Change in fire frequency drives a shift in species composition in native Eucalyptus regnans forests: Implications for overstorey forest structure and transpiration

The world's most iconic forests are under threat from climate change. Climate‐fire‐vegetation feedback mechanisms are altering the usual successional trajectories of forests. Many obligate seeder forests across the globe are experiencing regeneration failures and subsequent alterations to their recovery trajectories. For example, the persistence of Eucalyptus regnans F. Muell. forests in southeast Australia is highly vulnerable to the effects of climate‐driven increases in wildfire frequency. Shortening of the wildfire return interval from >100 years to < 20 years would inhibit or entirely stop regeneration of E. regnans, leading to replacement with understorey species such as Acacia dealbata Link. In this study, it is hypothesised that following such replacement, forest overstorey structure and transpiration will diverge. An experiment was designed to test this hypothesis by measuring and comparing overstorey transpiration and structural properties, including sapwood area and leaf area, between E. regnans and A. dealbata over a chronosequence (10‐, 20‐, 35‐ and 75‐/80‐year‐old forests). We found that overstorey structure significantly diverged between the two forest types throughout the life cycle of A. dealbata after age 20. The study revealed strikingly different temporal patterns of water use, indicating a highly significant eco‐hydrologic change as a result of this species replacement. Overall, the results provide a strong indication that after age 20, overstorey transpiration in Acacia‐dominated forests is substantially lower than in the E. regnans forests they replace. This difference may lead to divergence in water yield from forested catchments where this species replacement is widespread.


