiation, and moisture availability) which control plant productivity (Richerson and Lum 198(1. Wright 1983. Currie and Paquin 1987, Adams and Woodward 1989, Currie 1991. Austin et al. 1996). Although the decline we dernonstrate in tree speeies richness with decreasing humidity is a novel result, it is consistent with the general hypothesis already outlined, i.e., that diversity declines on sites whieh are unfavourable to tree growth. The absence of many tree species from sites with lower humidity most likely reflects the severe limitations placed on their growth, and perhaps survival, during extreme events sueh as fohn winds (Leathwiek unpubl.). During these, low humidity is often combined with high ternperatures and solar radiation, and strong winds, all of which Increase evaporative demands on trees (Landsberg 1986). Discussion Correlations between climate and tree richness Results from this study indicate that the maximum total species richness Ibr trees in New Zealand's forests 242 Fig. 6. Contours of broadleavcd tree species richness in relation lo mean annual temperature (T) and minimum monthly mean humidity (H) as predicted from environmental regressions to which a variable Itidicating :\'riiliofagm- presence/absenee has beet! added. The range of combinations of T and H satnplcd by the relevant data are showti by overlaid points in both graphs. Values for other environmental variables are set as in Kig. 3. a) Broadleaved trees Nothofagus absent b) Broadleaved trees Nothofagus present 10 Mean annual letnperature 12 Mean annual temperature The consistently high correlation between mean annual temperatitre and total species richness we demonstrate is also in accord with documented gradual declines along gradients of increasing altitude in both woody species richness {Wardle 1984. Ogden 1995) and tree richness (Wardle 1984) for New Zealand forests. The increase in diversity of all species with inereasing altitude reported from southwestern New Zealand by Wilson ct a!. (1990) appears contradictory. However, this probably reflects their grouping together of differing growth forms for analysis, an approach whieh can obscure important correlations because different mechanisms may control the richness of different groups (Hitston 1994), The correlation we demonstrate between species richness and solar radiation contrasts with the relatively low variation in species richness along latitudinal gradients demonstrated by Wardle (1984) and Ogdcn (1995). particularly given the strong correlation between latitude and solar radiation (Table 5). It may reflect a confounding of their analyses, which used latitude alone, by the east to west variation in solar radiation at any latitude due to interaction between predominantly westerly-borne cloud masses and New Zealand's axial mountain ranges. The pronounced variation between species groups demonstrated here is consistent with their documented environmental preferences, i.e., with gradual replacement of broadleavcd species by Nothofagus with progression towards less favourable growing conditions (e.g., Wardle 1964. Leathwiek 1995, Ogden et al. 1996. Leathwiek unpubl.). and by eonifers on Quaternary or rhyolitic substrates (e.g.. Wardle et al. 1983, Leathwiek and Mitchell 1992, Leathwiek 1995). The less satisfactory spatial prediction oX CCOCiRAI'llY 21:3 Noihofagus richness across New Zealand is consistent with conclusions from other studies that the disjunct distributions of all (bur species are poorly correlated with climate, and probably reflect very slow reoccupation of their potential ranges following displacement by major geomorphie disturbances or glaciation (e.g., Wardie 1984, McGlone 1985, Wardle 1991, Leathwiek ct al. 1996. Leathwiek unpubl.). Disturbance and tree species richness Although a lack of quantitative data prevented analysis of the effects of recent disturbance events ( <ca 200-500 yr), some clues as to the importance of older events are apparent from this analysis, particularly eoncerning the effects of large-scale events. For example, the lower total species riehness predicted for Quaternary and rhyolitic substrates (Fig. 3a), both of which have formed as a result oi geologically recent deposition of new parent material, suggests incomplete occupation of these sites by the range of speeies present on older sedimentary and granitic substrates in comparable climates. Some support for this explanation comes from the distribution of speeies groups across these substrates. For example, the Nothofagus species, which are noted both for their limited dispersal ability (e.g., Wardle 1980. Rogers 1989) and for their slowness to invade nonNothofagiis tbrest communities (Ogden et al. 1996), are predieted to reach their highest speeies riehness on granitie substrates (Fig. 3d). Such substrates oecur in parts of New Zealand which have been more geologically stable since the Oligocene (McGlone 1985), and in some cases, were subject to less severe glaciation (Wardle 1963). Conversely. Nothofu}^us richness is pre243 a) Total tree richness b) Broadleaved tree richness c) Conifer richness d) Nothofagus richness Fig. 7. Predicted lotal tree species richness on l).4 ha plots across New Zealand for a grid of points al a 5 km spacing, and with equivalent environmental data as that used