As a result of climate change warmer temperatures are projected through the 21st century and are already increasing above modelled predictions. Apart from increases in the mean, warm/hot temperature extremes are expected to become more prevalent in the future, along with an increase in the frequency of droughts. It is crucial to better understand the response of terrestrial ecosystems to such temperature extremes for predicting land-surface feedbacks in a changing climate. While land-surface feedbacks in drought conditions and during heat waves have been reported from Europe and the US, direct observations of the impact of such extremes on the carbon and water cycles in Australia have been lacking. During the 2012/2013 summer, Australia experienced a record-breaking heat wave with an exceptional spatial extent that lasted for several weeks. In this study we synthesised eddy-covariance measurements from seven woodlands and one forest site across three biogeographic regions in southern Australia. These observations were combined with model results from BIOS2 (Haverd et al., 2013a, b) to investigate the effect of the summer heat wave on the carbon and water exchange of terrestrial ecosystems which are known for their resilience toward hot and dry conditions. We found that water-limited woodland and energy-limited forest ecosystems responded differently to the heat wave. During the most intense part of the heat wave, the woodlands experienced decreased latent heat flux (23 % of background value), increased Bowen ratio (154 %) and reduced carbon uptake (60 %). At the same time the forest ecosystem showed increased latent heat flux (151 %), reduced Bowen ratio (19 %) and increased carbon uptake (112 %). Higher temperatures caused increased ecosystem respiration at all sites (up to 139 %). During daytime all ecosystems remained carbon sinks, but carbon uptake was reduced in magnitude. The number of hours during which the ecosystem acted as a carbon sink was also reduced, which switched the woodlands into a carbon source on a daily average. Precipitation occurred after the first, most intense part of the heat wave, and the subsequent cooler temperatures in the temperate woodlands led to recovery of the carbon sink, decreased the Bowen ratio (65 %) and hence increased evaporative cooling. Gross primary productivity in the woodlands recovered quickly with precipitation and cooler temperatures but respiration remained high. While the forest proved relatively resilient to this short-term heat extreme the response of the woodlands is the first direct evidence that the carbon sinks of large areas of Australia may not be sustainable in a future climate with an increased number, intensity and duration of heat waves.
Average temperatures in Australia have increased by 0.9
During summer 2012/2013, Australia experienced a record-breaking heat wave
that was deemed unlikely without climate change (Steffen, 2015). The
Australian summer 2012/2013 was nicknamed the “Angry Summer” or the “Extreme
Summer”, as an exceptionally extensive and long-lived period of high
temperatures affected large parts of the continent in late December 2012 and
the first weeks of January 2013 (Bureau of Meteorology, BOM, 2013). Record
temperatures were observed in every Australian state and territory, and the
record for the hottest daily average temperature (32.4
Heat waves are becoming hotter, they last longer, and they occur more often (Steffen, 2015). As many ecological processes are more sensitive to climate extremes than to changes in the mean state (Hanson et al., 2006), it is imperative to understand the effect of climate extremes in order to predict the impact on terrestrial ecosystems. Processes and sensitivities differ among biomes, but forests are expected to experience the largest detrimental effects and the longest recovery times from climate extremes due to their large carbon pools and fluxes (Frank et al., 2015). There is increasing evidence that climate extremes may result in a decrease in carbon uptake and carbon stocks (Zhao and Running, 2010; Reichstein et al., 2013). It is therefore crucial to better understand ecosystem responses to climate extremes. The role of climate extremes could be critical in shaping future ecosystem dynamics (Zimmermann et al., 2009), but the sporadic and unpredictable nature of these events makes it difficult to monitor how they affect vegetation through space and time (Mitchell et al., 2014).
List of OzFlux sites used in this study, abbreviations and site information. MW stands for Mediterranean woodlands, TW for temperate woodlands and TF for temperate forest. MAT and MAP are the mean annual temperature and precipitation for the years 1982–2013 (BIOS2).