| INTRODUCTION
Changes in forest structure resulting from changing forest age and dominant species may alter evapotranspiration. This would have two significant outcomes: fundamentally changing the ecohydrologic functioning of the affected landscape and the alteration of streamflow and other ecosystem services. While forest harvesting, defoliation from pest attack and drought mortality can impact on forest species mix and structure, fire is the most significant disturbance agent in temperate forest ecosystems.
Recent studies in Australia (Brookhouse et al., 2013;Feikema et al., 2013;Gharun et al., 2013;Lane et al., 2006Lane et al., , 2010Nolan et al., 2014Nolan et al., , 2015 have focussed on the hydrologic impact of regeneration of eucalypt forests after single fire events, where regeneration follows predictable recovery trajectories. In fire-sensitive ecosystems, however, the occurrence of high-intensity wildfires at short return intervals has the potential to replace the dominant overstorey species with co-occurring understorey species (Adams, 2013;Brown & Johnstone, 2012;Buma et al., 2013;Fairman et al., 2016). For example, climate-driven increases in wildfire frequency are resulting in replacement of previously dominant serotinous conifer forests such as black spruce forests in the Alaskan region (Johnstone et al., 2010), lodgepole pine in the Greater Yellowstone forests (Turner et al., 2019;Westerling et al., 2011), ponderosa pine in the southwestern United States (Savage & Mast, 2005) with broadleaf species and boreal forests in North America (Coop et al., 2020).
Similar climate-driven increases in fire frequency are resulting in species replacement in Australian native serotinous forests. Since 2002, an increase in the fire return interval from the historic >100 years to < 20 years has severely damaged about 1890 km 2 of obligate seeder, Eucalyptus forests across Victoria (Keenan & Nitschke, 2016). This includes important water supply catchments for Australia's second-largest city (Melbourne) and Australia's "food bowl," the Murray-Darling Basin. Increases in wildfire frequency and severity in the past two decades in southeast Australia are largely consistent with climate change projections for this region, suggesting changes to fire regimes are linked with climate change (Clarke et al., 2012;Oldenborgh et al., 2020). High yielding water supply catchments in southeast Australia are dominated by fire-sensitive Eucalyptus regnans F. Muell. and Eucalyptus delegatensis R.T. Baker forests. A single wildfire in these forests results in partial or complete stand replacement (Ashton, 1976a), causing long-term increases in evapotranspiration (Vertessy et al., 2001) and reductions in streamflow (Kuczera, 1987), but a second wildfire occurring before reproductive maturity at age 20 years (Ashton, 1976a) can result in local extirpation of these two obligate-seeder species (Bowman et al., 2014(Bowman et al., , 2016, and replacement by common understorey species, often Acacia dealbata Link (Fairman et al., 2016;Keeley et al., 1999).
This occurred over significant areas following large wildfires in 1926 and 1939.
A change from E. regnans to A. dealbata represents a dramatic shift in vegetation, perhaps one of the most dramatic globally outside of human-induced deforestation. E. regnans, the world's tallest Angiosperm, typically grows up to 70-100 m, which is second only to Sequoia sempervirens (California redwood). In contrast, A. dealbata is a short-lived species, growing to a maximum height of only 30 m (May, 1999). Replacement of the iconic E. regnans forests with A. dealbata forests opens a new path of successional trajectories in these forested catchments ( Figure 1). The ecological implications are very significant, as are the impacts on carbon sequestration, landscape flammability and management options. Hydrologically, we may see marked divergence in forest structure and overstorey transpiration due to differences in sapwood area, leaf area and canopy conductance. The hydrologic signature of these "states", while important for water resource planning, is crucial to our understanding of landscape responses to increased stressors that are a function of climate change.
E. regnans is a long-lived species (up to 500 years) with a very high growth rate and biomass continuing to increase up to at least 250 years of age (Volkova et al., 2018), whereas A. dealbata is a relatively short-lived (< 100 years) pioneer species, with biomass peaking F I G U R E 1 Changing successional pathways in E. regnans forests as a result of high-frequency fire 20-30 years after regeneration (Trouve et al., 2019). The differences in longevity and growth rates between these two forest types may cause differences in their overstorey transpiration rates throughout their life span. Forest overstorey transpiration is the product of sap velocity and sapwood area of species that are dominant in the overstorey (Dunn & Connor, 1993;Haydon et al., 1996;Kostner, 2001;Vertessy et al., 2001), and thus, stand-scale overstorey transpiration is determined by both mean sap velocity and sapwood area.
Forest sapwood area varies widely with species, age and environmental conditions (Cermak & Nadezhdina, 1998). This may occur due to variation in properties of the hydro-active sapwood among species (Wang et al., 1992). Different species can also show significantly different rates of sap velocity if they possess substantially different water-use strategies (Liu et al., 2017). Even coexisting species sharing similar environmental conditions may transpire differently because of differences in species-specific sensitivities to climate and soil water relations (Link et al., 2014). Sap velocity is driven by short-term variations in meteorological conditions, whereas the sapwood area is determined by stand structural attributes, such as tree diameter, sapwood thickness and stocking density, that change relatively slowly over time, suggesting that climate and forest structure both influence stand transpiration (Cermak & Nadezhdina, 1998;Ewers et al., 2002;Hernandez-Santana et al., 2015;Medhurst & Beadle, 2002). When a shift in overstorey species occurs, divergence in overstorey transpiration is possible because of changes either in climatic and physiological control of sap velocity or structural control of forest sapwood area.
Overstorey dominates the total evapotranspiration of most temperate forests. Wet E. regnans forests conform to this, except in their late senescing stage (Vertessy et al., 1998(Vertessy et al., , 2001. Consequently, measurement of overstorey dynamics can illuminate knowledge gaps in the structural, functional, eco-physiological and eco-hydrological effects of replacement of E. regnans forests with A. dealbata. Except in young forests, changes in sap velocity were studied well in mature and old-growth E. regnans forests. Previous studies based on mature and old-growth stages of E. regnans showed that sap velocity does not change significantly with the forest age (Benyon et al., 2017;Dunn & Connor, 1993). While much is known about the ecohydrology of E. regnans forests (Benyon et al., , 2017Kuczera, 1987;Vertessy et al., 1998Vertessy et al., , 2001, only one study has examined co-dominant stands of E. regnans and A. dealbata. Hawthorne et al. (2018) showed that mean sap velocity of 20-year-old E. regnans during summer (December to March) was 12.8 cm h À1 compared with 7.5 cm h À1 , in A. dealbata of the same age, indicating overstorey sap velocity of E. regnans was 71% higher than A. dealbata even under the same environmental conditions with similar LAI. This work suggested that there was a marked change in fluxes between the species at this age. However, this study was not designed to enable detailed comparison of water use between these two forest types over their life cycle.
On reflection, the knowledge gap for forest hydrological aspects for A. dealbata requires further illumination, as no studies examine sap velocity, sapwood area or transpiration throughout the life cycle of A. dealbata forests. Given predictions that more frequent burns in E. regnans forests will result in their replacement by A. dealbata, it is timely to compare ecohydrological parameters between the two forest types throughout their life cycles to determine any significant changes in forest structure and overstorey transpiration.
In this study, it is hypothesised that if E. regnans is replaced with A. dealbata due to climate-induced increases in wildfire frequency, forest overstorey structure and transpiration will be altered. The main objectives of this study are, therefore, to (i) determine the divergence in overstorey transpiration between E. regnans and A. dealbata forests over an age sequence and (ii) identify the major drivers of this divergence process. These objectives were achieved by measuring overstorey sap velocity, sapwood area, and forest structural variables at the plot-scale in 10-, 20-, 35-and 75-/80-year-old stands of E. regnans and A. dealbata. Structural control and climate control of transpiration were evaluated for understanding the major drivers of forest transpiration after this climate-induced species replacement. It should be noted that although E. regnans water use and streamflow has been measured in a number of studies (e.g., Benyon et al., 2015Benyon et al., , 2017Dunn & Connor, 1993;Kuczera, 1987;Vertessy et al., 1998Vertessy et al., , 2001, stocking densities and structural attributes vary considerably for a given age class. It may be argued that to some extent the work of Kuczera (1987), which produced a curve of streamflow as a function of age, has led to a conceptually attractive understanding of E. regnans hydrology, which may be far more variable in reality. Therefore, further measurements of evapotranspiration in these forests are very useful. The significance is that water supply planning, and forest management policy and operations are heavily influenced by the age-streamflow dynamics.

