Biodiversity and Conservation 11: 2217–2238, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
Assessing New Zealand fern diversity from spatial predictions of species assemblages A. LEHMANN∗ , J.R. LEATHWICK and J.McC. OVERTON
Manaaki Whenua – Landcare Research, Private Bag 3127, Hamilton, New Zealand; ∗ Author for correspondence: Present address: Swiss Centre for Faunal Cartography, Terreaux 14, CH-2000 Neuchâtel, Switzerland (e-mail:
[email protected]; fax: +41-32-7177969) Received 15 June 2001; accepted in revised form 25 March 2002
Abstract. The utility of explicit spatial predictions for biodiversity assessment is investigated with New Zealand fern flora. Distributions of 43 species were modelled from climatic and landform variables and predicted across New Zealand using generalised additive models (GAM). An original package of functions called generalised regression analysis and spatial prediction (GRASP) was developed to perform the analyses. On average, for the 43 models, the contributions of environmental variables indicate that mean annual temperature is the most important factor at this broad regional scale. Both annual solar radiation and its seasonality had higher correlations than temperature seasonality. Measures of water availability such as ratio of rainfall to potential evapotranspiration, air saturation deficit and soil water deficit presented significant contributions. Lithology was a better predictor than slope and drainage. These results are similar to those obtained from analyses of the distributions of New Zealand tree species and are consistent with the hypothesis that both tree and fern diversity are highest on sites conducive to high productivity. In order to identify hotspots of fern diversity, spatial predictions of individual species were summed up. The resulting map gave a very similar result to the direct prediction of their corresponding richness (number of species by plot out of 43 spp.). As a consequence, and where individual species models were not all available, the number of species within different species assemblages was directly modelled. Predicted richness hotspots of total species (out of 122 spp.), selected species (out of 43 and 21 spp.) and common species (out of 23 spp.) present very similar spatial patterns and are highly correlated. Richness of uncommon species (out of 39 spp.) was also accurately predicted, but presented a different spatial pattern. The number of rare species (out of 60 spp.) was not correctly modelled. Even though the lack of data for rare species clearly limits the application of this approach, fern community composition of more common species can be partially reconstructed from individual species predictions. This case study offers therefore a consistent approach not only for biodiversity hotspots identification, but also for setting targets to biodiversity assessment and restoration programs. Key words: Biodiversity hotspot, Climate, Generalized additive models, Generalized regression analysis and spatial predictions, Geographic information systems, Species distribution modelling Abbreviations: GAM – generalized additive model; GIS – geographic information systems; GRASP – generalized regression analysis and spatial predictions. Nomenclature: Brownsey and Smith-Dodsworth (1989).
2218 Introduction Biodiversity hotspots Understanding both the geographic and environmental distributions of species is central to ecology. With the increasing awareness of human impacts on both species survival and habitat integrity, ecologists are now asked to not only understand, but increasingly to predict habitat suitability (Grime 1988; Austin and Graywood 1994; Keddy 1994; Tilman 1994), in order to design effective conservation and research programs (Margules and Redhead 1995; Austin 1999). One common aim is to identify and protect areas that offer the likely highest contribution of biological diversity in the future (Margules et al. 1988; Prendergast et al. 1993; van Jaarsveld et al. 1998). These biodiversity hotspots are generally characterised by attributes such as high species richness, and high numbers of endemic, rare or threatened species. In a recent review, Reid (1998) identified two priorities in biodiversity conservation: (i) the identification of indicator species or species groups that describe patterns of the greater part of biodiversity for which data is lacking or not readily acquired, and (ii) the development of optimal methods for analysis of hotspot information in setting conservation priorities. There has been considerable interest in biodiversity hotspot identification and methodologies for prioritisation of conservation actions such as reserve network design (Margules et al. 1994; Williams et al. 1996; Barnard et al. 1998; Howard et al. 1998; Jaffre et al. 1998; van Jaarsveld et al. 1998). Margules and Nicholls (1994) defined conservation value of a site in a region as the contribution made to sampling regional biological diversity. In the lack of exhaustive biological data, this definition leads to compromises. It stresses out the notion of representativity, where even a species poor site, if it contributes a sample of regional diversity not contributed by many other sites, has a high conservation value. Hotspots are often spatially identified by counting occurrences of taxomic groups within arbitrary grid cells (Prendergast et al. 1993) or