The growth response of plants to elevated co2 under non-optimal conditions

Oecologia (2001) 129:1-20DOI 10.1007/s004420100736 Hendrik Poorter. Marta Pérez-Soba
The growth response of plants to elevated CO under non-optimal Received: 10 August 2000 / Accepted: 19 April 2001 / Published online: 25 July 2001 Springer-Verlag 2001 Abstract Under benign environmental conditions, plant
and decrease in others, resulting in an average interac- growth is generally stimulated by elevated atmospheric tion with elevated CO that was not significant. Under CO concentrations. When environmental conditions be- high ozone concentrations, the relative growth enhance- come sub- or supra-optimal for growth, changes in the ment by elevated CO was strongly increased, to the ex- biomass enhancement ratio (BER; total plant biomass at tent that high CO even compensated in an absolute way elevated CO divided by plant biomass at the current for the harmful effect of ozone on growth. No systematic CO level) may occur. We analysed literature sources difference in response was found between herbaceous that studied CO ×environment interactions on the growth and woody species for any of the environmental vari- of herbaceous species and tree seedlings during the vege- tative phase. For each experiment we calculated the dif-ference in BER for plants that were grown under 'opti- Keywords Nutrients. Water. Light. Temperature. Salt.
mal' and 'non-optimal' conditions. Assuming that inter- actions would be most apparent if the environmentalstress strongly diminished growth, we scaled the differ-ence in the BER values by the growth reduction due to The complex effect of elevated CO on plant growth the stress factor. In our compilation we found a largevariability in CO ×environment interactions between ex- The current increase in the atmospheric CO concentra- periments. To test the impact of experimental design, we tion has triggered a wide variety of botanical investiga- simulated a range of analyses with a plant-to-plant varia- tions during the last two decades, at a range of integra- tion in size common in experimental plant populations, tion levels. Notwithstanding this huge effort, we still in combination with a number of replicates generally have only a limited understanding about the effect of ele- used in CO ×environment studies. A similar variation in vated CO on plant growth. There is considerable varia- results was found as in the compilation of real experi- tion in the direction and magnitude of growth responses ments, showing the strong impact of stochasticity. We to elevated CO , partly depending on the duration of the therefore caution against strong inferences derived from exposure, plant development, species (e.g. species that single experiments and suggest rather a reliance on aver- differ in inherent growth rate or type of photosynthetic age interactions across a range of experiments. Averaged pathway) and the availability of primary resources over the literature data available, low soil nutrient supply (Kimball 1986a; Idso and Idso 1994; Poorter et al. 1996; or sub-optimal temperatures were found to reduce the Curtis and Wang 1998; Saxe et al. 1998). However, there proportional growth stimulation of elevated CO. In con- is still debate about when and where and to what extent trast, BER increased when plants were grown at low wa- these factors are important (Kimball 1986b; Idso and ter supply, albeit relatively modestly. Reduced irradiance Idso 1994; Lloyd and Farquhar 1996, 2000; Poorter or high salinity caused BER to increase in some cases 1998; Stitt and Krapp 1999). The situation becomes evenmore complex if we take into account that concomitantwith the increased level of CO , there are also increases in the level of air pollutants (ozone, nitrogen oxides, sul- Plant Ecophysiology, Utrecht University, P.O. Box 800.84, phur dioxide) and ultraviolet radiation. Enhanced deposi- 3508 TB Utrecht, The Netherlandse-mail: h.poorter@bio.uu.nl tion of air pollutants results in eutrophication and acidifi- Tel.: +31-30-2536859, Fax: +31-30-2518366 cation of natural ecosystems. Increased emissions ofCO , methane and chlorofluorocarbons might result in Alterra, P.O. Box 47, 6700 AA Wageningen, The Netherlands increased temperature and alterations in other climate parameters, such as the distribution and intensity of stress was then calculated as the reduction in total biomass at am- clouds (light) and precipitation (water). We therefore bient CO of plants grown at the non-optimal level compared to the total biomass of plants grown at the optimal level. We call this need to analyse how these changing environmental fac- the 'growth reduction due to stress' (GRS) and calculated it as: tors may modify the impact of elevated CO on plant A range of research papers and reviews has dealt with the interactions between elevated atmospheric CO con- with M and M being the total biomass of plants at the optimal level O and at a certain sub- or supra-optimal level S, respectively.
