Palaeofire Activity in Western Tasmania: Climate Drivers and Land-Cover Changes

Palaeofire activity in western Tasmania: climate drivers and land-cover changes

MICHELA MARIANI

(ORCID: 0000-0003-1996-3694)

Thesis submitted in total fulfilment of the requirements of the degree of

DOCTOR OF PHILOSOPHY

November 2017

School of Geography

Faculty of Science

The University of Melbourne

Al mio nonno Battista,

La tua malattia ha cancellato il mio nome nella tua memoria,

Ma tu rimarrai sempre nei miei ricordi migliori

…come quando andavamo a funghi a Valle

…o quando giocavamo a carte i pomeriggi d’estate

…o quando mi stringevi la mano e mi raccontavi le tue storie.

Per sempre con me.

To my grandpa Battista,

Your illness took my name away from your mind,

But you will always be in my best memories

…like when we used to go mushroom-hunting in the mountains

…or when we used to play cards on summer afternoons

…or when you used to hold my hand and tell me your stories.

Forever with me.

III

Abstract

Under the current changing climatic regime, in which wildfires are predicted to increase in frequency and magnitude, it is important we gain a better understanding on past climatic trends and fire activity to properly manage fires and landscapes, preserve valuable natural ecosystems and protect human lives and properties. Fire activity is especially projected to increase in temperate regions, such as Australia’s southeast. In this context, western Tasmania represents a key region where the environmental impacts of wildfires can be disastrous for the remnant pockets of fire- sensitive vegetation.

Climate influence on fire activity and vegetation dynamics operates at multiple timescales, from inter-annual to multi-millennial. Given the time limitation of historical records, we need to look at long-term records to gain a better understanding on what modulates fire activity and how changes in fire regimes influence ecosystem dynamics. This PhD project aimed to a) identify the climate drivers of short- and long-term fire variability in western Tasmania and b) quantify climate- and fire- driven vegetation changes in this region throughout the Holocene.

To understand the short-term drivers of fire activity in western Tasmania, I explored the relationship between the main climate modes of the Southern Hemisphere and a documentary record of ﬁre occurrence from this region. This analysis suggested that the Southern Annual Mode (SAM) -an index for the position and strength of SWW- is strongly correlated with inter-annual fire activity across western Tasmania during the last 25 years. Moreover, the persistent positive trend in SAM recorded during the last 500 years was found to be tightly coupled to increased biomass burning within the same region.

To understand the long-term landscape changes in western Tasmania, I combined high resolution pollen and charcoal analyses, coupled with recently developed mathematical modelling of pollen dispersal and productivity. Within this Thesis, I applied pollen dispersal models to calibrate the pollen-vegetation relationship for the first time in Australia. This method involves two steps: (1) a modern pollen analysis coupled with distance-weighted vegetation data to calibrate the present-day pollen-vegetation relationships and (2) an application of these relationships to a fossil pollen record to produce past vegetation cover estimates. The application of pollen dispersal models proved the biases inherent in previous interpretations of pollen spectra from western Tasmania. Specifically, the results from these analyses showed that this region was mostly dominated by treeless moorland vegetation, supporting the identification of western Tasmania as a cultural landscape. Moreover, my results showed that land-cover changes throughout the Holocene occurred in response to climatic change and a shift in fire regimes due to ENSO/SWW interactions.

Declaration

This is to certify that:

I. the Thesis comprises only my original work towards the PhD,

II. due acknowledgement has been made in the text to all other material if and when used,

III. the Thesis is less than 100,000 words in length, exclusive of tables, maps, bibliographies and appendices.

Publications arising from this PhD Thesis

Chapter 3.

Mariani M. & Fletcher M.-S. (2016) The Southern Annular Mode determines interannual and centennial-scale fire activity in temperate southwest Tasmania, Australia. Geophysical Research Letters, 43, 1702–1709.

Chapter 4.

Mariani M. & Fletcher M.-S. (2017) Long-term climate dynamics in the extra-tropics of the South Pacific revealed from sedimentary charcoal analysis. Quaternary Science Reviews, 173, 181-192.

Chapter 5.

Mariani M., Connor S.E., Theuerkauf M., Kuneš P., & Fletcher M.-S. (2016) Testing quantitative pollen dispersal models in animal-pollinated vegetation mosaics: An example from temperate Tasmania, Australia. Quaternary Science Reviews, 154, 214– 225.

Chapter 6.

Mariani M., Connor S.E., Fletcher M.-S., Theuerkauf M., Kuneš P., Jacobsen G., Saunders K.M., Zawadzki A. (2017) How old is the Tasmanian cultural landscape? A test of landscape openness using quantitative land-cover reconstructions. Journal of Biogeography, 10, 2410–2420.

VII

Acknowledgement of Collaboration

I hereby certify that part of the work included in this Thesis was performed in collaboration with other researchers. Below I have outlined the extent of collaboration, with whom and under what auspices.

Most charcoal records (11/13) used for the multi-site compilations presented in Chapter 3 and 4 were produced prior this PhD project by different analysts and were made available to me by my principal supervisor, Michael-Shawn Fletcher. Charcoal analyses were previously performed by Rita Attwood (Basin Lake), Anthony Romano (Lake Gaye), Alexa Benson (Lake Gwendolyn and Lake Vera), Rachael Fletcher (Lake Nancy, Hartz Lake, Lake Selina and Owen Tarn), Haidee Cadd (Square Tarn), William Rapuc (Lake Julia) and Michael-Shawn Fletcher (Lake Osborne). All the co-authors listed in the ‘Publications arising from this PhD thesis’ list commented on the corresponding manuscripts.

VIII

Acknowledgement of Authorship

I hereby certify that the work embodied in this Thesis contains published papers of which I am a joint author. I have included as part of the Thesis a written statement, endorsed by my Principal Supervisor, attesting to my contribution to the joint publications.

I, Michela Mariani, was the primary investigator and lead author of all the published manuscripts presented in this Thesis.

______

Michael-Shawn Fletcher (Principal Supervisor)

______

Michela Mariani

ACKNOWLEDGMENTS

I sincerely thank:

o My supervisors Michael-Shawn Fletcher and Simon Edward Connor for supporting my ideas throughout my PhD candidature. Without your inspiring guidance, I would have never reached this life milestone. Thanks to Michael-Shawn Fletcher for financial support during my fieldwork and for giving me the opportunity to attend the Southern Connection conference in January 2016 and the AQUA Conference in December 2016. Thanks to Simon E. Connor for being my guide in the vegetation survey fieldtrip and for teaching me all Tasmanian plants.

o Krystyna M. Saunders, Henk Heijnis, Geraldine Jacobsen, Patricia Gadd, Atun Zawadzki and Robert Chisari

for supporting my AINSE PGRA application, for hosting me at ANSTO several times through the course of my PhD and for giving me the opportunity to learn new methodological approaches and geochemical techniques.

o Martin Theuerkauf and Petr Kuneš

for their essential help with the first application of quantitative models for land-cover changes in Australia

o The AINSE team (Sandy O’ Connor, Rachel Caldwell, Michelle Durant, Nerissa Phillips)

for their help during my AINSE PGRA visits and for providing me travel support to attend the following conferences and workshops: AQUA 2016, ASLO 2017 and PALAEOCLIMATE OF THE SOUTHERN HEMISPHERE 2016

o The professional staff at the School of Geography (Joanne, Tina and Darren)

for their important help to solve administrative and technical issues

o My lab colleagues and friends (Haidee, Anthony, Kristen, Rachael, Coralie, Alexa and Bianca)

for reciprocal support through all the good and bad moments of our degrees and/or work during the last 3 years

o Valentina Vanghi

for being my italian best friend in Australia

o My family in Italy (my parents – Tiziana and Massimiliano, my grandparents – Giovanna, Battista, Stefano and Maria, my aunts and uncles – Adele, Deborah, Dorino, Silvana, Angela, Rossella, Marco, Fabio and Ilaria, my cousins – Luca, Silvia, Marco, Ivan, Federica, Nadia, my parents-in law – Irma and Alessandro)

for constantly reminding me where my home is

o My friends in Italy, UK and Spain

for being close friends regardless the distance that separates us

o La tata Donatella for taking care of my baby during the final writing stages

o My family in Australia (Lorenzo and Linneo)

per essere la parte piú importante di ogni mio giorno e per sostenermi in ogni mia scelta

I am also very grateful for the financial support provided for this PhD research project by the following organisations:

o Australian Research Council

Grants # IN140100050 and DI110100019

• Australian Institute of Nuclear Science and Engineering (AINSE)

PGRA- Post Graduate Research Award #12039

• The University of Melbourne, the Faculty of Science and the School of Geography

MIRS and MIFRS scholarships John and Allan Gilmour Science Award 2016 Michael Webber Doctoral Prize in Geography 2016

XII

TABLE OF CONTENTS

Chapter 1. Introduction and Thesis aims______1 1.1. Introduction______.1 1.2. Thesis aims______5 1.3. Organisation of the Thesis______11

Chapter 2. Background and Methods______12 2.1. PhD project workflow______12 2.2. Palaeoecology______14 2.3. Palaeoecology fundamentals______16 2.3.1. Coring techniques______16 2.3.2. Chronology______17 2.4. Fire activity and its reconstruction______20 2.4.1. Documentary records of fire activity______22 2.4.2. Charcoal analysis______23 2.4.3. Regional charcoal compilations______25 2.5. Pollen and palaeovegetation reconstruction______26 2.5.1. Fossil pollen______26 2.5.2. Pollen-vegetation calibration and modelling______28 2.5.3. Quantitative palaeovegetation reconstructions in Australia______34 2.6 Study area______36 2.6.1. Tasmania and its climatic context______36 2.6.2. Tasmanian vegetation and the legacy of fire______40 2.6.3. An ancient cultural landscape?______44 2.6.4. Study site locations______46

Chapter 3. Short-term climate and fire variability in western Tasmania______49 3.1. Introduction______50 3.2. Methods______53 3.3. Results______56

XIII

3.4. Discussion______58 3.5. Conclusion______61 3.6. Supplementary results and information______62

3.6.1. Additional information and results for the SEA______64 3.6.2. Palaeofire compilation and charcoal results______67

Chapter 4. Long-term climate variability and palaeofire activity in western Tasmania______70 4.1. Introduction______71 4.1.1. Spatiotemporal climate dynamics in the southern extra-tropics______74 4.1.2. Western Tasmania______77 4.2. Methods______79 4.2.1. Chronology and charcoal analysis______79 4.2.2. Palaeofire analysis______80 4.2.3. Statistical analyses______81 4.3. Results______82 4.4. Discussion______87 4.4.1. Western Tasmanian palaeofire reconstruction______87 4.4.2. Dominant drivers of western Tasmanian palaeofire dynamics______90 4.4.3. Implications for modern climate patterns and non-stationary teleconnections______92 4.5. Conclusions______93 4.6. Supplementary results and information______94 4.6.1. Additional details about the sites location______95 4.6.2. Additional information about cores chronology______96 4.6.3. Additional information about the Southern South America compilation___ 102

Chapter 5. Calibrating the pollen-vegetation data______103 5.1. Introduction______104 5.2. Pollen dispersal, production and relevant source area______106 5.3. Study area______108

XIV

5.4. Methods______110 5.4.1. Surface samples collection and vegetation surveys______110 5.4.2. Pollen analysis______114 5.4.3. Dispersal models______115 5.5. Results______119 5.5.1. Vegetation surveys______119 5.5.2. Surface samples pollen analysis and fall speeds______119 5.5.3. Relevant source area of pollen (RSAP)______120 5.5.4. Pollen productivity estimates (PPEs) ______121 5.6. Discussion______123 5.6.1. Relevant source area of pollen (RSAP) ______124 5.6.2. Pollen productivity estimates (PPEs)______125 5.6.3. Potential applications in Australian palaeoecology______127 5.7. Conclusion______129 5.8. Supplementary results and information______130 5.8.1. Additional details on the ERV method for PPEs and RSAP calculations______132 5.8.2. Additional results for RSAP and PPEs calculations______136

Chapter 6. A quantification of regional land-cover changes from western Tasmania______143 6.1. Introduction______144 6.1.1. Quantitative reconstruction of land-cover from pollen data______147 6.2. Methods______148 6.2.1. Study area______148 6.2.2. Pollen and charcoal analyses______150 6.2.3. Quantitative vegetation reconstruction and model validation______151 6.3. Results______153 6.3.1. Pollen and charcoal analyses______153 6.3.2. REVEALS validation______155 6.3.3. REVEALS application______156 6.4. Discussion______158

6.4.1. REVEALS model performance in western Tasmania______158 6.4.2. Holocene land-cover changes in western Tasmania______159 6.4.3. Did moorland dominate Holocene landscapes of western Tasmania? _____ 162 6.5.Conclusions______163 6.6. Supplementary results and information______164 6.6.1. Additional details about the area surrounding Dove Lake______165 6.6.2. Core extraction and chronology information______166 6.6.3. Additional information about the pollen record______169 6.6.4. Pollen zones descriptions for Dove Lake core______170 6.6.5. Further calculations of PPEs using different sets of atmospheric parameters______172 6.6.6. Further statistics supporting the REVEALS validation______173 6.6.7. Individual estimates for plant taxa and additional REVEALS model runs______175

Chapter 7. General discussion and approach limitations______178 7.1. Regional-scale drivers of fire occurrence in Tasmania______179 7.1.1. Limitations of charcoal records and multi-site compilations ______183 7.2. Climate- and fire- driven land-cover changes in western Tasmania______188 7.2.1. Advancements and limitations of model-based quantitative vegetation reconstructions______191

Chapter 8. Conclusions______196

References______198

Appendices______223 Appendix A1 – Additional research papers published during this PhD project___ 223 Appendix A2 – PhD research outputs (first authorship only)______238

XVI

Appendix A3 – List of all courses/workshops/laboratory attended during the PhD candidature______240 Appendix A4 – My PhD in numbers______242

XVII

LIST OF FIGURES with CAPTIONS

Figure 2-1. Diagram showing the gradient of landscapes with different levels of anthropogenic alteration (modified from Colombaroli et al., 2017)……………...... …15

Figure 2-2. Records of fire activity at different spatio-temporal scales reporting fire triangles (modified from Scott et al., 2013). a) Fire triangle at long temporal time scales (palaeofire triangle; Scott, 2000); b) Fire triangle for short temporal time scales (Pyne et al., 1996)…………………………………………………………………………...22

Figure 2-3. Pollen processing flow chart and some common pollen grains and spores found in Tasmanian samples……………………………………………………………...27

Figure 2-4. Map showing the current distribution of studies regarding the quantitative vegetation reconstruction technique adopted in this project (LRA). Europe is the key area where theories and applications of this methodology have developed during the last 15 years……………………………………………………….30

Figure 2-5. a) Mean annual rainfall map of Tasmania; b) Mean annual temperature map of Tasmania. The black line represents the 1250 mm isohyet, characterising the border of the western “superhumid” region according to Gentilli (1972). Maps were created by interpolating values of mean annual temperature (80 stations) and total annual precipitation (220 stations) during the period 1961-1990. CoKriging was used to interpolate the climatic data in ESRI ArcMap 9.3, taking into account elevation from a digital elevation model (100m resolution). Temperature and rainfall data were obtained from BOM (Bureau of Meteorology). Coordinates system: GDA 1994 Zone 55 (Grid resolution: 1.8 km)………………………………………………………...37

Figure 2-6. Rainfall anomaly correlation with a) SAM index (Marshall, 2003) and b) SOI (Southern Oscillation Index) (data from NOAA). Maps were created by calculating correlation coefficients (r) between the annual rainfall anomalies of the period 1961-1990 for 220 stations and the climate modes indices. Rainfall anomalies are the differences between the total precipitation of each year and the average total precipitation of the 30-year baseline period. The r values from the stations have been spatially interpolated using the Kriging method in ArcMap 9.3. Rainfall data from BOM. Coordinates system: GDA 1994 Zone 55 (Grid resolution: 1.8 km)……………39

Figure 2-7. Vegetation map of western Tasmania, indicated as the “superhumid” area with annual rainfall above 1250 mm, according to Gentilli (1972); data by TasVeg 3.0 (Government of Tasmania, 2013). Walter and Lieth climate diagram for Queenstown (1965-1994) and Lake St. Clair National Park stations (1990-2014)…….41

XVIII

Figure 2-8. Map showing the location of the sites used for the multi-site compilation of charcoal data (red triangles) and the site used for the regional vegetation reconstruction (green triangle). Background brown shading represents contour lines (100 m- interval) and yellow polylines indicate the road network to help readers with the geographical orientation………………………………………………………...47

Figure 3-1. a) Correlation map between zonal wind speed at 850 mb and the SAM index (all data sourced by NOAA) b) Map of the correlation between annual rainfall anomalies and annual SAM index across Tasmania. Solid line indicates the boundary of the SAM zone (r>0.3). Dots represent all the fires occurred between 1992 and 2014 within this area. White triangles indicates the sites used for the palaeofire analysis.………………....…………………………………………………………………...51

Figure 3-2. a) Annual SAM index (1992-2014), (Marshall, 2003) b) Number of fires and c) Area burnt in the SAM zone of influence in Tasmania (1992-2014). Black solid lines represent the respective weighted average of the annual SAM index and the number of fires…………………………………………………………………………..….56

Figure 3-3. Departures from mean values for annual SAM index obtained using SEA during a) fire years based on number of fires; b) non-fire years based on number of fires; c) fire years based on area burnt and d) non-fire years based on area burnt. Dark grey blocks represent significant correlations with p<0.05………………………………………………………………………………….….…57

Figure 3-4. a) Paleofire charcoal composite of the SAM zone (50 year interval); b) SAM index reconstruction by Villalba et al. (2012); c) SAM index reconstruction by Abram et al., 2014; grey solid line is the annual index, black solid line represents the 70-year LOESS smoothing of the yearly reconstructed SAM index…………………………………………………………………………………………58

Figure 3-S1. Map of the correlation between annual rainfall anomaly and a) annual SAM Index; b) annual SOI Index; c) annual PDO Index; d) annual IOD Index………………………………………………………………………………………....64

Figure 3-S2. Departures of the main other climate modes influencing rainfall variability in the Southern Hemisphere in relationship with fire occurrence in western Tasmania (number of fires and area burnt)...………………………………….65

Figure 3-S3. Departure of seasonal SAM Index in relationship with number of fires; a) fire years; b) non-fire years…….………………………………………………….……66

Figure 3-S4. Time-series of the CHAR records used in the palaeofire analysis…………………………………………………………………………………...….68

XIX

Figure 3-S5. Palaeofire composite of southwest Tasmania with confidence intervals………………………………………..………………………………………….…69

Figure 4-1. Climate maps showing the proxies and paleofire sites used in this study: a) Southern Hemisphere map showing the correlation between annual surface zonal wind speed (m/s) and annual rainfall (mm) from the ERA-40 Reanalysis dataset (NOAA); b) same as a) but the projection is Australia-centered; c) correlation map between annual rainfall in Tasmania and the SAM index; d) correlation map between annual rainfall in Tasmania and the SOI index. Rainfall data from the Australian Bureau of Meteorology (BOM), SOI index obtained from NOAA and SAM index provided by the British Antarctic Survey (Marshall, 2003). Stars represent the published records used in this study and triangles are the charcoal sites location. A more detailed map of the site locations is presented in Supporting Information Figure S1. Black solid line in c) and d) represents the border of the “SAM zone”, where a rainfall-SAM index correlation has an r value below -0.3 (Mariani and Fletcher, 2016) …………………………...... ……..74

Figure 4-2. Plot of all the charcoal records included in the paleofire analysis and listed in Table 4-1. Values are shown as charcoal accumulation rates (CHAR, particles/cm-2 yr-1) on the right column. Left column shows the transformed records after Min-Max, Box-Cox and Z-score transformations……………………………...….83

Figure 4-3. Palaeofire results showing a) palaeofire composite for western Tasmania at 100 yrs- resolution using the thirteen charcoal records shown in Figure 4-2 and Table 4-1. Shaded grey area correspond to 95% percentiles; b) paleofire composite for western Tamania at 100 yrs- resolution plotted with red shaded areas representing confidence intervals between 90 and 95% from circular block-bootstrap (circboot) analysis; c) plot of the total number of sites used for the calculation of the regional charcoal influx values in each time step………………………………………………………………………………..………….84

Figure 4-4. Generalized Additive Models splines for western Tasmania (green, top panel) and southern South America (blue, botton panel) regional charcoal influxes. Shaded areas show 95% confidence intervals…………………………………….……..85

Figure 4-5. a) First derivatives of the GAM splines presented in Fig. 4 (green=western Tasmania; blue= southern South America); b) Regional charcoal influx from western Tasmania (z-scores; grey solid line) and El Niño number of events/100yrs from Laguna Pallcacocha (red solid line; Moy et al., 2002); c) Wavelet coherence between the first derivatives of the regional charcoal influx from western Tasmania and southern South America; d) Wavelet coherence between the regional

charcoal influx from western Tasmania and the ENSO proxy reconstructed from Laguna Pallcacocha (Moy et al., 2002). Right arrows indicate an in-phase relationship (left = anti-phase). Arrows pointing up or down represent possible lead/lag relationships between the time-series (Grinsted et al., 2004). Black lines indicate significant power spectra (significance=0.9). Areas inside the cone of influence should not be considered for interpretation as they are affected by edge effects………………………………………………………………………………….……..86

Figure 4-6. Summary plot showing a) Modelled lake level at Lake Gnotuk (southeast Australia) (Wilkins et al., 2013); b) Lago Condorito Palaeovegetation Index (NPI) (Moreno, 2004), positive values are indicative of lower relative moisture; c) Regional charcoal influx from western Tasmania (five points- weighted average); d) Regional charcoal influx from southern South America (40-55°S) (five points- weighted average); e) Forest pollen (%) from Lake Vera, Tasmania (Macphail, 1979); f) El Niño proxy expressed as the number of events/100 yrs (Moy et al., 2002). Red fill in f) represents the values about 5 events/100yrs, considered as significant El Niño activity according to Moy et al. (2002)……………………………………………...…….89

Figure 4-S1. Topographic map of Tasmania showing the location of the sites analysed in this study. Lake names are reported (please refer to Table 1 for correspondent core codes). Yellow-black polylines are roads; brown lines are elevation contours (100m); blue polygons are water bodies…………………….……..95

Figure 4-S2a. Age depth models of the thirteen cores used in this study. All age- depth models have been performed using clam v.2 in R………………………..…….100

Figure 4-S2b. Age depth models of the thirteen cores used in this study. All age- depth models have been performed using clam v.2 in R………………………..…….101

Figure 5-1. Map of the sampling locations in Tasmania, Australia. A total of 27 surface samples, mostly located within the Cradle Mountain-Lake St. Clair National Park, were collected during the field campaign. Walter and Lieth climate diagrams show similar hydrothermal conditions for the study locations……………………………………………………………………….…………..112

Figure 5-2. Vegetation survey and ring data extraction design. A maximum distance of 50 km has been considered for the Lagrangian Stochastic Model and Gaussian Plume Model. A total of 9 quadrats for each sampling location have been surveyed (see Methods). Background is represented by elevation contour lines in (a) and vegetation polygons in (b) and (c)………………………………………………………114

XXI

Figure 5-3. Optimization function scores and estimated RSAP from the Gaussian Plume and Lagrangian Stochastic models; a) plot of the first 2 km; b) plot of the first 20 km. Total maximum radius of the analysis is 50 km (Supplementary Figure 5- S5)………………………………………………………………………………..………….121

Figure 5-4. Plot comparing pollen productivity estimates (PPEs) obtained using the Gaussian Plume Model (blue shaded) and Lagrangian Stochastic Model (orange shaded) using 2 and 10 km radii. Asterisk (*) indicates the reference taxon…………………………………………………………………………………..……122

Figure 5-S1. a) RSAP results obtained considering a 2 km vegetation radius. The likelihood function scores produced by the ERV-submodel 1 sharply decrease closer to the sampling point (~500 m) compared to the values obtained using the ERV- submodel 2 and ERV-submodel 3, which both have yielded a very similar LFS curve with estimated RSAP at ~900 m. b) RSAP results obtained considering a 10 km vegetation radius. The introduction of the regional vegetation into the dataset produced a sharp shift in the LFS starting from 4 km, suggesting a worse fit between the modelled vegetation and pollen data, likely due to a lack of a uniform distant vegetation………………………………………………………………………………….134

Figure 5-S2a,b,c. Scatterplots of distance weighted plant abundance (DWPA %) vs pollen abundance (%). The y-intercept in the plots represents background pollen component for the surveyed area (50 km radius).. …………………………………………….……………….136

Figure 5-S3. Optimization function scores using the GPM and LSM for 50 km vegetation radius………………………………………………………………………….139

Figure 5-S4. Comparison of PPEs from LSM and GPM for 2, 10 and 50 km vegetation radius……………………………………………………………………………………….140

Figure 5-S5. Standard deviations of abundances (area covered) of the 13 target taxa in this study. Standard deviations across all the vegetation sampling sites and rings employed in the analyses have been averaged to check whether the plant abundances are more or less variable within the study sites. This is used as a measure of the plant abundance gradient present in our vegetation dataset. Asterisk (*) indicates the target taxon...... 141

Figure 5-S6. Plot of the ratio between PPEs generated with GPM and LSM for 10 km vegetation radius (PPEs GPM/PPEs LSM) and pollen fall speeds (m/s). The GPM clearly over-estimates pollen productivity estimates for small grains with low fall speed (e.g. Eucryphia lucida) and under-estimates it for larger pollen grains with high fall speed (e.g. Gymnoschoenus sphaerocephalus)………………………………. ………142

XXII

Figure 6-1. Location map of the study site (Dove Lake, Tasmania – TAS1507SC2, red star). The green star indicates the validation site (Lake St Clair). Black box represents the vegetation extent reconstructed by the REVEALS model (100 x 100 km). Red (green) circle is the 50 km buffer around Dove Lake (Lake St Clair) used to extract modern plant cover from the TasVeg 3.0 vegetation maps (Government of Tasmania, 2013) for the REVEALS validation process (see Methods). Polygons in the right panel represent vegetation formations. Lake area is shown as shades of blue representing bathymetry in metres……………………………………………………………………..149

Figure 6-2. Diagram showing pollen % of key taxa used for the vegetation reconstruction. Macro- and micro- charcoal accumulation rates (particles/cm2 yr-1) are also shown. Colours represent the grouping of the taxa according to vegetation structure: green is used for taxa generally occurring within forests; red is used for plant taxa commonly found in non-forested environments. Dashed lines indicate statistically determined pollen zones (see Chapter 6.6.4. for zones’ description)……………………………………………………………………..………….154

Figure 6-3. Plot of the REVEALS validation PCA. REVEALS vegetation estimates from the surface samples of Dove Lake and Lake St. Clair (Tasmania) were compared to the actual modern plant cover around 50 km from each lake. REVEALS results from four model runs (REVEALS-GPM3m/s, REVEALS-GPM6.5m/s, REVEALS-LSM and REVEALS-LSM windy) are shown. Axis 1 represents 68.5% of the variance, whereas Axis 2 represents the 16.5%. Grey arrows highlight taxa with a correlation with the PCA Axes larger than r=0.5. Filled symbols indicate data from Lake St. Clair, hollow symbols represent Dove Lake………………….……………...155

Figure 6-4. Comparison of pollen data and quantitative reconstruction results: summary diagram showing a) pollen percentages; b) REVEALS vegetation estimates using the Gaussian plume model under neutral conditions (REVEALS-GPM); c) REVEALS vegetation estimates using the Lagrangian stochastic model (REVEALS- LSM). Dashed lines indicate statistically determined pollen zones (same as Figure 6- 2)…………………………………………………….………………………………………157

Figure 6-5. Summary figure showing major palaeoenvironmental trends, including a) reconstructed plant cover for Gymnoscheonus sphaerocephalus (%); b) reconstructed total rainforest cover (%); c) Reconstructed Eucalyptus plant cover; d) Macroscopic CHAR record; e) Fire activity based on charcoal influx from two alpine sites in western Tasmania (Fletcher et al., 2015). Orange-yellow shading highlights periods of relatively low moisture, enhanced fire activity and low forest cover. Green shading identifies the period with maximum forest cover and low fire activity, suggesting wetter conditions………………...………………………………………………………..160

XXIII

Figure 6-S1. Digital elevation model of the mountainous area surrounding Dove Lake, Tasmania. Roads, water bodies and toponyms are also shown……………………………………………………………………….……………...165

Figure 6-S2. Age-depth model performed using Bacon for core TAS1507SC2. Five 210Pb dates and eleven 14C date were employed. Black solid line indicates the median age used to plot all the results presented in this work. Dashed grey lines indicate the minimum and maximum ages according to 95% confidence intervals…………………………………………………………………………………....168

Figure 6-S3. Extended pollen diagram for Dove Lake core including all taxa with abundance >1%...... 169

Figure 6-S4. Bar plot of the difference between actual cover and REVEALS-estimated abundances for single plant taxa. The REVEALS-LSM results generally show the lowest deviations from the predicted values (actual plant cover), especially the surface sample results from Lake St Clair…………………………………………...…173

Figure 6-S5. Plot of the average vegetation cover errors from the REVEALS-GPM and REVEALS-LSM runs for Dove Lake (a) and Lake St. Clair (b). Lowest average errors are evident for the REVEALS-LSM runs compared to REVEALS-GPM. REVEALS-GPM performed with a wind speed of 6.5 m/s shows a slightly better performance. In general, Lake St. Clair shows the lowest error estimates………………………………………………………………………………...…174

Figure 6-S6. REVEALS-GPM results and error bars for single taxa. REVEALS was run using 6.5 m/s and PPEs calculated with the same parameters (from Mariani et al., 2016)……………………………………………………………………………………175

Figure 6-S7. REVEALS-LSM results and error bars using for single taxa. REVEALS was run with LSM unstable parameters (from Kuparinen et al., 2007) using the set of PPEs published in Mariani et al. (2016)…………………………………………………176

Figure 6-S8. REVEALS-GPM and REVEALS-LSM results using different sets of atmospheric parameters and PPEs. REVEALS-LSM shows more coherent results. REVEALS-GPM was run under neutral conditions with a wind speed of 6.5 m/s and 3 m/s and the REVEALS-LSM was run using the unstable and windy unstable parameters from Kuparinen et al. (2007). To evaluate the coherence between models and PPEs, REVEALS and PPEs settings were mixed in four other runs: REVEALS- GPM set with a wind speed of 6.5 m/s was run using PPEs calculated at 3 m/s (and vice versa). Likewise, REVEALS-LSM set with unstable conditions was run using PPEs derived using a windy unstable setting (and vice versa). A total of eight model

XXIV

runs were performed in the DISQOVER package for R (Theuerkauf et al., 2016)………………………………………………………………………………………...177

Figure 7-1. a) Southern Hemisphere map showing the correlation between annual surface zonal wind speed (m/s) and annual rainfall (mm) from the ERA-40 Reanalysis dataset (NOAA); b) Regional charcoal influx from western Tasmania (five-point weighted average); c) Regional charcoal influx from southern South America (40-55°S) (five points- weighted average); d) El Niño proxy expressed as the number of events/100 yrs (Moy et al., 2002). Red fill in d) represents the values above 5 events/100yrs, considered as significant El Niño activity according to Moy et al. (2002). Extracted from Mariani and Fletcher, 2017 (Chapter 4)…………………182

Figure 7-2. Summary plot of the main time-series extracted from Chapter 4 (Mariani and Fletcher, 2017) and 6 (Mariani et al., 2017) a) Rainforest cover (%) around Dove Lake; b) Regional charcoal influx from western Tasmania; c) Macroscopic-CHAR record from Dove Lake; d) Microscopic-CHAR record from Dove Lake; e) Eucalyptus cover (%) around Dove Lake; f) El Niño number of events (Moy et al., 2002)……..185

Figure 7-3. Summary diagram showing rainforest (RF) pollen abundance (%) from the closest sites to Dove Lake (see Colhoun, 1996 for location map) compared to Dove Lake RF pollen % and land-cover estimates derived from the applications of pollen dispersal models performed in this Thesis (Mariani et al., 2017). Only Nothofagus cunninghamii and Phyllocladus aspleniifolius were summed as indicators of rainforest. Asterisk (*) indicates that the RF pollen % from Dove Lake were recalculated using the same subset of pollen taxa as in Colhoun (1996). Land-cover estimates (%) from Dove Lake were taken from Chapter 6 (Figure 6-4) and no alteration to pollen types was made.……………………………………….…….....…..190

XXV

LIST OF TABLES with CAPTIONS

Table 2-1. List of the main vegetation types in western Tasmania with information about area covered, main species and fire interval characteristics……………………43

Table 3-S1. List of years used for the Superposed Epoch Analysis (SEA)...... 64

Table 3-S2. List of charcoal records used in the palaeofire analysis…..…………...…67

Table 4-1. List of sites in western Tasmania used for the palaeofire analysis (map shown in Figure 4-1)……………………………………………………………………….80

Table 4-S1a. Radiocarbon dates obtained on the 13 cores analysed for this study. Ages were calibrated in OxCal 4.2 (Ramsey, 2013) using the Southern Hemisphere calibration curve (Hogg et al., 2013)………………………………………………….…..96

Table 4-S1b. 210-Pb dates obtained on a subset of the 13 cores analysed for this study………………………………………………………………………………………...99

Table 4-S2. List of sites from the Global Charcoal Database (GCD; Power et al., 2010) used for the paleofire compilation from Southern South America (SSA) between 40 and 55°S……………………………………………………………………………………102

Table 5-1. List of sites used for the vegetation surveys and surface sample collection…………………………………………………………………………………...113

Table 5-2. Pollen measurements and estimated pollen fall speeds for the 13 target taxa in western Tasmania. The B-axis is not listed for spherical pollen grains……………………………………………………………………………………….119

Table 5-3. PPEs obtained from the Gaussian Plume (GPM) and Lagrangian Stochastic (LSM) models at 2 and 10 km vegetation source radii. Superscripts indicate assumed zoophilous (z) or anemophilous (a) pollinating plant taxa……………………………………………………………………………………….....123

Table 5-S1 - List of the TasVeg 3.0 vegetation types surveyed and target taxa abundances (% cover)……………………………………………………………...……..131

Table 5-S2. PPEs from ERV submodels and standard errors………………………...135

Table 6-1. Table showing fall speeds and PPEs for the ten key pollen taxa used for the vegetation reconstruction from Dove Lake (data from Mariani et al., 2016). GPM= Gaussian plume model; LSM= Lagrangian stochastic model. Superscripts f and nf are used to differentiate forest and non-forest plant taxa…………………………………153

XXVI

Table 6-S1a. Table listing all 210Pb dates obtained on core TAS1507SC2……...…....167

Table 6-S1b. Table listing all radiocarbon dates obtained from core TAS1507SC2. Asterisks indicate 14C dates excluded from the age-depth model. Calibrated ages given in the right-hand columns……………………………………………..………….167

Table 6-S2. Comparison of new PPEs calculations and published PPEs in Mariani et al. (2016). LSM parameters for unstable and unstable windy conditions are described in Kuparinen et al. (2007). Highlighted in bold are the published PPEs as in Mariani et al. (2016)………………………………………………………………………………....168

XXVII

LIST OF ACRONYMS

o AP: arboreal pollen o CHAR: Charcoal accumulation rates o CE: Common Era o ENSO: El Niño Southern Oscillation o ERV: Extended R-value o GAMs: Generalized Additive models o GDC: Global Charcoal Database o GPM: Gaussian plume model o IOD: Indian Ocean Dipole o ITCZ: Intertropical Convergence Zone o ka: thousands of years ago o kyr: thousands of years (duration) o LRA: Landscape Reconstruction Algorithm o LSM: Lagrangian stochastic model o NAP: non-arboreal pollen o NPI: Northern Patagonian (palaeovegetation) Index o PCA: Principal Component Analysis o PDO: Pacific Decadal Oscillation o PPEs: Pollen productivity estimates o REVEALS: Regional Estimates of VEgetation Abundance from Large Sites o RSAP: Relevant Source Area of Pollen o SAM: Southern Annular Mode o SEA: Superposed Epoch Analysis o SH: Southern Hemisphere o SOI: Southern Oscillation Index o SSA: Southern South America o SSTs: Sea surface temperatures o STH: Sub-Tropical High Pressure belt o SWW: Southern westerly winds o WTAS: Western Tasmania

XXVIII

Chapter 1. Introduction and Thesis aims

1.1. Introduction

Fire is one of the main drivers of terrestrial ecosystem dynamics, affecting vegetation distribution and structure around the globe (Bowman et al., 2009). Without fire, closed forests would likely double their land-cover, mostly at the expense of grasslands and shrublands (Bond et al., 2005). Additionally, wildfires have substantial economic and social impacts, which may be exacerbated by changing climatic conditions and human population growth (Lannom et al., 2014; Moritz et al., 2014; Westerling et al., 2016). During the last two decades in the developed world, extreme wildfires reported as being economically or socially catastrophic were concentrated in suburban areas intermixed with flammable forest, with a hotspot of disastrous events located in southeastern Australia (Sharples et al., 2016; Bowman et al., 2017). Despite the clear importance of wildfires in shaping vegetation and impacting human lives (Lannom et al., 2014; Sharples et al., 2016; Bowman et al., 2017), the climatic drivers of fire activity through time are poorly understood in many regions on Earth (e.g. Le Goff et al., 2007; Holz and Veblen, 2011; Román- Cuesta et al., 2014; Holz et al., 2017). Under the current changing climatic regime in which wildfires are predicted to increase in frequency and magnitude (McWethy et al., 2013; Power et al., 2013; Moritz et al., 2014), it is important we gain a better understanding of the link between past climatic trends and fire activity in order to properly manage fires and landscapes, preserve valuable natural ecosystems and protect human lives and properties.

Australia is the most fire-prone continent on Earth (Hennessy et al., 2005) and fire is pivotal in shaping vegetation patterns (e.g. Bradstock et al., 2002; Bowman, 2000) and in causing socio-economic disasters (e.g. Bowman et al., 2017). Fire activity is projected to increase in response to climate change, especially in the temperate regions of Australia (Moritz et al., 2012), where highly-flammable eucalypt forests are juxtaposed against a rapidly expanding ‘bush-urban interface’ (Sharples et al.,

1 |

2016; Bowman et al., 2017). These highly flammable eucalypt forests also abut ancient (Gondwanan) fire-sensitive vegetation formations: the rainforests of southeast Australia. These rainforests are currently restricted to tiny pockets of fire refugia (Bowman, 2000) that are under threat from recent climate and human-driven wildfires, which have decimated large tracts of this fire sensitive vegetation, for instance, in Tasmania (e.g. Holz et al., 2014).

Tasmania is a cool-temperate continental island that was separated from mainland Australia around 12,000 years ago when rising sea-level flooded the Bassian Plain forming the Bass Strait (Chappell and Thom, 1977). Tasmania hosts rainforests dominated by Gondwana-linked plant taxa such as Phyllocladus, Nothofagus and Athrotaxis (Hill, 1990). Species-poor remnants of these rainforests also occur in the southeast Australian mainland. These fire-sensitive rainforests have undergone a massive range contraction in response to increasing aridity and fire activity (Hill, 1994; Sniderman and Haberle, 2012) and the largest remnant pockets of these rainforests survive today within a landscape-scale mosaic of vegetation types in western Tasmania. This landscape pattern has been considered a product of its past fire history (Jackson, 1968; Brown and Podger, 1982a). While there has been considerable focus on the role of fire in shaping and reinforcing the current fine- scaled mosaic of fire-sensitive and fire-adapted vegetation states in western Tasmania (Jackson, 1968; Mount; 1979; Bowman and Jackson, 1981; Bowman, 2000; Bowman and Wood, 2009; Wood and Bowman, 2012), there is scant evidence for shifts between these vegetation states (Fletcher et al., 2014b). This lack of information exposes a gap in our understanding of the role of fire in driving shifts between fire- sensitive and fire-adapted vegetation states, increasing the vulnerability of this landscape to the potential effects of increasing climate-driven wildfires in this region.

Model-based projections of fire danger under a global warming scenario show a significant increase for Tasmania in the coming century (Fox-Hughes et al., 2014). While considerable effort has been made to understand what influences potential fire weather in Tasmania (e.g. Fox-Hughes et al., 2014; Grose et al., 2014), our current

2 |

understanding of the regional climatic controls of actual fire activity is severely limited. Ignitions in Tasmania are largely attributed to humans (who have occupied the region for >35,000 years), with lightning strikes accounting for less than 0.1% of ignitions during historical times (Ingles, 1985; Jackson and Bowman, 1982; Bowman and Brown, 1986). Given the constant source of ignition (i.e. human presence) and the abundant biomass in this landscape (i.e. fuel), the principal factor controlling the occurrence and spread of fire through time is thought to be climate and its effect on fuel moisture content – i.e. fires only occur when fuel is dry enough to burn (e.g. McWethy et al., 2013).

Climatic variability in Tasmania is driven by changes in two of the main climate features that operate in the Southern Hemisphere: the El Niño Southern Oscillation (ENSO) and the southern westerly winds (SWW) (Gillett et al., 2006; Hendon et al., 2007; Garreaud, 2009; Hill et al., 2009; Risbey et al., 2009). The short-term variability in the strength and position of the SWW is described by the Southern Annular Mode (SAM) index (Marshall, 2003). The warm phase of ENSO, El Niño, is linked to reduced rainfall and increased wildfire activity across southeast Australia, including Tasmania (Risbey et al., 2009; Cai et al., 2009; Mariani et al., 2016). As global temperature rises, El Niño activity is amplifying (Power et al., 2013; Cai et al., 2014), potentially resulting in an increase in the occurrence of droughts and wildfires in southeast Australia (Guilyardi, 2006). Likewise, the persistent positive trend in SAM over the last ca. 60 years (Perlwitz et al., 2008; Thompson et al., 2011) has seen storm- tracks embedded in the SWW move away (south) from southern Australia, progressively reducing moisture delivery to this region (Smith and Reynolds, 2005; Fogt et al., 2009). The potential influence of these ongoing climatic shifts over Tasmanian fire activity is unknown, adding to the uncertainty of future climate-fire- vegetation dynamics in this region.

While documentary records provide a fine-scale understanding of the spatial pattern and recent controls over fire activity, this information source is time limited (e.g. last 30 years in Tasmania). Due to the magnitude of projected change, inferring future ecosystem dynamics based only on short-term observations is difficult (Willis and

3 |

Birks, 2006), with time-limited perspectives potentially resulting in inadequate conservation strategies with unexpected consequences under new climatic regimes (Williams and Jackson, 2007; Gillson et al., 2013; Whitlock et al., 2017). Palaeoecological/palaeoclimatic information, spanning time scales from centuries to millennia, uncovers a variety of ecosystem processes that occur at different spatial and temporal scales. For instance, palaeoclimatic research reveals that many of the controls over the global climate system and fire regimes operate at scales ranging between inter-annual and multi-millennial (e.g. Verdon et al., 2004; Holz et al., 2011; Fletcher et al., 2014a; Holz et al., 2017). Hence, we need to look at longer-term datasets, such as fossil records, to gain a fuller understanding of what drives fire activity and how changes in fire regimes influence terrestrial ecosystems and their processes. This is particularly the case for slow acting systems, such as Tasmanian rainforests, in which the ecological response to fire is measured in centuries to millennia (e.g. Lindblah et al., 2013; Fletcher et al., 2014a).

Sedimentary records are an ideal source of long-term ecological data because of the broad temporal and spatial coverage they can provide, such as reconstruction of fire and vegetation history at scales ranging from local to global, and from inter-annual to millennial (e.g., Carcaillet et al., 2002; Iglesias and Whitlock, 2014; Marlon et al., 2016). While palaeoecological data retrieved from sedimentary archives is an abundant and well utilised data source for understanding long-term ecology, it suffers from a number of critical limitations that limit the potential for this information to inform ecosystem management (e.g. Björck and Wohlfarth, 2002; Blaauw, 2010; Faegri and Iversen, 1989; Gaillard et al., 2010). The most significant of these limitations is, arguably, the disparity between the fossil assemblages and actual ecosystem composition. For example, fossil pollen spectra are often dominated by a few plant taxa, masking the true nature of vegetation (e.g. Faegri and Iversen, 1989; Bunting, 2003).

4 |

This may be the case in western Tasmania, where a long-standing debate exists about the origin of a landscape dominated by treeless moorland (e.g. Colhoun, 1996; Fletcher and Thomas, 2010a), mostly due to the ‘silent’ signal of this vegetation type in pollen spectra from this region (e.g. Pickett et al., 2004; Fletcher and Thomas, 2007a). The moorland communities of western Tasmania (dominated by Gymnoschoenus sphaerocephalus) represent one of the largest peatland systems in the Southern Hemisphere (Gallego-Sala and Prentice, 2013) and are the product of an exceptional combination of high fire activity and high rainfall (Reid, 1999; Whinam and Hope, 2005). The projected temperature increase coupled with rainfall reduction threatens these systems with a severe reduction in cover (Gallego-Sala and Prentice, 2013), highlighting the necessity of developing informed conservation plans in this region. By obtaining a better understanding of the past moorland cover over long time-scales, we can provide a solid baseline reference to improve the management of these ecosystems.

1.2. Thesis aims

The overall Thesis aim is achieving a better understanding of climate-fire-vegetation dynamics at multiple time scales in western Tasmania. Specifically, this Thesis endeavours to achieve three aims, ranging from short-term (inter-annual) climate and fire activity relationships to long-term (multi-millennial) climate, fire and vegetation interactions at a regional scale. Further details about the overall methodological approach, workflow and background information are presented in Chapter 2.

Aim I. Understanding short-term climate drivers of fire activity in western Tasmania

Western Tasmania is characterised by high rainfall (up to 3500 mm/year) and high forest cover (i.e. high biomass availability). Because of the great fuel availability, fires in western Tasmania are generally regarded as being limited by fuel moisture (i.e. climate) (McWethy et al., 2013), but currently there are few empirical studies that

5 |

support this assertion (Styger and Kirkpatrick, 2015). Current fires in Tasmania are modulated by seasonal, inter-annual, and decadal variations in temperature and rainfall - i.e. ﬁres occur in response to hot and dry conditions (Nicholls and Lucas, 2007; Grose et al., 2014), but there are no data about the broad-scale (i.e. SWW, ENSO) climatic drivers of fire activity. Rainfall in western Tasmania is derived from the orographic uplift of the SWW over the dominant NW-SE trending mountain range and inter-annual rainfall variability was found to have a significant correlation with the SAM index (Hill et al., 2009), reinforcing the notion that the SWW represent the main control rainfall variability in this region. This fact prompts the question of whether the inter-annual variability in the SWW play an important role in modulating fire activity in this region.

To satisfy Aim I, I will (1) explore the relationship between the SAM index and historical records of ﬁre occurrence in western Tasmania and (2) test whether the persistent positive SAM trend over the last 500 years (Abram et al., 2014) has influenced ﬁre activity in this temperate region. Disentangling the large-scale climatic controls on fire activity in Tasmania is crucial to understand the past and current fire variability in the region and make informed projections on future fire activity. Moreover, the findings of this analysis are important for the interpretation of long-term fire activity (palaeofire) records in the western Tasmanian region (see Aim II).

Aim II. Reconstructing long-term climate and fire variability over the Holocene in western Tasmania

Adopting a long-term perspective to understand the cause of changing fire regimes has a critical importance for managing fire and ecosystems in the present and the future (e.g. Conedera et al., 2008; Marlon et al., 2016; Whitlock et al., 2017). Creating effective land management approaches requires a solid understanding of past landscape and fire dynamics, and it requires knowing the natural range of variability in the disturbances and processes that have shaped such ecosystems before the appearance of the modern landscape (Conedera et al., 2008).

6 |

The analysis of charcoal particles deposited in lake sediments is used to reconstruct past fire activity from multi-millennial to decadal time scales (Whitlock and Larsen, 2001; Whitlock and Bartlein, 2004) that we can compare in relationship to climate proxies. In the last decade, regional and global syntheses of sedimentary charcoal records have been used to examine broad-scale patterns in past fire activity and analyse the linkages among fire, climate, vegetation and humans (Carcaillet et al., 2002; Power et al., 2008; Daniau et al., 2012; Marlon et al., 2008; Marlon et al., 2013). Long-term climatic linkages with fire activity in Tasmania have been recently explored (e.g. Fletcher et al., 2014a; Fletcher et al., 2015), and the authors suggested a tight relationship between fire history and SWW and ENSO dynamics. Nevertheless, these studies only presented data from single locations (local-scale analyses) and a synthetic (regional) approach is necessary to have a better understanding of palaeofire activity in relationship to climatic fluctuations (e.g. Power et al., 2008). Integrating data from multiple sites mitigates against the influence of local-scale factors (e.g. topography) and allows a more robust understanding of regional trends, which are driven by broad-scale climatic controls (Power et al., 2008), such as shifts in the SWW or ENSO intensification. Given its geographical and physiographic characteristics, Tasmania constitutes an ideal location to disentangle the interplay between ENSO and SWW in modulating fire activity. The western portion of the island lies mostly under the influence of SWW, whereas the north and east of the island are under the ENSO-domain (Hill et al., 2009).

With the work presented in this Thesis, I aim to understand the role of SWW and ENSO activity in driving multi-millennial and sub-millennial changes in palaeofire regimes in this region. This knowledge forms a foundation for subsequent analysis of how these climatic and fire regimes changes affect the regional vegetation patterns across western Tasmania (Aim III). This thesis adopts a two-step approach to correct these biases and produce robust estimates of past land-cover.

7 |

Aim III. Reconstructing climate- and fire- driven regional vegetation changes in western Tasmania

The robust quantification of the effects of fire and climate change on terrestrial ecosystem depends on our ability to reconstruct past vegetation changes (e.g. Gaillard et al., 2010). The semi-quantitative nature of Australian palynology (Kershaw, 1986; Fletcher and Thomas, 2010a,b) has allowed objective inferences of vegetation change from pollen data, yet we remain uninformed about the actual degree of past vegetation change due to biases in pollen records.

Step 1: Calibrating pollen-vegetation data using dispersal models

Several factors influence the representation of plant taxa in pollen spectra (e.g. taphonomy, pollen productivity, dispersal capabilities) (Faegri and Iversen, 1989) and some plant species are found to be over- or under- represented in the pollen rain (e.g. Fletcher and Thomas, 2007a; de Nascimento et al., 2015). In turn, there is a nonlinear relationship between pollen percentages and plant cover (Prentice and Webb, 1986; Faegri and Iversen, 1989; Berglund, 1986), such that estimating palaeovegetation cover from sedimentary pollen composition requires advanced modelling of pollen-vegetation relationships (e.g. Sugita, 2007a,b; Bunting et al., 2013).

A key step in this approach is estimating the pollen production and dispersal for selected plant taxa using distance-weighted plant cover data and pollen proportions from modern surface samples. The incorporation of this information within pollen dispersal models provides the basic taxon-specific parameters for developing pollen- based vegetation reconstructions: pollen productivity estimates (PPEs) (e.g. Bröstrom et al., 2005; Abraham et al., 2012; Bunting et al., 2013; Bunting et al., 2015; Li et al., 2015). Within this Thesis, PPEs will be assessed for thirteen key Tasmanian pollen taxa in order to achieve a quantitative assessment of long-term vegetation changes in this region (see Step 2, below). This work represents the first application of dispersal models for the estimation of pollen productivity in Australia.

8 |

Step 2: Model-based regional vegetation reconstruction

The quantification of past vegetation (or past land-cover) changes and its incorporation into climate models is a priority in the current global climate modelling community (e.g. Meehl et al., 2007; van der Linden and Mitchell, 2009; Friend et al., 2014). Quantitative assessment of past landscape alterations is also pivotal for sustainable landscape management and conservation, archaeological reconstructions and for addressing biological/geomorphological issues (e.g. Gaillard et al., 2010). Conservation efforts must be directed depending on the past land-use of a certain landscape (Colombaroli et al., 2017). Landscapes characterised by a mixture of pristine vegetation and managed lands are challenging for conservationists because of the need to preserve both natural values and cultural heritage (Lindenmayer and Hunter 2010; Whitlock et al., 2017). The development of quantitative palaeovegetation estimates allows us to have a better understanding of the past landscape history in a certain area and appropriately direct management strategies. For example, quantitative reconstructions of past land-cover carried out in Northern Europe showed that pollen percentages severely underestimated non- arboreal cover in cultural landscapes (Brostrom et al., 2004; Gaillard et al., 2010; Soepboer et al., 2010; Trondman et al., 2015), affecting the appropriateness of nature conservation plans and the efficiency of climate models (Trondman et al., 2015).

A case in point is the origin and evolution of the present-day dominance of pyrogenic moorland in western Tasmania is the subject of a long-standing debate (see Colhoun, 1996; Fletcher and Thomas, 2010a,b; Macphail, 2010 and references therein). According to Macphail (1979) and Colhoun (1996), climate amelioration in Tasmania during the post-glacial turned most of the treeless landscape of the Last Glacial Maximum (LGM) into forest and woodlands. This model has been rejected by Fletcher and Thomas (2010a), who proposed that moorland became established across the region during the last glacial cycle (ca. 35 kyrs) and was then maintained by anthropogenic burning. Fletcher and Thomas (2010a) argued that the biases present in the pollen records are the cause of misinterpretation of the former Holocene vegetation models by Macphail (1979) and Colhoun (1996). Fletcher and

9 |

Thomas (2007a,b, 2010a,b) used modern pollen-vegetation analysis to show that Tasmanian pollen spectra are skewed towards plants that produce large amounts of well-dispersed pollen (i.e. Nothofagus cunninghamii – syn. Lophozonia cunninghamii - and Phyllocladus aspleniifolius), biasing the interpretation of pollen records in favour of rainforest. Thus, debate over the origin and development of the modern treeless landscape of western Tasmania centres on how pollen production and deposition biases are taken into account.

The state-of-the-art quantitative approach adopted in this Thesis will address pollen production and dispersal biases to provide more objective insights into the past land-cover of rainforest and treeless (moorland) communities in western Tasmania. I present the first application of mechanistic models for regional vegetation reconstruction to quantify landscape changes in the Southern Hemisphere. These reconstructions allow us to address biogeographical and methodological debates over (1) the efficiency of mechanistic models for vegetation reconstruction and (2) the timing and extent of long-term vegetation changes in western Tasmania.

10 |

1.3. Organisation of the Thesis

The Thesis is organised in accordance with the aims presented above. Following the introductory first chapter, the second chapter presents the overall methodology and workflow used in this PhD project, which is complementary to the methods outlined in later chapters (Chapters 3-6). In the third chapter, the short-term variability of historical fire activity in western Tasmania is documented, alongside the climate drivers, satisfying Aim I. The contents of this chapter were published in Geophysical Research Letters in 2016. The fourth chapter addresses Aim II and contains the results of the long-term analysis of regional fire activity published in Quaternary Science Reviews in 2017. Chapter five deals with the application of dispersal models to calibrate pollen-vegetation relationships in western Tasmania (Aim III – step 1). The contents of this chapter were published in Quaternary Science Reviews in 2016. Chapter six addresses Aim III (step 2) with a quantification of climate- and fire- driven land-cover changes through the Holocene from Dove Lake. These results were published in Journal of Biogeography in 2017. Chapter seven integrates the findings from all the papers and presents a critical analysis of the limitations of the methodologies applied with suggestions for future work. In the final, eighth chapter, a summary of the conclusions of this Thesis is presented.

11 |

Chapter 2. Background and methods

This PhD project was based on a three-phase development, involving a suite of palaeoecological approaches (charcoal and pollen analyses, radiometric dating), vegetation surveys and quantitative vegetation reconstruction. Each phase outlined below corresponds to the aims listed in Chapter 1.2.

2.1. PhD project workflow

Red bounded boxes represent methodological steps

PHASE 1 (AIM I): UNDERSTANDING CLIMATE DRIVERS OF SHORT-TERM FIRE ACTIVITY IN WESTERN TASMANIA (Chapter 3)

statistical analyses charcoal processing Documentary coring chronology and counting records of fire Multi-site compilation of charcoal occurrence records

Comparison with What drives short-term short-term climate climate- and fire- time series and variability in western

reconstructions Tasmania?

PHASE 2 (AIM II): RECONSTRUCTING LONG-TERM CLIMATE AND FIRE VARIABILITY IN WESTERN TASMANIA (Chapter 4)

charcoal processing coring chronology and counting

Multi-site compilation of charcoal records

Comparison with What drives long-term long-term climate climate- and fire- 12 | time series and variability in western reconstructions Tasmania? PHASE 3 (AIM III): RECONSTRUCTING REGIONAL VEGETATION CHANGES IN WESTERN TAMASMANIA (Chapters 5 and 6)

STEP 1: MODERN POLLEN-VEGETATION CALIBRATION (Chapter 5)

Vegetation surveys Analysis of modern + pollen samples

GIS vegetation extraction

How can we correct for

Application of pollen dispersal productivity biases in the models to obtain pollen Tasmanian pollen productivity estimates for records? Tasmanian plant taxa

STEP 2: REGIONAL VEGETATION RECONSTRUCTION (Chapter 6)

pollen processing pollen productivity coring chronology and counting estimates (Step 1)

Quantitative vegetation reconstruction using the REVEALS model

How did past changes in climate and fire activity impacted the landscape of western Tasmania?

13 |

2.2. Palaeoecology

Paleoecology aims to understand the past variability in ecosystem composition and distributions and it provides a long-term perspective on the present ecological systems to address specific conservation issues (e.g. Willis and Birks, 2006). Palaeoecology uses parameters that track changes in fossil organisms and geochemical characteristics (proxy data) from natural archives (e.g. lake sediments, ice cores, speleothems and other depositional environments) to reconstruct past ecosystem changes (Birks and Birks, 2006). Palaeoecological data can be derived from fossil and geological sedimentary sequences that provide insights into past vegetation, fire regimes and environmental changes spanning inter-annual to multi- millennial time scales (Mayle and Iriarte 2014). These can be used to deduce ecosystem variability, biotic and abiotic responses to disturbances, baseline conditions and thresholds (Willis and Birks, 2006; Seppä and Bennett, 2003; Willis et al., 2010; Lindbladh et al., 2013).

With long-term data, we can infer the naturalness of certain landscapes or ecosystems, for example by solving problems related to biological invasions, or by defining baselines for the impacts of wildfires occurrences and climatic change (e.g. Willis and Birks, 2006). Critically for landscape management and conservation, the short time scale recorded in historical ecological data (e.g. last decades), does not allow natural variability to be disentangled from other trends in the records (Willis and Birks, 2006). For instance, regarding wildfires, conservationists are concerned whether changes in fire frequency, severity and burning extent are part of natural variability or are anthropogenic (McKenzie et al., 2004). Establishing the natural variability of wildfires is important since it can be used as a benchmark against which to evaluate contemporary conditions and make alternative plans that can lead to more appropriate land and biodiversity management (Whitlock, 2004).