administrative boundaries (Dobson et al. 1997). However, most studies have been constrained by reliance on imprecise estimates of species distributions, as provided by patchily distributed point data from field survey or biological collections such as herbaria. The approach demonstrated in the present paper is quite different. It takes advantage of recent developments in statistical modelling (Bio et al. 1998; Franklin 1998; Guisan et al. 1998; Leathwick 1998; Lehmann 1998) to combine a large georeferenced dataset describing species distribution with explanatory environment variables. Species occurrences are first modelled in environmental space, and then predicted in geographic space as species potential distributions. From species models to species assemblages Although these techniques are being increasingly used by vegetation ecologists, few studies integrate species-level predictions to an assemblage-level. In one such exam-
2219 ple from Ecuador, palm richness was estimated by summing spatial predictions of species occurrence based on climatic factors (Skov and Borchsenius 1997). In a more extensive project, biogeographic regions in Tasmania were defined using principal component analysis to summarise the information contained in spatial predictions for a range of tree, bird, frog, reptile, and mammal species (Peters and Thackway 1998). Summed species’ predictions are used in our approach, together with cluster analysis and direct models of species richness, to assess spatial patterns in New Zealand’s fern biodiversity. This approach is consistent with the one advocated recently by Pausas and Austin (2001) who insist on using direct multivariate and non-linear environmental gradients to model species richness. Examples of such methods are becoming more and more frequent in the literature (e.g. Margules et al. 1987; Pausas 1994; Austin et al. 1996; Leathwick et al. 1998; Guisan et al. 1998). A similar approach can be found in the Gap Analysis Project (e.g. Scott et al. 1993; Scott and Jennings 1997) which also combined species distribution maps to calculate richness and find biodiversity hotspots. However, the nature of the spatial predictions in the Gap Analysis is based on inductive habitat suitability models from expert knowledge, instead of deductive ones from observational data. Fern ecology Ferns once dominated terrestrial plant communities over the entire globe (Vogel et al. 1999) and are still present in most terrestrial ecosystems (Odland et al. 1995). However, only a few attempts have been made to model fern distributions from environmental characteristics (Pakeman and Marrs 1996), or to investigate broad scale patterns in fern diversity. Most studies of ferns have taken place in humid tropical regions where more species are found (Poulsen and Nielsen 1995; Lwanga et al. 1998). Ecological studies tend to focus on one or a few species and generally at a fine geographical scale (e.g. Kelly 1994; McMaster 1994; Alonso-Amelot and RodulfoBaechler 1996; Russell et al. 1998), or examine subsets of species such as epiphytes (Hietz and Briones 1998) or rare taxa (Vazquez and Norman 1995). Biodiversity indicators of New Zealand fern flora Although ferns are a notable feature of the New Zealand flora (Brownsey and SmithDodsworth 1989), little has been published either on their ecology or on regional patterns of diversity (Parris 1976; Brownsey 2001). With 173 species, nearly half of which are endemic, ferns represent about 7.5% of the New Zealand indigenous vascular flora (total richness approx. 2300 species – Wardle 1991). Ferns are surprisingly well represented in New Zealand forest ecosystems given the relatively low average temperature of these islands. They are characterised by a diversity of growth forms, which range from small and delicate epiphytes to tall and robust trees. Even though Cyathea dealbata (silver fern) is considered as a national emblem, ferns in general
2220 receive relatively little attention from ecologists when compared to herbaceous vegetation, shrubs and trees in New Zealand. Many questions on fern ecology are therefore still open, e.g. the relative role of environment versus interspecific competition in structuring fern communities. In addition, Pteridophytes present interesting features that make them good candidates for spatial prediction from broad scale environmental factors: (i) they share a reproductive strategy based on high dispersal capabilities of spores and aquatic mobility of gametes; (ii) they are represented throughout New Zealand from 34 to 47◦ of latitude south, and from sea level to more than 1500 m of altitude (Wardle 1991); and, (iii) they exhibit marked variation in spatial pattern among species (Brownsey and Smith-Dodsworth 1989). Ferns are also good candidates with which to address important issues related to the monitoring and restoration of biodiversity (Pearson 1995), including: Can a selected subset of species be used to indicate regional fern diversity? How does fern diversity compare to tree and shrub diversity (Overton et al. 2000)? Or, can ferns act as a surrogate for diversity of other groups such as lichens and bryophytes (Pharo et al. 1999), or even faunal groups (Crisp et al. 1998)? Our study addresses the first of these questions by exploring the biodiversity and ecological information value of spatial predictions of assemblages for subsets of fern species.
