Abstract
Habitat Suitability Indices (HSI) are widely used in conservation and in pre-development surveying. We tested a commonly-used HSI to assess its effectiveness at predicting the presence of a European protected species, the great crested newt Triturus cristatus, at the edge of its range. This HSI is used to understand species’ conservation needs, and in assessing the need for, and designing, mitigation. Given the cost of surveying to developers, it is essential that they can have confidence in the index used in targeting work and in Environmental Impact Assessments. We found that nine of the ten factors which make up the HSI are robust in the region, even in a disjunct population believed to have been isolated for around 3000 years. However, we propose modification of the geographic factor, based upon an improved knowledge of the species’ distribution since the HSI was originally devised.
Introduction
The use of models to help us to understand species’ ecological needs has grown markedly since its beginning in the early 20th century (Breckling et al., 2011). Whilst many of these models are complex and require specialist software, there are also simple models such as the Habitat Evaluation Procedure (HEP; FWS, 1976). This scores habitats for a given species using the Habitat Suitability Index (HSI), built up from a series of indices, each representing a key habitat variable. Such HSI are easy to use, for example by citizen scientists, nature conservation agency staff and students which, in turn, allows them to be used for studies covering large geographic areas or time periods (Brooks, 1997; Burgman et al., 2001). They have been developed for a wide variety of taxa including plants (e.g. Zajac et al., 2015), invertebrates (e.g. Cake, 1983) and all vertebrate classes, including amphibians (e.g. Stevens et al., 2008).
Triturus cristatus, together with its breeding and resting places, are protected under European legislation (Annexes II and IVa of the Habitats Directive, Council Directive 92/43/EEC), which requires member states to designate protected sites and implement a system of strict protection. It is seen as a flagship species for wetland conservation, and a European conservation action plan has been produced for the species (Edgar and Bird, 2006). This species reaches the north-western limit of its global range in Scotland (Wielstra et al., 2014). Populations in southern and central Scotland are connected to the other British populations, but the disjunct Highland population is believed to have been separated for over 3000 years as a result of climate change (O’Brien and Hall, 2012), and shows a level of genetic differentiation consistent with isolation and fragmentation (O’Brien et al., 2015).
Study areas and number of ponds, indicating presence or absence of Triturus cristatus.
In Britain, the HSI for Triturus cristatus was devised by Oldham et al. (2000) and modified by Lee Brady (ARG-UK, 2010). It is part of the standard survey methodology mandated by government conservation agencies e.g. for reporting under Article 17 of the Habitats Directive and in Environmental Impact Assessments (NE, 2015), and is used extensively in Britain and elsewhere in Europe (e.g. Unglaub et al., 2015) in species survey and to guide habitat management (Jehle et al., 2011).
There are inherent risks in any expert system and “predictions must be treated with caution” (Oldham et al., 2000; and see Unglaub et al., 2015). This may be particularly true at the edge of the species’ range where, for example, different habitats might be preferred, or different constraints may be of greater importance than within the core range (Sexton et al., 2009), as has been suggested for this species in the Scottish Highlands (Miró et al., 2017). In addition, studies of other taxa (see discussion in Unglaub et al., 2015) do not always find a significant association between the habitat suitability score and simple species occurrence. This highlights the importance of a clear understanding of what is indicated by HSIs (e.g. suitable foraging or breeding habitat). Given the importance of T. cristatus from a conservation perspective, and the economic implications of newt surveying for development planning, there is ample justification for testing the validity of this HSI, particularly as new indices are developed and used for other species. This work is thus relevant to the protection of the individual species, as well as serving as an example of the importance of calibrating HSIs across a species’ range.
The aims of this study were to evaluate the applicability of the HSI in Scotland, by comparing sites where T. cristatus were known to be present with sites where repeated survey using the same protocols had been unable to locate them. We tested the strength of the relationship between the individual index factors and presence/absence of T. cristatus, and investigated the impact of modifying the geographic index factor to take account of our improved knowledge of the species’ distribution since the original index was published.
