Aerodynamic effects cause higher forest evapotranspiration and water yield reductions after wildfires in tall forests

Wildfires are increasing in frequency, intensity, and extent globally due to climate change and they can alter forest composition, structure, and function. The destruction and subsequent regrowth of young vegetation can modify the ecosystem evapotranspiration and downstream water availability. However, the response of forest recovery on hydrology is not well known with even the sign of evapotranspiration and water yield changes following forest fires being uncertain across the globe. Here, we quantify the effects of forest regrowth after catastrophic wildfires on evapotranspiration and runoff in the world's tallest angiosperm forest (Eucalyptus regnans) in Australia. We combine eddy covariance measurements including pre‐ and post‐fire periods, mechanistic ecohydrological modeling and then extend the analysis spatially to multiple fires in eucalypt‐dominated forests in south‐eastern Australia by utilizing remote sensing. We find a fast recovery of evapotranspiration which reaches and exceeds pre‐fire values within 2 years after the bushfire, a result confirmed by eddy covariance data, remote sensing, and modeling. Such a fast evapotranspiration recovery is likely generalizable to tall eucalypt forests in south‐eastern Australia as shown by remote sensing. Once climate variability is discounted, ecohydrological modeling shows evapotranspiration rates from the recovering forest which reach peak values of +20% evapotranspiration 3 years post‐fire. As a result, modeled runoff decreases substantially. Contrary to previous research, we find that the increase in modeled evapotranspiration is largely caused by the aerodynamic effects of a much shorter forest height leading to higher surface temperature, higher humidity gradients and therefore increased transpiration. However, increases in evapotranspiration as well as decreases in runoff caused by the young forest are constrained by energy and water limitations. Our result of an increase in evapotranspiration due to aerodynamic warming in a shorter forest after wildfires could occur in many parts of the world experiencing forest disturbances.


| INTRODUC TI ON
Forests provide vital ecosystem services (Costanza et al., 1997) and regulate the exchange of water, carbon, and energy between the land surface and the atmosphere.Globally, these areas are currently under threat from large forest disturbances, such as wildfires (Artés et al., 2019;Lindenmayer & Taylor, 2020;Liu et al., 2019) that are predicted to further increase in frequency, intensity, and extent with climate change in many parts of the world (Abatzoglou & Williams, 2016;Canadell et al., 2021;Gutierrez et al., 2021;Jolly et al., 2015;McColl-Gausden et al., 2022).Such a change in frequency of forest disturbances globally was shown to cause alterations in forest structure and age leading to shorter and younger ecosystems (McDowell et al., 2020).
In recent years, fires have destroyed large parts of forests in catchments crucial for the provision of freshwater resources, such as in California (e.g., Gutierrez et al., 2021;Maina & Siirila-Woodburn, 2020) and southern and eastern Australia (e.g., Canadell et al., 2021;Feikema et al., 2013;Lindenmayer & Taylor, 2020).Some of these forests are highly susceptible to frequent fires and if fire occurs more frequently due to global climate change than the age of maturity of the trees, there is an irreversible change in forest type, structure, and function (Bassett et al., 2015;Bowd et al., 2018;Bowman et al., 2016).The destruction of mature forests by fire and the subsequent regrowth of young forests can alter the exchange of energy, water, and carbon at the land surface (Liu et al., 2019;Seidl et al., 2014) and potentially lead to changes in evapotranspiration (ET) and thus discharge and water yield from these forested catchments (Bond-Lamberty et al., 2009;Kuczera, 1987;Li & Lawrence, 2017).Knowledge of these potential changes in ET following wildfires and their impact on runoff (Q) is crucial for future freshwater resources management, especially in water scarce regions of the world.
Many previous studies have analyzed the short-and long-term effects of wildfires associated with subsequent forest recovery on downstream water availability with direct measurements of ET and streamflow (e.g., Kuczera, 1987;Niemeyer et al., 2020), with remotely sensed vegetation indices and their derived ET products (e.g., Liu et al., 2019;Ma et al., 2020), and using hydrological and ecohydrological models (e.g., Bond-Lamberty et al., 2009;Li & Lawrence, 2017).Global scale modeling and remote sensing studies find an average decrease in ET and increase in Q caused by wildfires (Li et al., 2017;Li & Lawrence, 2017;Liu et al., 2019).
A large number of case studies using a range of measurements and modeling also report lower ET and higher discharge following wildfires in a broad range of climates and geographical locations such as in California (Boisramé et al., 2019;Ma et al., 2020;Maina & Siirila-Woodburn, 2020), the interior Pacific Northwest, USA (Niemeyer et al., 2020), in Montana, USA (Blount et al., 2020), and across the North American boreal forest (Bond-Lamberty et al., 2009).At the same time, several studies analyzing the effects of wildfires in south-eastern Australia report an increase in ET attributed to the vigorous regrowth of ash-type eucalypt forest post-fire, which can potentially impair freshwater resources availability (Feikema et al., 2013;Haydon et al., 1997;Jaskierniak et al., 2016;Kuczera, 1987;Vertessy et al., 2001).Additionally, some studies in mid-high latitudes also indicate the potential for an ET increase after forest disturbance (Biederman et al., 2014;Kettridge et al., 2019).
Past studies reporting a decrease in ET have mostly explained their results by a medium-to long-term reduction in vegetation canopy cover or a change in the type of vegetation regrowing after the fire (e.g., Blount et al., 2020;Boisramé et al., 2019;Li & Lawrence, 2017;Ma et al., 2020;Niemeyer et al., 2020).While studies reporting an increase in ET post-fire explained their findings by a denser vegetation canopy and higher transpiration (Haydon et al., 1997;Vertessy et al., 2001) or abiotic factors leading to higher evaporation (Biederman et al., 2014;Kettridge et al., 2019).However, a modeling study in Europe indicated that also the interplay between aerodynamic and biotic factors determines the sign of ET changes after afforestation (Breil et al., 2021).
With wildfire frequency, extent and intensity changing at the global level and even the sign of hydrological changes post-fire being uncertain, it is crucial to further quantify and understand the mechanisms leading to the observed short-and long-term ET and Q changes post-fire and their effect on water resources availability.
Hence, in this study, we shed light on the effects of forest recovery on the hydrological fluxes after wildfires in tall eucalypt-dominated forests in south-eastern Australia.To do so, we combined a unique data set of eddy covariance measurements, remote sensing products, and ecohydrological modeling pre-and post-fire.The case study entails the world's tallest (up to 90 m) angiosperm-dominated forests (Eucalyptus regnans) in south-eastern Australia (Kilinc et al., 2013;Van Pelt et al., 2004) before its destruction by a stand-replacing crown fire in February 2009 with catastrophic conditions within the region producing record-breaking maximums of fire danger and air temperature, impacting also Melbourne, which reached 46.4°C in air temperature (Teague et al., 2010).We further expanded the analysis to several larger fires using remote sensing products to test if the findings are generalizable to eucalypt forests in south-eastern Australia.We also performed numerical experiments with an ecohydrological model (Tethys-Chloris, Fatichi et al., 2012aFatichi et al., , 2012b) )  and Q response after forest fires?Answering these questions is expected to contribute to forest management practices after wildfires and a better assessment of water resource availability in a world with globally increasing forest disturbances.