| MATERIALS AND METHODS
Our primary research hypothesis, that if E. regnans is replaced with A. dealbata, forest overstorey structure and transpiration will diverge, was examined by testing five sub-hypotheses as shown in Table 1.
While Dunn and Connor (1993) and Benyon et al. (2017) worked in mature and old-growth forests, we undertook measurements in E. regnans stands for two reasons: to ensure we were measuring the two species under the same environmental conditions and to fill in data gaps for young stands. Given the expense and time commitment of sapflux work across multiple stand ages, we restricted our sapflow measurements to one to two plots per species per age to test subhypotheses 1, 3, and 4. As stand structural attributes that determine sapwood area are much more variable both spatially and with forest age (Benyon et al., 2017;Haydon et al., 1996;Vertessy et al., 2001), a larger number of inventory plots were used (Table S1) to test subhypotheses 2 and 5.

| Site description, sap velocity plots
Sap velocity measurements were taken from 10-, 20-, 35-and 75-/80-year-old E. regnans and A. dealbata forests ( Figure 2). During site selection, preference was given to forests which were regenerated purely with A. dealbata or E. regnans after stand-replacing fires. Both E. regnans and A. dealbata are evergreen forests. E. regnans is broad-leaved, whereas A. dealbata is a compound leaved species.
After high severity fires, these two forests regenerate from seeds as even-aged forests. To accurately account for differences in speciesspecific sap velocity responses to climate, adjacent forests of the two species were selected from each age class where possible. The selected sites were located in areas typical of the wetter forests of southeast Australia.   for each species was determined for the stand using nine larger inventory plots. As with the other sites used in the present study, the climate is temperate, with monthly mean maximum temperatures of 23 C in January and 9 C in July and mean annual rainfall of $1800 mm. The soil is deep, well-drained loam with high water holding capacity (Hawthorne et al., 2018).
Two years of sap velocity data for two E. regnans plots aged $75 years located in Maroondah catchment $80 km northeast of Melbourne were obtained from Benyon et al. (2017). The climate and soils are similar to the other sites, with mean annual rainfall of $1800 m.