to develop the regressions. No predictions are made for areas SIIOWLI in wliite. ;ts their environments tall outside the range sampled by the dala. dieted to be dramatically lower on rhyolitic substrates, which have been subject to a number of large-scale eruptions over the last 50 000 yr (Froggatt and Lowe 1990), despite the demonstrated climatie suitability of these sites (Leathwiek and Mitchell 1992. Leathwiek 1995). Similarly, the higher richness of the predominantly bird-dispersed eonifers on rhyolites (Fig. 3c) probably reflects in part the strong ability of many of these speeies to colonise new mineral soils (MeKelvey 1963). However, their higher diversity on this substrate may also be favoured by the predisposition to fire of the generally flat rhyolite plateau landscapes, because of a lack of topographic barriers to fire spread. Although natural fires are rare in New Zealand, firing has been extensive since Polynesian occupation 900-1000 yr BP (McGlone 1983). and some fire modifieation of the supposedly primary forests that grow there, particularly under lower rainfall regimes, cannot be ruled out. Conversely, the low richness predicted for eonifers on gran244 ites (Fig. 3c) probably reflects ihat the localised disturbance events through mass movement, which they exploit elsewhere (Ogden and Stewart 1995). arc less eommon on these substrates (Rcif and Allen 1988. Allen et al. 1991). However, the confounding effects of more subtle environmental differences between these substrates must also be considered alongside likely disturbance effects. For example, the richness of conifers on rhyolites may relkct not only their good dispersal abihty, but also a degree of exclusion of shade-tolerant but more eold-sensitivc broadleaved competitors from these prcdominiiiitly Hat and frost-prone sites (Leathwiek and Mitchell 1992). Similarly, on Quaternary substrates the lower richness of broadleaved trees and Nothofagus spp. and the higher richness of eonifers probably iti part also reflects their relative responses to edaphie differences, and particularly the poor drainage eonditions whieh develop as a consequence of" intense leaching and the relative absence of periodic rejuvenation F.COCiRM'llY :i:3 (IWK) through local tectonic activity (Wardle ct a!. 1983, Rcif and Allen 1988). Conclusions Both environmental and disturbance paradigms are important for understanding the distribution of tree species richness in New Zealand's primary forests. In common with studies from other countries, tree richness at a broad scale is correlated with - and probably in some manner controlled by - the environmental factors of temperature, solar radiation, and moisture availability in both the atmosphere and soil, which in turn control productivity. However, at large to intermediate spatial scales the efteets of cataclysmic but infrequent disturbances resulting from the dynamic nature of New Zealand's landscape, become more apparent. The effects of recent ( < ca 300 yr) disturbance could not be examined in this study, but are undoubtedly also important, particularly at smaller spatial scales. Although more precise partitioning of the relative effects of environment and this variatitjn in disturbance regime is desirable from a theoretical perspeetive, its analysis is problematic with these data given the difficulty in accurately quantifying recent disturbance histories for large numbers of sampie sites. Aekiunvledgmeiits - A number of people have coiUribiited lo ihe development of llic ideas presented in this paper, and in piirlicular M. P. Austin, ,1. Dale, and G. M. Rogers. The paper benefited greatly from ii review in i(s draft stages by W. Lee. Access Io part of the data used in the analysis was facilitated by L. R, Burrows and G, M. .1. Hull. The project was funded by the New Zealand Foundiitlon for Researeli. Science, and 7"eclinolojiy under Contract no. C09405. The influence of Nothofagus Less direct imprints of historical factors are evident in the negative relationship which we have demonstrated particularly between broadleaved richness and the presence of the patchily distributed Nothofagus spp. (see above). This suggests that on sites within the clitiiatic range of Nothofagus. bi'oadleaved tree richness is controlled not only by envirohment but also by whether Nothofagus species have arrived on site. Where (he lalter occur, the richness of other broadieaved trees can be expected to be significantly reduced, perhaps owing to the superior ability of (he ectomycorrhizal Nothofagus to capture nutrients (Wardle 1984. Leathwick unpubl.). Such ;i reduction in riehness consequent on the presenee ol competitive but poorly dispersed species is fundamental to the dynamic equilibrium model of community structure proposed by Huston (1979, 1994). and has been observed in both tropical (Hart et a!. 1989) and temperate settings (e.g.. Loucks 1970). In contrast, richness of conifers appears to be less aOeeted by the presence of Notbofagus. perhaps a consequence of their greater stature enabling them to coexist in at least some Nothofagus Ibrests as emergents. or of their greater tolerance of the infertile soils which ol'ten develop under a Nothofagus eanopy.

Journal

Ecography – Wiley

Published: Jun 1, 1998

There are no references for this article.