| Overview
In this study, we combined eddy covariance latent heat (LE) measurements (and the derived ET), remote sensing, and ecohydrological modeling to analyze and quantify the short-and long-term LE, ET, and Q response after forest fires in eucalypt-dominated forests in south-eastern Australia.
First, we assessed the short-and long-term recovery of the LE timeseries at the forest stand scale and its consistency over 18 years (2000-2017, including ca. 9 years pre-fire and post-fire) using eddy covariance measurements, remote sensing, and modeled with the ecohydrological model Tethys-Chloris (T&C; Fatichi et al., 2012aFatichi et al., , 2012b)).However, due to an extended drought period, resulting in below average rainfall in the decade before the wildfire in 2009 (Van Dijk et al., 2013) To minimize interannual climate variability effects, reference timeseries extracted from a buffer around each fire extent were subtracted from the analyzed variables.Subsequently, theoretical recovery curves were fitted to the LE, land surface temperature, and vegetation index timeseries of each fire.From these fitted curves, the time to recovery for each fire was calculated and the extrapolated long-term changes in LE or vegetation indices were predicted and compared among fires.

| Study site
Before the Black Saturday bush fires in February 2009, the Wallaby Creek eddy covariance tower site in Kinglake National Park, Victoria, Australia (Figure 1a) was home to the world's tallest angiospermdominated forest (Kilinc et al., 2013;Van Pelt et al., 2004).The park was predominantly managed for ecosystem conservation and was assigned the IUCN Category II (National Parks) of the United Nations' list of National Parks and protected areas.The flux tower site was located northeast of Melbourne on the southern edge of the Hume Plateau at an elevation of 738 (Beringer et al., 2016).The area was dominated by Mountain Ash (Eucalyptus regnans), which is the world's tallest flowering plant species (angiosperm).The oldgrowth Eucalyptus regnans forest had an age of around 300 years and an average canopy height of 76.6 m but some individual trees were taller than 90 m (Kilinc et al., 2013;Van Pelt et al., 2004).The understory comprised of a dense temperate rainforest with tree heights ranging from 10 to 18 m in combination with tree ferns (Kilinc et al., 2013;Van Pelt et al., 2004).However, a high-severity forest fire in February 2009, part of the Black Saturday bushfires, lead to high forest mortality and the destruction of this old, tall eucalypt forest, which triggered rapid forest regrowth from seedlings (Benyon & Lane, 2013).The site contained mostly krasnozemic soils with a depth of at least 200 cm and with high amounts of organic matter in the top 20-30 cm of the soil and increasing clay content with depth.The study area has a cool, temperate climate with the highest temperatures occurring in the summer months (average diurnal temperature range of 12-21°C measured at the study site during December to February, 2000February, -2017) ) and coolest temperatures in winter (average diurnal temperature range of 5-9°C measured at the study site during May to August, 2000August, -2017)).The long-term annual average rainfall is 1207 mm year −1 from 1885 to 2006 (Kilinc et al., 2013), however, the years preceding the Black Saturday bush fires experienced an extended drought in south-eastern Australia (Van Dijk et al., 2013) and the average rainfall measured close to the study site was only 925 mm year −1 from 2000 to 2009.

| Eddy covariance measurements
Eddy covariance measurements were analyzed, as one of the data sources in this study, to quantify the effects of forest regrowth on total ecosystem LE and ET as their changes affect the water budget and water resources availability.The gap-filled and expanded eddy covariance LE timeseries was analyzed at the monthly and annual scale and the average LE fluxes before (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) and after (2011)(2012)(2013)(2014)(2015)(2016)(2017) the fire (excluding the initial recovery period of 2 years following the fire) were statistically compared at the annual scale and in summer (average flux in December, January, and February).
Additionally, the eddy covariance measurements were used to evaluate the performance of the mechanistic ecohydrological model, Tethys-Chloris (T&C; Section 2.2), which was subsequently applied to explore the mechanisms of the observed LE, ET, and Q changes at the plot scale.
The eddy covariance site was located at Wallaby Creek in the Kinglake National Park, Victoria, Australia (37°25′44″ S, 145°11′14″ E, Figure 1a; Beringer et al., 2016;Kilinc et al., 2013).The site is part of the international FLUXNET network (ID is Au-WAC) and a member of the regional Australian Flux network (OzFlux; Beringer et al., 2016Beringer et al., , 2022)).The Wallaby Creek eddy covariance tower was located within the old-growth Mountain Ash (Eucalyptus regnans) forest (see Section 2.2).The initial tower, which had a height of 110 m, collected data from August 2005 to February 2009 when the original measurement set-up was destroyed by the Black Saturday bush fires.Prior to the fire, a separate understory flux system was installed in addition to the overstory eddy covariance system.The overstory measurement devices were located at a height of 95 m and 50% of the flux footprint was determined to be within ca.900 m upwind of the tower (Kilinc et al., 2013).The understory fluxes were measured by devices at 25 m height (Kilinc et al., 2013).Postfire, eddy covariance measurement devices were first installed at a height of 5 m and the height was gradually increased as vegetation grew keeping them approximately 5 m above the vegetation canopy height.In total, the data analyzed in this study were collected from August 2005 to February 2009 and from May 2010 to April 2013, resulting in ca 3.5 years of eddy covariance measurements before the fire and approximately 3 years after the fire.During the measurement period, LE data availability was 68%.However, the measured LE flux timeseries was gap filled and expanded to cover the continuous timespan from January 2000 to December 2017 using the OzFlux standard processing protocol (Beringer et al., 2017;Isaac et al., 2017).During the summer months from December 2005 to April 2006 overstory vegetation contributed to 61%, understory to 34%, and soil evaporation to 4%-6% of the total ET flux (Kilinc et al., 2013).These percentages were applied to the total ET (calculated from the measured LE) prior to the fire to compare measured to modeled ET sources in the T&C simulations.
In this study, the hourly latent heat (LE) and the corresponding ET were used.ET was calculated from the eddy covariance LE measurements as ET = LE/λ × dt, where λ (Jkg −1 ) is the latent heat of vaporization, dt (s) the time interval, and ET and LE have the units (kg m −2 or mm) and (Wm −2 ), respectively.Additionally, meteorological variables measured at the flux tower site were used as forcing data in the ecohydrological model, specifically air temperature, vapor pressure, and wind speed.More detailed information on the eddy covariance instrumentation set-up and data quality control is provided in the supplementary material.