Material and methods Fern data A subset of 19 875 ‘Reconnaissance’ Plots (RECCE plots, approx. 20 × 20 m, Allen 1992) was selected from the National Vegetation Survey Database (NVS, Wiser et al. 2001), which comprises more than 80 thousands plots. The selected plots were exclusively coming from primary indigenous forests. These plot data were mostly gathered in the 40-year period from 1950–1990 and contained occurrences for 122 fern taxa. The majority of taxa were recorded to species level, but 22 were recorded or subsequently interpreted at genus level due to uncertainties in taxonomic resolution. Having extracted presence/absence data for all fern taxa, we then selected 43 species on the basis of their high frequency in plots (representing 70% of presence observations), and likely robust identification during description of field plots (Tables 1 and 2, Figure 1). A more conservative subset of 21 species was subsequently selected on the basis of more conservative taxonomic criteria (P. Bellingham, personal communication), principally to address concerns about the influence of species misidentification on our results: these latter species comprised 32% of observations in the dataset. In the remainder of our analysis we test the ability of these two subsets to be used as potential indicators or surrogates of distributional and richness patterns for the entire forest fern flora. We compared spatial predictions of: (i) all 122 taxa; (ii)
2221 Table 1. ‘Common’ and ‘uncommon’ categories of fern species based on frequency of occurrence (respectively >10%, between 10 and 1%) in plots in New Zealand’s primary indigenous forests. Indicators
Count
Common species
*BLEDIS
9204 9198 8973 7595 7319 6675 6312 4865 4652 4367 4212 3559 3529 3104 3072 2984 2756 2519 2423 2265 2194 2080 1989
Blechnum discolor Grammitis billardierei Asplenium flaccidum Blechnum novae-zelandiae Cyathea smithii Phymatosorus pustulatus Dicksonia squarrosa Blechnum fluviatile Polystichum vestitum Asplenium bulbiferum Hymenophyllum demissum Hymenophyllum multifidum Trichomanes reniforme Cyathea dealbata Histiopteris incisa Blechnum procerum Rumohra adiantiformis Leptopteris superba Asplenium polyodon Blechnum chambersii Lastreopsis hispida Leptopteris Hymenophyllum flabellatum
ASPFLA BLENOV CYASMI *PHYPU DICSQU *BLEFL ASPBUL
*TRIREN CYADE *HISINC *RUMA LEPSUP *ASPPO BLECHA LASHIS LEPHYM
Indicators Count Uncommon species
HYMDIL
*BLEFIL HYMFE CYAME *ASPOB *STICUN *PHYSC *LYGAR TRIVEN *BLENI *PYREL *HYPMI *BLECO
*PNEPE DICLAN *PAESC *BLEFR *ANALA HYMLY POLRIC
TRISTR
DICFIB HYPRUF Total
105 846 75%
Total
1922 1752 1435 1422 1414 1398 1358 1245 1205 1204 1164 1127 1105 1021 975 904 886 861 787 734 729 656 573 546 536 477 410 389 389 375 369 340 332 261 258 254 237 237 202
Ctenopteris heterophylla Hymenophyllum revolutum Hymenophyllum dilatatum Lindsaea trichomanoides Cyathea colensoi Blechnum filiforme Blechnum penna-marina Hymenophyllum Hymenophyllum Cyathea medullaris Asplenium oblongifolium Sticherus cunninghamii Phymatosorus scandens Lygodium articulatum Trichomanes venosum Blechnum nigrum Pyrrosia elaegnifolium Hypolepis millefolium Blechnum colensoi Hymenophyllum scabrum Hymenophyllum rarum Pneumatopteris pennigera Dicksonia lanata Paesia scaberula Blechnum fraseri Anarthropteris lanceolata Hymenophyllum lyallii Pteridium esculentum Polystichum richardii Hymenophyllum flexuosum Blechnum vulcanicum Cyathea cunninghamii Trichomanes strictum Hymenophyllum Pteris macilenta Gleichenia dicarpa Dicksonia fibrosa Hypolepis rufobarbata Lastreopsis glabella
31 489 23%
Selected indicator species are identified in the table by six letter codes. * indicates 21 indicator species with likely higher identification accuracy. Source of data is the National Indigenous Vegetation Survey database administered by Landcare Research.
‘common species’, defined as being present in more than 10% of the plots (23 spp., 75% of observations); (iii) ‘uncommon species’, present at between 10 and 1% of the plots (39 spp., 23% of observations), and (iv) ‘rare species’, present in less than 1% of the plots (60 spp., 2% of observations) (Tables 1 and 2).
2222 Table 2. ‘Rare’ fern species identified on the basis of frequency of occurrence (