Method
Data used
We used three sources of data, giving a total of 481 ponds (table 1) surveyed using the same HSI protocol, training materials and scoring sheets as described in ARG-UK (2010) to examine the strength of the relationship between Triturus cristatus presence/absence and each of the individual index factors therein (table 2).
Eighty-six ponds were surveyed as part of the “National Amphibian and Reptile Recording Scheme” (NARRS) (Wilkinson and Arnell, 2013), a UK-wide citizen science survey programme for widespread species. Ponds were randomly selected from a 5 × 5 km grid centred on each surveyor’s address. This programme also provided data for 107 additional ponds, which were not randomly selected.
Targeted surveys of 200 ponds within 5 km of known Triturus cristatus sites were carried out in 2010 and 2011 under the sponsorship of the government’s statutory nature conservation body Scottish Natural Heritage (Wilkinson et al., 2014).
The third source was a study of T. cristatus in the Scottish Highlands (Miró et al., 2017). This comprised the survey of all 32 known ponds (excluding known introductions) in the Scottish Highlands with T. cristatus records since 1990, together with 55 sites where T. cristatus had not previously been recorded, chosen using a random number generator to select grid references.
Geographic index factor
Index factor 1 (SI1) is based on geographic location [zone A (optimal), zone B (marginal) and zone C (unsuitable)]; central Scotland and the southern coastal lowlands are characterised as zone B and given a value of 0.5, whereas all other areas of Scotland are classified as zone C and given a value of 0.01. Since the original publication of the index, we have developed an improved understanding of T. cristatus distribution in the low-lying coastal area around the Inner Moray Firth in the Highlands (hitherto classed as unsuitable). In addition, recent presence/absence surveys in central Scotland suggest that the boundaries of the established zones should be refined. We therefore varied this weighting for both of these areas.
Based on our findings, and on records of T. cristatus since 1990, we drew minimum convex polygons allowing a 1.5 km margin around the presumed range of the species in Scotland (i.e. approximating the greatest recorded dispersal distance in the species; Haubrock and Altrichter, 2017). We then modelled the data for the Moray Firth (Model 1 – standard HSI; Model 2 – using the modified polygons with SI1 scored as B ‘marginal’, rather than C ‘unsuitable’) and for the whole of Scotland (Model 3 – standard HSI; Model 4 using the modified polygons).
Numerical treatment
Previous work in south-east England (ARG-UK, 2010) has suggested that the HSI scores (Excellent (>0.8), Good (0.7-0.79), Average (0.6-0.69), Below Average (0.5-0.59) and Poor (<0.5)) can be translated into probabilities of ponds supporting T. cristatus. We used these to calculate the theoretical number of ponds likely to be occupied according to each model.
We evaluated the power of the HSI to discriminate between ponds with and without Triturus cristatus using empirical Kernel Probability Density Functions (PDF) and Generalized Linear Models (GLM). Numerical procedures were carried out with R statistical software (R Development Core Team, 2014), using the basic functions and the packages sfsmisc (Maechler, 2015) and pROC (Robin et al., 2011).
Results
Individual index factors
The relationship between each factor and the presence/absence of T. cristatus is shown in table 3. There were significant correlations between six of the factors and T. cristatus. No relationship was found with SI2 pond area, SI6 waterfowl, SI7 fish or SI8 pond density.
Comparison between SI scores according to presence/absence of Triturus cristatus (data for the whole of Scotland). P-values < 0.05 are shown in bold.
Geographic index factor
Using PDF models, we determined that a geographic index value of 0.5 (zone B: Marginal) for ponds in the Moray Firth area better discriminated between ponds with and without T. cristatus, for Scotland as a whole [Model 3 (fig. 1c) in comparison with Model 4 (fig. 1d)]. Using this value, the PDF of Model 4 showed reduced overlap between ponds with and without T. cristatus (Model 3 = 73%; Model 4 = 59%; fig. 1c and d). If Moray Firth ponds are included in zone B, the values of SI1 for ponds with or without T. cristatus were significantly different, whereas using the established zones failed to discriminate between them (table 3).