| Remote sensing data and recovery curves
Timeseries of vegetation indices normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), leaf area index (LAI), LE, and daytime land surface temperature (LST day ) were extracted from the Moderate Resolution Imaging Spectroradiometer (MODIS) products (Table 1) to quantify vegetation recovery in the years following the fire.The seasonal and annual average MODIS LE timeseries (Running et al., 2017) extracted within 1 km radius from the Wallaby Creek flux tower were compared to the eddy covariance measurements and the ecohydrological simulations.The same variables were also extracted and averaged over the total Wallaby Creek fire extent (ca.895 km 2 ) and for several additional large fires occurring between 2006 and 2014 in the state of Victoria to test if the pattern of vegetation recovery was generalizable to fires in southeastern Australia.The fires were selected according to the criteria listed in the following.Fire extents, as shown in Figure 1a, were defined based on the GlobFire database, which is based on the MODIS burnt area Collection 6 product (MCD64A1; Artés et al., 2019).
All selected fires (seven in total) had an area greater than 100 km 2 and more than 50% of the fire area was covered by medium or tall eucalypt forest based on the classification provided by Forests of Australia (Australian Bureau of Agricultural and Resource Economics and Sciences, 2018).Note that out of the seven analyzed fires, four contain large areas only burned once between 1995 and 2017, while three contained within the larger fire extent some smaller areas burned several times in the same time period (Lindenmayer & Taylor, 2020).While the different MODIS products have different temporal resolution, the extracted quality-controlled data were averaged to the monthly and yearly time scale.Interannual climate variability complicates the quantification of the forest recovery response on the observed variables (NDVI, EVI, LAI, LE, LST day ) as described previously (Nolan et al., 2015).Hence, we follow a similar approach to Liu et al. (2019) who chose unaffected reference areas in close proximity to the burned forest to remove the confounding role of year to year climate variability in the quantification of forest recovery.Specifically, we extracted the land surface variables within a buffer radius of 10 km from each fire extent and subtracted the annual timeseries of spatially averaged values of the unburned buffer from the annual timeseries of spatially averaged values of the burned land.Since landcover composition can differ between burned area and the unburned buffer area, there is an offset between average land surface variables of burned area and its buffer even prior to the fire.In the analyzed timeseries, this offset was removed by subtracting the time averaged pre-fire offset between buffer and burned area leading to an average zero deviation between burned and buffered area in the period prior to the fire.
To quantify and compare the recovery process of the MODIS land surface variables among the selected fire extents, a recovery curve (Cassottana et al., 2019) was fitted to each of the MODIS timeseries (buffer values removed and normalized as described above to ensure pre-fire zero average).Fitting such a recovery curve allows to predict and compare the time to recovery and the longterm projected degree of recovery for the different fire extents in a systematic way.The time to recovery was calculated based on the fitted curves assuming the values have reached recovery once they exceed average pre-fire value (which is zero) minus one standard deviation of the annual pre-fire variability.Further information and the equations of the fitted recovery curves are provided in the supplementary material.

| Discharge measurements
In combination with the eddy covariance LE and the derived ET, discharge measurements at the annual scale were used as an additional source to evaluate the water flux partitioning simulated by the ecohydrological model T&C (Fatichi et al., 2012a(Fatichi et al., , 2012b, Section 2.2).
The discharge measurements used in this study were obtained from three measurement locations (Table 2) which had their upstream catchments at least partially affected by the 2009 Black Saturday fires.The location and time period of the discharge measurements are summarized in Table 2. Note, because of computational requirements, we only performed a forest stand scale simulation with T&C as we do not aim to reproduce the short-term rainfall-runoff response but the long-term water balance partitioning.Previous research did not find an instantaneous increase in Q in catchments after the 2009 Black Saturday fires and they hypothesized that this is due to TA B L E 1 Summary of the vegetation indices (NDVI, EVI, LAI), latent heat (LE), and daytime land surface temperature (LST day ) extracted from the Moderate Resolution Imaging Spectroradiometer (MODIS) products for seven fires in the state of Victoria, Australia.the replenishing of a substantial soil water deficit following the fire (Feikema et al., 2013).Additionally, interannual climate variability, an extended drought period before the fire (Van Dijk et al., 2013), and relatively short measurement time periods confound the quantification of wildfire effects on streamflow response from timeseries alone.For example, results of Nolan et al. (2015) showed that the climatic conditions after the fire could lead to large differences in the resulting reported streamflow decrease attributed to the fire.

| Ecohydrological model
The where T s (K) is the surface temperature, T a (K) is the air temperature at atmospheric reference height, q sat (T s ) (−) is the saturated specific humidity calculated as function of the surface temperature, and q a (−) is the specific humidity at the atmospheric reference height.
is the specific heat capacity of air, ρ a (kg m −3 ) the air density, and λ (J kg −1 ) the latent heat of vaporization.T&C computes the aerodynamic resistance according to a simplified solution of the Monin-Obukhov similarity theory (Mascart et al., 1995).LAI is a prognostic variable in the model and calculated based on net carbon assimilation, its allocation to different carbon pools controlled by environmental conditions, phenology, and allometric constraints, and vegetation tissue turn-over in response to aging, and cold and drought stresses.Hence, LAI calculated with T&C is independent of the LAI estimated from remote sensing products also analyzed in this study.Further information on T&C, its parameterization and the parameters applied in the current study are described in the supplementary material.The model ability to simulate energy, water, and carbon fluxes was evaluated in various ecosystems and climates comparing model results with eddy covariance measurements and other ecohydrological datasets in multiple previous studies (e.g., Botter et al., 2021;Fatichi et al., 2012aFatichi et al., , 2012b;;Hutley et al., 2022;Manoli et al., 2018;Mastrotheodoros et al., 2017;Pappas et al., 2016;Paschalis et al., 2022;Fatichi et al., 2021).
In this study, a plot-scale T&C simulation was performed for the Wallaby Creek eddy covariance measurement site (37°25′44" S, Finally, to analyze the effects of energy and water limitations on the forest recovery process after wildfire, further simulations were (1)  T&C also modeled an increased LE flux of 13% at the annual scale and 10% during the summer months post-fire.However, it has to be noted that post-fire annual and summer rainfall had also increased by 20% (p-value of 0.15, t-test) and 34% (p-value of 0.28, t-test), respectively, compared to the average pre-fire period due to an extended drought before the fire (Van Dijk et al., 2013).This increased rainfall can cause higher water availability for ET and LE post-fire independent of forest recovery effects.Indeed, in the undisturbed T&C simulations the postfire annual and summer LE is larger by 2% and 6%, respectively, compared to the pre-fire period indicating that some part of the observed post-fire LE and ET increase is likely caused by interannual climate variability and cannot be attributed to the fire.Furthermore, it has to be considered that a new eddy covariance measurement system was set up after the fire and that eddy covariance measurements in itself have uncertainties and often do not close the energy budget (Mauder et al., 2020).Energy budget closure for all seasons at the Wallaby Creek flux tower site was 0.81 without the inclusion of storage fluxes (Kilinc et al., 2012).Hence, we caution to conclude on a statistically significant increase in LE flux after the Wallaby Creek wildfire due to forest regrowth based on eddy covariance measurements alone and we refer to the further mechanistic analysis with T&C in which forest regrowth effects can be isolated through the differences between the fire-on and undisturbed simulations (Section 3.4).