Today, conservation efforts to protect forested landscapes are challenged by climate change projections suggesting an important future rearrangement of vegetation and fire regimes (Diffenbaugh and Field 2013; Whitlock et al., 2017). Adopting a palaeoecological approach to take up this challenge is crucial, since palaeoecological

14 |

studies offer a baseline for interpreting current landscape conditions and help set goals for conservation and restoration of threatened environments. The amount of conservation efforts needed to preserve a certain landscape depends on its land-use history (Figure 2-1; Colombaroli et al., 2017). Between nearly pristine landscapes at one end and highly altered landscapes at the other, intermediate conditions are characterised by both natural and cultural components. These intermediate landscapes are the most challenging for conservationists because of the need to support both their natural values and their cultural heritage (Lindenmayer and Hunter 2010; Whitlock et al., 2017). The use of palaeoecology can inform us about the past land-use of a specific landscape, thus we become enabled to place this landscape correctly along the land-use gradient and evaluate management strategies appropriately.

Figure 2-1. Diagram showing the gradient of landscapes with different levels of anthropogenic alteration (modified from Colombaroli et al., 2017).

15 |

2.3. Palaeoecology fundamentals

Every sediment-based palaeoecological study has two pivotal elements: (1) sediment extraction and (2) chronological control. In this subchapter I present the coring and dating techniques adopted in this PhD project.

2.3.1 Coring techniques

Coring is the first necessary step to collect the material (e.g. lake sediments) to be analysed and it can be performed with a large variety of techniques (e.g. Livingstone, 1955; Wright et al., 1984; Renberg, 1991). In lakes, sediment accumulation is influenced by sediment ‘focusing’, implying the re-suspension and transport of sediment from the shallower to deeper zones of the lake (e.g. Davis 1971). Therefore, the deepest section of a lake is defined as the accumulation zone where no further focusing takes place and accumulation is greatest (Lehman, 1975; Davis, 1973; Blais and Kalff, 1995). At each location, a depth profile of the lake basin was performed to determine the coring site precisely, by looking for the deepest point of each lake basin. A simple bathymetric map was developed for all lakes using a Hondex© hand-held depth sounder.

Coring was performed at the deepest point of the lake basins using a UWITEC© Universal Gravity Corer from a platform or an inflatable boat. This corer allows the collection and preservation of the sediment-water interface, so the information preserved in the top centimetres of the core can be used to reconstruct palaeoenvironmental change of recent times, potentially up to the present days (Aaby and Digerfeldt, 1986).

16 |

2.3.2 Chronology

Establishing a good chronology is important for any palaeoecological work, since we must be able to compare our results to other records at either local, regional and global scales (e.g. Björck and Wohlfarth, 2002). Dating for this project was performed using the analysis of radioactive lead (210Pb) and carbon (14C) isotopes to achieve robust age-depth models for all cores.

Radiocarbon dating

Radiocarbon dating is based on the premise of radioactive decay of carbon isotopic ratios correlated to a known atmospheric carbon calibration curve at a particular time (Hedges, 1981; Björck and Wohlfarth, 2001; Walker and Lowe, 2000; Olsson, 1991). Radiocarbon (14C) is produced in the upper atmosphere by the bombardment of 14N atoms by cosmic-ray neutrons. It is then incorporated into living organisms through the photosynthetic process and, after death, 14C concentrations decrease exponentially following laws of radioactive decay. Given the 14C concentrations relative to 12C at the time of measurement and the known half-life of this isotope, it is possible to calculate the age of an organic-rich sample. Due to the relatively short half-life of 14C (5730 years) the dating range of this method is limited to the last ~50 kyrs (Bard et al., 2004). The highly organic sediments and the time-span covered by the selected Tasmanian lakes (11,700 kyrs) make radiocarbon dating the most suitable dating technique for this study. The Southern Hemisphere calibration curve was employed to calibrate 14C years to calendar years before present (cal yr BP, where BP=1950 AD) (Hogg et al., 2013).

Radiocarbon dating was performed at the Australian Nuclear Science and Technology Organisation (ANSTO) and at DirectAMS (Bothell, WA, USA) using Accelerator Mass Spectrometry (AMS) technique, which allows small amounts of organic material (1-3 mg) to be dated. Bulk sediment samples were used to date desired levels of cores, given the absence of fossil plant material (macrofossils). Sediments followed standard ABA (Acid-Base-Acid) pre-treatment procedures to

17 |

eliminate carbonates and potential contaminants, such as humic acids (e.g. Brock et al., 2010). The ABA pre-treatment involves washing the samples with hot hydrochloric acid (HCl) followed by multiple sodium hydroxide (NaOH) washes and a final HCl wash before sample drying (e.g. Brock et al., 2010).

Although careful pre-treatment is applied, bulk sediment dating may carry potential biases. For instance, the release of catchment-derived carbon into a lake basin (known as “reservoir effect”) would systematically offset the “true” age of deposition with “older” carbon (Bertrand et al., 2012). Other issues with bulk samples is the down-profile migration of humic acids, root penetration or bioturbation, which can make the radiocarbon ages erroneously young (Kaland et al., 1984).

A total of 120 radiocarbon dates were obtained on the cores for this project. The list of radiocarbon dates determined for the multi-site compilations are presented in Chapter 4. The list of radiocarbon dates obtained on the Dove Lake core is given in Chapter 6.

Lead-210

Nuclear weapons testing doubled the amount of radioactive carbon in the atmosphere in the late 1950s and early 1960s (e.g. Reimer et al., 2004). The alteration of atmospheric composition following these events limits the use of 14C dating for young samples (e.g. Reimer et al., 2004). Therefore, short-lived isotopes, such as 210Pb, are used for dating sediments younger than this age (Appleby, 2008). The 210Pb method is used to determine the accumulation rate of the uppermost sediment, going back in time approximately 100 years, based on the decay profile of this radioactive isotope (Appleby, 2008). Lead-210, which is part of the uranium-238 (238U) decay series, is absorbed onto sediment particles deposited in lakes (Oldfield and Appleby, 1984). The total 210Pb activity has two components: the supported 210Pb, deriving from in situ decay of the parent radionuclide 226Ra and the unsupported 210Pb, deriving from the atmospheric flux. By determining the

18 |

background levels of unsupported 210Pb activity and comparing the unsupported 210Pb activity measured for the core top, this method can provide an age estimate for a given depth (Appleby, 2008).

The 210Pb profile of several cores used in this project was determined by Alpha Spectrometry at ANSTO. A total of 43 210Pb measurements were obtained and the list is shown in Chapter 4 (for the multi-site comparison) and Chapter 6 (for Dove Lake). To account for potential sediment supply variations that may occur in response to catchment changes, the calibration of the 210Pb dates was done by using the constant rate of supply (CRS) model. The CRS model assumes a constant rate of 210Pb supply to sediments, but permits variations in sedimentation rate (Appleby and Oldfield, 1978).

Age-depth modelling

Creating age-depth models is a necessary step in order to interpolate the obtained radiocarbon ages to all sample depths analysed in the sediment cores. Two different packages for R (R Core Development Team, 2013) were used to model the chronologies of the data presented in Chapters 3, 4 and 6.

In Chapters 3 and 4, the 14C and 210Pb dates collected from thirteen sites across Tasmania were modelled using the clam package (Blaauw et al., 2010) in R (R Core Development Team, 2013). Clam performs classical age-depth modelling using a spline smoother. The same parameters were applied to all cores to maintain consistency among all records. Age-depth models for all the cores are presented in the Supplementary Data section of Chapter 4.

In Chapter 6, dates obtained from 14C and 210Pb analyses from the Dove Lake core were used to create a Bayesian age-depth model using the Bacon package (Blaauw and Christen, 2013) for R (R Core Development Team, 2013). Calibrated radiocarbon dates often result in widened, asymmetrical or multi-peaked calendar age estimates. Compared to classical age-depth modelling, which reduces calibrated 14C

19 | distributions to single point estimates with symmetric error distributions, Bacon provides an alternative to point estimates by considering the full probability of calibrated ages and conducts repeated random sampling of the calibrated distributions. Bacon incorporates the prior assumption that deeper sediments are always older than the upper sediments and automatically rejects models that present age reversals (Blaauw and Christen, 2013).

2.4. Fire activity and its reconstruction

Fire is a worldwide phenomenon that began soon after the appearance of terrestrial plants 420 million years ago, influencing ecosystem patterns and processes globally and affecting vegetation distribution and structure (Bowman et al., 2009). The impact of fire on a global scale has been demonstrated by Bond et al. (2005), who reported a replacement of substantial areas of fire dependent grass- and shrub-land by trees in a simulated world without fire. They concluded that biomes over much of the Earth have not reached their climatic potential and that fire is the only disturbance agent capable of reducing biomass at a global scale.

In Australia, fire is particularly pervasive in shaping vegetation patterns (Bradstock et al., 2002) and it has been suggested as the most critical factor controlling the fragmentary distribution of fire-sensitive rainforest within vast tracts of flammable sclerophyll vegetation from the monsoonal north to the temperate south of the country (Bowman, 2000). The vegetation mosaic of western Tasmania represents an example of the role of fire in determining vegetation landscapes in Australia, with large areas in which rainforest is the climatic potential vegetation, instead occupied by fire-promoted vegetation (e.g. sclerophyll, moorland). Western Tasmania has been categorised as an area with high biomass availability, where fire activity over time is modulated by moisture variability (Cochrane, 2003; Pausas and Ribeiro, 2013; Moritz et al., 2012; McWethy et al., 2013), but few empirical observations have been produced thus far (Styger and Kirkpatrick, 2015) and no studies investigating the role of regional-scale drivers (i.e. SWW, ENSO). In this Thesis, I address this

20 |

knowledge gap by analysing the short- and long-term variability of fire activity in this region and by comparing these trends with climate and palaeoclimate data.

Depending on the spatio-temporal scales of the proxies analysed, it is possible to infer different factors in determining the occurrence of fires over time and space (Figure 2-2). Historical (documentary) records can provide information such as absolute timing, total area burnt and ignition cause, thus represent a relatively accurate tool for understanding drivers of fire occurrence over short time frames (daily, annual or decadal; Figure 2-2). The information contained in historical databases is crucial if we aim to understand the controls on fire, since we can analyse time-series of actual fire occurrences along with observational climate data (see Chapter 2.3.1 and Chapter 3). At this time-scale, fire occurrence in any given place is determined by the confluence of sufficient fuel loads, an ignition trigger (i.e. heat source: lightning or humans), and suitable weather conditions (Figure 2-2b; Pyne et al., 1996; Scott, 2000; Scott et al., 2013). In this Thesis, I analyse documentary fire data spanning the last 30 years (i.e. available observational data) at seasonal and annual resolutions in western Tasmania to address Aim I (as introduced in Chapter 1). Long- term past fire activity can be reconstructed using charcoal particles in sedimentary records from lakes and wetlands (e.g. Whitlock and Larsen, 2001; see Chapter 2.3.2, Chapter 3 and 4) for periods spanning the last decades to the last million years (Figure 2-2). The long-term controls of fire are different than the short-term (Figure 2-2a): regional variables, such as climatic change and vegetation, become crucial over longer time scales and larger spatial extents. Further, at this long-term scale and over regional extents, studies suggest that we cannot distinguish the anthropogenic component of fire activity from climate or vegetation change, because humans are considered to operate at local and short temporal scales (Bowman et al., 2011; Scott et al., 2013; Vanniére et a., 2016). In this Thesis I reconstruct palaeofire activity at a regional scale in western Tasmania over the last 12,000 years (Holocene period) by using multi-decadal/centennial resolution charcoal records from lake sediments.

21 |

2.4.1. Documentary records of fire activity

Historical records of fire occurrence can be used to reconstruct fire variability over the short-term, from daily to decadal scales, and over spatial extents ranging from local to global (Figure 2-2, above). In this Thesis, documentary evidence of fire activity in Tasmania was gathered to analyse relationships between fire and interannual climatic variability (Aim I). Fire occurrence in Tasmania is constantly monitored and recorded by the Tasmanian Fire Service (TFS) and data were obtained from the Land Information System Tasmania (The List, www.thelist.tas.gov.au). Even though the fire record for western Tasmania goes

22 |

back to the early 1970’s, the total number of fires recorded before the 1990s is very low, likely due to the remoteness of this area, which precluded accurate fire detection at that time. Thus, only contiguous years (considered as fire ignition seasons - late spring/early autumn) with a total number of fires >25 across the island were included in the analysis presented in Chapter 3, i.e. the period between fire seasons of 1991/1992 and 2013/2014. Both human caused and natural fires were included in the analyses presented in Chapter 3, except for deliberate fires as categorised by the TFS (i.e. arson and prescribed/management fires).

2.4.2. Charcoal analysis

Sedimentary charcoal is used as a proxy for past fire activity (or palaeofire) in this Thesis (Chapters 3, 4 and 6 – Aim II). Charcoal is produced by the incomplete burning of organic material and analysis of sedimentary charcoal particles is commonly used to determine past fire occurrence (Whitlock and Larsen, 2001). The rate at which charcoal accumulates in a lake depends on the characteristics of the fire and the processes that transport and deliver charcoal to the lake (charcoal taphonomy). Primary charcoal refers to the material introduced during or shortly after a fire event (thus representing a proxy for fire occurrence), whereas secondary charcoal is introduced during non-fire years, as a result of surface runoff and lake- sediment mixing (Whitlock and Larsen, 2001), and thus cannot be used as a direct fire proxy.

Since primary charcoal particles can be carried aloft to great heights and transported great distances (Radtke et al., 1991; Andreae, 1991), the source of charcoal may be from regional (distant) fires, extra-local (nearby but not within the catchment) fires, or local (within the catchment) fires (Whitlock and Larsen, 2001). In general, large particles with a high ratio of volume to surface area tend to move shorter distances from the source fire. Thus, macroscopic charcoal (>125 µm) can be used as a proxy informative of past fire history from a local source area, whereas microscopic charcoal (<100 µm) is usually associated with regional-scale fires (Patterson et al.,

23 |

1987; Clark, 1988; Whitlock and Larsen, 2001; Higuera et al., 2007). Nevertheless, the size of charcoal particles can also be influence by the source biomass type (e.g. Fletcher et al., 2014a). For instance, at Lake Osborne, in southwest Tasmania, a shift from macroscopic-CHAR to microscopic-CHAR in correspondence to Eucalyptus spread was interpreted as a biomass-related change from rainforest to sclerophyll vegetation (Fletcher et al., 2014a).

Topography and vegetation density may influence the transport and deposition of charcoal particles in lakes. Steep slopes may increase the introduction of secondary charcoal through erosion (Swanson 1981, Meyer et al., 1995). On the one hand, the presence of riparian vegetation at the lake margin may trap some of this material and thus limit the input of secondary charcoal and provide a more accurate record of fire history (Whitlock and Millspaugh, 1996; Terasmae and Weeks, 1979). On the other hand, if the abundance of riparian vegetation is not constant through time, the charcoal record could suffer biases due to the differential input of secondary charcoal. Given the accumulation of secondary charcoal, the amount of microscopic- and macroscopic- charcoal preserved in the sediment does not correlate with fire intensity and the severity of a certain fire event, but only its occurrence (Whitlock and Larsen, 2001).

Within the present study, macroscopic charcoal was analysed for each core at a 0.5- cm resolution, except in the slowly accumulating Lake Selina sediments where a resolution of 0.25-cm was adopted. For each sample, a set volume of sediment was sub-sampled (1.25 ml) and immersed in 10% Sodium Hypoclorite for at least one week to digest organic matter (Aaby and Digerfelt, 1986; Whitlock and Larsen, 2001). After digestion, samples were rinsed using nested sieves of 125 µm and 250 µm (Whitlock and Larsen, 2001). The two different sizes were counted separately using Petri dishes under an Olympus SZ51 microscope at 40x magnification. Microscopic charcoal particles (<100 µm) were counted for the Dove Lake core during routine pollen analysis using a tally counter.

24 |

Thanks to the development of a robust age-depth model for all cores, macro- and micro-scopic charcoal counts were converted into CHAR values (charcoal accumulation rates, pieces/cm2/yr), using the following formula (F1). This calculation corrects the charcoal counts based on the sediment accumulation rate, which may vary throughout the core.

F1. Charcoal Accumulation Rates (pieces cm-2yr-1) = Macroscopic charcoal counted (number of particles) / Volume ( )

Depth interval (cm) * Time (yrs) 3 cm

2.4.3. Regional charcoal compilations

Charcoal data from single sites can be combined using the paleofire package (Blarquez et al., 2014) in R (R Core Development Team, 2013) to produce a regional charcoal synthesis (read: regional-scale palaeofire activity). Holocene charcoal series from 13 sites in Tasmania were compiled using this method (see Chapter 3 and 4 for more details). The aim of this approach is to obtain a regional reconstruction of biomass burning through a meta-analysis (Aim II), diminishing local-scale signals related to inter-site variability (e.g. Power et al., 2008). This aim follows the same approach as palaeofire multi-site compilations elsewhere (i.e. the creation of the paleofire package for R) and the Global Paleofire working group (PAGES, http://www.gpwg.paleofire.org). The approach has been widely applied in publications focusing on continental and global scales (e.g. Power et al., 2008; Mooney et al., 2011; Marlon et al., 2013), allowing data from this Thesis to be compared to previous research. A detailed methodology of palaeofire analysis is presented in Chapter 4.

25 |

2.5 Pollen and paleovegetation reconstruction

To quantify the long-term impacts of climatic change and fire activity on terrestrial ecosystems (Aim III), the analysis of fossil pollen coupled with modelling was employed in this Thesis. A necessary step to achieve this endeavour is the calibration of modern pollen-vegetation relationships (Aim III - Step 1), which then can be applied to the fossil record to quantitatively reconstruct land-cover changes through the Holocene (Aim III - Step 2).

2.5.1. Fossil pollen

The analysis of pollen stored in sedimentary archives (palynology) is the most widely used data source for describing and understanding long-term terrestrial ecosystem dynamics (Jackson and Lyford, 1999; Davis, 2000). Tasmania provides an excellent study area for reconstructing vegetation changes based on pollen analysis. High concentrations of well-preserved pollen grains are found in a wide range of sedimentary environments, allowing precise counting and the detection of environmental shifts over a long-term scale (Macphail, 1979).

Pollen, spores, non-pollen palynomorphs, algae and microscopic charcoal in each sample were isolated from a set volume of sediment using a modified version of well-established standard procedures (Faegri and Iversen, 1989). These essentially isolate the desired fraction via the disintegration of the organic and inorganic sediment matrix (Figure 2-3). Subsamples were spiked with a known number of exotic spores (Lycopodium), which allows for the calculation of pollen concentration (grains/cm3).

26 |

Figure 2-3. Pollen processing flow chart and some common pollen grains and spores found in Tasmanian samples.

All the fossil pollen samples analysed within this thesis were 0.5 cm3. Given the lower pollen concentration in modern surface samples, 2-5 cm3 were processed for analysis and stored in vials with glycerol for preservation. After isolation of the palynological fraction, samples were mounted on a microscope slide with a cover slip and counted using a compound light microscope (between 400-1000x magnification). A minimum of 300 pollen grains of terrestrial taxa were counted to achieve statistical reliability, with palynomorphs identified to species, genus or family level using the Macphail Tasmanian Pollen Reference Collection (Macphail, unpublished) and the Australasian Pollen and Spore Atlas (Australian National University, http://apsa.anu.edu.au/).

27 |

2.5.2. Pollen-vegetation calibration and modelling

One of the critical shortcomings of pollen analysis is that it is principally a qualitative or, at best, semi-quantitative, tool for inferring vegetation change through time. While quantitative vegetation reconstruction has been the aim of palynology since the beginning of the discipline 100 years ago (e.g. von Post, 1916; Davis, 1963; Andersen, 1970; Prentice and Parsons, 1983; Sugita, 1993, 1994; Prentice et al., 1996, 1998; Gaillard et al., 1998, 2008; Birks and Berglund, 2017), only recently the increased computer power, reduced analysis time and cost and the compilation of large pollen dataset conspired to allow the development of robust quantitative techniques (Gaillard et al., 2010; Birks and Berglund, 2017). Previous approaches to quantify pollen-vegetation relationships, such as the R-Value by Davis (1963), the correction factors by Andersen (1970), the Extended R-value (ERV; Parsons and Prentice, 1981; Prentice and Parsons, 1983; Sugita, 1994) and the pioneer studies on pollen-dispersal models by Tauber (1965, 1967, 1974, 1977) paved the way to the recent developments in Quaternary pollen analysis (Birks and Berglund, 2017). Many methods have been developed in the recent decades, including biomization (e.g. Prentice et al., 1996) and mechanistic models (e.g. Sugita, 2007a,b), to attempt the reconstruction of landscape change in a quantitative way. The biomization approach converts pollen spectra into biome types by assigning pollen taxa to plant functional types characteristic of each biome (e.g. Prentice et al., 1996, 1998; Prentice and Webb, 1998; Tarasov et al., 2013). Mechanistic models instead take advantage of the modern pollen-vegetation relationships to correct for biases in pollen production and dispersal (e.g. Sugita, 2007a; Mazier et al., 2012; Theuerkauf et al., 2013).

A number of factors influence the representation of vegetation in pollen spectra (e.g. taphonomy, pollen productivity, dispersal capabilities), such that some plant taxa may be over- or under- represented in the pollen rain (i.e. the composition of pollen in the atmosphere at a particular point in time) (e.g. Fletcher and Thomas, 2007a; de Nascimento et al., 2015). Thus, the relationship between pollen data (absolute counts or percentages) and vegetation cover is non-linear (Prentice and Webb, 1986; Faegri and Iversen, 1989; Berglund, 1986). Estimating vegetation from sedimentary pollen

28 |

assemblages requires empirical-based modelling of this relationship across vegetation types. Recent advances in palynology have allowed effective modelling of pollen productivity and dispersal (Sugita, 2007a,b; Gaillard et al., 2010; Theuerkauf et al., 2013), improving understanding of past vegetation dynamics and providing land-cover data to make a quantitative contribution to landscape management and palaeoclimate modelling (Gaillard et al., 2010). Emerging from these efforts is the Landscape Reconstruction Algorithm (LRA), an algorithm proposed by Sugita (2007a,b) to obtain estimates of vegetation abundance on local (<1 km2 up to 5 km2 – which employs the LOVE model) to regional (104-105 km2 – which employs the REVEALS model) scales. Together, these models represent the state-of-the-art in quantitative modelling of vegetation from pollen data.

To date, quantitative vegetation reconstructions have been carried out primarily in Europe (e.g. Hellman et al., 2008; Soepboer et al., 2010; Marquer et al., 2014; Trondman et al., 2015) and North America (Sugita et al., 2010), while the key pollen dispersal parameters that underpin these methodologies have also been measured in China (Li et al., 2015) and South Africa (Duffin and Bunting, 2008). In less than 10 years after the development of the Landscape Reconstruction Algorithm (LRA) model (Sugita, 2007a,b) - a cornerstone achievement in the quest to quantitatively reconstruct vegetation cover from pollen data - this technique has produced valid results in 4 continents (Figure 2-4).

29 |

Figure 2-4. Map showing the current distribution of studies using the quantitative vegetation reconstruction technique adopted in this project. Europe is the key area where theories and applications of this methodology have developed during the last 15 years.

The LRA approach introduced by Sugita (2007a,b) is applicable at local, regional and, potentially, continental scales (e.g. Sugita et al.,. 2008; Abraham et al., 2014; Trondman et al., 2015). Indeed, this modelling framework is still developing and improving, with recent up-scaled application of this method to a sub-continental scale in Europe (Trondman et al., 2015).

A first step for developing a quantitative vegetation reconstruction is to understand how the present vegetation is reflected in the pollen samples (Andersen, 1970; Bradshaw and Webb, 1985; Jackson, 1990; Sugita, 1994, 1998; Gaillard et al., 1998; Broström et al., 2008). A multitude of different techniques have been used to quantify pollen to land-cover relationships, such as indicator species (Behre, 1981; Gaillard et al., 1992), analogue matching (Overpeck et al., 1985), various ordination techniques (Gaillard et al., 1992; Odgaard and Rasmussen, 1998, 2000), linear regression (Andersen, 1970; Bradshaw, 1981; Webb et al., 1981) and other calibration techniques, such as the R-value (Davis, 1963) and Extended R-Value (ERV) methods (Parsons and Prentice, 1981; Prentice and Parsons, 1983; Sugita, 1994) and Partial Least Squares (PLS, Gaillard et al., 1998). Most of these methods involve using a dataset of modern pollen assemblages and environmental variables (mostly

30 |

temperature and rainfall), which produce semi-quantitative vegetation or climate reconstructions where modern analogues for past landscapes exist in the area. However, in the great majority of cases, this approach may perform poorly given that most modern landscapes are in disequilibrium with climate due to the effects of human activity (such as agriculture, industrialisation and frequent burning) - i.e. modern landscapes represent a poor analogue for past landscapes (Nielsen and Vad Odgaard, 2005). LRA differs from the ‘analogue’ approach by allowing quantitative reconstruction of past vegetation cover without directly relying on the contemporary landscape. Rather, the LRA seeks to understand the relationship between land-cover and pollen production and dispersal of individual plant species (e.g. Sugita et al., 2007a,b).

Pollen productivity and dispersal capabilities are two major factors controlling the representation of surrounding vegetation in pollen records and, thus, the interpretation of the pollen spectra in terms of vegetation cover (Prentice, 1985). Pollen productivity is a simple measure of the amount of pollen released for transport per unit area of pollen-producing organs (grains/m2/yr). However, this parameter is difficult to measure, so direct measurements of pollen productivity are highly time-consuming and, thus, rare (e.g. Saito and Takeoka, 1985). PPEs are usually estimated and expressed as a dimensionless ratio relative to a reference taxon, termed the Relative Pollen Productivity (e.g. Davis, 1963; Andersen, 1970, Broström et al., 2008). For the reference taxon the RPP is set to 1. A good reference taxon is present in both pollen and vegetation data from as many of the sites sampled as possible, has a wide range of values of both parameters across the whole dataset and is expected to have an intermediate absolute pollen production value (Bunting et al., 2013). To date, most studies have used the pollen taxon Poaceae as the reference taxon because of its ubiquity (e.g. Xu et al., 2016; Mazier et al., 2008; Soepboer et al., 2007; Bunting et al., 2005; Brostrom et al., 2008). However, since this pollen taxon originates from a wide range of plant species, each with different pollen productivities, it is likely that the same pollen type represents different plant species mixtures in different studies making inter-study comparison less reliable (Bunting et al., 2013). For instance, this taxon comprises species that occur in wetland vegetation

31 |

(e.g. Phragmites) – where surface samples are often collected -, thus the calculation of productivity estimates may be distorted (Brostrom et al., 2008). In this study, Eucalyptus was used as a reference taxon after assessing its abundance in the pollen and vegetation data (see Chapter 5).

To estimate PPEs, the pollen settling velocity (m/s) for each taxon is needed. Pollen sedimentation velocities may either be measured directly or estimated using Stoke’s Law (Chamberlain, 1975), which predicts the settling velocity of smooth spheres of a defined density with diameters between about 1 and 70 µm (i.e. the size of most anemophilous pollen). Experimental attempts have been used to measure settling velocity of pollen grains (e.g. Brush and Brush, 1972; Ferrandino and Aylor, 1984; Di Giovanni et al., 1995) and differences in the experiments’ results can be attributed to convection currents and electrostatic forces which differ between settling towers of different dimensions and construction materials (Jackson and Lyford, 1999). Despite the fact that some pollen grains are not smooth and spherical (Faegri and Iversen, 1989), Stoke’s Law is believed to be an accurate estimator of pollen settling rate (fall speed; Gregory 1973), which, for a spherical particle, is expressed as:

F2. = 2 g (p p)/ 2 0 where 𝒇𝒇𝒇𝒇𝒇𝒇𝒇𝒇 𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔𝒔 𝑟𝑟 − 9µ r = particle radius (cm) g = gravitational acceleration constant (981 cm/sec2)

3 3 P0 = particle density (g/cm )  Set to 1 g/cm according to Jackson and Lyford (1999) p = density of fluid (1.27 x 10-3 g/cm3 for air) µ = viscosity of fluid (1.8 x 10-4 g/cm/sec for air at 18°C)

In this study, for non-spherical pollen grains, the major (a) and the minor (b) axes were measured and the volume of the ellipsoid converted into the radius of an equivalent sphere in order to apply Stoke’s law. Accounting for taphonomic effects in the sampling locations is important, therefore botanical reference material was not used in this study. Moreover, pollen grains in reference material is subject to changes in size depending on the storage duration – pollen grains increase in size with time,

32 |

especially when mounted in glycerol or glycerine jelly. Therefore, measurements taken on reference material are less suitable for calculating PPEs compared to freshly collected pollen grains from surface samples (Moore et al., 1991). In the present study, measurements were done on at least 30 randomly chosen pollen grains for each chosen taxon using modern pollen samples from moss pollsters (see Chapter 5). For the taxa identified at a Genus or Family level (Eucalyptus and Cupressaceae), the mixture of pollen grains found in the pollen samples from different sites was considered to take into account most of the variability in these pollen types.

Another important component of the quantitative vegetation reconstruction approach is the dispersal model. The choice of dispersal model may have a major impact on the resulting PPEs and on the REVEALS plant cover estimates (Theuerkauf et al., 2013). To date, the most commonly used dispersal model is the Gaussian Plume Model (GPM), which underpins the LRA (Prentice, 1985; Sugita, 1993, 1994). The GPM is a relatively simple model based on Sutton's air pollutant plume dispersion equation (Sutton, 1953) and was calibrated using the concentration of particles (e.g. pollen) several hundred metres downwind from a point source as spreading outward from the centreline of the plume following a normal probability distribution. More recent approaches attempt to improve this methodology by using mechanistic models, including Lagrangian Stochastic Models (LSM) (Theuerkauf et al., 2013), which is a more elaborate model involving turbulence and updrafts of air flows (Andersen, 1991; Kuparinen et al., 2007). The LSM implies a large distance of dispersal for seeds, pollen and spores over several kilometres, whereas the GPM implies a locally-limited dispersal (Kuparinen et al., 2007). In fact, empirical and modelling studies have shown that dispersal is not limited to local scales only (Tackenberg, 2003; Schueler and Schlunzen, 2006), but regional sources should also be considered especially when approaching large spatial scale studies because of the complex atmospheric-surface dynamics (Kuparinen et al., 2007; Theuerkauf et al., 2013). Tasmania is characterised by a topographically complex landscape (high surface roughness) with relatively strong winds (Australian Bureau of Meteorology: http://www.bom.gov.au), so pollen dispersal is likely turbulent and influenced by extra-local atmospheric dynamics and the application of the simple GPM may not be

33 | adequate. In this study, in order to obtain realistic land-cover estimates for the Holocene in western Tasmania, the efficiency of the two models was compared (more details are provided in Chapter 5 and Chapter 6).

2.5.3. Quantitative palaeovegetation reconstructions in Australia

The first attempts towards a quantitative vegetation reconstruction approach in Australia did not arrive until the late decades of the 20th century with the pioneer work conducted by Kershaw and Hyland (1975) and Kershaw and Strickland (1990) on the Atherton Tablelands (Queensland). In these first attempts, modern pollen trapping was performed in order to understand the representation of rainforest pollen into fossil spectra, which was found to be significantly lower than in temperate forests (Kershaw and Hyland, 1975), possibly due to the lower, although variable, pollen productivity and the limited dispersal of pollen grains (Kershaw and Strickland, 1990). These findings undeniably highlight the issue of pollen productivity and dispersal biases in Australian pollen records, which is likely complicated by the abundance of animal-pollinated plant taxa in the landscape (Walker, 2000; Dufﬁn and Bunting, 2008).

More recently, the biomization approach (Prentice et al., 1996) was attempted in Australia (Pickett et al., 2004) by taking advantage of the high density of pollen studies in this continent (particularly the east coast). The biomization procedure translates pollen spectra into biome types based on an affinity score calculated using plant functional types (Prentice et al., 1996). Despite promising results on other continents, this approach was found to produce inconsistent results when reconstructing cool temperate rainforests and wet sclerophyll forests in Australia (Pickett et al., 2004). The explanation proposed in Pickett et al. (2004) for this inefficient performance is the ubiquitous presence of Eucalyptus in modern pollen samples from these vegetation types, which is a product of productivity and dispersal capability biases in the pollen spectra.

34 |

In this Thesis, I aim to achieve a quantitative regional reconstruction of past vegetation changes in western Tasmania by correcting for biases linked to the differential productivity and dispersal capability of several Tasmanian key plant taxa, including Eucalyptus (Chapter 5, Aim III).

In Tasmania, a semi-quantitative approach to take into account pollen production and dispersal biases was carried out for the first time by Fletcher and Thomas (2007). This study provided important information on the key pollen taxa, related vegetation types and assumed representation in the pollen rain, which was later used to objectively interpret pollen records (Fletcher and Thomas, 2010a,b). However, the methodology employed in their study does not satisfy the strict requirements of the mechanistic models-based quantitative land-cover reconstructions, which require a calibration of distance weighted plant abundance and modern pollen spectra. An important insight of the work conducted by Fletcher and Thomas (2007a,b) is the finding that moorland vegetation (dominated by Gymnoschoenus sphaerocephalus) is virtually invisible in the pollen records, due to the over-representation of anemophilous rainforest taxa.

35 |

2.6. Study area

2.6.1. Tasmania and its climatic context

Tasmania is a continental island featuring a temperate maritime climate with mild winters and cool summers. The mean annual temperature ranges from 6°C on the mountain tops, and the elevated Central Plateau, to 15°C in coastal regions (Figure 2- 5b). The main factors influencing temperature regimes in Tasmania are (1) the proximity to the ocean, (2) elevation and (3) cloudiness. In coastal areas, maximum temperatures are rarely below 10ºC, whereas inland at low elevations temperatures below 10ºC are common for most of the year (Langford, 1965). Temperature decreases on average by 0.6ºC every 1000 metres of elevation (Nunez, 1988), so highland regions are substantially colder than lowlands. The rising of westerly air stream over the NW-SE trending mountain range that bisects Tasmania (maximum elevation: 1,617 m) produces orographic cloudiness, which plays an important role in reducing day temperatures in the highlands of the western portion of the island. Likewise, the eastern portion of the State experiences less cloud cover and higher temperatures due to descending dry air masses (Föhn effect). Frosts occur with increasing frequency with greater distance from the coastline and inland areas above 300 m altitude experience frosts during the whole year.

An important precipitation gradient exists from western to eastern Tasmania (Figure 2-5a). Westerly air streams uplifted from the Southern Ocean represent the primary source of precipitation of Tasmania. A rain-shadow is produced by the orographic uplift of the prevailing westerly winds (SWW - Southern Westerly Winds) over the NW-SE trending mountain range. This phenomenon results in a super-humid west and sub-humid central and eastern regions. Average total annual rainfall is always above 1000 mm in western Tasmania, exceeding 3000 mm on the West Coast Range, declining sharply east of the ranges (to <400 mm/yr). Reduced evaporation, high cloud cover and rainfall amounts >1250 mm/year characterise the “superhumid” region according to Gentilli (1972) (solid line in Figure 2-5).

36 |

Inter-annual precipitation variability in Tasmania is currently modulated by the El Niño Southern Oscillation (ENSO) and the Southern Annular Mode (SAM - also known as AAO – Antarctic Oscillation) (Hill et al., 2009; Risbey et al., 2009). ENSO is centred on the Pacific Ocean basin and is the result of the interaction between the atmospheric and oceanic pressure systems in the tropical Pacific (Philander, 1983). El Niño refers to the extensive warming of the central and eastern tropical Pacific due to the trade winds weakening, leading to cooler than normal sea surface

37 |

temperatures (SSTs) in the western Pacific, and to the north of Australia. El Niño events are associated with an increased probability of droughts over eastern Australia. In the eastern Pacific, this phenomenon is linked with wetter than normal conditions and widespread flooding (particularly along the coast of South America). During the opposite phase, La Niña, trade winds bring clouds and moisture to eastern Australia and drier conditions prevail in South America.

SAM is an index that describes the north–south movement of the southern westerly wind (SWW) belt that circles Antarctica and is calculated as the zonal pressure difference between the latitudes of 40º and 65 º S (Marshall, 2003). Variations in the SWW drive rainfall and temperature variability across the mid- and high-latitudes of the Southern Hemisphere (e.g. Gillet et al., 2006; Hendon et al., 2007; Garreaud et al., 2009). A positive SAM phase indicates a poleward and a concomitant intensification of the SWW, moving the storm tracks embedded in the SWW away from southern Australia and facilitating the development of high pressure systems over southern Australia, thus resulting in less rainfall. Conversely, the negative phase of SAM indicates an expansion of the SWW belt towards the equator (and a concomitant weakening) which brings low pressure systems and their associated storm tracks over Southern Australia, thus resulting in increased rainfall (Hill et al., 2009; Risbey et al., 2009; Abram et al., 2014; Mariani and Fletcher, 2016). In Tasmania, the SAM index is negatively correlated with annual precipitation (r=-0.5) across the west (Figure 2-6a), whereas ENSO shows the strongest influence on rainfall (r=0.8) in the north and east of the island (Figure 2-6b) (Chapter 3).

38 |

Figure 2-6. Annual rainfall anomaly correlation with a) SAM index (Marshall, 2003) and b) SOI (Southern Oscillation Index) (data from NOAA). Maps were created by calculating correlation coefficients (r) between the annual rainfall anomalies of the period 1961-1990 for 220 stations and the climate mode indices. Rainfall anomalies are the differences between the total precipitation of each year and the average total precipitation of the 30-year baseline period. The r values from the stations have been spatially interpolated using the Kriging method in ArcMap 9.3. Rainfall data from BOM. Coordinates system: GDA 1994 Zone 55 (Grid resolution: 1.8 km).

39 |

2.6.2 Tasmanian vegetation and the legacy of fire

Western Tasmania's climate makes it a suitable location for the development of cool- temperate rainforest (i.e. climatic climax; Jackson, 1968; Brown and Podger, 1982a), however, a large proportion of the western Tasmanian landscape is instead occupied by moorland and sclerophyll vegetation (Figure 2-7). The distribution of these vegetation types is governed by interactions between topography, soil type and fire history, and boundaries between contrasting vegetation types are usually sharp (i.e. little or no ecotone) (Bowman, 2000b; Fletcher and Thomas, 2010a,b; Wood et al., 2011b; Wood and Bowman, 2012).

40 |

Figure 2-7. Vegetation map of western Tasmania, indicated as the “superhumid” area with annual rainfall above 1250 mm, according to Gentilli (1972); data by TasVeg 3.0 (Government of Tasmania, 2013). Walter and Lieth climate diagram for Queenstown (1965- 1994) and Lake St. Clair National Park stations (1990-2014).

41 |

Fire is the main determinant of the type of vegetation boundaries in western Tasmania (Gilbert, 1959; Jackson, 1968; Mount, 1979; Bowman and Jackson, 1981; Brown and Podger, 1982a; Bowman, 2000) and several models have been proposed to explain the existing landscape mosaic (e.g. Gilbert, 1959; Jackson, 1968; Mount, 1979; Bowman and Jackson, 1981; Brown and Podger, 1982a,b; Bowman, 2000; Thomas et al., 2010; Wood and Bowman, 2012; Wood et al., 2011; Fletcher et al., 2014a). Jackson (1968), Mount (1979), Bowman and Jackson (1981) and Brown and Podger (1982b) agreed that each structural formation in Tasmania has a characteristic average interval between fires (Table 2-1). Moorland vegetation has the lowest fire return interval (<20 years), whereas rainforests survive with very low fire frequencies (> 300-400 years).

Jackson (1968) proposed the ‘ecological drift’ or ‘alternative stable states’ model, which maintains that the different vegetation communities in Tasmania ‘drift’ along a continuum dictated by fire return interval. For instance, starting from a moorland community, with the decrease of fire frequency there is an increase in woody-shrub plant taxa. As the fire interval becomes longer, more trees develop until a closed forest is established. This model assumes that stochastic changes in fire regimes would result in vegetation shifts. On the contrary, Mount (1979) proposed the ‘stability’ or ‘stable fire cycles’ model, which maintains that fire occurrence is not random in space in time, but it is driven by the accumulation of fuel loads. For a certain time after a fire event, fuel loadings are too low to carry another fire event, thus, according to this model, each vegetation community has a characteristic fire return interval and has fixed boundaries. The alternative stable states theory proposed by Jackson (1968) recently found support in historical and palaeoecological observations (Thomas et al., 2010; Wood and Bowman, 2012; Fletcher et al., 2014a,b).

42 |

COVER FIRE FIRE FLAMMABILITY AREA % MAIN PLANT TAXA (Reid et al., SENSITIVITY INTERVAL TYPE (Pyrke and Marsden (Jackson, 1968; (WESTERN 1999; Fletcher and Thomas, 2007) (Pyrke and Marsden Smedley, 2005) Mount, 1979; Brown TASMANIA) Smedley, 2005) and Podger, 1982a)

Eucalyptus, Microstrobos niphophilus, Microcachrys tetragona, Nothofagus gunnii, From Moderate ALPINE 5.6 Moderate <100y Nothofagus cunninghamii, Astelia to Extreme alpina, Diselma archeri, Athrotaxis spp., Poaceae, Cyperaceae

Gymnoschoenus sphaerocephalus, Restionaceae, Cyperaceae, Ericaceae, Epacris, Sprengelia, Monotoca, MOORLAND 20.2 Melaleuca, Baumea, Baeckea, Drosera, Low Very High <20y Banksia, Agastachys, Richea, Dracophyllum, Leptospermum

DRY Eucalyptus, Allocasuarina, Banksia, From Moderate to SCLEROPHYLL 4.4 Bursaria, Ericaceae, Asteraceae, Moderate <50y High FOREST Poaceae, Fabaceae, Exocarpos

WET Eucalyptus, Leptospermum, Acacia, SCLEROPHYLL 23.3 Baeura, Pomaderris apetala, Phebalium, High Moderate 30-300y FOREST Olearia

Leptospermum, Melaleuca, Baeura, From Low to WET SCRUB 13.4 Banksia, Pomaderris apetala,Ericaceae, High 15-100y Proteaceae, Asteraceae Moderate

Nothofagus cunninghamii, Phyllocladus RAINFOREST aspleniifolius, Eucryphia lucida, 20.5 (lowland) Lagarostrobos franklinii, Atherosperma moschatum, Anodopetalum biglandosum Extreme Low >300-400 y

MONTANE Athrotaxis cupressoides, A. selaginoides, RAINFOREST 2.6 Nothofagus gunnii, Diselma archeri, (highland) Poa, Sphagnum

Agricultural lands, Native grasslands, Water bodies, Lichen-covered lithosphere, Urban areas, OTHER 9.9 Coastal vegetation, Non-vegetated areas

Table 2-1. List of the main vegetation types in western Tasmania with information about area covered, main species and fire interval characteristics.

43 |

2.6.3. An ancient cultural landscape?

The origin and evolution of the modern dominance of moorland in the western Tasmanian landscape is the subject of a long-standing debate (Jackson, 1968; Jones, 1969; Cosgrove et al., 1990; Thomas, 1993; Cosgrove, 1995; Thomas, 1995b; Thomas, 1995a; Colhoun, 1996a; Fletcher and Thomas, 2007a; Fletcher and Thomas, 2010a; Macphail, 2010; Colhoun and Shimeld, 2012). Models considering Late Pleistocene and Holocene climatic, edaphic and anthropogenic factors have been proposed to explain the landscape dominance by fire-promoted treeless moorland (Figure 2-7) in a landscape where rainforest is the predicted climax vegetation (Jackson, 1986). These models invoke either (1) a replacement of a rainforest dominated landscape by moorland in the late Holocene driven by a combination of climate, edaphics and/or people (Macphail, 1979; Colhoun, 1996) or (2) Late Pleistocene inheritance and subsequent maintenance of treeless vegetation resulting from deliberate human manipulation of fire regimes (Thomas, 1993; Thomas, 1995a,b; Fletcher and Thomas, 2010a).

Advocates of model (1) interpret the region-wide dominance of rainforest pollen types as indicating a climate-driven expansion of rainforest across the region from the Late Pleistocene to the mid-Holocene (Macphail, 1979; Markgraf et al., 1986; Harrison and Dodson, 1993; Colhoun, 1996; Cosgrove, 1995). Colhoun (1996) adapted Iversen’s (1958) Glacial-Interglacial Cycle to describe the post glacial phases in western Tasmania. In this work, Colhoun used pollen sums recalculated to include only 10 key pollen/spore types in Tasmania (Nothofagus cunninghamii, Phyllocladus aspleniifolius, Eucalyptus, Casuarina, Pomaderris, Dicksonia, Dodonaea, Poaceae, Asteraceae, Amaranthaceae), excluding indicators of moorland vegetation (e.g. Cyperaceae – Gymnoschoenus sphaerocephalus) that dominated the local vegetation at the study sites. In this way, values of forest pollen (%) in the studied Holocene records satisfy the thresholds proposed by Macphail (1979) of rainforest presence in southwest Tasmania (i.e. 70% of rainforest pollen abundance) and brought Colhoun to propose an overall forest dominance during this period (Colhoun, 1996; Colhoun and Shimeld, 2012).

44 |

The initial phase of Iversen’s cycle represents full glacial conditions (Cryocratic – glacial peak, 18-20 ka) and vegetation is characterised by grassland, herbland and sedgeland communities. The following phase (Protocratic) is described as an increase in temperature and precipitation that drives the development of alpine and subalpine woody vegetation. The penultimate phase (Mesocratic – interglacial peak, 7 ka) shows the advance of Nothofagus cunninghamii forests, whereas the last phase (Oligocratic, post 7 ka) sees a decline in rainforest in concert with moorland/heath development and with a higher abundance of sclerophyll vegetation. In this scenario proposed by Colhoun (1996), the open landscape in western Tasmania developed only after 7 ka.

Proponents of model (2) highlight the presence of biases in Tasmanian pollen records toward plants that produce large amounts of well-dispersed pollen (i.e. rainforest species), and use semi-quantitative ‘finger-printing’ of pollen sequences with modern pollen data to infer the persistence of open vegetation across western Tasmania for the last ca. 12,000 years (Fletcher and Thomas 2007a,b; 2010a,b). The analysis conducted by Fletcher and Thomas (2010a) shows a departure between the Holocene and previous interglacial pollen spectra and charcoal sequences. Holocene charcoal values are significantly higher than during any previous interglacial period (Fletcher and Thomas, 2010a), when climatic conditions were analogous. Thanks to the semi-quantitative study presented in Fletcher and Thomas (2007b), the authors argue for an over-estimation of rainforest pollen and suggest the persistence of an open landscape throughout the Holocene (Fletcher and Thomas, 2010a,b). The ‘anomalous’ stability of the treeless moorland vegetation during an interglacial period, instead of the climatic climax (i.e. rainforest), was interpreted as a result of the recurrent anthropogenic burning through the latest glacial–interglacial transition. These findings were interpreted as a profound impact of human arrival and fire use on the vegetation mosaic (Fletcher and Thomas, 2010a). This hypothesis finds support in the fact that ignition sources in western Tasmania are mostly anthropogenic, as lightning was found to have a very minor importance in this region (Bowman and Jackson, 1981). Lightning-strike induced fires have been

45 |

observed to account between 0.01% and 0.1% during recent times, with human- triggered fires the overwhelmingly dominant ignition source (Bowman and Jackson, 1981; Jackson and Bowman, 1982; Ingles, 1985; Jackson and Brown, 1986).

2.6.4 Study site locations

A total of 14 sites from this region were involved in the realisation of this project (Figure 2-8). To reconstruct the long-term regional fire activity in western Tasmania (Aim II), 13 high-resolution charcoal records from previously analysed cores were included in two multi-site compilations (further discussed in Chapters 3 and 4). These sites are all located within the zone of strongest influence of SWW in Tasmania (‘SAM zone’; Figure 2-6; see more details in Chapter 3) and one of the strongest in the SH (Gillet et al., 2006), providing an opportunity to test the influence of SWW influence on fire activity in this region.

46 |

To reconstruct regional land-cover changes (Aim III), pollen data from large lakes are needed (Sugita, 2007a). Palynologists have observed that fossil pollen assemblages from different sized basins represent past vegetation at different spatial scales (Jacobson and Bradshaw, 1981). In order to understand the selection choices of this project, it is important to specify the definitions of a “large” lake, according to Sugita (2007a,b). “Large” lakes are defined as “lakes among which pollen assemblages are not statistically different within a given regional landscape”.

47 |

In theory, any pollen record from a large lake can be used to reconstruct the vegetation at a regional scale. Au contraire, “small” lakes are defined as lakes among which pollen assemblages are statistically different within a given landscape. The boundary between large and small lakes is usually set to a lake surface area of 0.5 km2 (Sugita, 2007a; Trondman et al., 2015), but larger lakes (surface area > 1 km2) give estimates of regional vegetation composition with less statistical variability (Sugita, 2007a). According to this definition, for the quantitative reconstruction analysis one of the largest natural lakes in western Tasmania was selected: Dove Lake (surface area = 0.9 km2; green triangle in Figure 2-8). The other sites analysed in the region were much smaller (0.2-0.3 km2) and therefore unsuitable for regional- scale reconstruction using the modelling approach adopted in this Thesis (REVEALS, Sugita, 2007a).

48 |

Chapter 3. Short-term climate and fire variability in western Tasmania1

1 This chapter is an adapted version of the manuscript published in Geophysical Research Letters (Mariani and Fletcher, 2016): The Southern Annular Mode determines inter-annual and centennial-scale fire activity in temperate southwest Tasmania, Australia. The analyses here presented address Thesis Aim I (as stated in Chapter 1 - Introduction).

Abstract

Southern Annular Mode (SAM) is the primary mode of atmospheric variability in the Southern Hemisphere. While it is well established that the current anthropogenic-driven trend in SAM is responsible for decreased rainfall in southern Australia, its role in driving fire regimes in this region has not been explored. We examined the connection between fire activity and SAM in southwest Tasmania, which lies in the latitudinal band of strongest correlation between SAM and rainfall in the Southern Hemisphere. We reveal that fire activity during a fire season is significantly correlated with the phase of SAM in the preceding year using superposed epoch analysis. We then synthesized new 14 charcoal records from southwest Tasmania spanning the last 1000 years, revealing a tight coupling between fire activity and SAM at centennial timescales, observing a multi-century increase in fire activity over the last 500 years and a spike in fire activity in the 21st century in response to natural and anthropogenic SAM trends.

49 |

3.1. Introduction

Fire is a key Earth system process, driving global ecosystem patterns and processes, determining global vegetation distribution (Bond et al., 2005), modulating the carbon cycle (Liu et al., 2015), and influencing the climate system (Bowman et al., 2009). Despite the clear importance of fire, the drivers of fire activity through time are poorly understood in many regions on Earth. A case in point is the range of explanations invoked to account for the increase in fire activity in temperate forest ecosystems across the globe over recent decades (Holz and Veblen, 2011; Meyn et al., 2007; Parisien and Moritz, 2009; Moritz et al., 2012), which include climate change, human ignitions, land use change, and/or altered vegetation structure and patterns (McWethy et al., 2013). Fire activity over the last few centuries in the temperate forests of Patagonia, for example, has recently been linked to hydroclimatic variability associated with the Southern Annular Mode (SAM) (Holz and Veblen, 2011). SAM is the leading mode of Southern Hemisphere climatic variability (Fogt et al., 2009), prompting the question of whether the relationship between SAM and fire in temperate Patagonia holds across the entire Southern Hemisphere or whether it is a more localized southern South American phenomenon. In this paper, we (1) explore the relationship between SAM and fire occurrence in southwest Tasmania, Australia, a temperate region in which rainfall and temperature variability are controlled by SAM and (2) test whether the persistent trend toward a positive SAM state over the last 500 years, particularly over the 21st century (Abram et al., 2014), has influenced fire activity in this temperate region.

SAM describes the north-south movement of the southern westerly wind belt (SWW), a zonally symmetric climate feature that encircles Antarctica and which controls rainfall and temperature variability across the extratropics of the entire Southern Hemisphere (Garreaud, 2007; Gillett et al., 2006; Hill et al., 2009). In the positive phase of SAM, the SWW contract poleward facilitating the development of high-pressure systems over southern Australia and Tasmania, resulting in a decrease in rainfall. Conversely, the negative phase of SAM sees an expansion of SWW toward the equator, bringing low pressure systems and their associated storm tracks

50 | over Southern Australia and Tasmania, resulting in increased rainfall (Fogt et al., 2009; Garreaud et al., 2009; Hill et al., 2009; Risbey et al., 2009; Abram et al., 2014) (Figure 3-1). Interannual positive anomalies of SAM are associated with higher temperatures and lower precipitation across the Southern Hemisphere (Gillett et al., 2006; Hendon et al., 2007; Hill et al., 2009). Importantly, the last ~60 years is characterized by a trend toward extreme positive SAM in response to ozone depletion (Thompson and Solomon, 2002; Marshall, 2003; Perlwitz et al., 2008) that is associated with warmer and drier conditions across the southern extratropics (Smith and Reynolds, 2005; Fogt et al., 2009). Moreover, this trend is embedded within a longer centennial-scale trend toward positive SAM occurring over the last 500 years (Abram et al., 2014) and it is unknown what, if any, impact this has had over Southern Hemisphere ﬁre activity.

51 |

Fire occurrence and spread is determined by the confluence of sufficient fuel, an ignition source, and suitable weather: the fire triangle (Krawchuk et al., 2009). In areas of high biomass (read: abundant fuel), such as southwest Tasmania, fire occurrence through time is modulated by fuel moisture (i.e., climate) and ignitions (lightning and humans) (Cochrane, 2003; Pausas and Ribeiro, 2013; Bradstock, 2010; McWethy et al., 2013). Humans have actively used fire to modify the Tasmanian environment for more than 40,000 years (Cosgrove, 1999; Fletcher and Thomas, 2010; Jones, 1969), and along with lightning strike (which account for less than 0.1% of ignitions (Bowman and Brown, 1986), the constant source of ignition in this landscape effectively isolates climate variability as the principal factor modulating the occurrence of fire through time. Fires in Tasmania are driven by seasonal, interannual, and decadal variations in temperature and rainfall: i.e., fires occur in response to hot and dry conditions (Nicholls and Lucas, 2007). Rainfall in southwest Tasmania is derived entirely from the SWW and interannual variations in rainfall are controlled by SAM (Figure 3-1). We posit, then, that if fire activity in this landscape is modulated by climate, interannual fire activity should be correlated with SAM. Further, if this relationship exists, we hypothesize that the persistent 21st century trend toward extreme positive SAM phase will have increased the risk of fire in this landscape, placing highly fire sensitive endemic ecosystems in this region at risk of extinction.

Southwest Tasmania is a topographically complex landscape that hosts a number of extremely fire sensitive endemic vegetation systems that have suffered substantial fire-driven range contraction throughout the Holocene (Fletcher et al., 2013, 2014) and since European colonization (Cullen, 1987; Holz et al., 2014). Indeed, the distribution of rainforest in this region is, like much of the highly flammable Australian continent, restricted to fire refugia that are determined principally by topography and nonlinear feedbacks between vegetation type and flammability (Jackson, 1968; Bowman, 2000; Wood et al., 2011). Not only does the current SAM trend pose a potentially significant threat to the security of the remaining pockets of fire sensitive ecosystems via a shortening of the fire return interval, the potential reduction in rainfall associated with this trend in southern Australia and Tasmania

52 |

(Fyfe and Saenko, 2006; Miller et al., 2006) creates increasingly inhospitable climatic conditions for plant growth and recovery. This threefold impact of current climate trends, termed “interval squeeze” (Enright et al., 2015), threatens ﬁre sensitive ecosystems with extinction. Thus, it is critical that we attempt to understand the role that climate has in driving long-term ﬁre activity, so that realistic management options for our natural systems can be explored.

In this paper, we explore the relationship between climate and fire occurrence in southwest Tasmania, testing whether the reported relationship between SAM and fire activity in Patagonia is also manifest in Tasmania. We then draw on a database of new palaeofire records from this region spanning the last 1000 years to test for a link between SAM and long term (centennial scale) fire activity in southwest Tasmania. We specifically ask: (1) does SAM driven climate variability control contemporary fire activity in southwest Tasmania? (2) Does centennial-scale SAM variability control longer-term fire activity in southwest Tasmania? (3) Is there an upward spike in fire activity related to the current positive SAM trend driven by ozone depletion?

3.2. Methods

To identify the principal driver of rainfall in our study region we created a correlation map between annual rainfall anomalies and all of the main climate indices identified as important drivers of rainfall anomalies in southern Australia (SAM, the El Niño–Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), and the Pacific Decadal Oscillation (PDO)). We calculated correlation coefficients (r) between annual rainfall anomalies during the period 1961–1990 for 220 meteorological stations (data from Australian Bureau of Meteorology) and the annual climate indices for the Marshall (2003) SAM index (British Antarctic Survey), ENSO (SOI Index from NOAA), IOD (DMI Index, http://www.jamstec.go.jp/), and PDO (Index from NOAA) (Figures 3-1 and 3-S1). Climate modes operate at scales ranging from seasonal to centennial and we selected the average annual values of

53 |

the climate indices for this analysis. Rainfall anomalies are the differences between the total precipitation of each year and the average total precipitation of the 30 year baseline period (1961–1990). The r values from the stations have been spatially interpolated using the Universal Kriging method in ArcMap 9.3 (ESRI— Environmental Systems Resource Institute, 2009, Redlands, California). Coordinates system isGDA1994 Zone 55 and the grid resolution is 1.8 × 1.8 km. The results of this analysis clearly reveal SAM as the key driver of rainfall variability in SW Tasmania over the analysis period (Figure 3-S1), with all other indices displaying little or no explanatory power for rainfall anomaly in this area. Thus, we focus on SAM for the remainder of this paper. We restrict our analysis of fire occurrence to what we deem as the “SAM zone,” identified as the area with an r correlation coefficient > 0.3.

Fire occurrence data for the SAM zone were obtained from the Land Information System Tasmania (the List, Government of Tasmania). Since the total number of fires before the 1990s is very low, likely due to the remote-ness of this area precluding accurate fire detection at that time, only contiguous years (considered as fire ignition seasons—late spring/early autumn) with a total number of fires >25 across the island have been chosen, i.e., the period between fire seasons of 1991/1992 and 2013/2014. While this represents a relatively short period for correlation, we feel that this data set represents the best current data set for testing the important questions tackled by this paper, which are crucial for fire activity forecasting and management. Figure 3-1 presents the location of all fires used in our analysis plotted with the spatial correlation between fire season SAM and rainfall anomalies. We include both human caused and natural fires in the analyses, with the exception of deliberate management fires (i.e., prescribed/management fires).

