## Introduction

Plant respiration (*R*) is a significant component of the global carbon cycle and contributes ca. 60 Pg of carbon (C) released into the atmosphere each year (Prentice *et al*. 2001), approximately seven times more than C released by fossil fuel burning (Canadell *et al*. 2007). At the individual plant level, 30–70% of CO_{2} fixed by daily photosynthesis is released by *R*, with leaf *R* contributing to ≈50% of whole-plant *R* under favourable conditions (Atkin, Scheurwater & Pons 2007). The close association between leaf *R* and environmental factors makes it crucial that environment-induced variability in *R* is properly incorporated into coupled climate and C-cycle models (Gifford 2003; Atkin *et al*. 2005a); here, describing the temperature (*T*) response of *R* of plants growing in thermally contrasting environments is of central importance.

Several simulation models (Aber & Federer 1992; Schimel *et al*. 1997) and global dynamic vegetation models (White, Cannell & Friend 2000; Cox 2001; Cramer *et al*. 2001) have assumed that *R* can be modelled assuming a constant *Q*_{10} of 2.0 (i.e. the *T*-sensitivity of *R* remains constant, whereby rates double per 10 °C rise). However, there is growing acceptance that this assumption is not correct (Wythers *et al*. 2005; King *et al*. 2006; Atkin *et al*. 2008). Rather, the shape of the *T* response is dynamic, reflecting the fact that the functional form of the short-term temperature response curve of *R* departs significantly from a simple exponential. Changes in growth *T* that last several days can also alter the short-term *Q*_{10} (Atkin *et al*. 2005a), with *Q*_{10} values often varying seasonally (Atkin, Holly & Ball 2000; Zaragoza-Castells *et al*. 2008); there is also evidence that *Q*_{10} values are lower in tissues where substrates and/or energy demand limit *R* (Atkin & Tjoelker 2003). Finally, it is well established that *Q*_{10} values decrease with short-term increases in measuring *T* (James 1953; Forward 1960; Tjoelker, Oleksyn & Reich 2001). Theory and empirical evidence suggests that the declining *T*-sensitivity of leaf *R* (i.e. declines in *Q*_{10} of *R*) as measurement *T*'s rise is linked to shifts in the control exerted by maximum enzyme activity at low *T*, substrate limitations at moderate-high *T*, and loss of enzyme function at very high *T*'s (Atkin & Tjoelker, 2003 and references cited therein). Failure to account for the fundamental non-exponential form of the *R*–*T* relationship leads to respiratory CO_{2} release being overestimated over long periods (Wythers *et al*. 2005; King *et al*. 2006; Atkin *et al*. 2008).

When considering how to fit curves to a non-exponential form of the *R*–*T* relationship, several alternative approaches are available that differ from a constant *Q*_{10}. Using the Arrhenius approach, *T*-mediated variations in leaf *R* can be modelled using a constant activation energy (*E*_{a}) (e.g. Turnbull *et al*. 2005); this approach accounts for slight declines in the *Q*_{10} with increasing *T* (Zaragoza-Castells *et al*. 2008). However, the Arrhenius approach assumes that respiration is substrate saturated – in many cases, this assumption is invalid for intact leaves (Atkin, Bruhn & Tjoelker 2005b). Moreover, application of a single *E*_{a} value does not account for marked *T*-dependent declines in *Q*_{10} often exhibited by plants in natural ecosystems (Atkin *et al*. 2000; Tjoelker *et al*. 2001) or the fact that exposure to very high *T*'s results in rates of leaf *R* decreasing (Atkin & Tjoelker 2003; Hüve *et al*. 2011, 2012). Thus, other approaches may be needed to account for the non-exponential form of the *R*–*T* relationship. One approach is to use a modified Arrhenius algorithm formulated by Lloyd & Taylor (1994), which was developed to model the *T*-dependence of soil *R* but has since been used to model rates of leaf *R* in a range of species (Turnbull *et al*. 2001; Griffin, Turnbull & Murthy 2002; Dillaway & Kruger 2011). Another approach is to assume that *Q*_{10} declines with increasing measurement *T* in a near-linear manner (Tjoelker *et al*. 2001). To compare model fits using these different approaches, it is essential that comparisons be made using high-resolution *R*–*T* plots (i.e. where rates of leaf *R* are continuously measured as leaf *T* increases). In addition to providing a basis for elucidating response functions over physiologically relevant *T* ranges, high-resolution *R*–*T* plots also provide an opportunity to assess responses to extreme *T*'s where damage to leaf functioning is irreversible.