| Eucalypt forest recovery after wildfires measured at larger spatial scale in Victoria, Australia
Analyzing

| Model performance evaluation
The effects of forest recovery on the water fluxes (ET, Q) were further quantified with the use of the mechanistic ecohydrological model T&C.To evaluate T&C's performance, we first conducted a comparison of the measured and modeled LE, ET, and Q fluxes (Figure 3).The comparison of the hourly measured and modeled LE flux including the fire event, showed good agreement between T&C and the eddy covariance measurements with a R 2 of 0.80 (Figure 3b).
At the seasonal scale, T&C tended to slightly underpredict and overpredict the LE flux during winter and summer months, respectively, compared to the eddy covariance measurements (Figure 3d).Prior to the fire, the ET flux simulated by T&C was partitioned into 7% from soil, 29% from understory, and 63% from overstory vegetation (i.e., the eucalypt trees), which was very similar to the measured partitioning of the ET flux pre-fire (Kilinc et al., 2013; Figure 3a).Postfire, only one canopy layer was simulated in T&C, which consists of growing eucalypt trees.A qualitative comparison at the seasonal and annual scale showed good agreement in the timing and magnitude of the LE flux from the recovering forest ecosystem in the years following the fire (Figure 1c).Comparison of the modeled and measured hydrological partitioning of ET and Q showed that the magnitude was well reproduced by the model (Figure 3c).T&C simulated an annual average ET flux accounting for 89% and 88% of the annual precipitation amount, and a Q flux of 12% and 14% of the annual precipitation amount during the pre-fire (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) and postfire (2011-2017) period, respectively.Note, there were also small changes in soil water storage during these time periods, which led to the sum of ET and Q percentages to slightly differ from 100%.The simulated hydrological partitioning agreed with the eddy covariance and Q measurements, which showed an ET to precipitation ratio of 92% and 86% and a Q to precipitation ratio of 9% and 17% during the pre-fire (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) and post-fire (2011-2017) periods, respectively.Note that, even though we found an increase in the absolute ET magnitude post-fire as measured by the eddy covariance tower (Section 3.1), when analyzing ET and Q as a percentage of incoming precipitation, there was a decrease and increase in the ET to precipitation and Q to precipitation ratio, respectively.This indicates that forest recovery impacts on streamflow can be hard to disentangle with measurements alone in this case due to confounding interannual and long-term climate variability effects.Figure 3d shows that T&C predicted an increase in Q in the year following the fire, while the discharge measurements showed such a Q increase post-fire with a lag of 2 years.Feikema et al. (2013) concluded that the replenishing of a substantial soil moisture deficit caused by dry periods before the fire could be the reason for not detecting an instantaneous increase in the Q measurements post-fire.The current T&C simulation was performed at the point scale and hence, we did not aim to analyze changes in soil moisture storage at the catchment scale in the current study as was done by Feikema et al. (2013), which might result in an earlier Q response.

| Evapotranspiration and runoff changes after wildfire explained with mechanistic ecohydrological modeling
We quantified the effects of forest recovery on the hydrological fluxes after the wildfire through the difference between a fire-on and undisturbed T&C simulation.This allows to eliminate the role of interannual climate variability.We found that after an initial sharp decrease in total ET caused by the forest destruction during the fire year, ET recovered fast and started exceeding the ET of the mature undisturbed forest as early as 2 years after the fire.
The ET increase peaked 3 years after the fire at +152 mm year −1 and while it showed a consistent downward trend thereafter, ET was still elevated compared to the levels of the mature forest 8 years after the fire (Figure 4a).The ET increase was partly caused by a small increase in ground evaporation, but it was mainly due to an increase in transpiration from the regrowing forest (Figure 4c).
While evaporation from interception was decreased in the 3 years following the fire, it showed a negligible difference in the long term (Figure 4c).Q sharply increased in the year following the fire due to the decreased ET caused by forest destruction.Subsequently, Q decreased and stayed suppressed from 2 years post-fire to the end of the simulation period at 8 years post-fire due to the increased ET flux (Figure 4a).There was a short-lived peak in Q due to forest destruction which only occurred the year following the fire as soil  2).The hydrological fluxes are expressed as a ratio of the incoming annual precipitation.
moisture was replenished first, shown as a positive change of dS/ dt, during the fire year (Figure 4a).While changes in Q and changes in ET were of similar magnitude (as they should) when analyzed in absolute units, that is, mm year −1 , the percentual change of ET and Q showed a large difference due to the hydrological flux partitioning of this forest ecosystem with approximately 90% of the precipitation going into ET and only approximately 10% contributing to Q (see Section 3.3).Hence, small percentual changes in ET, which peaked at 20% 3 years post-fire, led to much larger relative changes in Q (Figure 4b).Note that maximum percentual Q changes did not occur at the time of maximum percentual ET changes, which are 3 years post-fire, but 5-7 years after the fire in our simulations.The 2 years directly following the fire experienced larger than average precipitation amounts and a larger absolute Q amount, which led to the smaller percentual Q changes (∆Q/Q in %) compared to later time periods.Furthermore, interannual climate variability influenced absolute Q changes more than absolute ET changes (Figure 4a).Longterm droughts in itself can shift precipitation-runoff patterns (e.g., Saft et al., 2015) and an undisturbed T&C simulation shows lower runoff given a certain precipitation amount during the extended drought before the fire compared to the time period following the fire (Figure S1).However, including the fire in the simulations leads to the reversed pattern indicating that the forest regrowth effects dominate (Figure S1).
While previous research suggested that the increased ET of the young regrowing forest was caused by a higher stand density compared to the old-growth mature eucalypt forest, the T&C model simulations showed that LAI drastically decreased during the fire year, but recovered to pre-fire values within 3 years after the fire and thereafter did not show any large differences compared to the mature old forest, which included the understory vegetation canopy (Figure 4e).Our simulations rather showed that the higher ET of the young forest was mainly caused by aerodynamic effects hindered the dissipation of sensible heat from the plant canopy and led to higher surface temperatures (Figure 4d,f).The increase in surface temperature promoted ET due to a higher humidity gradient between leaf interiors and the air, which was able to offset the negative effects of increased aerodynamic resistance on the LE flux (aerodynamic resistance is included in the denominator of Equation 2), leading to a net increase in ET.In the simulations, stomatal resistance was also slightly decreased in the years following the fire, but it returned to pre-fire values within 4 years after the fire (Figure 4d).Hence, the observed increase in ET was mostly caused by the change from an old-growth forest with a height of up to 90 m to a young and short forest, which led to a smoother land surface, increased aerodynamic resistance, higher surface temperatures, and in this climate, higher ET.