Empirical Kernel Probability Density Functions (PDF) for (a) Model 1, Moray Firth data, standard HSI, (b) Model 2, Moray Firth data, SI1 modified, (c) Model 3, all Scottish data, standard HSI, (d) Model 4, all Scottish data, SI1 modified. Vertical doted lines indicate the categories of T. cristatus predicted presence depending on HSI values (ARG-UK, 2010): excellent (HSI > 0.8), good (0.7-0.79), average (0.6-0.69), below average (0.5-0.59) and poor (<0.5).
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Empirical Kernel Probability Density Functions (PDF) for (a) Model 1, Moray Firth data, standard HSI, (b) Model 2, Moray Firth data, SI1 modified, (c) Model 3, all Scottish data, standard HSI, (d) Model 4, all Scottish data, SI1 modified. Vertical doted lines indicate the categories of T. cristatus predicted presence depending on HSI values (ARG-UK, 2010): excellent (HSI > 0.8), good (0.7-0.79), average (0.6-0.69), below average (0.5-0.59) and poor (<0.5).
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Empirical Kernel Probability Density Functions (PDF) for (a) Model 1, Moray Firth data, standard HSI, (b) Model 2, Moray Firth data, SI1 modified, (c) Model 3, all Scottish data, standard HSI, (d) Model 4, all Scottish data, SI1 modified. Vertical doted lines indicate the categories of T. cristatus predicted presence depending on HSI values (ARG-UK, 2010): excellent (HSI > 0.8), good (0.7-0.79), average (0.6-0.69), below average (0.5-0.59) and poor (<0.5).
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Newt occurrence was positively related to HSI scores. The GLM for all four models showed P-values < 0.001 (chi square test) and AUC values of: Model 1 = 0.8225, Model 2 = 0.8225, Model 3 = 0.677 and Model 4 = 0.7527. The discrimination power of Models 1 and 2 for the data from the Moray Firth area are identical, since the same modification was made for data for all sites in Model 2. Nevertheless, PDF showed a marked improvement of the fit between HSI values and presence/absence of T. cristatus [Model 1 (fig. 1a) in comparison with Model 2 (fig. 1b)].
Revised geographic zones. Zone A (optimal), SI = 1 (not found in Scotland); Zone B (marginal), SI = 0.5; Zone C (unsuitable), SI = 0.01. Each grid square is 100 km × 100 km.
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Revised geographic zones. Zone A (optimal), SI = 1 (not found in Scotland); Zone B (marginal), SI = 0.5; Zone C (unsuitable), SI = 0.01. Each grid square is 100 km × 100 km.
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Revised geographic zones. Zone A (optimal), SI = 1 (not found in Scotland); Zone B (marginal), SI = 0.5; Zone C (unsuitable), SI = 0.01. Each grid square is 100 km × 100 km.
Citation: Amphibia-Reptilia 38, 3 (2017) ; 10.1163/15685381-00003108
Using the occupancy probabilities (ARG-UK, 2010) predicted that only three ponds in the Moray Firth area should support T. cristatus using the established geographic zones (zone C). In contrast, if this area is included in zone B, 27 ponds would be expected to support the species, much closer to the 32 observed to hold breeding newts.
Discussion
The geographically modified HSI appears to be a good predictor of pond occupancy in Scotland, even when compared to a well-studied area (southern England; Oldham et al., 2000) with much higher rates of T. cristatus occurrence. In any study of this kind, imperfect detection can be a confounding variable (Sewell et al., 2010). This is less likely to be a major issue in the Moray Firth sites, which have been intensively surveyed over six years (2010-2015), using a range of techniques (egg-searching, dip-netting, torching and trapping) with at least three visits per year (Miró et al., 2017). Elsewhere in Scotland, structured surveys will be beneficial in confirming, or further extending, the known range of the species.
Individual index factors
It is, perhaps, surprising that our results show no significant effect of fish presence on presence or absence of newts; other studies have shown them to be highly significant (Beebee, 2007; Jarvis, 2010), although Miró et al. (2017) showed that T. cristatus can coexist with fish when there are predator refuge areas, such as well-vegetated shallows. However, the lack of significance may result from the difficulty of recording fish presence for non-specialists, rather than the absence of an effect per se (Dot Driver pers. comm.), and it is likely that this is a valid indicator but poorly recorded.
