On the 'temperature sensitivity' of soil respiration: Can we use the immeasurable to predict the unknown?

The temperature dependence of soil respiration (R(S)) is widely used as a key characteristic of soils or organic matter fractions within soils, and in the context of global climatic change is often applied to infer likely responses of R(S) to warmer future conditions. However, the way in which these temperature dependencies are calculated, interpreted and implemented in ecosystem models requires careful consideration of possible artefacts and assumptions. We argue that more conceptual clarity in the reported relationships is needed to obtain meaningful meta-analyses and better constrained parameters informing ecosystem models. Our critical assessment of common methodologies shows that it is impossible to measure actual temperature response of R(S), and that a range of confounding effects creates the observed apparent temperature relations reported in the literature. Thus, any measureable temperature response function will likely fail to predict effects of climate change on R(s). For improving our understanding of R(S) in changing environments we need a better integration of the relationships between substrate supply and the soil biota, and of their long-term responses to changes in abiotic soil conditions. This is best achieved by experiments combining isotopic techniques and ecosystem manipulations, which allow a disentangling of abiotic and biotic factors underlying the temperature response of soil CO(2) efflux.

Background

Soil CO2 efflux (or soil respiration, RS) is considered the largest source of CO2 from terrestrial ecosystems. Recent estimates indicate global soil CO2 emissions in the range of 98 ± 12 Pg y−1, with annual increases of 0.1 Pg that have been suggested to be temperature-associated (Bond-Lamberty and Thomson, 2010). At a global, regional and local scale, soil temperature (TS) and soil moisture have been considered the most important abiotic parameters determining RS and its underlying processes (Kutsch et al., 2009). Empirical response functions are commonly used to derive annual estimates of RS based on sporadic field measurements (e.g. Savage et al., 2008), whilst short-term (i.e. diurnal) deviations from an average abiotic response of RS have been interpreted as effects of photosynthesis on RS (Tang et al., 2005). Although temperature is undoubtedly one of the most important environmental factors affecting respiratory processes on a physiological scale, we argue that its direct influence on soil CO2 efflux can at best be approximated, which calls for more care in the interpretation and extrapolation of what is often assumed to be a TS–RS relationship. The response of RS to climate change is a critical component in predicting possible feedbacks between the global carbon cycle and the climate system, and simplistic temperature-based extrapolations will not advance our ability to forecast these changes (Davidson et al., 2006). In the following we demonstrate that several of the assumptions, on which the TS–RS relationship and its interpretation have often been based, are somewhat arbitrary and deserve careful reconsideration.

Incubation experiments – effects of substrate supply and depletion on the apparent temperature sensitivity of soil C turnover

Lab incubations of soil samples indicate generally consistent temperature response functions, illustrating the fact that in principle the decomposition process in homogeneous soils can be well described using soil temperature (e.g. Reichstein et al., 2005). Experimental warming of incubated soils has been found to lead only to a transient increase in soil CO2 production, with an apparent compensation for the increase in temperature by a reduction of temperature sensitivity (commonly expressed as Q10, representing the respiration rate change over a temperature shift by 10 °C). However, there is good evidence that such apparent thermal acclimation is caused by the depletion of substrate pools in the soil rather than an intrinsic ability of soils to “adapt” to changes in temperature conditions (Hartley and Ineson, 2008; Kirschbaum, 2004). The apparent acclimation does not however indicate per se that an intrinsic temperature sensitivity of RS is altered, as a range of environmental constraints to decomposition are temperature dependent in themselves and physico-chemical mechanisms of SOM stabilization and destabilization are confounded with the kinetic properties of substrates and enzymes (Davidson and Janssens, 2006). Furthermore, decomposition of more recalcitrant soil organic matter (SOM), whilst being of lower magnitude, may display a higher Q10 (Conant et al., 2008). It is therefore necessary to express soil CO2 efflux rates in warming experiments or lab incubation studies in relation to pool sizes of different substrate qualities. Furthermore, soil incubation experiments generally do not account for the fact that belowground carbon allocation (Litton et al., 2007) and its effects on root and rhizosphere respiration (Curiel Yuste et al., 2004) as well as priming of SOM decomposition (Fontaine et al., 2004; Kuzyakov, 2002) may alter soil C turnover and CO2 emissions at any given temperature.

Inherent problems related to in situ testing of temperature dependent and – independent effects on soil CO2 efflux

Whilst RS measurements in the field have the advantage of including all CO2 sources of intact soils (i.e. SOM decomposition as well as root and rhizospheric CO2 flux), the interpretation of annual or seasonal temperature relations requires some caution. Belowground C allocation in plants, which contributes around 40–60% of RS seasonally in most biomes (Subke et al., 2006), shows immense seasonal variation in the majority of ecosystems. Fig. 1 illustrates how the coincidence of peak rhizospheric CO2 flux with seasonal maxima in TS results in an apparently high TS–RS response, owing to increased plant C supply to the soil during summer (Fig. 1; see also Davidson et al., 2006; Reichstein and Beer, 2008).