| Climate effects on evapotranspiration and runoff changes after wildfires
Energy and water limitations can constrain LE and therefore, also the increase and decrease in ET and thus Q subsequent to forest recovery after the wildfire.This will occur independently of a theoretically higher or lower potential for ET in the modified forest ecosystem.Hence, we further analyzed how changes in meteorological conditions, which can lead to energy or water limitations, influenced the hydrological response of forest recovery after the Wallaby Creek wildfire.To do so, we performed simulations with T&C in which precipitation was increased and decreased by 20% (a proxy of changes in water availability) and a simulation in which incoming shortwave radiation was decreased by 20% (a proxy of energy constraints).The forest recovery effects were again isolated through the subtraction of an undisturbed simulation from a simulation including the wildfire for all the different climate scenarios.
Increasing precipitation by 20% did not lead to a change in the simulated ET increase after the Wallaby creek fire compared to the original climate forcing (Figure 5a), which indicated that ET in this scenario and in the original climate forcing was largely energy and not water limited.Decreases in Q were on average of a similar magnitude as simulated with the original climate forcing, but they were less influenced by climate variability and more consistent throughout the analyzed years (Figure 5b).Note that the relative Q decrease (∆Q/Q) was of smaller magnitude though due to the higher Q volume caused by higher rainfall.If precipitation was decreased by 20%, the post-fire ET increase was of much shorter duration and reached pre-fire values within 6 years after the fire (Figure 5a).However, the maximum ET increase was of the same magnitude as with the original climatic forcing.There was an above average rainfall amount during the 2 years following the fire which could have allowed the same increase in ET as in the original scenario as the system was likely still in an energy limited state during these years.A considerably smaller decrease in Q was simulated directly after the fire if analyzed in absolute units (i.e., mm year −1 ; Figure 5b).Due to the decrease in rainfall by 20%, the ecosystem tended toward a water limited state and the Q volume was low even without disturbance.Hence, the Q decrease after the fire was constrained by water availability and hence of smaller magnitude.
Note that the percentual Q decrease would be much larger though due to the already small discharge in the undisturbed scenario.
When incoming shortwave radiation was decreased by 20%, a smaller increase in post-fire ET was simulated which returned to prefire values at around 8 years after the fire (Figure 5a).Due to the smaller ET increase, Q decrease was also smaller post-fire in both absolute and relative units (Figure 5b).consider energy and water limitations in understanding the potential changes in ET of a recovering forest after wildfire.Even if the younger, shorter forest ecosystem could have higher ET rates due to aerodynamic effects than the mature forest, ET changes could be constrained by water availability.Similarly, the aerodynamic warming effects and ET increase of a shorter forest might be very marginal in climates with low energy inputs.

| Changes in LE and ET fluxes
Given the results of this study, we conclude that there is no prolonged decrease in ET after wildfires in eucalypt-dominated forests in south-eastern Australia, which is in stark contrast to the long-term decrease in ET reported after forest fires in many other parts of the world (e.g., Blount et al., 2020;Li et al., 2017;Li & Lawrence, 2017;Liu et al., 2019;Ma et al., 2020;Niemeyer et al., 2020).For example, Ma et al. (2020) found a persistent decrease in ET over 15 years following fires in California's Sierra Nevada and concluded that more frequent fires could even lead to higher downstream water availability.Similarly, decreased ET and elevated discharge were also reported several years after fires in other parts of the United States (Blount et al., 2020;Niemeyer et al., 2020).However, our analysis suggested that increased Q, which could benefit water resource availability in water scarce locations, was not occurring in south-eastern Australia after eucalypt forest fires.Our mechanistic ecohydrological modeling showed a rapid recovery in LAI and a 20% increase in ET 3 years after the Wallaby Creek fire along with a decrease in the simulated Q volume post-fire.These results aligned with the general patterns, albeit differ in magnitude and time of maximum effect, reported in previous research from southern Australia, which predicted a decrease in water yield due to ET increase after wildfires in catchments dominated by Mountain Ash eucalypt forest (e.g., Feikema et al., 2013;Haydon et al., 1997;Jaskierniak et al., 2016;Kuczera, 1987;Vertessy et al., 2001).

