Tree Core Research Zones in Tasmania:Spatial Analysis of Rainforest Conifer Species

Report to Resource Management and Conservation Division and Parks and Wildlife Service, Department of Primary Industries, Parks, Water and EnvironmentNick FitzgeraldNovember 2012

Background and ObjectiveDendrochronology research in Tasmania primarily involves obtaining tree ring samples by coring living trees of four rainforest conifer species. Cores and discs are also sampled from dead trees and logs of these species. Most sampling occurs on land managed by the Tasmanian Parks and Wildlife Service, includign the Tasmanian Wilderness World Heritage Area.This project aims to assist the land managers and permitting authority, PWS and RMC, in making decisions about the issuing of permits for dendrochronology sampling. The process is to create spatial layers identifying known dendrochronology research sites and zones of critical habitat for endemic rainforest conifer tree species, culminating in a spatial layer of areas where dendrochronology will not be permitted unless there is a compelling case.Four conifer species are included in the project brief: Athrotaxis cupressoides, A. selaginoides, Lagarostrobos franklinii, Phyllocladus aspleniifolius. These are fire-sensitive rainforest trees, all of which are endemic to Tasmania. These species have been and continue to be the target of dendrochronology research. Tree coring for dendrochronology mostly occurs in reserves Besides these four species of rainforest conifer there is an inter-specific hybrid, A. Xlaxifolia. There are only 35 records for A. Xlaxifolia in the Natural Values Atlas. Since such a small sample precludes useful spatial analysis and considering that this hybrid always co-occurs with the two Athrotaxis species it is assumed that critical habitat will be captured in the analysis of the parent species.In consultation with DPIPWE ecologists a spatial ruleset was devised to identify areas of particularly high conservation significance for each of the four conifer species. These zones will provide guidance to land management agencies in the permitting of tree ring sampling. The approach used is to identify the geographic extremes of the current distribution for each species, which are likely to be particularly vulnerable to change, as well as likely future refugia.AbbreviationsASR – annual solar radiationCTP – celery top pine (Phyllocladus aspleniifolius)DPIPWE – Department of Primary Industries, Parks, Water & EnvironmentHP – huon pine (Lagarostrobos franklinii)ITRDB – International Tree Ring DatabaseKBP – king billy pine (Athrotaxis selaginoides)NVA – Natural Values AtlasPP – pencil pine (Athrotaxis cupressoides)TCRZ – Tree Core Research ZoneTPI – topographic position index

2

1. Identifying Critical Habitat for Rainforest Conifers in Tasmania

1.1 Geographic range limits for each species1. For each species (except CTP) select the Tasveg communities dominated by that species (see Table 1) from the Tasveg 2.0 dataset and export to a new shapefile.2. Visually inspect Tasveg mapping for outlying polygons and verify or delete these (e.g. Tasveg 2.0 has RPF and RHP polygons in North-East Tasmania which are clearly erroneous).3. Download observations from Natural Values Atlas for each species and import into ArcGIS.4. Select records with Position Accuracy >1000 m and buffer these points using the value for Position Accuracy as the radius to create circular polygons of possible location.5. Select records by species where Position Accuracy ≤1000 m and buffer these points using the value for Position Accuracy as the radius to create circular polygons of possible location.6. Visually inspect outliers and range limits to identify erroneous or dubious records, remove these to derive Accepted Records.7. Combine buffered points (excluding dubious records) with mapped vegetation communities for each species to create Known Presence layer for the species .8. Create Minimum Convex Hull (also called Min Convex Polygon) to approximate Known Range.9. Compare Hull polygon with all buffered point observation records – if any observations are outside the Hull (i.e. they are outlying records with Position Accuracy >1000 m or they are centred within the Hull but uncertainty extends them beyond) flag these for record verification.10. For record verification, refer to NVA record details such as Recorder and Location fields, this may provide information which allows the Location or Location Accuracy to be improved. Where these records are old and/or have poor spatial accuracy and the area is relatively accessible it is assumed that the records are outdated or inaccurate.11. Adjust the Convex Hull if necessary following record verification.12. Create a buffer 5 km inside the Known Range; the resulting polygon is the Core Range for the species and any occurrences outside this polygon are Range Limits.

