RETRACTED: Propensities of Old Growth, Mature and Regrowth Wet Eucalypt Forest, and Eucalyptus nitens Plantation, to Burn During Wildfire and Suffer Fire-Induced Crown Death

Abstract

There are conflicting conclusions on how the flammability of wet eucalypt forests changes in the time after disturbances such as logging or wildfire. Some conclude that forests are most flammable in the decades following disturbance, while others conclude that disturbance has no effect on flammability. The comparative flammability of Eucalyptus nitens plantations in the same environment as wet eucalypt forest is not known. We determined fire incidence and fire severity in regrowth, mature and old growth wet eucalypt forest, and E. nitens plantation, in the Huon Valley, Tasmania after the January–February 2019 wildfire. To control for topographic variation and fire weather, we randomly selected sites within the fire footprint, then randomly located a paired site for each in different forest types in the same topographic environment within 3 km. Each pair of sites was burned on the same day. Old growth forest and plantations were the least likely to burn. Old growth and mature forest exhibited scorched eucalypt crowns to a much lesser degree than regrowth forests. In a comparison of paired sites, plantation forest was less likely to burn than combined mature and old growth forests, but in all cases of detected ignition the canopy of plantation was scorched. The lower flammability of older forests, and their importance as an increasing store of carbon, suggests that a cessation of logging outside plantations might have considerable benefits.

1. Introduction

Unplanned forest fires frequently kill people and destroy property. For example, the 2009 Black Saturday fires caused the loss of 179 human lives and more than 2000 homes [1]. Thirty-three people died (24th January 2020) during the 2019/2020 fires in Victoria and New South Wales [2]. After extreme events like these, there is often a short-lived political urgency to take actions to prevent future ferocious fire events. Under climate change, fire incidence is expected to increase over the next century in both the Southern and Northern hemispheres, particularly in wet forests [3,4,5]. Fire hazard in forests, human land use and policy are connected in socio-ecological systems [6]. The effect of disturbance history on fire hazard in forests is specific to vegetation characteristics and fire regime [7]. Understanding the ways in which forest management affects fire behavior and incidence is imperative. The present study investigates the effect on fire incidence and severity of two forest management practices common in Tasmania; clearfelling and plantation establishment.

The study area supports wet, temperate forests; specifically, wet Eucalyptus (predominantly Eucalyptus regnans and Eucalyptus obliqua) dominated forests which have an understorey of broad-leaved shrubs or rainforest trees. The wet forests that are dominated by large eucalypts more than 110 years old have not been substantially disturbed by logging so are here described as ‘old-growth’ [8]. Clearfelling and the closely related ‘aggregated-retention’ (in which patches in the clearfelled area are retained) are used to harvest these forests. Clearfelling is followed by high intensity burning and aerial seeding of local eucalypt species [9]. This results in extremely dense regeneration followed by self-thinning [10,11]. Our study area contains a patchwork of coupes which have undergone clearfell logging at various times in the last 60 years, eucalypt plantations and old growth forest [8].

In wet forests in temperate zones, fire weather, ignition events and fuel moisture content limit fire activity [12,13,14]. This contrasts with dry ecosystems, in which fuel load, rather than fuel moisture content, is often a limiting factor [4,14]. In south-eastern Australia there is high daily variability of fire weather. Landscape fire events can occur in wet forests when there is an ignition on one of the infrequent days of extreme fire weather [15,16]. Analysis of fire weather data and active fire detections from moderate-resolution imaging spectroradiometer (MODIS) instruments has shown that fire weather is a strong driver of fire variability both on diurnal and larger time scales [17]. Post-fire analysis of fire severity in dry sclerophyll forests has found that weather is a stronger predictor of crown fire than topography and time since last fire [13].

In Australia there is yet to be a consensus on the effects of natural disturbance or clearfell logging on fire regimes. An analysis on fire feedbacks in the Australian Alps found regenerating tall wet forests were 8.3 times more likely to burn in the 58 years after fire than mature forests that had not been burned for much longer [18]. Surface fine fuels and overall fuel hazard have been found to accumulate to an equilibrium after disturbance [19,20]. Regenerating forests have been found to have a more vertically and horizontally continuous fuel layer which increases ease of fire spread [21]. The understorey species composition can change to more flammable species in response to changing fire regimes [11,19,22]. Species composition determines flame dimensions and therefore fire behaviour [23,24]. In contrast to the above, modeling of fire behavior related to fuel availability and structure of different forest age classes in south-east Australia found the highest flame height, probability of crown fire, and rate of spread to be in mature tall-mixed forests [25]. Surface fuel moisture has also been found to be higher in the first decade after fire, suggesting that young forests should be less available to burn [26].

