Tropical forests play a major role in regulating global carbon (C) fluxes and stocks, and even small changes to C cycling in this productive biome could dramatically affect atmospheric carbon dioxide (CO2) concentrations. Temperature is expected to increase over all land surfaces in the future, yet we have a surprisingly poor understanding of how tropical forests will respond to this significant climatic change. Here we present a contemporary synthesis of the existing data and what they suggest about how tropical forests will respond to increasing temperatures. Our goals were to: (i) determine whether there is enough evidence to support the conclusion that increased temperature will affect tropical forest C balance; (ii) if there is sufficient evidence, determine what direction this effect will take; and, (iii) establish what steps should to be taken to resolve the uncertainties surrounding tropical forest responses to increasing temperatures. We approach these questions from a mass-balance perspective and therefore focus primarily on the effects of temperature on inputs and outputs of C, spanning microbial- to ecosystem-scale responses. We found that, while there is the strong potential for temperature to affect processes related to C cycling and storage in tropical forests, a notable lack of data combined with the physical, biological and chemical diversity of the forests themselves make it difficult to resolve this issue with certainty. We suggest a variety of experimental approaches that could help elucidate how tropical forests will respond to warming, including large-scale in situ manipulation experiments, longer term field experiments, the incorporation of a range of scales in the investigation of warming effects (both spatial and temporal), as well as the inclusion of a diversity of tropical forest sites. Finally, we highlight areas of tropical forest research where notably few data are available, including temperature effects on: nutrient cycling, heterotrophic versus autotrophic respiration, thermal acclimation versus substrate limitation of plant and microbial communities, below-ground C allocation, species composition (plant and microbial), and the hydraulic architecture of roots. Whether or not tropical forests will become a source or a sink of C in a warmer world remains highly uncertain. Given the importance of these ecosystems to the global C budget, resolving this uncertainty is a primary research priority.
More than any other biome, tropical forests dominate the global carbon (C) flux between the biosphere and the atmosphere (Saugier, Roy & Mooney, 2001). They account for 55% of forest biomass (Pan et al., 2011) and one third of the world's soil C (Jobbagy & Jackson, 2000). Due to the size of tropical C fluxes and stocks, even a slight change in the uptake and storage of C in these ecosystems could have substantial consequences for the global C cycle and present large feedbacks to future climate [e.g. efflux of 1% of the tropical soil C pool ≅ total annual anthropogenic carbon dioxide (CO2) emissions; IPCC (2007)]. Despite the importance of tropical forests to global C cycling, we have little understanding of how this vast amount of C will respond to future environmental change. Consistently warm temperatures and low intra- and inter-annual variability in temperature have led many researchers to focus their efforts on changes in precipitation as the primary driver of climate-change effects. New evidence challenges this paradigm, suggesting the potential for a strong response of tropical forests to subtle increases in temperature (Lewis et al., 2009a). Although controversial, these findings have reopened the debate as to the importance of temperature in tropical forested ecosystems (Corlett, 2011; Reed, Wood & Cavaleri, 2012).
Climatic conditions in the tropics are more diverse than anywhere else in the world (Brown & Lugo, 1982). According to the Holdridge Life Zone System, 66 of the world's 116 life zones are found in the tropics, 30 of which support forests ranging from lowland dry deciduous to montane wet evergreen (Holdridge, 1967; Brown & Lugo, 1982). Across tropical forests, mean annual temperature (MAT) can range from 10 to 28 °C, and mean annual precipitation (MAP) from 1000 to 6000 mm (Raich et al., 2006; FAO, 2009). These differences in climate, as well as in other factors such as soil type, have led to an incredible level of diversity in community composition and forest structure, both within and among tropical forest sites. As such, tropical forests are unlikely to exhibit a single response trajectory to a changing climate.
Global models project mean temperature increases in the tropics that are either in line with the global mean (1.7–3.9 °C) or warmer (1.8–5.0 °C) (A1B Scenario; Christensen et al., 2007). In contrast to the common perception that high-latitude areas face the most accelerated warming, recent work suggests that tropical systems are likely to shift to an unprecedented heat regime within the next two decades, where historical temperature extremes will become the norm (Christensen et al., 2007; Anderson, 2011; Diffenbaugh & Scherer, 2011). These recent predictions place tropical systems as top-priority research biomes for ecosystem-scale climate change research (Luo et al., 2011).
Here, we evaluate available data to ask three questions. First, is there enough evidence to support the conclusion that increased temperature will affect the tropical forest C balance? Second, if there is sufficient evidence, what direction will this effect take? Third, what steps can we take to resolve uncertainty surrounding tropical forest responses to increasing temperature? Given the broad scope of this topic, we place certain constraints on this review. First, we use a mass-balance approach by focusing primarily on the effects of temperature on inputs and outputs of C on hourly to decadal time scales and spanning microbial- to ecosystem-scale responses. Next, due to their predominance in the tropics (i.e. 88% of all tropical forests; FAO, 2009), we concentrate on the responses of lowland tropical forests, which are at the warm end of the temperature spectrum with MATs >20 °C. In addition, temperature is likely to interact with other abiotic variables such as precipitation, atmospheric CO2 concentration and light availability. Understanding the effects of multiple drivers simultaneously is clearly important; however, we currently lack the data to do so effectively. Therefore, we focus on temperature as an individual driver, independent of other abiotic variables. Finally, we limit the review to intact forests and exclude discussion of potential longer term shifts in forest community composition. Although both disturbance and changes in species composition are likely to be highly influential in determining tropical ecosystem C balance (Laurance, 1999; Enquist, 2002; Phillips et al., 2002; Wright, 2005), a thorough examination of these topics is beyond the scope of this review.
II. ECOSYSTEM C BALANCE
Several approaches are used to evaluate the potential effects of warming on whole-ecosystem C balance, including: elevation gradients, cross-site comparisons, eddy covariance, coupled C-climate models, and field warming experiments. We explore current predictions of the tropical forest C balance response to increased temperature using these methodological approaches.