To identify a relationship between the annual SAM index and fire occurrence in the SAM zone, we performed superposed epoch analysis (SEA) analysis in R v.3.0.3. This analysis allows assessing the significance of the departure from the mean for a given set of key event years (e.g., fire years) and lagged years (Lough and Fritts,1987).The fire occurrence data for “fire seasons” (number of fires and area burnt) and the SAM index were converted to z scores (using the entire series mean)

54 | prior to analysis and significant deviations from the mean were used to identify “fire years” and “non-fire years”. Fire seasons span the period between December and March and include ~80% of fires occurring in any 12 month period. The unique landscape-scale vegetation mosaic in SW Tasmania, which juxtaposes pyrophobic (fire retarding) and pyrogenic (fire promoting) vegetation types, exerts a major influence over the spread and extent of fires; thus, we hypothesized that changes in the number of fires will more accurately reflect changes in the broad-scale drivers of fire activity in this landscape than the more traditionally employed area burnt metric.

For our last 1000 year palaeofire analysis, we synthesized new sedimentary charcoal records analyzed by our research team and located within the SAM zone identified in our climate analysis (Figures 3-1 and S2). Chronology of the charcoal records is based on radiocarbon and Lead-210 assays (Table 3-S1), with age-depth modeling performed using Clam v2.1 (Blaauw, 2010). A charcoal composite curve for all 14 sites was performed using the Paleofire package in R (Blarquez et al., 2014). A 50 year interval for this analysis was chosen, since it represents the best achievable resolution in order to include the majority of records for the entire reconstruction period. The full list of the sites used in the palaeofire analysis is shown in Table 3-S1, along with the charcoal records for the last 1000 years (Figure 3-S2).

55 |

3.3 Results

The spatial climate correlation analysis shows a distinct pattern of correlation between SAM and rainfall anomalies across the island of Tasmania: a strong SAM- rainfall correlation in the southwest and no correlation in the northeast and east (Figure 3-1). A total of 368 fires (accidental human-ignited and naturally ignited) were identified in the SAM zone during the period 1992–2014 (Figure 3-1). The SEA reveals a statistically significant (p value <0.05) positive annual SAM departure occurring in the year preceding a fire season (Figure 3-3a). To support this result, we show that non-fire years (fire seasons with an anomalously low fire occurrence) correspond to a significant (p value <0.05) negative departure in SAM (Figure 3-3b). Area burnt (both fire years and non-fire years) did not show any relationship with the annual SAM Index (Figures 3-3c and 3-3d).

56 |

The palaeofire composite analysis of our new data set of 14 southwest Tasmanian charcoal records spanning the last 1000 years shows initially high fire activity around 1000 Common Era (C.E.), a sharp decline to minimum values at 1400 C.E. and a persistent increase toward the present, interrupted by a plateau between 1600 and 1800 C.E. and finally by a precipitous increase from 1800 C.E. to the present (Figure 3-4).

57 |

3.4. Discussion

Our analysis reveals, for the first time, that the phase of SAM preceding a fire season in SW Tasmania determines inter-annual fire activity in this landscape (Figures 3-2 and 3-3). Further, the results confirm our hypothesis that trends in the number of fires in the landscape of SW Tasmania are more reflective of changes in the climatic drivers governing fire activity than the area burnt. This finding is entirely consistent with the dominant influence that the fine-scale mosaic of juxtaposed pyrophobic and pyrogenic vegetation types has over the spread and extent of fires in this region (Jackson, 1968; Wood et al., 2011, 2012). The stark contrast in fuel moisture content, flammability, and fire sensitivity of vegetation types in this region (Pyrke and Marsden-Smedley, 2005) dictates that the relationship between the area burnt and climate is unlikely to be linear. Rather, our results confirm that where fires ignite in relation to vegetation boundaries, topographic divides and the prevalent westerly airflow are key determinants of fire spread and

58 |

extent, thereby, reducing the efﬁcacy of the area burnt metric for our present analysis.

Our results indicate that an increase (decrease) in fire activity during a fire season (December–March) is preceded by an anomalously dry (wet) year associated with a positive (negative) SAM phase. The 1 year lag we have identified between SAM years and fire seasons reflects the high moisture content of fuels in this perennially wet landscape and the time required to precondition fuels to burn. The same lag between SAM and fire occurrence was not identified in the drier temperate forests in Patagonia studied by Holz and Veblen (2011), who based their analysis on fires inferred from fire-scarred trees in forests located close to the Patagonian forest- steppe ecotone. The forest-steppe ecotone environment in Patagonia is considerably drier than southwest Tasmania (Garreaud et al., 2009; Sturman and Tapper, 2006), and while hosting a high biomass load that does not limit fire (Holz and Veblen,2011; Veblen et al., 1999), less time would be required to condition the fuel in that landscape to burn when compared with southwest Tasmania. Thus, our analysis identifies SAM as the main driver of interannual fire activity across a broad swath of the Southern Hemisphere. Our results are consistent with the pervasive influence of the North Atlantic Oscillation (NAO), the northern counterpart of SAM, over fire regimes in forest ecosystems in North America, where NAO driven shifts in the Northern Hemisphere westerlies modulate temporal fire activity via their influence on hydroclimate (Le Goff et al., 2007). Indeed, evidence is mounting that a number of climate modes play a pivotal role in modulating long-term fire activity in high biomass ecosystems globally (Le Goff et al., 2007; Holz and Veblen, 2011; Román- Cuesta et al., 2014; Fletcher et al., 2015) and these relationships must be considered when attempting to predict future climate fire trends (Moritz et al., 2012).

We identify a tight coupling between landscape-wide fire activity in southwest Tasmania and a recent SAM reconstruction for the last millennium (Figure 3-4). This coupling is entirely consistent with our findings of significant correlation between SAM and fire activity in southwest Tasmania, revealing a persistence of this relationship over longer timescales. Initially high charcoal values are consistent with

59 |

relatively dry conditions through the latter part of the Medieval Climate Anomaly (circa 1050–600 cal yr B.P.). A salient feature of our analysis is the persistent increase in fire activity since 1500 C.E., throughout the Little Ice Age (circa 600–100 cal yr B.P.). Comparison with the two leading proxy-based SAM reconstructions (Abram et al., 2014; Villalba et al., 2012) reveals a very tight synchronicity between hemispheric-scale reconstructions of SAM and southwest Tasmanian fire activity through the last 500 years. This period represents a phase in which SAM becomes progressively more positive, exceeding the range of SAM variability experienced over the last millennium (Abram et al., 2014; Villalba et al., 2012) and it is clear that this trend drove an increase in landscape burning in southwest Tasmania. The observed dramatic increase in fire in this region after 1800 C.E. is consistent with the timing of European colonization and a series of landscape-scale wildfires in the mid to late 1800s (Marsden-Smedley, 1998). Critically, the relationship between SAM and southwest Tasmanian fire activity persists through the 21st century, when anthropogenic activity induced a further positive shift in SAM (Perlwitz et al., 2008), despite a move toward greater fire regulation in this landscape. Our results reveal a high sensitivity of the Tasmanian environment to SAM driven shifts in the SWW and heralds a significant threat for fire sensitive ecosystems in this region.

Fire activity is predicted to increase in temperate forest biomes under projections of future climate scenarios (Moritz et al., 2012). Our revelation of a clear link between interannual and centennial-scale SAM dynamics and fire activity in southwest Tasmania (and across the Southern Hemisphere) introduces an additional variable that must be considered when projecting and planning for the future of these important ecosystems. While future trajectory and mean state of SAM is uncertain as ozone levels recover (Polvani et al., 2011; Perlwitz, 2011), it is imperative that we attempt to grasp Earth system teleconnections, such as climate-fire interactions. The implication that SAM drives hemisphere-wide fire activity adds to the vast array of natural systems that are influenced by this important component of the global climate system, such as stream discharge (Lara et al., 2008), rodent population fluctuations (Murúa et al., 2003), insect outbreaks (Paritsis and Veblen, 2011), and coastal and marine ecosystem dynamics (Forcada and Trathan, 2009; Schloss et al.,

60 |

2012; Alvain et al., 2013; Weimerskirch et al., 2012). Thus, the pervasive inﬂuence of SAM over the Earth system means that many SAM inﬂuenced or dependent systems may face deleterious effects resulting from the current anthropogenically driven SAM trend, underscoring the need for studies such as ours which attempt to elucidate climate-biosphere interactions.

3.5. Conclusion

This research constitutes the first attempt in disentangling the role of SAM in driving fire activity in Tasmania. We reveal that SAM is significantly linked with interannual fire occurrence (number of fires) in southwest Tasmania. Palaeofire analysis reveals a tight coupling between southwest Tasmanian fire activity and two proxy- based SAM reconstructions, revealing that SAM drives fire activity at multiple scales of time in this landscape. We observe a multi-century increase in fire activity in southwest Tasmania in tandem with a positive trend in SAM over the last 500 years, and importantly, we note a 21st century spike in fire activity in response to the anthropogenic influence on SAM brought by ozone depletion.

61 |

3.6. Supplementary results and information

The figures and tables displayed below are part of the supporting information for the paper ‘The Southern Annular Mode determines inter-annual and centennial-scale fire activity in temperate southwest Tasmania, Australia’ (Mariani and Fletcher, 2016).

62 |

Figure 3-S1. Map of the correlation between annual rainfall anomaly and a) annual SAM Index; b) annual SOI Index; c) annual PDO Index; d) annual IOD Index.

63 |

3.6.1. Additional information and results for the SEA

Table 3-S1. List of years used for the Superposed Epoch Analysis (SEA).

64 |

65 |

Figure 3-S3. Departure of seasonal SAM Index in relationship with number of fires; a) fire years; b) non-fire years.

66 |

3.6.2. Palaeofire compilation and charcoal results

Palaeofire analysis was carried out using the paleofire package in R (Blarquez et al., 2014). Fourteen charcoal records from southwest Tasmania have been selected for this analysis (Fletcher et al unpublished data) (see Table S1). Firstly, a transformation of the data was performed using the function pfTransform with MinMax, Box-Cox and Z-score methods. Transformation and standardization of different charcoal records is a highly recommended step in generating a synthesis (Blarquez et al., 2014). Here, we used the methodology proposed by Power et al. (2008) and involved a three-step data transformation including a min-max data-rescaling, variance homogenization using Box-Cox data transformation (Box and Cox, 1964), and final rescaling to Z-scores. The palaeofire composite was calculated using the function pfCompositeLF, consisting in a modified version of the methods proposed by Marlon et al. (2008) and Daniau et al. (2012), involving a two-stage smoothing method (including LOWESS; Cleveland, 1979) of the selected bins interval. In this case, 50 years-bins were used, since it represents the best achievable resolution in order to include the majority of charcoal records for the entire reconstruction period. Confidence intervals were obtained using the function circboot with 1000 repetitions, which applies a "moving" or "circular" block bootstrap method as proposed by Kunsch (1989) to test significance of changes in stationary time series.

Table 3-S2 List of charcoal records used in the palaeofire analysis.

67 |

Figure 3-S4. Time-series of the CHAR records used in the palaeofire analysis.

68 |

Figure 3-S5. Palaeofire composite of southwest Tasmania with confidence intervals.

69 |

Chapter 4. Long-term climate variability and palaeofire activity in western Tasmania2

2 This chapter is an adapted version of the manuscript published in Quaternary Science Reviews (Mariani and Fletcher, 2017): Long-term climate dynamics in the extra-tropics of the South Pacific revealed from sedimentary charcoal analysis. The analyses here presented address Thesis Aim II (as stated in Chapter 1 - Introduction).

Abstract

We synthesized 13 high-resolution charcoal records located within the current zone of strongest correlation between the southern westerly winds (SWW) and rainfall on Earth in an attempt to assess how shifts in the SWW drive climatic change in this region. High regional charcoal influx values are found during the early Holocene (12–8 ka), progressively decreasing and reaching a minimum during the mid-Holocene ( 5 ka). Wavelet coherence

analysis between regional charcoal influxes from southern South Amer∼ ica (SSA) and western Tasmania (WTAS) shows a tight periodicity coherence from 12 to 6ka, supporting

synchronous SWW-driven climatic change in these areas. The same analysis∼ between the regional Tasmania charcoal influx and an ENSO proxy suggests a coherent pattern of frequency variability between these records since 6 ka, highlighting the importance of

ENSO in altering fire regimes in this region. Our data∼ also provide insights into the non- stationarity of the climate system in space and time and highlights the potential limitations of modern climatic relationships for informing our understanding of the global climate system.

70 |

4.1 Introduction

Projections of an increase in fire activity across many of the Earth’s biomes in response to climate change (Moritz et al., 2012) have led to the necessity for an increased understanding the role of climate in driving fire activity (e.g. Girardin et al., 2013; Marlon et al., 2013; McWethy et al., 2013). Fire regimes are modulated by human activity, climate and biological processes that operate from local to global scale and across daily to multi-millennial timescales (Fletcher et al., 2015; Power et al., 2008; Whitlock et al., 2007). It is, thus, essential to develop a multi-scale perspective of how fire regimes change through time if we are to adapt to, manage and predict future fire activity (Marlon et al., 2013). The development and synthesis of high-resolution sedimentary charcoal records from within and across regions has facilitated a significant leap in our understanding of how fire activity changes through deep time. Such approaches have typically synthesised records within biomes and/or within global circulation fields (e.g. Inter-Tropical Convergence Zone – ITCZ; Sub-Tropical High Pressure belt – STH, circumpolar westerlies), revealing climate as the primary driver of long-term (supra-centennial scale) fire activity across much of the Earth (Marlon et al., 2013; Mooney et al., 2011; Power et al., 2008).

Critically, the biome and circulation field approach can be potentially insensitive to the role of important regional climate modes, such as the North Atlantic Oscillation (NAO), Southern Annular Mode (SAM), Indian Ocean Dipole (IOD) or the El Niño- Southern Oscillation (ENSO) in driving long-term climate-fire dynamics, as the influence of these modes often transgresses boundaries between biomes and circulation fields. A case-in-point is the spatially complex interaction between climate and inter-annual fire activity in extra-tropical southeast Australia (30-45°S) (Williamson et al., 2016). While this region is included within the STH (e.g. Mooney et al., 2011), the drivers of modern inter-annual fire activity in this region vary between ENSO in the east and north (Mariani et al., 2016) and SAM, the leading mode of variability in the circumpolar southern westerly wind belt (SWW), in the southwest (Mariani and Fletcher, 2016). Further, consideration of the effects of the existence of non-static relationships between local and remote climates (‘non-

71 |

stationarity’, sensu Gallant et al., 2013) and of changes in the dynamical relationship between fuel, climate and fire across landscapes and through time (i.e. the spectrum from fuel to climate-limited fire regimes) should precede any attempt to synthesise regional sedimentary charcoal data. Syntheses of sedimentary charcoal records that are guided by a climatic framework, rather than biomes or global circulation fields, then, can potentially yield more detailed information about how and why fire activity changes in a region through time. Here, we present a synthesis and analysis of 13 new sedimentary charcoal records from western Tasmania (WTAS), Australia, spanning the last 12,000 years (12 kyrs) in the southwest sector of the Pacific Ocean basin. We targeted sites located in temperate ecosystems and within in the zone of strongest correlation between the SWW and modern rainfall anomalies on Earth (Gillett et al., 2006) in an attempt to understand how and why fire activity has varied within this climate zone through the last 12 kyrs.

The extra-tropics of the South Pacific (30-60°S) is a cool and wet sector dominated by temperate ecosystems in which fires are limited by climate (i.e. moisture) (McWethy et al., 2013). Exploiting this relationship between fire and hydroclimate in temperate forest ecosystems using sedimentary charcoal records can provide significant insights into the relationship between fire activity and climate through time and, indeed, of inferring past climate dynamics (Fletcher et al., 2015; Fletcher and Moreno, 2012; Power et al., 2008). Our current understanding of terrestrial climatic change over the last 12 kyrs in the extra-tropics of the South Pacific region is largely derived from sedimentary charcoal records and pollen analysis, and can be summarised as being driven by a long-term interaction between ENSO and the SWW (Fletcher and Moreno, 2012; Rees et al., 2015): multi-millennial scale shifts in the strength and latitudinal position of the SWW dominate between ca. 14,000-6000 years before present (14-6 ka) - enhanced SWW flow between ca. 14-12 ka and 7-5 ka (low biomass burning), attenuated SWW flow between ca. 11-8 ka (high biomass burning). After ca. 6 ka, the onset of ‘modern’ ENSO variability (Moy et al., 2002) drove an increase in the frequency of hydroclimatic shifts and associated trends in biomass burning (Fletcher and Moreno, 2012; Fletcher et al., 2015; Rees et al., 2015).

72 |

However, key data supporting the SWW-ENSO model of long-term climate evolution from the Australian sector of extra-tropical South Pacific come from either poorly dated and coarse resolution sedimentary charcoal data of an obscure origin (principally pollen slide charcoal with often no unit of measurement) (Mooney et al., 2011) or, more recently, from three high resolution sedimentary charcoal and pollen records (Fletcher et al., 2015; Rees et al., 2015). Thus, there is a need for a targeted synthesis of high resolution and well dated records from this region if we are to advance our understanding of long term changes in climate and fire activity.

Western Tasmania is ideally situated to understand the dynamic interplay between the SWW and ENSO through time using sedimentary charcoal records because (1) Tasmania lies at the interface of the SWW and ENSO in the extra tropics (Figure 4- 1c,d) (Hill et al., 2009); (2) it is a temperate island in which fires are climate limited (Mariani et al., 2016; McWethy et al., 2013; Nicholls and Lucas, 2007) and (3) it hosts numerous lakes well suited to sedimentary charcoal analysis. Here, we employ a synthesis of 13 high-resolution sedimentary charcoal records from western Tasmania to assess the climatic paradigm of long-term climatic change in the southern extra- tropical latitudes advanced by Fletcher and Moreno (2012) that predicts a shift from multi-millennial SWW driven climate and fire trends to a higher frequency ENSO driven climate and fire regime after ca. 6 ka. In restricting our analysis to western Tasmania, we hope to constrain the potential climatic (ENSO and SWW) and non- climatic (i.e. fuel type and amount) influences over palaeofire activity during the last 12 kyrs, thus, permitting an assessment of the climate drivers modulating fire regimes throughout the reconstruction period.

73 |

4.1.1. Spatiotemporal climate dynamics in the southern extra-tropics

The SWW are a zonally symmetric climate feature that encircles Antarctica that control rainfall and temperature variability across the extra-tropics of the Southern Hemisphere (Garreaud, 2007; Gillett et al., 2006; Hill et al., 2009). The position and strength of the SWW are variable through time, with the modern ‘core’ of the SWW situated at ca. 50°S. When the SWW contract towards Antarctica (such as during the positive phase of SAM), the development of high-pressure systems over the southern extra-tropics (i.e. Tasmania, southeast Australia, northern Patagonia) is promoted, resulting in a decrease in rainfall north of the modern SWW core. Conversely, an expansion of the SWW belt towards the equator (such as during the

74 |

negative phase of SAM) brings low pressure systems and their associated storm tracks over this region, resulting in increased rainfall north of the modern SWW core (Abram et al., 2014; Fogt et al., 2009; Garreaud et al., 2009; Hill et al., 2009; Risbey et al., 2009). Further, the broadly north-south trending mountain ranges that bisect all southern extra-tropical landmasses intercept the SWW and impart a local orographic effect on rainfall patterns. This orographic effect results in the strongest rainfall anomalies associated with SWW on the west (windward) facing slopes of these regions (Garreaud, 2007). The modern variability of the strength and position of the SWW is described by SAM (Marshall, 2003). Critically, a strong link between contemporary fire activity and SAM has been detected in temperate regions of both northern Patagonia and western Tasmania (Holz and Veblen, 2012; Mariani and Fletcher, 2016) and this relationship is often exploited to interpret palaeofire records developed from sedimentary charcoal in terms of intensity or latitudinal variations in the SWW (Fletcher et al., 2015; Fletcher and Moreno, 2011, 2012; Lamy et al., 2010; Mariani and Fletcher, 2016; Moreno, 2004; Moreno et al., 2010; Moreno et al., 2009; Moreno et al., 2014).

ENSO is an irregularly periodic oceanic-atmospheric fluctuation which mostly affects the tropics and subtropics of the Southern Hemisphere (Philander, 1983). The climatic effects of ENSO are asymmetric (anti-phased) across the Southern Hemisphere (Philander, 1983; Garreaud et al., 2009; Risbey et al., 2009). El Niño is the warm phase of ENSO and is linked with negative moisture anomalies in the western Pacific (Australia, including Tasmania) and overall positive moisture anomalies in the eastern Pacific (South America), while La Niña (the cool phase of ENSO) results in the opposite hydroclimatic signature (Philander, 1983). Although ENSO does not show a significant correlation with modern annual rainfall variability in western Tasmania (Figure 4-1d), negative rainfall anomalies have been reported during El Niño years (Hill et al., 2009). This influence is manifest through the development of an anomalously high pressure system across the East Pacific during El Niño events between the tropics and as far south as southern Tasmania, resulting in negative moisture anomalies across this entire region (Hill et al., 2009).

75 |

Importantly, the spatio-temporal expression of ENSO events in southern South America (SSA) is complex, with climate-ENSO analyses revealing quite marked seasonal differences in the expression of ENSO and substantial non-stationarity in the influence of ENSO on rainfall over the historic period (Garreaud et al., 2009). The intensity and frequency of ENSO oscillations has not been stable through time (Conroy et al., 2008; Donders et al., 2005; Koutavas et al., 2002; McClymont and Rosell-Melé, 2005; McGlone et al., 1992; Moy et al., 2002; Riedinger et al., 2002; Rodbell et al., 1999), with changes in the ENSO system invoked to account for wide- ranging environmental change across the Pacific Rim and further afield (e.g. Donders et al., 2007; Fletcher et al., 2015; Fletcher and Moreno, 2012; Graham et al., 2010; Herweijer et al., 2007; Whitlock et al., 2007), including substantial changes in climate and fire activity in Australia (e.g. Black et al., 2008; Black et al., 2007; Dodson, 2001; Donders et al., 2008; Mariani et al., 2016; Nicholls and Lucas, 2007; Verdon et al., 2004). Proxy evidence for long-term changes in ENSO activity indicates a spike in the frequency/intensity of ENSO events after ca. 6-5 ka (Conroy et al., 2008; Donders et al., 2008; McGlone et al., 1992; Moy et al., 2002). Moy et al. (2002) report the initial onset of ‘modern’ El Niño-driven rainfall in the eastern Pacific during the Holocene at ca. 6.7 ka, with a major and persistent spike in El Niño frequency occurring between ca. 5.2-4.7 ka. Further, an integration of various proxy sources across the southwest Pacific indicate enhanced moisture (i.e. La Niña-like conditions) between 4.7 and 3.5 ka (Donders et al., 2008), while a range of proxy datasets indicate a pronounced increase in the frequency and/or intensity of El Niño events after ca. 3.5 ka (Conroy et al., 2008; Donders et al., 2008; Marchant and Hooghiemstra, 2004; McGlone et al., 1992; Moy et al., 2002; Sandweiss et al., 2001).

Importantly, proxy records of variability in climate features, such as the SWW and ENSO, indicate that the instrumental record does not capture the full range of variability experienced through time (Abram et al., 2014; Conroy et al., 2008; Moy et al., 2002; Yan et al., 2011) and an over-dependence on modern climatology to interpret proxies of past climatic change should be avoided. Rather, contemporary climate correlations provide a snapshot of the spatial manifestation of climatic

76 | variability within a landscape and such climatic ‘frameworks’ can provide a useful mechanism for selecting proxy sites to test specific questions. Likewise, non- stationarity in the teleconnections between components of the climate system (sensu Gallant et al., 2013) limit our ability to fully attribute observed past environmental changes to contemporary climate relationships. A case in point is the invocation of ENSO to account for climate-driven fire regime changes in parts of Tasmania in which ENSO has little or no statistically significant explanatory power over rainfall variability over the instrumental period (Fletcher et al., 2015; Rees et al., 2015), but in which ENSO is linked with negative rainfall anomalies during its warm phases (El Niño years) (Hill et al., 2009).

4.1.2. Western Tasmania

Tasmania (41-44°S) is a cool temperate island bisected by a NW-SE mountain range that intercepts the dominant westerly airflow across this latitude of the Southern Hemisphere, resulting in a steep W-E orographic precipitation gradient. In western Tasmania, temperatures are cool (5-7°C in winter and 14-16°C in summer) and precipitation (between 1200 and 3500 mm/year) exceeds evaporation for most of the year (Gentilli, 1972; Sturman and Tapper, 2006). Inter-annual climatic variability in Tasmania is driven by both SAM and ENSO (Figure 4-1c,d), with SAM being the only statistically significant driver of rainfall and fire variability in western Tasmania during the historical period (Mariani and Fletcher, 2016). Fires in this region are climate-limited, with an increase (decrease) in the occurrence of fire associated with drier (wetter) conditions possibly resulting from south (north) displacements of the SWW (Mariani and Fletcher, 2016). Ignition sources in the hyper-humid and perennially wet western Tasmania landscape include lightning and humans. Lightning-strike induced fires have been observed to account between 0.01% and 0.1% over the historical period (Bowman and Brown, 1986; Ingles, 1985; Jackson and Bowman, 1982), although recent (2016) catastrophic landscape-scale wildfires in this region were the result of a series of intense lightning strikes in summer that were unaccompanied by rain. Nevertheless, over the historical period, human-triggered

77 |

fires are the overwhelmingly dominant ignition source in this period and the long- term occupation of this landscape (>40 kyrs) (Cosgrove, 1999) indicates a constant ignition source throughout this time.

The vegetation landscape of western Tasmania is a cultural landscape resulting from the prolonged application of fire by Australian Aborigines (Fletcher and Thomas, 2010; Mariani et al., in press) and is characterised by a landscape-scale vegetation mosaic of pyrophobic temperate rainforest and pyrophitic sclerophyllous forest, scrub and moorland vegetation. The distribution of these vegetation types is governed by interactions between topography, soil type and fire history, and boundaries between contrasting vegetation types are usually sharp (i.e. little or no ecotone) (Bowman, 2000; Fletcher and Thomas, 2010; Wood and Bowman, 2012; Wood et al., 2011). Despite clear evidence for the role of people in the long-term evolution of the vegetation landscape of western Tasmania (Fletcher and Thomas, 2010), and the persistent human ignition source, long term (supra-centennial scale) trends in fire activity in this landscape are modulated by long-term climatic trends (Fletcher and Moreno, 2012). Moreover, the temporal and spatial resolution of the vast majority of sedimentary charcoal and pollen records in Tasmania (and indeed Australia) is predisposed to detecting changes occurring over large spatial and temporal scales. This feature of palaeoecological data, when coupled with the inherent and often large associated dating errors, renders most palaeoecological data insensitive to the small-scale landscape management practiced by Australian Aborigines today and in the past (Langton, 1998; Robinson and Plomley, 1966; Thomas, 1995; Yibarbuk, 1998). This limitation of palaeoecological data is exacerbated when synthesising data across vast regions that hosted economically, culturally and politically distinct groups of people occupying heterogeneous landscapes (AIATSIS, 1994).

78 |

4.2. Methods

4.2.1. Chronology and charcoal analysis

Chronologies were developed for all thirteen cores using a combination of radiocarbon and 210Pb analyses (Table 4-1). At least four radiocarbon dates were collected from each record and the age-depth models were developed using clam v.2 (Blaauw, 2010) applying the Southern Hemisphere calibration curve (SHCal13) (Hogg et al., 2013). This is a significant improvement over previous charcoal datasets employed in charcoal syntheses from this region (e.g. Mooney et al., 2011), some of which are not dated in the last 12 kyrs (e.g. Colhoun et al., 1999). The Constant Rate of Supply (CRS) model (Appleby and Oldfield, 1978) was used for calibrating 210Pb dates. Age depth models for individual sites are presented in Figure 4-S1. All sediment cores were subsampled for macroscopic charcoal analysis at 0.5-cm intervals (excluding TAS1402, which was subsampled at 0.25-cm intervals). All samples were digested in Sodium Hypochlorite (5%) until complete organic digestion (excluding charcoal; >2 weeks). Samples were then sieved using a 125 μm and 250 μm mesh (Whitlock and Larsen, 2001) and counted under a stereomicroscope. To account for changes in sediment accumulation rates, Charcoal Accumulation Rates (CHAR, particles cm-2 yr-1) were calculated using the following equation: (number of particles/volume) / deposition time (years per centimetre).

79 |

MAXIMUM ELEVATION MAXIMUM AGE of NUMBER OF 210Pb CORE CODE SITE NAME LATITUDE LONGITUDE BASIN DEPTH (m a.s.l.) the record (yrs BP) 14C DATES CHRONOLOGY (m)

TAS1104 BASIN 41° 58'50.96''S 145° 32'53.84''E 577 5.2 22709 10 Y TAS1102 GAYE 41° 49'35.12''S 145° 36'11.99''E 892 1.2 10858 4 N TAS1106 GWENDOLYN 42°15'44.58''S 145°49'23.11''E 923 30 11531 8 Y TAS1107 NANCY 42°15'31.56''S 145°49'37.62''E 1037 24.1 11291 7 Y TAS1108 VERA 42° 16'28.53''S 145° 52'47.73''E 571 48 18358 11 Y TAS1110 OSBORNE 43°12'58.37''S 146°45'33.46''E 920 9.5 14011 11 Y TAS1203 JULIA 41°53'21.22''S 145°34'34.09''E 616 12 10501 4 N TAS1205 SQUARE TARN 43°12'51.52''S 146°35'39.19''E 865 3.5 7918 5 Y TAS1207 HARTZ 43°14'17.12''S 146°45'23.62''E 952 40.5 5216 5 N TAS1402 SELINA 41°52'39.80''S 145°36'34.01''E 516 7.4 17393 15 N TAS1501 OWEN TARN 42°5'58.6''S 145°36'33.95''E 969 7 7525 11 Y TAS1503 ISLA 41°58'13.91''S 145°39'55.57''E 720 14 11935 9 Y TAS1504 ROLLESTON 41°55'17.35''S 145°37'29.12''E 560 42 4282 7 N Table 4-1. List of sites in western Tasmania used for the palaeofire analysis (map shown in Figure 4-1).

4.2.2. Palaeofire analysis

The 13 new high-resolution macroscopic charcoal records from western Tasmania were used for the Palaeofire analysis (Table 4-1) using the paleofire package (Blarquez et al., 2014) for R 3.3.1 (R Core Development Team, 2013). Firstly, due to the variability in the physiographic characteristics of sites, a transformation of the data was performed using the function pfTransform with MinMax, Box-Cox and Z-score methods. This process of transformation and standardization of different charcoal records is a highly recommended step in generating a synthesis (Blarquez et al., 2014). Here, we used the methodology proposed by Power et al. (2008), which involved a three-step data transformation including a min-max data-rescaling, variance homogenization using Box-Cox data transformation (Box and Cox, 1964), final rescaling to Z-scores. The palaeofire composite was calculated using the function pfCompositeLF, following the method proposed by Marlon et al. (2009) and Daniau et al. (2012). In this case, 100 years-bins were used, since it represents the best achievable resolution in order to include the majority of charcoal records for the entire reconstruction period. Confidence intervals (0.05, 0.1, 0.90 and 0.95 bootstrapped percentiles) were obtained using the function circboot with 1000 repetitions, which applies a "moving" or "circular" block bootstrap method, a

80 | modification of the method proposed by Kunsch (1989) to test significance of changes in the composite time series. This procedure is used for testing the significance of local minima or maxima in the composite time series. When a composite curve exceeds the confidence intervals, this means that those minima/maxima are significantly different from the long-term trends. Nevertheless, if the curve does not exceed the confidence levels, this fact still does not exclude the occurrence of important trends in the composite series (Blarquez et al., 2014).

To create a hemispheric comparison we synthesized a total of 23 southern South American charcoal records from the Global Charcoal Database (GDC; Power et al., 2010) located within the area where a positive correlation between local precipitation and SWW wind speed is observed (Figure 4-1a). The same methodology explained above for WTAS was employed to create the regional SSA charcoal composite. The list of sites used for this compilation is presented in Table 4-S2.

4.2.3. Statistical analyses

Generalized Additive Models (GAMs; Hastie and Tibshirani, 2006) were used to identify trends in the regional charcoal influx compilations for WTAS and SSA. The additive models do not involve a-priori parameter settings, instead allow the shape of the relationship to be determined from the data using penalised regression (Hastie and Tibshirani, 2006). Smooth term coefficients, model fixed effects and correlation matrix parameters were all estimated using restricted maximum likelihood (REML). All GAMs were fitted using the Mixed GAM Computation Vehicle (mgcv) package (Wood and Wood, 2007) for R 3.3.1 (R Core Development Team, 2013). To highlight periods of shifts in the regional charcoal influxes for WTAS and SSA, the first derivative of the GAM splines was calculated (github.com/gavinsimpson/tsgam). This approach allows measuring the slope at each point throughout a time-series, thus it enables us to extract trends in the palaeodata (e.g. Bennion et al., 2015).

81 |

To check for the existence of coherent frequency patterns between WTAS and SSA palaeofire records, we performed wavelet coherence analysis using the wavelet coherence toolbox for MATLAB (Grinsted et al., 2004). Analogously, we also tested if the onset of ENSO since the mid-Holocene is linked with a change in fire regime across WTAS by performing wavelet coherence between an ENSO proxy - binned to the same 100yrs- time windows as in the WTAS charcoal influx composite - (raw data from Moy et al., 2002) and the reconstructed charcoal influx from this region. Wavelet analyses have often been applied to climate data (e.g. Lau and Weng, 1995; Meyers et al., 1993; Wang and Wang, 1996) and provide a useful tool to reveal frequency localization in climate signals in time and for identifying coherence between two climate time series datasets through time (Torrence and Compo, 1998). Wavelet coherence analysis was chosen instead of cross-wavelet analysis because of the unsuitability of the latter to test for the interrelationship between two processes (Maraun and Kurths, 2004).

4.3. Results

A total of 2709 charcoal samples were analysed across thirteen locations in western Tasmania (Figure 4-2). The regional charcoal curve for western Tasmania obtained using the paleofire package in R shows maximum positive anomalies between 12-8 ka and a decline is observed since then, reaching minimum values at 5 ka. During the mid-Holocene a short-lived increase in charcoal influx is visible at ~6 ka, followed by slight decline and then a strong sharp increase since ~4 ka (Figure 4-3). High values peaks are observed at 9.2, 2.5, 1.5 ka and between 0.3 ka and the present, whereas negative charcoal dips are observed between 6.8 and 4.3 ka. These levels exceed the 90th confidence interval range given by the circular block bootstrap procedure (Figure 4-3), thus they are indicating that these minima and maxima are significantly different from linear long-term trends.

82 |

83 |

84 |

The GAM splines produced for WTAS and SSA highlight the significant trends in the data by efficiently removing the noise embedded in the time-series and thus allowing a better comparison of the palaeorecords (Figure 4-4).

A synchronous high fire activity period is evident between 12-7 ka in WTAS and SSA, followed by low fire activity between ~6 and ~4 ka. After ~4ka a sharp increase in charcoal influx in manifest in WTAS, whereas only a small increase up to around average values is present in the SSA composite. First derivative of GAMs splines shows remarkable similarities in the frequencies of regional charcoal influxes from WTAS and SSA (Figure 4-5a). The wavelet coherence analysis revealed a millennial- scale in-phase relationship (arrows pointing right) between the SSA charcoal influx and western Tasmanian palaeofire between 12-6 ka. A shift to a significant antiphase relationship between these two records is visible after 3 ka. Results from the

85 |

application of the wavelet coherence analysis to the regional charcoal influx from Tasmania and the ENSO proxy from Moy et al. (2002) clearly highlight a significant frequency coherence at millennial-scale periodicities between these two records during the period between 6.3 -1.7 ka (Figure 4-5d). Arrows pointing down can be interpreted as a 90° lead of X (in this case = WTAS charcoal influx) over Y (in this case = ENSO proxy) or a lag of 270° between the above mentioned time-series. Nevertheless, we believe that given the differences in temporal/spatial resolutions and chronological uncertainties within the records, we are unable to meaningfully test lead/lag relationships. Areas inside the “cone of influence,” where edge effects are present, have not been considered for the interpretation of the results.

86 |

indicate an in-phase relationship (left = anti-phase). Arrows pointing up or down represent possible lead/lag relationships between the time-series (Grinsted et al., 2004). Black lines indicate significant power spectra (significance=0.9). Areas inside the cone of influence should not be considered for interpretation as they are affected by edge effects.

4.4. Discussion

4.4.1. Western Tasmanian palaeofire reconstruction

Whilst the role of people using fire to deflect post-glacial vegetation development in western Tasmania and, thus, creating a cultural landscape, is clear (Fletcher and Thomas, 2010; Mariani et al., 2017), the spatial and temporal scale of our analysis far exceeds the spatial and temporal scale at which Australian Aboriginal land management with fire is and was undertaken (Robinson and Plomley, 1966; Thomas, 1995). Rather, the extent and impact of human landscape management using fire in Australia prior to British Invasion was most likely modulated over large spatiotemporal scales by climate. Given this assumption, we interpret our analysis of 13 new sedimentary charcoal influx (CHAR) records from western Tasmania as reflecting long-term trends in landscape-scale biomass burning that we interpret as substantial shifts in regional hydroclimate within the modern SWW-dominant zone of Tasmania. These data allow us to assess the paradigm depicting a shift from a dominance by multi-millennial scale shifts in the SWW between ca. 12-6 ka to a climate regime in which higher frequency (millennial to sub-millennial) ENSO activity is influential after ca. 6 ka (Fletcher and Moreno 2012; Rees et al. 2015). Importantly, our data reveal a shift from relatively high biomass burning between ca. 11-8.5 ka, with a strong multi-millennial frequency wavelength between ca. 10-8 ka, to a phase of persistently low biomass burning between ca. 8.5-6 ka. This was followed by a significant millennial/sub-millennial frequency wavelength after 6 ka and a sharp shift toward increased biomass burning after ca. 4 ka until the present. The persistently high biomass burning between ca. 11-8.5 ka indicates a prolonged phase of low relative moisture within the study region, while the persistent negative

87 | trend in biomass burning through the next few millennia signal a prolonged increase in relative moisture over western Tasmania, with a wet interval occurring between ca. 8.5-7 ka (Figure 4-3). The overall increase in biomass burning for the remainder of the record exceeds that recorded previously during the Holocene. Indeed, the most recent spike in fire activity in western Tasmania closely tracks reconstructions of SAM (Mariani and Fletcher, 2016), both supporting the notion of a close coupling between climate and fire activity in this region and signalling an unprecedented increase in fire activity in response to anthropogenic influences over the climate system. In terms of fuel type changes, it is important to note that shifts in regional fire activity occur independently of vegetation change recorded at Lake Vera (Figure 4-6; Macphail, 1979), one of the study sites used in this analysis, suggesting that fuel type and amount had little impact on the regional palaeofire trends.

88 |

89 |

4.4.2. Dominant drivers of western Tasmanian palaeofire dynamics

We compare our data to regional climatic proxies in an attempt to elucidate the drivers of palaeofire/hydroclimate throughout the last 12,000 years (Figure 4-6). The multi-millennial scale trends in biomass burning between ca. 12-7 ka in western Tasmania mirror both palaeofire trends from across southern South America (SSA, 40-55°S; Figure 4-6d) and pollen-inferred palaeomoisture trends derived from Lago Condorito (41°S; Figure 4-6b), Chile (Moreno, 2004), a lake within temperate forest that also lies within the latitudinal band of strongest correlation between modern rainfall and SWW on Earth (Gillett et al., 2006). Phases of high fire activity and low relative moisture in SSA (ca. 11.5-8 ka) correspond to phases of high biomass burning in western Tasmania (Fig.6 c,d) and a raft of evidence from around the hemisphere that indicates a prolonged phase of SWW attenuation in the mid to high latitudes of the Southern Hemisphere (Fletcher and Moreno, 2011, 2012; Lamy et al., 2010; Lisé-Pronovost et al., 2015; Moreno et al., 2010). Likewise, the shift toward low biomass burning across western Tasmania between ca. 8.5-7 ka is coeval with (1) a decrease in palaeofire activity in SSA (Figure 4-6d); (2) a sharp increase in moisture at Lago Condorito (Figure 4-6b); (3) a marked increase in reconstructed lake level at Lake Gnotuk (based on particle size), southwest Victoria, Australia (Figure 4-6a; Wilkins et al., 2013), a lake in receipt of predominantly SWW-derived rainfall; and (4) with other proxy evidence that indicate more SWW-derived moisture in the mid- latitudes/subtropical regions of the SH, such as increased moisture in central Chile (Jenny et al., 2003).

We note that lake level regression at Lake Gnotuk (38°S) commenced at ca. 7 ka, a timing that, while synchronous with a slight increase in charcoal in western Tasmania (between 41-44°S), coincides with an overall charcoal minima in western Tasmania and SSA between ca. 8-5.5 ka. One possible explanation for the divergence between the more northerly SWW-controlled Lake Gnotuk and the more southerly sites in western Tasmania and SSA is that the SWW began either shifting south or attenuating at ca. 7 ka, resulting in a decrease in SWW-derived moisture at more northerly locations while maintaining sufficiently wet climate in western Tasmania

90 |

and SSA to keep fire activity low. Unfortunately, our data cannot resolve whether these changes resulted from an attenuation of SWW flow across their latitudinal band (sensu Rojas and Moreno, 2010) or from latitudinal shifts (sensu Lamy et al., 2010).

Critically, our wavelet coherence analysis reveals a significant and in-phase relationship between western Tasmanian and SSA palaeofire between ca. 12-6 ka, followed by a shift to a significant anti-phase relationship after ca. 3 ka. Further, the wavelet coherence between western Tasmania palaeofire and the Moy et al. (2002) ENSO proxy reveals a significant coherence in the frequency of oscillations between these two proxies after ca. 6 ka (Figure 4-5d), concomitant with the onset of ‘modern’ ENSO variability inferred from the Moy et al. (2002) record (Rodbell et al., 1999; Moy et al., 2002; Koutavas et al., 2002). Critically, uncertainties over the response times and dating of sedimentary archives, coupled with the uncertainties inherent in interpreting leads/lags based on arrow phase direction in wavelet coherence plots (Grinsted, 2004), precludes confidence in interpreting the phase direction of the relationship between ENSO and western Tasmanian palaeofire. Nevertheless, the significant frequency coherence between ENSO and western Tasmanian palaeofire after ca. 6 ka adds additional empirical support for the role of ENSO in breaking down the symmetric multi-millennial scale SWW-driven trends in palaeofire activity between western Tasmania and SSA after ca. 6 ka. We also add to the significant body of evidence indicating that the onset of ENSO at ca. 6 ka was contemporaneous with a marked shift of the climate of the Pacific Ocean region (Black et al., 2008; Conroy et al., 2008; Fletcher and Moreno, 2012; Sandweiss et al., 2001; Shulmeister and Lees, 1995) and, indeed, for the interpretation of a global ENSO footprint (e.g. Herweijer et al., 2007; Graham et al., 2010).

91 |

4.4.3. Implications for modern climate patterns and non-stationary teleconnections

Our evidence for ENSO-driven hydroclimatic change in western Tasmania is inconsistent with the modern climatology of this SWW-dominated region (Figure 4- 1) (Hendon et al., 2007; Hill et al., 2009; Mariani and Fletcher, 2016), highlighting the limitations of modern climate analyses for interpreting proxies of past climatic change. While it is recognised that climate teleconnections vary in strength and direction through time (Fogt and Bromwich, 2006; Garreaud, 2007; Garreaud et al., 2009), so-called ‘non- stationarity’ (sensu Gallant et al., 2013), what is less appreciated is how significant changes in strength of particular components of the climate system are manifest climatically across the Earth. Modern climate data and our results reveal that phases of particularly strong ENSO variability can drive substantial climatic and environmental change in areas outside of their contemporary zone of influence. This finding is important, as systems such as ENSO and SWW are currently changing in response contemporary climate change (Mariani and Fletcher, 2016; Perlwitz et al., 2008; Power et al., 2013; Thompson et al., 2011) and there is an urgent need for an understanding of how climate signals change in response to changes in the strength and frequency of key components of the global climate system. A case in point are the catastrophic fires that have recently swept across the landscape of Tasmania (January 2016), threatening the existence of fire- sensitive natural systems. These fires are the product of an anthropogenic-driven positive trend in SAM over the last 60 years (Perlwitz et al., 2008; Mariani and Fletcher, 2016) during a particularly strong El Niño year embedded within a long- term phase of relatively muted ENSO activity (Moy et al., 2002; Conroy et al., 2008; Yan et al., 2011). Our data reveal a clear role of ENSO in driving landscape-scale biomass burning in western Tasmania, heralding a further potential threat to these systems in response to the current and potential future intensification ENSO system (Guilyardi, 2006; Lenton et al., 2008; Power et al., 2013). Further, there is an urgent need for long-term reconstructions of climate modes other than ENSO (e.g. SAM), if we are to tease apart the drivers of climatic and environmental change and tackle questions of stationarity in climate teleconnections through time.

92 |

4.5. Conclusions

Under a warming world scenario in which fire activity is predicted to increase in temperate regions (Moritz et al., 2012), it is essential to understand the role of climate dynamics (such as ENSO and SWW) in driving long-term fire activity. Our compilation of 13 new sedimentary charcoal records from temperate Tasmania, Australia, an island that lies within the band of strongest influence of the SWW on rainfall anomalies on Earth, has allowed the identification of clear long term shifts in biomass burning. These shifts are driven by hemispheric-scale climate dynamics resulting from the dynamic interplay between the SWW and ENSO systems. We identify a high fire activity period from 12 to 8 ka, a minimum between ca. 7-5 ka and a sharp increase since ~4 ka exceeding Holocene variability during the last 0.3 ka. A persistent, yet oscillating, increase in burning since the mid-Holocene coincides with a breakdown of zonal symmetry across the Southern Hemisphere and marks the influence of increased tropical ENSO variability (since ~6 ka). Wavelet coherence analysis suggests a significant in-phase relationship between western Tasmania and southern South America fire activity up to ~6 ka. A significant frequency coherence between El Niño events in Ecuador (Moy et al., 2002) and western Tasmanian palaeofire was also detected since ca. 6 ka, highlighting the effects of the ENSO onset on fire activity in this region. While recent (last millennium) fire activity in western Tasmania was driven by shifts in the SWW (Mariani and Fletcher, 2016), our data indicate the potential for marked increases in the frequency/intensity of ENSO variability to drive significant climatic change in areas considered beyond its zone of influence according to modern climatology.

93 |

4.6. Supplementary results and information

The figures and tables displayed below are part of the supporting information for the paper ‘Long-term climate dynamics in the extra-tropics of the South Pacific revealed from sedimentary charcoal analysis’ (Mariani and Fletcher, 2017).

94 |

4.6.1. Additional details about the sites location

95 |

4.6.2. Additional information about cores chronology

Table 4-S1a. Radiocarbon dates obtained on the 13 cores analysed for this study. Ages were calibrated in OxCal 4.2 (Ramsey, 2013) using the Southern Hemisphere calibration curve (Hogg et al., 2013).

Upper Lower Median CORE Depth 14C Error cal BP cal BP age (cal CODE (cm) age (yrs) (2σ) (2σ) BP) TAS1106 10.5 1519 23 1520 1345 1398 TAS1106 13.25 1880 23 1880 1736 1833 TAS1106 18.25 2030 27 2098 1899 1978 TAS1106 25.75 3327 28 3634 3478 3561 TAS1106 31.25 4249 30 4864 4662 4836 TAS1106 40.75 5818 31 6727 6508 6629 TAS1106 55.75 8410 37 9521 9312 9450 TAS1106 67.75 9531 39 11083 10696 10872

TAS1107 12.75 2029 27 2060 1898 1977 TAS1107 17.25 2773 31 2949 2789 2867 TAS1107 24.75 4619 34 5466 5292 5406 TAS1107 35.5 6508 33 7485 7325 7428 TAS1107 50.25 8306 38 9440 9141 9334 TAS1107 60.75 8977 37 10233 9930 10176 TAS1107 70.75 9801 38 11256 11179 11220

TAS1108 23.25 650 25 669 558 595 TAS1108 50.25 1260 30 1282 1086 1218 TAS1108 81.25 2040 30 2111 1904 1995 TAS1108 87.5 2250 25 2341 2158 2229 TAS1108 111.5 3260 30 3565 3403 3488 TAS1108 131.5 4590 25 5445 5085 5312 TAS1108 196.5 8090 30 9121 8985 9017 TAS1108 269.5 11200 35 13133 13005 13073 TAS1108 308.5 13300 60 16205 15779 15995 TAS1108 318.5 14900 65 18315 17919 18113 TAS1108 327.5 15500 70 18905 18604 18761

TAS1110 8 985 25 921 794 856 TAS1110 15 1440 30 1352 1274 1304 TAS1110 21 1720 25 1616 1530 1579 TAS1110 25 1940 25 1906 1747 1849 TAS1110 39 2960 25 3163 2961 3060 TAS1110 50 3500 25 3829 3639 3724 TAS1110 65 4260 25 4680 4628 4745 TAS1110 161 5520 30 6370 6195 6282 TAS1110 226 7280 40 8164 7971 8057 TAS1110 260 8230 35 9269 9022 9133

96 |

Upper Lower Median CORE Depth 14C Error cal BP cal BP age (cal CODE (cm) age (yrs) (2σ) (2σ) BP) TAS1402 8.25 2401 40 2698 2344 2437 TAS1402 11.25 2614 27 2771 2725 2749 TAS1402 13.75 3649 28 4083 3889 3965 TAS1402 17.5 4589 26 5447 5080 5311 TAS1402 23.25 6105 40 7157 6885 6981 TAS1402 30.75 7194 33 8152 7945 7998 TAS1402 38.25 8515 39 9543 9473 9511 TAS1402 42.25 8723 36 9886 9552 9673 TAS1402 52.25 10414 35 12517 12096 12282 TAS1402 62.25 11726 41 13710 13446 13529 TAS1402 67.75 12123 40 14134 13824 14001 TAS1402 72.25 12457 53 14985 14240 14604 TAS1402 77.25 12957 40 15696 15287 15485 TAS1402 82.25 13878 54 17039 16555 16813 TAS1402 84.5 14234 52 17523 17127 17335

TAS1503 10.75 1343 25 1306 1187 1283 TAS1503 23.75 2603 27 2765 2719 2746 TAS1503 40.75 4282 31 4959 4821 4849 TAS1503 56.25 5646 39 6499 6315 6427 TAS1503 70.75 7105 40 8005 7850 7939 TAS1503 80.5 7928 47 8982 8610 8771 TAS1503 90.75 8575 45 9654 9481 9539 TAS1503 107.25 9411 36 10738 10560 10640 TAS1503 120.25 9991 42 11699 11268 11456

TAS1203 1917 32 8.75 1890 1729 1812 TAS1203 3865 29 20.75 4405 4093 4218 TAS1203 4909 24 43.75 5658 5487 5606 TAS1203 8972 39 60.25 10200 9910 10046

TAS1501 790 40 18.25 739 571 686 TAS1501 1402 25 24.75 1311 1188 1286 TAS1501 1880 35 27.75 1874 1702 1777 TAS1501 2300 28 34.25 2348 2160 2247 TAS1501 2904 24 40.25 3073 2873 2978 TAS1501 3332 26 45.75 3608 3446 3516 TAS1501 5050 60 50.25 5902 5611 5750 TAS1501 5515 35 53.75 6393 6188 6276 TAS1501 6465 31 65.25 7422 7276 7357 TAS1501 6537 34 67.75 7482 7316 7401 TAS1501 7810 60 67.76 8700 8410 8543

97 |

Upper Lower Median CORE Depth 14C Error cal BP cal BP age (cal CODE (cm) age (yrs) (2σ) (2σ) BP) TAS1102 1406 33 15.75 1342 1185 1285 TAS1102 2467 22 40.25 2699 2352 2454 TAS1102 6128 31 76.25 7156 6799 6949 TAS1102 8102 28 89.25 9085 8777 8993

TAS1504 690 24 6.25 660 559 601 TAS1504 1355 35 12.75 1302 1177 1232 TAS1504 2000 40 20.75 2004 1828 1913 TAS1504 2648 30 29.25 2790 2520 2745 TAS1504 3240 30 40.25 3551 3351 3419 TAS1504 3449 27 44.25 3820 3568 3654 TAS1504 3677 28 49.75 4082 3849 3947

TAS1207 1397 33 20.25 1313 1185 1280 TAS1207 2388 29 60.25 2486 2312 2360 TAS1207 2947 34 100.3 3169 2925 3041 TAS1207 3992 34 140.3 4520 4250 4407 TAS1207 4599 33 180.3 5438 5050 5170

TAS1205 2258 29 28 2326 2154 2231 TAS1205 3488 26 55 3830 3615 3705 TAS1205 5550 29 93 6399 6217 6304 TAS1205 6600 31 114 7561 7421 7463 TAS1205 7127 37 133.4 7999 7832 7909

TAS1104 1750 25 22.25 1703 1545 1623 TAS1104 2470 25 34.25 2699 2352 2463 TAS1104 4900 30 72.25 5658 5482 5600 TAS1104 4815 40 88 5596 5329 5513 TAS1104 6447 38 166 7422 7266 7342 TAS1104 8800 60 238 10128 9548 9763 TAS1104 10120 57 289 11950 11334 11630 TAS1104 10952 69 311 12971 12697 12785 TAS1104 14268 38 341 17504 17130 17323 TAS1104 16930 80 350.5 20597 20107 20368

98 |

Table 4-S1b. 210-Pb dates obtained on a subset of the 13 cores analysed for this study.

CORE CODE Depth (cm) CRS age (yrs BP) CRS age error (yrs) TAS1106 0.75 -35.87 3.61 TAS1106 1.25 -21.33 6.64 TAS1106 1.75 -1.93 26.05 TAS1106 2.25 8.33 26.04 TAS1106 2.75 27.22 66.46 TAS1106 3.25 46.45 74.31

TAS1107 0.75 -44.97 1.57 TAS1107 1.25 -28.1 3.58 TAS1107 1.75 -6.5 9.88 TAS1107 2.25 7.96 12.21 TAS1107 2.75 38.5 36.72 TAS1107 3.25 72.77 99.03

TAS1503 0.25 -60.36 2.15 TAS1503 0.75 -47.99 4.12 TAS1503 1.25 -32.84 5.67 TAS1503 1.75 -22.00 6.56 TAS1503 2.25 -13.10 7.20

TAS1108 0.3 -54.14 6.9 TAS1108 0.8 -40.41 7 TAS1108 1.8 -12.95 7.7 TAS1108 2.8 14.51 8.8 TAS1108 5.3 83.15 12.5

TAS1110 0.5 -19 11 TAS1110 1.5 17 22

TAS1501 0.25 -63 2 TAS1501 1.25 -56 3 TAS1501 3.75 -48 4 TAS1501 5.25 -43 5 TAS1501 11.75 -5 8 TAS1501 14.25 17 9

TAS1205 0.5 -46 4 TAS1205 1 -33 6 TAS1205 1.5 -6 18 TAS1205 3.5 28 79

TAS1104 0.25 -57 4 TAS1104 2.25 -25 9 TAS1104 4.25 7 16 TAS1104 6.25 38 22

99 |

Figure 4-S2a. Age depth models of the thirteen cores used in this study. All age-depth models have been performed using clam v.2 in R.

100 |

Figure 4-S2b. Age depth models of the thirteen cores used in this study. All age-depth models have been performed using clam v.2 in R.

101 |

4.6.3. Additional information about the Southern South America compilation

Table 4-S2. List of sites from the Global Charcoal Database (GCD; Power et al., 2010) used for the paleofire compilation from Southern South America (SSA) between 40 and 55°S.

Number of Dates Number of Site ID (GCD) Latitude Longitude Elevation Min Age Max Age dates interval samples Lake Facil 75 -44.325 -74.2833 10 0 15966.2 10 1596.62 101 Laguna Venus 76 -45.5333 -72.01667 600 -34 1650 2 842 58 Trebol Lake 92 -41.07125 -71.4916 758 -52 22588 15 1509.3333 577 Mosquito Lake 93 -42.496 -71.4029 556 126.7039 9378.959 20 462.61275 1456 El Salto 143 -41.64 -73.0961 67 288.8 16069 12 1315.0167 430 Lago Condorito 144 -41.75 -73.1167 60 180.7 14335.4 18 786.37222 249 Lepue 147 -42.8 -73.7 152 324.6 18356.1 24 751.3125 1131 Laguna Oprasa 149 -44.35556 -73.65556 50 0 17295.1 12 1441.2583 91 Lago Pollux 150 -45.6756 -71.8629 640 -3.21 17695.18 18 983.24401 1020 Vegaandi 152 -50.93278 -72.76528 200 109.9 12431.8 10 1232.19 170 Torres del Paine2 154 -51.08333 -73.06667 100 57.43746 13126.68 9 1452.1379 86 Lago Guanaco 155 -51.13306 -73.10972 60 -39.1 3697 9 415.12222 164 Potrok Aike 157 -51.96667 -70.38333 100 -42 437 NA NA 13 Rio Rubens 158 -52.1375 -71.88139 220 -12 21367 NA NA 994 Punta Arenas 160 -53.15 -70.95 75 7.716152 17289.69 7 2468.8527 81 Puerto del Hambre 162 -53.6 -70.91667 3 47.197 19222.17 7 2739.2824 74 Puerto Haberton 170 -54.88333 -67.16667 20 -40 15422.92 14 1104.4943 105 Canal de la Puntilla 554 -40.95 -72.9 120 -50 23503 24 981.375 72 Laguna Azul 622 -52.12047 -69.52269 100 -14 1151 13 89.61538 33 Laguna Lincoln 1069 -45.36667 -74.06667 19 41 15937 5 3179.2 91 Laguna Lofel 1070 -44.92884 -74.32533 13 -48 15930 6 2663 70 Laguna Six Minutes 1071 -46.41667 -74.3333 15 -3 16852 5 3371 82 Laguna Stibnite 1072 -46.41667 -74.4 15 684 16867 10 1618.3 103

102 |

Chapter 5. Calibrating the pollen-vegetation data3

3 This chapter is an adapted version of the manuscript published in Quaternary Science Reviews (Mariani et al., 2016): Testing quantitative pollen dispersal models in animal-pollinated vegetation mosaics: An example from temperate Tasmania, Australia. The analyses here presented address part of Thesis Aim III (Step 1; as stated in Chapter 1 - Introduction).

