public read cn=submitters,o=unaffiliated,dc=ecoinformatics,dc=org all cn=pn_vc_tall_eucalypt,o=unaffiliated,dc=ecoinformatics,dc=org read Mr Laurence Berry Fenner School of Environment and Society, Australian National University PhD student

Geography Building 48a, Linnaeus way Fenner School of Environment and Society, Australian National University, Acton Canberra, ACT 0200 Australia

+61 (02) 6125 7653 Laurence.berry@anu.edu.au Mr John Stein The Fenner School of Environment and Society Senior Manager 1 (Research)

Frank Fenner Building 141, Linnaeus Way The Australian National University Canberra ACT 0200 Australia

+61-2-6125 4669 john.Stein@anu.edu.au Professor Brendan Mackey Griffith University Director, Griffith Climate Change Response Program

G01, Academic 1 Building, room 2.25 Griffith University, Gold Coast campus Queensland 4222

Originator 1427086270560 Researcher We used a case study in an Australian wet montane forest to establish how predictive fire simulation models can be interpreted as management tools to identify potential fire refuges. We tested the ability of a topographically based fire prediction model developed by Mackey et al (2002) in the O’Shannassy and Maroondah water catchments, NE north-east of Melbourne, Australia, with fire severity data collected following a large wildfire in 2009 in the same area. We derived our fire severity data from a larger map created by the Department of Sustainability and Environment (2009), using SPOT satellite imagery and the normalised-burnt ratio. We examined the relationship between the probability of fire refuge occurrence as predicted by an existing fire refuge model and fire severity experienced during a large wildfire. We also examined the extent to which local fire severity was influenced by fire severity in the surrounding landscape. We used a combination of statistical approaches including generalised linear modelling, variogram analysis and receiver operating characteristics and area under the curve analysis (ROC AUC). We found that the amount of unburnt habitat and the factors influencing the retention and location of fire refuges varied with fire conditions. Under extreme fire conditions, the distribution of fire refuges was limited to only extremely sheltered, fire-resistant regions of the landscape. During extreme fire conditions, fire severity patterns were largely determined by stochastic factors that could not be predicted by the model. When fire conditions were moderate, physical landscape properties appeared to mediate fire severity distribution. Our study demonstrates that land managers can employ predictive landscape fire models to identify the broader climatic and spatial domain within which fire refuges are likely to be present. It is essential that within these envelopes, forest is protected from logging, roads and other developments so that the ecological processes related to the establishment and subsequent use of fire refuges are maintained. Department of Sustainability and Environment (2009) Remote sensing guideline for assessing landscape-scale fire severity in Victoria’s forest estate. Unpublished technical manual., Department of Sustainability and Environment, Melbourne. Mackey, B., D. Lindenmayer, M. Gill, M. McCarthy, and J. Lindesay. 2002. Wildlife, Fire and Future Climate: A Forest Ecosystem Analysis. CSIRO publishing, Collingwood. Fire Fragmentation LTERN Monitoring Themes 0501 0502 ANZSRC-FOR Earth Science Services > Hazards Management > Hazards Planning GCMD TERN-BY-SA Special Condition If this data is accepted for publication, the metadata should be made live as soon as possible. However, the data should be embargoed until appropriate approval to publish has been received from whoever holds the rights to reproduce the fire severity data originally created by DSE 2009 (John Stein). Approval to reproduce the Mackey et al (2002) data can be sought from David Lindenmayer, a co-author on the original paper. These conditions must be included when licensing any derivative works. The data were collected in the O’Shannassy and Maroondah Water Catchments ~ approximately 80km NE of Melbourne in Victoria, Australia 145.5 145.7311 -37.5843 -37.6579 2002 2009 1408532875740

The fire severity data presented in this data package was originally created by DSE (2009) using the normalized-burn ratio from SPOT satellite data. We trimmed the DSE fire severity layer to within the extent of the O’Shannassy and Maroondah water catchments only. The file represents a raster grid with 20 x 20 m cells. Within each cell fire severity is recorded on a scale of 1-5. 1) high severity crown fire 2) crown scorch 3) moderate crown fire (canopy survives) 4) understorey burn only 5) no fire in the crown or the understorey.