Table 1. Tasveg mapping units which indicate the presence of each target conifer species.Species Tasveg CommunitiesAthrotaxis cupressoides (PP) RPF, RPW, RPPAthrotaxis selaginoides (KBP) RKP, RKF, RKSLagarostrobos franklinii (HP) RHPPhyllocladus aspleniifolius (CTP) n/aPoor spatial accuracy of observations can be a problem if the record is potentially representative of a range limit or outlying population. For example, 76 of the 204 NVA records for A. cupressoides have an accuracy of 1,000 metres or poorer including many herbarium records at 2,000 or 20,000 m.For the 4 species, 458 out of the 3494 NVA records have position accuracy >1000 m. For the hybrid

A. Xlaxifolia there are 35 records of which 32 have an accuracy of 1000 m or better. 3

Table 2. Summary of distribution data used for spatial analysis.SPECIES TOTAL AREA

TASVEG (ha)MEAN AREA TASVEG POLYGON (ha)

Accepted NVA records

SAMPLE POINTS FROM TASVEG POLYGONSHP 13831 20 124 975KBP 28627 18 293 2167PP 24290 17 145 2054CTP - - 2442

1.2 Altitudinal Limits1. This process samples elevation values from a 25m DEM corresponding to the mapped presence of each species.2. Tasveg polygons and NVA records are combined into a single point coverage by creating randomly sampled points within each Tasveg polygon and combining these with the Accepted Records NVA points.3. This assumes unbiased observations of species across the elevation range, however there should be sufficient observations to constitute a reasonable sample size, moreover the inclusion of Tasveg communities (except for CTP, which has no Tasveg mapping but does have a greater number of NVA observations than the other species) adds more certainty that the core elevation range is representative of the species.4. Extract altitude value from tasDEM25 for each sample point. Get Statistics for the elevation values (see Table 2).5. Determine Altitudinal Range for each species based on minimum and maximum values.6. Core Altitudinal Range is here defined as the range between 2 standard deviations either side of the mean elevation.7. Altitudinal Limits are then the parts of a species' altitudinal range more than 2 standard deviations from the mean elevation.8. For each species the Core Altitudinal Range is derived from TasDEM25 and converted to a polygon.

Table 3. Altitude statistics for 4 conifer species (metres above sea level).SPECIES POINTS MIN MAX RANGE MEAN STD -2 STD +2 STDCTP 2441 1 1267 1266 404 278 -152 960KBP 2514 2 1296 1294 760 207 346 1174PP 2296 354 1438 1084 1142 118 906 1378HP 434 5 1010 1005 230 256 -282 7421.3 Climatic LimitsThere are a wide range of species distribution models (SDMs) available which utilise statistical methods to model the ecological niche or habitat suitability for a species in relation to environmental variables (Franklin 2009). The climatic niche for a species can be modelled by mathematically relating point records for the species with climate layers in a GIS. Modelled bioclimatic envelopes are sensitive to the accuracy of the distributional data used and the climatic limits may be underestimated if the observations for a species do not cover the entire geographic range (Hughes 2003). 4