(1) Elevation gradients and cross-site comparisons
Temperature gradients within the tropics provide a valuable opportunity to assess the potential response of tropical forests to increased temperature by substituting space for time. Meta-analyses of the effects of temperature on C storage and fluxes across a broad range of tropical forest sites have found that total net primary production (NPP), litter production, tree growth, and below-ground C allocation all increase with increasing MAT (range = 10–28 °C; Silver, 1998; Raich et al., 2006) and the temperature to precipitation ratio (Brown & Lugo, 1982). However, soil C decomposition and turnover time also increase with increasing MAT, indicating that atmospheric C uptake via increased forest productivity could be offset by increased soil C loss with warming (Raich et al., 2006). Although cross-site comparisons can provide a window into how tropical forest C balance might respond to increased temperature, they are not without limitations. Other environmental variables in addition to temperature vary along elevation gradients and among forest sites, including light availability and plant community composition, confounding the elucidation of single drivers and relationships. Cross-site comparisons also do not account for the consequences of perturbing a system away from its ‘native’ temperature. For example, a forest with a MAT of 24 °C is likely not comparable to a forest with a MAT of 20 °C after increasing temperature by 4 °C. Tropical communities develop over evolutionary time scales, and the rapid rate at which climate change is occurring may not enable sufficient time for these communities to adapt. Given that most tropical forests may be moving beyond temperatures historically experienced by these systems (Diffenbaugh & Scherer, 2011), there is high potential for tropical forests to surpass temperature thresholds. A temperature threshold/thermal optimum is defined as the maximum temperature for a given reaction or process, beyond which rates decline.
Tropical forests that maintain high MATs may be limited by factors other than temperature, such as nutrient availability (Townsend et al., 2011). For example, if we restrict the meta-analyses by Raich et al. (2006) to sites with temperatures >20 °C, the relationships between temperature and tropical forest C stocks are no longer significant. Hence, while analyses of the C balance along temperature gradients within the tropics suggest that temperature may have little effect on the net C balance of tropical forests (Raich et al., 2006), we cannot say with certainty that these trends will continue as the world continues to warm.
(2) Eddy covariance
Tower-based eddy covariance techniques can be used to correlate net ecosystem exchange (NEE) of CO2 with temperature over large tracts of forest (500–100 ha). Eddy covariance data from the Brazilian Amazon combined with a simple gas-exchange model suggest that NEE is quite sensitive to temperature increases, and that the forest may switch from a C sink to a source with only a 1.2 °C increase in MAT (Grace et al., 1996). An Amazonian eddy covariance dataset over 1 year showed an overall trend of C source behaviour at temperatures above approximately 27 °C (Doughty & Goulden, 2008), and a 3-year dataset from a rain forest in Costa Rica showed reduced C sink behaviour at air temperatures above 20 °C (Loescher et al., 2003). Interestingly, a longer term (4 years) eddy covariance study in the Amazon found the strongest controls of the components of NEE to be phenology and light, rather than temperature (Hutyra et al., 2007). Hence longer term data sets (3+ years) may be needed to predict accurately the consequences of warming for net C exchange in tropical forests.
(3) Coupled C-climate models
Models enable us to project the potential response of tropical forests to increased temperature over the next century. In contrast to most of the results discussed above, numerous atmosphere-biosphere modeling studies suggest that warmer tropical forests will likely become an increasing source of C to the atmosphere as a result of increased heterotrophic respiration (Cramer et al., 2001), increased plant respiration (White, Cannell & Friend, 2000), decreased NPP or photosynthesis (White et al., 2000; Cramer et al., 2001), and/or forest dieback (White et al., 2000). While there is general agreement among models that tropical forests are likely to become a net source of C as temperatures increase, they disagree as to which variables are likely to control this response. Models are only as good as the available input data, however, and the lack of a mechanistic understanding of how tropical forests will respond to warming significantly constrains these endeavours.
(4) Field warming experiments
Field warming of both plants and soils provides a concomitant estimate of above- and below-ground responses to increased temperature, enabling a comprehensive analysis of the ecosystem warming response. Although a valuable tool for predicting temperature effects on ecosystem C balance, a large-scale warming experiment that incorporates both soils and the forest canopy does not currently exist for any forested ecosystem, due in large part to the challenges and expense of warming such a large area (Aronson & McNulty, 2009; Amthor et al., 2010; Kimball, Conley & Lewin, 2011). Field experiments that measure temperature responses of individual components (e.g. soils) have been useful for evaluating the response of temperate forests to warming (e.g. Farnsworth et al., 1995; Nakamura et al., 2010; Melillo et al., 2011); however, data from such experiments have yet to be published for the tropics.
(5) Synthesis: temperature effects on net C balance
Cross-site comparisons suggest that, over evolutionary time scales, an increase in forest NPP due to increased temperature will offset increased soil C loss, with no net change in the C balance (Raich et al., 2006). Coupled C-climate models and eddy covariance studies predict that tropical forests will become a net source of C; however, the models do not always agree as to which variables will drive the increased C loss, and eddy covariance results vary among studies. Resolving the potential effects of temperature on the net C balance of tropical forests will likely require an improved understanding of temperature controls on the flow of C into and out of the system over multiple time scales (Fig. 1). Carbon fluxes include, but are not limited to, temperature effects on net CO2 uptake and release by plants, biomass turnover rates, biomass accrual, soil C inputs, decay rates, and mobilization of C stocks. Field warming experiments in the tropics would complement existing methodologies and would be incredibly valuable in elucidating these individual fluxes.
III. ABOVE GROUND
(1) Growth and turnover rates
Over the last several decades, long-term (5+ years) inventory plots in tropical forests around the globe have revealed contrasting trajectories of stand-level growth and biomass turnover. Datasets in tropical Africa and the Amazon have shown increasing growth rates (Phillips et al., 2004; Lewis et al., 2009b), while plots in Costa Rica, Panama, and Malaysia revealed decelerating growth (Clark et al., 2003; Feeley et al., 2007). A recent meta-analysis of large (16–52 ha), long-term tropical forest inventory plots across three continents (America, Africa and Asia) showed an increase in biomass over 20 years in seven of ten plots analyzed (Chave et al., 2008). Both tree recruitment and mortality have also increased in Amazonian forests, with recruitment outpacing mortality (Phillips et al., 2004). Possible drivers of increasing biomass increment include: CO2 fertilization (Lloyd & Farquhar, 2008; Lewis et al., 2009a), recovery from past disturbance (Chave et al., 2008), or a shift in community composition towards faster growing species (Laurance et al., 2004). Observed growth declines have been attributed to temperature-induced increases in plant respiration rates (Clark et al., 2003; Feeley et al., 2007); decreased net photosynthesis as a result of increasing temperature beyond the thermal optimum (Doughty & Goulden, 2008); or light limitation from increased liana shading of canopy trees, or from global dimming (Feeley et al., 2007). Further evidence of a temperature effect on decreasing productivity includes increased mortality rates in several forests around the world during strong El Niño Southern Oscillation (ENSO) years, which exhibit higher temperatures and lower rainfall than normal years (Clark, 2004, and sources within). It is difficult to tease apart effects of temperature and precipitation when correlating growth to the large-scale climatic effects of ENSO, although temperature increases will likely coincide with lowered water availability through enhanced evapotranspiration rates. In a recent review, Clark (2004) focused on temperature increase as a major driver of tropical forest change, and suggested that as global temperatures increase tropical rain forests will likely show decreased productivity and increased mortality rates, thus shifting to a net C source.