Abstract

Reconstructing past vegetation abundance and land-cover changes through time has important implications in land management and climate modelling. To date palaeovegetation reconstructions in Australia have been limited to qualitative or semi-quantitative inferences from pollen data. Testing pollen dispersal models constitutes a crucial step in developing quantitative past vegetation and land cover reconstructions. Thus far, the application of quantitative pollen dispersal models has been restricted to regions dominated by wind- pollinated plants (e.g. Europe) and their performance in a landscape dominated by animal- pollinated plant taxa is still unexplored. Here we test, for the ﬁrst time in Australia, two well-known pollen dispersal models to assess their performance in the wind- and animal- pollinated vegetation mosaics of western Tasmania. We focus on a mix of wind- (6 taxa) and animal- (7 taxa) pollinated species that comprise the most common pollen types and key representatives of the dominant vegetation formations. Pollen Productivity Estimates and Relevant Source Area of Pollen obtained using Lagrangian Stochastic turbulent simulations appear to be more realistic when compared to the results from the widely used Gaussian Plume Model.

103 |

5.1. Introduction

The development of realistic estimates of vegetation and land cover change from sub-fossil pollen data is crucial for attempts to both understand how vegetation responds to factors such as climatic change and disturbance. Terrestrial vegetation is an important component of the Earth System that is both influenced by climate and influences climate through biogeochemical and bio-geophysical processes/feedbacks (e.g. Foley et al., 2003). Indeed, recognition of this mutual influence has led to the coupling of dynamic vegetation models (DVMs) with climate models (e.g. Smith et al., 2011) and fire models (Thonicke et al., 2001) in an attempt to better model Earth System dynamics. Critically, DVMs coupled with climate and fire models only simulate climate-induced potential vegetation and do not take account actual past land-cover changes (Smith et al., 2011; Thonicke et al., 2001). Thus, it stands that the development of quantitative land-cover estimates through time is of critical importance for improving models’ performance (Strandberg et al., 2014; Trondman et al., 2015).

Pollen data are often times dramatically skewed in favour of a few abundant pollen types, a fact that has hampered quantitative vegetation reconstruction since the beginning of the palynological research (von Post, 1946). This gap in our understanding not only prevents attempts to fully understand how vegetation systems respond to environmental change, it runs the risk of mismanagement of ecosystems whose baseline variability is poorly known. Not until the most recent decade have increased computer power, reduced analysis time and the compilation of large pollen datasets conspired to allow the development of robust quantitative techniques suitable for Quaternary pollen data (Sugita, 2007a,b; Gaillard et al., 2010). Application of these cutting-edge advancements is largely restricted to Europe and North America (e.g. Sugita et al., 1999; Brostrom et al., 2004; Bunting et al., 2005; Rasanen et al., 2007; Soepboer et al., 2007, 2008; Filipova-Marinova et al., 2010; Poska et al., 2011; Abraham and Kozakova, 2012; Hjelle and Sugita, 2012; Li et al., 2015), with virtually no attempts at applying these approaches in the Southern Hemisphere (Dufﬁn and Bunting, 2008). The ﬁrst step toward generating robust estimates of

104 |

vegetation from pollen data is the development of suitable pollen dispersal models (e.g. Theuerkauf et al., 2013). Here we test empirical pollen dispersal models in western Tasmania, Australia, a region that substantially differs from the models’ previous applications in the Northern Hemisphere because of the abundance of animal-pollinated plant taxa in the vegetation canopy.

The Landscape Reconstruction Algorithm (LRA) (Sugita, 2007a,b), developed in Europe, represents the state-of-the-art model for generating quantitative land cover estimates from pollen data. The LRA is underpinned by pollen dispersal models developed, to date, principally from Northern Hemisphere plants that are assumed to be largely wind-dispersed, especially canopy trees from the Betulaceae, Fagaceae and Pinaceae families (Tauber, 1965; Andersen, 1970, 1974; Andersen, 1974; Prentice, 1985; Sugita, 1993, 1994, 2007a,b; Calcote, 1995). The assumption of wind-dispersal from the canopy is a critical limitation that restricts the application of the Landscape Reconstruction Algorithm to vegetation types dominated by wind-dispersed canopy species. In landscapes with animal-mediated pollination, the applicability of these models for vegetation reconstruction is unknown and has been questioned (Walker, 2000; Dufﬁn and Bunting, 2008).

In Australia, the most important canopy dominants (e.g. Eucalyptus, Acacia, Melaleuca) are pollinated by insects, marsupials, birds, insects and bats (zoophilous pollination) (Andersen, 1970; Kershaw and Strickland, 1990). The lack of quantitative information about pollen dispersal parameters of these taxa constitutes a signiﬁcant knowledge gap. While the limited application of semi-quantitative techniques in Australian palynology has allowed objective inferences of vegetation change from pollen data (e.g. Kershaw, 1979; Fletcher and Thomas, 2010a,b), the lack of truly quantitative approaches to vegetation reconstruction severely limits the understanding of long-term vegetation change in this region. A case-in-point is the long-standing debate over if and when moorland replaced rainforest in western Tasmania during the mid to late Holocene (Thomas, 1995; Colhoun, 1996; Fletcher and Thomas, 2010a,b; Thomas et al., 2010; Macphail, 2010). Despite dominating the landscape, moorland vegetation is virtually invisible in the pollen spectra (Fletcher

105 |

and Thomas, 2007). Indeed, over-representation of anemophilous rainforest taxa in pollen spectra has led to the routine ignorance of the existence of treeless vegetation in this landscape (Colhoun, 1996; Pickett et al., 2004; Colhoun and Shimeld, 2012) and there is a need to quantitatively address the biases inherent in pollen data to better understand long term vegetation dynamics in landscapes such as this.

In this paper, we aim to test, for the ﬁrst time in Australia, the application of pollen dispersal models in order to quantify past land-cover changes in western Tasmania, Australia. Speciﬁcally, we test two of the leading dispersal models, the Gaussian Plume Model (GPM) and the Lagrangian Stochastic Model (LSM), to assess their performance in a landscape dominated by animal- and insect-pollinated plant taxa, a mix that differs substantially from the vegetation in which they were developed. Testing these models has important implications for 1) extending the geographical scope of quantitative land-cover reconstructions to the Southern Hemi-sphere and to areas which have been previously ignored because of the abundance of animal- pollinated plant taxa and 2) providing an objective test of competing theories of Holocene landscape evolution in such areas (e.g. Tasmania).

5.2. Pollen dispersal, production and relevant source area

Pollen dispersal is one of the major factors in determining the representation of vegetation in pollen spectra, along with taphonomy, pollen productivity and pollination mode (Faegri and Iversen, 1989). These issues often result in a non-linear relationship between pollen percentages and vegetation cover (Fagerlind, 1952; Davis, 1963; Faegri and Iversen, 1989). Pollen productivity varies among species depending on biological and ecological parameters such as their pollination system, plant life forms, ﬂower traits, vegetation dynamics, structure and climate (Faegri and Iversen, 1989). Over-representation of pollen types occurs because pollen is produced abundantly or disperses easily; wind-pollinated plants are typically better represented in pollen rain than animal-pollinated species (Jacobson and Bradshaw, 1981). Therefore, pollen representation is often related to the pollination mode (e.g.

106 |

wind, animals) (e.g. Fletcher and Thomas, 2007; de Nascimento et al., 2015). Little work has been carried out on the pollen vectors of Australian plant taxa, with the existing literature revealing a large variety of biotic mechanisms controlling the pollination of Australian's ﬂora (e.g. Armstrong, 1979; Serventy and Raymond, 1974; Paton and Ford, 1977; Ford et al., 1979; Hopper, 1979; Hopper and Moran, 1981; Irvine and Armstrong, 1990). The pollination behaviour of the most studied plant taxa in Australia remains uncertain: for instance, Eucalyptus (Australia's most common canopy dominant) is mostly commonly believed to be a zoophilous plant taxon pollinated by birds, mammals and insects (Armstrong, 1979; Regal, 1982), but anemophily has been suggested to occur in a few species (Pryor, 1976). Furthermore, while zoophily has also been suggested for many other sclerophyllous taxa in this region (e.g. Proteaceae and Myrtaceae) (Regal, 1982), no information is available for the key components of much of the Australian ﬂora including the temperate rainforests that occupy the wettest parts of Australia's southeast.

Implicitly or explicitly, all interpretations of pollen diagrams make assumptions about pollen dispersal and source area. These parameters have been modelled in various ways (Prentice, 1985; Sugita, 1993, 1994; Kuparinen et al., 2007; Theuerkauf et al., 2013), with the Gaussian Plume Model underpinning the LRA (Prentice, 1985; Sugita, 1993, 1994) being the most widely-used model of pollen dispersal. The GPM is based on Sutton's air pollutant plume dispersion equation (Sutton, 1953). The model has been calibrated using the concentration of particles (e.g. pollen) several hundred meters downwind from a point source as spreading outward from the centreline of the plume following a normal probability distribution. The GPM fails to realistically predict particle dispersal over longer distances, which is largely governed by vertical airﬂows and updrafts (Kuparinen et al., 2007). Fully mechanistic dispersal models such as the Lagrangian Stochastic Model (LSM) describe pollen dispersal more realistically and may be therefore more suitable to model pollen deposition in lakes (Theuerkauf et al., 2013). In contrast to the Gaussian Plume Model, when considering pollen spectra from lake surface sediments, the Lagrangian Stochastic Model gives greater importance to pollen arriving from 10 to 100 km distance (Theuerkauf et al., 2013).

107 |

Pollen dispersal models are used to estimate two key parameters of the quantitative vegetation reconstruction algorithm: Pollen Productivity Estimates (PPEs) and Relevant Source Area of Pollen (RSAP). PPEs are a measure of the amount of pollen released for transport per unit area of pollen-producing vegetation (grains/m2/yr). PPEs are derived from surface sample studies and usually expressed as a dimensionless ratio relative to a reference taxon (Relative Pollen Productivity) (Bunting et al., 2013). Notwithstanding the assumed pollination mechanisms adopted by plant taxa, pollen productivity may be effectively estimated using a dataset of modern pollen and distance-weighted vegetation abundance (e.g. Brostrom et al., 2008). The RSAP is defined by a radial distance from a sampling point, beyond which the relation-ship between pollen and cumulative distance- weighted vegetation data does not improve (Sugita, 1994). The RSAP is affected by a variety of factors, including basin type and size (e.g. lake or bog) and the structure of the surrounding vegetation mosaic (Sugita, 1994; Bunting et al., 2004). The RSAP specifies the smallest spatial scale at which site-to-site variation in the vegetation is reflected in the pollen assemblages and it assumes that pollen coming from beyond the RSAP (“background”) is virtually constant at all sites within a given region.

5.3. Study area

Tasmania is an appropriate location for testing pollen dispersal models in the Australian context because it is geographically constrained (limiting the pollen source-area), has a relatively simple ﬂora (improving pollen-taxonomic precision) and has more wind-dispersed canopy trees than most Australian forests, enabling the effect of pollination strategy to be analysed.

Tasmania has a temperate maritime climate with mild winters and cool summers. A rain-shadow is produced by the prevailing westerly winds (SWW - Southern Westerly Winds) that rise over the NW-SE trending mountain range that bisects Tasmania, resulting in a super-humid west and a sub-humid east (Sturman and Tapper, 2006). Mean annual temperature ranges from 6°C on the mountain tops and

108 |

the elevated Central Plateau region of the west to 15°C in coastal regions. Average total annual rainfall is above 1000 mm in western Tasmania, exceeding 3500 mm on the West Coast Range, and declining sharply east of the ranges (falling to <400 mm/yr) (Bureau of Meteorology, BOM). This sharp rainfall boundary differentiates western Tasmania from the eastern portion of the island in terms of climate and vegetation distribution (Gentilli, 1972; Sturman and Tapper, 2006). The persistently high moisture levels and the abundance of lakes in this formerly glaciated region makes it an ideal location for fossil pollen analysis. Indeed, the majority of high- resolution palaeoecological data collected from Tasmania come from the western bioclimatic province. These records reveal revealing a tight link between ﬁre regimes and landscape dynamics (e.g. Fletcher et al., 2014). This area is characterised by a landscape-scale vegetation mosaic of pyro-phobic temperate rainforest and pyrophitic sclerophyllous forest, scrub and moorland vegetation (Kirkpatrick and Dickinson, 1984). The distribution of these vegetation types is governed by interactions between topography, soil type and ﬁre history, and boundaries between contrasting vegetation types are usually sharp (i.e. little or no ecotone) (Bowman, 2000; Fletcher and Thomas, 2010b; Wood et al., 2011; Wood and Bowman, 2012).

In western Tasmania, regionally important tree species, Lophozonia cunninghamii (syn. Nothofagus cunninghamii) and Phyllocladus aspleniifolius, have been found to be over-represented in the pollen rain, while other communities have been found to be under-represented (Macphail, 1979; Fletcher and Thomas, 2007). For instance, while beech forests are considered over-represented palynologically, buttongrass moorland (dominated by Gymnoschoenus sphaerocephalus), one of the dominant vegetation communities, has gone undetected in attempts to predict regional vegetation from pollen spectra (Pickett et al., 2004; Fletcher and Thomas, 2007; 2010a,b). This highlights the need for objective and quantiﬁed vegetation cover estimates. The ﬁrst quantitative assessment of pollen-vegetation relationships in western Tasmania was performed by Fletcher and Thomas (2007), providing useful information on the key pollen taxa, related vegetation types and assumed representation in the pollen rain. The methodology used during their study, however, does not satisfy the strict data format requirements to obtain pollen

109 |

productivity estimates through the application of pollen dispersal models, which require distance weighted plant abundance. The present work applies distance- weighted plant abundance to determine modern pollen-vegetation relationships, which is essential to calculate the key parameters of the Landscape Reconstruction Algorithm (PPEs and RSAP), and thus quantitatively estimate plant cover changes using pollen-based models.

5.4. Methods

5.4.1. Surface samples collection and vegetation surveys

Across 27 sampling locations (Table 5-1), thirteen pollen taxa (Table 5-2) have been selected for calibrating modern pollen-vegetation relationships in western Tasmania. Selection of these pollen types was based on: 1) their ecological importance in modern vegetation and 2) their representation in Tasmanian fossil assemblages (i.e. collectively these 13 taxa constitute at least 75% of the pollen sum in the cores studied thus far in this region, see Fletcher et al., 2014). Approximately 25-40 pollen taxa are routinely identiﬁed in the rainforest/moorland areas of Tasmania, contrasting with much greater taxonomic richness in the tropical rainforest zone of Northern Australia (e.g. Kershaw and Strickland, 1990). In this paper the genus name Nothofagus sensu lato will be used when referring to Lophozonia cunninghamii and Fuscospora gunnii (syn. Nothofagus gunnii), in order to maintain consistency with the fossil records (Hill et al., 2015).

Vegetation survey methods and sampling strategies have an important bearing on the resultant PPEs (Bunting and Hjelle, 2010; Twiddle et al., 2012). According to Bunting et al. (2013), sampling sites should be spaced at least one RSAP distance from each other to avoid problems associated with spatial auto-correlation. Since we currently have little information on the pollen source area of the selected taxa in western Tasmania, and in order to minimize spatial autocorrelation, we selected

110 |

sites at least 1 km from each other. Another key step in calibration is the selection of at least twice the number of sampling locations as the number of target pollen taxa (Soepboer et al., 2007; Sugita, 2007b; Theuerkauf et al., 2013) - i.e 26 sites are required for 13 taxa. Moss cushions are commonly used as surface samples in studies of pollen-vegetation relation-ships (e.g. Brostrom et al., 2004; Bunting et al., 2005). We chose to sample brown moss and Sphagnum cushions since they are assumed to record an average of several years of pollen deposition (e.g. Heim, 1970; Bradshaw, 1981), which is recommended to avoid annual variations in pollen productivity (Hicks, 2001; Rasanen et al., 2004; Van der Knaap et al., 2001). A total of 27 surface sample locations were surveyed in February 2016 (Figure 5-1) and multiple moss/Sphagnum samples were collected within a 1 m radius and combined into a composite sample to avoid local over-representation. Sampling took place in vegetation clearings >10 m diameter to avoid the effects of pollen addition through routes other than aerial deposition, such as insects or anther fall from the canopy, which can cause spikes in the percentage of one taxon (Bunting et al., 2013). Vegetation surveys were carried out using nested rings, in order to capture the required information about vegetation along transects at increasing distances from the sampling point (Figure 5-2). In the ﬁeld, three rings (0-5m, 10-15 m and 100 m) were surveyed directly. For the intermediate (10-100 m) and outer rings (100-2000 m) vegetation cover was estimated from GIS using detailed vegetation maps available for the study area (TasVeg 3.0; Government of Tasmania, 2013). The ﬁrst two rings were surveyed in 5 x 5 m quadrats (Central, N, S, E, W), whereas the 100 m ring was surveyed in 10 x 10 m (Figure 5-2). Studies from the US have shown tight pollen- vegetation relationships using basal area (Calcote, 1995; Davis, 1998; Parshall; 2002; Sugita et al., 2010). According to more recent empirical studies by Bunting and Hjelle (2010) and Bunting et al. (2013), the cover method of vegetation survey is most similar to a pollen sample's view of the vegetation, i.e. the pollen signal of a taxon is most strongly related to visual estimates of plant cover instead of vegetation data obtained using other methods (such as basal area). No empirical evidence of this relationship is available for Tasmania.

111 |

112 |

Site Code Region Area Lat (°S) Long (°E) Elevation Material Vegetation type SS1 Central Highlands Cradle Mountain -41.509858 145.989802 766 Sphagnum Wet eucalypt forest and woodland SS2 Central Highlands Cradle Mountain -41.63866 145.9502217 882 Moss Scrub, heathland and coastal complexes SS3 Central Highlands Cradle Mountain -41.6656235 145.9565994 918 Moss Rainforest and related scrub SS4 Central Highlands Cradle Mountain -41.583448 145.9346271 906 Moss Rainforest and related scrub SS5 Central Highlands Cradle Mountain -41.5547521 145.9194861 900 Moss Native grassland SS6 Central Highlands Cradle Mountain -41.5345163 145.908478 830 Sphagnum Highland and treeless vegetation SS7 Central Highlands Cradle Mountain -41.6356672 145.9396304 1040 Moss Rainforest and related scrub SS8 Central Highlands Cradle Mountain -41.6505244 145.9636482 934 Moss Moorland, sedgeland, rushland and peatland SS9 Central Highlands Cradle Mountain -41.5526827 145.8969687 920 Sphagnum Highland and treeless vegetation SS10 Central Highlands Cradle Mountain -41.6612326 145.9382112 1270 Sphagnum Rainforest and related scrub SS11 Central Highlands Cradle Mountain -41.65177286 145.9447763 1054 Moss Moorland, sedgeland, rushland and peatland SS12 Central Highlands Cradle Mountain -41.63526113 145.9600009 955 Sphagnum Scrub, heathland and coastal complexes SS13 Central Highlands Cradle Mountain -41.6297628 145.9487202 870 Moss Dry eucalypt forest and woodland SS14 Central Highlands Cradle Mountain -41.59612141 145.9342721 800 Moss Moorland, sedgeland, rushland and peatland SS15 West Coast Anthony Road -41.8786241 145.612585 530 Moss Moorland, sedgeland, rushland and peatland SS16 West Coast Anthony Road -41.9505829 145.5460319 540 Moss Moorland, sedgeland, rushland and peatland SS17 West Coast Lyell Highway -42.12458172 145.8340678 420 Sphagnum Moorland, sedgeland, rushland and peatland SS18 West Coast Lyell Highway -42.21033898 145.9736369 400 Moss Moorland, sedgeland, rushland and peatland SS19 West Coast Lyell Highway -42.2160537 146.0197213 420 Moss Wet eucalypt forest and woodland SS20 Central Highlands Lake St. Clair -42.1178224 146.1358916 1060 Sphagnum Wet eucalypt forest and woodland SS21 Central Highlands Lake St. Clair -42.0936565 146.114189 990 Moss Moorland, sedgeland, rushland and peatland SS22 Central Highlands Lake St. Clair -42.04295167 146.1373087 850 Moss Rainforest and related scrub SS23 Central Highlands Lake St. Clair -42.033753 146.129543 740 Moss Rainforest and related scrub SS24 Central Highlands Lake St. Clair -42.1156107 146.1687046 770 Moss Native grassland SS25 Central Highlands Lake St. Clair -42.1092421 146.1619722 770 Moss Rainforest and related scrub SS26 Central Highlands Lake St. Clair -42.1225668 146.1954308 760 Sphagnum Dry eucalypt forest and woodland SS27 Central Highlands Lake St. Clair -42.12097938 146.2076087 760 Moss Dry eucalypt forest and woodland Table 5-1. List of sites used for the vegetation surveys and surface sample collection.

To incorporate the vegetation data into the models, multiple quadrat data from the same vegetation type (according to TasVeg 3.0 - Government of Tasmania, 2013) were combined to create local (0-10 m), extra-local (10-100 m) and regional (100-2000 m) averages of plant abundance in each site for the key taxa in this study. Since the Lagrangian Stochastic Model assumes a greater dispersal distance compared to the Gaussian Plume Model due to atmospheric turbulence, vegetation cover percentages were also extracted for rings from 2 km up to 50 km from the sampling points. Abundances of the vegetation types for each ring at each site were extracted from TasVeg 3.0 using ArcMap (ESRI). The resulting vegetation cover percentages for each ring were then converted into plant abundances for the 13 selected taxa by multiplying the percentage value of each vegetation type by the plant abundance averages obtained by merging the plant cover percentages from multiple quadrats belonging to the same vegetation category.

113 |

5.4.2. Pollen analysis

All the surface samples were processed using the standard protocols (Faegri and Iversen, 1989) and at least 300 pollen grains of the 13 selected taxa were counted for each sample. In order to estimate both PPEs and RSAP, the pollen settling velocity (m/s) is needed. Pollen sedimentation velocities may either be measured directly or estimated using Stoke's Law (Chamberlain, 1975) which predicts the settling velocity of smooth spheres of a deﬁned density with diameters between about 1 and 70 mm

114 |

(i.e. the size of most anemophilous pollen). Despite the fact that some pollen grains are not smooth and spherical (Faegri and Iversen, 1989), Stoke's Law is believed to be an accurate estimator of pollen settling rate (Gregory, 1978). For non-spherical pollen grains, the major (a) and the minor (b) axes were measured and the volume of the ellipsoid converted into the radius of an equivalent sphere in order to apply Stoke's law.

Accounting for taphonomic effects on pollen size is important, therefore botanical reference material was not used in this study. Size measurements to calculate the pollen fall speeds were taken on at least 30 randomly chosen pollen grains for each target taxon using modern pollen samples from three different sampling areas (e.g. Cradle Mountain, Lake St. Clair and West Coast Range). Identiﬁcation of pollen grains relied on the Australasian Pollen and Spore Atlas (http://apsa.anu.edu.au/) and a reference collection (Macphail, unpublished). For the taxa identiﬁed at a Genus or Family level (Eucalyptus and Cupressaceae), the mixture of pollen grains found in the pollen samples from different sites has been considered to take into account most of the variability (e.g. morphological and size differences) in these pollen grains within the region.

5.4.3. Dispersal models

Dispersal model selection greatly impacts on the resulting pollen productivity estimates (Theuerkauf et al., 2013). To date, the most commonly used dispersal model is the Gaussian Plume Model, but more recent approaches attempt to improve this methodology by using mechanistic models, including the Lagrangian Stochastic Model (Andersen, 1991; Kuparinen et al., 2007). In this study, the Gaussian Plume and Lagrangian Stochastic models were tested on the pollen and vegetation data collected in western Tasmania in order to assess the suitability of the models' assumptions and the reliability of the PPEs produced by both models in a context where zoophilous plants are abundant. Mazier et al. (2012) advised the exclusion of strictly zoophilous taxa from reconstructions, since the models assume

115 | that all pollen grains are airborne. Nevertheless, many important Australian plant taxa are animal-pollinated and their pollen is well-documented in the fossil record (e.g. Eucalyptus, Leptospermum, Acacia), probably because they are partly wind- dispersed (as is European Tilia; Mazier et al., 2012). Given the ecological importance of such taxa and their abundance in pollen records, we consider their exclusion would prevent robust and meaningful quantitative regional vegetation reconstructions in Australia.

When vegetation abundance is properly evaluated (considering inter-taxonomic differences in pollen dispersal and basin size and type), pollen loadings of individual taxa at similarly-sized sites are linearly related to the plant abundances of those taxa surrounding study sites; the slope of the relationship represents the pollen productivity estimate for individual taxa (Sugita, 1994). PPEs are usually expressed as a dimensionless ratio relative to a reference taxon, termed the Relative Pollen Productivity (e.g. Davis, 1963; Andersen, 1970; Brostrom et al., 2008). According to Bunting et al. (2013) the reference taxon should be a medium pollen producer abundantly represented in both pollen and vegetation PPEs. In addition, differences in pollen productivity of plant species are also one of the major factors influencing pollen productivity estimates, as few pollen taxa can be identified to the species level and a pollen taxon may therefore correspond to more than one plant species. While the allocation of a taxon-specific PPE is necessary to calibrate pollen data against vegetation data, it is perhaps deserving of more critical attention. An ideal reference taxon should be relatively abundant in both the plant community and pollen assemblage. For example, Poaceae possesses these characteristics and has been chosen as a reference taxon in many European studies. However, Poaceae has the clear disadvantage that it is a Family taxon that includes plant species likely to have a large variety of pollen productivity and dispersal characteristics, including wetland species. This has prompted some researchers to choose other reference taxa, such as Quercus (Bunting et al., 2005; Li et al., 2015), Juniperus (Sugita et al., 1999) and Cerealia (Nielsen and Odgaard, 2005). We have chosen Eucalyptus as the pollen reference in our study because 1) its pollen can be found in all modern moss samples and the corresponding vegetation quadrats and 2) it has been found to be well-

116 |

represented in the pollen spectra according to Fletcher and Thomas (2007), and can be considered a medium pollen producer.

One of the first approaches introduced to estimate pollen productivity is the Extended R-Value (ERV) method (Prentice and Parsons, 1983), which was also tested in this study (Supplementary Information 5-S1-S3). The main assumption of the ERV method is that the background pollen component is the same for all the samples included in the analysis; therefore similarly-sized sites are preferably to be selected within a homogeneous landscape. Nevertheless, with heterogeneous and patchy vegetation, the ERV method can still produce reliable results, as long as major gradients in species composition and vegetation structure do not exist (Sugita, 1994). We found that the regional vegetation in our sampling locations is not homogenous, has a rather diverse structure and the pollen composition is variable, which likely affects the performance of this method in our study region. In contrast to the ERV method, the application of the Gaussian Plume Model in this study incorporates site- speciﬁc regional vegetation. Further explanations of the ERV method and results are given in the Supplementary Information 5-S1-S3.

Over the past few years, the Lagrangian Stochastic Model (LSM) has frequently been suggested as a way of producing realistic descriptions of atmospheric airﬂows and particle dispersal (Soons et al., 2004; Jarosz et al., 2004). A previous test on lakes located in NE Germany has proven that the LSM of Kuparinen et al. (2007) better describes observed pollen deposition than the Gaussian Plume Model (Theuerkauf et al., 2013). Furthermore, fall speed of pollen has far less impact on the dispersal pattern in the LSM compared to the GPM (Theuerkauf et al., 2013).

We calculated pollen productivity estimates from both dispersal models using optimization with the DeOptim package in R (Mullen et al., 2009). Optimization is applied upon simulations of pollen deposition at each site. Pollen deposition is calculated by multiplying distance weighted plant abundances (DWPA) for each taxon at each site with a respective PPE value. Optimization starts with a set of random PPEs and then searches for the set of PPEs which gives the most similar simulated pollen deposition at each site compared to the empirical pollen

117 | deposition. This method does not take into account pollen background component, which was considered negligible across the region, as shown from preliminary studies (y-intercepts in Figure 5-S2).

DWPA is calculated by multiplying the cover of each taxon in each ring with a respective, taxon speciﬁc weighting factor. The weighting factors are calculated based on the symmetry argument (Chamberlain, 1967): in uniform vegetation the number of pollen grains travelling a given distance to reach a point must equal the number of grains travelling the same distance away from the point. For a given ring (e.g. 100-200 m radius) the weighting factor is then calculated as the proportion of pollen released at a source that is deposited within that ring around the source, i.e. between 100 and 200 m distance from the source. Weighting factors are calculated for each taxon with its speciﬁc fall speed of pollen separately. Calculations used the original formula for the Gaussian Plume Model (Sutton, 1953) and a lookup function from the R package DISQOVER for the LSM (http://disqover.botanik.uni- greifswald.de/the-launch/; Theuerkauf et al., 2016). Optimization then seeks greatest similarity between the simulated and empirical pollen data. Differences between simulated and empirical pollen data are calculated as the weighted least square distance and the sum is minimised by the DeOptim function.

To explore the suitable size of the sampling area, calculations start with DWPA calculated from only the innermost ring and then successively include further rings during each run. We tested the performance of the GPM and LSM using three different maximum radii: 2, 10 and 50 km. The DeOptim function also produces Optimization Functions Scores (OFS), similar to the ERV method (Likelihood Function Scores-LFS), which allow the identiﬁcation of the relevant source area of pollen. In practice, decreasing OFS indicates a better ﬁt between simulated and empirical pollen data. In order to produce error estimates of pollen productivity estimates, calculations were repeated 100 times.

118 |

5.5. Results

5.5.1. Vegetation surveys

A total of 265 vegetation quadrats were surveyed around the 27 surface samples collected during the ﬁeld campaign (Figure 5-1, Table 5-1). The percent cover of the 13 key taxa in all quadrats was merged based on vegetation types, according to the information available in TasVeg 3.0 (Government of Tasmania, 2013) and the resulting averages for the target taxa of this study are shown in Supplementary Table 5-S1.

5.5.2. Surface samples pollen analysis and fall speeds

Modern surface samples were used to determine the size of the pollen grains and relative fall speeds. The fall speeds calculated by applying Stoke's Law (Gregory, 1978) to measurements of 390 pollen grains from the 13 target taxa range between 0.002 and 0.072 m/s (Table 5-2). As expected, the largest pollen grains, such as Nothofagus cunninghamii and Gymnoschoenus sphaerocephalus yielded the highest fall speeds, whereas the smallest values were obtained for Eucryphia lucida and Bauera rubioides, which are the smallest pollen grains in the Tasmanian pollen ﬂora (<15 mm).

TAXON A-axis (mean μm) B-axis (mean μm) FALL SPEED (m/s) Eucalyptus spp. 23.772 11.518 0.008 Bauera rubioides 14.014 10.496 0.004 Cupressaceae 31.787 0.031 Ericaceae 37.809 0.043 Eucryphia lucida 8.933 8.222 0.002 Gymnoschoenus sphaerocephalus 48.736 0.072 Leptospermum spp. 19.604 8.590 0.005 Monotoca spp. 26.786 18.348 0.015 Nothofagus cunninghamii 46.966 26.325 0.037 Nothofagus gunnii 31.858 20.481 0.020 Phyllocladus aspleniifolius 36.766 23.094 0.026 Poaceae 31.613 0.030 Sprengelia spp. 29.192 0.026

Table 5-2. Pollen measurements and estimated pollen fall speeds for the 13 target taxa in western Tasmania. The B-axis is not listed for spherical pollen grains.

119 |

From the GPM, the only pollen productivity estimate higher than Eucalyptus (reference taxon, PPE ¼ 1) was found for Eucryphia lucida, whereas the PPEs for all the other taxa are substantially lower (Table 5-3, Figure 5-4). The lowest values were found for Ericaceae and G. sphaerocephalus. Estimated errors are in general relatively low, with highest errors for E. lucida, Leptospermum, and Cupressaceae.

The application of Lagrangian Stochastic simulations produced substantially different results compared to the Gaussian Plume Model, especially for some taxa (e.g. E. lucida, N. cunninghamii) (Table 5-3, Figure 5-4). Considering the results from 2 km radius, PPEs relative to Eucalyptus are higher for Cupressaceae and N. cunninghamii. In line with the results from the GPM, the lowest pollen productivity estimates were obtained for Ericaceae and G. sphaerocephalus.

Pollen productivities estimated using 10 km radii produced quite distinct values within both GPM and LSM estimate sets, with substantial differences for some taxa (e.g. Phyllocladus aspleniifolius, Cupressaceae, Sprengelia) (Figure 5-4). Comparison of all the results from 2, 10 and 50 km are shown in Figure 5-S3. In general, the GPM clearly over-estimates pollen productivities for small grains with low fall speed (e.g. E. lucida) and under-estimates it for larger pollen grains with high fall speed (e.g. G. sphaerocephalus) (Figure 5-S5).

5.5.3. Relevant source area of pollen (RSAP)

RSAP estimates using the Gaussian Plume Model and Lagrangian Stochastic Model are virtually identical considering a vegetation sampling radius of 2 and 10 km radius (Figure 5-3). Likewise, the Optimization Function Scores (OFS) are also very similar when comparing results using 50 km radius (Figure 5-S3). From the analysis using a radius of 2 km, both GPM and LSM suggest an RSAP of 1 km (Figure 5-3a), which is closely related to the ERV results (Figure 5-S3). Extending the analysis to 10 km shows that function scores further decline (i.e. increase of the goodness of ﬁt between estimated and empirical pollen data) to reach their optimum only at around 10 km (Figure 5-3b).

120 |

5.5.4. Pollen productivity estimates (PPEs)

In contrast with the similarities obtained for RSAP estimates, the application of the Gaussian Plume Model and Lagrangian Stochastic Model produced signiﬁcantly different sets of PPEs, also high-lighting a signiﬁcantly different performance when comparing results from 2 km to 10 km (Table 5-3, Figure 5-4). Major differences are evident for the smallest (e.g. E. lucida) and the largest pollen grains (e.g. N. cunnighamii, G. sphaerocephalus).

121 |

122 |

GPM 2km GPM 2km SD GPM 10km GPM 10km SD LSM 2km LSM 2km SD LSM 10km LSM 10km SD Bauera rubioides z 0.137 0.024 0.165 0.025 0.049 0.010 0.067 0.011 Cupressaceaea 0.800 0.110 1.003 0.124 1.115 0.178 1.337 0.198 Ericaceaeaz 0.037 0.004 0.043 0.004 0.057 0.008 0.067 0.007 Eucalyptus spp.z 1.000 0.000 1.000 0.000 1.000 0.000 1.000 0.000 Eucryphia lucida z 1.888 0.224 1.459 0.113 0.440 0.070 0.340 0.043 Gymnoschoenus sphaerocephalus a 0.020 0.002 0.027 0.003 0.037 0.004 0.050 0.006 Leptospermum spp.z 0.244 0.027 0.284 0.027 0.110 0.014 0.145 0.016 Monotoca spp.z 0.387 0.057 0.424 0.049 0.439 0.068 0.484 0.060 Nothofagus cunninghamii a 0.847 0.064 0.830 0.055 1.521 0.121 1.281 0.084 Nothofagus gunnii a 0.180 0.025 0.251 0.029 0.271 0.037 0.385 0.033 Phyllocladus aspleniifolius a 0.439 0.097 0.631 0.099 0.476 0.143 0.842 0.163 Poaceaea 0.218 0.019 0.242 0.019 0.407 0.032 0.414 0.031 z Sprengelia spp. 0.078 0.009 0.135 0.016 0.097 0.014 0.209 0.024

5.6. Discussion

Our results reveal that the selection of pollen dispersal model has a large impact on pollen productivity estimates in Tasmania's mixed wind- and animal-pollinating flora. Importantly, the ability of the Lagrangian Stochastic Model (LSM) to incorporate a range of complex physical features that affects the dispersal process gives more realistic estimates than the Gaussian Plume Model. LSM predicts the trajectory of each dispersing particle under turbulent conditions, which depend on the degree of atmospheric (in)stability and the vertical structure of the atmospheric boundary layer. Intuitively, atmospheric conditions may have larger impact on pollen with low fall speed than on pollen with high fall speed. However, previous analyses have shown that dispersal of pollen with low fall speed is hardly affected by atmospheric conditions as its falling velocity is typically lower than average vertical turbulent flows (Kuparinen et al., 2007). In contrast, dispersal of “heavier” pollen grains depends on strong turbulent flows that are capable of carrying such pollen across longer distances (Kuparinen et al., 2007). Pollen is primarily released under unstable atmospheric conditions with strong turbulent flows (Jackson and Lyford, 1999), implying that differences in pollen dispersal are largely independent

123 |

of fall speed. Therefore, the LSM is preferable in modelling the dispersal of pollen grains over both short and long distances (Theuerkauf et al., 2013, 2016) and our study conﬁrms this view.

5.6.1. Relevant source area of pollen (RSAP)

The size of the relevant source area of pollen is mostly influenced by spatial patterns of vegetation, such as size, structure and distribution of plant communities and the size of the basin used for the analyses (Sugita, 1994; Bunting et al., 2004; Brostrom et al., 2005; Nielsen and Sugita, 2005). These factors largely explain the variability of results obtained in the previous studies around the globe (Sugita et al., 1999; Rasanen et al., 2007; von Stedingk et al., 2008; Mazier et al., 2008; Duffin and Bunting, 2008; Li et al., 2015). Regarding sample selection, even when only moss cushions are used, RSAP estimates can vary between different bioclimatic regions with distinct vegetation patterns. For instance, an RSAP of 1000 m was estimated from a pine and birch forest tundra in Finland (Rasanen et al., 2007); a 500 m RSAP was obtained from a spruce and birch forest tundra in west-central Sweden (von Stedingk et al., 2008); a 300 m RSAP was estimated in open pasture woodland in Switzerland (Mazier et al., 2008); a RSAP of 600e900 m was estimated in South African savannas (Duffin and Bunting, 2008) and a RSAP of 2000-2500 m has been obtained in Northern China (Li et al., 2015).

Regarding atmospheric conditions, the effect of wind speed has been excluded as an important factor in previous simulation studies (Nielsen and Sugita, 2005), but atmospheric turbulence has been found to play a major role in determining the pollen source area from lakes (Theuerkauf et al., 2013). Another important factor in determining the RSAP is the dispersal model used in the calculations. By applying dispersal models simulations from lakes in Europe, it has been found that the GPM for neutral conditions predicts that only 15-35% of the total deposition would arrive from distance greater 10 km, whereas the LSM predicts that 50-60% of the total deposition arrives from such greater distances (Theuerkauf et al., 2013).

124 |

Interestingly, here we observe no significant difference in RSAP between the two models (Figure 5-3). This result is in contrast with previous findings in NE Germany and it may be due to the use of moss/Sphagnum polsters instead of lake surface sediments combined with a substantially different vegetation structure and pollination vectors. Our results from the 2 km-radius analysis fall within the range of the previously estimated RSAP values, being especially close, for instance, to the RSAP size estimated in South African savannas (Duffin and Bunting, 2008), where, similarly to Tasmania, the vegetation features a considerable number of animal- pollinated plants (e.g. Acacia). The tests of GPM and LSM using further distances exhibit an optimum of the goodness of fit between modelled and empirical pollen data at around 10 km, which greatly extends the relevant source area of pollen. This relates to the fact that vegetation composition even in the largest rings samples (48- 50 km) differs, i.e. regional vegetation is not uniform.

5.6.2. Pollen productivity estimates (PPEs)

The main factors affecting the RSAP outlined above are known to also strongly influence PPEs (Brostrom et al., 2005; Nielsen and Sugita, 2005; Bunting et al., 2013; Theuerkauf et al., 2013). Pollen dispersal model selection has been found to influence PPEs estimates substantially when applied to the same dataset (Theuerkauf et al., 2013). The study here presented confirms the observations by Theuerkauf et al. (2013), who found that PPEs obtained using Lagrangian Stochastic simulations are more realistic than the PPEs obtained using the GPM under neutral atmospheric conditions. For instance, with the GPM we obtained the highest PPE for Eucryphia lucida, which is generally assumed to be an entomophilous taxon (e.g. it is the basis of the honey industry in Tasmania) and was previously found to be under- represented in Tasmanian pollen spectra (Fletcher and Thomas, 2007). In contrast, assumed anemophilous plants, such as Nothofagus cunninghamii, which is regarded as over-represented by Fletcher and Thomas (2007), are estimated to be low pollen producers by the GPM (Figure 5-4, Table 5-3). Surprisingly, by comparing estimates from different source radii (2, 10 and 50 km), some taxa have shown to

125 |

perform in a substantially different way. This is the case of P. aspleniifolius, Cupres- saceae and Sprengelia, which show higher PPEs with an increasing radius of analysis. When comparing plant cover abundance across all rings and sites, these taxa have the lowest standard deviations (Figure 5-S5). This means that they occur at all the sites in rather similar abundances (i.e. short abundance gradients) a fact that makes the analysis more sensitive and likely requiring more investigation. To address this issue, a closer look at the Optimization Function Scores in Figure 5-3b allows us to identify an optimum at 10 km (Figure 5-3b), suggesting that PPEs at this distance are likely more reliable than the ones obtained at 2 km. We believe that the most reliable pollen productivity estimates for P. aspleniifolius, Cupressaceae and Sprengelia are obtained using a vegetation radius of 10 km. Indeed, these three taxa are assumed to be wind-pollinated and over-represented in the pollen records (Fletcher and Thomas, 2007). Across the two distances, both models generally agree on the taxa with the lowest PPEs, such as Ericaceae and Gymnoschoenus sphaerocephalus, which were previously found to be strongly under-represented in the pollen spectra (Fletcher and Thomas, 2007).

The zoophilous species used in this study can be compared to the cases of Tilia in Europe (Brostrom et al., 2008) and Acacia in South Africa (Dufﬁn and Bunting, 2008), which have been found to have relatively low pollen productivity estimates. Similarly, all the assumed animal-pollinated plant taxa in this study are characterised by low PPEs with a general agreement of the two models, except for Eucryphia lucida in the GPM results (Figure 5-4, Table 5-3).

Wind dispersal of animal-pollinated plant taxa occurs frequently (Paw and Hotton, 1989; Andersen, 1970; Kershaw and Strickland, 1990; Dufﬁn and Bunting, 2008; Mazier et al., 2012) and this can be effectively modelled, according to the ﬁndings presented here. Given the low errors and the coherence between the two models, we consider pollen productivity estimates for zoophilous taxa obtained from this study suitable for the implementation in pollen-based quantitative vegetation reconstructions in western Tasmania. Further improvements of the estimated pol-len productivities could be achieved with additional vegetation surveys across the

126 |

vegetation types present in the region (e.g. TasVeg 3.0; Government of Tasmania, 2013).

5.6.3. Potential applications in Australian palaeoecology

Our results represent the first attempt to quantify pollen productivity and dispersal for Australian plant species. We provide a quantitative correction to the inherent biases in pollen data that will enable realistic estimates of vegetation from this predominantly animal-pollinated landscape-scale vegetation mosaic. Our results quantitatively support previous (semi-quantitative) studies on the representation of pollen types in western Tasmania (Macphail, 1979; Fletcher and Thomas, 2010b) and pave the way for an objective test of competing hypotheses on the evolution of this intriguing landscape. The ability to understand actual vegetation response to past changes in climate and disturbances will allow the development of realistic management approaches to conservation efforts under future climate change projections. Given predictions of increasing forest fires (Moritz et al., 2012), it is important that we gain an understanding of how fire drives and shapes ecosystems through time so that we may better predict and manage the effects of fire on threatened Australian ecosystems. Importantly, our models will be critical in estimating the impact of fire and the potential fate of critically endangered ecosystems in Tasmania, which, despite decades of research and the use of cutting- edge technologies (e.g. Yospin et al., 2015), we are still unable to fully comprehend. Understanding how these ecosystems have responded to past periods of climatic and fire regime change will enable better decision-making about their future management.

In Australia, fire is particularly pervasive in shaping vegetation patterns (Bradstock et al., 2002) and it has been implicated as the most critical factor controlling the fragmentary distribution of fire-sensitive rainforest within considerable tracts of flammable sclerophyll vegetation from the monsoonal north to the temperate south (Bowman, 2000). Tasmania is an exemplar of the role of fire in shaping vegetation

127 |

landscapes in Australia, with vast areas in which rainforest is the climatic potential vegetation but are occupied instead by fire-promoted vegetation. In western Tasmania, inter-annual precipitation variability and fire activity are driven by changes in the main climate mode of the Southern Hemisphere: the Southern Annular Mode (SAM) (Risbey et al., 2009; Mariani and Fletcher, 2016). This climate mode is experiencing significant change due to anthropogenic activity (Mariani and Fletcher, 2016), placing the existing pockets of fire-sensitive vegetation under potential threat. Under a changing climate regime in which fires are predicted to increase in frequency and magnitude (McWethy et al., 2013; Power et al., 2013; Moritz et al., 2012), it is important to use all available information on how fire- sensitive ecosystems respond to changes in fire activity. By quantifying the impact of past fire activity on fire-sensitive ecosystems, informed conservation and management plans may be developed.

With this ﬁnal projection, in this study we estimated, for the ﬁrst time in Australia, quantitative pollen productivity estimates and pollen source area in order be able to apply quantitative vegetation reconstructions algorithms (LRA- Sugita, 2007a,b) to the numerous fossil pollen records from across western Tasmania.

128 |

5.7. Conclusion

The results obtained in this study are the first quantitative estimate of pollen productivity in Australia and represent a significant advancement for palaeoecology in this region as they set the basis for quantitative vegetation reconstruction. The final product of this innovative technique will be the incorporation of fire-driven vegetation changes into global climate models, as vegetation affects climate through biogeochemical and biogeophysical feedbacks.

Since pollen dispersal models have been developed basing on wind-pollinated plant taxa, the application of these models in Australia is challenging due to a large representation of animal-pollinated plant taxa. To overcome this problem we tested two pollen dispersal models in order to achieve pollen productivity estimates for thirteen plant taxa in Tasmania. Our ﬁndings show a better performance of the Lagrangian Stochastic Model compared to the Gaussian Plume Model in the animal- pollinated vegetation mosaics of western Tasmania. Indeed, Pollen Productivity Estimates obtained with this model are more realistic when compared with previous pollen-vegetation relationship studies and the assumed pollen dispersal of the taxa. We thus suggest the implementation of these results in future applications of quantitative palaeovegetation reconstruction in the western Tasmanian region. We recommend that these results only to be applied to Holocene records, given potential changes in pollen production under glacial periods. Further work should aim to spatially extend these analyses to obtain pollen productivity estimates for mainland Australia's major pollen taxa, allowing the realization of quantitative vegetation reconstructions to quantify land-cover changes through time at a regional/continental scale.

Since the results from the two pollen dispersal models are considerably different for some key plant taxa in our case-study, there are also implications for future work elsewhere. Indeed, the Gaussian Plume Model is the most widely used pollen dispersal model in Europe and current quantiﬁcations of past vegetation cover are based on this model. We suggest that future work should incorporate a test of both

129 |

pollen dispersal models, as the choice of pollen productivity estimates may result in substantial differences when estimating past vegetation cover.

5.8. Supplementary results and information

The figures and tables displayed below are part of the supporting information for the paper ‘Testing quantitative pollen dispersal models in animal-pollinated vegetation mosaics: An example from temperate Tasmania, Australia’ (Mariani et al., 2016).

130 |

Table 5-S1 - List of the TasVeg 3.0 vegetation types surveyed and target taxa abundances (% cover). 0.1 0.0 0.0 1.0 0.0 0.0 0.9 0.0 4.6 0.0 0.4 3.2 0.5 0.0 0.0 0.7 0.0 0.0 0.3 0.3 0.0 0.6 0.0 0.0 0.0 0.0 0.0 2.0 1.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.7 0.0 5.0 0.0 0.0 21.7 10.0 Sprengelia 0.1 0.0 0.0 1.0 0.7 0.1 0.7 4.0 0.0 0.3 1.5 1.0 0.0 1.7 0.0 0.0 0.0 0.0 0.0 0.4 0.0 1.0 0.0 2.5 5.0 0.3 1.5 1.0 0.0 9.0 0.0 5.0 1.3 1.3 6.9 1.0 0.3 1.0 75.0 75.0 40.0 53.8 14.3 80.0 50.0 12.5 Poaceae 4.6 1.0 1.2 1.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.5 0.0 0.0 0.0 1.0 0.1 0.3 0.3 0.0 0.0 2.5 5.8 0.0 0.0 2.0 0.0 1.5 1.0 0.0 0.0 0.0 0.6 0.0 0.0 0.0 0.0 0.0 1.0 0.7 0.0 10.0 10.0 11.1 Phyllocladus 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 1.5 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 42.9 60.0 60.0 N.gunnii 6.1 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.6 0.0 0.0 0.0 0.0 0.0 0.7 8.7 0.0 0.0 0.0 5.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 8.6 0.0 0.0 0.0 0.0 0.0 15.0 27.0 10.0 15.0 21.8 80.0 75.0 33.6 61.7 80.0 N.cunninghamii 0.0 0.5 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 5.0 1.0 0.0 0.0 0.0 1.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.9 0.0 0.0 0.0 0.0 0.8 0.0 1.1 0.0 0.0 0.0 11.2 15.0 15.0 25.0 10.0 Monotoca 0.0 0.0 2.4 0.0 0.0 8.4 0.0 0.0 6.4 2.2 3.5 0.0 9.5 0.0 0.0 0.0 0.0 9.0 5.0 1.0 0.0 7.3 0.0 3.5 0.5 0.0 0.0 0.0 5.2 0.0 0.3 0.0 15.0 15.0 10.0 10.0 31.5 31.5 15.0 60.0 15.3 40.0 39.3 27.5 14.4 85.0 Leptospermum 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.5 2.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.6 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 35.7 34.4 36.8 41.7 45.8 19.2 35.0 45.0 19.3 Gymnoschoenus 0.0 0.0 0.0 3.6 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 2.5 2.5 5.0 0.0 0.0 0.0 5.3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 45.0 11.0 60.0 17.9 Eucryphia 0.1 2.5 5.0 5.0 0.0 0.0 0.0 0.4 0.0 0.1 1.4 0.2 0.0 2.5 0.0 2.0 5.0 7.9 1.0 3.5 0.1 5.0 0.0 2.5 0.0 2.7 0.0 0.0 0.3 5.0 10.2 26.5 26.5 15.0 20.0 16.2 12.5 35.0 30.0 80.0 20.0 25.0 40.0 45.0 14.6 32.5 Eucalyptus 1.0 4.6 1.0 0.4 5.0 8.0 0.9 1.0 3.5 0.0 0.0 7.0 0.8 0.8 1.0 0.0 3.0 0.0 1.0 5.0 0.0 1.5 0.0 1.0 9.8 2.0 2.2 9.7 5.0 27.6 14.0 28.7 21.0 17.1 27.9 39.8 22.8 25.0 72.5 81.0 21.0 13.3 32.5 11.0 16.0 12.3 Ericaceae 1.0 2.4 0.0 5.0 0.1 0.0 0.0 0.0 0.0 1.4 0.0 0.0 0.0 2.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.0 0.0 0.2 0.0 5.0 5.0 0.0 10.2 10.0 Cupressaceae 1.0 0.0 2.4 0.0 0.0 1.1 0.0 3.5 0.0 9.2 0.0 0.0 0.0 2.0 0.0 0.0 9.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.1 5.4 0.0 0.0 0.0 0.0 12.5 50.0 15.0 32.8 32.8 10.0 90.0 60.0 15.0 10.0 Bauera rubioides Bauera (RKP) Athrotaxis selaginoides rainforest selaginoides Athrotaxis (RKP) (RMS) Nothofagus - Phyllocladus short rainforest Phyllocladus-short Nothofagus (RMS) (RKS) Athrotaxis selaginoides subalpine scrub (RKF) Athrotaxis selaginoides - Nothofagus gunnii short rainforest selaginoidesgunniiAthrotaxisshort (RKF) Nothofagus - (RFS) gunnii Nothofagus scrub and rainforest (MRR) Restionaceae rushland (MRR) (NAD) Acacia dealbata forest (MGH) Highland grassy sedgeland grassy Highland (MGH) (MBW) Western buttongrass moorland buttongrass Western (MBW) Subalpine(MDS) latifolia Diplarrena rushland (MBE) Eastern buttongrass moorland buttongrass Eastern (MBE) with moorland emergent shrubs Buttongrass (MBS) (undifferentiated) moorland Buttongrass (MBU) (HSE) Eastern alpine Eastern sedgeland(HSE) (HHE) Eastern alpine Eastern heathland (HHE) (HCH) Alpine(HCH) coniferous heathland (WSU) Eucalyptus subcrenulata forest and woodland forest and subcrenulata Eucalyptus (WSU) (WOR) Eucalyptus obliqua forest over Eucalyptus rainforest(WOR) (WDL) Eucalyptus delegatensis Eucalyptus (WDL) forest over Leptospermum nitida forest over Leptospermum Eucalyptus (WNL) nitida forest over(WNR) rainforest Eucalyptus (WOU) Eucalyptus obliqua wet Eucalyptus forest (undifferentiated)(WOU) (WDU) Eucalyptus delegatensis Eucalyptus wet(WDU) forest (undifferentiated) (WDR) Eucalyptus delegatensis Eucalyptus (WDR) forest over rainforest (GPH) Highland Poa grassland (FPL) Plantations for silviculture for Plantations (FPL) (WDB) Eucalyptus delegatensis shrubs broad-leafwithforest Eucalyptus (WDB) (DPD) Eucalyptus pauciflora forest and woodland forest dolerite on and pauciflora (DPD) Eucalyptus land Agricultural (FAG) (WDA) Eucalyptus dalrympleana forest (WDA) Eucalyptus (DNI) Eucalyptus nitida woodland forest(DNI) and dry Eucalyptus (SWW) Western wet scrub (DGW) Eucalyptus gunnii woodland Eucalyptus (DGW) (SLL) Leptospermum lanigerum Leptospermum scrub (SLL) (SMM) Melaleuca squamea heathland (RMT) Nothofagus - Atherosperma rainforest Atherosperma - Nothofagus (RMT) (DDE) Eucalyptus delegatensis Eucalyptus woodland forest(DDE) and dry (RMU) Nothofagus rainforest (undifferentiated) rainforest Nothofagus (RMU) (AHL) Lacustrine herbland Lacustrine (AHL) woodland forest dolerite on and amygdalina (DAD) Eucalyptus woodland forest and coccifera (DCO) Eucalyptus DESCRIPTION (RSH) Highland low Highland (RSH) scrub and rainforest (RPF) Athrotaxis cupressoides - Nothofagus gunnii short rainforest gunniishort Nothofagus -cupressoides (RPF)Athrotaxis (RPP) Athrotaxis cupressoides rainforest (SHS) Subalpine heathland (RPW) Athrotaxis cupressoides open woodland RKP RMS RKS RKF RFS MRR NAD MGH MBW MDS MBE MBS MBU HSE HHE HCH WSU WOR WDL WNL WNR WOU WDU WDR GPH FPL WDB DPD FAG WDA DNI SWW DGW SLL SMM RMT DDE RMU AHL DAD DCO TASVEG 3.0 CODE RSH RPF RPP SHS RPW

131 |

5.8.1. Additional details on the ERV method for PPEs and RSAP calculations

The ERV (Extended R-Value) method (Prentice and Parsons, 1983) is the most commonly used method to estimate PPEs and RSAP in Europe (e.g. Bröstrom et al.,

2004; Sugita, 2007b). The main assumption of the ERV method is that the background pollen component is the same for all the samples included in the analysis, therefore a set of sample points all need to be collected within an essentially homogeneous landscape region. Nevertheless, we found that the regional vegetation in our sampling locations is not homogenous and this likely creates an adequate performance of the ERV method in our study region (see further). Thus, we argue that this methodology commonly used elsewhere is not applicable in our case study, as the results are not reliable.

The ERV program calculates ‘‘likelihood function scores’’ (LFS) (Prentice and Webb

III, 1986; Sugita, 1994), which are plotted as a function of distance from the sampling point. The spatial patterns of vegetation, such as patchiness and heterogeneity, are among the factors affecting the size of the RSAP and the changes in the likelihood function scores (Sugita, 1994). The smaller the LFS, the better the fit of the pollen– vegetation data to the model. In general the LFS decreases with the increasing range of vegetation survey and the curve will eventually reach an asymptote. This asymptote represents the ‘‘relevant distance’’, or the radius of the RSAP (Sugita,

1994, 1998, 2007b; Sugita et al. 1999). Three ERV submodels have been developed

(Parsons and Prentice, 1981; Prentice and Parsons, 1983; Sugita 1994) and will be tested in this study. Relative plant cover is a recalculated value to sum up the

132 | coverage of all taxa selected for PPEs as 100% in submodel1 and submodel2, whereas absolute plant cover (%) is used for sub-model 3. For all the ERV model runs we used a wind speed of 6.5 m/s, which was obtained from the average value at the closest meteorological station to the majority of sites (Cradle Valley,

Australian Bureau of Meteorology - BOM). We tested here the performance of the

ERV using a 2 km and 10 km radius.

133 |

Figure 5-S1. a) RSAP results obtained considering a 2 km vegetation radius. The likelihood function scores produced by the ERV-submodel 1 sharply decrease closer to the sampling point (~500 m) compared to the values obtained using the ERV-submodel 2 and ERV- submodel 3, which both have yielded a very similar LFS curve with estimated RSAP at ~900 m. b) RSAP results obtained considering a 10 km vegetation radius. The introduction of the regional vegetation into the dataset produced a sharp shift in the LFS starting from 4 km, suggesting a worse fit between the modelled vegetation and pollen data, likely due to a lack of a uniform distant vegetation.

134 |

Table 5-S2. PPEs from ERV submodels and standard errors.