| Mechanisms for increasing ET
The increase in ET has been historically explained by a higher stand density and an increased sapwood area of the young Mountain Ash eucalypt forest, which led to higher transpiration rates and decreased Q compared to the old-growth forest (Benyon et al., 2015;Haydon et al., 1997;Jaskierniak et al., 2016;Vertessy et al., 2001).
For example, Vertessy et al. (2001) measured several vegetation indices and water balance components in multiple Mountain Ash stands to explain the empirically observed relationship between stand age and water yield in these forests (e.g., Kuczera, 1987).
They found that sap flux velocity did not vary significantly with stand age, however, the sapwood area index declined at older stands, which led to a decrease in transpiration in old-growth Mountain Ash forests (Vertessy et al., 2001).Our simulation results with T&C also showed elevated transpiration rates and decreased Q during the decade following the Wallaby Creek forest fire.However, we showed with the simulation results that these elevated ET rates were not caused as much by a change in vegetation LAI, but instead by aerodynamic effects.The destruction of the 90 m tall Mountain Ash forest and the regrowth of the juvenile trees led to decreased aerodynamic roughness (increased aerodynamic resistance) in the model, which hindered the turbulent transport of sensible heat resulting in higher surface temperatures and an increased humidity gradient between the tree leaves and the air.This higher humidity gradient caused higher transpiration rates in the model.Note that higher aerodynamic resistance would also hinder the turbulent transport of latent heat (aerodynamic resistance is included in the denominator of both Equations ( 1) and (2)), however, its effect on the leaf-to-air humidity gradient dominated and led to higher LE and transpiration in the Wallaby Creek forest.The T&C results did not solve for the planetary boundary layer and do not simulate land-atmospheric feedback effects which could be caused by extended landcover changes.However, atmospheric conditions at model forcing height were taken directly from pre-and post-fire observations, and thus might include some level of land-atmospheric feedback, if it occurred.Changes in albedo due to fires has also been shown previously (Liu et al., 2019) to lead to a land surface temperature increase.In the T&C simulations, albedo is modeled as a function of LAI and soil properties and does not show any prolonged changes post-fire.Furthermore, T&C does not account for any potential soil water hydrophobicity effects caused by the fire, and thus infiltration patterns are not dissimilar in pre-and post-fire conditions.The difference in mechanism explaining the increase in ET after forest fires is important as it implies increasing ET is not only confined to the regrowth of young Mountain Ash forests but could also occur with the regrowth of different vegetation species if the overall vegetation amount and cover, for example, LAI, does not significantly change.
For example aerodynamic effects were also reported in a recent study by Breil et al. (2021) which modeled the effects of afforestation in Europe and found increased ET rates from short grasslands due to lower aerodynamic roughness (increased aerodynamic resistance) leading to higher surface temperature and higher leafto-air humidity gradients compared to forests in large parts of south and central Europe.Similarly, Li and Lawrence (2017) found a significant increase in ET in equatorial African rainforests due to fires in a modeling study and attributed the increased ET to an increase in transpiration caused by higher leaf temperatures as a result of larger scale weather pattern changes caused by fires.While case studies in many parts of the world showed a decrease in ET and potentially an increase in downstream water availability (e.g., Blount et al., 2020;Ma et al., 2020;Niemeyer et al., 2020), these modeling studies (Breil et al., 2021;Li & Lawrence, 2017) showed that an increase in ET after forest destruction due to higher leaf temperatures was not only confined to Australia, where this effect was confirmed here providing mechanistic explanations for findings reported in previous research (e.g., Kuczera, 1987;Vertessy et al., 2001).Hence, further research is needed to assess the counterintuitive effect of decreased water availability in the years (not immediately though) following wildfires potentially also occurring in other parts of the world.This is especially important since globally occurring forest disturbances are predicted to lead to shorter and younger forest in the future (McDowell et al., 2020).

| Regional responses and impact on water availability
The increased ET and reduced discharge after the eucalypt forest fires in south-eastern Australia might put pressure on already scarce water resources.In the analyzed Mountain Ash eucalypt forest, almost 90% of the incoming precipitation was used for ET.
Hence, even small relative changes in ET can lead to a large relative change in Q.It has to be noted that we modeled the destruction of an extremely tall and old forest stand at the plot scale.
In the last two decades, a large increase in wildfire frequency was observed in Australia during which many of its eucalypt forests burned (Canadell et al., 2021;Lindenmayer & Taylor, 2020;McColl-Gausden et al., 2022).Hence, south-eastern Australia might not observe the magnitudes of ET increase and discharge reduction after wildfires shown here or in previous research in the future, because many of the Australian eucalypt forests will likely be younger and shorter during future fires.Additionally, extended droughts, such as the millennial drought occurring before the fire, can in itself shift the rainfall-runoff relationship leading to a decreased runoff for a given rainfall in comparison to historical conditions (Saft et al., 2015).When analyzed over a larger extent, we also saw here that the remotely sensed estimate of LE after multiple forest fires did not show a large increase in the years following the fire likely as the LE was computed over the total fire extent and forest age was not controlled for.Similarly, Feikema et al. (2013) concluded a lower Q decrease after wildfires due to heterogeneity in forest burn intensity and forest age at the catchment scale as not the whole catchment was covered by old-growth Mountain Ash forest.Additionally, higher frequency of forest fires due to climate change (Canadell et al., 2021) decreases not only forest age, but could also lead to changes in forest structure and species composition (Hill & Field, 2021), which in turn could alter canopy function.For example, Eucalyptus regnans trees take around 20 years to reach reproductive maturity and fires occurring more frequently than that will likely lead to a change in forest type (Bassett et al., 2015;Bowd et al., 2018;Bowman et al., 2016).A decrease in LAI and a reduction of vegetation cover were common explanations for the decrease in ET and increase in Q found in previous research in other parts of the world (e.g., Bart et al., 2016;Boisramé et al., 2019;Bond-Lamberty et al., 2009;Li & Lawrence, 2017).Hence, the impacts of higher frequency forest fires on water resource availability in south-eastern Australia in the future might differ from the findings presented here and from past research.

| The role of water and energy limitations
We further showed in this study that the maximum increase in ET and decrease in Q are constrained by energy and water limitations in the ecosystem.Similar findings were also reported in literature from different locations globally showing that water limitations can constrain both ET increase and decrease after forest disturbances (e.g., Biederman et al., 2014;Feikema et al., 2013;Saksa et al., 2017).For example, Boisramé et al. (2019) found a smaller decrease in ET and increase in Q, respectively, due to forest disturbances in water-limited catchments and Saksa et al. (2017) report that the benefits of forest thinning on downstream water availability did not necessarily occur if water became limiting.
Furthermore, Biederman et al. (2014) reported higher ET due to forest mortality in North America, however, this ET increase vanished during drier years due to water limitations.Such water and energy limitation aspects warrant future studies as they are likely essential in quantifying forest disturbance effects on water resources and explaining differences across climates and ecosystems globally.This is even more relevant in a climate change context as ecosystems are expected to become more water limited in the future (Denissen et al., 2022;Jiao et al., 2021) while also undergoing change due to increased forest disturbances (McDowell et al., 2020) and associated hydrological processes (McDowell et al., 2023).