centration and environmental factors (e.g. Kimball We thereby assume that the higher the GRS, i.e. the larger the dif- 1986b; Gifford 1992; Idso and Idso 1994; Curtis and ference in biomass between the optimal and a non-optimal level, Wang 1998; Poorter 1998; Luo and Mooney 1999). In the stronger was the stress experienced by the plants. Because ra- most experiments, the CO effect is analysed at two lev- tios are ln-normally distributed by nature, we first ln-transformed the BER values obtained under optimal and non-optimal condi- els of another environmental factor, sometimes with tions, and then scaled the difference between these two values by quite contrasting results that hinder generalisations the growth reduction observed because of the interacting stress across experiments (Rawson 1992). Differences in re- sponse between species might be responsible for differ-ent results. Far less attention has been paid to the possi- bility that these differences are merely due to chance. In where SLB is an acronym for 'slope of the line connecting the two the first part of this paper, we analyse the degree of vari- BER values'. A graphical example of our method is given in ability in the results of a CO ×environment interaction Fig. 1. If plant biomass is as specified in the insert, then the ratio when we repeatedly sample a limited number of plants of plant biomass at elevated CO relative to ambient CO (BER) is from the same experimental population.
2 at the optimal level and 1.5 at the sub-optimal level. At ambientCO , we assume that the treatment with the highest biomass is op- In the second part, we try to obtain an overall picture timal, with a GRS of 0 as the x-value at which we plot the BER of of the interaction between elevated CO and environ- 2. The growth reduction due to the sub-optimal level is 0.6, the x- mental factors, such as primary resources, temperature value at which we plot the BER of 1.5. These values result in an and air pollutants. We will restrict our analysis to indi- SLB of -0.48. A negative SLB indicates that at a given non-opti-mal level of the interacting factor, the relative growth response to vidually grown plants in the vegetative stage. Apart from elevated CO is smaller than under optimal conditions. Note that the stochastic variation mentioned above, another factor in most of this paper we will focus on the relative growth re- may hinder generalisations across experiments, i.e. the sponse; the absolute growth response will almost always be lower range of environmental growth conditions applied in dif- A weak point in this approach is that we assume that BER ferent experiments, which most likely stress plants to changes linearly from optimal to non-optimal levels and that the different degrees. Therefore, we follow a method that environmental condition at which plants show the largest growth links the growth stimulation due to elevated CO to the growth reduction at ambient levels of CO due to the stress factor. That is, the severity of the applied environ-mental stress, as evident from the growth reduction inthe control CO plants, is used to scale the change in bio- mass response to elevated CO. This allows one to, at least partly, correct for differences between experiments.
An additional advantage of this approach is that we cancompare interactions between elevated CO and a range of growth-limiting environmental factors at the samescale.
SLB, a parameter to quantify CO ×environment interactions The minimal experimental design to analyse CO ×environment in- teractions requires an orthogonal combination of two CO concen- trations (ambient and elevated) and two levels of the other envi- Fig. 1 Example to show the method used to calculate the effect of
ronmental factor (optimal and non-optimal for growth). To quanti- limiting factors on the biomass enhancement ratio (BER). The x- tatively analyse those experiments, we used a method based on axis represents the reduction in total biomass at ambient CO of two main parameters. The first is an indicator of the stimulating plants grown at the sub- or supra-optimal level when compared to effect of elevated CO on total plant biomass (sum of above- and the total biomass of plants grown at the optimal level (growth re- belowground biomass) and is calculated as the ratio of plant bio- duction due to stress, GRS). The y-axis represents the ratio of mass at elevated and at ambient CO levels. We call this the 'bio- plant biomass at elevated and ambient CO levels. The positive, mass enhancement ratio', using BER as an acronym. The second zero or negative sign of the slope of the line connecting the two parameter is an indicator of the stress experienced by plants due to BER values indicates the type of interaction (see text). For the cal- a non-optimal level of the environmental factor under study. For culations, all BER values have to be ln-transformed prior to any each experiment, we considered as the 'optimal' level, the treat- statistical analysis, as ratios are ln-normally distributed by nature.
ment that resulted in the highest total biomass. The intensity of the treatments. The slope calculated to determine a CO ×environment interaction is based on the biomass of at least four differentlytreated groups of plants, each with its own variability in total bio-mass. Consequently, the estimate of the slope is affected by theadded variability in all four experimental groups (cf. Poorter et al.