Earlier studies have highlighted the importance of source ponds in population persistence (e.g. Halley et al., 1996). The lack of a relationship with pond density may be an artefact of a delayed population response to the overall reduction in pond numbers across the country over the 20th and 21st centuries. Swan and Oldham (1993) found a decline of 38% across Britain since 1945, and a study in the Glasgow area found that 35% of ponds shown on Ordnance Survey maps had disappeared over 12 years, mainly through landfilling with waste, and 48% of those remaining were considered to be threatened (Swan, 1994). This can be expected to lead to isolated populations in areas which previously had more ponds.
The lack of a relationship between waterfowl and T. cristatus presence is harder to explain. Many waterfowl eat plants which are used by newts for egg-laying or as a predator refuge, and the dabbling feeding method of Anas spp. increases turbidity, so they might be expected to be a negative factor. Data for the Highlands found that the most common waterfowl in T. cristatus ponds were, with the exception of mallard Anas platyrhincus, mainly diving species, feeding on fish or invertebrates, such as the grebes Podiceps auritus and Tachybaptus ruficollis and tufted duck Aythya fuligula (Miró et al., 2017). Whilst these birds may well take some newts, particularly larvae, they are less likely to lead to the heavily-browsed plant communities or turbid waters associated with waterfowl in other areas. Unfortunately, we do not have detailed bird species data for elsewhere in Scotland.
The absence of a relationship with pond size may be an artefact of surveyors targeting ponds which they perceive to be a typical size for the species, though further studies of this possible effect may be limited by the relatively small number of T. cristatus ponds in Scotland.
Geographic index factor
The geographic modification to the HSI provides valuable improvements in discrimination and overlap of the empirical Kernel function, and is relatively simple to impart to trainee surveyors. The original study was based on 72 ponds in England (Oldham et al., 2000) and, while it drew from a larger Britain-wide survey (Swan and Oldham, 1993), the coverage of Scotland was still limited. Since 2000 there have been several Scotland-wide surveys of T. cristatus (Wilkinson et al., 2014; O’Brien et al., 2015; Miró et al., 2017), greatly increasing our knowledge of its range. It seems sensible, therefore, to use these data to evaluate the application of the HSI to Scotland and revise the boundary of the geographic zones in central Scotland and the Moray Firth (fig. 2 and Appendix).
Application
The widespread adoption of the T. cristatus HSI by surveyors, and its perceived effectiveness, has led to calls for the development of indices for other species of herpetofauna (e.g. Brady and Phillips, 2012; Limburn, 2012). The advantages of a simple tool to help target scarce resources during a species survey, however, should not blind researchers and conservation agencies to the limitations of the HSI approach. Indeed, their exponents are often the first to point out the potential limitations and weaknesses (Oldham et al., 2000; Brady and Phillips, 2012). The HSI score may indicate factors other than mere occurrence. For example, Unglaub et al. (2015) tested this HSI in Germany, and found that high scores predicted probability of breeding rather than occurrence or survival. Furthermore, combining several (10) indices into a single HSI could potentially be misleading, as a single negative variable (e.g. high fish or waterfowl abundance) could have a major impact on newt presence, abundance or breeding success, even though the overall score may be relatively high. Testing established HSI should, therefore, be seen as an essential process if we are to have confidence in their continued use.
It is testimony to the robustness of the original methodology that it has stood the test of time with only minor modification, and appears to work well at the species north-western limits, even for populations displaying different terrestrial habitat preferences to conspecifics in the core range (Miró et al., 2017). Given the conservation importance of protecting species at the edge of their ranges, particularly those with distinct genetic histories (Lesica and Allendorf, 1995; Eckert et al., 2008), it is vital that nature conservation organisations can be confident in the tools they use. Furthermore, the species’ preference for lowland habitats has often brought it into conflict with developers wishing to build in such areas. Developers, and the agents they engage to survey for European protected species, must similarly have effective tools at their disposal if they are to avoid wasted effort and potentially expensive mitigation work. We believe that our proposed modifications to the Habitat Suitability Index for T. cristatus will be of benefit to both the protection of the species and economic development.