Taylor et al. [11] found a strong relationship between stand age in Eucalyptus regnans forest in Victoria and the severity of fire, with the highest severity fires occurring in 7–36 year old stands. In the same forest type Bowman et al. [27] found that stand age only had a minor influence on fire severity compared to the strong influence of fire weather. Attiwill et al. [28] likewise claimed that there is no relationship between timber harvesting and fire extent and severity.

Our study analyses the incidence and severity of the 2019 Riveaux Road fire in Southern Tasmania in four forest types: regrowth, mature forest, old growth forest and plantation. Fire severity (the effect of fire on the vegetation) rather than fire intensity (the energy produced by the fire) was analysed in three broad categories of unburnt, understorey fire or crown fire [29]. Fire severity is considered more pertinent than fire intensity in our analysis as it relates to how the forest responds to fire and therefore creates feedbacks. Fire severity can also indicate fire intensity when comparing similar vegetation types [30]. We consider regrowth forest to be undergoing the process of self-thinning after logging, while mature forest has largely completed self-thinning. Old-growth forest was defined as that which had less than 30 crowns per hectare and crowns of irregular shape indicating mature and senescing trees [31]. The plantations in the study are Eucalyptus nitens plantations of various ages. They have a uniform structure of evenly spaced trees and little to no understorey [8]. We expected to find that old-growth forests burnt at a lower rate and lower severity than younger forests.

2. Materials and Methods

2.1. Study Area

The study area was 17,406 ha of the Huon Valley in southern Tasmania (Figure 1). The mean daily minimum and maximum temperatures for January are 8.4 °C and 19.6 °C, respectively, and for July are 2.4 °C and 8.6 °C, respectively. The average annual rainfall is 1762 mm (Warra weather station, 43.06° S, 146.70° E; Australian Bureau of Meteorology, Jan 2020). The natural vegetation of the study area is wet eucalypt forest, mostly dominated by Eucalyptus regnans or Eucalyptus obliqua. Old-growth forests in this region often contain extremely tall eucalypt trees (mature trees often between 60 and 100 m) and have an understorey dominated by cool temperate rainforest species (mixed forests) or broad-leaved tall shrubs and small trees (wet sclerophyll forests) [32]. The area has been managed for timber extraction for many decades and contains a mosaic of old-growth forests, native forest regrowth and Eucalyptus nitens plantations [8,32]. Sixty-four thousand hectares were burned in the Riveaux Road fire, which started on 15th of January and continued burning until mid-February 2019 [33]. Lightning strikes ignited several fires to the west of Geeveston which amalgamated and spread to the east and south. The fires of the 2018–19 season in Tasmania caused the destruction of six homes and burned 14% of Tasmania’s tall forests as well as significant areas of Gondwanan myrtle-beech forests and alpine vegetation [33]. The region experienced record low rainfall during January 2019 and summer temperatures 1.6 °C above average [33].

2.2. Study Design

Sampling was designed to minimise the effects of variation in topography and fire weather on the relationships between disturbance history and both fire incidence and fire severity. This outcome was achieved by randomly selecting 50 pairs of topographically similar sites. The study area was gridded in 8 km² cells. Initial sample points were randomly located, with a maximum of one in each grid cell. Sample points were located on pre-fire imagery [34] from which we characterised the forest type. The slope, slope position, aspect and elevation were determined from Google Maps. A paired random location in another forest type was sought within a 3 km radius of the initial selection. Random selections were made until the second site had no more than 40 m difference in elevation, 90° difference in aspect and 20% difference in slope. The pairs also had to occur in the same slope position (ridge, upper slope, lower slope, valley) or an adjacent one.

Plantations were identified through TasVeg Live [35]. Other forest types were determined from high resolution aerial photography. Regrowth forests were defined as those with either more than 100 visible crowns per hectare or very young regenerating stands with ground visible between the trees. Mature forests had 30 to 99 eucalypt crowns per hectare and rounded regular canopies indicating even age. Old growth forests had some (up to 29 per ha) emergent, unevenly shaped eucalypt crowns and smaller trees beneath.

Post-fire aerial ESRI images [36] were analysed to determine fire severity in the 0.1 ha around the point and check for any changes of forest type. Completely green forest was regarded as ‘unburned’. Plots labelled ‘understorey fire’ had at least some remaining green in the tree crowns but were brown or black beneath. Completely brown or black forest was labelled as ‘crown fire’. Some of the apparent crown scorch in these cases may have occurred between the fire and the taking of the image, and not have resulted from fire in the canopy, but rather from heat from below or the effects of ringbarking by fire.