Map indicating the locations of the OzFlux sites used in this study. The sites are grouped into three distinct climate and ecosystem types, indicated by red dots for Mediterranean woodlands (MW), light green dots for temperate woodlands (TW) and a dark green dot for the temperate forest (TF).
Australian forest and woodland ecosystems are strongly influenced by large
climatic variability, characterised by recurring drought events and heat
waves (Beringer et al., 2016; Mitchell et al., 2014).
The large spatial extent of the heat wave in early 2013 across Australia and direct observations from the OzFlux network enable us for the very first time to analyse the effect of extreme hot and dry conditions on the carbon, water and energy cycles of the major woodland and forest ecosystems across southern Australia. In this study, we combined eddy-covariance measurements from seven woodland and forest sites with model simulations from BIOS2 (Haverd et al., 2013a, b) to investigate the impact of the 2012/2013 summer heat wave and drought on the carbon and water exchange of terrestrial ecosystems across climate zones in southern Australia and to assess the influence of land-surface feedbacks on the magnitude of the heat wave.
We compared hourly data from seven OzFlux sites (Fig. 1, Table 1), measured during the heat wave period 1–18 January 2013, to observations from a background reference. We used eddy-covariance data to compare hourly data and the daily cycle of latent and sensible heat as well as carbon fluxes. We used the measured hourly data of a background period (BGH) one year later from 2 to 6 January 2014. During these time periods all towers were actively taking measurements, although data gaps were present after 18 January in 2013. The reference period was shorter than the heat wave period because another significant heat wave event affected southeastern Australia in late January 2014 during a time period when not all sites had comparable data available in 2013. Temperatures during the background reference period were also somewhat warmer than average climatology (Fig. 2). We therefore expect the relative severity of the effects of the heat wave to appear smaller than they otherwise would when compared against a climatological reference. To ensure the representativeness of our results, we also compared daily data against a climatology derived from daily BIOS2 (see below) output for the time period 1982–2013 (background climatology, BGC). BIOS2 results for the whole time period were only available as daily values.
We analysed data from seven southern Australian sites (Beringer et al., 2016), grouped into three distinct ecosystem and climate types: Mediterranean woodlands (MW), temperate woodlands (TW) and temperate forests (TF; Fig. 1, Table 1).
MW sites included (i) a coastal heath
The sites fall into the classifications “Mediterranean forests, woodland
and scrub” (AU-Gin, AU-GWW and AU-Cpr) or the “temperate broadleaf and
mixed forest” (AU-Wom, Au-Cum, AU-Whr and AU-Tum) classifications of IBRA
(Interim Biogeographic Regionalisation for Australia v. 7; Environment,
2012). In temperate Australia both woodlands and forests are mainly
dominated by
We analysed data collected by the OzFlux network (
The coupled carbon and water cycles were modelled using BIOS2 (Haverd
et al., 2013a, b) constrained by multiple observation
types, and forced using remotely sensed vegetation cover and daily AWAP
meteorology (Raupach et al., 2009), downscaled to half-hourly time resolution
using a weather generator. BIOS2 is a fine-spatial-resolution (0.05
All data analyses were performed on Jupyter notebooks using Python 2.7.11
and the Anaconda (4.0.0) distribution by Continuum Analytics. Differences
between heat waves and reference periods were determined by calculating
We use the terminology and concepts as introduced by Chapin et al. (2006), where net and gross carbon uptake by vegetation (net ecosystem production (NEP) and gross primary production; GPP) are positive directed toward the surface and carbon loss from the surface to the atmosphere (ecosystem respiration; ER) is positive directed away from the surface.
The heat wave event commenced on 25 December 2012 with a build-up of extreme heat in the southwest of Western Australia. A high-pressure system in the Great Australian Bight and a trough near the west coast directed hot easterly winds over the area (BOM, 2013). From 31 December the high pressure system started moving eastward, and it entered the Tasman Sea off eastern Australia on 4 January. The northerly winds directed very hot air into southeastern Australia. Temporary cooling was observed in the eastern states after 8 January, but a second high pressure system moved into the bight in the meantime, starting a second wave of record-breaking heat across the continent. The heat wave finally ended on 19 January, when southerly winds brought cooler air masses to southern Australia.