Acknowledgements
We would like to thank the volunteers across Scotland who have contributed to the NARRS Surveys and without whom this work would have been impossible, Scottish Natural Heritage for licensing and funding the T. cristatus surveys and ARC Trust for access to data. We thank Karen Frake and Tina Ross of SNH for preparing the maps. Furthermore, we would like to thank Robert Jehle for his comments on the project design and on the draft paper, as well as his ongoing championing of amphibian conservation research in Scotland. Finally, we thank the editor and reviewers for their helpful comments.
References
ARG-UK (2010): Great crested newt Habitat Suitability Index. ARG UK Advice Note 5, Amphibian and Reptile Groups of the United Kingdom.
Beebee T. (2007): Thirty years of garden ponds. Herpetol. Bull. 99: 23-28.
Brady L.D., Phillips M. (2012): Developing a ‘Habitat Suitability Index’ for reptiles. ARC Research Report 12/06.
Breckling B., Jopp F., Reuter H. (2011): Historical background of ecological modelling and its importance for modern ecology. In: Modelling Complex Ecological Dynamics, p. 29-40. Jopp F., Reuter H., Breckling B., Eds, Springer, Berlin, Germany.
Brooks R.P. (1997): Improving habitat suitability index models. Wildl. Soc. Bull. 25: 163-167.
Burgman M.A., Breininger D.R., Duncan B.W., Ferson S. (2001): Setting reliability bounds on habitat suitability indices. Ecol. Appl. 11: 70-78.
Cake E.W. Jr. (1983): Habitat suitability index models: Gulf of Mexico American oyster. Biological Report 82(10.57), US Fish and Wildlife Service, Department of the Interior, Washington, DC, USA.
Eckert C.G., Samis K.E., Lougheed S.C. (2008): Genetic variation across species’ geographical ranges: the central-marginal hypothesis and beyond. Mol. Ecol. 17: 1170-1188.
Edgar P., Bird D.R. (2006): Action plan for the conservation of the crested newt Triturus cristatus species complex in Europe. Convention on the Conservation of European Wildlife and Natural Habitats.
FWS (1976): Habitat evaluation procedures. Division of Ecological Services, US Fish and Wildlife Service, Department of the Interior, Washington, DC, USA.
Haubrock P.J., Altrichter J. (2017): Northern crested newt (Triturus cristatus) migration in a nature reserve: multiple incidents of breeding season displacements exceeding 1 km. Herpetol. Bull. 138: 31-33.
Jarvis L.E. (2010): Non-consumptive effects of predatory three-spined sticklebacks (Gasterosteus aculeatus) on great crested newt (Triturus cristatus) embryos. Herpetol. J. 20: 271-275.
Jehle R., Thiesmeier B., Foster J. (2011): The Crested Newt: a Dwindling Pond-Dweller. Laurenti, Bielefeld, Germany.
Lesica P., Allendorf F.W. (1995): When are peripheral populations valuable for conservation? Conserv. Biol. 9: 753-760.
Limburn B. (2012): Conference report: developing a Habitat Suitability Index for the smooth snake (Coronella austriaca) in Dorset. Herpetol. J. 22: 4.
Maechler M. et al., (2015): sfsmisc: utilities from ‘Seminar fuer Statistik’ ETH Zurich. R package version 1.0-28. http://CRAN.R-project.org/package=sfsmisc.
Miró A., O’Brien D., Hall J., Jehle R. (2017): Habitat requirements and conservation needs of peripheral populations: the case of the great crested newt (Triturus cristatus) in the Scottish Highlands. Hydrobiologia. DOI:10.1007/s10750-016-3053-7.
NE (2015): Standing advice for local planning authorities who need to assess the impacts of development on great crested newts. Natural England. https://www.gov.uk/guidance/great-crested-newts-surveys-and-mitigation-for-development-projects. Accessed 21st January 2017.