| CON CLUS ION
Forest disturbances, such as wildfires are widespread and are expected to increase in frequency and extent due to climate change (Abatzoglou & Williams, 2016;Canadell et al., 2021).Forest destruction due to wildfires and the subsequent regrowth of vegetation can impact water resource availability due to changes in ET.We provided a multi-data analysis over an old-growth Mountain Ash (Eucalyptus regnans) forest that was destroyed during the Black Saturday bushfires in 2009 by combining 18 years of eddy covariance measurements, remote sensing products, and ecohydrological modeling.We showed that LAI and ET recovered extremely fast.By extending the analysis spatially with the use of remote sensing to multiple fires, we showed that such a fast ET recovery is likely generalizable for eucalypt forests after bushfires in south-eastern Australia and no medium-to longterm increase in water resources availability can be expected as was found by research in other parts of the world (e.g., Blount et al., 2020;Li & Lawrence, 2017;Ma et al., 2020;Niemeyer et al., 2020).In contrast, at the plot scale, modeled ET exceeded pre-fire values by 20% 3 years after the fire and stayed elevated in the decade following the fire.Mechanistic ecohydrological modeling showed that the increased ET is caused by aerodynamic effects of a much shorter forest stature post-fire, which led to higher surface temperatures and humidity gradients, which caused increased transpiration in an ecosystem which was largely energy limited.This was in contrast to previous research which explained the reduction in streamflow post-fire due to a higher stand density (e.g., Kuczera, 1987;Vertessy et al., 2001).The mechanistic explanation has important implications for the anticipated hydrological changes after forest disturbances, which are predicted to lead to a shorter and younger forest in other parts of the world (McDowell et al., 2020).As the ET increase is caused by aerodynamic effects, similar results could also occur in other parts of the world where differences in vegetation height pre-and post-disturbance are very pronounced with negative impacts on downstream water availability in water scarce regions.However, we also showed in a numerical experiment that the ET increase and Q decrease post-fire could be constrained by energy and water limitations in the ecosystem, remarking the complex picture of global hydrological responses following forest disturbances.Mechanistic modeling such as presented here could be applied regionally and globally to assess the effects of forest disturbance on land atmosphere feedbacks and implications for hydrology worldwide.
to explain the mechanistic effects leading to the observed ET and Q responses after the wildfire and to remove the confounding effects of interannual climate variability.With this analysis, we aimed to answer research questions such as: What is the short-and long-term observed ET response after the destruction of a tall, old-growth Mountain Ash eucalypt forest (Eucalyptus regnans) in south-eastern Australia and do model results concur?What are the general patterns of eucalypt forest recovery after fire in south-eastern Australia as shown by remote sensing?What are the mechanisms leading to the observed changes and could this have global implications for ET , the observed LE changes post-fire were difficult to separate from climate variability effects.Hence, in a second step, we quantified the hydrological changes after the fire with targeted T&C simulations to remove the confounding effects of interannual climate variability.This was done by analyzing the difference in the hydrological fluxes between a simulation including the wildfire and a simulation without any forest disturbance.Ecohydrological modeling was further used to explore and explain the mechanisms leading to the simulated LE, ET, and Q changes by analyzing the difference in modeled vegetation indices, evapotranspiration sources, and aerodynamic variables governing the turbulent transport of energy.Third, we performed numerical experiments with T&C by perturbing the climate forcing to quantify the effects of energy and water limitations in the ecosystem in the modeled ET and Q response after the wildfire.Prior, the model performance was evaluated against the LE eddy covariance measurements and against discharge measurements from catchments affected by the bushfire.Last, to determine if the observed and modeled LE and ET changes at the Wallaby Creek flux tower site are generalizable to eucalypt forest recovery at the larger scale, we analyzed timeseries of LE, surface temperature, and vegetation indices for the entire Wallaby Creek fire extent (ca.895 km 2 ) and several other large wildfires occurring from 2006 to 2014 in the state of Victoria, Australia.

F
I G U R E 1 (a) Location of the Wallaby creek eddy covariance tower in the state of Victoria, Australia.Further shown is the location and extent of seven fires for which vegetation indices, latent heat flux, and land surface temperature were extracted from remote sensing products.Five of the analyzed fires occurred in the Australian summer of 2009 while one fire occurred in the summer of 2006-2007 and one occurred in 2014.(b) Monthly average and (c) annual timeseries of latent heat (LE) including the fire in 2009 for eddy covariance measurements (FT), a remote sensing product (MODIS) and the T&C simulations (T&C) at the Wallaby creek forest site.The MODIS LE timeseries is extracted within a radius of 1 km from the eddy covariance flux tower site and the T&C simulations were performed at the forest stand scale.
ecohydrological model Tethys-Chloris (T&C;Fatichi et al., 2012aFatichi et al., , 2012b) ) was used to quantify and explain the mechanisms leading to the observed and modeled changes in the hydrological fluxes postfire without the confounding effects of interannual climate variability.Additionally, numerical experiments were performed with T&C to quantify the effects of water and energy limitations in the ecosystem on post-fire hydrological changes.T&C(Fatichi et al., 2012a(Fatichi et al., , 2012b) is a mechanistic ecohydrological model that resolves the coupled dynamics of energy, water, and carbon fluxes at the land surface at the hourly scale.T&C includes all main hydrological fluxes, resolves for variably saturated soil across multiple layers, and accounts for biophysical and ecophysiological properties of vegetation.The model calculates a prognostic land surface temperature by solving the energy budget accounting for absorbed shortwave and longwave radiation, sensible heat (H), LE, and the conductive heat flux in the soil.H and LE are calculated applying a series of resistances (r i ): 145°11′14″ E) for the time period of 2000-2017.The simulations included the wildfire occurring in February 2009 which led to complete forest destruction and subsequently modeled regrowth of a dense canopy of young ash-type eucalypt forest.Note, the T&C simulation prior to the fire was designed to include both the overstory and understory vegetation canopy, while post-fire, only one vegetation canopy was defined.The fire was prescribed in the model and occurred over 1 day during which all aboveground biomass was removed.Part of the nutrients (i.e., 25% of nitrogen) stayed in the model as ashes and the regrowth of the young forest was triggered by the seed bank included in the model.A simulation without fire was also performed as a baseline scenario to compare with the effects of forest regrowth after the wildfire on the hydrological fluxes without the confounding effects of interannual climate variability.Model performance was evaluated through the comparison of simulations with the eddy covariance data presented in this study.T&C's water partitioning between ET and Q, was further evaluated against discharge measurements (Table2).The changes in the hydrological fluxes (ET, Q, water storage change over the time period dt, dS/dt) caused by the fire were quantified by subtracting the simulations in which no fire occurred, in the following called "undisturbed simulation", from the simulations in which fire occurred.The changes in ET and Q were analyzed in absolute (mm year −1 ) and in relative terms (%, change in flux divided by total flux of the undisturbed simulation).Changes in the total ET were further attributed to changes from vegetation transpiration, evaporation from interception, and soil evaporation.Additionally, modeled changes in leaf area index (LAI), land surface temperature, and resistances governing the turbulent transport of heat and water vapor were analyzed to explain the mechanisms leading to the simulated changes in ET (Section 3.4).