| Forest overstorey transpiration
Forest overstorey transpiration is the product of sap velocity and sapwood area of species that are dominated in the overstorey (Kostner, 2001) and thus, stand scale overstorey transpiration is strongly linked with both sap velocity and sapwood area of the forest (Equation 1).
where Overstorey T = Transpiration from the forest overstorey, Mean SV_os = Mean sap velocity of a representative sample of trees in the overstorey, SA_os = Stand sapwood area (total sapwood crosssectional area of the forest overstorey per unit of ground area).
Sub-hypothesis 3, which stand mean transpiration varies significantly with species and age, was evaluated by comparing stand mean overstorey transpiration across species-age combinations. To do this, SA_OS in Equation (1) was estimated for each species/age as the mean of all inventory plots in which SA was determined (plots in Table S1), whereas SV_OS was the mean of the one or two sap velocity measurement plots in each species/age class (plots in Table 2).

| Selection of sample trees for sap velocity measurements
Sap velocity was measured concurrently in between five and eight sample trees per plot in each species/age class combination (Table 2).
Sample trees were randomly selected from circular plots that were

| Measurement and estimation of sap velocity
This study used the heat ratio method (ICT International, Armidale, NSW) to measure sap velocity in all species/age classes, consistent with the studies of Hawthorne et al. (2018) and Benyon et al. (2017).
Each HRM (heat ratio method) sap flow meter (SFM) includes a heater probe located between two sensor probes positioned 5 mm above and below the heater. Each sensor probe contains sensors located 7.5 and 22.5 mm from the probe tip. Sap velocity near to the heartwood was recorded by the inner sensor and sap velocity near the cambium was recorded by the outer sensor. The middle needle generated a 25-J heat pulse every 30 min. Following each heat pulse, the ratio of temperature increase in the downstream (upper) sensor compared with the upstream (lower) sensor was measured by the logger. Heat pulse velocity (Vh) was then estimated by the logger using this ratio, as shown in Equation (2) (Marshall, 1958).
where k is the thermal diffusivity (cm 2 s À1 ), x is the distance (cm) between the heater and the two temperature sensors, and v 1 and v 2 are temperature increases from initial temperature ( C) in the downstream and upstream sensors, respectively.
Thermal diffusivity (k) (Marshall, 1958), thermal conductivity (Kw) of dry wood (Swanson & Whitfield, 1981), moisture content (mc), green density ρ ð Þ, basic density (ρb) and wound correction factors (Burgess et al., 2001) were estimated to correct the sap flow velocity in each sample tree. Sap velocity in the xylem tissues can be determined using corrected heat pulse velocity and other estimated and available factors in the literature (Equation 3).
where SV is the corrected sap flow velocity, Vc is the heat pulse velocity corrected for the wound effect, ρb is the basic density of wood (kg m À3 ), Cw is the specific heat capacity of the wood (1200 J kg À1 C À1 at 20 C; Becker & Edwards, 1999;Bleby et al., 2004), Cs is the specific heat capacity of sap (4182 J kg À1 C À1 at 20 C; Bleby et al., 2004), mc is moisture content of wood and ρs is the density of sap.
When the downstream and upstream probe holes are drilled exactly the same distance above and below the heater, Equations (2) and (3)  gov.au/silo) for the last 116 years  to identify anomalies of rainfall during the data collection period and also to verify the proper functioning of weather stations at the study sites.  (Welles & Norman, 1991). The exposure of the camera was calibrated at each LAI measurement plot as described by Macfarlane et al. (2007). All hemispherical images were analysed using the Hemi view 2.1 (Delta-T devices, Ltd., Cambridge, UK) image processing software package.