ERV1 ERV1 SE ERV2 ERV2 SE ERV3 ERV3 SE Bauera rubioides 0.310 0.035 0.268 0.012 0.252 0.079 Cupressaceae 0.975 0.099 0.619 0.083 0.258 0.065 Ericaceae 0.035 0.003 0.031 0.001 0.027 0.006 Eucalyptus spp. 1.000 0.000 1.000 0.000 1.000 0.000 Eucryphia lucida 2.662 0.109 2.731 0.114 2.399 0.373 Gymnoschoenus sphaerocephalus 0.039 0.003 0.037 0.001 0.034 0.010 Leptospermum spp. 0.399 0.050 0.344 0.093 0.443 0.162 Monotoca spp. 0.257 0.056 0.286 0.053 0.369 0.021 Nothofagus cunninghamii 0.164 0.014 0.170 0.023 0.190 0.022 Nothofagus gunnii 0.287 0.040 0.232 0.015 0.182 0.054 Phyllocladus aspleniifolius 0.456 0.046 0.298 0.031 0.363 0.028 Poaceae 0.195 0.019 0.196 0.015 0.102 0.006 Sprengelia spp. 0.132 0.020 0.081 0.010 0.134 0.044

135 |

5.8.2. Additional results for RSAP and PPEs calculations

Bauera rubioides Cupressaceae 30 30

25 25

20 20 y = 1.1867x - 0.7555 15 15 R² = 0.2696 y = 1.6768x + 2.5017 10 10 R² = 0.0912 Pollen abundance (%) abundance Pollen

5 (%) abundance Pollen 5

0 0 0 5 10 15 0 1 2 3 4 DWPA (%) DWPA (%)

Ericaceae Eucalyptus 35 120

30 100 y = 2.8227x + 5.2817 25 R² = 0.7101 y = 0.1611x + 0.2827 80 20 R² = 0.222 60 15 40 10 Pollen abundance (%) abundance Pollen 5 (%) abundance Pollen 20 0 0 0 20 40 60 80 0 10 20 30 40 DWPA (%) DWPA (%)

Eucryphia lucida

50 Figure 5-S1b. Scatterplots of distance weighted plant abundance (DWPA %) vs 40 pollen abundance (%). The y-intercept in 30 y = 7.9005x - 1.7924 the plots represents background pollen 20 R² = 0.891 component for the surveyed area (50 km radius). Pollen abundance (%) abundance Pollen 10

0 0 2 4 6 8 DWPA (%)

136 |

Gymnoschoenus Leptospermum sphaerocephalus 60 50 50 40 40

30 30 y = 0.1862x - 0.3122 y = 0.3421x + 3.9453 20 R² = 0.3257 20 R² = 0.005 10 (%) abundance Pollen 10 Pollen abundance (%) abundance Pollen 0 0 0 20 40 60 80 0 2 4 6 8 10 DWPA (%) DWPA (%)

Monotoca Nothofagus cunninghamii 12 70

10 60 y = 0.2099x + 24.253 50 8 y = 0.888x + 1.4581 R² = 0.0685 R² = 0.1119 40 6 30 4 20 Pollen abundance (%) abundance Pollen Pollen abundance (%) abundance Pollen 2 10 0 0 0 1 2 3 4 5 0 20 40 60 80 DWPA (%) DWPA (%)

Nothofagus gunnii Poaceae 40 35 35 y = 0.9856x - 0.3788 30 y = 0.5218x + 2.3677 R² = 0.9873 R² = 0.4081 30 25 25 20 20 15 15 10 10 Pollen abundance (%) abundance Pollen 5 (%) abundance Pollen 5 0 0 0 10 20 30 40 0 10 20 30 40 50 DWPA (%) DWPA (%)

Figure 5-S2b. Scatterplots of distance weighted plant abundance (DWPA %) vs pollen abundance (%). The y-intercept in the plots represents background pollen component for the surveyed area (50 km radius).

137 |

Phyllocladus aspleniifolius Sprengelia 45 16 40 14 y = 0.5969x - 0.332 y = 1.2311x + 3.8736 35 12 R² = 0.9019 R² = 0.6339 30 10 25 8 20 6 15 10 4 Pollen abundance (%) abundance Pollen Pollen abundance (%) abundance Pollen 5 2 0 0 0 10 20 30 0 5 10 15 20 25 DWPA (%) DWPA (%)

Figure 5-S2c. Scatterplots of distance weighted plant abundance (DWPA %) vs pollen abundance (%). The y-intercept in the plots represents background pollen component for the surveyed area (50 km radius).

138 |

Figure 5-S3. Optimization function scores using the GPM and LSM for 50 km vegetation radius.

139 |

Figure 5-S4. Comparison of PPEs from LSM and GPM for 2, 10 and 50 km vegetation radius.

140 |

141 |

4.50 Eucryphia lucida 4.00

3.50

3.00

2.50 Bauera rubioides 2.00 Leptospermum 1.50 Nothofagus Gymnoschoenus 1.00 cunninghamii spaherocephalus 0.50

Relative pollen productivity ratio GPM/LSM Ericaceae 0.00 0.000 0.010 0.020 0.030 0.040 0.050 0.060 0.070 0.080 Fall speed (m/s)

142 |

Chapter 6. A quantification of regional land- cover changes from western Tasmania4

4 This chapter is an adapted version of the manuscript published in Journal of Biogeography (Mariani et al., 2017): How old is the Tasmanian cultural landscape? A test of landscape openness using quantitative land-cover reconstructions. The analyses here presented address Thesis Aim III (Step 2; as stated in Chapter 1 - Introduction).

Abstract

Aim: To test competing hypotheses about the timing and extent of Holocene landscape opening using pollen-based quantitative land-cover estimates.

Location: Dove Lake, Tasmanian Wilderness World Heritage Area, Australia.

Methods: Fossil pollen data were incorporated into pollen dispersal models and corrected for differences in pollen productivity among key plant taxa. Mechanistic models (REVEALS— Regional Estimates of VEgetation Abundance from Large Sites) employing different models for pollen dispersal (Gaussian plume and Lagrangian stochastic models) were evaluated and applied in the Southern Hemisphere for the first time.

Results: Validation of the REVEALS model with vegetation cover data suggests an overall better performance of the Lagrangian stochastic model. Regional land-cover estimates for forest and non-forest plant taxa show persistent landscape openness throughout the Holocene (average landscape openness ~50%). Gymnoschoenus sphaerocephalus, an indicator of moorland vegetation, shows higher values during the early Holocene (11.7–9 ka) and declines slightly through the mid-Holocene (9–4.5 ka) during a phase of partial landscape afforestation. Rain forest cover reduced (from ~40% to ~20%) during the period between 4.2–3.5 ka.

143 |

Main conclusions: Pollen percentages severely under-represent landscape openness in western Tasmania and this bias has fostered an over-estimation of Holocene forest cover from pollen data. Treeless vegetation dominated Holocene landscapes of the Dove Lake area, allowing us to reject models of landscape evolution that invoke late- Holocene replacement of a rain forest-dominated landscape by moorland. Instead, we confirm a model of Late Pleistocene inheritance of open vegetation. Rapid forest decline occurred after c. 4 ka, likely in response to regional moisture decline.

6.1. Introduction

The ability to accurately model and predict Earth system behaviour is crucial for developing effective management strategies for our future environment (e.g. Anderson et al., 2010; Gaillard et al., 2010). Recognition of the reciprocal relationship between vegetation cover and climate via biogeochemical and biogeophysical processes/feedbacks (e.g. Foley et al., 2003) has fostered development of coupled dynamic vegetation and climate models (e.g. Smith et al., 2011). A critical point is the scarce information on past land cover currently incorporated into such models (e.g. Kaplan et al., 2002; Sitch et al., 2003; Smith et al., 2011). Quantitative estimates of past plant cover are rare (e.g. Trondman et al., 2015), with the Southern Hemisphere particularly under-represented. Here we apply, for the first time in the Southern Hemisphere, mechanistic models for regional vegetation reconstruction to quantify landscape changes in western Tasmania, Australia. These reconstructions allow us to address biogeographical and methodological debates over (1) the efficiency of mechanistic models in landscapes dominated by both wind- and animal-pollinated plants; and (2) the timing and extent of long-term vegetation changes in western Tasmania.

Fire is considered the most critical non-climatic factor controlling land-cover changes on Earth (Bond, Woodward, and Midgley, 2005). Fire and vegetation are strongly linked in Australia (Bradstock, Williams, and Gill, 2002), with, for example, the distribution of fire-sensitive rain forest among flammable sclerophyll

144 |

vegetation considered to be a result of the past fire history (Bowman, 1998, 2000). Tasmania (Figure 6-1) is an exemplar of the role of fire in decoupling vegetation from climate, with large areas in which rain forest is the potential vegetation currently occupied by fire- promoted vegetation (Brown and Podger, 1982a; Jackson, 1968; Wood and Bowman, 2011). Indeed, the juxtaposition of fire-sensitive rain forest against pyrogenic species (e.g. Eucalyptus) in Tasmania has puzzled ecologists for decades (Bowman and Jackson, 1981; Brown and Podger, 1982b; Yospin et al., 2015) and the origin and evolution of the present-day dominance of pyrogenic moorland is the subject of long-standing debate (see Colhoun, 1996; Fletcher and Thomas, 2010a; Macphail, 2010 and references therein). This debate is split between those that argue for late-Holocene replacement of a forested landscape by open vegetation (Colhoun, 1996; Macphail, 1979), while others argue an open landscape persisted throughout the Holocene (Fletcher and Thomas, 2010a; Thomas, 1995a).

Proponents of the former model interpret the region-wide dominance of rain forest pollen types as indicating a climate-driven expansion of rain forest from the Late Pleistocene to the mid-Holocene (e.g. Colhoun, 1996; Macphail, 1979). In contrast, Fletcher and Thomas (2007a,b, 2010a,b) highlight the bias in the pollen record towards plants that produce large amounts of well-dispersed pollen (i.e. rain forest species in Tasmania), and use modern pollen spectra to infer the persistence of open vegetation across western Tasmania for the last c. 12,000 years (12 kyr). They note a departure between Holocene and previous interglacial floras and charcoal sequences, citing human arrival during the last glacial cycle (~35 kyr) as the probable cause for this, thus concluding that western Tasmania represents an ancient cultural landscape. Macphail (2010) has critiqued their model, arguing that the nature of pollen deposition at many of the mires studied by Fletcher and Thomas (2010a) biases pollen records towards local bog vegetation, which is palynologically similar to pyrogenic moorland. Thus, the debate over the origin and development of the modern treeless and pyrogenic vegetation landscape of western Tasmania centres on how pollen production and deposition biases are resolved.

145 |

Here, we produce quantitative land-cover estimates to understand vegetation change and test competing models of landscape evolution in western Tasmania. We employ recently reported pollen productivity estimates (PPEs) for this region (Mariani et al., 2016) to develop quantitative land-cover estimates using the REVEALS (Sugita, 2007b) approach.

6.1.1 Quantitative reconstruction of land-cover from pollen data

Quantitative reconstruction of past vegetation cover has long been a major objective of palynology (e.g. Brostrom et al., 2004; Gaillard, Birks, Ihse, and Runborg, 1998; Sugita, 1993). Numerous factors influence the over- or under-representation of vegetation in pollen spectra (e.g. taphonomy, pollen productivity, dispersal capabilities) (Fletcher and Thomas, 2007b; de Nascimento et al., 2015). Recent advances enable effective modelling of pollen productivity and dispersal (Gaillard et al., 2010; Sugita, 2007a,b), improving understanding of past vegetation dynamics and providing quantitative land-cover data to landscape management and palaeoclimate modelling. Emerging from these efforts is the landscape reconstruction algorithm (LRA: Sugita, 2007a,b) for estimating vegetation abundance on regional (104–105 km2 - REVEALS model) and local (<1 km2 up to 5 km2—LOVE model) scales. These mechanistic models employ PPEs to correct productivity bias, and models of dispersal and deposition of small particles in the air, to correct differences in pollen dispersion.

The main assumption behind REVEALS, supported by simulations and empirical studies, is that pollen deposited in large sites (>50-100 ha) originate from a larger source-area (Jacobson and Bradshaw, 1981), and can be used to estimate regional vegetation cover (Hellman, Gaillard, Brostrom, and Sugita, 2008; Sugita, 2007b). A critical choice is the pollen dispersal model. The most commonly employed dispersal model in the REVEALS approach is the Gaussian plume model

146 |

(REVEALS-GPM) (Prentice, 1985; Sugita, 1993, 1994, 2007a), while the emerging Lagrangian stochastic model (REVEALS-LSM) has only been employed in Germany thus far (Kuparinen, Markkanen, Riikonen, and Vesala, 2007; Theuerkauf, Kuparinen, and Joosten, 2013). The GPM uses Sutton’s air pollutant plume dispersion equation (Sutton, 1953) and describes the concentration of particles downwind from a point source as spreading outward from the centreline of the plume following a normal distribution. This model may be less reliable for particle dispersal over longer distances, due to the influence of vertical airflows and turbulence (Kuparinen et al., 2007). The LSM is fully mechanistic dispersal model that more realistically describes pollen dispersal and deposition in lakes (Theuerkauf et al., 2013). Both models were previously employed in Tasmania to derive PPEs; LSM was found to perform more realistically (Mariani et al., 2016).

Quantifying landscape openness from palynological data is a long- awaited goal (Faegri and Iversen, 1989). In Europe, the ratio between Arboreal Pollen (AP) and Non-Arboreal Pollen (NAP) has traditionally been used to qualitatively describe human impact on landscapes. However, pollen percentages and AP:NAP relationships are not linearly related to vegetation composition and landscape patterns (e.g. Hellman et al., 2008; Sugita, Gaillard, and Brostrom, 1999) and the proportion of unforested land is strongly underestimated by NAP data (Hellman et al., 2008; Kunes, Odgaard, and Gaillard, 2011; Kunes et al., 2015). An attempt to correct for this bias in southern Sweden revised NAP-based estimates of 30%–40% open vegetation to 60%–80% using REVEALS (Gaillard et al., 2010; Hellman et al., 2008). REVEALS development and validation have mainly occurred in the Northern Hemisphere (e.g. Abraham and Kozakova, 2012; Hellman et al., 2008; Hultberg, Gaillard, Grundmann, and Lindbladh, 2015; Mazier et al., 2012; Sugita, Parshall, Calcote, and Walker, 2010; Trondman et al., 2015), where most plant species are wind-pollinated and the goal has been quantifying human impact on vegetation. To date, no attempt has been made to quantitatively reconstruct land-cover changes in the Southern Hemisphere using pollen dispersal models, nor in vegetation with

147 |

numerous animal-pollinated plants (e.g. Eucalyptus), such as in Tasmania.

In this paper, we apply and validate REVEALS-LSM and REVEALS-GPM to reconstruct Holocene vegetation/land cover in the Southern Hemisphere for the first time.

6.2. Methods

6.2.1 Study area

Tasmania is a cool temperate island located around 300 km south of mainland Australia. The island is bisected by a north-west/south-east trending mountain range, which produces a steep orographic precipitation gradient with a wet west and dry east (Gentilli, 1972; Sturman and Tapper, 2006). Mean annual temperature in the west varies from 5–7°C (winter) to 14–16°C (summer) and precipitation values exceed 3000 mm (Australian Bureau of Meteorology, http://www.bom.gov.au/). Presently, western Tasmania is dominated by treeless pyrogenic vegetation (moorland), whereas rain forest communities are restricted by topography and fire protection (Wood et al., 2011). The main component of current moorland vegetation in western Tasmania is buttongrass (Gymnoschoenus sphaerocephalus Brown/Hooker [Cyperaceae]).

148 |

149 |

Dove Lake (41°39034.27″S, 145°57035.14″E; Wee bone ne tinker in the local indigenous language) lies beneath the iconic Cradle Mountain (War loun dig er ler) within the World Heritage-listed Cradle Mountain-Lake St. Clair National Park. Humans settled here>35 ka (Cosgrove, 1999) and Cradle Mountain was the last place in Tasmania’s west occupied by Tasmanian Aborigines prior to their exile in the 19th century (Plomley, 1966). Dove Lake is medium- sized (90 ha), subalpine lake (940 m a.s.l.) of glacial origin (Figure 6-1, Figure 6-S1) and surrounded by steep topography (Figure 6-S1). Dove Lake is one of the largest natural lakes in western Tasmania and its size makes it suitable for regional vegetation reconstruction using REVEALS (e.g. Hellman et al., 2008; Sugita, 2007b). Mean annual rainfall at the nearest meteorological station (Cradle Valley; 41°38024.23″S, 145°56024.28″E; 903 m a.s.l.) is 2698 mm (Figure 6-1).

The vegetation around Dove Lake is dominated by sclerophyllous heath and Eucalyptus woodlands on the eastern and western flanks; moorland is found at the northern edge and rain forest only occupies the south-west corner (Figure 6-1b). Dominant species in the modern landscape are Eucalyptus coccifera Hook., Lophozonia cunninghamii Hook. (syn. Nothofagus cunninghamii), Gymnoschenous sphaerocephalus and various ericaceous shrubs. Typical montane rain forest trees are also found, such as Fuscospora gunnii Hook. (syn. Nothofagus gunnii), Athrotaxis selaginoies D.Don and A. cupressoides D.Don (Cupressaceae). In this paper, the genus name Nothofagus will be used when referring to Lophozonia and Fuscospora in order to maintain consistency with fossil records (Hill et al., 2015).

6.2.2. Pollen and charcoal analyses

Pollen, spore and microscopic charcoal sample preparation followed standard protocols (Faegri and Iversen, 1989) at 1-cm intervals. Relative pollen data were calculated from a pollen sum of at least 300 terrestrial pollen grains per sample (excluding wetland taxa and ferns). Microscopic charcoal was counted on the microscope slides during pollen counting. Macroscopic charcoal was analysed at 5 mm- resolution to reconstruct local fire history. A set amount of sediment (1.25 cm3)

150 |

per sample was digested in 5% sodium hypoclorite (bleach) for at least 2 weeks and sieved using 125 and 250-μm diameter meshes (Whitlock and Larsen, 2001). To account for variations in sediment deposition, charcoal counts were converted into Charcoal Accumulation Rates (CHAR, particles cm2/yr).

6.2.3. Quantitative vegetation reconstruction and model validation

The choice of dispersal model in REVEALS is crucial (Theuerkauf et al., 2013). The Gaussian plume model (Prentice, 1985; Sugita, 1993, 1994) and the Lagrangian stochastic model (Kuparinen et al., 2007) are the two models currently implemented in PPE calculations and REVEALS runs (REVEALS-GPM and REVEALS-LSM). Compared to the GPM, the LSM gives greater importance to pollen arriving from longer distances (Kuparinen et al., 2007; Theuerkauf et al., 2013). Pollen fall speed (terminal velocity) also has a large effect on predicted dispersal in the GPM, yet a small effect in the LSM (Theuerkauf et al., 2013).

We validated the models’ vegetation estimates from two surface sediment samples from large lakes (Dove Lake and Lake St Clair- Hopf, Colhoun, and Barton, 2000) against the actual vegetation cover around the same lakes and pollen data (see Hellman et al., 2008) using principal component analysis (PCA; Jolliffe, 2014). Lake St Clair (Figure 6-1) is the largest natural lake (4,500 ha) in western Tasmania and expected to register regional vegetation information. REVEALS modelling employed recently published PPEs and fall speeds for 13 key pollen taxa in western Tasmania (Mariani et al., 2016). In addition to the regional average wind speed of 6.5 m/s, we calculated additional GPM-based PPEs for a wind speed of 3 m/s to compare with published estimates. For the LSM, we calculated PPEs with parameters for windy conditions (see Kuparinen et al., 2007) to compare with the published LSM-based PPEs calculated with parameters for unstable conditions (see Mariani et al., 2016). Thus we assess four modelled atmospheric scenarios: GPM 3 m/s, GPM 6.5 m/s, LSM unstable and LSM windy unstable.

151 |

All model runs were performed in the “disqover” package (Theuerkauf et al., 2016) for R v.3.3.1 (The R foundation for Statistical Computing, 2016). Plant cover percentages around the target lakes were calculated for a 50 km-radius (approximately the area covered by the REVEALS reconstruction) using the vegetation map, TasVeg 3.0 (http://maps. thelist.tas.gov.au/listmap/) in ARCMAP 10.2 (ESRI—Environmental Systems Resource Institute, 2009, Redlands, CA) and plant survey data (Mariani et al., 2016). Validation employed PCA in PC-ORD FOR WINDOWS 4.27 (McCune and Mefford, 1999). Detrended correspondence analysis was performed in the R package “vegan” (Oksanen et al., 2015) to confirm that PCA was appropriate (Birks, Lotter, Juggins, and Smol, 2012). Pollen percentages from the two surface samples were also added to the analysis to gauge model improvement. Mean deviation from actual plant cover (%) was also calculated for each taxon.

For REVEALS application to the Dove Lake fossil record, all taxa from Mariani et al. (2016) were used (excluding infrequent <1% taxa: Monotoca, Sprengelia and Eucryphia), accounting for at least 80% of the terrestrial pollen counts per sample. To assess the performance of REVEALS under different atmospheric scenarios, REVEALS was applied to the Holocene pollen data from Dove Lake using the four model settings mentioned above. Plant taxa were grouped into forest and non-forest categories based on structure (Table 6-1).

152 |

Fall speed (m/s) GPM 10km GPM 10km SD LSM 10km LSM 10km SD Bauera rubioides nf 0.004 0.165 0.025 0.067 0.011 Cupressaceaef 0.031 1.003 0.124 1.337 0.198 Ericaceaeanf 0.043 0.043 0.004 0.067 0.007 Eucalyptus spp.f 0.008 1.000 0.000 1.000 0.000 Gymnoschoenus sphaerocephalus nf 0.072 0.027 0.003 0.050 0.006 Leptospermum spp. nf 0.005 0.284 0.027 0.145 0.016 Nothofagus cunninghamii f 0.037 0.830 0.055 1.281 0.084 Nothofagus gunnii f 0.020 0.251 0.029 0.385 0.033 Phyllocladus aspleniifolius f 0.026 0.631 0.099 0.842 0.163 nf Poaceae 0.030 0.242 0.019 0.414 0.031

6.3. Results

6.3.1. Pollen and charcoal analyses

Pollen, spores and microscopic charcoal were analysed in 109 samples. A reduced pollen diagram showing only key taxa for vegetation reconstruction appears in Figure 6-2. For comparison with REVEALS results key taxa were re-scaled to 100%. The full, non-rescaled pollen diagram with pollen zone descriptions appears in Figure 6-S3 and 6-S4. Pollen abundances of montane rain forest indicators (Cupressaceae and Nothofagus gunnii) show a persistent decline from early to late Holocene. Pollen of the main rain forest component today, Nothofagus cunninghamii, shows high values (~50%), until c. 4 ka, and then declines to 25%. Pollen of Phyllocladus aspleniifolius Hooker increases from the early to mid-Holocene, with an evident decline since c. 5 ka. Eucalyptus pollen is relatively abundant throughout the sequence, with a stepwise increase at around c. 5 ka. Pollen percentages of non- forest taxa are consistently low (<20%): Poaceae and Bauera rubioides increase after

153 |

c. 5 ka, together with pollen of Gymnoschoenus sphaerocephalus. G. sphaerocephalus pollen is most abundant during the early Holocene (~8%), somewhat rarer throughout the mid-Holocene, and again more abundant after c. 5 ka. Total forest pollen increases from early to mid-Holocene, reaching 85% of the total terrestrial pollen spectrum. During the late Holocene, rain forest pollen declines substantially to 40%– 60%.

Macroscopic CHAR values are generally low throughout the core, with substantial peaks at 3.7, 4.5 ka and relatively high macroscopic charcoal influx between 6.5 and 7.5 ka (Figure 6-2). Micro-CHAR peaks are mainly synchronous with macro-CHAR peaks, both showing a sharp increase during the last 160 years preceded by periods of high microscopic charcoal influx between 3.4–4, 5.8–6.3 ka and peaks at 7.2 and 10.8 ka (Figure 6-2).

Figure 6-2. Diagram showing pollen % of key taxa used for the vegetation reconstruction. Macro- and micro- charcoal accumulation rates (particles/cm2 yr-1) are also shown. Colours represent the grouping of the taxa according to vegetation structure: green is used for taxa generally occurring within forests; red is used for plant taxa commonly found in non- forested environments. Dashed lines indicate statistically determined pollen zones (see Chapter 6.6.4. for zones’ description).

154 |

6.3.2. REVEALS validation

PPEs calculations under atmospheric settings different from the published material (Table 6-1) appear in Table 6-S3. Differences between REVEALS estimates and predicted vegetation cover (%) are presented in Figures 6-S4 and S5. PPEs calculated with GPM 3 m/s and GPM 6.5 m/s show large differences, whereas PPEs from LSM under unstable and windy conditions are quite similar. Vegetation cover errors (%) are generally smaller for the REVEALS-LSM runs, compared to REVEALS-GPM (Figures 6-S4 and 6-S5).

155 |

REVEALS validation appears in Figure 6-3. PCA axis 1 clearly isolates pollen percentages, modern plant cover and REVEALS results for GPMs and LSMs. Gymnoscheonus sphaerocephalus an Eucalyptus show the strongest correlations with Axis 1, r = .98 and r = -.89 respectively, whereas Ericaceae and Nothofagus cunninghamii are correlated with Axis 2, r = .85 and r = -.71 respectively.

The majority of pollen taxa align along Axis 2, clearly separating anemophilous (skewed towards N. cunninghamii) and zoophilous taxa (skewed towards Ericaceae). Zoophilous taxa are positioned close to results for REVEALS- LSM and REVEALS- GPM 6.5 m/s and modern plant cover, whereas wind-pollinated taxa are skewed towards pollen percentages. REVEALS-LSM (both unstable and windy conditions) and REVEALS- GPM (6.5 m/s) estimates for Lake St. Clair are close to the actual plant cover extracted from a 50-km buffer from this site. The PCA biplot shows REVEALS- GPM performs better using 6.5 m/s wind speed than 3 m/s. Current plant cover around Dove Lake and Lake St Clair is more closely related with REVEALS-LSM results than REVEALS-GPM results on Axis 1, suggesting an overall better performance of LSM. Interestingly, the two runs for REVEALS-LSM are virtually identical, whereas the REVEALS-GPM runs are separated on the PCA biplot.

6.3.3. REVEALS application

Figure 6-4 summarizes pollen percentages and REVEALS estimates obtained using GPM and LSM for forest and non-forest taxa. REVEALS estimates for single taxa are presented in Figures 6-S6 and 6-S7, alongside effects of different atmospheric parameters (Figure 6-S8). REVEALS-LSM provides more coherent cover estimates than REVEALS-GPM (Figure 6-S8). Given the validation results, we focus on the results of REVEALS-GPM (6.5 m/s) and REVEALS-LSM (unstable) to address long- term landscape dynamics at Dove Lake. Gaussian plume model estimates show high land-cover for G. sphaerocephalus (solid red, Figure 6-4), a moorland indicator, reaching more than 60%. Rain forest cover (light green, Figure 6-4) is relatively low, accounting for <40% throughout the Holocene. Eucalyptus (dark green, Figure 6-4), an indicator of sclerophyll forests, covers <5% throughout the reconstruction period.

156 |

LSM dispersal simulations produce markedly different results: G. sphaerocephalus covers up to 45% of total reconstructed land cover, forest cover varies between ~20% and ~40% and Eucalyptus cover is fairly constant between 5% and 10% throughout the reconstruction period. Both model runs show maximum forest cover between 10 and 4 ka. After ~4 ka, a sharp decrease in forest cover (from ~40% to ~20%) is observed, corresponding to a 50% decline of tree cover.

Figure 6-4. Comparison of pollen data and quantitative reconstruction results: summary diagram showing a) pollen percentages; b) REVEALS vegetation estimates using the Gaussian plume model under neutral conditions (REVEALS-GPM); c) REVEALS vegetation estimates using the Lagrangian stochastic model (REVEALS-LSM). Dashed lines indicate statistically determined pollen zones (same as Figure 6-2).

157 |

6.4. Discussion

6.4.1. REVEALS model performance in western Tasmania

REVEALS corrects for dispersal and productivity biases in pollen data, hence the choice of the dispersal model alters the reconstructions. Application of REVEALS- GPM and REVEALS-LSM produced considerably better results than the raw percentages of key pollen taxa (Figures 6-3 and 6-4. The PCA biplot (Figure 6-3) shows the LSM outperformed the more commonly used GPM. Improvement is more evident for Dove Lake, likely due to its small size (~90 ha) relative to Lake St. Clair (4,500 ha). REVEALS-LSM results for unstable and windy unstable conditions are very consistent, whereas REVEALS- GPM results are more sensitive to wind parameters (Figure 3). REVEALS-LSM estimates for Dove Lake are close to actual plant cover, and, thus, appear realistic. REVEALS-GPM performs similarly with high wind speed, however, PPEs calculated with the GPM are unrealistic (Mariani et al., 2016). Our results suggest that LSM is a better model for pollen dispersal and deposition in western Tasmania. Pollen dispersal over long distances the most important for pollen deposition in larger lakes is mainly carried by turbulent flows and updrafts (Jackson and Lyford, 1999). GPMs do not describe such asymmetric airflows and therefore appear only suited to predict short-distance dispersal (Theuerkauf et al., 2016). The Lagrangian stochastic model simulates updrafts and may be more suitable to model pollen deposition in lakes (Theuerkauf et al., 2013). Updraft velocities are typically much higher than the fall speed of pollen, which explains the greater difference between estimates produced by REVEALS-GPM and REVEALS-LSM for taxa characterized by very large pollen grains, such as N. cunninghamii and G. sphaerocephalus (Figures 6-S7 and 6-S8).

We conclude that the REVEALS model can be successfully applied to Holocene fossil pollen records in western Tasmania. This represents an important advance in Southern Hemisphere palaeoecology and for systems in which animal pollination

158 |

is common. A critical limitation of model application in our study region is the scarcity of large (>1 km2) natural lakes. However, model applicability can be tested by substituting single large lakes with multiple small sites, as proposed in southern Sweden (Trondman et al., 2015) or using other simulation-based reconstruction approaches (Bunting and Middleton, 2009; Mrotzek et al., 2017).

6.4.2. Holocene land-cover changes in western Tasmania

Our land-cover estimates reveal a landscape dominated by open vegetation through the entire Holocene (Figure 6-5). Early Holocene forest cover was approximately 40%, increasing towards a rain forest maximum in the mid-Holocene, a period when forest and non-forest equally shared 50% of the landscape. This vegetation mix corresponds with peak rain forest pollen content in pollen records from western Tasmania (e.g. Colhoun, 1996; Fletcher and Moreno, 2012; Macphail, 1979; Markgraf, Bradbury, and Busby, 1986) and a minimum in regional fire activity (Figure 6-5; Fletcher and Moreno, 2012; Fletcher et al., 2015). This period is synchronous with a phase (9.2–5 ka) of surplus moisture promoting speleothem growth in Lynds Cave (Xia, Zhao, and Collerson, 2001), approximately 25 km NNW of Dove Lake. Wetter conditions in Tasmania during this phase are linked to an enhancement and/or northward-displacement of the prevailing mid- latitude Southern Westerly Winds (SWW) (Fletcher and Moreno, 2011, 2012), which drove lake level fluctuations in southern Australia and hydroclimatic change across the Southern Hemisphere mid-latitudes (Fletcher and Moreno, 2012; Wilkins, Gouramanis, De Deckker, Fifield, and Olley, 2013).

159 |

160 |

A late-Holocene rain forest decline around Dove Lake occurred in response to decreased moisture availability and a concomitant increase in fire activity across western Tasmania (Beck et al., 2017; Fletcher and Moreno, 2012; Fletcher et al., 2014; Fletcher, Wolfe, et al., 2014; Fletcher et al., 2015; Rees et al., 2015). These changes are linked to increasingly frequent El Niño events in the tropical Pacific (Donders, Wagner-Cremer, and Visscher, 2008; McGlone, Kershaw, and Markgraf, 1992; Moy, Seltzer, Rodbell, and Ander- son, 2002), which are associated with drier conditions and increased fire activity in south-east Australia today (Mariani and Fletcher, 2016; Nicholls and Lucas, 2007). The sharp decline in forest cover at 4.1 ± 0.1 ka is, considering age uncertainties, virtually synchronous with an increase in fire activity observed across western Tasmania at 4.0 ka (Fletcher et al., 2015; Figure 6-5e). While it is impossible to distinguish between anthropogenic and climatic drivers of this fire activity, historical ignitions in Tasmania are almost entirely of human origin (Bowman and Brown, 1986). The long occupation (>35 kyr) and people’s historically documented use of fire to manage landscapes (Plomley, 1966; Thomas, 1995b) argues strongly for a climatically modulated fire regime in which humans were the primary ignition source.

Comparison with another record from a small site (0.51 ha) near Dove Lake, Wombat Pool (~1.5 km north-west of Dove Lake) (Stahle, Whitlock, and Haberle, 2016) offers insights into local fire and vegetation dynamics. A peak in macroscopic charcoal is recorded at Wombat Pool and Dove Lake at 3.7 ka, indicating enhanced local fire activity. Paradoxically, this peak in local fire activity occurs after the main vegetation change around Dove Lake (Figure 6-5). Regional vegetation change therefore occurred prior the local increase in fire activity surrounding Dove Lake, likely in response to regional hydroclimatic and fire regime change. The close match between regional vegetation shifts at Dove Lake and regional charcoal influx from western Tasmania (Figure 6-5e) suggests the Dove Lake land-cover record reflects climate-vegetation dynamics that were mediated by humans and fire activity.

161 |

6.4.3. Did moorland dominate Holocene landscapes of western Tasmania?

Our land-cover estimates prove the persistence of treeless vegetation through the entire Holocene, with G. sphaerocephalus accounting for 15%–45% (REVEALS-LSM) of land cover throughout the reconstruction period (Figure 6-5). Pollen percentages for this taxon vary between 2% and 8% (Figure 6-2), highlighting severe under-representation (Fletcher and Thomas, 2007b). Tasmanian moorland vegetation, although dominated by G. sphaerocephalus (Jarman, Kantivalis, and Brown, 1988), includes other taxa, including Leptospermum, Poaceae, Restionaceae, Melaleuca, Sprengelia and Gleichenia (Jarman et al., 1988). Most of these are also found within other vegetation formations (e.g. heath, Eucalyptus woodlands, sclerophyll forests), making them less reliable indicators of moorland. Hence, our moorland abundance reconstructions based on G. sphaerocephalus must be considered as absolute minimum values for western Tasmania moorlands. Other approaches, such as the Multiple Scenario Approach (Bunting and Middleton, 2009), should be implemented in future to test the possible range of abundances of these communities.

The landscape-scale decoupling of vegetation and climate in western Tasmania has intrigued ecologists and archaeologists for decades (Colhoun, 1996; Colhoun and Shimeld, 2012; Cosgrove, 1995; Fletcher and Thomas, 2007a,b; Fletcher and Thomas, 2010; Jackson, 1968; Macphail, 2010; Thomas, 1993, 1995a,b). We find no evidence for rain forest (the climatic climax vegetation) dominance in the landscape at any time through the Holocene around Dove Lake, enabling us to reject models that invoke late- Holocene replacement of a rain forest-dominated landscape by moorland (Colhoun, 1996; Macphail, 1979). Instead, we support the inheritance of Late Pleistocene landscape openness through the Holocene to the present (Fletcher and Thomas, 2010a). This model con- tends that the people’s arrival to a largely treeless landscape during the Last Glacial Cycle (ca. 35 kyrs) and their subsequent manipulation of fire regimes restricted Holocene expansion of typical interglacial vegetation (i.e. rain forest) (Colhoun and van der Geer, 1998) to areas protected from fire. Instead, plants tolerant of frequent burning and an interglacial climate (i.e. moorland and associated

162 | sclerophyllous species) expanded at the Pleistocene-Holocene transition and remained dominant through the Holocene. Indeed, G. sphaerocephalus is absent from pre- Holocene interglacial pollen spectra (Colhoun, Pola, Barton, and Heijnis, 1999; Colhoun and van der Geer, 1998), and Holocene sequences have significantly more charcoal than pre-Holocene interglacials (Fletcher and Thomas, 2010a). Thus, we provide empirical support for the Late Pleistocene inheritance model and the notion that western Tasmania constitutes an ancient cultural landscape (Fletcher and Thomas, 2010a).

6.5. Conclusions

Statistical analysis of modern pollen and plant cover estimates in western Tasmania shows a more realistic performance of REVEALS estimates based on the Lagrangian stochastic model (LSM) when compared to widely used Gaussian plume model. REVEALS-LSM effectively corrects for biases of productivity and dispersal, allowing better quantification of palynologically underrepresented plant taxa. Our reconstruction quantifies change in landscape openness in a defined geographic area for the first time in the Southern Hemi- sphere. Results indicate a landscape dominated by treeless vegetation throughout the last 12 kyr. We identify a regional shift in regional vegetation at c. 4 ka, manifest as a rapid halving of rain forest cover from 40% to 20% landscape cover. This phase likely reflects a modulation of the effects of regional-scale anthropogenic burning in response to regional climatic change. Finally, evidence for a persistently open landscape supports the notion that this region represents an ancient cultural landscape resulting from the influence of anthropogenic burning through the last glacial cycle and its influence over post-glacial vegetation development.

163 |

6.6. Supplementary results and information

The figures and tables displayed below are part of the supporting information for the paper ‘How old is the Tasmanian cultural landscape? A test of landscape openness using quantitative land-cover reconstructions’ (Mariani et al., 2017).

164 |

6.6.1. Additional details about the area surrounding Dove Lake

Figure 6-S1. Digital elevation model of the mountainous area surrounding Dove Lake, Tasmania. Roads, water bodies and toponyms are also shown.

165 |

6.6.2. Core extraction and chronology information

Methodology for core extraction and chronology

A bathymetric map of Dove Lake was created using Universal Kriging in ArcGIS 10.2 (ESRI - Environmental Systems Resource Institute, 2009, Redlands, California) of data acquired with an ultrasonic depth sounder. A 122 cm-long organic sediment core (TAS1507SC2) was subsequently retrieved from Dove Lake (90 ha) at a depth of 43 m using a 6.8-cm diameter polycarbonate chamber attached to a Universal Gravity Corer (http://www.aquaticresearch.com; Fig. 1b). The core was split subsampled at 5 mm resolution for charcoal and pollen analysis. Core chronology is based on a combination of 11 210Pb and 5 14C. Lead radioisotope activities were determined using alpha-spectrometry at the Australian Nuclear Science and Technology Organization (ANSTO; Table 6-S2a), employing Constant Rate of Supply (CRS) model. Radiocarbon dating was performed at ANSTO and Direct AMS (Bothell, WA, USA) and calibrated to calendar years before present (cal. yr BP; 1950 CE) using the Southern Hemisphere calibration curve (Hogg et al., 2013; Table 6- S2b). Duplicate bulk sediment and macrofossil (wood) ages were obtained where available (n=2; 50.5 and 53 cm) to assess reservoir effects. Age-depth modelling was performed using the Bacon package for R (Blaauw and Christen, 2011).

Results for lake bathymetry and core chronology

A bathymetric map of Dove Lake indicating coring location is shown in Figure 6- S1b. Radioactive lead and radiocarbon ages are presented in Table 6-2a,b. We restrict our current analysis to the Holocene period (11.7 ka to present). The age-depth model (Figure 6-S2) shows slow accumulation rates throughout the core, with a median rate of ~0.0125 cm yr-1. Comparison of bulk sediment and macrofossil ages for the only two identified macrofossils to rule out reservoir effects. The coupled samples collected from 50.5 cm show only 100 year-difference, whereas the coupled dates obtained at 53 cm show an offset of 900 years. Ages of the two wood pieces at 50.5 and 53 cm are, within errors, identical. An explanation of this issue is the possible downcore movement of the macrofossil collected at 53 cm, and this date

166 | was excluded from the age-depth model. No other macrofossils were found in the core TAS1507SC2. Uncertainty in the stratigraphic integrity of macrofossil fragments led us to adopt an age-depth model with no corrections for reservoir effect. Further investigations on the lake geochemistry and the development of palaeolimnological proxies from this and other cores (sensu Bertrand et al., 2012) will help refine the chronology.

Table 6-S1a. Table listing all 210Pb dates obtained on core TAS1507SC2.

Dry bulk density CRS Age CRS Error CRS Mass Accumulation Lab code Material Depth (g/cm3) (years BP) (years) Rates (g/cm2/year) 2016rc0054a - S393 Bulk sediment 0.25 0.19 -49 1 0.0023 ± 0.000 2016rc0054a - S394 Bulk sediment 0.75 0.16 -22 5 0.0055 ± 0.001 2016rc0054a - S395 Bulk sediment 1.75 0.15 -3 8 0.013 ± 0.004 2016rc0054a - S396 Bulk sediment 2.75 0.14 7 10 0.017 ± 0.007 2016rc0054a - S397 Bulk sediment 3.75 0.15 16 12 0.014 ± 0.006

Table 6-S1b. Table listing all radiocarbon dates obtained from core TAS1507SC2. Asterisks indicate 14C dates excluded from the age-depth model. Calibrated ages given in the right- hand columns.

Median age Lower age Upper age Lab code Material Depth (cm) Δ13c (per mil) 14C age 14C age error (cal yr BP) (cal yr BP) 2σ (cal yr BP) 2σ ANSTO OZU240 Bulk sediment 17.75 -26.5 2840 30 2900 2792 2980 D-AMS 015337 Bulk sediment 26.5 -23.7 3841 27 4185 4012 4349 D-AMS 015338 Bulk sediment 35.5 -21.6 4609 29 5211 5053 5442 ANSTO OZU241 Bulk sediment 40.75 -26.1 5085 35 5810 5663 5903 ANSTO OZU225 * Bulk sediment 51 -25.7 5560 35 6316 6218 6403 ANSTO OZU227 Macrofossil 51 -26.4 5475 35 6239 6123 6306 D-AMS 015339 Bulk sediment 53.5 -21.3 6338 34 7222 7158 7316 D-AMS 015336 * Macrofossil 53.5 -22.3 5460 30 6235 6034 6297 D-AMS 015340 Bulk sediment 58.5 -27.4 6548 31 7426 7323 7484 ANSTO OZU242 Bulk sediment 70.5 -26.8 7910 35 8662 8549 8965 D-AMS 015341 Bulk sediment 84.5 -28.8 8984 52 10052 9900 10229 ANSTO OZU243 Bulk sediment 100.75 -27.1 9705 45 11076 10787 11204 D-AMS 015342 Bulk sediment 119.5 -20.8 10650 39 12599 12436 12681

167 |

168 |

6.6.3. Additional information about the pollen record

Figure 6-S3. Extended pollen diagram for Dove Lake core including all taxa with abundance >1%.

169 |

6.6.4. Pollen zones descriptions for Dove Lake core

ZONE D1 (108-83 cm; 11.7-10 ka): this zone is characterised by high pollen abundance of N. cunninghamii (20-40%), which shows a constant increase from the bottom of the zone, reaching maximum values at the top. P. aspleniifolius pollen shows the minimum values of the entire study period (<7%) following an increasing trend towards zone D2, suggesting gradually increasing moisture conditions from 11.7 to 10 ka. Pollen taxa showing an opposite trend to P. aspeniifolius are Amaranthaeae, Poaceae and Asteraceae. Pollen indicators of montane rainforest, Cupressaceae and N. gunnii show a relatively stable abundance, respectively around 10% and 6% of the pollen spectrum. The combined abundance of these pollen taxa during this period is the highest of the entire pollen record. G. sphaerocephalus shows the highest pollen abundance (~ 7%) of the entire reconstruction period. Eucalyptus pollen maintains stable values around ~10%.

ZONE D2 (83-25 cm; 10-4 ka): this zone is characterised by a persistently high pollen abundance of N. cunninghamii (~40%). P. aspleniifolius pollen shows a gradually increasing trend, reaching maximum values at 7 ka, suggesting enhanced moisture conditions at this stage. G. sphaerocephalus, Amaranthaeae, Poaceae and Asteraceae follow an opposing trend. Eucalyptus pollen maintains stable values around ~10%, with an increase up to 20% at the top of the pollen zone. Pollen indicators of montane rainforest, Cupressaceae and N. gunnii, follow a gradual decline up to combined values <5% at 4ka.

ZONE D3 (25-0 cm; 4ka-present): this zone marks a major restructuring of the landscape surrounding Dove Lake. Pollen abundance of N. cunninghamii declines by 50% (from 40% in zone D2 to 20% in zone D3). P. aspleniifolius pollen gradually decreases, reaching minimum values at 0.5 ka (~5%), highlighting a reversal in the climatic trend towards drier conditions. This interpretation is supported by Eucalyptus pollen, which shows the highest values of the entire pollen spectrum. Furthermore, G. sphaerocephalus, Bauera rubioides, Amaranthaeae, Poaceae and

170 |

Asteraceae (all indicators of open landscape) show an increase in abundance. The littoral fern Isoëtes, a lake level indicator (sensu Pesce and Moreno, 2014), shows the highest spore abundance values of the reconstruction period, suggesting a decrease in lake level.

171 |

6.6.5. Further calculations of PPEs using different sets of atmospheric parameters

GPM neutral 3 m/s GPM neutral 6.5 m/s LSM unstable windy LSM unstable Bauera 0.125 0.165 0.053 0.067 Cupressaceae 1.614 1.003 1.117 1.337 Ericaceae 0.083 0.043 0.064 0.067 Eucalyptus 1.000 1.000 1.000 1.000 Eucryphia 1.188 1.459 0.382 0.340 G. sphaerocephalus 0.067 0.027 0.043 0.050 Leptospermum 0.225 0.284 0.127 0.145 Monotoca 0.493 0.424 0.416 0.485 N. cunninghamii 1.858 0.830 1.352 1.280 N. gunnii 0.325 0.251 0.269 0.385 P. aspleniifolius 0.989 0.631 0.596 0.842 Poaceae 0.385 0.242 0.407 0.410 Sprengelia 0.156 0.135 0.127 0.209

172 |

6.6.6. Further statistics supporting the REVEALS validation

173 |

a) DOVE LAKE

GPM 6.5 m/s

GPM 3 m/s

LSM WINDY UNSTABLE

LSM UNSTABLE

0 2 4 6 8 10 Average error (cover %)

b) LAKE ST. CLAIR

GPM 6.5 m/s

GPM 3 m/s

LSM WINDY UNSTABLE

LSM UNSTABLE

0 1 2 3 4 5 6 Average error (cover %)

174 |

6.6.7. Individual estimates for plant taxa and additional REVEALS model runs

Figure 6-S6. REVEALS-GPM results and error bars for single taxa. REVEALS was run using 6.5 m/s and PPEs calculated with the same parameters (from Mariani et al., 2016).

175 |

176 |

Figure 6-S8 REVEALS-GPM and REVEALS-LSM results using different sets of atmospheric parameters and PPEs. REVEALS-LSM shows more coherent results. REVEALS-GPM was run under neutral conditions with a wind speed of 6.5 m/s and 3 m/s and the REVEALS- LSM was run using the unstable and windy unstable parameters from Kuparinen et al. (2007). To evaluate the coherence between models and PPEs, REVEALS and PPEs settings were mixed in four other runs: REVEALS-GPM set with a wind speed of 6.5 m/s was run using PPEs calculated at 3 m/s (and vice versa). Likewise, REVEALS-LSM set with unstable conditions was run using PPEs derived using a windy unstable setting (and vice versa). A total of eight model runs were performed in the DISQOVER package for R (Theuerkauf et al., 2016).

177 |

Chapter 7. General discussion and approach limitations

This chapter aims to aggregate the key findings of the research papers presented in Chapters 3-4-5-6 to address the Thesis’ aims outlined in Chapter 1. Chapter 3 showed how changes in the position and strength of the SWW have had major implications for fire occurrence in western Tasmania during the last 30 years and the last millennium (Aim I). Chapter 4 revealed that the SWW and ENSO were important drivers of fire occurrences at sub- millennial/millennial time-scales through the last 12,000 years (Aim II). Chapter 5 presented the application of pollen dispersal models to calculate pollen productivity estimates (PPEs) for thirteen key plant taxa in Tasmania (Aim III – step 1). The results of Chapter 5 were then used in Chapter 6 to reconstruct land-cover changes around Dove Lake (Aim III – step 2), which revealed a landscape dominated by treeless vegetation through the entire Holocene.

The following discussion integrates the findings of the separate chapters to better understand climate-fire-vegetation dynamics in western Tasmania, both in the present and the past. I focus the discussion around two overarching themes: drivers of fire occurrence (short- and long-term), and the impacts of past climatic change and fires on regional land cover. Firstly, I wish to elaborate upon the finding of the SWW being the most important climatic control of fire activity over the short- and long-term in western Tasmania (Chapter 7.1.). In this section I will also present a critical analysis of the limitations of charcoal records. Secondly, I will discuss Holocene land-cover changes in western Tasmania as driven by the interplay of climate, fire and humans. I will then discuss the advantages and limitations of the model- based quantitative vegetation reconstruction approach (presented in Chapter 5 and 6). A summary of the findings of this PhD project follows in Chapter 8 (Conclusions).

178 |

7.1. Regional-scale drivers of fire occurrence in Tasmania

In this Thesis, I investigated the role of SWW in modulating fire occurrence at multiple time scales in the western Tasmanian region. The short-term latitudinal movement of the SWW belt is described using the SAM index (Marshall, 2003). SAM is significantly correlated with rainfall and temperature variability across the extra- tropics of the entire Southern Hemisphere (Garreaud, 2007; Gillett et al., 2006; Hill et al., 2009). Western Tasmania exhibits one of the strongest correlations between rainfall variability and SAM on Earth (Gillet et al., 2006), hence changes in the strength of SWW flow over this region can potentially trigger a variety of rainfall- dependent environmental impacts in this region, such as drought and fire.

The results presented in Chapter 3 highlight a statistically significant relationship between SAM and actual fire occurrence in areas that display a strong negative correlation (<-0.3) between the SAM index and rainfall variability in western Tasmania – i.e. ‘the SAM zone’. My results indicate that a fire season with a high number of events is preceded by an anomalously low rainfall year caused by the sustained southward displacement of the SWW (i.e. positive SAM phase). In line with this finding, enhanced moisture availability due to a northward migration of the SWW (i.e. negative SAM phase) was found to precede fire season years with few events. These findings probably reflect the high moisture content of fuels in this landscape (e.g. rainforests, wet sclerophyll forests), thus prolonged drying may be required to precondition fuels to burn (with a lagged relationship). Indeed, fire occurrences in Tasmanian rainforest were recently found to be dependent on antecedent (previous month) rainfall (Styger and Kirkpatrick, 2015). This discovery of a SWW modulation of the short-term fire activity in western Tasmania introduces a potential key variable that must be considered when projecting and planning for the future of fire management and for the preservation of Tasmania’s fire-sensitive ecosystems.

179 |

I identified a significant increase in fire activity over the last 500 years using a compilation of sedimentary charcoal records spanning the last millennium. This increase occurred in tandem with a sustained positive trend in SAM (Villalba et al., 2012; Abram et al., 2014), indicating fire as a probable consequence of the SWW weakening over this region. Likewise, a spike in fire activity was recorded during the 20th and 21st centuries (Figure 3-4 in Chapter 3), a period when the reconstructed SAM index becomes progressively more positive and exceeds the range of natural variability experienced over the last millennium (Villalba et al., 2012; Abram et al., 2014). Embedded in this dramatic increase in fire activity is the European colonization of Tasmania after 1800 C.E., which produced a drastic change in the fire regimes in this region: from frequent low-intensity fires in buttongrass moorland to an increase in fire intensity and frequency in all vegetation types (Marsden-Smedley, 1998).

The sharp positive trend in SAM observed during the late 20th and early 21st centuries likely occurred in response to the anthropogenic inﬂuence on SWW shifts brought by ozone depletion and global warming (Thompson et al., 2011; Perlwitz, 2011), in turn reducing rainfall amounts and enhancing biomass burning in western Tasmania. Although the future trajectory of SAM is uncertain due to ozone recovery (Polvani et al., 2011; Perlwitz et al., 2013), the recognition of its role in modulating inter-annual fire activity is critical for land managers in view of projected climatic change. In fact, there are projections of increased positive SAM phases under future scenarios featuring higher greenhouse gas concentrations (Thompson et al., 2011), thus, it is crucial to take this climate mode into account when addressing future projections of fire activity in western Tasmania and at a broader extent, the areas of the Southern Hemisphere where climate and fire variability are under SWW influence (e.g. southern Chile; Holz et al., 2017). While these climate controls are relatively well understood, the human dimension of fire remains difficult to predict and will continue to be a source of uncertainty in predictions.

180 |

Over the longer term (last 12,000 years), the research findings presented in Chapter 4 suggest significant coherence of millennial-scale fire activity between western Tasmania and southern South America. My results indicate that long-term changes in strength of SWW flow over these latitudes drove synchronous hydroclimatic fluctuations in Tasmania and southern South America between ~12-6 ka (Figure 7- 1b,c). Fire activity in both regions is strongly correlated (positively) with SAM (Holz and Veblen, 2011; Mariani and Fletcher, 2016; Holz et al., 2017), suggesting a dominant control of SWW over the rainfall and fire variability across the mid- latitudes of the SH. High biomass burning was reconstructed in both regions between 12-9 ka, a finding supported by evidence from around the SH that indicates a prolonged phase of SWW attenuation (Fletcher and Moreno, 2011, 2012; Lamy et al., 2010; Moreno et al., 2010; Wilkins et al., 2013; Lisé-Pronovost et al., 2015). Palaeoclimate reconstructions suggest this period was also warm, as documented by increasing sea-surface temperatures (SSTs) around Tasmania (Sikes et al., 2009; Calvo et al., 2007), reaching maximum values at ~11 ka (Sikes et al., 2009). These findings indicate that the period between 11-9 ka was relatively warm and dry. The mid-Holocene was probably still warm, but wetter, as implied by palaeoclimate records of SSTs and SWW strength. A phase of SWW enhancement over the SH mid- latitudes (Moreno et al., 2010; Fletcher and Moreno, 2011, 2012) occurred in concert with warm conditions at the Murray Canyons (NW of Tasmania), recording maximum SSTs at 7.5 ka (Calvo et al., 2007). Although warmer climates are thought to be more conducive to fire (e.g. Westerling et al., 2011), the wet conditions recorded during this period likely counteracted temperature increase and yielded decreased fire activity in western Tasmania. The strengthening of SWW flow with enhanced moisture delivery at the mid-latitudes of the SH explains the low fire activity in both Tasmania and SSA.

181 |

After ~6 ka, the increased magnitude of regional biomass burning in western Tasmania and the decoupling with southern South America is linked to the onset of ‘modern’ El Niño activity (Figure 7-1d; Donders et al., 2008; McGlone et al., 1992; Moy et al., 2002). El Niño events are associated with drier conditions and increased fire activity in south-east Australia (Mariani et al., 2016; Nicholls and Lucas, 2007) and wetter conditions in South America (Garreaud et al., 2007; Garreaud, 2009). Overall, in this period regional climate conditions were drier (likely due to El Niño influence) and cool, as invoked by regional SSTs records (Calvo et al., 2007; Sikes et al., 2009).

The proposed link between ENSO variability and fire activity in western Tasmania suggests that ENSO phases (and future ENSO projections) may represent a key factor to consider when projecting potential future directions of fire occurrence in

182 |

this region, even though the short-term rainfall variability is strongly linked to SWW shifts (i.e. SAM variability). ENSO is projected to amplify under global warming scenarios (Guilyardi, 2006; Lenton et al., 2008; Power et al., 2013; Cai et al., 2014), threatening not only the survival of the ancient fire-sensitive Tasmanian ecosystems, but also, at a broader extent, Australian water security and population safety (e.g. Bowman et al., 2017). For instance, catastrophic fires impacted the landscape of Tasmania in January 2016, threatening fire-sensitive ecosystems and human lives and properties. These fires occurred during a particularly strong El Niño year (Australian Bureau of Meteorology, http://www.bom.gov.au/climate/enso/enlist ) embedded within the persistent positive trend in the SAM (Marshall, 2001; Thompson et al., 2011), highlighting the importance of both climate modes in controlling biomass burning in the region and emphasizing the necessity for further investigations about their interplay and past dynamics.

With the results from the analyses of short- and long-term fire activity presented in this Thesis, I was able to demonstrate a tight link between fire occurrences and broad-scale climatic features (SWW and ENSO). This finding represents an empirical proof for the hypothesized climate-modulation of fire activity western Tasmania, where biomass (fuel) is perennially abundant and humans are a constant source of ignition (e.g. Cochrane, 2003; Bradstock, 2010; Moritz et al., 2012; Pausas and Ribeiro, 2013; McWethy et al., 2013).

7.1.1 Limitations of charcoal records and multi-site compilations

The findings discussed in Chapter 3, 4 and 7.1. are mostly derived from the analysis of sedimentary charcoal data. However, interpreting charcoal records may not be straightforward in some instances. As discussed in Chapter 2, charcoal accumulation rates in a lake are thought to depend on (1) the characteristics of the fire, (2) the characteristics of the fuels burning and (3) the processes delivering charcoal particles to the lake (Whitlock and Larsen, 2001).

183 |

Single-site charcoal records

In this Thesis, a total of 14 macroscopic-CHAR records from western Tasmania are presented in Chapters 3, 4 and 6. The records used for the regional charcoal compilations in Chapter 3 and 4 present overall similar trends consistent with the output regional compilation (except Lake Osborne and Square Tarn; Figure 4-2). However, at Dove Lake (Chapter 6), the macroscopic-CHAR record appears to be uncoupled from the regional shifts in fire activity recorded in the ‘SAM zone’ (Figure 7-2b,c,d). In fact, at this location, high macroscopic-CHAR influx is observed during a phase (10-5 ka) of generally high biomass availability (high relative forest cover) and relatively wet climate.

The evidence of higher charcoal influx coupled with higher biomass availability finds analogies in studies from Kenya (Colombaroli et al., 2014), Spain (Gil-Romera et al., 2014) and Arizona (Brunelle et al., 2010), where biomass-limitation was pointed out as the most plausible explanation. However, the Holocene of western Tasmania was characterised by a minimum forest cover of 30% (see Chapter 6), thus a shortage of fuel availability around Dove Lake is unlikely. In a recent case-study from Tasmania, Lake Osborne, a divergence between macroscopic- and microscopic- CHAR was interpreted as a particle size variation in response to biomass availability changes: a shift from high-biomass rainforests to low-biomass sclerophyll vegetation corresponded to a shift from high macroscopic-CHAR to high microscopic-CHAR (Fletcher et al., 2014a). At Dove Lake, the microscopic and macroscopic CHAR records do not show such variation, but they are instead coupled, probably reflecting a common source of these particles. The microscopic-CHAR at Dove Lake follows the overall trend and the peaks in the macroscopic-CHAR curve. This finding questions the widely believed assumption of charcoal particle size being an indicator of charcoal source distance (Whitlock and Larsen, 2001). This assumption was in fact recently questioned by an analysis of modern lake sediment traps in Europe, where only macroscopic-CHAR particles > 600 µm were found to be indicative of local fire activity (Adolf et al., in press).

184 |

185 |

The charcoal records at Dove Lake may then indicate that (1) the fire history at Dove Lake is the product of localised burning (within catchment); (2) fires occurring in high biomass settings (e.g. higher rainforest cover between 10 and 5 ka) produces greater amounts of charcoal particles compared to more sclerophyll-rich open landscapes; (3) rain episodes after fire events produced high in-wash of sediment and charcoal particles to the lake (high charcoal = high rainfall); and (4) stochasticity and/or variability in anthropogenic influence on fire occurrence and spatial distribution.