(2) Individual tree growth and morphology
Tree ring analysis can be used to relate historical annual diameter growth to climatic variability. However, this methodology has limited application in tropical forests because few tropical trees develop annual growth rings, and even when present, tropical tree rings are usually the result of strong drought or dry season, making them difficult to interpret. Nevertheless, a 37-year tree ring record of a fast growing secondary forest tree species common to the dry forests of Mexico suggested that, although tree growth was related to temperature, dry-season rainfall was the primary driver of tree growth in this forest (Brienen et al., 2010). This is not surprising given the degree of water limitation in this system, and highlights the difficulty in teasing apart individual climatic drivers with this technique.
Patterns of plant growth depend on the partitioning of photosynthate, but growth cannot always be predicted by rates of photosynthesis. Tropical forest species in Australia showed optimum growth under much higher temperatures than what was optimal for photosynthesis, indicating that growth can continue to increase even as photosynthesis declines with increasing temperature (Cunningham & Read, 2003a). Conversely, increased photosynthesis as a result of elevated CO2 in a mature tropical rain forest in Panama did not increase growth, but instead increased levels of non-structural carbohydrates in canopy tissue (Körner, 2003). Because source/sink feedbacks may have more control over tree C balance than either photosynthesis or respiration, a more mechanistic understanding of the response of tropical tree growth to increasing temperature should include investigating the effects of temperature on cell division and expansion (Ryan, 2010).
A recent global meta-analysis of the effects of temperature on tree growth found warming to increase foliage biomass, decrease root biomass, and produce taller, thinner stems, although tropical tree growth decreased overall when compared to either boreal or temperate species (Way & Oren, 2010). By contrast, a study of tropical tree seedlings in Australia showed increased root:shoot ratios under warming treatments (Cunningham & Read, 2003a). Warming may therefore alter the morphology and biomass allocation patterns of trees in addition to affecting diameter growth rate. As such, long-term studies (5+ years) in tropical forests showing growth patterns based solely on diameter may be misleading with respect to total biomass increment. Future studies in tropical forests should take into account the possible effects of warming on tree morphology and above- versus below-ground C allocation.
(3) Whole-plant water use and hydraulic architecture
Warming has the potential to alter the water balance of tropical forests, either by increasing or decreasing whole-plant transpiration, or by conferring competitive advantages to species with particular hydraulic architecture. In a Costa Rican tropical rain forest, sap flow was better correlated to vapour pressure deficit (VPD) than to soil moisture (O’Brien, Oberbauer & Clark, 2004), suggesting that transpiration may be more sensitive to changes in air temperature than to precipitation when soils are consistently moist (but see Williams et al., 1998).
Tropical tree sap flow generally increases with increasing temperature (i.e. evaporative demand) up to a threshold, and then decreases as stomata close to maintain leaf water potential above a critical threshold (O’Brien et al., 2004). Whole-canopy transpiration, however, depends upon both stomatal conductance and boundary layer conductance (Meinzer et al., 1997). While stomatal conductance is primarily influenced by temperature, humidity, and solar radiation, boundary layer conductance is primarily driven by wind speed and leaf morphology (Meinzer et al., 1997). As a result, global warming may affect stomatal conductance, but have little effect on boundary layer conductance, which may have greater influence than stomatal behaviour on whole-tree transpiration in tropical forest canopies (Meinzer et al., 1997).
Hydraulic architecture may also have a strong effect on how species react to global climate change. Trees with shallow roots may be at a competitive disadvantage compared to deeper rooted species (Stork et al., 2007). However, in a Panamanian tropical forest, Meinzer et al. (1999) found that smaller trees tapped deeper water than larger trees. In Amazonian forest, deep-rooted trees passively transfer water at night from deep moist soil to shallow dry soil, or from moist shallow to dry deep soil layers, a phenomenon called ‘hydraulic redistribution’ (Oliveira et al., 2005). If many tropical forests exhibit this redistribution behaviour, the effects of increased evaporative demand with warming may be buffered for shallow-rooted species.
Variation in xylem architecture may also affect species' ability to adapt to warmer temperatures. Lianas (woody vines) can transpire more water than trees of the same diameter, but are more susceptible to drought-induced embolisms (Fisher & Ewers, 1995). In the Neotropics, lianas have increased in abundance over the last two decades, possibly as a result of increasing atmospheric CO2 (Phillips et al., 2002; Schnitzer & Bongers, 2011), which could mean that forests as a whole may become more susceptible to greater water stress with the increasing dominance of lianas. On the other hand, high liana cover has been found to buffer host tree sap flow responses to climatic extremes in a Costa Rican rain forest (O’Brien et al., 2004). The diversity of root architecture, soil water partitioning, and hydraulic architecture are not well explored in tropical forests, and would be important baseline data for our understanding of how water balance may be affected by the increased evaporative demand that would come hand-in-hand with increased temperatures.
(4) Photosynthesis and stomatal conductance
Photosynthesis increases with increasing measurement temperature (temperature of air, soil, or plant material measured at the time of sampling) up to a thermal optimum, then decreases (Berry & Bjorkman, 1980). Thermal optima greatly depend upon genetic factors and growth temperatures (temperature at which plants/microbes are grown or incubated), with tropical tree species showing sharp declines in leaf-level photosynthesis between measurement temperatures of 33 and 40 °C (Fig. 2; Koch, Amthor & Goulden, 1994; Lerdau & Keller, 1997; Doughty & Goulden, 2008; Doughty, 2011). Midday depression of photosynthesis has been measured in tropical forest tree canopies at the leaf-level (Koch et al., 1994), coinciding with the hottest part of the day and the highest leaf-to-air VPD. The direct cause of this midday depression of net photosynthesis is often difficult to determine because the effects of elevated temperature on photosynthesis can be broken down into two primary driving mechanisms: mesophyll (direct) and stomatal (indirect) effects (Lloyd & Farquhar, 2008).