| Statistical analysis
Sub-hypothesis 1 that stand mean sap velocity varies significantly with species and age was tested using two-sample t tests (Dunn & Connor, 1993) and Mann-Whitney test (Becker, 1996). Samples comprised of long-term average sap velocity of individual trees in each species-age class. The effects of species and age on sap velocity were tested using data from all sites at long term and daily time resolutions.
To avoid any bias due to the seasonal effects on sap velocity and transpiration, statistical analysis was confined to the 151 days of the year (October-February), when sap velocity data were available for all spe- We made an assumption that the extra data from the Hawthorne et al. (2018) and Benyon et al. (2017) for the same 151-day period of the year (but in different years) could be used in these analyses. While environmental conditions were clearly not identical, they were not out of the "ordinary" (i.e., no drought/heat waves or extreme wet periods).
As well as comparing plot mean sap velocities directly, to remove the effects of site and year to year differences in climatic conditions, tree and stand mean sap velocities were adjusted to a common set of climatic conditions. When sap velocity is not water limited, as was the case in this study, VPD is the main external driver of transpiration via its effect on sap velocity. We compared current and previously measured mean sap velocity data from all species and age class combinations after normalising mean sap velocities to account for differences in VPD. This was done by fitting a regression of daily mean sap velocity of each sample tree against daily mean VPD at each site. These fitted relationships were then used to estimate daily mean sap velocity for each tree using the daily VPD averaged across all sites: for each of the 151 days of the year common to all sites, the average VPD across all sites was calculated and then the tree-specific relationship between sap velocity and VPD was used to estimate the daily sap velocity for each tree. This enabled a direct comparison of plot mean sap velocity for the 151-day period as if all sites had experienced exactly the same sequence of daily VPDs.
Age-related structural divergence was tested using a two-way analysis of variance (ANOVA) and Tukey's post hoc analysis (for multiple pairwise comparisons). Then, the statistical significance of species and age effects on stand mean dbh, stand basal area and stand density was evaluated.
Differences in daily transpiration between the two forest types across the age classes were tested with a two-factor repeated measure ANOVA. Multiple comparisons across species and ages were undertaken with Tukey's HSD post hoc test (H 3 ).

| Examining structural control of transpiration
To evaluate which variable (sap velocity or stand sapwood area) is more important in determining stand transpiration in the two forest types, sub-hypothesis 4 was tested by comparing the strength of regression relationships between stand mean daily transpiration and stand sapwood area and between stand mean daily transpiration and stand mean sap velocity. Further, multiple linear regression was used to determine the statistical significance and relative importance of sap velocity (SV) and sapwood area (SA) in driving stand mean daily transpiration (T).
Sub-hypothesis 5 that stand sapwood area is controlled by stand structural properties was tested using regression relationships between stand sapwood area, and stand basal area in all water use plots. Further, the variations in stand density, stand basal area and stand sapwood thickness between species along the age sequence were evaluated. After adjustment of the plot mean sap velocity data to remove the effect of differences in VPD, the results were similar ( Although the species-specific difference in sap velocity was relatively small at longer-term time steps, species-specific differences in sap velocity were evident at daily time steps. Analysis of relationships between meteorological variables and daily mean sap velocity at each site indicated that VPD is the major climatic driver of sap velocity in both forest types (Figure 3). Differences in quantified relationships between mean daily sap velocities and mean daily VPD suggest that species and age contributed to variability in sap velocity ( Figure 4). However, these two factors were not strong enough to cause distinctively different sap velocity responses between the two forest types.