Fig. 1

Heuristic example of seasonal soil CO2 efflux dynamics, based on simulated data representative of a temperate ecosystem setting with clear seasonality. (A) Seasonal flux contributions from heterotrophic decomposition (RH; solid black line), root derived CO2 (roots and rhizosphere; RR, dashed red line), and resulting total soil CO2 efflux (RS; dotted blue line). (B) Same monthly fluxes as in panel A, plotted against typical monthly temperatures, and showing exponential regression fits. RR flux dynamics are governed by plant productivity changes over the season, and cause a strong apparent temperature “response” of RS, with an excellent exponential fit (R2 = 0.95), but only a fraction of the flux response is directly influenced by temperature changes. For examples of actual field data, please see partitioning studies (e.g. Gaumont-Guay et al., 2008; Fig. 5 in Moyano et al., 2008) illustrating the same temperature response relations as described here. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

A further problem for identifying temperature-related and temperature-independent effects on RS in situ relates to the fact that in ecosystems TS is rarely constant across the soil profile (Fig. 2), and its change with depth may vary from diel to seasonal timescales. Accordingly, the choice of the soil depth used for inferring the temperature sensitivity of RS may strongly influence the shape of the temperature response curve, and thus Q10 (Fig. 3; Pavelka et al., 2007; Reichstein and Beer, 2008). It has been shown that commonly used temperature measurement depths in field experiments are likely to result in an underestimation of temperature sensitivity and that an arbitrary selection of a reference depth can produce an unrealistic range of Q10 values (Graf et al., 2008). Even the maximum R2 depth method, which helps identify a reference depth yielding a minimum bias, can only provide rough approximates, which may change if there are shifts in respiratory activity or diffusivity across the soil profile. Errors in apparent Q10 as related to temperature measurement depth are further increased by a pronounced and heterogeneous horizon of respiration activity, a low thermal and CO2 diffusivity of the soil and a low annual temperature amplitude (Graf et al., 2008).

Fig. 2

Abiotic and biotic changes throughout the soil profile. (A) Soil temperature (red lines; solid = mid-day, dashed = midnight) and moisture (blue dotted line). (B) Soil organic matter content (triangle width) and quality (shading indicates differences in complexity and molecular weight of carbon compounds). (C) CO2 production (white bars: root and rhizospheric sources, dark brown bars: heterotrophic sources, light brown bar: mineral weathering). (D) CO2 diffusion between different depths resulting from CO2 production. The image in panel C is reproduced with kind permission from USDA – Natural Resources Conservation Service, Lincoln, Nebraska. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Fig. 3

(A) Simulated temperature data exemplifying diurnal fluctuations at different soil depths (scaled on left–hand axis – see panel B for colour code of temperature depth) and simulated concurrent soil CO2 efflux (RS: green line and right-hand axis) calculated assuming a Q10 of 2 and prescribed distributions of organic matter in the soil profile. (B) Apparent temperature dependencies for the same data, showing calculated Q10 values. Note how the reduction in range and time shift in temperature dynamics with increasing soil depth cause an apparent increase in temperature sensitivity and the occurrence of hysteresis “loops”. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).

Also the assessment of ‘temperature-independent’ effects on a diel timescale from observed hysteresis in the TS–RS relationship (Fig. 3B; see also Liu et al., 2006; Tang et al., 2005; Vargas and Allen, 2008) faces the major drawback that it does not consider shifts in phase and amplitude in TS with soil depth, and may thus be confounded by an arbitrary selection of the soil depth at which temperature is measured and to which CO2 efflux is related (Bahn et al., 2008; Reichstein et al., 2005). Moreover, besides temperature, a range of further factors may strongly influence an apparent diurnal TS–RS relationship, or any deviation from it (compare also Fig. 2): 1) soil moisture and CO2 diffusivity at a single point in space and time, and their respective diurnal changes; 2) the vertical distribution of roots and microbes, their specific respiration rates and TS responses; 3) changes in the quality of SOM and its accessibility to microbes and enzymes across the soil profile. 4) Effects of fresh photoassimilates on root and rhizosphere respiration, incl. priming effects (see above), may potentially also cause deviations from a simple diurnal Ts – RS relationship. However, due to a range of likely confounded effects (see above) it is not possible to consistently infer such ‘temperature-independent’ effects of photosynthesis on RS. Conversely, changes in abiotic and biotic conditions across the soil profile may alter both the (immeasurable) actual and the (generally reported) apparent temperature response of soil respiration.

Outlook and conclusions

We conclude that any measureable temperature response function will likely fail to predict effects of climate change on Rs. For improving our understanding of RS in changing environments a shift in focus from simplistic abiotic response relationships to modelling of ecosystem processes including biotioc and abiotic interactions is needed. In particular, we require a considerable improvement in our understanding of assimilate allocation to belowground, phloem transport processes, assimilate storage dynamics in different plant organs, and exudation controls in the rhizosphere (Bahn et al., 2010). Ecosystem models have started to incorporate these more complex interactions, and consider issues of C allocation between plant and soil (Sitch et al., 2008 and citations therein), but significant challenges remain. Experimentally, a range of recent isotopic labelling experiments have provided critical new insights into soil C turnover processes (see e.g. review by Paterson et al., 2009). For future experimental work, we think that a combination of isotopic tracer studies with environmental manipulations (e.g. soil or ecosystem warming, throughfall-displacement, or pollution/deposition experiments) hold the best promise to elucidate C pathways and identifying specific mechanisms under changed environmental conditions by tracing C molecules from assimilation to their respiratory “use”. To be able to account for long-term effects of warming on soil carbon losses, such experiments should also consider the physico-chemical stabilization and destabilization of SOM fractions, as we urgently require an integration of slow and fast components of soil C turnover (Bahn et al., 2010) in order to obtain more realistic estimates of soil CO2 efflux in a warmer climate.