|
Latent heat dynamics including wildfire at the flux tower site LE timeseries from eddy covariance measurements, from remote sensing, and modeled with T&C were analyzed to assess the effects of forest recovery after the Wallaby Creek wildfire on LE and therefore ET (Figure1).As expected, all data sources showed a sharp drop in LE during the fire year.Annual average LE decreased by 45%, 35%, and 45% in 2009 compared to the long-term pre-fire annual average for eddy covariance measurements, T&C simulations, and remote sensing derived LE, respectively.However, the recovery is rapid and LE reached pre-fire values as soon as 2 years after the fire in 2011 (Figure1).These 2 years following the fire (2009 and 2010) during which rapid recovery occurred were subsequently excluded from the statistical comparison as they are not representative of the longer term LE changes following the fire.The annual average post-fire (2011-2017) LE from eddy covariance measurements was 4% larger at the yearly scale than pre-fire(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008) with a p-value of 0.04 (t-test) while, the average summer season (D, J, F) LE was 20% larger post-fire compared to pre-fire (pvalue of <10 −4 , t-test).Similarly, the LE derived from the MODIS remote sensing product within a 1 km radius of the Wallaby Creek flux tower site also showed a 5% increased LE post-fire at the annual scale (p-value of 0.08, t-test) and an average summer season LE, which was 14% larger post-fire compared to pre-fire (p-value of <10 −3 , t-test).
timeseries of remotely sensed vegetation indices (NDVI, EVI, LAI), LE, and daytime land surface temperature (LST day ) for the Wallaby Creek fire extent (ca.895 km 2 ) and six additional fires affecting medium and tall eucalypt forests in the state of Victoria, Australia showed that a fast recovery of vegetation indices and LE flux to pre-fire values was generalizable to eucalypt forests in south-eastern Australia.Fitting of recovery curves on the remote sensing data suggested that none of the seven analyzed fire areas had experienced a long-term decrease or increase in NDVI, EVI, or LAI after the fire at the spatial average of their fire extents (Figure2).Vegetation indices reached pre-fire levels in less than 10 years for the majority of the analyzed fire areas (Figure2) and the predicted long-term degree of recovery was within ±6% of the absolute prefire NDVI, EVI, or LAI values.Similarly, LE had recovered fast and the increase or decrease in LE was predicted to be less than 5% of the absolute pre-fire value in the long-term for the analyzed fire areas.In contrast to the analysis of the eddy covariance footprint at the forest stand scale (extracted within 1 km radius from the flux tower, Section 3.1), the fitting of the recovery curve to the MODIS derived LE flux averaged over the total Wallaby Creek fire extent (ca.895 km 2 ) rather predicted a slightly decreased long-term LE flux at around −4% of the pre-fire LE.Daytime land surface temperature recovery time showed a higher variability than the other analyzed variables with certain fire areas recovering fast while others taking more than 10 years.

F
Recovery of vegetation indices (a) normalized difference vegetation index (NDVI), (b) enhanced vegetation index (EVI), (c) leaf area index (LAI), (d) latent heat flux (LE), and (e) daytime land surface temperature (LST day ) after forest fires analyzed for seven different fire areas using remote sensing products.The data and fitted recovery curve for the Wallaby Creek (WB) fire extent is highlighted in red and purple.Vertical black bars show the predicted time of recovery for the individual fire areas.

F
I G U R E 3 T&C model performance assessment.(a) Comparison of the modeled (T&C) and measured (FT) partitioning of the evapotranspiration (ET) pre-fire into evapotranspiration from soil (E s ), evapotranspiration from the understory (ET u ), and evapotranspiration from the overstory (ET o ).ET e,pre-fire and ET e,post-fire show the average modeled and measured total ecosystem evapotranspiration pre-fire (2000-2008) and post-fire (2011-2017), respectively.Black bars show one standard deviation of the annual mean.(b) comparison of the modeled hourly latent heat flux with T&C (LE T&C ) against eddy covariance measurements (LE FT ).R 2 denotes the coefficient of determination, MBE the mean bias error, and RMSE the root mean square error.The red line displays the 1:1 standard line and the color bar shows the density of data in the plot.(c) comparison of the seasonal modeled (T&C) and measured (FT) latent heat flux (LE).Points show the daily average LE flux for each day of the year and lines show the 2-week moving average.(d) comparison of the annual modeled hydrological flux partitioning into runoff (Q T&C ), evapotranspiration (ET T&C ) and water storage change (dS/dt T&C ) against the measured eddy covariance evapotranspiration flux (ET FT ), and runoff measurements at King Parrot Creek (Q King Parrot ), Sunday Creek (Q Sunday ) and Westcott Creek (Q Westcott ; Table

(
Figure 4d) as explained in the following.Due to the much shorter forest stature of the young eucalypt forest, aerodynamic roughness decreased leading to an increase in aerodynamic resistance which F I G U R E 4 Change in hydrological fluxes, aerodynamic and stomatal resistances, leaf area index, and surface temperature between the T&C simulation including forest destruction due to fire and subsequent regrowth of young eucalypt forest and a T&C simulation of an old undisturbed mature eucalypt forest.(a) Change in evapotranspiration (∆ET), runoff (∆Q), and water storage (∆dS/dt) in absolute units (mm year −1 ), (b) Change in evapotranspiration and runoff expressed as the percentage of the evapotranspiration and runoff flux of the undisturbed simulation, (c) Change in total ecosystem evapotranspiration (∆ET tot ), transpiration from vegetation (∆T veg ), evaporation from interception (∆E in ), and ground evaporation (∆E G ), (d) Change in aerodynamic resistance between vegetation canopy and reference height of observations (∆r aero,veg ), aerodynamic resistance between vegetation and ground surface (∆r aero,soil ), soil resistance (∆r soil ), and stomatal resistance (∆r stomata ), (e) change in leaf area index (∆LAI), and (f) change in average surface skin temperature (∆T s ).
Due to the decrease in incoming shortwave radiation, the increase in surface temperature caused by the aerodynamic effects of forest structure change was less pronounced, which led to the smaller simulated ET increase.In summary, our additional simulations showed that it is important to F I G U R E 5 Changes in (a) evapotranspiration (∆ET) and (b) runoff (∆Q) between the T&C simulation including forest destruction due to the Wallaby Creek fire and subsequent regrowth of young eucalypt forest and the T&C simulation of an old undisturbed mature eucalypt forest for different climate forcing scenarios.Orig denotes the original climate forcing, P r−20% denotes that precipitation was reduced by 20%, Pr +20% denotes that precipitation was increased by 20%, and SWR −20% denotes that incoming solar shortwave radiation was reduced by 20%.
r i , VIC: State of Victoria (Department of Environment, Land, Water and Planning), downloaded from http:// www.bom.gov.au/ water data/ .