Many biochemical processes in the mesophyll are directly affected by increasing temperature (Sage & Kubien, 2007). Under historical atmospheric CO2 concentrations (<380 ppm) and above a thermal optimum for plants (>35 °C), the primary limiting factor for photosynthesis is a reduction in Rubisco's carboxylation activity (Farquhar, Von Caemmerer & Berry, 1980). Under current and future CO2 concentrations (>380 ppm), however, the primary limitations on C3 photosynthesis above a thermal optimum are electron transport rate and Rubisco activase function (Sage & Kubien, 2007). Electron transport rates of tropical species decline between approximately 35 and 40 °C, but this process is largely reversible (Sage & Kubien, 2007; Lloyd & Farquhar, 2008). The irreversible denaturation of Rubisco activase (a key enzyme that regulates the function of Rubisco) may be a primary cause of photosynthesis limitation above approximately 42 °C (Sage & Kubien, 2007). The biochemical acclimation potential of photosynthesis under warmer growth temperatures has been associated with either an increase in heat stability of Rubisco activase or an increased capacity for electron transport (Berry & Bjorkman, 1980; Sage & Kubien, 2007). Biochemical acclimation is a specific form of thermal acclimation whereby plants or microbes equilibrate to new measurement temperatures via biochemical adjustments such as enzyme or membrane stabilization. We define thermal acclimation as the adjustment of the rate of a plant or microbial process in response to a new growth temperature. Mechanisms may include: biochemical acclimation, substrate limitation, and/or a change in microbial community structure.
As temperature increases, so too does VPD, which indirectly decreases the rate of photosynthesis by inducing stomatal closure to avoid water stress (Lloyd & Farquhar, 2008). Teasing apart mesophyll versus stomatal effects can be challenging (Lewis, Malhi & Phillips, 2004; Doughty & Goulden, 2008). A simple leaf model using known temperature sensitivities of tropical species showed reductions in photosynthesis above thermal optima to be almost entirely due to indirect stomatal responses (30%), as opposed to direct biochemical effects on photosynthetic machinery (2%) (Lloyd & Farquhar, 2008). On the other hand, a recent canopy leaf warming experiment in the Amazon found reductions in photosynthetic capacity to be primarily the result of irreversible damage to photosynthetic machinery, rather than CO2 limitation due to stomatal closure (Doughty, 2011).
Leaf-level gas-exchange studies in Australia have shown that tropical species have lower photosynthetic thermal acclimation potential than temperate species, suggesting that tropical species may be more susceptible to climate change because they develop under lower seasonal and day-to-day temperature variation (Cunningham & Read, 2003b). This theory is further supported by in situ leaf warming experiments in the Amazon showing no evidence of photosynthetic thermal acclimation in response to 13 weeks of 2–5 °C warming of existing leaves (Doughty, 2011). Given the recent evidence, tropical plants appear to be unable to thermally acclimate their photosynthetic machinery, suggesting a decrease in overall forest C sink potential with global warming. It is unclear, however, whether or not plants will be able to thermally acclimate under naturally warming conditions, which occur more slowly than experimental warming conditions.
The eddy covariance datasets mentioned previously found that temperature thresholds of canopy-level photosynthesis were substantially lower (approximately 27 °C) than temperature thresholds of leaf-level photosynthesis (approximately 33 °C, Fig. 2), suggesting an important disconnect between leaf-level and whole-canopy level C assimilation sensitivities to temperature. Highlighting this phenomenon, a recent study in Amazonia found approximately 10 °C difference in the temperature thresholds of leaf-level versus canopy-level photosynthesis (Doughty & Goulden, 2008). The disconnect was attributed to vertical heterogeneity of warming within the canopy, where sunlit leaves can warm to well above bulk canopy air temperature, while shaded leaves remain relatively cool (Doughty & Goulden, 2008). The warmed, well-lit leaves can exceed leaf temperature thresholds for photosynthesis, causing overall decreases in C gain. Because these upper leaves contribute disproportionately more to whole-canopy C flux, the net effect is a reduction in canopy CO2 exchange even when the bulk air temperature is well below leaf temperature thresholds for photosynthesis (Fig. 2; Doughty & Goulden, 2008).
Doughty & Goulden (2008) argue that their results support the theory that Amazonian tropical forests are currently near a high-temperature threshold. Further evidence of this theory is found in palaeoclimate and palaeoecological studies which find that the Amazon is presently experiencing climate outside its historic range of variability (Maslin et al., 2005), and even moderate warming has the potential to push the system beyond a critical threshold and induce positive warming feedbacks (Cowling et al., 2004). However, Lloyd & Farquhar (2008) dispute this theory using a leaf-level model of photosynthesis to show that tropical forests are in fact well below their temperature threshold. These interpretations may not be mutually exclusive, however, because Lloyd & Farquhar (2008) modeled leaf-level photosynthesis and leaf temperature, not canopy-level photosynthesis and air temperature. If Doughty & Goulden (2008) are correct in their assessment of leaf versus canopy disconnects, then bulk air temperature may in fact be well below the leaf-level photosynthesis temperature threshold and still have a negative effect on C assimilation for the entire canopy (Fig. 2).
Foliar respiration increases exponentially with measurement temperature in the short term (minutes to hours), with overall mean foliar Q10 values ranging from 1.8 to 2.3 for tropical forest trees (Fig. 2; Meir, Grace & Miranda, 2001; Cavaleri, Oberbauer & Ryan, 2008). However, these short-term responses may not be applicable to predicting long-term temperature responses (days to years), because respiratory thermal acclimation may be primarily driven by substrate supply, which is controlled by net photosynthesis (Saxe et al., 2001). In a cross-biome analysis, Saxe et al. (2001) found that the long-term effects of temperature are ultimately driven by the thermal acclimation of photosynthesis rather than respiration. However, a recent meta-analysis across global functional types and biomes showed that leaf-level measurements of photosynthesis did not acclimate to temperature, while leaf-level respiration did (Way & Oren, 2010). This study, together with the limited ability of tropical photosynthesis to acclimate thermally (Cunningham & Read, 2003b; Doughty, 2011), suggests that tropical species may not be able to maintain positive C balance under warmer conditions. To date, there have been no investigations of the long-term effects of warming on tropical tree respiration.
(6) Biogenic volatile organic compound emission
Biogenic volatile organic compounds (BVOCs), especially isoprene, have been found to diffuse from stomata of 30–50% of tested tropical tree species throughout the world (Lerdau & Keller, 1997; Harley et al., 2004). Tropical forest isoprene emission is the most important individual source of photochemically reactive compounds in the atmosphere globally, and isoprene emissions increase exponentially with increasing measurement temperatures at the leaf level in the short term (minutes to hours) (Fig. 2; Lerdau & Keller, 1997). Above approximately 38 °C, isoprene emission from tropical forests may be a significant source of assimilated C lost to the atmosphere (Clark, 2004; Harley et al., 2004). One of the primary functions of foliar isoprene is to increase thermo-tolerance of photosynthetic machinery (Sharkey, Wiberley & Donohue, 2008); however, isoprene may not help plants tolerate long-term sustained warming (days to years) (Sharkey et al., 2008). Like foliar respiration, thermal acclimation potential of BVOC emission is likely directly linked to net photosynthetic uptake over the long term, although this has not been investigated in tropical forests.