Read and Busby (1990) suggest that low summer rainfall is the primary limitation for A. selaginoides based on bioclimatic modelling, while their physiological research suggests high summer temperatures are directly limiting, at least at lower elevations where rainfall is adequate. Other rainforest conifer species are likely to respond to similar limiting climatic factors.Whilst the difficulty of interpreting the climatic niche is compounded by the substantial influence of fire and slow dispersal on the realised niche and the possibility that present distributions of conifer vegetation may reflect past climatic events (Read & Busby 1990), given sufficient distribution records it should be possible to determine climatic limits, assuming that many locations the species distribution will reflect climatic limits. Good distribution data is available for the target species, with Tasveg mapping at 1:25 000 scale, combined with NVA records which appear to comprehensively cover the range of the species.Boosted Decision Trees and Maximum Entropy performed best in comparisons of several widely used modelling methods (Elith et al. 2006; Guisan et al. 2007). These methods cope well with complex and interacting factors and noise in the data (Elith et al. 2006). The online Atlas of Living Australia provides a powerful tool which uses Maximum Entropy (MaxEnt) modelling (Phillips et al. 2006) to create a predictive surface based on known records for a species and a suite of environmental variables. Bioclim provides a suite of 35 climatic variables which are relevant to ecological modelling (Nix 1986) and are available in the ALA at a useful resolution for reflecting species distributions (1km2).1. In order to combine Tasveg mapping (polygons) and NVA observations (points), the sampling points used for the elevation analysis are used for input to the ALA (i.e. a combination of accepted NVA records and randomly sampled points within Tasveg polygons – see Section 1.2).2. Import location points for each species into ALA [http://spatial.ala.org.au].3. Define an 'active area' for each species – this is the geographic extent of the MaxEnt modelling. Create an active area which includes all of Tasmania, excluding King and Furneaux islands).4. Use Predict tool to perform MaxEnt model for each species using all 35 Bioclim variables.5. Choose the most effective and least correlated predictors for each species based on the MaxEnt results. Two variables are sufficient for each species.6. Use Predict tool to perform MaxEnt model for each species using the selected variables (Table 3).7. Import prediction surface into ArcGIS and convert into percentiles by area for Tasmania. The highest 10% of predicted probabilities of occurrence based on the climatic model are then designated as Core Climatic Range for each species.8. Huon pine has specific habitat preferences that are not related to climate so climatic modelling for Huon pine was deemed ineffective and therefore no Core Cliamtic Range ahs been defined for this species.The assumption here is that climatic factors are an important determinant of the species range, which is likely to be true for these species since they are not substrate-specific and although they may have a distribution limited by migration ability, fire history and competition, the distribution is still likely to include sites with a representative range of suitable climates.

5

Table 4. Bioclim variables used in the initial analyses, with selected variables for final analyses shown for each species.CODE CLIMATE VARIABLE CTP KBP PPBIO1 Annual Mean Temperature BIO2 Mean Diurnal Range(Mean(period max-min)) BIO3 Isothermality (BIO2/BIO7)BIO4 Temperature Seasonality (ANUCLIM C of V) BIO5 Max Temperature of Warmest PeriodBIO6 Min Temperature of Coldest PeriodBIO7 Temperature Annual Range (P5-P6) BIO8 Mean Temperature of Wettest Quarter XBIO9 Mean Temperature of Driest Quarter BIO10 Mean Temperature of Warmest Quarter X XBIO11 Mean Temperature of Coldest Quarter BIO12 Annual Precipitation XBIO13 Precipitation of Wettest Period BIO14 Precipitation of Driest Period BIO15 Precipitation Seasonality(Coefficient of Variation)BIO16 Precipitation of Wettest Quarter BIO17 Precipitation of Driest Quarter BIO18 Precipitation of Warmest Quarter BIO19 Precipitation of Coldest Quarter BIO20 Annual Mean Radiation BIO21 Highest Period Radiation BIO22 Lowest Period Radiation BIO23 Radiation Seasonality (Coefficient of Variation) BIO24 Radiation of Wettest Quarter BIO25 Radiation of Driest Quarter BIO26 Radiation of Warmest Quarter BIO27 Radiation of Coldest Quarter BIO28 Annual Mean Moisture Index BIO29 Highest Period Moisture Index BIO30 Lowest Period Moisture Index XBIO31 Moisture Index Seasonality (Coefficient of Variation) BIO32 Mean Moisture Index of Highest Quarter MIBIO33 Mean Moisture Index of Lowest Quarter MI BIO34 Mean Moisture Index of Warmest Quarter XBIO35 Mean Moisture Index of Coldest Quarter