| Species and age effects on sapwood area
Sub-hypothesis 2 was supported with a high degree of confidence.
Based on the larger sample of inventory plots (Table S1), A. dealbata exhibited a four-fold decline in overstorey sapwood area (SA) over time, from a mean of 16.3 m 2 ha À1 at age 10 to 3.8 m 2 ha À1 at age 80, whereas the SA of E. regnans declined by only half between age 10 and 75 (15.5 m 2 ha À1 at age 10 to 7.5 m 2 ha À1 at age 75; Figure 5). At age 10, the difference in mean SA between species was not statistically significant (P > 0.05), but it was at ages 20, 35, and 75/80 (P < 0.05).

| Comparison of overstorey transpiration between E. regnans and A. dealbata forests
Sub-hypothesis 3.0, which stand mean overstorey transpiration (T) varies significantly with species and age, was supported with a high degree of confidence. Over the period of the year when sap velocity data were available from all water use plots (October-February), mean daily T differed significantly among species and age classes after age 10 (P < 0.001). Further, there was a four-fold variation in daily T across the eight species/age combinations (Figure 6).
At age 10, overstorey T was slightly higher in A. dealbata (4.0 mm day À1 and 3.6 mm day À1 for A. dealbata and E. regnans, respectively), although the difference was not statistically significant.
However, overstorey T of the two forest types began to diverge from 20 years onwards (lower in Acacia, Figure 6). This divergence continued to increase up to age 75/80 (1.0 mm day À1 in A. dealbata vs. 2.2 mm day À1 in E. regnans) and was statistically significant (P < 0.005) in all age classes from 20 years onwards. The results suggest that SA area exerts a strong control on stand T and reduces with stand age, but at a faster rate in A. dealbata.

| Drivers of differences in overstorey transpiration between E. regnans and A. dealbata forests
Sub-hypothesis 4.0, which stand mean T is controlled more by stand SA than by mean SV, was supported with a high degree of confidence.
Large differences in mean overstorey T between species and age classes evident in Figure 6 were mainly correlated with the overstorey SA. The relationship between stand mean T and SA was far stronger than between T and SV: variation in stand sapwood area explained 95% of the variation in stand mean T, whereas the T versus SV relationship was weak (R 2 = 0.04). Further, multiple linear regression analysis suggests that SA was the only statistically significant determinant of stand mean T (P < 0.05). SA alone explained 95% of the variability in T between plots, while sap velocity explained additional variability of only 0.002%, confirming that stand SA is the main driver of stand mean T.
3.5 | The link between the divergence in forest structure and overstorey sapwood area in the two forest types A strong relationship between tree SA and tree basal area strongly supports acceptance of sub-hypothesis 5.0: stand structure controls stand sapwood area (Figure 7) in the two forest types.
SA in E. regnans forests diverges dramatically from that of A. dealbata forests over time because E. regnans forests have significantly higher stand basal area from age 20 onwards (Figure 8). Average stand density in A. dealbata forests declined from 3940 trees ha À1 at age 10 to 308 trees ha À1 by age 80. In E. regnans, the rate of selfthinning was proportionally greater: stocking declined from only slightly (4%) less than A. dealbata at 10 years (3763 trees ha À1 ) to 57% less than A. dealbata (131 trees ha À1 ) by age 75. However, the dominant A. dealbata overstorey begins to die off rapidly at around 75 to 80 years of age, suggesting A. dealbata stands significantly reduce stand density after age 75 compared with E. regnans forest.
This stand density comparison suggests that A. dealbata should have a higher stand sapwood area than E. regnans because up to age 80 the acacias always have higher stand density. However, in fact, after age 10, the opposite is true: E. regnans has a higher stand sapwood area after age 10. This is mainly because a significantly higher stand basal area exerts a stronger influence in the opposite direction to reduce the rate of reduction in stand sapwood area in E. regnans compared with in A. dealbata after age 10 ( Figure 8).
Stand mean dbh and basal area both diverge markedly between species but in the opposite direction to stocking density (Figure 8).
F I G U R E 5 Variation of overstorey sapwood area between the two forest types with age (vertical bars represent the upper and lower 95% confidence intervals of stand sapwood area for each species-age class) F I G U R E 6 Comparison of overstorey transpiration for the October to February period between A. dealbata (AD) and E. regnans (ER) forests with stand age (vertical bars show upper and lower 95% confidence intervals of daily transpiration for each species-age group). Transpiration for 20-year-old forests was estimated using data from Hawthorne et al. (2018) and for the 75-year old E. regnans from Benyon et al. (2017) From age 20 onwards there is a statistically very large difference in both mean dbh and stand basal area between the two species. This divergence increases up to age 75/80, by when median dbh and stand basal area are both almost three times larger in E. regnans Results suggest that stand dbh and stand basal area start to diverge from age 20 onwards (lower in A. dealbata forests, Figure 8).  Figure 10). Transpiration per LAI was significantly higher in E. regnans after age 20 (Table 4). This was due to higher sapwood area per LAI in E. regnans for a given age (Table 4). Therefore, differences in LAI between the two species alone cannot explain the proportionally much larger differences in stand mean transpiration evident in the present study.  Vertessy et al. (1995Vertessy et al. ( , 2001, Benyon et al. (2017) and Dunn and Connor (1993) are largely in concert with our findings.