(7) Synthesis: above-ground processes and drivers
The apparent effects of warming on above-ground tropical forest tree physiology differ depending upon the spatial and temporal scale of investigation. At the stand level, the diversity of both growth response and suggested climatic drivers may indicate that there is no single response of tropical forests to increasing temperatures, and conclusions vary among locations due to differing measurement techniques. If above- versus below-ground C allocation, whole-tree morphology, and source-sink feedbacks can be affected by warming, then the standard diameter at breast height (DBH)-based forest inventory methods may not be sufficient to detect effects of changing temperatures. A more holistic approach to stand-level measurements may be warranted, including a combination of DBH and height measurements with more ecophysiological mass-balance approaches of measuring C allocation, both above and below ground.
A better understanding of both leaf-level and canopy-level responses to warming also warrants a multi-scale approach. The disconnects between stomatal and whole-canopy conductance, and between leaf- and canopy-level photosynthesis, indicate that experiments which combine the two approaches can yield emergent patterns that may not be seen with either approach alone. Photosynthetic temperature thresholds, for example, should be investigated both in controlled chamber experiments and whole-canopy flux experiments. When investigating canopy-scale fluxes, temporal scale becomes important as well. For example, more and more long-term eddy flux studies (3+ years) are revealing inter-annual patterns that were not detected by shorter term investigations. A multi-scale approach will be crucial for the success of modeling efforts.
While it is important to understand individual drivers of climate change (e.g. warming), often it is impossible to tease apart the effects of increased temperature from the effects of water stress. The diversity of root architecture, soil water partitioning, and hydraulic architecture are not well explored in tropical forests, and would be important baseline data for our understanding of how water balance may be affected by the increased evaporative demand that would come hand-in-hand with increased temperatures. At the leaf level, experiments investigating the relative importance of both stomatal closure due to water stress and direct temperature effects on mesophyll processes will help us gain a better understanding of the relative importance of warming versus water stress on canopies, especially with respect to thermal acclimation potential of photosynthesis, which likely controls long-term thermal acclimation of both respiration and isoprene emission (Fig. 1).
IV. BELOW GROUND
(1) Litter and root production
Temperature-induced changes in forest productivity, below-ground C allocation, and stress-induced leaf drop could all contribute to increased inputs of C into soils. While cross-site comparisons suggest a significant positive effect of increased temperature on below-ground C allocation and litterfall production (Silver, 1998; Raich et al., 2006), we found no conclusive data within individual tropical forest sites to support a strong positive effect of temperature on soil C inputs, with the exception of a 7-year study of root dynamics that shows a relatively weak correlation between MAT and dead root biomass in a wet tropical forest in Costa Rica (Espeleta & Clark, 2007). Given the low variability in temperature in tropical ecosystems, longer time series (10+ years) are likely needed to reveal relationships between litterfall, root production and temperature in tropical forests (Clark & Clark, 2011), nevertheless, these datasets are rare.
In turn, an increase in fresh litter inputs and deep rooting could also stimulate additional CO2 loss from soils (Kuzyakov, 2010). In seasonal forests in Brazil, both the mass and distribution of litter and roots have been shown to explain a large degree of the spatial variability in soil respiration (Sotta et al., 2006; Metcalfe et al., 2007). In addition, a twofold litter addition experiment in a seasonal forest in Panama significantly increased soil respiration rates, which suggests a priming of recalcitrant C decomposition in response to the additional fresh litter inputs (Sayer, Powers & Tanner, 2007). Despite the potential for priming to significantly affect C storage and fluxes in tropical soils, the importance of this mechanism for the overall C budget remains uncertain (Subke, Inglima & Cotrufo, 2006; Kuzyakov, 2010). The high spatial and temporal variability in litter and root inputs at the small scale (e.g. millimeters) could create ‘hot spots' of elevated soil respiration that are difficult to measure at large scales (e.g. hectares) (Kuzyakov, 2010). Variability in litter inputs could additionally affect the rate of C stabilization due to indirect effects of litter on the soil environment, influencing factors such as soil pH, soil moisture, soil organic matter (SOM) chemistry and nutrient availability (Sayer, 2006; but see Wood & Lawrence, 2008). Overall, increased C inputs may affect a range of ecological processes related to C cycling and storage, and resolving how litter and root production will respond to elevated temperatures is necessary to predict accurately the effects of warming on C dynamics in tropical forests.
(2) Heterotrophic soil respiration
Kinetic theory states that reaction rates increase with increasing temperature up to a thermal optimum, at which point either microbial function or substrate availability become limiting, and reaction rates decline (Fig. 3A; Davidson & Janssens, 2006). We might therefore expect temperature to limit decomposition rates in cooler climates and C availability and/or the surpassing of microbial thermal optima to limit decomposition in warmer climates (Fig. 3A; Davidson & Janssens, 2006). There are currently no published in situ soil-warming experiments in the tropics; however, experimental warming in temperate and high-latitude ecosystems find that soil respiration initially increases in response to soil warming, but typically returns to control values within a few years of treatment (Fig. 3B; Oechel et al., 2000; Luo et al., 2001; Kirschbaum, 2004). Various theories have been put forth to explain this observed thermal acclimation of soil respiration (Kirschbaum, 2000; Davidson & Janssens, 2006; Bradford et al., 2008; Kleber, 2010), yet how these theories apply to tropical forested ecosystems remains highly uncertain. Soil incubation experiments suggest that tropical microbes have thermal optima higher than their cooler climate counterparts (Balser & Wixon, 2009), and temperature thresholds are not readily observed in short-term incubations (Holland et al., 2000). In particular, laboratory experiments using tropical soils have revealed elevated soil respiration rates at growth temperatures as high as 55 °C, well beyond the predicted temperature increases for tropical regions (Holland et al., 2000; Balser & Wixon, 2009). These data suggest it is unlikely that increased temperature will push microbial respiration past its biochemical thermal optimum, and the size and composition of the soil C pool (i.e. substrate availability) may be more likely to constrain the temperature sensitivity of decomposition in tropical forests.