The fact that the trends in rainforest cover from Dove Lake follows the shifts in the regional charcoal influx from western Tasmania (Figure 7-2a,b), instead of the local macroscopic-CHAR record, supports the contention that the pollen signal reflects regional (extra-catchment) vegetation dynamics and the local importance of the fire activity reconstruction at Dove Lake (Figure 7-2c). Another possible explanation lies in biomass availability changes: the amount of surface fuels (e.g. leaf and woody litter) is a function of tree cover (Walker, 1981) and thus is higher in rainforests and dense wet sclerophyll forests compared to more open vegetation types, such as dry sclerophyll forests and woodlands (e.g. Bradstock. 2010; Bradstock et al., 2014).

These possible explanations are not mutually exclusive; in fact, charcoal produced during a fire in rainforest within the catchment, will likely be delivered to the lake during post-fire rain events, which are frequent in Tasmania (Bridle et al., 2003; Pemberton, 1988). Unfortunately, at the current stage of the research, it is not possible to have a full understanding of the charcoal production and deposition processes at Dove Lake. To date there is no available literature dealing with a quantification and particle-size distribution of charcoal produced by rainforest, sclerophyll or moorland vegetation in Tasmania. Further work at Dove Lake involving multiple proxies (e.g. geochemical analyses and granulometric profiles) is needed to investigate the possibility of secondary charcoal deposition. Investigations are also needed to address the issue of biomass changes in modulating charcoal production and charcoal particle size variability.

186 |

Multi-site charcoal compilations

For broad-scale reconstructions of fire activity, multi-site charcoal compilations have been employed for different spatio-temporal extents (Power et al., 2008; Mooney et al., 2011; Marlon et al., 2013; Mariani and Fletcher, 2017). The advantage of multi-site compilations is their mitigation against the influence of local-scale factors, allowing regional climatic trends to be isolated from palaeoclimate data (Power et al., 2008). Indeed, studies based on only one location or record may be strongly biased by local geographic factors (e.g. topography, wind patterns, charcoal taphonomy) and human influence in some cases.

Prior to my study, a charcoal compilation for the all of Australasia was carried out by Mooney et al. (2011) to synthesize temporal and spatial variability in fire regimes throughout the Late Quaternary. In that study, 223 sedimentary charcoal records from a large variety of sites featuring distinct bioclimatic characteristics (i.e. biomes, dominant climatic features) were combined to reconstruct a continental-scale fire history of the last 70 kyrs. A limitation of that approach was that it did not take into account the highly complex climate-fire-vegetation feedbacks in the whole region. Another important flaw of this approach is the low-resolution sampling in many of the sites selected, coupled with large errors in chronology, which likely altered the output synthesis curves to a substantial degree.

The study presented in Chapter 4 overcomes most of these issues by compiling high- resolution and well-dated charcoal records from within the same bioclimatic region (superhumid zone within western Tasmania). However, this meta-analysis does not take into account local patterns of fire-biomass-soil feedbacks, given the current lack of pollen and geochemical data from the other sites within the region (research currently underway) and the impossibility of accomplishing these additional time- consuming tasks within the time-frame of this PhD project.

187 |

7.2. Climate- and fire- driven land-cover changes in western Tasmania

The application of REVEALS to the pollen data collected from Dove Lake allowed a robust quantification of land-cover changes by correcting for pollen production and dispersal biases (Chapter 6). Thus far, theories of landscape evolution in western Tasmania centred on the semi-quantitative interpretation of pollen proportions between rainforest and treeless moorland taxa (Macphail, 1979; Colhoun et al., 1996; Pickett et al., 2004; Fletcher and Thomas, 2007a,b; 2010a,b). Due to the high abundance of over-represented rainforest taxa in the pollen spectra, such as Nothofagus cunninghamii and Phyllocladus aspleniifolius, moorland vegetation has gone undetected in attempts to reconstruct the Tasmanian landscape from pollen data (Macphail, 1979; Colhoun et al., 1996; Pickett et al., 2004; Fletcher and Thomas, 2007; 2010a,b). My results confirmed the hypothesis proposed by Fletcher and Thomas (2007a,b; 2010a,b), who identify western Tasmania as an ancient cultural landscape. This landscape evolution model considers buttongrass moorland as inherited from the Late Pleistocene and maintained by Aboriginal burning, as opposed to the former notion of a more recent (Late Holocene) origin of this vegetation due to coupled soil-vegetation feedbacks and anthropogenic burning (Macphail, 1979; Colhoun et al., 1996).

The regional analysis conducted by Colhoun (1996) using multiple sites across Tasmania shows high proportions of rainforest pollen through Early and mid- Holocene (60-80%), which was interpreted as post-glacial climate-driven rainforest expansion with a peak development at ~7 ka (Figure 7-3). However, Colhoun (1996) recalculated pollen sums by including only 10 regionally important pollen/spore taxa in Tasmania (Nothofagus cunninghamii, Phyllocladus aspleniifolius, Eucalyptus, Casuarina, Pomaderris, Dicksonia, Dodonaea, Poaceae, Asteraceae, Amaranthaceae) and excluding indicators of local vegetation (i.e. moorland - Gymnoschoenus sphaerocephalus). With these recalculated pollen spectra, he argued that the landscape of western Tasmania was dominated by forest during this period, possibly

188 | reinforcing a false dichotomy between the two models of landscape evolution in this region.

Mirroring the results of the compilation by Colhoun (1996), high rainforest pollen abundance was recorded at Dove Lake during Early and mid-Holocene, ranging between 50% and 70% and reaching maximum values between ~8-5 ka (Figure 7-3). However, my quantitative vegetation reconstructions showed relatively stable Holocene landscape-scale dynamics with rainforest cover reaching maximum ~40% between ~8-5 ka, thus revealing an overall dominance of treeless moorland vegetation through this period (Figure 7-3). This finding confirms the cultural landscape model proposed by Fletcher and Thomas (2010a,b), which states that anthropogenic burning through the Late Glacial deflected the development of typical interglacial vegetation (i.e. rainforest), instead facilitating the expansion of moorland. Corroborating evidence of human influence on Tasmanian landscape evolution is found in the higher charcoal content of Holocene records compared to previous interglacials (Fletcher and Thomas, 2010a), probably indicating the introduction of an ignition source that was not present then. Burning through the transition from the Last Glacial Maximum to the Holocene prevented rainforest expansion and favoured the spread of moorland vegetation.

189 |

Although the conclusion of my study and Colhoun (1996) are divergent in terms of dominant vegetation types (moorland vs rainforest, respectively), both studies highlight a mid-Holocene forest expansion. The period with maximum rainforest expansion around Dove Lake (~8-5 ka) corresponds with minimum fire activity recorded in western Tasmania (Figure 7-2a,b), coinciding with a phase of high moisture availability in this region. This finding supports the theories of Macphail (1979) and Colhoun (1996), who contended that climate was the first-order control of post-glacial vegetation changes in western Tasmania. My research reconciles the two models of landscape evolution for this region: while humans shaped the distribution

190 |

of open vegetation such as moorland, climate played a key role in determining Holocene vegetation structure and composition.

The sharp decline in forest cover since ~4.2 ka is, considering age uncertainties, synchronous with the start of the increase in fire regimes observed across the western Tasmanian region (Figure 7-2a,b). Even though fire activity in the western Tasmanian region has shown a tight coupling with SAM during the last 1000 years (Mariani and Fletcher, 2016) and SWW are considered to be responsible for multi- millennial moisture shifts during the Holocene (e.g. Rees et al., 2015; Fletcher et al., 2015; Mariani and Fletcher, 2017), intensified ENSO has been proposed as a possible driver of fire regime shifts in this region during at least the last 6 ka (Rees et al., 2015; Fletcher et al., 2015; Beck et al., 2016; Mariani and Fletcher, 2017). Based on the results presented in Chapter 4, confirming a coupling of millennial-scale El Niño events and regional fire activity in western Tasmania, I believe that the forest decline observed since ~4.2 ka around Dove Lake is probably linked with increased El Niño activity, resulting in a moisture reduction in this region, which favoured the expansion of fire-prone sclerophyll vegetation (e.g. Eucalyptus; Figure 7-2d,e).

7.2.1. Advancements and limitations of model-based quantitative vegetation reconstructions

The work presented in Chapter 6 constitutes the first REVEALS application in Australia and the Southern Hemisphere. Prior to this, only the biomization method (Prentice et al., 1996) was attempted to achieve a quantitative vegetation reconstruction on the Australian continent (Pickett et al., 2004). However, this approach proved to perform inefficiently for wet sclerophyll forests and moorland vegetation because of its inability to overcome pollen production and dispersal biases. The cause for this failure was especially attributed to Eucalyptus species, which constitute the most important canopy component across much of Australia and can produce great quantities of pollen that can be transported by the wind (e.g. Dodson, 1983; Kershaw and Strickland, 1990). In this work, Eucalyptus pollen was

191 |

indeed found to be omnipresent in modern pollen samples from a variety of biomes (Pickett et al., 2004), thus impeding correct biome classification. In support of the explanation proposed by Pickett et al. (2004) for the limited success of this approach, the results presented in Chapter 6 show a clear alignment of Eucalyptus alongside the most prolific wind-pollinated producers in Tasmania (Phyllocladus aspleniifolius and Nothofagus cunninghamii).

The results presented in Chapter 5 prove the adaptability of REVEALS, a wind- based modelling approach, in areas where animal pollination occurs alongside wind pollination, such as western Tasmania and, by extension, the Australian continent. This achievement paves the way to the application of the same modelling approach to the rest of Australia, where there is a critical over-representation of some canopy trees, including Casuarina and Eucalyptus (Dodson, 1983; Kershaw and Strickland, 1990; Black et al., 2006). Moreover, the application of such models is pivotal to achieve a quantification of land-cover that takes into account human agency, since anthropogenic landscapes are usually palynologically silent (Gaillard et al., 1998; Hellman et al., 2009; Gaillard et al., 2010).

My results showed that the quality of REVEALS performance is strictly dependent on the choice of the dispersal model used (Theuerkauf et al., 2013). The application of REVEALS in Germany yielded more realistic vegetation estimates using the Lagrangian Stochastic Model compared to the Gaussian Plume Model (Theuerkauf et al., 2013; Theuerkauf et al., 2015). Similarly, the results presented in Chapter 5 and Chapter 6 suggest that this modelling framework can be more successfully applied to Tasmanian pollen records when using the LSM. In this regard, there are important international implications because (1) the results of the PPE calculations from the two pollen dispersal models were divergent for some important plant taxa in our case-study (e.g. rainforest trees) and (2) the validation using modern plant cover around two validation sites showed important differences. In both cases, the LSM showed more realistic results. Nevertheless, the Gaussian plume model is in fact the most widely used pollen dispersal model in Europe and current quantiﬁcations of

192 |

the majority of past land cover reconstructions are based on this model (e.g. Mazier et al., 2012; Abraham et al., 2012; Trondman et al., 2015). Future work should aim to incorporate a test of both pollen dispersal models to find the optimal approach and avoid poor reconstruction results.

Even though the results presented in Chapter 6 suggest that the REVEALS model can be successfully applied in western Tasmania, there are critical limitations of this modelling approach:

• The scarcity of large (>1 km2) natural lakes for regional vegetation reconstructions. Only 5.3% of lakes in Tasmania are larger than 1 km2, whereas less than 1% of lakes in western Tasmania exceed this size. However, model applicability can be tested by substituting single large lakes with a combination of multiple small sites, as proposed in southern Sweden (Trondman et al., 2016). In this regard, a potential issue would be site selection criteria. Fossil pollen deposited in small lakes normally registers vegetation changes within a small radius, potentially only limited to the lake catchment. At this small scale, fire occurrence is not only modulated by climate, but also fine-scale topographic characteristics, vegetation density and/or human activity (e.g. Pitkanen et al., 2002; Clark and Royall, 1996). Thus, ideally, to avoid large errors in the reconstruction outputs, only sites with similar pollen records should be considered for this application. In the work conducted by Trondman et al. (2016), only similarly-sized lakes were recommended for this compilation analysis. As per the charcoal multi-site compilation limitations, another potential problem is the age accuracy of the records to be combined: robust age-depth models (depending on the analysed time frames) must be obtained for all the fossil records before attempting this approach.

• Variability in wind direction was not taken into account. In line with the current stage of the REVEALS framework development, the work conducted in Chapter 5 and 6 assumes uniform wind blowing evenly from all directions.

193 |

At present, the only way to incorporate this information within the modelling would be during the pollen productivity estimate calculations, by weighting the vegetation buffers according to fine-scale wind frequency information at each sample location (e.g. 27 sample sites in Chapter 5). However, wind direction data are available only for limited locations in Tasmania (Australian Bureau of Meteorology: http://www.bom.gov.au), with only two stations located in western Tasmania. At Cape Sorell (42.2°S; 145.17°E), winds blow mainly from the N and SW, whereas at Strathgordon (42.2°S; 145.17°E) the dominant wind directions are N and NW. Given the high topographic complexity of western Tasmania, basing a wind-weighted vegetation data correction based only on the information from these two locations is likely to produce major biases in the results. Further, given the month-to-month variability in wind patterns, another issue would be represented by the choice of the reference period, which should match the flowering season of the selected plant taxa (or at least the majority of them). No implementation of variable wind direction parameters has currently been made for REVEALS, thus reducing the potential benefits of a correction effort at this stage.

• The potential inapplicability of derived PPEs to pre-Holocene pollen records. Due to the high variability of Pleistocene climate, oscillating between glacial and interglacial periods, plant taxa likely suffered variations in pollen production due to variable thermal/rainfall conditions (e.g. Autio and Hicks, 2004). The pollen productivity estimates obtained in Chapter 5 are based on the modern pollen-vegetation assemblages and they are assumed to be constant through the Holocene period given the absence of extreme climate fluctuations. At the current stage of the research, the application of these PPEs to pre-Holocene records is not recommended. Further work directed to calibrate pollen-vegetation relationships under different climatic settings should be attempted, for example by examining pollen spectra and vegetation cover along elevational gradients.

194 |

• The vegetation survey method is only restricted to a small portion of the landscape. Due to logistical and technical reasons, only a relatively small portion of the vegetation can be effectively surveyed on the field, and a large proportion of the landscape is normally extrapolated using GIS (e.g. von Stedingk et al., 2008; Abraham et al., 2012; Bunting et al., 2013). This extrapolation is based on the assumption that the plant cover in the surveyed plots for each vegetation type is constant across the study area. The more accurate and extensive is the field survey, the more reliable are the pollen productivity estimates obtained by applying pollen dispersal models (Bunting et al., 2013). To overcome this issue, in the study presented in Chapter 5, different replicate quadrats (minimum of 3) of the same vegetation types were surveyed in order to minimize the errors when calculating the average plant cover for ring data extrapolations using GIS.

To conclude, further work on several aspects of the research topics outlined in this Thesis is still needed to achieve a better understanding of fire-climate-vegetation dynamics. Despite these limitations, the research conducted during this PhD project represents a major advancement of the discipline in Australia with broader implications at an international level.

195 |

Chapter 8. Conclusions

The work conducted during this PhD project greatly improved understanding of (1) the short- and long-term climate drivers of fire activity (Aims I and II) and (2) the long-term landscape dynamics in response to climate and fire regime changes in western Tasmania (Aim III). Overall, the findings of this work constitute an important advancement in the understanding of the linkages between fire and vegetation under variable climatic conditions that characterised the Earth system during the last 12,000 years.

The results presented in Chapter 3 reveal the Southern Annular Mode (SAM - an index describing the modern position and strength of SWW) as the main climatic driver of inter-annual ﬁre occurrence in western Tasmania (Aim I). Moreover, the multi-site charcoal compilation for the last 1,000 years from the same region reveals a close match between western Tasmanian ﬁre activity and two proxy-based SAM reconstructions, suggesting a tight coupling with SAM at multiple scales of time in this landscape.

Due to the limited temporal scale of historical records, which are unlikely to register the full spectrum of fire activity variability, longer-term analyses of fire history were undertaken using sedimentary charcoal records. These analyses revealed the linkages between fire occurrences in western Tasmania and the leading climate features of the mid-latitudes of the Southern Hemisphere: ENSO and SWW (Aim II). This effort is particularly important under a warming world scenario in which fire activity is predicted to increase in magnitude and frequency, especially within temperate regions, such as western Tasmania. The multi-site compilation of 13 sedimentary charcoal records from this region has allowed the identification of clear long-term shifts in biomass burning driven by the dynamic interplay between the SWW and ENSO systems. A period of high fire activity was observed between from 12 to 8 ka, declining to a minimum between ca. 7-5 ka. Over the late Holocene, a sharp increase since ca. 4 ka was observed and during the last 0.3 kyr, the entire

196 |

range of Holocene fire variability was exceeded. The persistent, yet oscillating, increase in fire activity observed during the late Holocene coincides with the timing of increased tropical ENSO variability.

In order to understand the vegetation dynamics in relation to the climatic changes and fire regime shifts described above, fossil pollen analysis was coupled with advanced mathematical modelling (Aim III). The results presented in Chapter 6 constitute the first quantitative reconstruction of land-cover in the Southern Hemisphere, proving the applicability of wind-based dispersal models in landscapes where animal pollination is prevalent. Results from the PPEs calculations (Chapter 5) and the statistical analysis of modern pollen and plant cover estimates (Chapter 6) showed a more realistic performance of REVEALS estimates based on the Lagrangian Stochastic Model when compared to the Gaussian Plume Model. This finding has major implications at an international scale, given the widespread use of the Gaussian Plume Model in vegetation reconstructions across the Northern Hemisphere. The successful application of these pollen dispersal models allowed a more robust quantification of land-cover changes compared to classical pollen analysis, proving the biases inherent in previous interpretations of pollen spectra from this region. Indeed, from the results of this approach, it is evident that the landscape of western Tasmania has been persistently open since the last glacial cycle, supporting the notion that this region represents an ancient cultural landscape. This finding has important implications in enforcing conservation strategies for the threatened Tasmanian buttongrass moorland (Balmer et al., 2007; Gallego-Sala and Prentice, 2013).

The results of this PhD project allowed the achievement of a better understanding of climate-fire-vegetation interactions in the western Tasmanian region through the last 12,000 years. The linkages found between climatic change, fire history and vegetation have major implications in the development of suitable fire management and ecosystem conservation strategies, which are a key priority of present-day societies facing the imminent consequences of global warming and fire activity intensification.

197 |

References

Aaby, B., Digerfeldt, G., 1986. 1986: Sampling techniques for lakes and bogs. In Berglund, BE, editor, Handbook of Holocene Palaeoecology and Palaeohydrology, Chichester: John Wiley, 181–194.

Aboriginal, A.I.o., Studies, T.S.I., Horton, D., 1994. The encyclopaedia of Aboriginal Australia: Aboriginal and Torres Strait Islander history, society and culture. Aboriginal Studies Press.

Abraham, V., Kozáková, R., 2012. Relative pollen productivity estimates in the modern agricultural landscape of Central Bohemia (Czech Republic). Review of Palaeobotany and Palynology 179, 1-12.

Abram, N.J., Mulvaney, R., Vimeux, F., Phipps, S.J., Turner, J., England, M.H., 2014. Evolution of the Southern Annular Mode during the past millennium. Nature Climate Change.

Adolf C., Wunderle S., Colombaroli D., Weber H., Gobet E., Heiri O., van Leeuwen J.F.N., Bigler C., Connor S.E., Gałka M., La Mantia T., Makhortykh S., Svitavská-Svobodová H., Vannière B., Tinner W. In press. The sedimentary and remote-sensing reflection of biomass burning in Europe. Global Ecology and Biogeography.

Alvain, S., Le Quéré, C., Bopp, L., Racault, M.-F., Beaugrand, G., Dessailly, D., Buitenhuis, E.T., 2013. Rapid climatic driven shifts of diatoms at high latitudes. Remote Sensing of Environment 132, 195- 201.

Andersen, S., 1970. The Relative Pollen Productivity and Pollen Representation of North European Trees, and Correction Factors for Tree Pollen Spectra. Danmarks Geologiske Undersogelse, II 96, 1-96.

Andersen, S.T., 1974. Wind conditions and pollen deposition in a mixed deciduous forest: I. Wind conditions and pollen dispersal. Grana 14, 57-63.

Andersen, T.B., 1991. A model to predict the beginning of the pollen season. Grana 30, 269-275.

Anderson, N.J., Bugmann, H., Dearing, J.A. and Gaillard, M.-J., 2006. Linking palaeoenvironmental data and models to understand the past and to predict the future. Trends in Ecology and Evolution, 21, 696-704.

Andreae, M.O., 1991. Biomass burning: its history, use, and distribution and its impact. Global Biomass Burning: Atmospheric, Climatic, and Biospheric Implications [JS Levine (Ed.)]. Cambridge, MA, MIT Press.

Appleby, P., 2008. Three decades of dating recent sediments by fallout radionuclides: a review. The Holocene 18, 83-93.

Appleby, P., Oldfield, F., 1978. The calculation of lead-210 dates assuming a constant rate of supply of unsupported< sup> 210 Pb to the sediment. Catena 5, 1-8.

Armstrong, J., 1979. Biotic pollination mechanisms in the Australian flora—a review. New Zealand journal of botany 17, 467-508.

Autio, J., Hicks, S., 2004. Annual variations in pollen deposition and meteorological conditions on the fell Aakenustunturi in northern Finland: potential for using fossil pollen as a climate proxy. Grana 43, 31-47.

198 |

Balmer, J., 2007. Introducing'Conservation and Management of Buttongrass Moorland'. Australasian Plant Conservation: Journal of the Australian Network for Plant Conservation 16, 1.

Bard, E., Rostek, F., Ménot-Combes, G., 2004. Radiocarbon calibration beyond 20,000 14 C yr BP by means of planktonic foraminifera of the Iberian Margin. Quat. Res. 61, 204-214.

Beck, K.K., Fletcher, M.-S., Gadd, P., Heijnis, H. and Jacobsen, G., 2017. An early onset of ENSO influence in the extra-tropics of the southwest Pacific inferred from a 14,600 year high resolution multi-proxy record from Paddy's Lake, northwest Tasmania. Quaternary Science Reviews, 157, 164- 175.

Berger, A.L., 1978. Long-term variations of daily insolation and Quaternary climatic changes. Journal of the Atmospheric Sciences 35, 2362-2367.

Berglund, B., 1986. The cultural landscape in a long-term perspective. Methods and theories behind the research on land-use and landscape dynamics. Striae 24, 79-87.

Bertrand, S., Araneda, A., Vargas, P., Jana, P., Fagel, N., Urrutia, R., 2012. Using the N/C ratio to correct bulk radiocarbon ages from lake sediments: Insights from Chilean Patagonia. Quaternary Geochronology 12, 23-29.

Birks, H.H., Birks, H.J.B., 2006. Multi-proxy studies in palaeolimnology. Vegetation history and Archaeobotany 15, 235-251.

Birks, H.J.B., Berglund, B.E., 2017. One hundred years of Quaternary pollen analysis 1916–2016. Vegetation History and Archaeobotany, 1-39.

Birks, J.B., Lotter, A.F., Juggins, S. and Smol, J.P., 2012. Tracking environmental change using lake sediments: data handling and numerical techniques. Springer, Dordrecht.

Björck, S., Wohlfarth, B., 2002. 14C chronostratigraphic techniques in paleolimnology, Tracking environmental change using lake sediments. Springer, pp. 205-245.

Blaauw, M., 2010. Methods and code for ‘classical’ age-modelling of radiocarbon sequences. Quaternary Geochronology 5, 512-518.

Blaauw, M., Christen, J., 2013. Bacon manual–v2. 2.: 1–11.

Black, M., Mooney, S., Attenbrow, V., 2008. Implications of a 14,200 year contiguous fire record for understanding human—climate relationships at Goochs Swamp, New South Wales, Australia. The Holocene 18, 437-447.

Black, M., Mooney, S., Martin, H., 2006. A> 43,000-year vegetation and fire history from Lake Baraba, New South Wales, Australia. Quaternary Science Reviews 25, 3003-3016.

Black, M.P., Mooney, S.D., Haberle, S.G., 2007. The fire, human and climate nexus in the Sydney Basin, eastern Australia. The Holocene 17, 469-480.

Blais, J.M., Kalff, J., 1995. The influence of lake morphometry on sediment focusing. Limnology and Oceanography 40, 582-588.

199 |

Blarquez, O., Vannière, B., Marlon, J.R., Daniau, A.-L., Power, M.J., Brewer, S., Bartlein, P.J., 2014. paleofire: an R package to analyse sedimentary charcoal records from the Global Charcoal Database to reconstruct past biomass burning. Computers and Geosciences 72, 255-261.

Bond, W.J., Woodward, F.I. and Midgley, G.F., 2005. The global distribution of ecosystems in a world without fire. New Phytologist, 165, 525-537.

Bowman, D.M., Williamson, G.J., Abatzoglou, J.T., Kolden, C.A., Cochrane, M.A., Smith, A.M., 2017. Human exposure and sensitivity to globally extreme wildfire events. Nature Ecology and Evolution 1, 0058.

Bowman, D.M.J.S. and Brown, M., 1986. Bushfires in Tasmania: a botanical approach to anthropological questions. Archaeology in Oceania, 21, 166-171.

Bowman, D.M.J.S. and Jackson, W.D., 1981. Vegetation succession in southwest Tasmania. Search, 12, 358-362.

Bowman, D.M.J.S., 1998. The impact of Aboriginal landscape burning on the Australian biota. New Phytologist, 140, 385-410.

Bowman, D.M.J.S., 2000. Australian rainforests: islands of green in a land of fire, First ed. Cambridge University Press, Cambridge.

Bowman, D.M.J.S., Balch, J., Artaxo, P., Bond, W.J., Cochrane, M.A., D’Antonio, C.M., DeFries, R., Johnston, F.H., Keeley, J.E., Krawchuk, M.A., 2011. The human dimension of fire regimes on Earth. Journal of Biogeography 38, 2223-2236.

Bowman, D.M.J.S., Balch, J.K., Artaxo, P., Bond, W.J., Carlson, J.M., Cochrane, M.A., D'Antonio, C.M., DeFries, R.S., Doyle, J.C., Harrison, S.P., Johnston, F.H., Keeley, J.E., Krawchuk, M.A., Kull, C.A., Marston, J.B., Moritz, M.A., Prentice, I.C., Roos, C.I., Scott, A.C., Swetnam, T.W., van der Werf, G.R., Pyne, S.J., 2009. Fire in the Earth System. Science 324, 481-484.

Bowman, D.M.J.S., Wood, S.W., 2009. Fire driven land cover change in Australia and W.D. Jackson's theory of the fire ecology of southwest Tasmania, in: Cochrane, M.A. (Ed.), Tropical fire ecology: climate change, land use and ecosystem dynamics. Springer-Praxis, Heidelberg pp. 87-111.

Box, G.E., Cox, D.R., 1964. An analysis of transformations. Journal of the Royal Statistical Society. Series B (Methodological), 211-252.

Bradshaw, R., Webb, T., 1985. Relationships between contemporary pollen and vegetation data from Wisconsin and Michigan, USA. Ecology 66, 721-737.

Bradshaw, R.H., 1981. Modern pollen-representation factors for woods in south-east England. The Journal of Ecology, 45-70.

Bradstock, R., Penman, T., Boer, M., Price, O., Clarke, H., 2014. Divergent responses of fire to recent warming and drying across south‐eastern Australia. Global change biology 20, 1412-1428.

Bradstock, R.A., 2010. A biogeographic model of fire regimes in Australia: current and future implications. Global Ecology and Biogeography 19, 145-158.

200 |

Bradstock, R.A., Williams, J.E., Gill, M.A., 2002. Flammable Australia: the fire regimes and biodiversity of a continent. Cambridge University Press.

Bridle, K., Cullen, P., Russell, M., 2003. Peatland hydrology, fire management and Holocene fire regimes in southwest Tasmanian blanket bogs. Nature Conservation Branch, Department of Primary Industries, Water and Environment, Hobart, Tasmania.

Brock, F., Higham, T., Ditchfield, P., Ramsey, C.B., 2010. Current pretreatment methods for AMS radiocarbon dating at the Oxford Radiocarbon Accelerator Unit (ORAU). Radiocarbon 52, 103-112.

Broström, A., Nielsen, A.B., Gaillard, M.-J., Hjelle, K., Mazier, F., Binney, H., Bunting, J., Fyfe, R., Meltsov, V., Poska, A., 2008. Pollen productivity estimates of key European plant taxa for quantitative reconstruction of past vegetation: a review. Vegetation history and archaeobotany 17, 461-478.

Broström, A., Sugita, S., Gaillard, M.-J., 2004. Pollen productivity estimates for the reconstruction of past vegetation cover in the cultural landscape of southern Sweden. The Holocene 14, 368-381.

Broström, A., Sugita, S., Gaillard, M.-J., Pilesjö, P., 2005. Estimating the spatial scale of pollen dispersal in the cultural landscape of southern Sweden. The Holocene 15, 252-262.

Brown, M.J. and Podger, F.D., 1982a. On the apparent anomaly between observed and predicted percentages of vegetation types in south-west Tasmania. Australian Journal of Ecology, 7, 203-295.

Brown, M.J. and Podger, F.D., 1982b. Floristics and fire regimes of a vegetation sequence from sedgeland-heath to rainforest at Bathurst Harbour, Tasmania. Australian Journal of Botany, 30, 659- 676.

Brunelle, A., Minckley, T., Blissett, S., Cobabe, S., Guzman, B., 2010. A 8000 year fire history from an Arizona/Sonora borderland ciénega. Journal of Arid Environments 74, 475-481. ∼ Brush, G.S., Brush, L.M., 1972. Transport of pollen in a sediment-laden channel; a laboratory study. American Journal of Science 272, 359-381.

Bunting, M.J. and Middleton, R., 2009. Equifinality and uncertainty in the interpretation of pollen data: the Multiple Scenario Approach to reconstruction of past vegetation mosaics. The Holocene, 19, 799-803.

Bunting, M.J., 2003. Pollen-vegetation relationships in non-arboreal moorland taxa. Review of Palaeobotany and Palynology 125, 285-298.

Bunting, M.J., Armitage, R., Binney, H.A., Waller, M., 2005. Estimates of ‘relative pollen productivity’ and ‘relevant source area of pollen’ for major tree taxa in two Norfolk (UK) woodlands. The Holocene 15, 459-465.

Bunting, M.J., Farrell, M., Broström, A., Hjelle, K., Mazier, F., Middleton, R., Nielsen, A.B., Rushton, E., Shaw, H., Twiddle, C., 2013. Palynological perspectives on vegetation survey: a critical step for model-based reconstruction of Quaternary land cover. Quaternary Science Reviews 82, 41-55.

201 |

Bunting, M.J., Gaillard, M.-J., Sugita, S., Middleton, R., Broström, A., 2004. Vegetation structure and pollen source area. The Holocene 14, 651-660.

Bunting, M.J., Hjelle, K.L., 2010. Effect of vegetation data collection strategies on estimates of relevant source area of pollen (RSAP) and relative pollen productivity estimates (relative PPE) for non- arboreal taxa. Vegetation History and Archaeobotany 19, 365-374.

Cai, W., Borlace, S., Lengaigne, M., Van Rensch, P., Collins, M., Vecchi, G., Timmermann, A., Santoso, A., McPhaden, M.J., Wu, L., 2014. Increasing frequency of extreme El Niño events due to greenhouse warming. Nature climate change 4, 111-116.

Cai, W., Cowan, T., Raupach, M., 2009. Positive Indian Ocean dipole events precondition southeast Australia bushfires. Geophysical Research Letters 36.

Cai, W., Van Rensch, P., Borlace, S., Cowan, T., 2011. Does the Southern Annular Mode contribute to the persistence of the multidecade‐long drought over southwest Western Australia? Geophysical Research Letters 38.

Calcote, R., 1995. Pollen source area and pollen productivity: evidence from forest hollows. Journal of Ecology, 591-602.

Calvo, E., Pelejero, C., De Deckker, P., Logan, G.A., 2007. Antarctic deglacial pattern in a 30 kyr record of sea surface temperature offshore South Australia. Geophysical Research Letters 34.

Carcaillet, C., Almquist, H., Asnong, H., Bradshaw, R., Carrion, J., Gaillard, M.-J., Gajewski, K., Haas, J., Haberle, S., Hadorn, P., 2002. Holocene biomass burning and global dynamics of the carbon cycle. Chemosphere 49, 845-863.

Chamberlain, A., 1975. The movement of particles in plant communities. Vegetation and the Atmosphere 1, 155-203.

Chappell, J. A. and Thom, B. G., 1977. Sea Levels and Coasts. Sunda and Sahul: prehistoric studies in south eastern Asia, Melanesia and Australia (eds. J. Allen, J. Golson and R. Jones) pp. 274-291. Academic Press, London.

Clark, J.S., 1988. Stratigraphic charcoal analysis on petrographic thin sections: application to fire history in northwestern Minnesota. Quat. Res. 30, 81-91.

Clark, J.S., Royall, P.D., 1996. Local and regional sediment charcoal evidence for fire regimes in presettlement north-eastern North America. Journal of Ecology, 365-382.

Cleveland, W.S., 1979. Robust locally weighted regression and smoothing scatterplots. Journal of the American statistical association 74, 829-836.

Cochrane, M.A., 2003. Fire science for rainforests. Nature 421, 913-919.

Colhoun, E.A. and van der Geer, G., 1998. Pollen analysis of 0-20m at Darwin Crater, western Tasmania, Australia. International Project of Paleolimnology and Late Cenozoic Climate, 11, 68-89.

Colhoun, E.A., 1996. Application of Iversen's glacial-interglacial cycle to interpretation of the late last glacial and Holocene vegetation history of western Tasmania. Quaternary Science Reviews 15, 557- 580.

202 |

Colhoun, E.A., 2000. Vegetation and climate during the last interglacial–glacial cycle in western Tasmania, Australia. Palaeogeography, Palaeoclimatology, Palaeoecology 155, 195-209.

Colhoun, E.A., Pola, J.S., Barton, C.E. and Heijnis, H., 1999. Late Pleistocene vegetation and climate history of Lake Selina, western Tasmania. Quaternary International, 57-58, 5-23.

Colhoun, E.A., Shimeld, P.W., 2012. Late-Quaternary vegetation history of Tasmania from pollen records. Peopled Landscapes: Archaeological and Biogeographic Approaches to Landscapes. Terra Australis 34, 29.

Colombaroli, D., Ssemmanda, I., Gelorini, V., Verschuren, D., 2014. Contrasting long‐term records of biomass burning in wet and dry savannas of equatorial East Africa. Global change biology 20, 2903- 2914.

Colombaroli, D., Whitlock, C., Tinner, W., Conedera, M., 2017. Paleo records as a guide for ecosystem management and biodiversity conservation. Past Global Changes Magazine 25, 78-79.

Conedera, M., Tinner, W., Neff, C., Meurer, M., Dickens, A.F., Krebs, P., 2009. Reconstructing past fire regimes: methods, applications, and relevance to fire management and conservation. Quaternary Science Reviews 28, 555-576.

Corbett, S., Balmer, J., 2007. Buttongrass Moorland in Tasmania-What and Where? Australasian Plant Conservation: Journal of the Australian Network for Plant Conservation 16, 3.

Cosgrove, R., 1995. Late Pleistocene behavioral variation and time trends: the case from Tasmania. Archaeology in Oceania, 30, 83 - 104.

Cosgrove, R., 1999. Forty-two degrees south: the archaeology of Late Pleistocene Tasmania. Journal of World Prehistory, 13, 357-402.

Cosgrove, R., Allen, J., Marshall, B., 1990. Palaeo-ecology and Pleistocene human occupation in south central Tasmania. Antiquity 64, 59-78.

Daniau, A.L., Bartlein, P., Harrison, S., Prentice, I., Brewer, S., Friedlingstein, P., Harrison‐Prentice, T., Inoue, J., Izumi, K., Marlon, J., 2012. Predictability of biomass burning in response to climate changes. Global Biogeochemical Cycles 26.

Davis, M.B., 1963. On the theory of pollen analysis. American Journal of Science 261, 897-912.

Davis, M.B., 1973. Redeposition of pollen grains in lake sediment. Limnology and Oceanography 18, 44-52.

Davis, M.B., Calcote, R.R., Sugita, S., Takahara, H., 1998. Patchy invasion and the origin of a hemlock– hardwoods forest mosaic. Ecology 79, 2641-2659.

Davis, M.B., 2000. Palynology after Y2K-Understanding the Source Area of Pollen in Sediments. Annual Review of Earth and Planetary Sciences 28, 1-18. de Nascimento, L., Nogué, S., Fernández-Lugo, S., Méndez, J., Otto, R., Whittaker, R.J., Willis, K.J., Fernández-Palacios, J.M., 2015. Modern pollen rain in Canary Island ecosystems and its implications for the interpretation of fossil records. Review of Palaeobotany and Palynology 214, 27-39.

203 |

Dearing, J.A., Braimoh, A.K., Reenberg, A., Turner, B.L. and van der Leeuw, S., 2010. Complex land systems: the need for long time perspectives to assess their future. Ecology and Society, 15, 21.

Diffenbaugh, N.S., Field, C.B., 2013. Changes in ecologically critical terrestrial climate conditions. Science 341, 486-492.

Di-Giovanni, F., Kevan, P., Nasr, M., 1995. The variability in settling velocities of some pollen and spores. Grana 34, 39-44.

Dodson, J.R., 1983. Modern pollen rain in southeastern New South Wales, Australia. Review of Palaeobotany and Palynology 38, 249-268.

Dodson, J.R., 2001. A vegetation and fire history in a subalpine woodland and rainforest region, Solomons Jewel Lake, Tasmania. The Holocene 21, 111 - 166.

Donders, T.H., Haberle, S., Hope, G., Wagner, F. and Visscher, H., 2007. Pollen evidence for the transition of the Eastern Australian climate system from the post-glacial to the present-day ENSO mode. Quaternary Science Reviews, 26, 1621-1637.

Donders, T.H., Wagner, F., Dilcher, D.L., Visscher, H., 2005. Mid-to late-Holocene El Nino-Southern Oscillation dynamics reflected in the subtropical terrestrial realm. Proceedings of the National Academy of Sciences of the United States of America 102, 10904-10908.

Donders, T.H., Wagner-Cremer, F., Visscher, H., 2008. Integration of proxy data and model scenarios for the mid-Holocene onset of modern ENSO variability. Quaternary Science Reviews 27, 571-579.

Duffin, K.I., Bunting, M.J., 2008. Relative pollen productivity and fall speed estimates for southern African savanna taxa. Vegetation History and Archaeobotany 17, 507-525.

Enright, N.J., Fontaine, J.B., Bowman, D.M., Bradstock, R.A., Williams, R.J., 2015. Interval squeeze: altered fire regimes and demographic responses interact to threaten woody species persistence as climate changes. Frontiers in Ecology and the Environment 13, 265-272.

Faegri, K., Iversen, J., 1989. Textbook of pollen analysis. Wiley, New York.

Fagerlind, F., 1952. The real signification of pollen diagrams.

Ferrandino, F.J., Aylor, D.E., 1984. Settling speed of clusters of spores. Phytopathology 74, 969-972.

Filipova-Marinova, M.V., Kvavadze, E.V., Connor, S.E., Sjögren, P., 2010. Estimating absolute pollen productivity for some European Tertiary-relict taxa. Vegetation History and Archaeobotany 19, 351- 364.

Fletcher, M.-S. and Moreno, P.I., 2011. Zonally symmetric changes in the strength and position of the Southern Westerlies drove atmospheric CO2 variations over the past 14 k.y. Geology, 39, 419-422.

Fletcher, M.-S. and Moreno, P.I., 2012. Have the Southern Westerlies changed in a zonally symmetric manner over the last 14,000 years? A hemisphere-wide take on a controversial problem. Quaternary International, 253, 32–46.

204 |

Fletcher, M.-S. and Thomas, I., 2007b. Holocene vegetation and climate change from near Lake Pedder, south-west Tasmania, Australia. Journal of Biogeography, 34, 665-677.

Fletcher, M.-S., Benson, A., Heijnis, H., Gadd, P.S., Cwynar, L.C., Rees, A.B.H., 2015. Changes in biomass burning mark the onset an ENSO-influenced climate regime at 42°S in southwest Tasmania, Australia. Quaternary Science Reviews 122, 222-232.

Fletcher, M.-S., Thomas, I., 2007a. Modern pollen–vegetation relationships in western Tasmania, Australia. Review of Palaeobotany and Palynology 146, 146-168.

Fletcher, M.-S., Thomas, I., 2010a. The origin and temporal development of an ancient cultural landscape. Journal of Biogeography 37, 2183–2196.

Fletcher, M.-S., Thomas, I., 2010b. A Holocene record of sea level, vegetation, people and fire from western Tasmania, Australia. The Holocene 20, 351-361.

Fletcher, M.-S., Wolfe, B.B., Whitlock, C., Pompeani, D., Heijnis, H., Haberle, S., Gadd, P., Bowman, D.M.J.S., 2014a. The legacy of mid Holocene fire on a Tasmanian montane landscape. Journal of Biogeography.

Fletcher, M.-S., Wood, S.W. and Haberle, S.G., 2014b. A fire-driven shift from forest to non-forest: evidence for alternative stable states? Ecology, 95, 2504-2513.

Fogt, R.L., Perlwitz, J., Monaghan, A.J., Bromwich, D.H., Jones, J.M., Marshall, G.J., 2009. Historical SAM variability. Part II: twentieth-century variability and trends from reconstructions, observations, and the IPCC AR4 models*. Journal of Climate 22, 5346-5365.

Foley, J.A., Costa, M.H., Delire, C., Ramankutty, N., Snyder, P., 2003. Green surprise? How terrestrial ecosystems could affect earth’s climate. Frontiers in Ecology and the Environment 1, 38-44.

Forcada, J., Trathan, P.N., 2009. Penguin responses to climate change in the Southern Ocean. Global Change Biology 15, 1618-1630.

Ford, H.A., Paton, D.C., Forde, N., 1979. Birds as pollinators of Australian plants. New Zealand journal of botany 17, 509-519.

Fox-Hughes, P., Harris, R., Lee, G., Grose, M., Bindoff, N., 2014. Future fire danger climatology for Tasmania, Australia, using a dynamically downscaled regional climate model. International journal of wildland fire 23, 309-321.

Friend, A.D., Lucht, W., Rademacher, T.T., Keribin, R., Betts, R., Cadule, P., Ciais, P., Clark, D.B., Dankers, R., Falloon, P.D., 2014. Carbon residence time dominates uncertainty in terrestrial vegetation responses to future climate and atmospheric CO2. Proceedings of the National Academy of Sciences 111, 3280-3285.

Fyfe, J.C., Saenko, O.A., 2006. Simulated changes in the extratropical Southern Hemisphere winds and currents. Geophysical Research Letters 33.

Gaillard, M.-J., Birks, H.J.B., Ihse, M. and Runborg, S., 1998. Pollen/landscape calibrations based on modern pollen assemblages from surface-sediment samples and landscape mapping-a pilot study in South Sweden. Palaeoclimate Research, 27, 31-52.

205 |

Gaillard, M.-J., Sugita, S., Mazier, F., Trondman, A.-K., Brostrom, A., Hickler, T., Kaplan, J.O., Kjellström, E., Kokfelt, U., Kunes, P., 2010. Holocene land-cover reconstructions for studies on land cover-climate feedbacks. Climate of the Past 6, 483-499.

Gallant, A.J., Phipps, S.J., Karoly, D.J., Mullan, A.B., Lorrey, A.M., 2013. Nonstationary Australasian teleconnections and implications for paleoclimate reconstructions. Journal of Climate 26, 8827-8849.

Gallego-Sala, A.V., Prentice, I.C., 2013. Blanket peat biome endangered by climate change. Nature Climate Change 3, 152-155.

Gammage, B., 2011. The biggest estate on earth. How Aboriginals made Australia.

Garreaud, R.D., 2007. Precipitation and circulation covariability in the extratropics. Journal of Climate 20, 4789-4797.

Garreaud, R.D., Vuille, M., Compagnucci, R., Marengo, J., 2009. Present-day South American climate. Palaeogeography Palaeoclimatology Palaeoecology 281, 180-195.

Gentilli, J., 1972. Australian climate patterns. The Griffin Press, Adelaide.

Gilbert, J.M., 1959. Forest Succession in the Florentine Valley, Tasmania. Papers and Proceedings of the Royal Society of Tasmania 93, 129-151.

Gillett, N.P., Kell, T.D., Jones, P.D., 2006. Regional climate impacts of the Southern Annular Mode. Geophysical Research Letters 33, L23704.

Gillson, L., Dawson, T.P., Jack, S., McGeoch, M.A., 2013. Accommodating climate change contingencies in conservation strategy. Trends in Ecology and Evolution 28, 135-142.

Gil-Romera, G., González-Sampériz, P., Lasheras-Álvarez, L., Sevilla-Callejo, M., Moreno, A., Valero- Garcés, B., López-Merino, L., Carrión, J.S., Sanz, A.P., Aranbarri, J., 2014. Biomass-modulated fire dynamics during the Last Glacial–Interglacial Transition at the Central Pyrenees (Spain). Palaeogeography, Palaeoclimatology, Palaeoecology 402, 113-124.

Girardin, M.P., Ali, A.A., Carcaillet, C., Gauthier, S., Hély, C., Le Goff, H., Terrier, A., Bergeron, Y., 2013. Fire in managed forests of eastern Canada: Risks and options. Forest Ecology and Management 294, 238-249.

Graham, N.E., Ammann, C.M., Fleitmann, D., Cobb, K.M., Luterbacher, J., 2010. Support for global climate reorganization during the “Medieval Climate Anomaly”. Climate Dynamics 2010, 1-29.

Green, D.G., 1981. Time series and postglacial forest ecology. Quat. Res. 15, 265-277.

Gregory, P., 1978. Distribution of airborne pollen and spores and their long distance transport. Pure and Applied Geophysics 116, 309-315.

Grinsted, A., Moore, J.C., Jevrejeva, S., 2004. Application of the cross wavelet transform and wavelet coherence to geophysical time series. Nonlinear processes in geophysics 11, 561-566.

Grose, M.R., Fox-Hughes, P., Harris, R.M., Bindoff, N.L., 2014. Changes to the drivers of fire weather with a warming climate–a case study of southeast Tasmania. Climatic Change 124, 255-269.

Guilyardi, E., 2006. El Niño–mean state–seasonal cycle interactions in a multi-model ensemble. Climate Dynamics 26, 329-348.

206 |

Harrison, S.P., Dodson, J.R., 1993. Climates of Australia and New Guinea since 18,000 yr B.P., in: Wright Jr, H.E., Kutzbach, J.E., Webb III, T., Ruddiman, W.F., Street-Perrot, F.A., Bartlein, P.J. (Eds.), Global climates since the Last Glacial Maximum. University of Minnesota Press, Minneapolis, MN, pp. 265–293.

Hedges, R., 1981. Radiocarbon dating with an accelerator: review and preview. Archaeometry 23, 3- 18.

Hellman, S., Gaillard, M.J., Broström, A. and Sugita, S., 2008. The REVEALS model, a new tool to estimate past regional plant abundance from pollen data in large lakes: validation in southern Sweden. Journal of Quaternary Science, 23, 21-42.

Hellman, S., Gaillard, M.J., Broström, A., Sugita, S., 2008. The REVEALS model, a new tool to estimate past regional plant abundance from pollen data in large lakes: validation in southern Sweden. Journal of Quaternary Science 23, 21-42.

Hendon, H.H., Thompson, D.W.J., Wheeler, M.C., 2007. Australian rainfall and surface temperature variations associated with the Southern Hemisphere annular mode. Journal of Climate 20, 2452-2467.

Hennessy, K., Lucas, C., Nicholls, N., Bathols, J., Suppiah, R., Ricketts, J., 2005. Climate change impacts on fire-weather in south-east Australia, Climate Impacts Group, CSIRO Atmospheric Research and the Australian Government Bureau of Meteorology, Aspendale.

Hicks, S., 2001. The use of annual arboreal pollen deposition values for delimiting tree-lines in the landscape and exploring models of pollen dispersal. Review of Palaeobotany and Palynology 117, 1- 29.

Higuera, P.E., Peters, M.E., Brubaker, L.B., Gavin, D.G., 2007. Understanding the origin and analysis of sediment-charcoal records with a simulation model. Quaternary Science Reviews 26, 1790-1809.

Hill, K.J., Santoso, A., England, M.H., 2009. Interannual Tasmanian Rainfall Variability Associated with Large-Scale Climate Modes. Journal of Climate 22, 4383-4397.

Hill, R., 1994. 16 The history of selected Australian taxa. History of the Australian vegetation: Cretaceous to Recent, 390.

Hill, R.S., 1990. Sixty million years of change in Tasmania's climate and vegetation. Tasforests, Forestry Commission, Tasmania 2, 89-98.

Hill, R.S., Jordan, G.J., Macphail, M.K., 2015. Why we should retain Nothofagus sensu lato. Australian Systematic Botany 28, 190-193.

Hjelle, K.L., Sugita, S., 2012. Estimating pollen productivity and relevant source area of pollen using lake sediments in Norway: How does lake size variation affect the estimates? The Holocene 22, 313- 324.

Hogg, A.G., Hua, Q., Blackwell, P.G., Niu, M., Buck, C.E., Guilderson, T.P., Heaton, T.J., Palmer, J.G., Reimer, P.J., Reimer, R.W., 2013. SHCal13 Southern Hemisphere calibration, 0–50,000 cal yr BP. Radiocarbon.

Holz, A., Paritsis, J., Mundo, I.A., Veblen, T.T., Kitzberger, T., Williamson, G.J., Aráoz, E., Bustos- Schindler, C., González, M.E., Grau, H.R., 2017. Southern Annular Mode drives multicentury wildfire activity in southern South America. Proceedings of the National Academy of Sciences, 201705168.

207 |

Holz, A., Veblen, T.T., 2011. Variability in the Southern Annular Mode determines wildfire activity in Patagonia. Geophysical Research Letters 38, L14710.

Holz, A., Wood, S.W., Veblen, T.T., Bowman, D.M., 2014. Effects of high severity fire drove the population collapse of the subalpine Tasmanian endemic conifer Athrotaxis cupressoides. Global change biology.

Hopf, F.V.L., Colhoun, E.A. and Barton, C.E., 2000. Late-glacial and Holocene record of vegetation and climate from Cynthia bay, Lake StClair, Tasmania. Journal of Quaternary Science, 15, 725-732.

Hopper, S., Moran, G., 1981. Bird pollination and the mating system of Eucalyptus stoatei. Australian Journal of Botany 29, 625-638.

Hopper, S.D., 1979. Biogeographical aspects of speciation in the southwest Australian flora. Annual review of ecology and systematics 10, 399-422.

Hultberg, T., Gaillard, M.-J., Grundmann, B. and Lindbladh, M., 2015. Reconstruction of past landscape openness using the Landscape Reconstruction Algorithm (LRA) applied on three local pollen sites in a southern Swedish biodiversity hotspot. Vegetation History and Archaeobotany, 24, 253-266.

Iglesias, V., Whitlock, C., 2014. Fire responses to postglacial climate change and human impact in northern Patagonia (41–43° S). Proceedings of the National Academy of Sciences 111, E5545-E5554.

Ingles, A., 1985. Fire. Environmental Impact statement into the continuation of woodchip exportsfrom Tasmania after 1990. Working paper No.

Irvine, A.K., Armstrong, J.E., 1990. Beetle pollination in tropical forests of Australia. Reproductive ecology of tropical forest plants 7, 135-149.

Iversen, J., 1958. The bearing of glacial and interglacial epochs on the formation and extinction of plant taxa. Uppsala Universiteit Arsskrift 6, 210 - 215.

Jackson, S.T., 1990. Pollen source area and representation in small lakes of the northeastern United States. Review of Palaeobotany and Palynology 63, 53-76.

Jackson, S.T., Lyford, M.E., 1999. Pollen dispersal models in Quaternary plant ecology: assumptions, parameters, and prescriptions. The botanical review 65, 39-75.

Jackson, W., Bowman, D., 1982. Reply: Ecological drift or fire cycles in south-west Tasmania. Search 13, 175-176.

Jackson, W.D., 1968. Fire, air, water and earth – an elemental ecology of Tasmania. Proceedings of the Ecological Society of Australia, 3, 9-16.

Jacobson, G.L., Bradshaw, R.H., 1981. The selection of sites for paleovegetational studies. Quaternary Research 16, 80-96.

Jarman, S.J., Kantivalis, G. and Brown, M.J. (1988) Buttongrass moorlands in Tasmania. Tasmanian Forest Research Council Research Report No. 3. Tasmanian Forest Research Council, Hobart.

Jarosz, N., Loubet, B., Huber, L., 2004. Modelling airborne concentration and deposition rate of maize pollen. Atmospheric Environment 38, 5555-5566.

208 |

Jenny, B., Wilhem, D., Valero-Garces, B.L., 2003. The Southern Westerlies in Central Chile: Holocene precipitation estimates based on a water balance model for Laguna Aculeo (33°50'S). Climate Dynamics 20, 269-280.

Jolliffe, I., 2002. Principal component analysis. Wiley Online Library.

Jones, R., 1969. Fire-stick Farming. Australian Natural History 16, 224 - 228.

Kaland, P.E., Krzywinski, K., Stabell, B., 1984. Radiocarbon‐dating of transitions between marine and lacustrine sediments and their relation to the development of lakes. Boreas 13, 243-258.

Kaplan, J.O., Prentice, I.C., Knorr, W. and Valdes, P.J., 2002. Modeling the dynamics of terrestrial carbon storage since the Last Glacial Maximum. Geophysical Research Letters, 29, 22

Kershaw, A., 1979. Local pollen deposition in aquatic sediments on the Atherton Tableland, North‐ Eastern Australia. Australian Journal of Ecology 4, 253-263.

Kershaw, A., Hyland, B., 1975. Pollen transfer and periodicity in a rain-forest situation. Review of Palaeobotany and Palynology 19, 129-138.

Kershaw, A., Strickland, K., 1990. A 10 year pollen trapping record from rainforest in northeastern Queensland, Australia. Review of Palaeobotany and Palynology 64, 281-288.

Kershaw, P.A., 1986. Climatic Change and Aboriginal burining in north-east Australia during the last two glacial/interglacial cycles, Nature, pp. 47-49.

Kilian, R., Lamy, F., 2012. A review of Glacial and Holocene paleoclimate records from southernmost Patagonia (49–55°S). Quaternary Science Reviews 53, 1-23.

Kirkpatrick, J., Dickinson, K., 1984a. Vegetation of Tasmania 1: 500 000. Forestry Commission Tasmania.

Kirkpatrick, J.A. and Dickinson, K.J.M., 1984b. The impact of fire on Tasmanian vegetation and soils. Australian Journal of Botany, 32, 613-629.

Kirkpatrick, J.B., 1999. A continent transformed: human impact on the natural vegetation of Australia. Oxford University Press.

Koutavas, A., Lynch-Stieglitz, J., Marchitto, T.M., Sachs, J.P., 2002. El Nino-like pattern in ice age tropical Pacific sea surface temperature. Science 297, 226-230.

Krawchuk, M.A., Moritz, M.A., Parisien, M.-A., Van Dorn, J., Hayhoe, K., 2009. Global pyrogeography: the current and future distribution of wildfire. PloS one 4, e5102.

Kuneš, P., Odgaard, B.V. and Gaillard, M.J., 2011. Soil phosphorus as a control of productivity and openness in temperate interglacial forest ecosystems. Journal of Biogeography, 38, 2150-2164.

Kuneš, P., Svobodová-Svitavská, H., Kolář, J., Hajnalová, M., Abraham, V., Macek, M., Tkáč, P. and Szabó, P., 2015. The origin of grasslands in the temperate forest zone of east-central Europe: long-term legacy of climate and human impact. Quaternary Science Reviews, 116, 15-27.

Kunsch, H.R., 1989. The jackknife and the bootstrap for general stationary observations. The Annals of Statistics, 1217-1241.

209 |

Kuparinen, A., Markkanen, T., Riikonen, H., Vesala, T., 2007. Modeling air-mediated dispersal of spores, pollen and seeds in forested areas. Ecological modelling 208, 177-188.

Kyaw Tha Paw U., Hotton, C., 1989. Optimum pollen and female receptor size for anemophily. American Journal of Botany, 445-453.

Lamy, F., Kilian, R., Arz, H.W., Francois, J.P., Kaiser, J., Prange, M., Steinke, T., 2010. Holocene changes in the position and intensity of the southern westerly wind belt. Nature Geoscience, 695-699.

Langford, J., 1965. Weather and climate. Atlas of Tasmania, 2-11.

Lannom, K.O., Tinkham, W.T., Smith, A.M., Abatzoglou, J., Newingham, B.A., Hall, T.E., Morgan, P., Strand, E.K., Paveglio, T.B., Anderson, J.W., 2014. Defining extreme wildland fires using geospatial and ancillary metrics. International Journal of Wildland Fire 23, 322-337.

Lara, A., Villalba, R., Urrutia, R., 2008. A 400-year tree-ring record of the Puelo River summer–fall streamflow in the Valdivian Rainforest eco-region, Chile. Climatic Change 86, 331-356.

Lau, K., Weng, H., 1995. Climate signal detection using wavelet transform: How to make a time series sing. Bulletin of the American Meteorological Society 76, 2391-2402.

Le Goff, H., Flannigan, M.D., Bergeron, Y., Girardin, M.P., 2007. Historical fire regime shifts related to climate teleconnections in the Waswanipi area, central Quebec, Canada. International Journal of Wildland Fire 16, 607-618.

Lehman, J.T., 1975. Reconstructing the rate of accumulation of lake sediment: the effect of sediment focusing. Quat. Res. 5, 541-550.

Lenton, T.M., Held, H., Kriegler, E., Hall, J.W., Lucht, W., Rahmstorf, S., Schellnhuber, H.J., 2008. Tipping elements in the Earth's climate system. Proceedings of the National Academy of Sciences 105, 1786-1793.

Li, Y., Nielsen, A.B., Zhao, X., Shan, L., Wang, S., Wu, J., Zhou, L., 2015. Pollen production estimates (PPEs) and fall speeds for major tree taxa and relevant source areas of pollen (RSAP) in Changbai Mountain, northeastern China. Review of Palaeobotany and Palynology 216, 92-100.

Lindbladh, M., Fraver, S., Edvardsson, J., Felton, A., 2013. Past forest composition, structures and processes–How paleoecology can contribute to forest conservation. Biological Conservation 168, 116- 127.

Lindenmayer, D., Hunter, M., 2010. Some guiding concepts for conservation biology. Conservation Biology 24, 1459-1468.