Active fire detections obtained from moderate resolution imaging spectroradiometer (MODIS) instruments on the Terra and Aqua satellites were used to estimate the time frame in which each point burnt [37]. Each sample point was estimated to have burnt between the earliest and latest dates of the five closest MODIS hotspot detections. Cloud cover can result in false negative detections, and this has, in some cases, resulted in a broad potential burn period [38]. If the sample point occurred within 300 m of a MODIS hotspot detection it was estimated to have burnt on the same day/night as the hotspot detection, as fire severity has been found to be spatially dependent within this distance [13]. Forest Fire Danger Index ratings were calculated in three hourly increments using wind velocity, temperature and relative humidity data from Warra weather station (Warra weather station, 43.15° S, 146.90° E; Australian Bureau of Meteorology, Mar 2020). Drought factor was calculated from data from Geeveston (Fourfoot Rd Weather Station, 43.06° S, 146.70° E; Australian Bureau of Meteorology, Jan 2020), using soil dryness index (SDI) as the drought factor input. The mean and maximum FFDIs over the estimated time frame at which each point was burned were determined.

2.3. Statistical Analyses

The data were analysed as a full set, and for individual pairs of forest types. Chi² was used to determine whether the relationship between fire severity classes and forest types deviated from random. It was also used to determine if the relationship between topographic position classes and forest types deviated from random. One-way analysis of variance was used to determine if forest types differed in elevation, slope, aspect, latitude and longitude, as well as maximum and mean FFDI. Minitab18 was used for analyses.

3. Results

There were no significant differences between forest types in any of the environmental conditions for the data set as a whole (

Table 1. Fire severity and environmental conditions by forest type, showing significance of differences (Chi² (C) and ANOVA). Column percentages are shown for the Chⁱ² analyses.

	Old Growth	Mature	Regrowth	Plantation	P
Unburned (%)	41	21	17	40
Understorey fire (%)	48	50	22	0
Crown fire (%)	11	29	60	60
					0.001 (C)
Valley (%)	22	32	29	50
Lower slope (%)	56	46	54	40
Upper slope (%)	22	21	17	10
					0.783 (C)
Elevation (m)	233	187	197	150	0.185
Slope (%)	21	17	24	16	0.822
Aspect (°)	128	144	137	113	0.922
Average maximum potential FFDI	20.32	21.13	24.54	19.22	0.17
Mean potential FFDI	11.07	11.05	12.15	12.94	0.51

) and for any of the paired forest types. There were no significant differences in mean or maximum FFDI experienced in sampled forest types (Table 1). The maximum potential FFDI for old growth, regrowth and mature forest was 38.78, and the minimum possible was 0.52. For plantations the maximum potential FFDI was 25.3 and the minimum was 0.54.

For the full data set, there was strong variation in severity between forest types. Old growth forests burned less than mature forests, which in turn burned slightly less than regrowth (Table 1, Figure 2). Plantations burned in the same proportion as old growth (Table 1, Figure 2). However, in both plantation and regrowth, fires crowned to a much greater extent than in mature and old growth forest (Table 1, Figure 3). There were few crown fires in old growth forest (Table 1, Figure 3). In the paired comparisons between forest types, old growth forest burned less than mature forest but not regrowth forests (

Table 2. Percentage of stands that were burned and percentage that had crown fires by forest type, showing significance of differences between forest types (Chi²).

	Old Growth	Mature	Regrowth	Plantation	P
Old growth vs. mature (n = 10)
Burned	0	60	na	na	0.038
Crown fire	0	0	na	na	1.000
Old growth vs. regrowth (n = 38)
Burned	68	na	79	na	0.461
Crown	11	na	53	na	0.005
Mature and old growth vs. plantation (n = 20)
Burned	100		na	60	0.025
Crown	40		Na	60	0.371
Mature vs. regrowth (n = 32)
Burned	na	75	88	na	0.365
Crown	Na	31	69	na	0.034

). However, old growth forest experienced fewer crown fires than regrowth (Table 2). Mature forests also experienced fewer crown fires than regrowth (

Table A1. List of sample points with coordinates and forest type.