There is large disagreement as to which C pool characteristics control the temperature sensitivity of soil C decomposition (Davidson & Janssens, 2006; Kirschbaum, 2006; von Lützow & Kogel-Knabner, 2009). One hypothesis is that soil C loss will be constrained by the size of the labile C pool, and that once this pool is depleted respiration rates will slow and ultimately be maintained by new soil C inputs (Fig. 3B; Giardina & Ryan, 2000). Others argue that, because recalcitrant C is more sensitive to increased temperature than labile C, we could expect a more substantial temperature-induced loss of soil C in the long term (decades) than previously predicted (Fig. 3B; Davidson & Janssens, 2006; Conant et al., 2008). There is also evidence that labile and recalcitrant C do not differ in their sensitivity to temperature (Fang et al., 2005). Taken together, tropical soil incubation experiments suggest that soil C loss will increase with temperature (Holland et al., 2000; Knorr et al., 2005); however, the magnitude and scale (spatial and temporal) of this loss remains unknown. Given that tropical forests store more soil C than any other terrestrial biome—containing 40% more soil C than temperate and boreal forests combined (Jobbagy & Jackson, 2000)—any increase in decomposition rates in these systems is likely to affect the global C balance.
In addition to substrate limitation, soil respiration could be affected by changes in the composition and/or overall function of the microbial community (i.e. thermal acclimation; Bradford et al., 2008; Balser & Wixon, 2009). A cross-site comparison of three forests (taiga, temperate and tropical) found that despite similar microbial biomass, taiga respiration rates were significantly lower than tropical forest rates when soils were incubated at the same temperature, suggesting inherent differences in basal metabolic activity of the microbial communities (Balser & Wixon, 2009). Additionally, microbes may respond to increased temperature by altering C allocation (Allison, Wallenstein & Bradford, 2010). Microbes release enzymes that break down SOM. The acquired resources can then be used to produce more enzymes, increase microbial biomass, and/or to maintain microbial activity (Allison et al., 2010). Shifting C allocation among these pathways can affect the fate of soil C and the rate of microbial respiration, consequently determining the loss of C from soil (Allison et al., 2010). Due to the amount of C cycled by soil microbes, a change in microbial C use efficiency (the proportion of microbially processed C respired as CO2versus the proportion allocated to microbial biomass) could have significant effects on C stored and lost from tropical forests. Finally, microbes could acclimate biochemically to increased temperature (e.g. produce enzymes with higher temperature thresholds), thereby showing an initial temperature-induced increase in respiration, followed by a rapid return to initial rates (Fig. 3B; Bradford et al., 2008). If soil microbial biochemical acclimation proves common across tropical ecosystems, it could significantly reduce the potential stimulatory effect of increased temperature on decomposition rates. At this time however, the response of tropical forest microbial community composition and function to increased temperature remains untested outside of the laboratory.
(3) Abiotic effects of temperature on SOC
In addition to biotic controls on soil organic C (SOC), temperature could also affect abiotic processes related to C availability in a variety of ways. Higher temperatures could decrease the sorption of SOC to the mineral surface, resulting in an increase in C availability (von Lützow et al., 2006). Conversely, higher temperatures may also increase SOC protection from decomposition, which would reduce C availability (Thornley & Cannell, 2001). Temperature may further influence C dynamics by increasing the diffusion of C substrates into the soil-water phase. This increased diffusion could increase microbial access to SOC and thereby increase microbial decomposition (Davidson, Janssens & Luo, 2006). However, soils also tend to become drier as temperature increases, and lower soil water content could restrict microbial access to these compounds and thus increased temperature could also reduce microbial decomposition via indirect effects of temperature on soil water availability (Grant & Rochette, 1994; Davidson et al., 2006). Finally, increased temperature may also influence soil pH (e.g. increasing concentrations of dissolved gases in soil pore water which alters the pH). Soil pH is known to influence enzyme activation energies and soil microbial community composition, and consequently SOC decomposition rates (Kleber, 2010). Currently, little is known about temperature effects on these processes in tropical forests, nor their potential to affect SOC decomposition in a warmer world (Davidson & Janssens, 2006).
(4) Autotrophic soil respiration
Root respiration (autotrophic respiration) is estimated to account for 24–38% of total soil respiration in some tropical forests (Silver et al., 2005; Sayer & Tanner, 2010), and litter and root biomass can explain as much as 44% of observed spatial variability in tropical forest soil respiration rates (Metcalfe et al., 2007). Moreover, differences in the relative time scale and response magnitude of autotrophic versus heterotrophic components could affect interpretations of total soil respiration responses to temperature (Subke & Bahn, 2010).
Findings from temperate forests suggest that root respiration is more sensitive to temperature than heterotrophic respiration (Epron et al., 2001). However, more recent research suggests that differences in root respiration may be driven largely by variability in substrate supply from photosynthesis (Ruehr & Buchmann, 2010). This rapid link between canopy and soil processes is illustrated by the rapid recovery of 13CO2 and 14CO2 in the soil CO2 efflux within days of canopy labeling in temperate forests, as well as a 37% decline in soil respiration within days of tree girdling in a boreal forest (Högberg & Read, 2006). Given evidence of strong ties between root respiration and photosynthesis, root respiration will likely acclimate to longer term warming as a result of substrate limitation if warming causes decreased overall photosynthesis rates.
Spatial variation in root abundances can also influence our interpretation of temperature effects on soil respiration. In a wet tropical forest in Costa Rica, Schwendenmann et al. (2003) found a strong positive relationship between temperature and CO2 concentration in the deeper soils (e.g. below the rooting zone), while both soil moisture and temperature were important controlling factors of soil respiration in the top 1 m (where the majority of roots are concentrated). Davidson et al. (2000) found a strong trend of increasing CO2 efflux with temperature in a tropical pasture in Brazil; however, the diurnal fluctuations of soil respiration were greater than seasonal variation, suggesting the temperature response was actually driven by diurnal changes in photosynthate allocation to roots rather than a direct response to temperature. Despite its importance to our overall understanding of total soil respiration, there are currently no studies we know of that directly investigate the thermal acclimation potential of root respiration in tropical forests. Differences in the relative time scale and magnitude of the responses of autotrophic versus heterotrophic components, and investigations into biochemical acclimation and substrate limitation could strongly affect interpretations of total soil respiration responses to temperature (Subke & Bahn, 2010).
(5) Synthesis: controls on below-ground C storage
Whether or not warming will have a positive effect on soil C inputs is vital to determining the net effect of temperature on soil C storage. While multiple cross-site comparisons suggest temperature will lead to an increase in C inputs, we found little evidence within tropical forest sites to support this result. Much longer time series (10+ years) are likely needed to reveal these relationships; however, these data sets do not currently exist for tropical forests. Our ability to predict temperature effects on C inputs will therefore depend heavily on our ability to resolve temperature sensitivities of above-ground processes (e.g. C allocation in plants).