Statistics of radiation and energy exchange for the
ecosystems Mediterranean woodlands (MW), temperate woodlands (TW) and
temperate forests (TF) and the variables flux of shortwave downward
radiation (Fsd), shortwave upward radiation (Fsu), longwave downward
radiation (Fld), longwave upward radiation (Flu), net radiation (Fn), latent
heat (Fe), sensible heat (Fh), ground heat (Fg) and the energy imbalance
(
Time series of daily maximum temperature (
Figure 3 shows the meteorological conditions at the sites during the heat
wave. Maximum temperatures as high as 46.3
During HW1 temperatures were generally more than 1.5–2 standard deviations
(
Box plot of energy fluxes for Mediterranean woodlands (MW, red),
temperate woodlands (TW, light green) and temperate forests (TF, dark
green). Energy fluxes are incoming shortwave radiation (Fsd), reflected
shortwave radiation (Fsu), downward longwave radiation (Fld), emitted
longwave radiation (Flu), net radiation (Fn), latent heat flux (Fe),
sensible heat flux (Fh), ground heat flux (Fg) and energy imbalance
(
The
Diurnal course of net radiation (Fn, light amber), sensible (Fh,
red) and latent (Fe, blue) heat at the Mediterranean woodlands (MW, top
row), the temperate woodlands (TW, middle row) and the temperate forest (TF,
lowest row) for the background period BGH (2–6 January 2014), and the
first and second period of the heat wave (HW1, 1–9 January 2013; HW2,
10–18 January 2013). Filled areas indicate the range of smoothed
Incoming and reflected short-wave radiation were significantly increased by
only 70 and 3 W m
Average daytime Bowen ratio measured over Mediterranean woodlands (MW, left panel), the temperate woodlands (TW, middle panel) and the temperate forest (TF, right panel) for BGH (green line), HW1 (light amber) and HW2 (dark amber).
Figure 5 demonstrates how remarkably different the energy partitioning was at MW, TW and TF sites, as we would expect given their large climatological and biogeographic differences (Beringer et al., 2016). While similar fractions of energy went into latent and sensible heat at the TF site, more energy was directed into sensible heat at TW sites. This energy flux partitioning toward sensible heat was more pronounced at MW sites, where both the mean and the variability of latent heat flux were very small due to severe water limitations. Most of the available energy was transferred as sensible heat and hence contributed to the warming of the atmosphere which was also observed for BGH.
During HW1, the generally small latent heat flux at the MW sites (38 W m
With values exceeding 7, the observed ratio of sensible to latent heat, the
Bowen ratio (
Diurnal course of net ecosystem productivity (NEP, light green)
and gross primary productivity (GPP, dark green) at the Mediterranean
woodlands (MW, top row), the temperate woodlands (TW, middle row) and the
temperate forest (TF, lowest row) for the background period (BHG), and the
first and second period of the heat wave (HW1, HW2). Filled areas indicate
the range of smoothed
Measured daily latent heat fluxes and
Patterns of carbon fluxes were similar to between-site patterns of energy
fluxes (Fig. 7, note differences in
Carbon uptake was significantly reduced at MW and TW during HW1 (Fig. 8)
with daytime averages decreasing from 4.6 to 3.1
in MW and from 11.2 to 6.2
Measured GPP and ER showed the same responses in carbon uptake and losses during the heat waves as the flux climatology derived with BIOS2 (BGC, Fig. A2): GPP was reduced during HW1 in woodland ecosystems and increased in the forest during both heat wave periods. ER was increased at all sites and during HW1 and HW2 compared to the long-term climatology.