| DISCUSSION
However, a limitation of these previous studies was their focus almost exclusively on the hydrological behaviour of mature and old-growth forests. Prior to Hawthorne et al. (2018), the only measurement of transpiration in E. regnans < 50 years old was in a single 15-year-old stand measured for only a short period from September to November by Vertessy et al. (1995). Their reported mean sap velocity of 11.9 cm h À1 compares to our mean at age 10 of 9.7 cm h À1 , albeit for a different period of the year. When adjusted to a common set of mean daily VPD data, our mean SV for 10 year old E. regnans was similar to that of A. dealbata (11.8 vs. 11.6 cm h À1 ) and to the mean SV of 75-year-old E. regnans (11.9 cm h À1 ). Our measurements are more representative of young stands of E. regnans as we included forests in different ecological conditions and for longer measurement periods.
The transpiration rate per unit of leaf area was highest in young stands of both species (Table 4) due to very high sapwood area in young stands ( Figure 5). The transpiration estimated from this study (including the Benyon et al., 2017, andHawthorne et al., 2018, data) is largely a function of sapwood area.
To our knowledge, this is the first study to investigate hydrological parameters in pure A. dealbata forests over a chronosequence.
Since the long-term average daily mean sap velocity in A. dealbata is relatively constant with forest age (Table 3), annual overstorey transpiration of A. dealbata forests can be predicted solely as a function of stand sapwood area. SA to LAI ratio (or Huber value) decreased with stand age in A. dealbata forests which is consistent with previous findings in a range of species (Kostner et al., 2002;Mokany et al., 2003;Watson et al., 1999). In general, age-related reductions in overstorey transpiration and overstorey sapwood area in A. dealbata forests follow a similar trend to that of E. regnans forests but compressed into a much shorter time scale.
The data show the forest structure and overstory hydrologic functioning of E. regnans and A. dealbata forests are closely matched during the first 10 years of their life but then begin to diverge. Overstorey transpiration and overstorey sapwood area in A. dealbata forests peak around age 10-20 and then rapidly decline over the next 60 years. However, both of those variables decline only gradually between age 10 and 75 in E. regnans forests, resulting in a significant contrast in forest structure and water use between the two forest types after age 20. This is possibly because A. dealbata forests have a regeneration-oriented life cycle, so that A. dealbata forests put more effort into reproduction and accumulation of a large seed store in the soil to ensure persistence over the long run (May, 1999