Studies that directly investigate the temperature sensitivity of soil C loss from tropical forests are notably rare. As such, the majority of our predictions regarding the temperature sensitivity of soil C decomposition are primarily theoretical. Nevertheless, the general consensus is that increased temperature will result in additional C respiratory losses from tropical soils. There is considerable uncertainty however as to the magnitude and time scale of this loss. While additional laboratory experiments will be useful for evaluating the temperature sensitivity of specific mechanisms, they are short in duration (≤1 year), involve significant disturbance of the soil matrix, exclude autotrophic responses, and cannot account for the effect of altered C inputs. The effects of warming on below-ground processes are likely to be varied in both time and space (Rustad et al., 2001). We therefore see great benefit in the establishment of long-term (10+ years), field-based temperature manipulation experiments in tropical forests. Given evidence of a strong link between above- and below-ground processes, these experiments could greatly benefit from incorporating the whole-system response to warming (Fig. 1). However, due to the challenges of scaling a warming manipulation to the forest level, mesocosm experiments that include both plant and soil warming could provide a more immediate way forward.
V. INDIRECT CONTROLS ON C BALANCE
(1) Litterfall quality
Litterfall is the major pathway for transferring C and nutrients to the soil. It follows that changes in litter quality could significantly affect a suite of processes associated with C cycling and storage in tropical forests (Vitousek, 1984). For example, increased inputs of low-quality litter (e.g. high lignin:N ratios) could reduce decomposition rates, resulting in a greater proportion of litterfall C being stabilized within the forest floor. By contrast, increased litter quality (e.g. lower lignin:N ratios) could lead to greater microbial respiration and loss of C as CO2 (Vitousek, 1984). These changes in litter quality could also feed back to affect future forest productivity, resulting in more C stored in plant biomass (Wood et al., 2009). Analyses of total fine litterfall across five tropical forest elevation gradients showed a strong positive relationship between litter N and temperature (Silver, 1998). By contrast, a 7-year study in secondary forest in Costa Rica found total fine litter N to decline in response to small increases in minimum temperature (Tully & Lawrence, 2010), suggesting the possibility of a negative effect of temperature on foliar (i.e. live leaf) N. Studies in other ecosystems (temperate forest, grassland, tundra) have found both positive and negative effects of temperature on foliar N concentrations (Arft et al., 1999; Hobbie et al., 2001; An et al., 2005). Overall, studies that report temperature effects on litterfall quality in tropical forests are rare, and we found none that reported effects on nutrients other than N (e.g. P or Ca).
(2) Soil nutrient availability
Temperature-induced changes to soil nutrient availability could indirectly affect the C balance of tropical forests (Hungate et al., 2003). Soil nutrient availability has been shown to affect forest productivity and foliar nutrient concentrations, as well as rates of net photosynthesis and dark respiration (Meir et al., 2001). Wardle et al. (2005) suggested that the availability of nutrients strongly limits productivity in highly weathered soils, such as those found in many tropical forests, and that forests on such soils have the potential to be more productive if their soils were more nutrient rich. McKane et al. (1995) also suggested that C:nutrient interactions will constrain increases in C storage, resulting in a maximum increase of 16% above present-day stocks within the next 200 years. Nutrient availability also affects the rate of soil C loss via effects on soil respiration and decomposition rates (Cleveland & Townsend, 2006). Accordingly, temperature-nutrient interactions have the potential to affect C cycling in tropical forest soils. A long-term N fertilization study in Puerto Rico found that recalcitrant C in N-fertilized soils had much longer turnover times than in unfertilized soils; however, under elevated temperatures the N-fertilized soils lost significantly more recalcitrant C than unfertilized soils (Cusack et al., 2010). Taken together, these results suggest that C efflux and long-term stability of C stocks in tropical soils depend upon nutrient availability and temperature as well as their interaction.
Temperature may affect multiple aspects of N and P cycling in the tropics, including N fixation, nitrification, denitrification (Sierra, 2002; Houlton et al., 2008), and phosphatase activity (Gholz et al., 2000; Sierra, 2002). While there are no published field warming experiments in the tropics, evidence from soil-warming experiments in temperate and high-latitude ecosystems have found large increases (46%) in N mineralization rates in response to warming (Rustad et al., 2001). By contrast, laboratory results synthesized by Houlton et al. (2008) suggested that N inputs from N fixation may decline at temperatures above 26 °C. If this response holds under field conditions, increased temperatures in many tropical forests could result in decreased N inputs. Increased temperatures could also affect the relative proportions of different forms of soil N (e.g. NO3−versus NH4+). For example, if nitrification rates increase relative to other N cycling pathways, soils may maintain higher NO3−:NH4+ ratios. Variation in the proportion of different soil N pools has been shown to affect a variety of ecosystem characteristics, such as decomposition rates, plant cover and, ultimately, C cycling (Austin, Sala & Jackson, 2006). Warmer temperatures could also stimulate additional N losses (in particular NO3−) via increased denitrification rates. As a set of enzymatic processes, denitification is sensitive to changes in temperature; however, this and many N cycling responses are likely to vary by soil type (Stanford et al., 1975). Overall, there are a suite of mechanisms through which temperature increases could affect soil N cycling, N pools and the relative proportions of different N molecules.
Increased temperature is also likely to affect P availability, which could be especially important for tropical forests given the likelihood of P limitation in these systems (Vitousek, 1984; Vitousek & Farrington, 1997; Cleveland, Reed & Townsend, 2006; McGroddy & Silver, 2011). Laboratory studies suggest that increased temperature can increase phosphatase activity, which could in turn increase P availability and thus stimulate plant growth and/or soil respiration (Vitousek & Farrington, 1997; Cleveland & Townsend, 2006). Increased temperature could further increase P availability by reducing the sorption of phosphate (Barrow, 1984). However, if the binding of inorganic and organic P to soil decreases in response to high temperature, loss of P to leaching could also increase (Hedin, Vitousek & Matson, 2003). Overall, despite the potential for strong regulatory effects, our understanding of the potential response of nutrient availability to increased temperature is not well understood for tropical forests. Accordingly, our limited perspective hinders predictions of how temperature may affect tropical forest C cycling indirectly via effects on nutrient cycles.
(3) Temperature interactions with water, CO2 and light
Temperature is likely to interact with other environmental variables such as water availability, atmospheric CO2 and solar radiation levels. While a full exploration of these interactions is beyond the scope of this review, we briefly mention here the primary uncertainties in the literature surrounding temperature interactions with other abiotic factors.