Persistent anticyclonic conditions during the “Angry Summer of 2012/13” led to a heat wave by transporting warm air from the interior of the continent to southern Australia. Such synoptic conditions are the most common weather pattern associated with Australian heat waves (Steffen et al., 2014). However, these weather patterns did not result in increased amounts of available energy at the surface, which was in contrast to heat waves observed in Europe and the USA (see Sect. 4.4). Instead, in our study the energy available for turbulent heat fluxes was similar to or even smaller than background conditions. Background conditions over Australia tend to have large available energy fluxes, even during very cyclonic periods (e.g. the 2010–2011 fluvial; Cleverly et al., 2013). Thus, differences in latent and sensible heat fluxes at the Australian sites used in this study were due to anomalous temperature and soil moisture content rather than to changes in available energy.
Boxplot of daytime values (09:00–16:00 local standard time) of gross primary productivity (GPP), ecosystem respiration (ER) and net ecosystem productivity (NEP) for the background period (BGH) and the first and second period of the heat wave (HW1, HW2). Daytime average values (DTA) are given below boxes and symbols indicate that they are significantly different from the background period (o) or not (x). Daily averages (DA, 00:00–23:00, local standard time) and their significance are also given. Colours as in Fig. 1.
During the heat wave, available energy preferentially increased sensible
heat flux and led to a subsequent increase of
Heat waves and drought can affect photosynthesis (Frank et al., 2015). By
means of stomatal regulation, plants exert different strategies to balance
the risks of carbon starvation and hydrological failure (Choat et al., 2012).
These strategies particularly come into play during extreme events (Anderegg
et al., 2012). While the ecosystem response during heat waves is linked to
plant stress from excessively high temperatures and increased evaporative
demand (i.e. higher vapour pressure deficit), drought stress occurs when
soil water supply can no longer meet the plant evaporative demand. The
former will lead to reduced carbon uptake through e.g. stomatal closure and
disruptions in enzyme activity – the latter can have direct impacts on carbon
uptake by reducing stomatal and mesophyll conductance, the activity and
concentrations of photosynthetic enzymes (Frank et al., 2015, and references
therein). Apart from these almost instantaneous responses, additional lagged
effects can further impact the carbon balance. If high temperatures were to
occur in isolation we would expect to observe a decrease in GPP. During the
2012/2013 heat waves in Australia, we observed a diurnal asymmetry in GPP at
all sites and in all measurement periods. This is expected in ecosystems
that exert some degree of stomatal control to avoid excessive reductions in
water potential (e.g. in the afternoon), during higher atmospheric demand
and when there is a reduced ability of the soil to supply this water to the
roots because of lower matrix potentials and hydraulic conductivity (Tuzet
et al., 2003). Daily average carbon uptake at MW and TW was reduced by up to
32 and 40 %, respectively. At the TF site, however, daily averaged
carbon uptake did not change significantly, and daytime carbon uptake was
significantly increased during both periods of the heat wave (see also Fig. 7). This can be explained partly by the very dry conditions during the
background period at this site, which could also have caused below average
carbon uptake, although comparing the site data against the long-term
climatology confirmed an increased carbon uptake during the heat wave (not
shown). Although air temperatures clearly exceeded the ecosystem scale
optimum of 18
Heat waves and drought not only affect photosynthesis but also have an impact on respiration (Frank et al., 2015). Increases in ER during the heat wave seem intuitive, given the exponential response of respiration to temperature (e.g. Richardson et al., 2006). Drought can also override the positive effect of warmer temperatures and lead to reduced respiration due to water limitations, as observed during the 2003 heat wave (Reichstein et al., 2007) or the 2011 spring drought (Wolf et al., 2013) in Europe. However, during the observed heat waves in Australia, increased air and soil temperatures led to significantly increased ecosystem respiration at all sites, indicating that the thermal response of respiration was undiminished despite soil moisture deficits.