None

We sourced data which mapped the probability of an area being retained as a fire refuge for fauna vulnerable to the effects of fire (probability of area remaining unburnt), from a study published by Mackey et al. (2002). As the original data used to create their model was unavailable, we rasterised a high quality digital image of the final published map showing the predicted distribution of areas likely to remain unburnt following a fire. This was trimmed to within the extent of the O’Shannassy and Maroondah water catchments only. The grid cells produced were 20 m x 20 m and shared identical co-ordinates with the DSE fire severity grid cells. This enabled a direct comparison to be made. The predicted severity data takes the form of a scale from 1-9, with 1 being a low probability that a location will remain unburnt and 9 being a high probability that a location will remain unburnt. Both sets of data were merged to form a large table of co-ordinates, severity values and predicted values for each catchment. To add depth to our analyses we also added environmental variables to each grid cell; topographic wetness index (twi), elevation, precipitation and aspect. To conduct analyses of spatial autocorrelation, we also included a spatially lagged response variable (SLRV) in our dataset (fm4, fm8). These were calculated in ArcMap using the focal mean tool, which produces a severity value for each cell based on the mean severity values in the surrounding 4 (rook) and 8 (queen) cells.

None The predicted severity values were created in 2002 by Mackey et al (2002). The fire severity data were collected by DSE (2009) following the 2009 ‘Black Saturday’ bushfires in the region. As the region surrounding our study area is subject to heavy logging pressure, we chose to sample the O’Shannassy and Maroondah catchments only to remove any potential influence of logging on fire severity. See Mackey, B., D. Lindenmayer, M. Gill, M. McCarthy, and J. Lindesay (2002) Wildlife, Fire and Future Climate: A Forest Ecosystem Analysis. CSIRO publishing, Collingwood. Department of Sustainability and Environment (2009) Remote sensing guideline for assessing landscape-scale fire severity in Victoria’s forest estate. Unpublished technical manual., Department of Sustainability and Environment, Melbourne. Professor David Lindenmayer Fenner School of Environment and Society, The Australian National University Principal Investigator

Fenner School of Environment and Society, The Australian National University Frank Fenner Building (Building 141), Fenner School of Environment and Society, The Australian National University Canberra ACT 0200