1996; Hedges et al. 1999; Jasienski and Bazzaz 1999). The preci-sion of the slope is co-determined by the number of plants harvest-ed per treatment. Because of constraints on space and labour, thenumber of replicates harvested per treatment in experiments thatstudy CO effects in combination with other factors will generally be low. This is unfortunate, because it decreases precision where-as, in fact, due to the added variability in four plant groups, ahigher number of replicates would have been required than in asingle-factor experiment with two groups of plants.
To what extent might plant-to-plant variability explain the ob- served variation in SLBs as in Fig. 2? Because we do not know allthe details of each experiment, we can only answer this questionby a simulation of the most likely situation. From the specifica-tions of the CO ×nutrient experiments provided by the authors, we know that the median number of plants harvested per treatmentwas five. A low number is four, and a high number is ten, asjudged from the 20th and 80th percentile, respectively, of the com- Fig. 2 Frequency distribution of CO ×nutrient interactions. Bars
piled number of plants harvested in these experiments. We do not indicate SLB values derived from 123 published observations. The know the variability in the plant populations under investigation.
bold line indicates the distribution of SLBs after simulating a Poorter and Garnier (1996) used the standard deviation in ln-trans- range of experiments with a low (n=4), an intermediate (n=5) and ) as a way to characterise variability in ex- a high (n=10) number of replicates per treatment, harvesting plant perimental plant populations. From their compilation of a range of =0.51) variability in dry mass. The aver- low value of 0.21 (20th percentile) and a high value of 0.51 (80th age mass for the four different treatments was chosen so that both percentile). Assuming now that the true GRS and SLB values GRS and SLB were exactly the same as the average values in the were the average of the 123 experimental observations (0.55 and compiled data set. More information is given in the text -0.41, respectively), and that plant-to-plant variability is not al-tered by elevated CO , we simulated experiments in which we ran- domly 'harvested' four, five or ten plants out of three artificial response is truly optimal; this may not necessarily be the case. An of 0.21, 0.31 or 0.51, respectively. In this advantage is that the same method can be applied to different en- way, we arrived at nine different scenarios, and for each of these vironmental variables, since the interactive effect with elevated CO is related to the growth reduction caused by the non-optimal assume that the aggregated distribution of calculated slopes gives level and not to the environmental level itself. This enables a com- us a reasonable estimate of the extent to which slopes vary due to parison of different treatments, using the growth reduction due to random variation in biomass alone. The simulated distribution of the stress factor as a biological yardstick.
SLB values is shown as the continuous line in Fig. 2. Although the Biomass responses were analysed based on a compilation of 'true' (average) SLB value was negative, positive interactions published and unpublished experiments on individually grown were observed in 22% of the simulations. Moreover, variation was herbaceous and woody C species (see Appendix 1 and 2). C spe- largely similar to that observed in the literature. Based on this sim- cies were excluded, because the low number of CO ×environment ulation, we conclude that the relatively low number of plants har- studies conducted with these plants hardly allows any generalisat- vested from rather variable populations can explain most of the ion. In addition, we did not consider those studies in which the observed variability in CO ×nutrient interactions. We do not doubt non-optimal treatment caused a growth reduction of less than that variation in SLB is also partly due to differences between spe- 10%, both because we felt that such a treatment was not stressful cies or growth conditions. However, in our opinion, support for for the plants and because the GRS would become too small to ac- these alternative explanations has to be found in an a posteriori curately determine the slope of the line in Eq. 2. Following the analysis of a range of experiments and not in the mere observation above method, we calculated the SLBs for a range of factorial ex- that species A in experiment 1 responded differently from species periments, restricting the analysis to plants in the vegetative phase.