While all sites remained carbon sinks during daytime hours in both heat wave
periods, reduced carbon uptake in the woodlands turned them to a net a
source of carbon on a daily average. It can hence be concluded that
increased ER combined with decreased or unchanged GPP likely turned large
areas of southern Australia from carbon sinks to sources. Unlike the
Mediterranean woodlands, the temperate woodlands recovered quickly after
rain but the response of these ecosystems to a short though intense heat
wave indicates that future increases in the number, intensity and duration
of heat waves can potentially turn the woodlands into carbon sources,
leading to a positive carbon–climate feedback. Heat waves can also induce a
transition from energy-limited to water-limited ecosystems (Zscheischler et al., 2015). Transitioning toward water limitation, especially for
energy-limited forests, can exacerbate the detrimental effects of extreme
events. Recurrent non-catastrophic heat stress can also lead to increased
plant mortality, the impact of which would be more evident over longer
timescales (McDowell et al., 2008) and as an increase in the frequency of
fires (Hughes, 2003). Similarly, legacy or carry-over effects of drought
result in increased mortality and shifts in species composition during
subsequent years (van der Molen et al., 2011). Future climate change is
likely to be accompanied by increased plant water-use efficiency due to
elevated CO
Intervening rain events led to differentiated responses in energy fluxes and
lower air temperatures, but soil moisture content remained mostly low during
HW2 (see Sect. 3.1). Available energy was significantly lower (compared to
BGH) during HW2 at MW but remained similar at TW and TF. At TF the latent
heat flux in HW2 was still enhanced compared to BGH yet smaller than during
HW1. Following rainfall the energy partitioning at the MW sites changed
toward latent heat flux, with fractions similar to or larger than background
conditions. This indicates that soil moisture feedbacks which inhibit
warming of the lower atmosphere largely led to a return to standard
conditions. At TW,
During HW1, the time of maximum carbon uptake at the woodland sites was earlier in the morning than during BGH, and we observed strongly reduced carbon uptake throughout the day. During HW2, however, the shift of maximum GPP toward earlier hours of the day was less pronounced at MW and TW; thus daytime carbon uptake was not significantly reduced. This was in response to the intermittent precipitation and lower temperatures, which led to a reduction in vapour pressure deficit and increased soil water availability. Increased ER at all sites and during both HW periods was dominated by warmer temperatures more than soil moisture limitations. Increased ER combined with decreased or unchanged GPP likely turned large areas of southern Australia from carbon sinks to sources, an effect that was reduced but not offset by the intermittent precipitation.
When carbon losses exceed carbon gains over a long time period (e.g. through increased respiration) mortality can result as a consequence of carbon starvation. Eamus et al. (2013) identified an increased vapour pressure deficit as detrimental to transpiration and net carbon uptake, finding that increased vapour pressure deficit is more detrimental than increased temperatures alone – with or without the imposition of drought. A recent study by Sulman et al. (2016) confirmed that episodes of elevated vapour pressure deficit could reduce carbon uptake regardless of changes in soil moisture. Here, all ecosystems responded with increased carbon uptake to the precipitation events and the associated lower temperatures and vapour pressure deficit. The improved meteorological conditions thus likely decreased the risk of mortality during HW2. As heat waves increase in frequency, duration and intensity in the future (Trenberth et al., 2014), however, we expect a decline in the ameliorating effects of intermittent rain events and an increased risk of mortality.
Anticyclonic conditions also caused the intense 2003 European heat wave
(Black et al., 2004) as well as the even more intense and widespread heat
wave that reached across eastern Europe, including western Russia, Belarus,
Estonia, Latvia, and Lithuania in 2010 (Dole et al., 2011). Less cloud cover
and more clear sky conditions strongly increased incoming radiation and
available energy during the European heat wave and drought in 2003 (Teuling
et al., 2010), as well as during the recent drought and heat in California
(Wolf et al., 2016), in contrast to the current study. Teuling et al. (2010) observed that surplus energy led to increases in both latent and
sensible heat fluxes: over grassland, the energy was preferentially used to
increase the latent heat flux, thereby decreasing
Stomatal control and reductions in GPP at the dry sites (MW and TW) were consistent and of similar magnitude with observations made during e.g. the 2003 European heat wave (Ciais et al., 2005), the 2010 European heat wave (Guerlet et al., 2013), the 2012 US drought (Wolf et al., 2016) and the 2013 heat wave and drought that affected large parts of southern China (Yuan et al., 2015). During these heat waves and droughts, carbon uptake was strongly reduced in general and biosphere–atmosphere feedbacks from reduced vegetation activity further enhanced surface temperatures. This contrasts with the wet site (TF), where local drought effects were observed only toward the end of the study. We found that the response of carbon fluxes of Australian woodland (dry) ecosystems were similar to comparable heat waves on other continents, whereas the detrimental effects of the heat wave were largely ameliorated in wet, energy-limited Australian ecosystems.