| Shift in species composition due to climate change impacts
Regeneration success is one of the factors that determine species distribution at the landscape and regional scale (Mok et al., 2012). Climate warming is likely to affect seed production (Redmond et al., 2012), seedling recruitment (Boucher et al., 2020;Johnstone et al., 2010), abundance and distribution of species (Mok et al., 2012) and tree survival  and in many fire-prone forest systems the stress of increased fire frequency may produce tipping points.
These compound disturbances of drought and fire have the potential to drive widespread disruption to eco-hydrologic functioning (Mirus et al., 2017), resulting in a shift in eco-hydrologic regime (Blount et al., 2020). There are growing numbers of studies focused on ecological aspects of species replacements such as changes in climate-fire-vegetation interactions, succession pathways, carbon stocks and ecological resilience in response to perturbations. Quantification of change in evaporative fluxes as a consequence of vegetation change is still very limited, despite the importance of understanding how these systems respond to repeated disturbance, and how vegetation-water feedbacks may be affected. Pfautsch et al. (2010) reported that E. regnans forests containing mid and understory strata had higher water use in comparison to single storey stands. This indicates that water use tends to be changed with the compositional changes in these forests. During the present study, we found distinct water use signals from pure patches of E. regnans and A. dealbata.
The finding in this study that VPD is the principal climatic driver of transpiration is germane (though not surprising) not only to the current transpirative regime, but for future climate where higher average VPD is predicted for southeast Australia (DELWP, 2019) and many other environments (e.g., SW USA). We note that rainfall via soil moisture availability was for the most part not a limiting factor in our study. Clearly, high soil moisture deficits will also limit transpiration.
The combination of high atmospheric demand and low water supply is most likely to exacerbate water stress. This link between rising temperatures, decreasing rainfall and lower humidity is of course the climatic combination that drives wildfire, with four "mega fire" events in southeast Australia since 2003, including the unprecedented "Black Summer" fires of 2019-2020 that burnt almost 5 million ha over four states. It is this apparent change in fire regime that results in the species change studied in this paper.
Extensive wildfires could change watershed outputs across the globe (Robinne et al., 2018). In the context of climate-induced vegetation change in southeast Australia, the present study provides a strong indication of lower overstorey transpiration in Acacia dominated forests. At the senescing stage of A. dealbata, understorey may off-set the loss of overstorey transpiration and eventually shift the evapotranspiration regime towards shrubby forests. These alterations may lead to divergence in catchment water yield. This study clearly demonstrates a link between differences in overstorey transpiration and differences in forest structure between the two forest types over the chronosequence, suggesting that time since forest replacement can be used as a proxy for detecting water use change and streamflow change over time in response to climate change and multiple burns.

| CONCLUSIONS
Increases in wildfire frequency due to climate change may induce changes in dominant species. Our research has shown that this could have important long-term consequences for water supplies. Replacement of E. regnans with A. dealbata in water supply catchments in southeast Australia would result in a reduction in overstorey transpiration after about 20 years, due mainly to divergence in sapwood area. In both forest types, sapwood area reduced with stand age after age 20, but at a faster rate in A. dealbata. In E. regnans forests, reduction in stand sapwood area from age 10 to age 75 was relatively low compared with that of A. dealbata. Therefore, overstorey transpiration in A. dealbata declined markedly between age 10 and age 80, whereas overstorey transpiration in E. regnans declined much less between the same ages.
This study also reveals that mean sap velocity was always higher in E. regnans compared with A. dealbata from age 20 onwards, but only statistically significantly so at age 20. Species has a significant effect on sap velocity at three different time resolutions, namely, long-term (seasonal), monthly and daily time resolutions. Although the species-specific difference in sap velocity was relatively small at longer-term time steps such as seasonal and mean annual scale, species-specific differences in sap velocity were evident at daily and sometimes monthly time steps. VPD is the main driver of sap velocity in both A. dealbata and E. regnans forests. Increases in mean VPD in future may increase mean transpiration rates, off-setting the effects of species change from E. regnans to A. dealbata.

DATA AVAILABILITY STATEMENT
Data that support the findings of this study are available from the corresponding author upon reasonable request.