B in experiment 2 (see also General discussion below). In the The ambient CO concentration ranged between 300 and 400 µl l-1, analysis to follow, we will consider the average response across all and the elevated CO concentration between 550 and 1,100 µl l-1, observations, and only test for possible differences between herba- except for one experiment with high salinity.
ceous and woody species in general, unless otherwise stated.
How precisely can an interaction be determined? The SLB values may differ substantially between experiments. Anexample is given in Fig. 2, where we plotted the distribution of SLB values for 123 observations of plant species grown in a facto-rial combination of elevated CO and nutrient supply (grey bars).
In some cases, strong positive interactions were reported (e.g.
From the literature data listed in Appendix 1 and 2 and Whitehead et al. 1997: SLB>1); in other cases, strong negative plotted in Fig. 2, we obtained the distribution of the SLBs were found (e.g. Heath and Kerstiens 1997: SLB<-2). Most slopes represented by the boxplots of Fig. 3. On average, discussions almost automatically assume that such contrasting re- the SLB for nutrients was negative (P<0.001), with no sponses are due to the fact that different experiments use different indication of a difference between herbaceous and species, another level of the stress factor, or simply a differentcombination of growth conditions (e.g. Lloyd and Farquhar 2000).
woody species (P>0.5). This implies that a decrease in A factor that has received less attention is plant variability within nutrient availability reduces the relative growth response Theoretically, the relative stimulation of photosynthesisby elevated CO is strongest close to the light compensa- tion point (Kimball 1986a), and this has indeed been ob-served (Idso and Idso 1994). At low light, plant growthis strongly carbon limited, and therefore one would ex-pect this stimulation of photosynthesis by elevated CO2 to be translated into increased growth. However, analysisof the limited information (Fig. 3; 19 observations)shows that this interaction is small: the average SLBdoes not deviate significantly from zero, although itcomes close (0.05<P<0.1). Similar results have beenfound for crop yield (Kimball 1986a). Although not sig-nificant (P>0.3), there seems to be a tendency for tree Fig. 3 Distribution of slopes (SLB), indicating the strength of the
interaction between elevated CO and the primary resources (nu-
seedlings to have positive SLB values, whereas the her- trients, irradiance and water) on plant growth. For each of the en- baceous plants in our compilation showed - on average - vironmental factors, data are separated for herbaceous species no response. One might expect tree seedlings to be gen- (open boxplots) and tree seedlings (shaded boxplots). Data are erally more shade-tolerant than the five crop species that based on a literature review of factorial experiments with combi-nations of elevated CO and nutrients (n=51 and n=72 for herba- represent the herbaceous plants in this case. Such obser- ceous and woody species, respectively, in 83 papers), irradiance vations would be in line with the conclusion of Kerstiens (n=11 and n=8, respectively, in 8 papers) and water (n=12 and (1998) that within the group of woody species, the n=30, respectively, in 25 papers). An explanation of SLB values is shade-tolerant ones are the strongest in their growth re- given in Methodology and Fig. 1. Numbers in the graph are the sponse. He suggests that shade-tolerant species have a 10%-trimmed means of SLB values for herbs and woody speciestogether. Boxplots indicate the distribution of a range of observa- lower leaf area per unit leaf mass, which is less reduced tions. The lower part of the box shows the 25th percentile. The than in other tree species at elevated CO. In addition, highest part of the box gives the 75th percentile, and the line in species-specific differences in response in tree seedlings between, the median (50th percentile). The whiskers indicate the may change with small increases in light availability 10th (lower) and 90th (higher) percentile (Hättenschwiler and Körner 2000). Clearly, the numberof experiments with low light is far too limited to allow of plants to elevated CO. Similar conclusions have been any firm conclusion. Moreover, other factors like the drawn for CO -enriched crops (Kimball 1986a) and veg- quality of light used in the experiments may play a role etations (Stöcklin et al. 1998). Overall, the pattern of re- sponse was not affected by the type of nutrient in shortsupply, as judged from the similarity in interaction be-tween experiments where nitrogen, phosphorus or all nu- trients together were modified (Poorter 1998). Althoughthe average SLB is negative, positive slopes are found in Overall, the results obtained for a range of different her- 20% of the experiments. As discussed below, more de- baceous and woody species confirmed that a reduced tailed research, including a range of nutrient levels, water supply modestly enhances the relative growth re- should show whether these positive slopes are merely caused by chance or are a systematic response of specific P<0.05), with again a small but non-significant differ- ence between herbs and trees (0.05<P<0.1). As in the At low nutrient levels, growth is apparently not re- case of nutrients, 20% of the observations show an inter- stricted by carbon availability, since high concentrations action deviating from the general trend.