6

1.4 Core RangeThe Core Range for each species is determined by intersection of the three layers: Core Geographic Range, Core Altitudinal Range, Core Climatic Range (the first two layers only are used for Huon pine and celery top pine, since climatic models were deemed to be less useful for these species).1.5 Fire RefugiaFire refugia are sites that are subject to much lower fire frequency than the surrounding landscape. These are typically determined by landscape features which influence fire spread and intensity (Wood et al. 2011). The likelihood of fire ignitions also varies spatially and therefore influences the distribution of fire in the landscape. However given the widespread distribution of both natural and anthropogenic fire starts the locations most likely to avoid fire over the long-term will be topographically protected sites.Fire refugia are important for maintaining highly fire sensitive flora and vegetation, such as rainforest conifers. Areas topographically protected from fire are likely to maintain relatively low fire frequencies and intensities compared to future changes in regional fire regimes driven by climatic factors. A Tasmanian Government project is currently developing a series of spatial layers to represent different elements of fire refugia.Given the biogeographic differences between the four rainforest conifers different parts of the landscape will function as fire refugia for each species. For example, pencil pine occurs largely on elevated sites with low relief and field observations suggest that Sphagnum bogs, blockstreams and cliffs support stands of this species by providing protection from fire. Huon pine occupies river valleys where the most fire-protected sites will be gorges and sites with high moisture levels, such as floodplains.Table 5. Spatial definition of fire refugia for rainforest conifer species.SPECIES FIRE REFUGIA LAYER CRITERIAA. cupressoides Island refugia (lakes) all islands > 2 haA. cupressoides Sphagnum bogs Tasveg MSP >1 haA. cupressoides Boulderfields and cliffs Tasveg ORO >2 haA. selaginoides Island refugia (lakes) all islands > 2 haA. selaginoides Annual Mean Solar Radiation below -2 std dev from mean, >50 haL. franklinii Annual Mean Solar Radiation below -2 std dev from mean, >50 haL. franklinii Topographic Position Index (200m) below -2 std dev from mean, >20 haL. franklinii Floodplains theLIST hydarea 'floodplains'P. aspleniifolius Island refugia (ocean) all islands > 2 haP. aspleniifolius Island refugia (lakes) all islands > 2 haP. aspleniifolius Annual Mean Solar Radiation below -2 std dev from mean, >50 ha1. To determine the lowest solar radiation and topographic position values within the distribution of each species the Known Distribution layer is used to mask the statewide 7

raster layer and then statistics are calculated for that combination of topographic variable and species distribution.2. A 50 ha minimum patch size is used to select the larger areas of potential fire refugia. Smaller patches will function as fire refuges in some situations but are less likely to be viable in the long term and are less practical for management purposes (e.g. mapping hundreds of small refuges; fire suppression).3. Islands within waterbodies and offshore islands are derived from 1:25 000 topographic mapping.4. The combination of the identified layers results in a Fire Refugia layer for each species.Solar radiation is used here to identify the least insolated parts of the landscape, such as gullies, gorges and steep south-facing slopes. These are likely fire refugia because they provided topographic protection from fire and maintain relatively high fuel moisture. Topographic position index (Beier & Brost 2009) is used to identify the deepest gullies and gorges.It is noted that Annual Solar Radiation (ASR) can vary considerably over small distances due to topography and therefore the polygons allocated to NVA records, especially those with a radius of 1000 m, may contain ASR values which are not representative of the microhabitat in which the species occurs in the location. However in this analysis the lowest ASR values are of interest and given the shade tolerance of these conifer species these low ASR sites (e.g. gullies, S-facing slopes) are likely to be occupied or are potential fire refuges for the relevant species (in some situations the highest ASR values in the polygon could represent exposed sites which have poor habitat or refuge value).Table 6. Statistics for topographic variables used to model fire refugia. Annual solar radiation is the total of 12 monthly mean solar radiation values given in watt hours per square metre (WH/m2) based on topography without cloud cover. Topographic position index gives deviation from the mean elevation value for an analysis neighbourhood of 200 m radius.SPECIES VARIABLE MIN MAX RANGE MEAN STD DEV - 2 STD DEVPhyllocladus aspleniifolius ASR 702017 1579120 877100 1261730 128304 1005122Athrotaxis selaginoides ASR 545442 1590240 1044800 1231180 150285 930610Lagarostrobos franklinii ASR 676423 1438270 761844 1150500 71201 1008099Lagarostrobos franklinii TPI -120 60 180 -4 12 -281.6 Tree Core Research ZoneIt is recommended that tree coring is not permitted, unless there is a compelling case, in areas which represent the range limits for a species, in key fire refugia and within the Wilderness Zone of the Tasmanian Wilderness World Heritage Area, as defined in the TWWHA Management Plan.A Tree Core Research Zone (TCRZ) for each species has been derived by removing fire refugia and the Wilderness Zone from the Core Range and adding existing tree ring sampling sites from the Dendrochronology Sampling Sites Database described below, each point record buffered to create 8