Figure Type	Latitude	Longitude
Mature Forest N = 28	−43.08303	146.64068
	−43.1454	146.6975
	−43.18016	146.69161
	−43.07995	146.71744
	−43.09978	146.73659
	−43.14320117	146.7217038
	−43.11016	146.76972
	−43.14243	146.75797
	−43.11317	146.82207
	−43.12675716	146.7633651
	−43.0801363	146.7491124
	−43.11205417	146.6818051
	−43.19338837	146.706203
	−43.15118863	146.7092791
	−43.09835272	146.813395
	−43.10181905	146.6846046
	−43.0960364	146.7176345
	−43.16487128	146.8008163
	−43.12446248	146.805289
	−43.12257778	146.732741
	−43.05575735	146.7370589
	−43.06846053	146.7907608
	−43.06377419	146.8190625
	−43.15994797	146.8146598
	−43.14499515	146.803759
	−43.09941536	146.6808368
	−43.08424	146.67764
	−43.07956	146.77419
Old Growth N = 27	−43.08515	146.64824
	−43.12263	146.6476
	−43.0655074	146.7038219
	−43.08431	146.70658
	−43.12503	146.70664
	−43.08882475	146.6577918
	−43.19767106	146.6904315
	−43.17274107	146.7948133
	−43.06194675	146.7333377
	−43.1395778	146.7008108
	−43.1061844	146.6928751
	−43.12869	146.75099
	−43.05016	146.78821
	−43.1146543	146.6837872
	−43.17519516	146.7040771
	−43.20208056	146.7122196
	−43.21392569	146.6989478
	−43.04989754	146.7596096
	−43.09795	146.75448
	−43.13735028	146.7635581
	−43.05699434	146.7330971
	−43.08764566	146.8224413
	−43.07408108	146.7430955
	−43.07691053	146.7238735
	−43.17294907	146.8107889
	−43.08509	146.68944
	−43.07502	146.7679
Plantation N = 10	−43.15255	146.69509
	−43.18547	146.69224
	−43.11973	146.70522
	−43.10448	146.73063
	−43.11764	146.81243
	−43.07688353	146.749029
	−43.11737	146.74405
	−43.05044	146.80162
	−43.1008982	146.8223385
	−43.06859812	146.7831244
Regrowth N = 35	−43.11947	146.65409
	−43.06204	146.70617
	−43.13761358	146.7234643
	−43.11815	146.76958
	−43.15569	146.75947
	−43.08713092	146.665777
	−43.12284072	146.7610805
	−43.20281548	146.693155
	−43.16940257	146.7962776
	−43.0633208	146.7311239
	−43.13596986	146.6994332
	−43.11101999	146.7008032
	−43.18401673	146.6963299
	−43.21111571	146.7092367
	−43.21876749	146.696354
	−43.18973554	146.7034153
	−43.15551434	146.7100481
	−43.05176174	146.7614738
	−43.1013635	146.6893879
	−43.09937212	146.7055105
	−43.10931557	146.7421104
	−43.15635164	146.7914153
	−43.14272755	146.7683749
	−43.12690012	146.8009437
	−43.11625033	146.7345929
	−43.05748329	146.7374861
	−43.05095175	146.7339691
	−43.06637026	146.8211688
	−43.08214941	146.8213097
	−43.15929051	146.8167518
	−43.14340951	146.8013612
	−43.07417457	146.7459314
	−43.07860165	146.7252177
	−43.17019969	146.8031068
	−43.09698709	146.6768909

). Pooled with mature forest, old growth forest burned more than plantations, although the incidence of crown fire did not differ (Table 2).

4. Discussion

Much of the world’s forest estate (and 12% of Tasmania’s land mass) is managed for timber production in a context in which unplanned forest fires are likely to increase in frequency as a result of climate change [3,4,5,39]. We sought to consolidate understanding of the effects of both the absence of clearfelling and the time since clearfelling on the fire hazard of commercial and reserved wet eucalypt forests. Using a methodological approach that minimised the effects of topographic variation and fire weather on our comparisons, we addressed the question ‘does clearfelling affect flammability?’ asked by Taylor et al. [11], Bowman et al. [27] and Attiwill et al. [28] in similar forest types. Our results are consistent with those of Taylor et al. [11], in that they show a higher incidence of canopy scorch or consumption by fire in regrowth forests than in both mature and old growth forests.

As the majority of our samples burned under relatively moderate fire weather conditions (average maximum FFDIs did not exceed 25), our results pertain to moderate rather than extreme fire weather conditions. Fire behaviour is less sensitive to variations in fuel under extreme conditions [13,40]. For example, fuel age was found to have no effect on fire severity in extreme fire weather conditions during Victoria’s Black Saturday fires [41]. This may explain the inconsistency of our results from those of Bowman et al. [27] and Attiwill et al. [28], as those studies find the effects of stand age minimal in comparison to the full range of fire weather.