of starch and other non-structural carbohydrates are usu- ally found in nutrient-limited plants. Therefore, we do 30-60% on average (Morison 1993), which in turn re- not expect an increase in carbon fixation to lead to a sim- duces water loss in the plant. Consequently, CO may al- ilar stimulation in growth, unless plants at elevated CO leviate plant water stress by reducing water use. Howev- would acquire more nutrients or use them more efficient- er, plants that are stimulated in growth by high CO will ly (BassiriRad et al. 2001). In the case of N, one of the have an increased leaf area. This will result in increased ways to use the acquired nutrients more efficiently is to transpiration at the whole-plant level, thereby moderat- invest less of the available N into Rubisco, and more into ing the interaction (Samarakoon and Gifford 1996). The other compounds that limit growth. Interestingly, this effect of CO on stomatal conductance is observed in does not happen (Medlyn 1996; Makino et al. 2000). We are only beginning to understand the mechanism by throughout plant development, with little evidence for which plants with a low nutrient status adjust their acclimation. There is growing experimental evidence growth and how this limits the response to elevated CO suggesting that elevated CO may have small or insignif- icant effects on stomatal conductance of many forest tree species, especially conifers (Curtis 1996). Hence, the re-duced use of water in coniferous forests growing underelevated CO and the subsequent growth response may be smaller than predicted. In our compilation, however,we did not find a difference in the strength of the interac-tion between conifers and hardwoods (P>0.7).
Interaction with temperature and salinity Our analysis shows that the average SLB is negative forsub-optimal temperatures, which indicates that at close-to-optimal temperatures, the relative biomass increase byelevated CO is higher than at low temperatures (Fig. 4; 59 observations, P<0.001). This result is in agreementwith results from previous analyses, which also conclud- Fig. 4 Distribution of SLB values, indicating the strength of the
interaction between elevated CO and sub-optimal temperature,
ed that low temperature reduced the growth response to supra-optimal temperature and salinity. Data are based on a litera- elevated CO (Idso et al. 1987; Rawson 1992; Curtis and ture review (sub-optimal temperature: n=48 and n=11 for herba- Wang 1998), although, again, 20% of the observations ceous and woody species, respectively, in 24 papers; supra-opti- differ in direction from the other experiments, with a mal: n=5 and n=4 in 6 papers; salinity: n=16 and n=2 in 12 pa-pers). Because of the low number of observations for supra-opti- BER higher at low temperature. No statistical difference mal temperatures and for woody species at high salinity, we only was detectable between herbs and woody species calculated the average values (open circles herbaceous plants, (P>0.15). In a few experiments, the highest temperature closed circles woody plants). For more information see the legend was supra-optimal for growth. In those cases, the largest growth response was at the highest temperature as well,although the difference was not statistically significant(Fig. 4; 9 observations, P>0.15).
negative effects of a supra-optimal salt (NaCl) concen- There are at least two explanations for the CO ×tem- tration on growth. This has indeed been found in a num- perature interaction. In the short term, an increase in am- ber of cases, but not all, and the mean SLB does not de- bient CO concentration results in increased photosyn- viate significantly from zero (Fig. 4; 18 observations, thesis in C species, not only by increasing the concen- P>0.4). Hardly any data have been published for woody tration of substrate but also by suppressing oxygenation species. Munns et al. (1999) suggested a positive (Long 1994). An increase in temperature promotes oxy- CO ×salt interaction at low salinity, but no CO effect at genation relative to carboxylation through decreases in high salinity. From the present compilation we conclude the affinity of the enzyme Rubisco for CO. Moreover, that most halophytes have a higher BER at supra-optimal the solubility of CO decreases faster than that of O at salinity, whereas most glycophytes have a lower BER high temperature, diminishing the relative abundance of under these conditions (Appendix 1 and 2). However, the CO in the chloroplasts (Jordan and Ogren 1984). There- few observations available preclude any firm conclusion fore, the stimulating effect of elevated CO on photosyn- thesis is strongest under warmer conditions. An alterna-tive explanation for the low response at low temperaturesis that growth is more impaired by sub-optimal tempera- tures than photosynthesis (Körner 1991; Rawson 1992).