Temperature anomalies during the 2012/2013 heat wave in Australia were less
extreme (
We have shown that extreme events such as the “Angry Summer” of 2012/2013 can alter the energy balance and therefore dampen or amplify the event. During this event the woodland sites reduced latent heat flux by stomatal regulation in response to the warm and dry atmospheric conditions. Stronger surface heating in the afternoons then led to an amplification of the surface temperatures. Only the forest site AU-Tum had access to readily available soil water and showed increased latent heat flux. The increased latent heat flux mitigated the effect of the heat wave but continuously depleted the available soil water. The generally increased atmospheric and soil temperatures led to increased respiration but unchanged net ecosystem productivity. The woodlands turned from carbon sinks into carbon sources and while the temperate woodlands recovered quickly after rain, the Mediterranean woodlands remained carbon sources throughout the duration of the heat wave. This demonstrates that there is potential for positive carbon–climate feedbacks in response to future extreme events, particularly if they increase in duration, intensity or frequency.
We have used measurements of a reference period during the same season but one year after the 2012/2013 heat wave occurred. Ideally we would have used a climatology derived from observations but OzFlux is a relatively young flux tower network. The first two towers started in 2001 and even globally, very few flux towers have been measuring for more than 15 years, which is relatively short compared to typical climatology records of 30 years. To ensure the representativeness of our results we have therefore compared daily data against a climatology derived from BIOS2 output for the time period 1982–2013.
Parameters of robust linear model fit between observations and BIOS2 output for all sites, the variables latent heat flux (Fe), gross primary productivity (GPP), ecosystem respiration (ER) and the time interval 1 January–31 December 2013.
Table A1 shows the agreement between BIOS2 output for all sites and the time period 1 January to 31 December 2013. Agreement was generally very good, even more so for the latent heat flux than for the carbon fluxes. Carbon fluxes, and more specifically respiration at the dry Mediterranean woodlands, showed stronger disagreement. It is likely that this to some degree reflects nighttime issues with the eddy-covariance method (e.g. van Gorsel et al., 2009) and with the partitioning of the measured fluxes. This may also be an indication that the model was underestimating drought-tolerance at these sites. The low modelled carbon uptake corresponded to periods of low soil water. There were long periods when the modelled soil water was below wilting point within the entire root zone of 4 m. Underestimation could occur if roots were accessing deeper water, the wilting point parameter was too high or the modelled soil water was too low, relative to the wilting point.
Left panel: boxplot of the ratio of observed latent heat (Fe(obs)) to the BIOS2 climatology of the latent heat flux (Fe(BGC)) during the first and second period of the heat wave (HW1, HW2). Right panel: same as left but for the Bowen ratio. Colours as in Fig. 1.
Left panel: boxplot of the ratio of observed gross primary productivity (GPP(obs)) to the climatology of GPP (GPP(BGC)) during the first and second period of the heat wave (HW1, HW2). Right panel: same as left but for the ER. Colours as in Fig. 1.
Figure A1 shows that during HW1 the latent heat flux at the MW and TW sites was reduced. During HW2, precipitation and temporarily increased water availability brought the latent heat flux back to levels observed during BGH for the woodland sites. At the temperate forest, however, the latent heat flux strongly increased, particularly during HW1. Increasingly reduced soil water and lower temperatures reduced the effect during HW2.
Figure A2 shows that carbon uptake was decreased at MW and TW during HW1 and similar to background conditions during HW2. At TF, the forest site, carbon uptake was increased. Respiration (Fig. A2b) was increased at all locations and during both heat wave periods.
This work utilised data from the OzFlux network which is supported by the
Australian Terrestrial Ecosystem Research Network (TERN;