a circle of 500 m radius to capture the sampling sites.Table 7. Area of mapped Tasveg communities for each conifer species within Permitted Zone.SPECIES TASVEG (ha) TREE CORE RESEARCH ZONE % of mapped Tasveg in TCRZHuon pine 13831 4684 33.9Pencil pine 24290 12797 52.7King billy pine 28627 7424 26.01.7 Spatial DataThe input layers and results of the spatial analysis are provided as ESRI shapefiles accompanied by business rules explaining their derivation.Table 8. Summary of datasets used in the development of TCRZ layer for each species.Layer CTP HP KBP PPCore Geographic Range X X X XCore Altitudinal Range X X X XCore Climatic Range X XWHA Wilderness Zone X X X XDendrochronology Sites X X X XAnnual Solar Radiation X X XTopographic Position Index XOffshore Islands XInland Islands X X XSphagnum bogs XRock (boulderfields, cliffs) XFlood plains X

9

2 Spatial Database of Dendrochronology Sampling Sites in TasmaniaLocations of dendrochronology sampling sites in Tasmania have been compiled from the Natural Values Atlas [http://www.naturalvaluesatlas.tas.gov.au] as at October 2012 and the International Tree Ring Database [http://www.ncdc.noaa.gov/paleo/treering.html] archived to March2012.It is recommended that this database be updated annually.

Table 9. Dendrochronology database structureFIELD NAME FIELD TYPE DESCRIPTIONPERMIT_NO NUMBER Unique number from RMC permit database.PERMIT_FROM DATE Commencement date for permit.PERMIT_TO DATE End date for permit.SURNAME STRING Surname of researcher whom permit issued to.TITLE STRING Title and first name of researcher.ORGANISATN STRING Name of organisation or research institution.SITE_NAME STRING Short descriptive name for sampling siteA_CUP_CORE NUMBER Number of trees cored for live A. cupressoides. -999 indicates unspecified number .A_SEL_CORE NUMBER Number of trees cored for for live A. selaginoides. -999 indicates unspecified number .L_FRA_CORE NUMBER Number of trees cored for for live L. franklinii. -999 indicates unspecified number .P_ASP_CORE NUMBER Number of trees cored for live P. aspleniifolius. -999 indicates unspecified number .DEAD_CONIF NUMBER Number of cores and/or discs for dead conifers (standing or logs). -999 indicates unspecified number of cores/discs.OTHER_SPP STRING Name of other species covered by permit and max number of cores if specified.LOCAT_ACC NUMBER 1 = point sampling location with GPS accuracy <100m2 = point location derived from map, publication, etc. >100.3 = polygon representing sampling area (more specific than a reserve)4 = polygon representing a reserve where permit appliesNVA_ACC NUMBER Location accuracy in metres from NVA record.NVA_PROJ STRING Project name from NVA.SAMPLE_DATE DATE Date sampling conducted (from NVA).TO_COLLECT STRING Material to be collected, as specified in permit.FROM_WHERE STRING Location specified in permit.SOURCE STRING Source of dendrochronology site information, e.g. RMC permit database, NVA, ITRDB.NOTES STRING Any additional information (e.g. notes from NVA record).10

3 Recommendations1. Tree Core Research Zones made available online via the NVA, Conservation Information System or similar publicly accessable interactive mapping platform.2. Land managers and the permit issuing authority should consult the Tree Core Research Zones when considering proposed permits for dendrochronology research.3. Permits for tree core sampling should not be issued for locations outside the relevant Tree Core Research Zone for each species, unless a compelling case exists.4. Permit applications should contain specific location details.5. Permits should include a condition that the permit holder provide timely and accurate details of samples obtained for inclusion in the NVA and the Dendrochronology Sampling Sites Database.6. The Dendrochronology Sampling Sites Database should be updated at least annually and published online via the NVA, Conservation Information System or similar publicly accessable interactive mapping platform.