02 61250654 david.lindenmayer@anu.edu.au Data Owner This project was not directly funded by any sources. Laurence Berry was supported by an ARC discovery scholarship as part of the Victorian Central Highlands (VCH) long term research project. Field observations used to cross-validate the DSE (2009) fire severity data were collected from long term monitoring sites across the VCH by project field staff. These observations were made in 2009. Since 2012 the Plot Network infrstructure has been funded as part of the Long Term Ecological Research Network (LTERN). LTERN is a Facility within the Terrestrial Ecosystem Research Network (TERN). TERN is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy. lvic_berry_fire_severity_maroondahcatchment_p363t617.csv lvic_berry_fire_severity_maroondahcatchment_p363t617.csv 1 0 \r\n column , ecogrid://knb/ltern2.289.3 cn=submitters,o=unaffiliated,dc=ecoinformatics,dc=org all cn=allusers,o=unaffiliated,dc=ecoinformatics,dc=org read cn=pn_vc_tall_eucalypt,o=unaffiliated,dc=ecoinformatics,dc=org read row Row location in raster grid number col Column location in raster grid number cell Cell number in grid (1 is top right) number xcord x co-ordinate AGD66. Co-oridinates for the centre of each 20 m grid cell meter whole ycord y co-ordinate AGD66. Co-oridinates for the centre of each 20 m grid cell number whole x Latitude WGS84 decimal degrees meter real y Longitude WGS84 decimal degrees meter real fsev Fire severity meaure 1 crown scorch 2 crown burn 3 moderate crown scorch 4 light or no crown scorch, understorey burnt 5 no crown scorch, no understorey burnt pref Probability of area remaining unburnt following fire from Mackey et al (2002). These values were generated as the final product of the Mackey et al (2002) modelling process. These values represent the probability of a grid cell remaining unburnt following a fire. As these values represent a scale of probability and are not categorical we can only provide a qualitative description of the characteristics associated with values at each end of the scale. These are as follows; 1 can be interpreted as = Low percentile mean fire interval (<100 years), low probability of multi-agedness (<25%). lower mean topographic wetness index (TWI), higher elevation percentile, higher annual mean temperature. 9 can be interpreted as = 90-100% percentile mean fire interval (>500 years), high probability of multi-agedness (>65%), higher mean topographic wetness index (TWI), lower elevation percentile, lower annual mean temperature. Character ele Elevation meters (ASL). 20 m DEM, created using Vicmap 10m contours, spot heights and watercourses meter whole twi Topographic wetness index. Index of contributing catchment divided by slope at each grid cell of the 20 m DEM TWI is a continuous terrain-based measure of potential wetness that indicates position in the landscape. It ranges from negative values on hill tops and ridges (with no contributing catchment) then upper slopes (small contributing catchment/steep slope) to increasingly higher positive values through lower slopes, valley flats and eventually drainage lines. dimensionless real slope Slope based on neighbourhood geometry of the 20 m DEM. degree real aspe East component of aspect based on sine of aspect angle (continuous value from -1 to +1) from neighbourhood geometry of the 20 m DEM dimensionless real aspn North component of aspect based on cosine of aspect angle (continuous value from -1 to +1) from neighbourhood geometry of the 20 m DEM mole real fm4 Mean fire severity value of the 4 adjacent neighbourhood cells Character fm8 mean fire severity value of all 8 neighbourhood cells Character crownfire Fire severity re-classified into binomial measure of crownfire = 1 or other = 0 0 other - combined DSE fire severity classes 3 + 4 + 5 1 crownfire - combined DSE fire severity classes 1 + 2 lowsev Fire severity re-classified into binomial measure of low severity fire = 1 or other = 0 0 other- combined DSE fire severity classes 1 + 2 1 low severity fire - combined DSE fire severity classes 3 + 4 + 5 lvic_berry_fire_severity_oshannassycatchment_p363t618.csv lvic_berry_fire_severity_oshannassycatchment_p363t618.csv 1 0 \r\n column , ecogrid://knb/ltern2.290.3 cn=submitters,o=unaffiliated,dc=ecoinformatics,dc=org all cn=pn_vc_tall_eucalypt,o=unaffiliated,dc=ecoinformatics,dc=org read cn=allusers,o=unaffiliated,dc=ecoinformatics,dc=org read row Row location in raster grid number col Column location in raster grid number cell Cell number in grid (1 is top right) number xcord x co-ordinate AGD66. Co-oridinates for the centre of each 20 m grid cell meter whole ycord y co-ordinate AGD66. Co-oridinates for the centre of each 20 m grid cell number whole x Latitude WGS84 decimal degrees meter real y Longitude WGS84 decimal degrees meter real fsev Fire severity meaure 1 crown scorch 2 crown burn 3 moderate crown scorch 4 light or no crown scorch, understorey burnt 5 no crown scorch, no understorey burnt pref Probability of area remaining unburnt following fire from Mackey et al (2002). These values were generated as the final product of the Mackey et al (2002) modelling process. These values represent the probability of a grid cell remaining unburnt following a fire. As these values represent a scale of probability and are not categorical we can only provide a qualitative description of the characteristics associated with values at each end of the scale. These are as follows; 1 can be interpreted as = Low percentile mean fire interval (<100 years), low probability of multi-agedness (<25%). lower mean topographic wetness index (TWI), higher elevation percentile, higher annual mean temperature. 9 can be interpreted as = 90-100% percentile mean fire interval (>500 years), high probability of multi-agedness (>65%), higher mean topographic wetness index (TWI), lower elevation percentile, lower annual mean temperature. Character ele Elevation meters (ASL). 20 m DEM, created using Vicmap 10m contours, spot heights and watercourses meter whole twi Topographic wetness index. Index of contributing catchment divided by slope at each grid cell of the 20 m DEM TWI is a continuous terrain-based measure of potential wetness that indicates position in the landscape. It ranges from negative values on hill tops and ridges (with no contributing catchment) then upper slopes (small contributing catchment/steep slope) to increasingly higher positive values through lower slopes, valley flats and eventually drainage lines. dimensionless real slope Slope based on neighbourhood geometry of the 20 m DEM. degree real aspe East component of aspect based on sine of aspect angle (continuous value from -1 to +1) from neighbourhood geometry of the 20 m DEM dimensionless real aspn North component of aspect based on cosine of aspect angle (continuous value from -1 to +1) from neighbourhood geometry of the 20 m DEM dimensionless real fm4 Mean fire severity value of the 4 adjacent neighbourhood cells Character fm8 mean fire severity value of all 8 neighbourhood cells Character crownfire Fire severity re-classified into binomial measure of crownfire = 1 or other = 0 0 other - combined DSE fire severity classes 3 + 4 + 5 1 crownfire - combined DSE fire severity classes 1 + 2 lowsev Fire severity re-classified into binomial measure of low severity fire = 1 or other = 0 0 other- combined DSE fire severity classes 1 + 2 1 low severity fire - combined DSE fire severity classes 3 + 4 + 5 Victorian Tall Eucalypt Forest Plot Network