Our results are also consistent with some other observations of changes in flammability after disturbance from mainland Australia. Extremely dense regrowth, up to 200,000 stems per ha in the first 7 years, undergoes thinning in the next 40 to 70 years, resulting in the build-up of dry fuel [42,43]. Younger stands of Eucalyptus trees have a more vertically continuous fuel load than older forests as they have a lower canopy, resulting in more frequent canopy consumption than in older stands [11,22,42]. Fuel load reaches an equilibrium after 40–70 years [19], but fuel moisture may continue to increase with age as the understorey trees and shrubs form closed canopies and water-holding moss mats and fern beds develop [19,44]. As fuel moisture is the limiting factor for fire in wet forests [4], fire hazards will stabilise or continue to decrease after the fuel load has stabilised. However, in a study with a small sample size, Cawson et al. [26] observed that fuel moisture was higher in forests recently burnt in a high severity fire than in long unburnt forests and forests recently burnt at low intensity. Modelling using fuel structure as the input has also predicted higher fire severity in mature tall mixed forests [25]. These discrepancies indicate the difficulty in predicting the effect of disturbance history on fire using single predictors.

Landscape traps are phenomena whereby due to a multitude of feedbacks between human and natural disturbance regimes, ecosystems in a landscape change to, and are maintained in, a new and compromised structural and functional state [44]. This concept relates to Jackson’s ‘alternative stable states’ model of vegetation and fire dynamics in south west Tasmania, in which a vegetation type will remain in a landscape due to stabilising feedbacks up to a threshold level of fire disturbance (or lack thereof), after which it will switch into a new stable state, in turn maintained by stabilising feedbacks [45]. Our study predicts a positive feedback mechanism by which the high intensity fire that is used to regenerate clearfelled coupes in wet forests increases the severity of fire in regenerating forests. This feedback will cause wet eucalypt forests to be maintained in a younger age class. If the frequency of canopy-consuming fire increases to more than two fires within 20–30 years, this may cause the localised extinction of obligate seeders, such as E. regnans [21,27,44].

We found that plantations were less likely to burn than other forest types, but if ignition was detected, the incidence of crown scorch or consumption was 100%. Eucalyptus globulus plantations in southern Australia can support a crown consuming fire at six years after planting. At this age there is ample leaf litter and ribbons of oil-rich bark hanging from the trees [46,47]. Eucalyptus nitens trees have similarly ribboning bark. The low observed occurrence of fire in plantations in our study may reflect the highly managed state of plantation forests, with surrounding roads and tracks acting as fire breaks and little understorey and no self-thinning, as well as the recency of establishment of some of the plantations. The dense canopy of older plantations may have reduced ability to detect understorey fires from aerial imagery.

Our study has a number of caveats. We assume random distribution of different fire weather conditions (quantified as FFDI values) across the different forest types sampled to make our conclusions on the relative impacts of fire on forests of different disturbance histories. Our assumption appears to be correct, though the precision of data on fire progression, and therefore the FFDI at which each point burnt, is limited. High resolution mapping of fire progress, severity and behavior in conjunction with meteorological and environmental factors has strengthened our understanding of recent Australian fires such as the 2017 Dunally fires [48] and the 2009 Black Saturday fires [49]. Such detailed research on the Riveaux Road fire would be highly valuable.

There may be some error in the detection of understorey fires due to reliance on satellite imagery. Undergrowth fires were identified by death of understorey species. This may have been missed when obscured by a dense canopy, or in plantations where understorey vegetation is sparse. The distinctions between different forest types are arbitrary, but consistent and are based on understood processes in aging eucalypt forests.

5. Conclusions

Our results show a clear relationship between disturbance history and fire incidence and severity under moderate fire weather conditions. Fire severity is related to intensity, which is important for ease of suppression [29,30]. Thus, the retention of older forests across the landscape may decrease fire risk. Allowing eucalypt forests to mature further than the normal cycles of 40 to 90 years could help reduce fire hazards. This is especially pertinent around built assets and vegetation types vulnerable to fire. Our results are consistent with the ‘landscape trap’ theory [21,44].

In addition to providing a valuable damper on fire incidence and severity, old growth Eucalyptus regnans wet forests store the highest density of carbon of any forests in the world [50]. Carbon continues to be captured in the aging old growth forests, rather than reaching an asymptote [51]. Soil carbon in wet eucalypt forest is a substantial store, its magnitude related to the above ground biomass of the forest [52]. The slow release of carbon from the soil as a result of previous conversion from primary forest to production forest, can only be prevented by allowing forest regrowth to continue to old age [52].

To refine our understanding of the relationship between disturbance history and fire we need to ask: how localised is the effect of patches of more flammable regrowth in a landscape—does a patchwork of forest types make a whole region more susceptible to fire; at what point after clearfell logging does a forest become less flammable; and, is there a way to manage the forest up to this point to reduce fire hazards while maintaining natural values?