While there is currently no CO2-addition experiment in any tropical forest, results from studies in temperate forests show CO2 fertilization effects in response to elevated atmospheric CO2 levels, including higher C production, greater water-use efficiency, and lower stomatal conductance (Ainsworth & Long, 2005). While some researchers suggest this CO2 fertilization effect is also occurring in tropical forests, whether or not tropical forests have experienced increased growth with elevated atmospheric CO2 levels remains controversial (Lloyd & Farquhar, 2008; Lewis et al., 2009a). Nevertheless, if increased CO2 does result in increased growth and water-use efficiency, the negative effects of increased temperature (and possible changes to rainfall regimes) would likely be reduced.
Finally, the amount of incoming solar radiation can significantly affect biological processes related to C cycling, most notably, photosynthesis. There is evidence that solar radiation varies on decadal time scales (Wild, 2009) and researchers have found evidence that an increase in anthropogenic aerosols in the atmosphere may result in global dimming (Rotstayn & Lohmann, 2002; Wielicki et al., 2002; Lewis et al., 2004; Wild, 2009). If light is indeed decreasing as temperature increases, we may expect reduced C uptake by plants due to lower photosynthetic rates (from lower light) and additional C loss from soils in response to increased temperature. Taken together, these concurrent changes could significantly reduce any net tropical forest C sink. However, given that light and temperature often co-vary on seasonal time scales, teasing apart their individual effects on ecosystem processes represents a significant research challenge. Experimental warming could help refine our understanding of the differential effects of these key drivers of ecosystem processes.
(4) Synthesis: indirect controls on C cycling
Current modeling efforts highlight the large role nutrient cycling could play in the response of net C flux to climate change (e.g. Wang & Houlton, 2009), and the few tropical fertilization studies that exist support this conclusion. In fact, temperature effects on nutrient cycling and availability could have a stronger influence on tropical C cycling than the direct effect of temperature change; however, the magnitude and extent of such changes remain unknown. The complexity of these interacting controls (i.e. temperature and nutrient availability) further confounds our ability to forecast future C cycling. Explicit studies in field settings would greatly facilitate accurate predictions of how temperature will affect C, N and P cycles that are tightly coupled in the real world. That said, laboratory and greenhouse studies could elucidate specific mechanisms behind temperature effects on nutrient availability so that we can begin to parcel out the likelihood and extent of these potential controls. The effects of increased temperature on N and P cycles when natural microbial communities, plant responses (e.g. C allocation), and other such complicating factors are included remain wholly unknown (Fig. 1).
Global warming is likely to be accompanied by other environmental changes (e.g. rising atmospheric CO2 concentrations), and these changes will interact with temperature in complex ways. However, while multi-factor experiments are valuable for testing concepts, these experiments can be difficult to design, conduct, and interpret. Single-factor experiments that focus on increasing our understanding of mechanisms may provide a reasonable way forward and a complement to both modeling and future large-scale endeavours (Norby & Luo, 2004).
1Over the last decade, debates regarding the primary forces driving changes in tropical forest C balance have been vigorous (e.g. Wright, 2005). Even with the full weight of the current evidence, an all-encompassing theory about what is happening and what will happen to tropical forests with global warming remains elusive.
2We suggest three main reasons for this. First, there is a general lack of data at and across the appropriate scales. While there is evidence to suggest that temperature will significantly affect important processes related to C cycling and storage in tropical forests, we found insufficient support in the literature to conclude with any certainty what the direction of this response is likely to be. Second, research conducted at different spatial and temporal scales often leads to contradicting conclusions. Large-scale observational studies confound multiple climatic, edaphic, and biotic factors, while small-scale studies cannot address the interconnectedness of above- and below-ground processes and are challenging to scale to the ecosystem. Third, tropical forests vary enormously in factors such as soil type and community composition (Townsend, Asner & Cleveland, 2008), which are known to influence ecosystem function. Thus, the likelihood that all tropical forests will respond similarly is low.
3We propose the following recommendations for future research that may help resolve some of the current uncertainties: (i) the majority of tropical forests already experience temperatures >20 °C. As such, our ability to use temperature gradients within the tropics to predict accurately their response to increased temperature is limited. There is currently no large-scale field-warming manipulation anywhere in the tropics, and we submit there is a dire need to attempt this in order to tease apart effects of temperature versus those of other environmental variables. (ii) The assortment of trajectories in tropical forests may be as diverse as the forests themselves, and the primary drivers of change may depend on multiple limiting factors. It is therefore important that we incorporate a range of tropical forest sites into our experiments and models. Ideally this would involve a network of tropical forest sites with coordinated efforts to ensure that methodological differences do not hinder comparisons [e.g. Center for Tropical Forest Science (CTFS), Amazon Forest Inventory Network (RAINFOR)]. (iii) Evidence suggests a tight interconnectedness between above- and below-ground processes (e.g. photosynthesis and root respiration). Warming may also affect C allocation patterns (e.g. tree morphology, C allocation of microbes) that could be missed by focusing on above- or below-ground components alone. Large-scale forest warming experiments would ideally be established to address these uncertainties. However, mesocosm experiments that include both plant and soil warming could accomplish some of the same goals. (iv) Long-term experiments are crucial (Luo et al., 2011). The recalcitrant C pool is likely to respond to environmental change on decadal time scales. Thermal acclimation of photosynthesis and/or microbial processes are also likely to occur over time scales longer than a season. As such, short-term manipulations (≤1 year) are unlikely to capture these critical responses. (v) There currently exists a large disconnect between findings obtained at different spatial scales. For example, temperature thresholds of photosynthesis depend on whether one is looking at the leaf-level (leaf temperature) or the canopy level (air temperature). Similarly, warming may differentially affect heterotrophic versus autotrophic soil respiration. Incorporating a range of scales into experiments will greatly improve our ability to understand mechanistically what is happening to tropical forests under warming conditions (see Fig. 1). (vi) Finally, there are some areas of tropical forest research where almost no field data on temperature response are available, including nutrient cycling, heterotrophic versus autotrophic respiration, thermal acclimation versus substrate limitation of plant and microbial communities, below-ground C allocation, temperature effects on species composition (plant and microbial), hydraulic architecture of roots, and the sensitivity of soil C loss to temperature.
4Whether or not tropical forests will become a source or a sink of C in a warmer world remains highly uncertain. Given the importance of these ecosystems to the global C budget, resolving this uncertainty remains a critical research priority.
We gratefully acknowledge M.G. Ryan, S.J. Van Bloem, A.E. Lugo, and two anonymous reviewers for their insightful comments and edits on earlier drafts of this manuscript. Support for this research was provided by the USDA Forest Service International Institute of Tropical Forestry.