Lisé-Pronovost, A., St-Onge, G., Gogorza, C., Haberzettl, T., Jouve, G., Francus, P., Ohlendorf, C., Gebhardt, C., Zolitschka, B., Team, P.S., 2015. Rock-magnetic proxies of wind intensity and dust since 51,200 cal BP from lacustrine sediments of Laguna Potrok Aike, southeastern Patagonia. Earth and Planetary Science Letters 411, 72-86.

210 |

Liu, W., Lu, J., Leung, L.R., Xie, S.-P., Liu, Z., Zhu, J., 2015. The de-correlation of westerly winds and westerly-wind stress over the Southern Ocean during the Last Glacial Maximum. Climate Dynamics, 1-12.

Livingstone, D., 1955. A lightweight piston sampler for lake deposits. Ecology 36, 137-139.

Lough, J., Fritts, H., 1987. An assessment of the possible effects of volcanic eruptions on North American climate using tree-ring data, 1602 to 1900 AD. Climatic Change 10, 219-239.

Lowe, J.J., Walker, M.J., 2000. Radiocarbon Dating the Last Glacial-Interglacial Transition (Ca. 14–9 14C Ka Bp) in Terrestrial and Marine Records: The Need for New Quality Assurance Protocols1. Radiocarbon 42, 53-68.

Macphail, M.K., 1979. Vegetation and climates in southern Tasmania since the last glaciation. Quaternary Research 11, 306-341.

Macphail, M.K., 2010. The burning question: Claims and counter claims on the origin and extent of buttongrass moorland (blanket moor) in southwest Tasmania during the present glacial-interglacial. Terra Australis 32, 323-339.

Mann, M.E., Zhang, Z., Rutherford, S., Bradley, R.S., Hughes, M.K., Shindell, D., Ammann, C., Faluvegi, G., Ni, F., 2009. Global signatures and dynamical origins of the Little Ice Age and Medieval Climate Anomaly. Science 326, 1256.

Marchant, R., Hooghiemstra, H., 2004. Rapid environmental change in African and South American tropics around 4000 years before present: a review. Earth-Science Reviews 66, 217-260.

Mariani, M., Connor, S., Theuerkauf, M., Kuneš, P. and Fletcher, M.-S., 2016. Testing quantitative pollen dispersal models in animal-pollinated vegetation mosaics: An example from temperate Tasmania, Australia. Quaternary Science Reviews, 154, 214-225.

Mariani, M., Connor, S.E., Fletcher, M.S., Theuerkauf, M., Kuneš, P., Jacobsen, G., Saunders, K.M., Zawadzki, A., 2017. How old is the Tasmanian cultural landscape? A test of landscape openness using quantitative land‐cover reconstructions. Journal of Biogeography.

Mariani, M., Fletcher, M.S., 2016. The Southern Annular Mode determines inter‐annual and centennial‐scale fire activity in temperate southwest Tasmania, Australia. Geophysical Research Letters.

Mariani, M., Fletcher, M.-S., 2017. Long-term climate dynamics in the extra-tropics of the South Pacific revealed from sedimentary charcoal analysis. Quaternary Science Reviews 173, 181-192.

Mariani, M., Fletcher, M.S., Holz, A., Nyman, P., 2016. ENSO controls interannual fire activity in southeast Australia. Geophysical Research Letters 43, 10,891–810,900.

Markgraf, V., Bradbury, J.P. and Busby, J.R., 1986. Palaeoclimates in southwestern Tasmania during the last 13,000 years. Palios, 1, 368-380.

Marlon, J.R., Bartlein, P.J., Carcaillet, C., Gavin, D.G., Harrison, S.P., Higuera, P.E., Joos, F., Power, M., Prentice, I., 2008. Climate and human influences on global biomass burning over the past two millennia. Nature Geoscience 1, 697-702.

211 |

Marlon, J.R., Bartlein, P.J., Daniau, A.-L., Harrison, S.P., Maezumi, S.Y., Power, M.J., Tinner, W., Vanniére, B., 2013. Global biomass burning: a synthesis and review of Holocene paleofire records and their controls. Quaternary Science Reviews 65, 5-25.

Marlon, J.R., Kelly, R., Daniau, A.-L., Vannière, B., Power, M.J., Bartlein, P., Higuera, P., Blarquez, O., Brewer, S., Brücher, T., 2016. Reconstructions of biomass burning from sediment charcoal records to improve data-model comparisons. Biogeosciences (BG) 13, 3225-3244.

Marquer, L., Gaillard, M.-J., Sugita, S., Trondman, A.-K., Mazier, F., Nielsen, A.B., Fyfe, R.M., Odgaard, B.V., Alenius, T., Birks, H.J.B., 2014. Holocene changes in vegetation composition in northern Europe: why quantitative pollen-based vegetation reconstructions matter. Quaternary Science Reviews 90, 199-216.

Marsden-Smedley, J.B., 1998. Changes in southwestern Tasmanian fire regimes since the early 1800's. Papers and Proceedings of the Royal Society of Tasmania 132, 15-29.

Marshall, G.J., 2003. Trends in the Southern Annular Mode from observations and reanalyses. Journal of Climate 16, 4134-4143.

Matthias, I., Nielsen, A.B., Giesecke, T., 2012. Evaluating the effect of flowering age and forest structure on pollen productivity estimates. Vegetation history and archaeobotany 21, 471-484.

Mayle, F.E., Iriarte, J., 2014. Integrated palaeoecology and archaeology–a powerful approach for understanding pre-Columbian Amazonia. Journal of Archaeological Science 51, 54-64.

Mazier, F., Broström, A., Gaillard, M.-J., Sugita, S., Vittoz, P., Buttler, A., 2008. Pollen productivity estimates and relevant source area of pollen for selected plant taxa in a pasture woodland landscape of the Jura Mountains (Switzerland). Vegetation History and Archaeobotany 17, 479-495.

Mazier, F., Gaillard, M.-J., Kuneš, P., Sugita, S., Trondman, A.-K., Broström, A., 2012. Testing the effect of site selection and parameter setting on REVEALS-model estimates of plant abundance using the Czech Quaternary Palynological Database. Review of Palaeobotany and Palynology 187, 38-49.

McClymont, E.L., Rosell-Melé, A., 2005. Links between the onset of modern Walker circulation and the mid-Pleistocene climate transition. Geology 33, 389-392.

McCune, B. and Mefford, M.J., 1999. PC-Ord for Windows. MjM Software, Gleneden Beach.

McGlone, M.S., Kershaw, P., Markgraf, V., 1992. El Nino/Southern Oscillation climatic variability in Australasian and South American paleoenvironmental records, in: Diaz, H.F., Markgraf, V. (Eds.), El Nino: historical and paleoclimatic aspects of the southern oscillation. Cambridge University Press, Cambridge.

McKenzie, D., Gedalof, Z.e., Peterson, D.L., Mote, P., 2004. Climatic change, wildfire, and conservation. Conservation biology 18, 890-902.

McWethy, D., Higuera, P., Whitlock, C., Veblen, T., Bowman, D., Cary, G., Haberle, S., Keane, R., Maxwell, B., McGlone, M., 2013. A conceptual framework for predicting temperate ecosystem sensitivity to human impacts on fire regimes. Global Ecology and Biogeography 22, 900-912.

212 |

Meehl, G.A., Stocker, T.F., Collins, W.D., Friedlingstein, A., Gaye, A.T., Gregory, J.M., Kitoh, A., Knutti, R., Murphy, J.M., Noda, A., 2007. Global climate projections.

Meyer, G.A., Wells, S.G., Jull, A.T., 1995. Fire and alluvial chronology in Yellowstone National Park: climatic and intrinsic controls on Holocene geomorphic processes. Geological Society of America Bulletin 107, 1211-1230.

Meyers, S.D., Kelly, B.G., O'Brien, J.J., 1993. An introduction to wavelet analysis in oceanography and meteorology: With application to the dispersion of Yanai waves. Monthly weather review 121, 2858- 2866.

Miller, R., Schmidt, G., Shindell, D., 2006. Forced annular variations in the 20th century intergovernmental panel on climate change fourth assessment report models. Journal of Geophysical Research: Atmospheres (1984–2012) 111.

Mooney, S., Harrison, S., Bartlein, P., Daniau, A.-L., Stevenson, J., Brownlie, K., Buckman, S., Cupper, M., Luly, J., Black, M., 2011. Late Quaternary fire regimes of Australasia. Quaternary Science Reviews 30, 28-46.

Mooney, S.D., Harrison, S.P., Bartlein, P., Stevenson, J., 2012. The prehistory of fire in Australasia. Flammable Australia: Fire regimes, biodiversity and ecosystems in a changing world, 3-25.

Moore, P., Webb, J., Collinson, M., 1991. 1991: Pollen analysis. Oxford: Blackwell Scientific Publications.

Moreno, P.I., 2004. Millennial-scale climate variability in northwest Patagonia over the last 15000 yr. Journal of Quaternary Science 19, 35-47.

Moreno, P.I., Francois, J.P., Moy, C.M., Villa-Martinez, R., 2010. Covariability of the Southern Westerlies and atmospheric CO2 during the Holocene. Geology 38, 727-730.

Moritz, M.A., Batllori, E., Bradstock, R.A., Gill, A.M., Handmer, J., Hessburg, P.F., Leonard, J., McCaffrey, S., Odion, D.C., Schoennagel, T., 2014. Learning to coexist with wildfire. Nature 515, 58-66.

Moritz, M.A., Parisien, M.-A., Batllori, E., Krawchuk, M.A., Van Dorn, J., Ganz, D.J., Hayhoe, K., 2012. Climate change and disruptions to global fire activity. Ecosphere 3, art49.

Mount, A.B., 1979. Natural regeneration processes in Tasmanian forests. Search 10, 180-186.

Moy, C.M., Seltzer, G.O., Rodbell, D.T. and Anderson, D.M., 2002. Variability of El Nino/Southern Oscillation activity at millennial timescales during the Holocene. Nature, 420, 162-165.

Murúa, R., González, L.A., Lima, M., 2003. Population dynamics of rice rats (a Hantavirus reservoir) in southern Chile: feedback structure and non‐linear effects of climatic oscillations. Oikos 102, 137- 145.

Nicholls, N., Lucas, C., 2007. Interannual variations of area burnt in Tasmanian bushfires: relationships with climate and predictability. International Journal of Wildland Fire 16, 540-546.

Nielsen, A.B., Møller, P.F., Giesecke, T., Stavngaard, B., Fontana, S.L., Bradshaw, R.H., 2010. The effect of climate conditions on inter-annual flowering variability monitored by pollen traps below the canopy in Draved Forest, Denmark. Vegetation History and Archaeobotany 19, 309-323.

213 |

Nielsen, A.B., Odgaard, B.V., 2005. Reconstructing land cover from pollen assemblages from small lakes in Denmark. Review of Palaeobotany and Palynology 133, 1-21.

Nielsen, A.B., Sugita, S., 2005. Estimating relevant source area of pollen for small Danish lakes around AD 1800. The Holocene 15, 1006-1020.

Nuñez, M., 1988. A Regional Lapse Rate for Tasmania. Papers and Proceedings of the Royal Society of Tasmania 122, 53 - 57.

Odgaard, B.V., Rasmussen, P., 1998. The use of historical data and sub-recent (AD 1800) pollen assemblages to quantify vegetation/pollen relationships. Paläoklimaforschung/Palaeoclim Res 27, 67-75.

Odgaard, B.V., Rasmussen, P., 2000. Origin and temporal development of macro‐scale vegetation patterns in the cultural landscape of Denmark. Journal of Ecology 88, 733-748.

Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Stevens, M., Oksanen, M. and Suggests, M., 2015. Vegan community ecology package: ordination methods, diversity analysis and other functions for community and vegetation ecologists. Version 2.3-1.

Oldfield, F., Appleby, P., 1984. Empirical testing of 210 Pb-dating models for lake sediments, Lake sediments and environmental history.

Olsson, I., 1991. Conventional Radiocarbon Dating and Some Problems of 14C Dating. Scientific Dating Methods: Brussels, Kluwer Academic Publishers, 15-35.

Overpeck, J.T., Web III, T., Prentice, I.C., 1985. Quantitative Interpretation of Fossil Pollen Spectra: Dissimilarity Coefficients and the Method of Modern Analogs. Quat. Res. 23, 87-108.

Paritsis, J., Veblen, T.T., 2011. Dendroecological analysis of defoliator outbreaks on Nothofagus pumilio and their relation to climate variability in the Patagonian Andes. Global Change Biology 17, 239-253.

Parshall, T., 2002. Late Holocene stand‐scale invasion by hemlock (Tsuga canadensis) at its western range limit. Ecology 83, 1386-1398.

Parsons, R., Prentice, I., 1981. Statistical approaches to R-values and the pollen—vegetation relationship. Review of Palaeobotany and Palynology 32, 127-152.

Paton, D., Ford, H., 1977. Pollination by birds of native plants in South Australia. Emu 77, 73-85.

Patterson, W.A., Edwards, K.J., Maguire, D.J., 1987. Microscopic charcoal as a fossil indicator of fire. Quaternary Science Reviews 6, 3-23.

Pausas, J.G., Ribeiro, E., 2013. The global fire–productivity relationship. Global Ecology and Biogeography 22, 728-736.

Pemberton, M., 1988. Soil erosion between Brichs Inlet and Elliot Bay, southwest Tasmania. Papers and Proceedings of the Royal Society of Tasmania 122, 109-114.

Perlwitz, J., 2011. Atmospheric science: Tug of war on the jet stream. Nature Climate Change 1, 29-31.

Perlwitz, J., Pawson, S., Fogt, R.L., Nielsen, J.E., Neff, W.D., 2008. Impact of stratospheric ozone hole recovery on Antarctic climate. Geophysical Research Letters 35.

214 |

Philander, S.G.H., 1983. El Niño southern oscillation phenomena. Nature 302, 295-301.

Pickett, E.J., Harrison, S.P., Hope, G., Harle, K., Dodson, J.R., Kershaw, P., Prentice, I.C., Backhouse, J., Colhoun, E.A., D'Costa, D., Flenley, J., Grindrod, J., Haberle, S., Hassell, C., Kenyon, C., Macphail, M., Martin, H., Martin, A.H., McKenzie, M., Newsome, J.C., Penny, D., Powell, J., Raine, J.I., Southern, W., Stevenson, J., Sutra1, J.-P., Thomas, I., van Der Kaars, S., Ward, J., 2004. Pollen-based reconstructions of biome distributions for Australia, Southeast Asia and the Pacific (SEAPAC region) at 0, 6000 and 18,000 14C yr BP. Journal of Biogeography 31, 1381-1444.

Pitkänen, A., Huttunen, P., Jungner, H., Tolonen, K., 2002. A 10 000 year local forest fire history in a dry heath forest site in eastern Finland, reconstructed from charcoal layer records of a small mire. Canadian Journal of Forest Research 32, 1875-1880.

Plomley, N.J.B., 1966. Friendly Mission: The Tasmanian Journals and Papers of George Augustus Robinson 1829-1834 (Kingsgrove: Tasmanian Historial Research Association).

Polvani, L.M., Previdi, M., Deser, C., 2011. Large cancellation, due to ozone recovery, of future Southern Hemisphere atmospheric circulation trends. Geophysical Research Letters 38.

Poska, A., Meltsov, V., Sugita, S., Vassiljev, J., 2011. Relative pollen productivity estimates of major anemophilous taxa and relevant source area of pollen in a cultural landscape of the hemi-boreal forest zone (Estonia). Review of Palaeobotany and Palynology 167, 30-39.

Power, M.J., Marlon, J., Ortiz, N., Bartlein, P.J., Harrison, S.P., Mayle, F.E., Ballouche, A., Bradshaw, R.H.W., Carcaillet, C., Cordova, C., Mooney, S., Moreno, P.I., Prentice, I.C., Thonicke, K., Tinner, W., Whitlock, C., Zhang, Y., Zhao, Y., Ali, A.A., Anderson, R.S., Beer, R., Behling, H., Briles, C., Brown, K.J., Brunelle, A., Bush, M., Camill, P., Chu, G.Q., Clark, J., Colombaroli, D., Connor, S., Daniau, A.L., Daniels, M., Dodson, J., Doughty, E., Edwards, M.E., Finsinger, W., Foster, D., Frechette, J., Gaillard, M.J., Gavin, D.G., Gobet, E., Haberle, S., Hallett, D.J., Higuera, P., Hope, G., Horn, S., Inoue, J., Kaltenrieder, P., Kennedy, L., Kong, Z.C., Larsen, C., Long, C.J., Lynch, J., Lynch, E.A., McGlone, M., Meeks, S., Mensing, S., Meyer, G., Minckley, T., Mohr, J., Nelson, D.M., New, J., Newnham, R., Noti, R., Oswald, W., Pierce, J., Richard, P.J.H., Rowe, C., Goni, M.F.S., Shuman, B.N., Takahara, H., Toney, J., Turney, C., Urrego-Sanchez, D.H., Umbanhowar, C., Vandergoes, M., Vanniere, B., Vescovi, E., Walsh, M., Wang, X., Williams, N., Wilmshurst, J., Zhang, J.H., 2008. Changes in fire regimes since the Last Glacial Maximum: an assessment based on a global synthesis and analysis of charcoal data. Climate Dynamics 30, 887-907.

Power, S., Delage, F., Chung, C., Kociuba, G., Keay, K., 2013. Robust twenty-first-century projections of El Niño and related precipitation variability. Nature 502, 541-545.

Prentice, C., Guiot, J., Huntley, B., Jolly, D., Cheddadi, R., 1996. Reconstructing biomes from palaeoecological data: a general method and its application to European pollen data at 0 and 6 ka. Climate Dynamics 12, 185-194.

Prentice, I., Parsons, R., 1983. Maximum likelihood linear calibration of pollen spectra in terms of forest composition. Biometrics, 1051-1057.

Prentice, I., Webb, T., 1986. Pollen percentages, tree abundances and the Fagerlind effect. Journal of Quaternary science 1, 35-43.

Prentice, I.C., 1985. Pollen Representation, Source Area, and Basin Size: Toward a Unified Theory of Pollen Analysis. Quat. Res. 23, 76-86.

215 |

Pryor, L.D., 1976. The biology of eucalypts. The biology of Eucalypts.

Pyne, S.J., Andrews, P.L., Laven, R.D., 1996. Introduction to Wildland Fire. Wiley.

Pyrke, A.F., Marsden-Smedley, J.B., 2005. Fire-attributes catefories, fire sensitivity, and flammability of Tasmanian vegetation communities. Tasforests 16, 35-46.

Radke, L.F., Hegg, D.A., Hobbs, P.V., Nance, J.D., Lyons, J.H., Laursen, K.K., Weiss, R.E., Riggan, P.J., Ward, D.E., 1991. Particulate and trace gas emissions from large biomass fires in North America. Global biomass burning: Atmospheric, climatic, and biospheric implications, 209-224.

Räsänen, S., Hicks, S., Odgaard, B.V., 2004. Pollen deposition in mosses and in a modified ‘Tauber trap’from Hailuoto, Finland: what exactly do the mosses record? Review of Palaeobotany and Palynology 129, 103-116.

Räsänen, S., Suutari, H., Nielsen, A.B., 2007. A step further towards quantitative reconstruction of past vegetation in Fennoscandian boreal forests: pollen productivity estimates for six dominant taxa. Review of Palaeobotany and Palynology 146, 208-220.

Rees, A.B., Cwynar, L.C., Fletcher, M.-S., 2015. Southern Westerly Winds submit to the ENSO regime: A multiproxy paleohydrology record from Lake Dobson, Tasmania. Quaternary Science Reviews 126, 254-263.

Regal, P.J., 1982. Pollination by wind and animals: ecology of geographic patterns. Annual Review of Ecology and Systematics 13, 497-524.

Reid, J.B., Hill, R.S., Brown, M. and Hovenden, M.J., 1999. Vegetation of Tasmania. Australia Biological Resources Study, Hobart.

Reimer, P.J., Brown, T.A., Reimer, R.W., 2004. Discussion: reporting and calibration of post-bomb 14C data. Radiocarbon 46, 1299-1304.

Renberg, I., 1991. The HON-Kajak sediment corer. Journal of Paleolimnology 6, 167-170.

Riedinger, M.A., Steinitz-Kannan, M., Last, W.M., Brenner, M., 2002. A 6100 14 C yr record of El Niño activity from the Galápagos Islands. Journal of Paleolimnology 27, 1-7. ∼ Risbey, J.S., Pook, M.J., McIntosh, P.C., Wheeler, M.C., Hendon, H.H., 2009. On the remote drivers of rainfall variability in Australia. Monthly Weather Review 137, 3233-3253.

Rodbell, D.T., Seltzer, G.O., Anderson, D.M., Abbott, M.B., Enfield, D.B., Newman, J.H., 1999. An~ 15,000-year record of El Niño-driven alluviation in southwestern Ecuador. Science 283, 516-520.

Rojas, M., Moreno, P.I., 2010. Atmospheric circulation changes and neoglacial conditions in the Southern Hemisphere mid-latitudes: insights from PMIP2 simulations at 6 kyr. Climate Dynamics DOI 10.1007/s00382-010-0866-3.

Román‐Cuesta, R., Carmona‐Moreno, C., Lizcano, G., New, M., Silman, M., Knoke, T., Malhi, Y., Oliveras, I., Asbjornsen, H., Vuille, M., 2014. Synchronous fire activity in the tropical high Andes: an indication of regional climate forcing. Global change biology 20, 1929-1942.

216 |

Saito, H., Takeoka, M., 1985. Pollen production rates in a young Japanese red pine forest. Nihon Seitai Gakkaishi 35, 67-76.

Sandweiss, D.H., Maasch, K.A., Burger, R.L., Richardson Iii, J.B., Rollins, H.B., Clement, A., 2001. Variation in Holocene El Nino frequencies: Climate records and cultural consequences in ancient Peru. Geology 29, 603.

Schloss, I.R., Abele, D., Moreau, S., Demers, S., Bers, A.V., González, O., Ferreyra, G.A., 2012. Response of phytoplankton dynamics to 19-year (1991–2009) climate trends in Potter Cove (Antarctica). Journal of Marine systems 92, 53-66.

Schueler, S., Schlünzen, K.H., 2006. Modeling of oak pollen dispersal on the landscape level with a mesoscale atmospheric model. Environmental Modeling & Assessment 11, 179.

Scott, A.C., 2000. The Pre-Quaternary history of fire. Palaeogeography, Palaeoclimatology, Palaeoecology 164, 281-329.

Scott, A.C., Bowman, D.M., Bond, W.J., Pyne, S.J., Alexander, M.E., 2013. Fire on earth: an introduction. John Wiley and Sons.

Seppä, H., Bennett, K., 2003. Quaternary pollen analysis: recent progress in palaeoecology and palaeoclimatology. Progress in Physical Geography 27, 548-579.

Sharples, J.J., Cary, G.J., Fox-Hughes, P., Mooney, S., Evans, J.P., Fletcher, M.-S., Fromm, M., Grierson, P.F., McRae, R., Baker, P., 2016. Natural hazards in Australia: extreme bushfire. Climatic Change 139, 85-99.

Shulmeister, J., Lees, B.G., 1995. Pollen evidence from tropical Australia for the onset of an ENSO- dominated climate at c. 4000 BP. The Holocene 5, 10-18.

Sikes, E., Howard, W., Samson, C., Mahan, T., Robertson, L., Volkman, J., 2009. Southern Ocean seasonal temperature and Subtropical Front movement on the South Tasman Rise in the late Quaternary. Paleoceanography 24.

Sitch, S., Smith, B., Prentice, I.C., Arneth, A., Bondeau, A., Cramer, W., Kaplan, J., Levis, S., Lucht, W. and Sykes, M.T. (2003) Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology, 9, 161-185.

Smith, B., Samuelsson, P., Wramneby, A., Rummukainen, M., 2011. A model of the coupled dynamics of climate, vegetation and terrestrial ecosystem biogeochemistry for regional applications. Tellus A 63, 87-106.

Smith, T.M., Reynolds, R.W., 2005. A global merged land-air-sea surface temperature reconstruction based on historical observations (1880-1997). Journal of Climate 18, 2021-2036.

Sniderman, J., Haberle, S.G., 2012. Fire and vegetation change during the Early Pleistocene in southeastern Australia. Journal of Quaternary Science 27, 307-317.

217 |

Soepboer, W., Sugita, S., Lotter, A.F., van Leeuwen, J.F., van der Knaap, W.O., 2007. Pollen productivity estimates for quantitative reconstruction of vegetation cover on the Swiss Plateau. The Holocene 17, 65-77.

Soepboer, W., Vervoort, J.M., Sugita, S., Lotter, A.F., 2008. Evaluating Swiss pollen productivity estimates using a simulation approach. Vegetation History and Archaeobotany 17, 497-506.

Soons, M.B., Heil, G.W., Nathan, R., Katul, G.G., 2004. Determinants of long‐distance seed dispersal by wind in grasslands. Ecology 85, 3056-3068.

Stahle, L.N., Whitlock, C., Haberle, S.G., 2016. A 17,000-Year-Long Record of Vegetation and Fire from Cradle Mountain National Park, Tasmania. Frontiers in Ecology and Evolution 4, 82.

Strandberg, G., Kjellström, E., Poska, A., Wagner, S., Gaillard, M.-J., Trondman, A.-K., Mauri, A., Kaplan, J., Birks, H., Bjune, A., 2014. Regional climate model simulations for Europe at 6 and 0.2 k BP: sensitivity to changes in anthropogenic deforestation. Climate of the Past, 661-680.

Sturman, A.P., Tapper, N.J., 2006. The weather and climate of Australia and New Zealand. Oxford University Press, USA.

Styger, J., Kirkpatrick, J.B., 2015. Less than 50 millimetres of rainfall in the previous month predicts fire in Tasmanian rainforest, Papers and Proceedings of the Royal Society of Tasmania, pp. 1-5.

Sugita, S., 1993. A model of pollen source area for an entire lake surface. Quat. Res. 39, 239-244.

Sugita, S., 1994. Pollen representation of vegetation in Quaternary sediments: theory and method in patchy vegetation. Journal of Ecology, 881-897.

Sugita, S., 2007a. Theory of quantitative reconstruction of vegetation I: pollen from large sites REVEALS regional vegetation composition. The Holocene 17, 229-241.

Sugita, S., 2007b. Theory of quantitative reconstruction of vegetation II: all you need is LOVE. The Holocene 17, 243-257.

Sugita, S., Gaillard, M.-J., Broström, A., 1999. Landscape openness and pollen records: a simulation approach. The Holocene 9, 409-421.

Sugita, S., Parshall, T., Calcote, R. and Walker, K., 2010. Testing the Landscape Reconstruction Algorithm for spatially explicit reconstruction of vegetation in northern Michigan and Wisconsin. Quaternary Research, 74, 289-300.

Sutton, O., 1953. Micrometeorology. McGraw-Hill, New York

Swanson, F.J., 1981. Fire and geomorphic processes. Mooney, HA; Bonnicksen, TM; Christensen, NL; Lotan, JE, 401-444.

Tackenberg, O., 2003. Modeling long‐distance dispersal of plant diaspores by wind. Ecological Monographs 73, 173-189.

Tarasov, P., Andreev, A., Anderson, P., Lozhkin, A., Leipe, C., Haltia, E., Nowaczyk, N., Wennrich, V., Brigham-Grette, J., Melles, M., 2013. A pollen-based biome reconstruction over the last 3.562 million years in the Far East Russian Arctic–new insights into climate–vegetation relationships at the regional scale. Climate of the Past 9, 2759-2775.

218 |

Tauber, H., 1965. Differential pollen dispersion and the interpretation of pollen diagrams, with a contribution to the interpretation of the elm fall.

Tauber, H., 1967. Investigations of the mode of pollen transfer in forested areas. Review of Palaeobotany and Palynology 3, 277-286.

Tauber, H., 1974. A static non‐overload pollen collector. New Phytologist 73, 359-369.

Tauber, H., 1977. Investigations of aerial pollen transport in a forested area [Gribskov, Sealand, Denmark]. Dansk botanisk forening.

Team, R.C.D., 2013. R: A language and environment for statistical computing, R Foundation for Statistical Computing Vienna Austria.

Terasmae, J., Weeks, N., 1979. Natural fires as an index of paleoclimate. Canadian Field-Naturalist 93, 116-125.

Theuerkauf, M., Couwenberg, J., Kuparinen, A., Liebscher, V., 2016. A matter of dispersal: REVEALSinR introduces state-of-the-art dispersal models to quantitative vegetation reconstruction. Vegetation History and Archaeobotany, 1-13.

Theuerkauf, M., Kuparinen, A., Joosten, H., 2013. Pollen productivity estimates strongly depend on assumed pollen dispersal. The Holocene, 0959683612450194.

Thomas, I., 1993. Late Pleistocene Environments and Aboriginal Settlement Patterns in Tasmania. Australian Archaeology, 36, 1-11.

Thomas, I., 1995a. "Where have all the Forests Gone?" New Pollen Evidence from Melaleuca Inlet in Southwestern Tasmania, in: Dixon, G., Aitkin, D. (Eds.), IAG 1993 Conference Proceedings. Monash Publications in Geography, Melbourne, pp. 295-301.

Thomas, I., 1995b. Ethnohistoric sources for the use of fire by Tasmanian Aborigines: a preliminary compilation and investigation of the evidence. Dept. of Geography and Environmental Studies, University of Melbourne.

Thomas, I., Cullen, P., Fletcher, M.-S., 2010. Ecological drift or stable fire cycles in Tasmania: A resolution?, in: Haberle, S., Stevenson, J., Prebble, M. (Eds.), Altered Ecologies: Fire, Climate and Human Influence on Terrestrial Landscapes. ANU E-Press, Canberra, The Australian National University.

Thompson, D.W., Solomon, S., 2002. Interpretation of recent Southern Hemisphere climate change. Science 296, 895-899.

Thompson, D.W., Solomon, S., Kushner, P.J., England, M.H., Grise, K.M., Karoly, D.J., 2011. Signatures of the Antarctic ozone hole in Southern Hemisphere surface climate change. Nature Geoscience 4, 741-749.

Thonicke, K., Venevsky, S., Sitch, S., Cramer, W., 2001. The role of fire disturbance for global vegetation dynamics: coupling fire into a Dynamic Global Vegetation Model. Global Ecology and Biogeography 10, 661-677.

219 |

Torrence, C., Compo, G.P., 1998. A practical guide to wavelet analysis. Bulletin of the American Meteorological society 79, 61-78.

Trondman, A.K., Gaillard, M.J., Mazier, F., Sugita, S., Fyfe, R., Nielsen, A.B., Twiddle, C., Barratt, P., Birks, H.J.B., Bjune, A.E., 2015. Pollen‐based quantitative reconstructions of Holocene regional vegetation cover (plant‐functional types and land‐cover types) in Europe suitable for climate modelling. Global change biology 21, 676-697.

Trondman, A.-K., Gaillard, M.-J., Sugita, S., Björkman, L., Greisman, A., Hultberg, T., Lagerås, P., Lindbladh, M., Mazier, F., 2016. Are pollen records from small sites appropriate for REVEALS model- based quantitative reconstructions of past regional vegetation? An empirical test in southern Sweden. Vegetation history and archaeobotany 25, 131-151.

Turney, C.S., Kershaw, A.P., Clemens, S.C., Branch, N., Moss, P.T., Fifield, L.K., 2004. Millennial and orbital variations of El Nino/Southern Oscillation and high-latitude climate in the last glacial period. Nature 428, 306-310.

Twiddle, C.L., Jones, R.T., Caseldine, C.J., Sugita, S., 2012. Pollen productivity estimates for a pine woodland in eastern Scotland: the influence of sampling design and vegetation patterning. Review of Palaeobotany and Palynology 174, 67-78.

Van der Knaap, W., van Leeuwen, J.F., Ammann, B., 2001. Seven years of annual pollen influx at the forest limit in the Swiss Alps studied by pollen traps: relations to vegetation and climate. Review of Palaeobotany and Palynology 117, 31-52.

Van der Linden, P., Mitchell, J., editors, 2009. ENSEMBLES: Climate change and its impacts-Summary of research and results from the ENSEMBLES project.

Vannière, B., Blarquez, O., Rius, D., Doyen, E., Brücher, T., Colombaroli, D., Connor, S., Feurdean, A., Hickler, T., Kaltenrieder, P., 2016. 7000-year human legacy of elevation-dependent European fire regimes. Quaternary Science Reviews 132, 206-212.

Verdon, D.C., Kiem, A.S., Franks, S.W., 2004. Multi-decadal variability of forest fire risk-eastern Australia. International Journal of Wildland Fire 13, 165-171.

Villalba, R., Lara, A., Masiokas, M.H., Urrutia, R., Luckman, B.H., Marshall, G.J., Mundo, I.A., Christie, D.A., Cook, E.R., Neukom, R., 2012. Unusual Southern Hemisphere tree growth patterns induced by changes in the Southern Annular Mode. Nature Geoscience 5, 793-798. von Post, L., 1946. The Prospect For Pollen Analysis In The Study Of The Earth's Climatic History. New Phytologist 45, 193-193. von Stedingk, H., Fyfe, R.M., Allard, A., 2008. Pollen productivity estimates from the forest—tundra ecotone in west-central Sweden: implications for vegetation reconstruction at the limits of the boreal forest. The Holocene 18, 323-332.

Walker, D., 2000. Pollen input to, and incorporation in, two crater lakes in tropical northeast Australia. Review of palaeobotany and palynology 111, 253-283.

Walker, J., 1981. Fuel dynamics in Australian vegetation. Fire and the Australian biota, 101-127.

Wang, B., Wang, Y., 1996. Temporal structure of the Southern Oscillation as revealed by waveform and wavelet analysis. Journal of Climate 9, 1586-1598.

220 |

Webb, T., Howe, S., Bradshaw, R., Heide, K., 1981. Estimating plant abundances from pollen percentages: the use of regression analysis. Review of Palaeobotany and Palynology 34, 269-300.

Weimerskirch, H., Louzao, M., De Grissac, S., Delord, K., 2012. Changes in wind pattern alter albatross distribution and life-history traits. Science 335, 211-214.

Westerling, A.L., Hidalgo, H.G., Cayan, D.R., Swetnam, T.W., 2006. Warming and earlier spring increase western US forest wildfire activity. science 313, 940-943.

Westerling, A.L., Turner, M.G., Smithwick, E.A., Romme, W.H., Ryan, M.G., 2011. Continued warming could transform Greater Yellowstone fire regimes by mid-21st century. Proceedings of the National Academy of Sciences 108, 13165-13170.

Whinam, J., Hope, G., 2005. The peatlands of the Australasian region. Stapfia 85, 397-433.

Whitlock, C., 2004. Land management: Forests, fires and climate. Nature 432, 28-29.

Whitlock, C., Bartlein, P.J., 2004. Holocene fire activity as a record of past environmental change. The Quaternary period in the United States. Elsevier, Amsterdam, 479-490.

Whitlock, C., Colombaroli, D., Conedera, M., Tinner, W., 2017. Land‐use history as a guide for forest conservation and management. Conservation Biology.

Whitlock, C., Larsen, C., 2001. Charcoal as a fire proxy. Smol, JP, Birks, HJB and Last, WM (eds.) Tracking environmental change using lake sediments Vol. 3, Terrestrial, algal, and Siliceous Indicators: 75-97. Kluwer Academic Publishers.

Whitlock, C., Millspaugh, S.H., 1996. Testing the assumptions of fire-history studies: an examination of modern charcoal accumulation in Yellowstone National Park, USA. The Holocene 6, 7-15.

Wilkins, D., Gouramanis, C., De Deckker, P., Fifield, L.K., Olley, J., 2013. Holocene lake-level fluctuations in Lakes Keilambete and Gnotuk, southwestern Victoria, Australia. The Holocene, 23(6), 784-795.

Williams, A.N., Mooney, S.D., Sisson, S.A., Marlon, J., 2015. Exploring the relationship between Aboriginal population indices and fire in Australia over the last 20,000 years. Palaeogeography, Palaeoclimatology, Palaeoecology 432, 49-57.

Williams, J.W., Jackson, S.T., 2007. Novel climates, no-analog communities, and ecological surprises. Frontiers in Ecology and the Environment 5, 475-482.

Williamson, G.J., Prior, L.D., Jolly, W.M., Cochrane, M.A., Murphy, B.P., Bowman, D.M., 2016. Measurement of inter-and intra-annual variability of landscape fire activity at a continental scale: the Australian case. Environmental Research Letters 11, 035003.

Willis, K., Bailey, R., Bhagwat, S., Birks, H., 2010. Biodiversity baselines, thresholds and resilience: testing predictions and assumptions using palaeoecological data. Trends in Ecology and Evolution 25, 583-591.

Willis, K.J., Birks, H.J.B., 2006. What is natural? The need for a long-term perspective in biodiversity conservation. Science 314, 1261-1265.

221 |

Wood, S.W., Bowman, D.M.J.S., 2012. Alternative stable states and the role of fire–vegetation–soil feedbacks in the temperate wilderness of southwest Tasmania. Landscape Ecology, 1-16.

Wood, S.W., Hua, Q. and Bowman, D.M.J.S., 2011a. Fire-patterned vegetation and the development of organic soils in the lowland vegetation mosaics of south-west Tasmania. Australian Journal of Botany, 59, 126-136.

Wood, S.W., Murphy, B.P., Bowman, D.M.J.S., 2011b. Firescape ecology: how topography determines the contrasting distribution of fire and rain forest in the south‐west of the Tasmanian Wilderness World Heritage Area. Journal of Biogeography 38, 1807-1820.

Wright, H.E., Mann, D., Glaser, P., 1984. Piston corers for peat and lake sediments. Ecology 65, 657- 659.

Xia, Y., Zhao, J. and Collerson, K.D., 2001. Early-Mid Holocene climatic variations in Tasmania, Australia: multi-proxy records in a stalagmite from Lynds Cave. Earth and Planetary Science Letters, 194, 177-187.

Xu, Q., Zhang, S., Gaillard, M.-j., Li, M., Cao, X., Tian, F., Li, F., 2016. Studies of modern pollen assemblages for pollen dispersal-deposition-preservation process understanding and for pollen-based reconstructions of past vegetation, climate, and human impact: A review based on case studies in China. Quaternary Science Reviews 149, 151-166.

Yospin, G.I., Wood, S.W., Holz, A., Bowman, D.M., Keane, R.E., Whitlock, C., 2015. Modeling vegetation mosaics in sub-alpine Tasmania under various fire regimes. Modeling Earth Systems and Environment 1, 1-10.

222 |

Appendix A1 - Additional research papers published during this PhD project

PAPER 1:

Mariani, M., Fletcher, M.S., Holz, A., Nyman, P., 2016. ENSO controls interannual fire activity in southeast Australia. Geophysical Research Letters 43, 10,891–810,900.

Contribution: Data collection: 100%; Data analysis: 100%; Manuscript writing: 80%; Ideas: 80%

PAPER 2:

Mariani, M., Fletcher, M.-S., Drysdale, R.N., Saunders, K.M., Heijnis, H., Jacobsen, G., Zawadzki, A., 2017. Coupling of the Intertropical Convergence Zone and Southern Hemisphere mid-latitude climate during the early to mid-Holocene. Geology.

Contribution: Data collection: 70%; Data analysis: 100%; Manuscript writing: 70%; Ideas: 60%

223 PUBLICATIONS

Geophysical Research Letters

RESEARCH LETTER ENSO controls interannual ﬁre activity in southeast 10.1002/2016GL070572 Australia

Key Points: M. Mariani1, M.-S. Fletcher1, A. Holz2, and P. Nyman3 • ENSO modulates interannual fire activity in southeast Australia 1School of Geography, University of Melbourne, Parkville, Victoria, Australia, 2Department of Geography, School of the • Fire occurrence in southeast Australia 3 is strongly dependent on seasonal Environment, Portland State University, Portland, Oregon, USA, School of Ecosystem and Forest Science, University of variations of ENSO Melbourne, Parkville, Victoria, Australia • Decadal-scale frequencies in ENSO are linked with fire activity in southeast Australia Abstract El Niño–Southern Oscillation (ENSO) is the main mode controlling the variability in the ocean-atmosphere system in the South Pacific. While the ENSO influence on rainfall regimes in the South Supporting Information: Pacific is well documented, its role in driving spatiotemporal trends in fire activity in this region has not been • Supporting Information S1 rigorously investigated. This is particularly the case for the highly flammable and densely populated southeast Australian sector, where ENSO is a major control over climatic variability. Here we conduct the first Correspondence to: fi fi M. Mariani, region-wide analysis of how ENSO controls re activity in southeast Australia. We identify a signi cant [email protected] relationship between ENSO and both fire frequency and area burnt. Critically, wavelet analyses reveal that despite substantial temporal variability in the ENSO system, ENSO exerts a persistent and significant influence fi Citation: on southeast Australian re activity. Our analysis has direct application for developing robust predictive Mariani, M., M.-S. Fletcher, A. Holz, and capacity for the increasingly important efforts at fire management. P. Nyman (2016), ENSO controls interannual fire activity in southeast Australia, Geophys. Res. Lett., 43, 1. Introduction doi:10.1002/2016GL070572. Fire is a ubiquitous Earth system process that began soon after the appearance of terrestrial plants 420 million Received 24 JUL 2016 years ago [Scott and Glasspool, 2006]. At the global scale, fire affects ecosystem patterns and processes, deter- Accepted 28 SEP 2016 mines vegetation distribution [Bond and Keeley, 2005; Bond et al., 2005], impacts the climate system [Bowman Accepted article online 30 SEP 2016 et al., 2009], and contributes to the carbon cycle [Liu et al., 2015; Santín et al., 2016]. Moreover, climate-driven changes in fire activity are a significant threat for human populations living in flammable biomes [e.g., Parisien, 2016] and for ecosystem services and function [Calder et al., 2015]. The predicted increase in fire activity in many regions on Earth in response to climate change [Parisien and Moritz, 2009; Moritz et al., 2012; Westerling et al., 2006], therefore, represents a significant challenge for attempts of a sustainable management of the Earth system. Despite the clear importance of fire and recognition of climate as a key component controlling this process, the relationship between climate and fire is still poorly understood in many regions on Earth. In southeast Australia, one of the most fire-prone areas on Earth [Hennessy et al., 2005], the main climate mode controlling key determinants of fire weather (moisture and temperature variability) is the El Niño–Southern Oscillation (ENSO) [Risbey et al., 2009]. In general, El Niño events, the warm phase of ENSO, starve southeast Australia of rainfall and promote drought and wildfire [Kiem and Franks, 2004; Verdon et al., 2004; Nicholls and Lucas, 2007; Murphy and Timbal, 2008; Hill et al., 2009]. The current amplification of El Niño activity due to anthropogenic climate change [Power et al., 2013] then heralds a ser- ious threat to southeast Australia’s water security, remnant fire-sensitive ecosystems and the ever expanding flammable bush-urban interface [Guilyardi, 2006; Lenton et al., 2008; Sharples et al., 2016]. Here we present a novel analysis of climate-fire dynamics in southeast Australia in an attempt to elucidate the dominant climatic drivers of fire in both space and time in this heavily populated region. An enhanced understanding of the climatic precursors of increased fire activity can lead to improved predictive power [e.g., Verdon et al., 2004; Nicholls and Lucas, 2007; Harris et al., 2014]. This is particularly important for areas such as southeast Australia where the quasiperiodic frequency of oscillations in key climate phenomena, such as ENSO, makes fire prediction a difficult task. ENSO is a particularly important control over fire regimes across the entire Pacific region, and the spatiotemporal variability of ENSO is evident from the range of analyses that attempt to elucidate the role of ENSO in governing fire activity. Indeed, significant correlation between ENSO and fire activity has been reported for the Florida everglades between 1948 and 1999 [Beckage et al., 2003], California during the last 150 years [Herweijer et al., 2007], tropical Mexico between

©2016. American Geophysical Union. 1984 and 1999 [Román-Cuesta et al., 2003], southwest United States between 1700 and 1905 [Swetnam All Rights Reserved. and Betancourt, 1990], Indonesia over the last 250 years [van der Kaars et al., 2010], and southern South

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 224 1 Geophysical Research Letters 10.1002/2016GL070572

America over the most recent centuries [Veblen et al., 1999; Holz and Veblen, 2012; Holz et al., 2012a]. While there is evidence that El Niño events increase the chance of fire weather in southeast Australia [e.g., Fox- Hughes et al., 2014; Grose et al., 2014; Williams et al., 2001], few analyses from this region assess the influence of ENSO over actual fire occurrence (area burnt and/or number of fires). Of these, anecdotal evidence of an ENSO control over fire activity has been reported from the island of Tasmania, where fires over the historical period are significantly related to sea surface temperatures (SSTs) in the Coral Sea, a location influenced by ENSO [Nicholls and Lucas, 2007]. More concrete evidence comes from the south coast of the region (state of Victoria), where a significant correlation between ENSO and both area burnt and number of fires has been reported between 1972 and 2012 [Harris et al., 2014]. While ENSO is most often considered the dominant control over southeast Australian climate and fire activity, the Indian Ocean Dipole (IOD) and the Southern Annular Mode (SAM), both important controls over climatic variability in this region, have been shown to influence fire activity within the southeast Australian sector [e.g., Cai et al., 2009; Mariani and Fletcher, 2016], suggesting that multiple modes of climatic variability must be considered when attempting to determine the drivers of fire activity in this region. A critical limitation of previous analyses of ENSO-fire relationships in southeast Australia is the lack of a regional synthesis of actual fire activity conducted through a rigorous climatological framework. A case in point is the work on the role of ENSO in governing fire activity in Tasmania [Nicholls and Lucas, 2007]. Tasmania is a topographically complex midlatitude island in which the climatic determinants of fire weather and activity are polarized between ENSO in the north and east and the SAM in the south and west [Mariani and Fletcher, 2016]. SAM relates to changes in the position and intensity of the circumpolar southern westerlies, a system that, while periodically synergistic with ENSO through time [Fogt et al., 2011], primarily acts independently to ENSO. Indeed, recent evidence reveals that fire activity in the south and west of Tasmania is determined by SAM, with ENSO having little explanatory power over fire activity through the recent and deeper past [Mariani and Fletcher, 2016]. Further, analyses of the climatic determinants of fire activity in southeast Australia have seldom employed actual fire data [Harris et al., 2014; Nicholls and Lucas, 2007; Mariani and Fletcher, 2016], rather, most employ indices of fire weather to ascertain the link between climate and the potential for fire occurrence [Fox-Hughes et al., 2014; Grose et al., 2014; Williams et al., 2001]. Moreover, of those that do employ actual fire occurrence data, only those focused on the island of Tasmania exclude planned burns from their analysis [Nicholls and Lucas, 2007; Mariani and Fletcher, 2016], an important source of noise that reduces the efficacy of the fire occurrence data set for correlation with climate variability. Here we present a multiscale analysis of the role of ENSO in governing actual fire activity in southeast Australia. Our specific aims are to (1) explore the relationship between ENSO and fire occurrence in the densely populated and highly flammable southeast of Australia; (2) synthesize the seasonal importance of ENSO in determining interannual fire activity in this region; and (3) test whether or not ENSO and fire occurrences have coherent time-frequency domains. We conducted our climate-fire analysis using the Southern Oscillation Index (SOI), an index of ENSO variability that strongly affects precipitation in southeast Australia [Risbey et al., 2009]. We employed observed fire history data filtered to include unplanned fires only and conducted correlation function and wavelet analyses to determine long-term seasonal-scale associations and time frequencies coherence between SOI and fire activity through time.

2. Methods To visualize the spatial pattern of the inﬂuence of ENSO on southeast Australian precipitation, we created a correlation map between annual and seasonal precipitation totals and the Southern Oscillation Index (SOI). SOI is calculated as the normalized difference between mean sea level pressure at Tahiti and Darwin, and it is negative during El Niño events and positive during La Niña events [Trenberth, 1984]. Although there are several indices for ENSO (e.g., Niño 4, Niño 3.4, ENSO-Modoki Index, EMI), we used the SOI because it has shown to have the highest correlation values with rainfall in Australia compared to the other indices [Risbey et al., 2009]. This is probably because the SOI is more closely related to the rainfall process in southeast Australia through its relationship with large-scale surface pressure, compared to ocean-based indices which rely on sea surface temperature (SST) [Risbey et al., 2009]. We calculated Pearson correlation coefﬁcients (r) between seasonal/annual rainfall during the period 1961–1990 for 1208 meteorological stations across the southeast Australia and the seasonal/annual SOI. Precipitation data were obtained from the Australian

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 225 2 Geophysical Research Letters 10.1002/2016GL070572

Bureau of Meteorology (BOM). The r values from the stations were spatially interpolated using the Universal Kriging method in ArcMap 10.2 [ESRI-Environmental Systems Resource Institute, 2011, Redlands, California]. To account for fuel limitations, an essential parameter in determining fire activity [Krawchuk et al., 2009; Cochrane, 2003; Pausas and Ribeiro, 2013; Bradstock, 2010; McWethy et al., 2013; Boer et al., 2016], we constrain our analysis to the cool temperate forest biome of southeast Australia (latitude between 28°S and 44°S; longitude between 140°E and 155°E), where biomass is considered to be always abundant and in which climate is the principal determinant of temporal changes in fire activity [Bradstock et al., 2014]. This screening removed fires from the Australian Alps, as much of this landscape approaches biomass limitation with respect to fire [Bradstock et al., 2014]. Observed fire occurrence data were obtained through local administrative databases and span the period between the fire seasons of 1951/1952 and 2013/2014 (hereon, we define a fire season year by the year in which the fire season ends—e.g., fire season 1951/1952 = 1952). We determine the “fire season” as the period between December and March, during which the majority of fire events occur [Williamson et al., 2016]. For Tasmania, only the eastern and northern side of the state was considered, as fire activity and precipitation in the west of this island have been linked to variability in the SAM, with no identified relationship between fire activity and ENSO [Mariani and Fletcher, 2016]. Fire activity is represented in this study as two metrics: number of occurrences and area burnt, representing different components of a fire regime. Area burned represents the outcome of a fire season and includes ignitions as well as fire weather and fuels that determine fire spread. Fire occurrence is mainly linked to ignitions and is more sensitive to artifacts linked to detection. The fire activity data (number of fires and area burnt) were first filtered to include all fires except deliberate management fires and then normalized by the standard deviations (z scores). Z scores values higher (lower) than 0.5 (-0.5) were used to identify significant “fire years” (“nonfire years”)[Mariani and Fletcher, 2016]. The list of years used in the analysis is shown in Table S1 in the supporting information. A persistent increase in the number of fires toward the present likely reflects improving detection and recording through time; thus, we applied linear detrending to the time series to extract the interannual variability. Figure 1 presents the location of all fires recorded in southeast Australia from 1951 to 2014. To identify a relationship between the annual and seasonal SOI and fire activity in the southeast Australia, we performed Superposed Epoch Analysis (SEA) analysis in R package (dplR package [Bunn et al., 2016]). This analysis allows for assessing the significance of the departure from the mean for a given set of key event years (e.g., fire years) and lagged years using bootstrapped confidence intervals [Lough and Fritts, 1987]. To analyze the relationship between fire activity and the SOI, we used significant “fire years” and “nonfire years” during the 1951–2014 period and annual and seasonal SOI values. The seasonal indices of SOI were defined as Winter (June–July–August of the fire season end year), Spring (September–October–November of the fire season end year), Summer (December of the previous year and January and February of the fire season end year), and Autumn (March–April–May of the fire season end year). Given the importance of the SAM and IOD in the climate-fire dynamics of the Southern Hemisphere [e.g., Cai et al., 2009; Risbey et al., 2009; Hill et al., 2009; Garreaud et al., 2009; Mariani and Fletcher, 2016], the same analysis using SEA has been also conducted with these climate modes to test whether they have a significant relationship with fire activity in the study region. To test whether the fire frequency and area burnt and the SOI show a similar periodicity, wavelet coherence analysis was undertaken using the wt() and wtc() functions respectively in the R package “Biwavelet” [Gouhier et al., 2016]. For this analysis, the spring-summer period was chosen as it represents the season with the highest significant correlation (p value < 0.05) between the SOI and the two fire activity metrics (Table S2 in the supporting information). Wavelet analyses have often been applied to climate data [e.g., Meyers et al., 1993; Lau and Weng, 1995; Wang and Wang, 1996] and provides a useful tool to reveal frequency localization in climate signals in time. The Morlet continuous wavelet transform was applied and the data were padded with zeros at each end to reduce wraparound effects [Torrence and Webster, 1999].

3. Results A total of 25,390 unplanned fires with a total area of 37,291,374 ha were recorded in the temperate biome region of southeast Australia during the period 1951–2014 (Figure 1a). A total of eight significant fire years were determined for fire occurrence, whereas 22 significant fire years were found using the area burnt metric

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 226 3 Geophysical Research Letters 10.1002/2016GL070572

Figure 1. (a) Location of the 25,390 ﬁre occurrences recorded in southeast Australia and correlation map between (b) annual and (c–f) seasonal rainfall and the SOI during the reference period 1961–1990. Grid resolution is 0.05° × 0.05°. Signiﬁcance has been tested across the 1208 stations used for the interpolation across southeast Australia and only areas with a p value < 0.05 are shown.

(Figures 2c and 2d). The total of significant nonfire years was 10 and 19 for the number of fires and area burnt, respectively (Figures 2c and 2d). The list of years used in the SEA is presented in Table S1. The spatial climate correlation analysis shows a complex pattern of correlation between SOI and rainfall anomalies across southeast Australia using annual and seasonal values (Figures 1b–1f). The strongest positive correlations between SOI and rainfall across most of the study region occur in winter and spring, while weak negative correlations occur in summer and autumn. An exception to this is the northeast of the study region, in which the strongest positive correlations occur during summer and a neutral/negative correlation is present during winter and spring (Figures 1c and 1d). Together, this variable seasonal pattern renders the annual SOI-rainfall correlations signature near neutral in most of the region, except northeastern Tasmania (Figure 1a). Importantly, negative correlations shown in the maps are always weak (i.e., r value < À0.3; p value > 0.05), indicating a probable nonsignificance of these correlations during the reference period (1961–1990). Testing both number of fires and area burnt metrics, the SEA revealed a statistically significant (p value < 0.05) negative SOI departure (El Niño conditions) occurring in winter and spring of the preceding year and during summer and autumn of the same year (Figures 3a and 3b); i.e., fire activity is significantly related to SOI in the full seasonal cycle leading up to and including a fire season. To support this result, we show that

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 227 4 Geophysical Research Letters 10.1002/2016GL070572

nonfire years (fire seasons with an anomalously low fire occurrence) correspond to a significant (p value < 0.05) positive departure in SOI (La Niña conditions) during the same seasons (Figures 3a and 3b). At an interannual scale, significant negative annual SOI departures (El Niño) were found during the fire years using the number of fires, whereas area burnt shows negative but not significant departures for annual SOI (Figure 3b). During the nonfire years for area burnt and number of fires, a significant positive SOI departure (La Niña) was found at lag zero. The analyses of both SAM and IOD seasonal and annual indices with the two fire activity metrics did not produce significant results (see Figure S1 and S2 in the supporting information), while analyses using both Nino 3.4 and the EMI are consistent with the SOI results, thus reinforcing the robustness of our results (Figure S3 and S4). Wavelet coherence indicates that negative SOI (El Niño) occurred in concert with positive anomalies in fire occurrence and area burnt at the respective time frequency during the time periods (Figure 4). The number of fire metrics shows a strong anticorrelated coherence Figure 2. Summary figure for the period 1951–2014 showing (a) rainfall pattern (i.e., a leftward arrow indicated anomaly for the southeastern Australian region (data from BOM); (b) that x and y are anticorrelated) with the annual SOI, red (blue) bars represent El Niño (La Niña) anomalies; (c) z spring-summer SOI with a short-period scores of the area burnt values; and (d) detrended z scores of the number – of fire occurrences. frequency (1 4 years) between 1963 and 1978, while area burnt shows a strong antiphase frequency between 2 and 5 years over the same 1963–1978 period. During the 1998–2010 period a strong antiphase coherence pattern with the spring-summer SOI with a periodicity between 1 and 5 years is observed in the area burnt coherence plot. A persistent strong antiphase coherence pattern with periodicity between 10 and 18 years was found both in the area burnt and fire occurrence metrics throughout the entire analysis period. Areas outside the “cone of influence,” where edge effects are present, have not been considered for the interpretation of the results.

4. Discussion Our analysis constitutes the first synthesis of the fire-climate teleconnections over the last 60 years across temperate southeast Australia, a highly flammable and densely populated region that is subject to frequent catastrophic wildfires that have major societal, cultural, and economic impacts [Sharples et al., 2016]. The SEA clearly indicates a multiseasonal control of ENSO on fire activity in this region, most likely due to ENSO-related precipitation anomalies. The significant negative departure of SOI during winter and spring of the year preceding the fire season (Figure 3), and the strong relationship between precipitation and ENSO (Figures 1 and 2), indicates that a reduction in water availability in this period is crucial for successful ignition during the following fire season. Likewise, another important factor emerging from this analysis is the ENSO state occurring during

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 228 5 Geophysical Research Letters 10.1002/2016GL070572

Figure 3. Results from the superposed epoch analysis (SEA) (a) between annual and seasonal SOI and number of fires; and (b) between annual and seasonal SOI and area burnt during fire years and nonfire years. Dark bars indicate lags with a significant SOI departure (p value < 0.05).

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 229 6 Geophysical Research Letters 10.1002/2016GL070572

the fire season: an El Niño condition present during summer and autumn is more likely to result in an increase in both number of fires and area burnt. Moreover, significant nonfire years are correlated with La Niña (high moisture) conditions during the preceding winter and spring and current summer and autumn (Figures 3a and 3b). Interestingly, the results for nonfire years using the number of fires show a significant SOI positive departure at lag À1 and 0. This may relate to the general persistence of a La Niña conditions during consecutive years (Figure 2b). Additionally, we find no evidence for a dominant role of the IOD in governing fire activity across the temperate forest biome of southeast Australia through our analysis period, in direct contrast to the study of Cai et al. [2009]. The lack of consistency between our results Figure 4. Results from the wavelet coherence analysis for the period and that of Cai et al. [2009] likely results – 1951 2013 showing (a) coherence between spring-summer SOI and from the low number of fire seasons number of fires; (b) coherence between spring-summer SOI and area (n = 21), narrow analysis period (1950– burnt. Black solid lines indicate areas exceeding the 90% confidence intervals. Leftward arrows indicate that x and y are anticorrelated; i.e., 2008) and use of spring-summer Niño negative SOI (El Niño) occur in concert with high number of events and 3.4 ENSO index, an index shown to have with positive anomalies in area burnt at the respective time frequency lower correlation values with Australian fi during the time periods shown in the gures. Dark grey solid line repre- rainfall relative to the SOI [Risbey et al., sents the “cone of influence,” where edge effects become important. 2009], in the latter analysis.