O’Brien C.D., Hall J.E. (2012): A hypothesis to explain the distribution of the great crested newt Triturus cristatus in the Highlands of Scotland. Herpetol. Bull. 119: 9-14.
O’Brien C.D., Hall J.E., Orchard D., Barratt C.D., Arntzen J.W., Jehle R. (2015): Extending the natural range of a declining species: genetic evidence for native great crested newt (Triturus cristatus) populations in the Scottish Highlands. Eur. J. Wildl. Res. 61: 27-33.
Oldham R.S., Keeble J., Swan M.J.S., Jeffcote M. (2000): Evaluating the suitability of habitat for the great crested newt (Triturus cristatus). Herpetol. J. 10: 143-155.
R Development Core Team (2014): R: a Language and Environment for Statistical Computing, Version 3.1.2. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/.
Robin X., Turck N., Hainard A., Tiberti N., Lisacek F., Sanchez J.-C., Müller M. (2011): pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics 12: 77.
Sewell D., Beebee T.J.C., Griffiths R.A. (2010): Optimising biodiversity assessments by volunteers: the application of occupancy modelling to large-scale amphibian surveys. Biol. Conserv. 143: 2102-2110.
Sexton J.P., McIntyre P.J., Angert A.L., Rice K.J. (2009): Evolution and ecology of species range limits. Annu. Rev. Ecol. Evol. Syst. 40: 415-436.
Stevens S.D., Page D., Prescott D.R.C. (2008): Habitat suitability index for the northern leopard frog in Alberta: model derivation and validation. Alberta species at risk report 132, Alberta Sustainable Resource Development, Fish and Wildlife Division.
Swan M.J.S., Oldham R.S. (1993): National amphibian survey. English Nature Research Report 38.
Swan M.J.S. (1994): Glasgow amphibian survey. Scottish Natural Heritage Survey 484, Clydebank.
Unglaub B., Steinfartz S., Drechsler A., Schmidt B.R. (2015): Linking habitat suitability to demography in a pond-breeding amphibian. Front. Zool. 12: 9.
Wielstra B., Sillero N., Voeroes J., Arntzen J.W. (2014): The distribution of the crested and marbled newt species (Amphibia: Salamandridae: Triturus) – an addition to the New Atlas of Amphibians and Reptiles of Europe. Amphibia-Reptilia 35: 376-381.
Wilkinson J.W., Arnell A.P. (2013): NARRS report 2007-2012: establishing the baseline. ARC Research Report 13/01.
Wilkinson J.W., Arnell A.P., Driver D., Driver B. (2014): Elaborating the distribution of the great crested newt in Scotland (2010-2011). Scottish Natural Heritage Commissioned Report 793.
Zajac Z., Stith B., Bowling A.C., Langtimm C.A., Swain E.D. (2015): Evaluation of habitat suitability index models by global sensitivity and uncertainty analyses: a case study for submerged aquatic vegetation. Ecol. Evol. 5: 2503-2517.
Appendix
List of 10 × 10 km British National Grid squares (Ordnance Survey) in Scotland where HSI factor 1 should be considered ‘Marginal’ SI1 = 0.5. Records from NBN and Highland Biological Recording Group with erroneous records and known introductions removed.
NH45, NH54, NH55, NH64, NH74, NH84, NH85, NH91, NH95, NH96
NJ05
NN50, NN60
NO11, NO42
NR72
NS21, NS39, NS46, NS48, NS56, NS57, NS58, NS66, NS67, NS68, NS76, NS77, NS79, NS87, NS88, NS95, NS97
NT06, NT07, NT09, NT17, NT25, NT26, NT28, NT29, NT36, NT42, NT51, NT52, NT57, NT58, NT67, NT94, NT95
NX15, NX26, NX33, NX34, NX35, NX36, NX43, NX44, NX45, NX46, NX55, NX64, NX65, NX66, NX67, NX68, NX75, NX76, NX89, NX95, NX98
NY00, NY16, NY26, NY36
Footnotes
Associate Editor: Francesco Ficetola.