Recently, a continuous heating protocol was developed that delivers high-resolution data on the short-term *T-*response of leaf *R* over a wide range of *T*'s (Hüve *et al*. 2011, 2012); using this method, leaf *R* is continuously recorded as leaves are heated at near 1 °C min^{−1}. From the resultant *R*–*T* curves, a ‘break-point’ *T* can be identified, where rates of *R* rapidly increase before reaching a maximum at *T*_{max}, after which rates rapidly decline; a similar phenomenon was observed earlier by Björkman (1975). The ‘break-point’ *T* corresponds to a similar pattern seen in the *T-*response of the steady state of minimal fluorescence of photosystem II (PSII) (*F*_{o}) in photosynthesis (Schreiber & Berry 1977; Havaux, Greppin & Strasser 1991; Knight & Ackerly 2002; Hüve *et al*. 2012). In studies assessing the *T*-response of *F*_{o}, a two-stage pattern is typically observed, whereby *F*_{o} rises slowly with increasing *T* until it reaches a critical temperature (*T*_{crit}), after which *F*_{o} rapidly increases (Knight & Ackerly 2002). As correlations have been identified between the *T*_{crit} of *F*_{o} and changes in the temperature response of *R* (Hüve *et al*. 2012), it seems likely that *T*_{crit} can be used as a proxy for the *T* at which disruption of both electron transport in PSII and respiratory biochemical machinery occurs in organelles. In *Populus tremula*, *T*_{crit} occurs at ca. 49 °C (Hüve *et al*. 2011), while various desert and costal Californian species exhibit *T*_{crit} at *c*. 42–54 °C (Knight & Ackerly 2002). The cause of changes associated with *T*_{crit} could include biochemical factors such as increases in cell membrane leakiness (Raison *et al*. 1980; Hazel 1995; Bukhov *et al*. 1999; Schrader *et al*. 2004), respiratory uncoupling and/or increased ATP demand (Skulachev 1998; Vacca *et al*. 2004; Hüve *et al*. 2011), or increased drought stress at high *T*'s, leading to changes in photosynthetic and respiratory metabolism (Lawlor & Cornic 2002; Rennenberg *et al*. 2006; Slot, Zaragoza-Castells & Atkin 2008; Atkin & Macherel 2009). Importantly, application of the high-resolution method of Hüve *et al*. (2012, 2012,2011) has been limited to studies on biochemical processes underpinning photosynthetic and respiratory *T*-responses in controlled-environment plants, with no study having applied the method to assess the non-exponential form of the *R*–*T* relationship or to elucidate ecological phenomena under field conditions.

Past studies considering the extent to which the short-term *T*-dependence of leaf *R* changes through time (e.g. acclimation to seasonal changes in the abiotic environment) and/or differs among plants growing in thermally contrasting sites have relied on low-resolution datasets [e.g. *R* is measured at wide *T* intervals (Bolstad, Mitchell & Vose 1999; Tjoelker *et al*. 2001; Bolstad, Reich & Lee 2003; Lee, Reich & Bolstad 2005; Atkin, Scheurwater & Pons 2006)] and/or via reliance on rates of leaf *R* measured at different diurnal time points (Atkin *et al*. 2000; Zaragoza-Castells *et al*. 2008; Rodriguez-Calcerrada *et al*. 2010; Crous *et al*. 2011). Such datasets provide insights into approximate patterns of acclimation [i.e. modulation of the long-term *T*-sensitivity of *R* in response to shifts in growth *T* (Larigauderie & Körner 1995; Atkin & Tjoelker 2003)], but do not elucidate the extent to which changes in growth *T* result in ‘fine-scale’ changes in the shape of the short-term *R*–*T* relationship of field-grown plants. Measuring high-resolution *T*-response functions would enable assessment of the stability or sensitivity of *T*_{crit} and *T*_{max} in plants in thermally contrasting environments.

The aims of our study were to (1) assess potential sources of variation in *T*_{crit} and *T*_{max} in individual plants of snow gum (*Eucalyptus pauciflora*) in thermally contrasting environments, particularly whether altitudinal and seasonal shifts in growth *T* alter the shape of the *T*-response curve of leaf *R*; (2) explore the extent to which sampling and leaf heating protocols [e.g. sample chilling and storage, leaf heating rate, duration of high-temperature exposure and relative humidity (RH)] influence the characterization and reproducibility of high *T*-response features (*T*_{crit}, *T*_{max}) of leaf *R*; and (3) assess the *R–T* relationship and its mathematical description using high-resolution data at subcritical *T*'s. Given that several simulation models (Aber & Federer 1992; Schimel *et al*. 1997) and global dynamic vegetation models (White *et al*. 2000; Cox 2001; Cramer *et al*. 2001) assume an exponential/near-exponential increase in leaf *R* with rising *T*, this final aim allowed us to compare and contrast the efficacy of exponential fits (using *Q*_{10} and Arrhenius approaches) with that of non-exponential forms of the *R*–*T* relationship.