11

ReferencesBALMER, J., WHINAM, J., KELMAN, J., KIRKPATRICK, J. & LAZARUS, E. 2004. A review of the floristic values of the Tasmanian Wilderness World Heritage Area. Nature Conservation Report. Hobart: Department of Primary Industries, Water and Environment.BEIER, P. & BROST, B. 2010. Use of land facets to plan for climate change: conserving the arenas, not the actors. Conservation Biology, 24, 701-10.DOBROWSKI, S. Z. 2011. A climatic basis for microrefugia: the influence of terrain on climate. Global Change Biology. 17, 1022–1035.ELITH, J., GRAHAM, C., ANDERSON, R., DUD K, M., FERRIER, S., GUISAN, A., HIJMANS, R., HUETTMANN, F., LEATHWICK, J. & LEHMANN, A. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography, 29, 129-151.FRANKLIN, J. 2009 Mapping species distributions. Cambridge University Press, Cambridge.GUISAN, A., ZIMMERMANN, N. E., ELITH, J., GRAHAM, C. H., PHILLIPS, S. & PETERSON, A. T. 2007. What matters for predicting the occurrences of trees: techniques, data, or species' characteristics? Ecological Monographs, 77, 615-630.HUGHES, L. 2003. Climate change and Australia: trends, projections and impacts. Austral Ecology, 28, 423-443.JORDAN, G. 2012. Report for the Independent Verification Group of the Tasmanian ForestsIntergovernmental Agreement (IGA) on palaeo-endemic plants (primitive, relictual and ancient plant groups). IVG Forest Conservation Report 3B.Nix, H.A. 1986. A biogeographic analysis of Australian Elapid Snakes. In. Atlas of Elapid Snakes of Australia. (ed.) R. Longmore pp. 415. Australian Flora and Fauna Series Number 7. Australian Government Publishing Service: Canberra. PEARSON, R. G. 2006. Climate change and the migration capacity of species. Trends in Ecology & Evolution 21, 111-3.READ, J. & BUSBY, J. R. 1990. Comparative response to temperature of the major canopy species of Tasmanian cool temperate rainforest and their ecological significance. II. Net photosynthesis and climate analysis. Australian Journal of Botany, 38.STYGER, J., BROWN, M. & WHINAM, J. 2010. Monitoring for the effects of climate change on the flora values of the Tasmanian Wilderness World Heritage Area. Papers and Proceedings of the Royal Society of Tasmania, 144, 21-28.PHILLIPS, S.J., ANDERSON, R.P. & SCHAPIRE, R.E. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190: 231-259.WOOD, S. W., MURPHY, B. P. & BOWMAN, D. M. J. S. 2011. Firescape ecology: how topography determines the contrasting distribution of fire and rain forest in the south-west of theTasmanian Wilderness World Heritage Area. Journal of Biogeography, 38, 1807-1820.

12

Appendix 1: Previous Tree Coring Sites

Sites from Dendrochronology Sampling Sites Database, as at October 2012. Point locations shown with a 5km radius buffer. Note that any sites not recorded in publicly available databases are not shown.

13

Appendix 2: Tree Core Research Zones

14

15

16

17

Appendix 2: Accepted distribution records for rainforest conifer species

Only uncertain outlying record is mouth of of Lagoon R on W Coast (18,500 m accuracy). Without further info or records, assume this is dubious.Also old King Is record, but considered extinct there.18

Several records with low accuracy and no location notes at the northern boundary,. All but two coincident outlying records (accuracy 18,500 m) are within 2 km of mapped known range based on better records and have accuracy 1350 m. Leave boundary as is, since the outlying records will be considered range limits anyway. Likewise the records in the upper Gordon valley, and at Melaleuca Lagoon, all within 1 km of known range.Several unverified records in the Kermandie R catchment near Geeveston. A record from Mt Field NP picnic area, and one from South Hobart are presumably in cultivation. The 1940 herbarium record labelled shores of Lake St Clair (but mapped in the NVA near Lake Will) has been excluded.19

Deleted 2 RHP polygons in the NE.One erroneous record - location labelled 'Moonlight Ridge' is placed on Bruny Is (evidently wrong easting).20

One erroneous record – for location 'Lake Mackenzie' is placed in north-east Tas (evidently wrong easting). One record from Liffey Valley with accuracy 10 km and some confusion re location in herbarium notes – ignore this given lack of other records from this area.Several records from West Coast Range, all accuracy >1000 m., including two records from 1951 which have location info (Henty R bridge; Lynchford) but without more recent and accurate records – and given the environmental history of these areas – it is assumed that the species is locally extinct.21