The strong coherency we have identified in our wavelet analysis at a 12 year periodicity between spring- summer El Niño events and fire activity (area burnt and fire number) throughout most of the analysis period (1963–2000) (Figure 4) indicates, for the first time, a clear and persistent decadal-scale modulation of fire activity in southeast Australia by ENSO. Given the irregular frequency found in the ENSO time series [An and Wang, 2000; Wang, 2001], it is remarkable that the observed decadal-scale modulation of fire activity by ENSO is fixed throughout the analysis period, highlighting the persistently dominant influence of ENSO over the climate of this highly flammable region, also given the predicted increase in extreme El Niños by the end of this century [Cai et al., 2014]. Further, the wavelet analysis also reveals a clear correlation between temporal shifts in ENSO variability and actual fire activity in southeast Australia. The higher frequency coherence (1–5 years) between 1963 and 1978 also corresponds to a period of maximum ENSO frequency [An and Wang, 2000]. Thus, our results reveal that southeast Australia is highly sensitive to shifts in the frequency of variability in the ENSO system and underscores the potential threat of increased fire activity in this region in response to the recent intensification of the ENSO system [Power et al., 2013]. Interestingly, there is an apparent decoupling between the coherence of area burnt and number of fires and the SOI (Figure 4)—i.e., ENSO seems to be having a control on the spread of fires between 1995 and 2003, but another factor is controlling the ignition during this period. It is possible that cultural factors, such as increased awareness, predictive capacity, and subsequent fire suppression efforts (e.g., construction of fire breaks, increased hazard reduction burning, and increased public awareness) actively suppressed ignition events in the latter part of the record. Importantly, the impact of ENSO on preconditioning the landscape for fire spread once a fire starts appears to have been less affected by these cultural adaptations. Models of future fire activity predict an increase in fire activity under projections of climate change scenarios in temperate forest biomes, such as southeast Australia [Moritz et al., 2012]. Moreover, global temperature increase has the potential to alter the thermal balance in the equatorial Pacific Ocean, driving an increase in the strength and frequency of El Niño events [Timmermann et al., 1999; Guilyardi, 2006], a process that

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 230 7 Geophysical Research Letters 10.1002/2016GL070572

is, indeed, already underway [Power et al., 2013]. Our identification of both a persistent and nonstationary influence of ENSO over fire activity in southeast Australia has direct and significant implications for managing ecosystem processes and human populations within the ENSO climate domain. Moreover, extreme variability in the Australian climate complicates attempts to understand the impacts of anthropogenic climate change [Jolly et al., 2015], and our analysis provides information on the magnitude and temporal variability of climate-driven fire activity that will significantly enhance the detection of persistent changes in a highly variable system. Our identification of a multiseasonal link between supraannual ENSO dynamics and fire activity in southeast Australia reveals that the relationship between ENSO and fire must be considered at multiple scales and for all seasons when planning for future fire activity in this densely populated region. Our results provide additional evidence for the long-term influence of ENSO over the Earth system. Over the past millennia, ENSO has been implicated in a variety of processes that include terrestrial ecosystem dynamics in eastern Australia [Donders et al., 2007] and southwest Tasmania [Fletcher et al., 2014, 2015]; storm patterns in New Zealand [Gomez et al., 2011]; coral reef systems in the eastern Pacific[Toth et al., 2012]; and human cultural change [Sandweiss et al., 2001; Magilligan and Goldstein, 2001; Turney et al., 2006] and fire activity in temperate southern South America [Whitlock et al., 2007; Holz et al., 2012b] and Tasmania [Fletcher et al., 2015; Rees et al., 2015]. Importantly, modern ENSO variability is muted, relative to the last few millennia [e.g., Moy et al., 2002; Conroy et al., 2008; Yan et al., 2011], and it is critical that we gain an understanding of how ENSO variability drives changes locally, regionally, and globally if we are to sustainably manage the Earth systems.

5. Conclusion Under a warming world scenario and a predicted future increase in fire activity across temperate biomes [Moritz et al., 2012], ENSO activity is projected to intensify [Timmermann et al., 1999; Guilyardi, 2006; Lenton et al., 2008; Power et al., 2013; Kim et al., 2014]. This research constitutes the first attempt in disentangling the role of SOI in driving fire activity across the entire southeastern Australian region during the past 60 years using observed wildfire activity data sets. We reveal that El Niño phases (negative SOI) are significantly linked with fire occurrence (number of fires) and area burnt in this region at the interannual scale throughout all seasons (i.e., winter and spring of the year preceding the fire season and during summer and autumn of the fire season year) and a coherent pattern of decadal-scale frequencies over the spring-summer period. The results obtained in this study reveal the seasonal ENSO states as important parameters to consider in fire forecasting and in ecological and palaeoecological interpretations.

Acknowledgments References Research was supported by ARC grants DI110100019 and IN140100050. Andrés An, S.-I., and B. Wang (2000), Interdecadal change of the structure of the ENSO mode and its impact on the ENSO frequency*, J. Clim., 13(12), – Holz was in part supported by the 2044 2055. Australian Research Council (grant Ashok, K., S. K. Behera, S. A. Rao, H. Weng, and T. Yamagata (2007), El Niño Modoki and its possible teleconnection, J. Geophys. Res., 112, DP110101950) and the U.S. National C11007, doi:10.1029/2006JC003798. fl fi Science Foundation Awards (grant Beckage, B., W. J. Platt, M. G. Slocum, and B. Panko (2003), In uence of the El Nino Southern Oscillation on re regimes in the Florida – 0966472). We thank the Victorian Everglades, Ecology, 84(12), 3124 3130. Department of Environment, Land, Boer, M. M., D. M. Bowman, B. P. Murphy, G. J. Cary, M. A. Cochrane, R. J. Fensham, M. A. Krawchuk, O. F. Price, V. R. De Dios, and R. J. Williams fi Water and Planning, the New South (2016), Future changes in climatic water balance determine potential for transformational shifts in Australian re regimes, Environ. Res. Wales Office of Environment and Lett., 11(6), 065002. ‘ ’ fl Heritage, the ACT Parks and Wildfire, Bond, W. J., and J. E. Keeley (2005), Fire as a global herbivore : The ecology and evolution of ammable ecosystems, Trends Ecol. Evol., 20(7), – and TasList (Government of Tasmania) 387 394. fi – for providing the fire occurrence data. Bond, W. J., F. I. Woodward, and G. F. Midgley (2005), The global distribution of ecosystems in a world without re, New Phytol., 165(2), 525 537. – We also thank Karl Braganza and Alex Bowman, D. M. J. S., et al. (2009), Fire in the Earth system, Science, 324(5926), 481 484. fi Evans from the Australian Bureau of Bradstock, R., T. Penman, M. Boer, O. Price, and H. Clarke (2014), Divergent responses of re to recent warming and drying across south- – Meteorology for providing the rainfall eastern Australia, Global Change Biol., 20(5), 1412 1428. fi – data from all the stations across south- Bradstock, R. A. (2010), A biogeographic model of re regimes in Australia: Current and future implications, Global Ecol. Biogeogr., 19, 145 158. east Australia. Data elaborated with Bunn, A., M. Korpela, F. Biondi, F. Campelo, P. Mérian, F. Qeadan, and C. Zang (2016), dplR: Dendrochronology Program Library in R. R package these analyses are available upon version 1.6.4. [Available at http://CRAN.R-project.org/package=dplR.] fi request to M.M. (mmariani@student. Cai, W., T. Cowan, and M. Raupach (2009), Positive Indian Ocean dipole events precondition southeast Australia bush res, Geophys. Res. Lett., unimelb.edu.au). 36, L19710, doi:10.1029/2009GL039902. Cai, W., S. Borlace, M. Lengaigne, P. Van Rensch, M. Collins, G. Vecchi, A. Timmermann, A. Santoso, M. J. McPhaden, and L. Wu (2014), Increasing frequency of extreme El Niño events due to greenhouse warming, Nat. Clim. Change, 4(2), 111–116. Calder, W. J., D. Parker, C. J. Stopka, G. Jiménez-Moreno, and B. N. Shuman (2015), Medieval warming initiated exceptionally large wildfire outbreaks in the Rocky Mountains, Proc. Natl. Acad. Sci. U.S.A., 112(43), 13,261–13,266. Cochrane, M. A. (2003), Fire science for rainforests, Nature, 421(6926), 913–919.

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 231 8 Geophysical Research Letters 10.1002/2016GL070572

Conroy, J. L., J. T. Overpeck, J. E. Cole, T. M. Shanahan, and M. Steinitz-Kannan (2008), Holocene changes in eastern tropical Pacific climate inferred from a Galápagos lake sediment record, Quat. Sci. Rev., 27(11–12), 1166–1180. Donders, T. H., S. Haberle, G. Hope, F. Wagner, and H. Visscher (2007), Pollen evidence for the transition of the Eastern Australian climate system from the post-glacial to the present-day ENSO mode, Quat. Sci. Rev., 26, 1621–1637. ESRI (2011), ArcGIS Desktop: Release 10, Environ. Syst. Res. Inst., Redlands, Calif. Fletcher, M.-S., B. B. Wolfe, C. Whitlock, D. Pompeani, H. Heijnis, S. Haberle, P. Gadd, and D. M. J. S. Bowman (2014), The legacy of mid Holocene fire on a Tasmanian montane landscape, J. Biogeogr., 41(3), 476–488. Fletcher, M.-S., A. Benson, H. Heijnis, P. S. Gadd, L. C. Cwynar, and A. B. H. Rees (2015), Changes in biomass burning mark the onset an ENSO- influenced climate regime at 42°S in southwest Tasmania, Australia, Quat. Sci. Rev., 122, 222–232. Fogt, R. L., D. H. Bromwich, and K. M. Hines (2011), Understanding the SAM influence on the South Pacific ENSO teleconnection, Clim. Dyn., 36(7-8), 1555–1576. Fox-Hughes, P., R. Harris, G. Lee, M. Grose, and N. Bindoff (2014), Future fire danger climatology for Tasmania, Australia, using a dynamically downscaled regional climate model, Int. J. Wildland Fire, 23(3), 309–321. Garreaud, R. D., M. Vuille, R. Compagnucci, and J. Marengo (2009), Present-day South American climate, Palaeogeogr. Palaeoclimatol. Palaeoecol., 281(3-4), 180–195. Gomez, B., L. Carter, A. R. Orpin, K. M. Cobb, M. J. Page, N. A. Trustrum, and A. S. Palmer (2011), ENSO/SAM interactions during the middle and late Holocene, Holocene, doi:10.1177/0959683611405241. Gouhier, T. C., A. Grinstead, and V. Simko (2016), biwavelet: Conduct univariate and bivariate wavelet analyses (Version 0.20.10). [Available from http://github.com/tgouhier/biwavelet.] Grose, M. R., P. Fox-Hughes, R. M. Harris, and N. L. Bindoff (2014), Changes to the drivers of fire weather with a warming climate—A case study of southeast Tasmania, Clim. Change, 124(1-2), 255–269. Guilyardi, E. (2006), El Niño–mean state–seasonal cycle interactions in a multi-model ensemble, Clim. Dyn., 26(4), 329–348. Harris, S., N. Nicholls, and N. Tapper (2014), Corrigendum to: Forecasting fire activity in Victoria, Australia, using antecedent climate variables and ENSO indices, Int. J. Wildland Fire, 23(2), 290–293. Hennessy, K., C. Lucas, N. Nicholls, J. Bathols, R. Suppiah, and J. Ricketts (2005), Climate change impacts on fire-weather in south-east Australia, 88 pp. Herweijer, C., R. Seager, E. R. Cook, and J. Emile-Geay (2007), North American droughts of the last millennium from a gridded network of tree- ring data, J. Clim., 20(7), 1353–1376. Hill, K. J., A. Santoso, and M. H. England (2009), Interannual Tasmanian rainfall variability associated with large-scale climate modes, J. Clim., 22, 4383–4397. Holz, A., and T. T. Veblen (2012), Wildfire activity in rainforests in western Patagonia linked to the Southern Annular Mode, Int. J. Wildland Fire, 21(2), 114–126. Holz, A., T. Kitzberger, J. Paritsis, and T. T. Veblen (2012a), Ecological and climatic controls of modern wildfire activity patterns across southwestern South America, Ecosphere, 3(11), 1–25. Holz, A., S. Haberle, T. Veblen, R. D. Pol-Holz, and J. Southon (2012b), Fire history in western Patagonia from paired tree-ring fire-scar and charcoal records, Clim. Past, 8(2), 451–466. Jolly, W. M., M. A. Cochrane, P. H. Freeborn, Z. A. Holden, T. J. Brown, G. J. Williamson, and D. M. Bowman (2015), Climate-induced variations in global wildfire danger from 1979 to 2013, Nat. Commun., 6, 7537. Kiem, A. S., and S. W. Franks (2004), Multi-decadal variability of drought risk, eastern Australia, Hydrol. Processes, 18(11), 2039–2050. Kim, S.-J., J. Lü, and B.-M. Kim (2014), The Southern Annular Mode (SAM) in PMIP2 simulations of the last glacial maximum, Adv. Atmos. Sci., 31(4), 863–878. Krawchuk, M. A., M. A. Moritz, M.-A. Parisien, J. Van Dorn, and K. Hayhoe (2009), Global pyrogeography: The current and future distribution of wildfire, PLoS One, 4(4), e5102. Lau, K., and H. Weng (1995), Climate signal detection using wavelet transform: How to make a time series sing, Bull. Am. Meteorol. Soc., 76(12), 2391–2402. Lenton, T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf, and H. J. Schellnhuber (2008), Tipping elements in the Earth’s climate system, Proc. Natl. Acad. Sci. U.S.A., 105(6), 1786–1793. Liu, W., J. Lu, L. R. Leung, S.-P. Xie, Z. Liu, and J. Zhu (2015), The de-correlation of westerly winds and westerly-wind stress over the Southern Ocean during the Last Glacial Maximum, Clim. Dyn., 45, 3157–3168. Lough, J., and H. Fritts (1987), An assessment of the possible effects of volcanic eruptions on North American climate using tree-ring data, 1602 to 1900 AD, Clim. Change, 10(3), 219–239. Magilligan, F. J., and P. S. Goldstein (2001), El Niño floods and culture change: A late Holocene flood history for the Rio Moquegua, southern Peru, Geology, 29(5), 431–434. Mariani, M., and M. S. Fletcher (2016), The Southern Annular Mode determines inter-annual and centennial-scale fire activity in temperate southwest Tasmania, Australia, Geophys. Res. Lett., 43, 1702–1709, doi:10.1002/2016GL068082. Marshall, G. J. (2003), Trends in the Southern Annular Mode from observations and reanalyses, J. Clim., 16, 24. McWethy, D., P. Higuera, C. Whitlock, T. Veblen, D. Bowman, G. Cary, S. Haberle, R. Keane, B. Maxwell, and M. McGlone (2013), A conceptual framework for predicting temperate ecosystem sensitivity to human impacts on fire regimes, Global Ecol. Biogeogr., 22(8), 900–912. Meyers, S. D., B. G. Kelly, and J. J. O’Brien (1993), An introduction to wavelet analysis in oceanography and meteorology: With application to the dispersion of Yanai waves, Mon. Weather Rev., 121(10), 2858–2866. Moritz, M. A., M.-A. Parisien, E. Batllori, M. A. Krawchuk, J. Van Dorn, D. J. Ganz, and K. Hayhoe (2012), Climate change and disruptions to global fire activity, Ecosphere, 3(6), 1–22. Moy, C. M., G. O. Seltzer, D. T. Rodbell, and D. M. Anderson (2002), Variability of El Nino/Southern Oscillation activity at millennial timescales during the Holocene, Nature, 420, 162–165. Murphy, B. F., and B. Timbal (2008), A review of recent climate variability and climate change in southeastern Australia, Int. J. Climatol., 28(7), 859–879. Nicholls, N., and C. Lucas (2007), Interannual variations of area burnt in Tasmanian bushfires: Relationships with climate and predictability, Int. J. Wildland Fire, 16(5), 540–546. Parisien, M.-A. (2016), Science can map a solution to a fast-burning problem, Nature, 534(7607), 297. Parisien, M.-A., and M. A. Moritz (2009), Environmental controls on the distribution of wildfire at multiple spatial scales, Ecol. Monogr., 79(1), 127–154. Pausas, J. G., and E. Ribeiro (2013), The global fire–productivity relationship, Global Ecol. Biogeogr., 22(6), 728–736.

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 232 9 Geophysical Research Letters 10.1002/2016GL070572

Power, S., F. Delage, C. Chung, G. Kociuba, and K. Keay (2013), Robust twenty-first-century projections of El Niño and related precipitation variability, Nature, 502(7472), 541–545. Rees, A. B., L. C. Cwynar, and M.-S. Fletcher (2015), Southern Westerly Winds submit to the ENSO regime: A multiproxy paleohydrology record from Lake Dobson, Tasmania, Quat. Sci. Rev., 126, 254–263. Risbey, J. S., M. J. Pook, P. C. McIntosh, M. C. Wheeler, and H. H. Hendon (2009), On the remote drivers of rainfall variability in Australia, Mon. Weather Rev., 137(10), 3233–3253. Román-Cuesta, R. M., M. Gracia, and J. Retana (2003), Environmental and human factors influencing fire trends in ENSO and non-ENSO years in tropical Mexico, Ecol. Appl., 13(4), 1177–1192. Saji, N., B. Goswami, P. Vinayachandran, and T. Yamagata (1999), A dipole mode in the tropical Indian Ocean, Nature, 401(6751), 360–363. Sandweiss, D. H., K. A. Maasch, R. L. Burger, J. B. Richardson Iii, H. B. Rollins, and A. Clement (2001), Variation in Holocene El Niño frequencies: Climate records and cultural consequences in ancient Peru, Geology, 29(7), 603. Santín, C., S. H. Doerr, E. S. Kane, C. A. Masiello, M. Ohlson, J. M. Rosa, C. M. Preston, and T. Dittmar (2016), Towards a global assessment of pyrogenic carbon from vegetation fires, Global Change Biol., 22(1), 76–91. Scott, A. C., and I. J. Glasspool (2006), The diversification of Paleozoic fire systems and fluctuations in atmospheric oxygen concentration, Proc. Natl. Acad. Sci. U.S.A., 103(29), 10,861–10,865. Sharples, J. J., G. J. Cary, Fox-Hughes, S. Mooney, J. P. Evans, M.-S. Fletcher, M. Fromm, P. F. Grierson, R. McRae, and P. Baker (2016), Natural hazards in Australia: extreme bushfire, Clim. Change, doi:10.1007/s10584-016-1811-1. Swetnam, T. W., and J. L. Betancourt (1990), Fire-southern oscillation relations in the southwestern United States, Science (Washington), 249(4972), 1017–1020. Timmermann, A., M. Latif, A. Grötzner, and R. Voss (1999), Modes of climate variability as simulated by a coupled general circulation model. Part I: ENSO-like climate variability and its low-frequency modulation, Clim. Dyn., 15(8), 605–618. Torrence, C., and P. J. Webster (1999), Interdecadal changes in the ENSO-monsoon system, J. Clim., 12(8), 2679–2690. Toth, L. T., R. B. Aronson, S. V. Vollmer, J. W. Hobbs, D. H. Urrego, H. Cheng, I. C. Enochs, D. J. Combosch, R. van Woesik, and I. G. Macintyre (2012), ENSO drove 2500-year collapse of eastern Pacific coral reefs, Science, 337(6090), 81–84. Trenberth, K. E. (1984), Signal versus noise in the Southern Oscillation, Mon. Weather Rev., 112(2), 326–332. Turney, C. S. M., S. Haberle, D. Fink, A. P. Kershaw, M. Barbetti, T. T. Barrows, M. Black, T. J. Cohen, T. Correge, and P. P. Hesse (2006), Integration of ice-core, marine and terrestrial records for the Australian Last Glacial Maximum and Termination: A contribution from the OZ INTIMATE group, J. Quat. Sci., 21(7), 751–761. van der Kaars, S., N. Tapper, and E. J. Cook (2010), Observed relationships between El Niño–Southern Oscillation, rainfall variability and vegetation and fire history on Halmahera, Maluku, Indonesia, Global Change Biol., 16(6), 1705–1714. Veblen, T. T., T. Kitzberger, R. Villalba, and J. Donnegan (1999), Fire history in northern Patagonia: The roles of humans and climatic variation, Ecol. Monogr., 69(1), 47–67. Verdon, D. C., A. S. Kiem, and S. W. Franks (2004), Multi-decadal variability of forest fire risk-eastern Australia, Int. J. Wildland Fire, 13(2), 165–171. Wang, B., and Y. Wang (1996), Temporal structure of the Southern Oscillation as revealed by waveform and wavelet analysis, J. Clim., 9(7), 1586–1598. Wang, C. (2001), On the ENSO mechanisms, Adv. Atmos. Sci., 18(5), 674–691. Westerling, A. L., H. G. Hidalgo, D. R. Cayan, and T. W. Swetnam (2006), Warming and earlier spring increase western US forest wildfire activity, Science, 313(5789), 940–943. Whitlock, C., P. I. Moreno, and P. Bartlein (2007), Climatic controls of Holocene fire patterns in southern South America, Quat. Res., 68(1), 28–36. Williams, A. A., D. J. Karoly, and N. Tapper (2001), The sensitivity of Australian fire danger to climate change, Clim. Change, 49(1-2), 171–191. Williamson, G. J., L. D. Prior, W. M. Jolly, M. A. Cochrane, B. P. Murphy, and D. M. Bowman (2016), Measurement of inter- and intra-annual variability of landscape fire activity at a continental scale: The Australian case, Environ. Res. Lett., 11(3), 035003. Yan, H., L. Sun, Y. Wang, W. Huang, S. Qiu, and C. Yang (2011), A record of the Southern Oscillation Index for the past 2,000 years from precipitation proxies, Nat. Geosci., 4(9), 611–614.

MARIANI ET AL. ENSO AND FIRE ACTIVITY IN SE AUSTRALIA 233 10 Coupling of the Intertropical Convergence Zone and Southern Hemisphere mid-latitude climate during the early to mid-Holocene

Michela Mariani1, Michael-Shawn Fletcher1, Russell N. Drysdale1, Krystyna M. Saunders2, Henk Heijnis2, Geraldine Jacobsen2, and Atun Zawadzki2 1School of Geography, University of Melbourne, Carlton, Victoria 3053, Australia 2Australian Nuclear Science and Technology Organisation, Kirrawee DC, New South Wales 2232, Australia

ABSTRACT sectors of the SH, which inhibits robust regional Conceptual models predict a tight coupling between the Intertropical Convergence Zone compilations. Indeed, the majority of previous (ITCZ) and the Southern Westerly Winds (SWW) in response to glacial-interglacial transi- studies present data from only one location or tions, yet little is known about this relationship under Holocene boundary conditions. Here record (e.g., Griffiths et al., 2009; Moreno et al., we present a synthesis of Holocene pollen data from the southwest Pacific mid-latitudes that 2014), which may be strongly biased by local tracks changes in the SWW. Comparison of our SWW paleoclimate records with data track- geographic factors (e.g., topography). ing the ITCZ, oceanic circulation, and insolation reveals clearly synchronous and in-phase Integrating data from multiple sites mitigates ITCZ-SWW dynamics between 12 and 5 ka, indicating a tight coupling between the tropics against the influence of local-scale factors and and southern mid-latitudes in response to ocean circulation and insolation. An apparent permits an understanding of regional trends decoupling of the SWW and ITCZ in the Pacific region after 5 ka is attributable to the over- driven by broad-scale drivers, such as climate. riding influence of the El Niño–Southern Oscillation (ENSO) over the proxy data. Here we integrate proxy data from multiple sites across western Tasmania, Australia, in an INTRODUCTION the early Holocene (11.7–9 ka), ice sheets were attempt to address critical gaps in our under- Hydroclimatic variability in the Southern retreating and NH summer insolation was high standing of Holocene ITCZ-SWW dynamics. Hemisphere (SH) is strongly modulated by shifts (Berger, 1978), resulting in a northward shift We exploit the tight coupling between climate, in both the southern margin of the Intertropical of the ITCZ (Haug et al., 2001). A subsequent vegetation, and fire activity, and the strong cor- Convergence Zone (ITCZ) (Koutavas and Lynch- weakening of NH summer insolation, in concert relation between climate and SWW speed in this Stieglitz, 2004) and the Southern Westerly Winds with the shift in the timing of Earth’s perihelion region. Indeed, the terrestrial climate of western (SWW) (Hendon et al., 2007). The ITCZ is a during the austral summer since the Holocene Tasmania hosts one of the strongest correlations belt of intense heating and deep atmospheric thermal maximum (9–5.5 ka) (Warren and Kutz- on Earth between changes in SWW speed and convection that marks the boundary of low-level bach, 1992), has resulted in a progressive south- rainfall anomalies (Gillett et al., 2006). We pres- confluence of north and south trade winds (e.g., ward migration of the boreal-summer ITCZ ent a compilation of new high-resolution pollen Schneider et al., 2014). The SWW are a promi- (Haug et al., 2001). Further south, evidence from data from five lakes within this region (Fig. 1) nent, zonally symmetric feature of SH climate hemisphere-wide paleoclimate proxy records and synthesize these new data with a recent that encircles Antarctica and controls rainfall and and transient simulations suggest zonally sym- western Tasmania paleofire reconstruction (Mar- temperature variability, as well as ocean circula- metric, reduced SWW flow between ca. 11 and iani and Fletcher, 2017) and with global paleo- tion, within a band spanning ~30°–70°S (Gar- 8 ka and enhanced SWW flow between 8 and 5 climate proxies to test the relationship between reaud et al., 2013; Gillett et al., 2006). ka at ~40°S (e.g., Fletcher and Moreno, 2012). ITCZ and SWW dynamics through the Holo- Coupling of the SWW and ITCZ has been An apparent breakdown in SWW zonal sym- cene. We focus on two key pollen taxa (Phyllo- proposed to explain the response of SH extra- metry indicated by paleoclimatic data after ca. cladus aspleniifolius and Eucalyptus) that track tropical climate during glacial-interglacial tran- 6–5 ka has been linked to the onset of mod- regional vegetation dynamics through time: (1) sitions to changes in the oceanic heat transfer ern-day El Niño–Southern Oscillation (ENSO) P. aspleniifolius is a hygrophilous rainforest tree via the Atlantic Meridional Overturning Circula- variability (Fletcher and Moreno, 2012), which (Allen et al., 2001) that has been shown to track tion (AMOC)and North Atlantic winter sea-ice imparts an asymmetric (anti-phased) hydrocli- millennial-scale trends in hydroclimate (Beck extent (Denton et al., 2010; Toggweiler et al., matic signature across the Pacific Ocean basin et al., 2017); and (2) Eucalyptus is a pyrophyte 2006). According to these conceptual models, a (Philander, 1983). that dominates under drier conditions (Bowman, southward shift in both the ITCZ and SWW at We currently rely on paleoclimatic proxy 2000) and is favored in western Tasmania by the commencement of glacial-interglacial tran- records of largely local and regional changes increased fire activity under a relatively dry and sitions occurs synchronously with a phase of from the tropics (e.g., Griffiths et al., 2009; variable hydroclimatic regime (Beck et al., 2017; prolonged AMOC weakening. A link between Kuhnt et al., 2015) to temperate regions (e.g., Fletcher et al., 2014). AMOC changes and movement of the ITCZ and Mariani and Fletcher, 2017; Lamy et al., 2010; SWW during the last deglaciation, and through- Moreno et al., 2010) to form the basis of our METHODS out the last glacial cycle, also finds support from current understanding of Holocene SH climate The five lake sediment cores were retrieved climate simulations (Montade et al., 2015) and dynamics. The lack of a hemispheric synthesis from Lake Isla (TAS1503; 41°58′13.91″S, proxy reconstructions (Markle et al., 2017). precludes an understanding of the possible cou- 145°39′55.57″E), Lake Selina (TAS1402; The applicability of the above conceptual pling between the low and high latitudes through 41°52′39.80″S, 145°36′34.01″E), Lake Nancy model of ITCZ-SWW dynamics under Holocene this period. This knowledge gap is in part due (TAS1107; 42°15′31.56″S, 145°49′37.62″E), boundary conditions remains untested. During to the lack of high-resolution data from key Lake Gwendolyn (TAS1106; 42°15′44.58″S,

GEOLOGY, December 2017; v. 45; no. 12; p. 1083–1086 | Data Repository item 2017385 | https://doi.org/10.1130/G39705.1 | Published online 27 October 2017 ©GEOLOGY 2017 Geological | Volume Society 45 | ofNumber America. 12 For | www.gsapubs.org permission to copy, contact [email protected]. 234 1083 binned into 100 yr classes prior to the regional compilation. Time resolution varies across the selected sediment records from median values of 33 yr/sample (TAS1108) to 170 yr/sample (TAS1106). The mean value across the five sites for each time step is considered representative of the regional abundance of the plant taxa (Euca- lyptus and P. aspleniifolius). Error envelopes were calculated as the standard deviation of the five records from the regional mean at each time step. All individual records for these taxa at each site are presented in Figs. DR2 and DR3.

RESULTS Here, despite the long human occupation (>40 k.y.) and clear evidence for an anthropogenic footprint over the vegetation landscape of western Tasmania (e.g., Mariani et al., 2017), we interpret millennial-scale trends in the pollen (and paleofire) data as a proxy for climatic change. Our rationale is based on the contention that sedimentary charcoal and pollen records from lake sediments are predisposed to detecting changes that occur over large spatial and temporal scales, whereas fire management by Australian Aborigines is (and was) uniformly small scale (e.g., Langton, 1998; Plomley, 1966). Rather than antagonistic, we argue that millennial-scale climatic change and human activity are synergistic through time, with trends in hydroclimate modulating the efficacy and impact of fire-management practices over these long time scales, a contention supported by the tight coupling between climate and paleofire activity in western Tasmania over the last Figure 1. A–C: Map of modern austral summer position of Intertropical Convergence Zone ~12 k.y. (Mariani and Fletcher, 2017). (ITCZ) plotted using gridded precipitation (A, B), and Southern Westerly Winds (SWW) plot- The regional compilation for P. aspleniifolius ted using zonal wind speed (C). D: Correlation map of SWW and rainfall in Tasmania with shows minimum values at ca. 10.5 ka followed pollen records analyzed in present work. Red stars show: 1—western Tasmania (40°–43°S); by a substantial increase during the mid-Holo- 2—Lake Gnotuk, southeastern Australia (38°S); 3—Lago Condorito, northern Patagonia (41°S); 4—El Niño proxy, Ecuador (3°N); 5—Cariaco Basin, Venezuela (10°N). White triangles cene (Fig. 2E). High values persisted until ca. 4 in D represent sites from western Tasmania analyzed in present study: a—Lake Selina; ka, after which a sharp decline is observed, cul- b—Lake Isla; c—Lake Gwendolyn; d—Lake Nancy; e—Lake Vera. Global rainfall and wind minating in minimum values at ca. 1.5 ka. Euca- speed data used for A–C are from European Centre for Medium-Range Weather Forecasts lyptus displays relatively high values at the start (ECMWF) ERA-Interim data set; correlation map for Tasmania (D) was developed using of the Holocene that reflect a glacial landscape precipitation data from Australian Bureau of Meteorology and zonal wind speed from U.S. National Oceanic and Atmospheric Administration (NOAA). dominated by cool climate eucalyptus among other alpine and subalpine taxa (Macphail, 1979). Stable, but low values of Eucalyptus persist 145°49′23.11″E), and Lake Vera (TAS1108; present (cal. yr B.P.) using the SH calibration between ca. 10.5 and 5 ka (Fig. 2F), followed 42°16′28.53″S, 145°52′47.73″ E) using a uni- curve (Hogg et al., 2013) (Table DR1 in the GSA by an increase after 5 ka. Error envelopes of versal gravity corer. The chronology of the sedi- Data Repository1). Age-depth modeling was per- the regional P. aspleniifolius composite curve ment cores analyzed in this study is based on formed in the CLAM package for R (Blaauw, range between 0.2 and 1.6 standard deviations a combination of 210Pb and 14C. Lead radioiso- 2010) and previously published by Mariani and (Fig. 2E). The range of errors recorded in the tope activity was determined on a total of 21 Fletcher (2017) (Fig. DR1). A total of 592 sam- Eucalyptus regional curve is between 0.13 and samples across four sites (TAS1503, TAS1106, ples from all the cores were analyzed for pollen 1.5 standard deviations (Fig. 2F). TAS1107, TAS1108) using alpha spectrometry and spores. Percentages for Eucalyptus and P. at the Australian Nuclear Science and Tech- aspleniifolius were extracted from the entire pol- DISCUSSION nology Organisation (ANSTO). A total of 51 len data sets. Data spanning the last 12 k.y. from bulk sediment samples across all the five sites all cores were standardized, interpolated, and SWW Dynamics in the Southwest Pacific were analyzed at the DirectAMS 14C laboratory during the Holocene (Bothell, Washington, USA), the NoSAMS 14C The composites of pollen and charcoal data 1 GSA Data Repository item 2017385, support- laboratory (Woods Hole, Massachusetts, USA), ing information, is available online at http://www reveal substantial millennial-scale variations 14 and the C laboratory at ANSTO. Radiocarbon .geosociety.org/datarepository/2017/ or on request that we interpret as indicating changes in SWW- dates were calibrated to calendar years before from [email protected]. derived moisture in western Tasmania during the

1084 www.gsapubs.org | Volume 45 | Number235 12 | GEOLOGY 2016). These shifts covary with proxy-inferred the clear evidence for an attenuation of the SWW paleoclimatic trends in Patagonia (e.g., Lamy et at their northern edge between 11.7 and 11 ka al., 2010; Mariani and Fletcher, 2017; Moreno, implies a southward displacement of both the 2004), in southeast mainland Australia (Wilkins et ITCZ and the northern limits of the SWW belt at al., 2013) (Fig. DR4), and across a broad swathe of the start of the Holocene, followed by a synchro- SWW-dominant rainfall zones in the SH (Fletcher nous northward migration of these two systems and Moreno, 2012), revealing hemisphere-wide through the early to mid-Holocene. SWW trends through this period. Our pollen and So what caused this apparent coupling of charcoal influx compilations indicate a shift to a the SSW and ITCZ through the Holocene? An drier, more fire-affected regional vegetation in extension of the Denton et al. (2010) model western Tasmania after ca. 5 ka, in concert with for glacial-interglacial transitions provides the persistently low lake levels in the SWW-rainfall most plausible explanation for synchronicity of zone of southeast Australia (Wilkins et al., 2013) ITCZ-SWW behavior during the early Holo- and a marked increase in the frequency of El Niño cene. Warming of the NH during the Holocene events (the warm phase of ENSO) recorded in the insolation maximum (between 11.7 and 11 ka) east Pacific (Fig. 2H) (Moy et al., 2002). ENSO drove increased ice-sheet melting. AMOC was exerts a major influence over modern climate in still in a relatively weakened state between 11.7 the Pacific Ocean basin and beyond, and there is and 11.0 ka following the meltwater flux associ- widespread recognition that ENSO has been an ated with the Younger Dryas (Fig. 2A), which important factor influencing moisture patterns and promoted the spread of winter sea ice in the fire regimes across the southwest Pacific since North Atlantic which, in turn, pushed the ITCZ the mid-Holocene (e.g., Donders et al., 2008; and SWW southward. A strengthening of the McGlone et al., 1992). The trends observed in AMOC between 11 and 7.5 ka is associated our western Tasmania pollen compilation, and with a northward movement of both the ITCZ in other records from the southwest Pacific, after and SWW, followed by relative stability at ca. ca. 5 ka may reflect either (1) shifts in intensity 7.5 ka (McManus et al., 2004) as ice-sheet melt- and/or latitudinal position of the westerlies or (2) ing was completed. This suggests an oceanic- ENSO-related moisture changes. forced northward shift of the ITCZ coupled with a shift of the northern margin of the SWW Coupling between ITCZ and SWW from the early to mid-Holocene. At ca. 8 ka, Latitudinal Shifts from Early to Mid- perihelion precessed toward its current austral Holocene (11.7–7 ka) summer position, reversing the trend in the inter- To understand the link between tropical and hemispheric temperature contrast (Marcott et al., Figure 2. Summary plot of records tracking the behavior of the ITCZ (Intertropical mid-latitude climate dynamics, we compare 2013) (Fig. 2D). Since then, insolation has likely Convergence Zone) and SWW (Southern our Tasmanian pollen data with existing paleo- remained the only driver of ITCZ shifts, result- Westerly Winds). Age axis is in thousands of climatic records from both the Pacific and the ing in a progressive southward displacement as calendar years before present (cal. kyr B.P.). Atlantic regions. A precipitation maximum at the SH has warmed (Schneider et al., 2014). 231 230 A: Pa/ Th as a proxy for Atlantic Meridi- the start of the Holocene in Indonesia and Bra- Records tracking ITCZ behavior since ca. 5 onal Overturning Circulation (AMOC) from North Atlantic (McManus et al., 2004). B: June zil, at the southern limit of the ITCZ (Cruz et ka seem to suggest a relative stability of a south- insolation at 60°N (Berger, 1978). SH—South- al., 2009; Kuhnt et al., 2015), in concert with erly displaced ITCZ belt, whereas we observe ern Hemisphere; NH—Northern Hemisphere. precipitation minima in southern China, at the large shifts in the Tasmanian pollen and paleo- C: Titanium concentration from Cariaco Basin, northern limit of the ITCZ (Yuan et al., 2004), fire data that suggest a trend toward an overall Venezuela (10°N) (Haug et al., 2001).D: Temper- ature contrast between Northern and Southern suggests a more southerly position of the ITCZ drier and more variable hydroclimate. We inter- Hemispheres (Marcott et al., 2013). E: Com- at this time relative to today (Fig. DR5). A grad- pret the apparent decoupling of the ITCZ and posite Phyllocladus aspleniifolius curve from ual increase in precipitation at 10°N in South SWW after 5 ka in our proxy data as the result western Tasmania (WTAS). F: Composite Euca- America between 11.7 and 6 ka, inferred from of an increasing importance of ENSO over the lyptus curve from western Tasmania. Shaded changes in elemental Titanium (Ti) (Fig. 2C) , Pacific region concurrent with a period of rela- areas in E and F represent composite error envelopes. G: Regional charcoal influx from suggests a northward migration and stabilization tive AMOC stability. The strong influence of western Tasmania (Mariani and Fletcher, 2017). of the ITCZ through this period in the east Pacific ENSO on precipitation variability in Indonesia H: El Niño number of events per 100 yr (Moy sector (Haug et al., 2001). This shift in the ITCZ (Hendon, 2003) suggests that ENSO fluctuations et al., 2002). Red fillings indicate periods with through the early to mid-Holocene is synchro- over the last ~5 k.y., since the “switching on” of a frequency of events > 5 per 100 yr. nous with changes in AMOC strength (McMa- the “modern” ENSO regime (Moy et al., 2002), nus et al., 2004) (Fig. 2A): a weaker AMOC in has likely confounded the signals of ITCZ-SWW first half of the Holocene. A drier climate between the early Holocene (11.7–11 ka) when ITCZ is behavior in our proxy data. Indeed, the apparent 11.7 and 9 ka is indicated by concurrent minima in displaced southward, with a shift to a stronger southward migration of the ITCZ since 8–7 ka regional P. aspleniifolius pollen and high regional AMOC toward the mid-Holocene (until ca. 7.5 recorded in the Cariaco Basin (Venezuela) (Fig. charcoal influx, implying a reduction in SWW ka) contemporaneous with a northward displace- 2B) is evident in neither the Timor Sea (Kuhnt et flow over Tasmania and high fire activity in the ment of the ITCZ. Our data from 40° to 44°S in al., 2015) nor on Flores in Indonesia (Griffiths et west. The subsequent gradual increase in P. asple- Tasmania, at the northern margin of the SWW, al., 2009). The relative stability of these proxies niifolius in concert with a decrease in charcoal reveal remarkable synchronicity between shifts from the Indonesian region during the last 5 k.y. influx (Fig. 2) signals an increase in available in both the ITCZ and SWW between 12 and 7 ka. may reflect that any southward shift in the ITCZ moisture that is consistent with increasing SWW In combination with proxy data from 40° to 43°S (which would enhance regional rainfall) was buff- flow over the region (e.g., Mariani and Fletcher, in southern South America (e.g., Moreno, 2004), ered by the rainfall-reducing influence of ENSO.

GEOLOGY | Volume 45 | Number 12 | www.gsapubs.org 236 1085 ACKNOWLEDGMENTS linked to early Holocene sea-level rise: Nature McGlone, M.S., Kershaw, P., and Markgraf, V., 1992, Research was supported by Australian Research Geoscience, v. 2, p. 636–639, https://doi.org/10 El Niño/Southern Oscillation climatic variability Council grants (DI110100019, IN140100050, and .1038/ngeo605. in Australasian and South American paleoenvi- IN170100062) and two Australian Institute for Haug, G.H., Hughen, K.A., Sigman, D.M., Peterson, ronmental records, in Diaz, H.F., and Markgraf, Nuclear Science and Engineer (AINSE) awards L.C., and Röhl, U., 2001, Southward migration of V., eds., El Niño: Historical and Paleoclimatic (ANALGRA15029 and ALNGRA16024). MM was the Intertropical Convergence Zone through the Aspects of the Southern Oscillation: Cambridge, also supported by an AINSE Postgraduate Research Holocene: Science, v. 293, p. 1304–1308, https:// UK, Cambridge University Press, p. 435–462. Award (12039). We thank two anonymous reviewers doi.org/10.1126/science.1059725. McManus, J.F., Francois, R., Gherardi, J.-M., Keigwin, for comments on an earlier draft of this manuscript. Hendon, H.H., 2003, Indonesian rainfall variability: L.D., and Brown-Leger, S., 2004, Collapse and Impacts of ENSO and local air–sea interaction: rapid resumption of Atlantic meridional circu- REFERENCES CITED Journal of Climate, v. 16, p. 1775–1790, https:// lation linked to deglacial climate changes: Na- Allen, K.J., Cook, E.R., Francey, R.J., and Michael, doi.org/10.1175/1520-0442(2003)016<1775: ture, v. 428, p. 834–837, https://doi.org/10.1038 K., 2001, The climatic response of Phyllocladus IRVIOE>2.0.CO;2. /nature02494. aspleniifolius (Labill.) Hook. f in Tasmania: Jour- Hendon, H.H., Thompson, D.W.J., and Wheeler, Montade, V., Kageyama, M., Combourieu-Nebout, N., nal of Biogeography, v. 28, p. 305–316, https:// M.C., 2007, Australian rainfall and surface tem- Ledru, M.-P., Michel, E., Siani, G., and Kissel, C., doi.org/10.1046/j.1365-2699.2001.00546.x. perature variations associated with the Southern 2015, Teleconnection between the Intertropical Beck, K.K., Fletcher, M.-S., Gadd, P., Heijnis, H., and Hemisphere annular mode: Journal of Climate, Convergence Zone and southern westerly winds Jacobsen, G., 2017, An early onset of ENSO in- v. 20, p. 2452–2467, https://doi.org/10.1175 throughout the last deglaciation: Geology, v. 43, fluence in the extra-tropics of the southwest Pa- /JCLI4134.1. p. 735–738, https://doi.org/10.1130/G36745.1. cific inferred from a 14,600 year high resolution Hogg, A.G., et al., 2013, SHCal13 Southern Hemi- Moreno, P.I., 2004, Millennial-scale climate variabil- multi-proxy record from Paddy’s Lake, northwest sphere calibration, 0–50,000 years cal BP: Ra- ity in northwest Patagonia over the last 15000 yr: Tasmania: Quaternary Science Reviews, v. 157, diocarbon, v. 55, p. 1889–1903, https://doi.org Journal of Quaternary Science, v. 19, p. 35–47, p. 164–175, https://doi.org/10.1016/j.quascirev /10.2458/azu_js_rc.55.16783. https://doi.org/10.1002/jqs.813. .2016.12.001. Koutavas, A., and Lynch-Stieglitz, J., 2004, Variabil- Moreno, P.I., Francois, J.P., Moy, C.M., and Villa- Berger, A.L., 1978, Long-term variations of daily inso- ity of the marine ITCZ over the eastern Pacific Martinez, R., 2010, Covariability of the South-

lation and Quaternary climatic changes: Journal during the past 30,000 years, in Diaz, N.F., and ern Westerlies and atmospheric CO2 during the of the Atmospheric Sciences, v. 35, p. 2362–2367, Bradley, R.S., eds., The Hadley Circulation: Pres- Holocene: Geology, v. 38, p. 727–730, https://doi https://doi .org /10 .1175 /1520 -0469 (1978)035 ent, Past and Future: Dordrecht, Netherlands, .org/10.1130/G30962.1. <2362:LTVODI>2.0.CO;2. Springer, p. 347–369, https://doi.org/10.1007 Moreno, P.I., Vilanova, I., Villa-Martínez, R., Gar- Blaauw, M., 2010, Methods and code for ‘classical’ /978-1-4020-2944-8_13. reaud, R., Rojas, M., and De Pol-Holz, R., 2014, age-modelling of radiocarbon sequences: Quater- Kuhnt, W., Holbourn, A., Xu, J., Opdyke, B., De Southern Annular Mode-like changes in south- nary Geochronology, v. 5, p. 512–518, https://doi Deckker, P., Röhl, U., and Mudelsee, M., 2015, western Patagonia at centennial timescales over .org/10.1016/j.quageo.2010.01.002. Southern Hemisphere control on Australian mon- the last three millennia: Nature Communications, Bowman, D.M.J.S., 2000, Australian Rainforests: Is- soon variability during the late deglaciation and v. 5, 4375, https://doi .org/10.1038 /ncomm s5375. lands of Green in a Land of Fire: Cambridge, UK, Holocene: Nature Communications, v. 6, 5916, Moy, C.M., Seltzer, G.O., Rodbell, D.T., and Ander- Cambridge University Press, 360 p., https://doi https://doi.org/10.1038/ncomms6916. son, D.M., 2002, Variability of El Niño/South- .org/10.1017/CBO9780511583490. Lamy, F., Kilian, R., Arz, H.W., Francois, J.P., Kaiser, ern Oscillation activity at millennial timescales Cruz, F.W., Vuille, M., Burns, S.J., Wang, X., Cheng, J., Prange, M., and Steinke, T., 2010, Holocene during the Holocene: Nature, v. 420, p. 162–165, H., Werner, M., Edwards, R.L., Karmann, I., Auler, changes in the position and intensity of the south- https://doi.org/10.1038/nature01194. A.S., and Nguyen, H., 2009, Orbitally driven east– ern westerly wind belt: Nature Geoscience, v. 3, Philander, S.G.H., 1983, El Niño southern oscillation west antiphasing of South American precipitation: p. 695–699, https://doi.org/10.1038/ngeo959. phenomena: Nature, v. 302, p. 295–301, https:// Nature Geoscience, v. 2, p. 210–214, https://doi Langton, M., 1998, Burning Questions: Emerging doi.org/10.1038/302295a0. .org/10.1038/ngeo444. Environmental Issues for Indigenous Peoples in Plomley, N.J.B., ed., 1966, Friendly Mission: The Tas- Denton, G.H., Anderson, R.F., Toggweiler, J.R., Ed- Northern Australia: Darwin, Australia, Northern manian Journals and Papers of George Augustus wards, R.L., Schaefer, J.M., and Putnam, A.E., Territory University Centre for Indigenous Natu- Robinson, 1829–1834: Hobart, Tasmanian His- 2010, The last glacial termination: Science, ral and Cultural Resource Management, 89 p. torical Research Association, 1166 p. v. 328, p. 1652–1656, https://doi.org/10.1126 Macphail, M.K., 1979, Vegetation and climates in Schneider, T., Bischoff, T., and Haug, G.H., 2014, Mi- /science.1184119. southern Tasmania since the last glaciation: Qua- grations and dynamics of the intertropical con- Donders, T.H., Wagner-Cremer, F., and Visscher, H., ternary Research, v. 11, p. 306–341, https://doi vergence zone: Nature, v. 513, p. 45–53, https:// 2008, Integration of proxy data and model scenar- .org/10.1016/0033-5894(79)90078-4. doi.org/10.1038/nature13636. ios for the mid-Holocene onset of modern ENSO Marcott, S.A., Shakun, J.D., Clark, P.U., and Mix, A.C., Toggweiler, J.R., Russell, J.L., and Carson, S.R., 2006,

variability: Quaternary Science Reviews, v. 27, 2013, A reconstruction of regional and global Midlatitude westerlies, atmospheric CO2, and p. 571–579, https://doi.org/10.1016/j.quascirev temperature for the past 11,300 years: Science, climate change during the ice ages: Paleocean- .2007.11.010. v. 339, p. 1198–1201, https://doi.org/10.1126 ography, v. 21, PA2005, https://doi.org/10.1029 Fletcher, M.-S., and Moreno, P.I., 2012, Have the /science.1228026. /2005PA001154. Southern Westerlies changed in a zonally sym- Mariani, M., and Fletcher, M.-S., 2016, The Southern Warren, L.P., and Kutzbach, T.J.E., 1992, Sensitiv- metric manner over the last 14,000 years? A hemi- Annular Mode determines interannual and cen- ity of the Indian monsoon to forcing parameters sphere-wide take on a controversial problem: Qua- tennial‐scale fire activity in temperate southwest and implications for its evolution: Nature, v. 360, ternary International, v. 253, p. 32–46, https://doi Tasmania, Australia: Geophysical Research Let- p. 647–652, https://doi.org/10.1038/360647a0. .org/10.1016/j.quaint.2011.04.042. ters, v. 43, p. 1702–1709, https://doi .org /10 .1002 Wilkins, D., Gouramanis, C., De Deckker, P., Fifield, Fletcher, M.-S., Wolfe, B.B., Whitlock, C., Pompeani, /2016GL068082. L.K., and Olley, J., 2013, Holocene lake-level D., Heijnis, H., Haberle, S., Gadd, P., and Bow- Mariani, M., and Fletcher, M.-S., 2017, Long-term fluctuations in Lakes Keilambete and Gnotuk, man, D.M.J.S., 2014, The legacy of mid-Holo- climate dynamics in the extra-tropics of the southwestern Victoria, Australia: The Holo- cene fire on a Tasmanian montane landscape: South Pacific revealed from sedimentary charcoal cene, v. 23, p. 784–795, https://doi.org/10.1177 Journal of Biogeography, v. 41, p. 476–488, analysis: Quaternary Science Reviews, v. 173, /0959683612471983. https://doi.org/10.1111/jbi.12229. p. 181–192, https://doi.org/10.1016/j.quascirev Yuan, D., Cheng, H., Edwards, R.L., Dykoski, C.A., Garreaud, R., Lopez, P., Minvielle, M., and Rojas, .2017.08.007. Kelly, M.J., Zhang, M., Qing, J., Lin, Y., Wang, M., 2013, Large-scale control on the Patagonian Mariani, M., Connor, S.E., Fletcher, M.S., Theuerkauf, Y., and Wu, J., 2004, Timing, duration, and transi- climate: Journal of Climate, v. 26, p. 215–230, M., Kuneš, P., Jacobsen, G., Saunders, K.M., and tions of the last interglacial Asian monsoon: Sci- https://doi.org/10.1175/JCLI-D-12-00001.1. Zawadzki, A., 2017, How old is the Tasmanian ence, v. 304, p. 575–578, https://doi .org /10 .1126 Gillett, N.P., Kell, T.D., and Jones, P.D., 2006, Regional cultural landscape? A test of landscape openness /science.1091220. climate impacts of the Southern Annular Mode: using quantitative land‐cover reconstructions: Geophysical Research Letters, v. 33, L23704, Journal of Biogeography, https://doi .org /10 .1111 https://doi.org/10.1029/2006GL027721. /jbi.13040 (in press). Manuscript received 28 June 2017 Griffiths, M.L., Drysdale, R.N., Gagan, M., Zhao, J.- Markle, B.R., et al., 2017, Global atmospheric tele- Revised manuscript received 12 September 2017 x., Ayliffe, L., Hellstrom, J.C., Hantoro, W., Fri- connections during Dansgaard-Oeschger events: Manuscript accepted 14 September 2017 sia, S., Feng, Y.-x., and Cartwright, I., 2009, In- Nature Geoscience, v. 10, p. 36–40, https://doi creasing Australian–Indonesian monsoon rainfall .org/10.1038/ngeo2848. Printed in USA

1086 www.gsapubs.org | Volume 45 | Number237 12 | GEOLOGY Appendix A2 - PhD research outputs (first authorship only)

PUBLISHED(P)/UNDER REVIEW OUTPUT TITLE/LOCATION (UR)/COMPLETED (C)/ TYPE IN PREPARATION (PR) How old is the Tasmanian cultural landscape? A test of landscape openness using Research P quantitative land-cover article reconstructions. JOURNAL OF BIOGEOGRAPHY Testing quantitative pollen dispersal models in animal- pollinated vegetation mosaics: Research P An example from temperate article Tasmania, Australia QUATERNARY SCIENCE REVIEWS The Southern Annular Mode determines inter-annual and centennial-scale fire activity Research P in temperate southwest article Tasmania, Australia GEOPHYSICAL RESEARCH LETTERS Long-term climate dynamics in the extra-tropics of the Research South Pacific revealed from P article sedimentary charcoal analysis QUATERNARY SCIENCE REVIEWS ENSO controls interannual fire activity in southeast Research P Australia. article GEOPHYSICAL RESEARCH LETTERS Coupling of the Intertropical Convergence Zone and the Research Southern Hemisphere mid- P article latitude climate during the Holocene GEOLOGY Research Climate change amplifications UR article of fire-climate teleconnections

238 in the Southern Hemisphere PNAS Biogeochemical responses to Holocene hydro-climate fluctuations in the Wilderness Research UR World Heritage Area of article Tasmania, Australia JOURNAL OF GEOPHYSICAL RESEARCH SOUTHERN CONNECTION Conference CONFERENCE, PUNTA C talk ARENAS, CHILE - JANUARY 2016 PALAEOCLIMATE OF THE Workshop SOUTHERN HEMISPHERE, C talk SANTIAGO, CHILE – NOVEMBER 2016 AUSTRALASIA QUANTERNARY Conference C CONFERENCE, AUCKLAND, talk NEW ZEALAND – DECEMBER 2016 AUSTRALASIA QUANTERNARY Conference C CONFERENCE, AUCKLAND, poster NEW ZEALAND – DECEMBER 2016 ASLO 2017 – MOUNTAINS TO Conference C THE SEA, HONOLULU, USA - talk FEB/MARCH 2017 Conference PAGES 2017, ZARAGOZA, C poster SPAIN – MAY 2017 WILDFIRE PIRE WORKSHOP, Workshop C CHICO SPRINGS, MONTANA, talk USA - MAY 2017

239 Appendix A3 - List of all courses/workshops/ laboratory attended during the PhD candidature

• Workshop Wildfire PIRE (organized by Cathy Whitlock and the PIRE group), Chico Springs, Montana, USA – 22-25th May 2017 • Hydrogen Pyrolysis procedure for radioactive carbon (14C) sample pre- treatment at James Cook University – 10-13th April 2017, supervised by Jordahna Haig • Environmental Reconstruction Workshop (as part of the organizers), University of Melbourne – 23rd March 2017 • R COURSE led by Gavin Simpson, University of Adelaide - 13-17th February 2017 • Workshop Palaeoclimate of the Southern Hemisphere (organized by SHAPE and SWEEP), Santiago, Chile – 2nd/4th November 2016 • Carbon (13C) and Nitrogen (15N) stable isotopes sample pre-treatment at ANSTO – 7/10th November 2016, supervised by Robert Chisari • Radioactive Carbon (14C) sample pre-treatment at ANSTO – 7/11th November 2016, supervised by Geraldine Jacobsen and Alan Williams • R COURSE - Statistical modelling with R (provided by Jumping Rivers) - University of Newcastle, Newcastle upon Tyne (UK) – 13th September 2016 • R COURSE - Programming with R (provided by Jumping Rivers) - University of Newcastle, Newcastle upon Tyne (UK) – 14th September 2016 • R COURSE – Efficient programming with R (provided by Jumping Rivers) - University of Newcastle, Newcastle upon Tyne (UK) – 15th September 2016 • POLQUANT Summer School for quantitative vegetation reconstruction (CNRS), Moulis, France – 28th Aug/4thSep 2016 • Carbon (13C) and Nitrogen (15N) stable isotopes sample pre-treatment at ANSTO – 1st/5th August 2016, supervised by Robert Chisari

240 • Exploratory Spatial Data Analysis and Geostatistical Interpolation at ESRI Australia – 28th July 2016 • ArcGIS III: Performing Analysis (10.3) at ESRI Australia – 2nd/3rd June 2016 • Radioactive Lead (210Pb) sample pre-treatment at ANSTO – 2nd/6th May 2016, supervised by Atun Zawadzki and Brodie Cutmore • Radioactive Carbon (14C) sample pre-treatment at ANSTO – 2nd/6th May 2016, supervised by Geraldine Jacobsen and Fiona Bertuch • Landscape Reconstruction Algorithm (LRA) – University of Melbourne (Presented by Petr Kuneš) – 15th February 2016 • Radioactive Lead (210Pb) sample pre-treatment at ANSTO – 31stAug/4th Sep 2015, supervised by Atun Zawadzki and Jack Goralewski • Assessment of PaleoEnvironments (SHAPE) – University of New South Wales (Presented by Steven Phipps) – 25th/28th Feb 2015

241 Appendix A4 - My PhD in numbers

Number of PhD days until submission 1056

Number of pollen samples counted (incl. work extra- 942 PhD)

Number of pollen grains counted 276,200

Number of charcoal samples analysed (incl. work extra- 1,156 PhD)

Number of charcoal particles counted 20,377

Number of days spent in the field 38

Estimated number of days spent in the lab 280

Estimated number of hours spent at the pollen 380 microscope

Number of references in my EndNote 2,210

Number of R scripts created/edited 51

Number of pollen samples processed (incl. work extra- 1216 PhD)

Estimated number of lunches had in the corridor ~200 outside the lab sitting on the floor Estimated number of kilometers traveled for Fieldwork/Conferences/Workshops/Laboratory work 54,017 (one way)

242

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Mariani, Michela

Title: Palaeofire activity in western Tasmania: climate drivers and land-cover changes

Date: 2017

Persistent Link: http://hdl.handle.net/11343/208771

File Description: Palaeofire activity in western Tasmania: climate drivers and land-cover changes

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.