As in the case of low nutrient supply, this will result in the accumulation of non-structural carbohydrates. Withsink strength being so crucial for the growth response of Of all factors considered here, ozone shows the strongest plants (e.g. Reekie et al. 1998), plants at low temperature interaction with CO. The slope is positive (Fig. 5; 29 are probably not able to profit much from an increased observations, P<0.001), and this is true for 95% of the sugar supply due to elevated CO (Greer et al. 2000).
observations, with no indication of a difference between woody and herbaceous species (P>0.7). This implies thatCO strongly ameliorates the detrimental effect of ozone.
There is good evidence that in plants in which stomatalconductance is reduced by CO enrichment, O flux into Salinity has a negative effect on both the water status the leaf interior is reduced and this contributes to reduc- and the photosynthetic apparatus of plants (Ball and ing the injurious impact of O on plant growth and phys- Munns 1992). As elevated CO has exactly the opposite iology (Turcsányi et al. 2000). Three major questions re- effects, one might expect elevated CO to ameliorate the main with regard to the protection against O damage generally increases the concentrations of soluble phenol-ic compounds (Poorter et al. 1997; Peñuelas and Estiarte1998), some of which are known to decrease plant sensi-tivity to UV-B. Most results to date have been obtainedunder artificial-environment conditions, which could re-sult in stronger damage than in the field situation. First,the UV-B levels used in the experiments are generallyvery high (Rozema 1993). Second, leaves developed un-der high light adapt morphologically and physiologicallyin a way that may also confer protection against UV-B(Teramura and Murali 1987). Consequently, plants ingrowth chambers, in which the daily irradiance is abouttwo times lower than under field conditions (Garnier andFreijsen 1994), may be more sensitive to UV-B thanplants in the field.
Fig. 5 Distribution of SLB values, indicating the intensity of the
interaction between elevated CO
based on a literature review of interactions with O (n=16 and n=13 for herbaceous and woody species, respectively, in 19 pa- pers), UV-B (n=2 and n=6 in 8 papers) or SO (n=3 and n=0 in 2 papers). Because of the low number of observations for UV-B and The very few data available on the combined effects of SO , we only calculated the average values (open circles herba- elevated CO and supra-optimal SO (3 observations) ceous plants, closed circles woody plants). For more information show a positive interaction, with high SLB values. Thissuggests that CO enrichment reduces the adverse effects of SO on plant growth. SO is probably used as a source provided by elevated CO. First, does elevated CO in- of sulphur and assimilated to proteins and other organic duce other advantageous mechanisms in addition to sto- matal closure, such as detoxification or repair processes higher metabolic rates that may stimulate the sulphur as- (J. Cardoso-Vilhena, personal communication)? Second, similation and accelerate repair processes (Rao and De what is the combined effect of elevated CO and O on Kok 1994). In addition, high CO decreases stomatal the growth and productivity of species in which the sto- conductance, which in turn may reduce the SO flux into mata are less responsive to CO enrichment, such as the leaf. However, when SO levels are very high, as in many conifers? Data indicate that for these species, there many East European countries, elevated CO may not be may be similar effects of O at ambient and elevated able to counteract the detrimental effect of SO.
CO , or at least much less amelioration of O damage than observed in herbaceous species (Pérez-Soba et al.
1995). However, the data on conifers in the literature are at present too sparse to be conclusive at this stage. Andthird, what is the combined effect of elevated CO and O on photosynthesis? Long-term exposure to elevated CO is accompanied by a decrease in Rubisco activity or We would like to make a strong case for meta-analysis as amount of Rubisco protein in many species (Drake et al.
a tool that allows generalisation across a wide range of 1997). Likewise, both short-term exposures to peak con- experiments (Gurevitch and Hedges 1999). It provides a centrations of O and to high background concentrations framework to judge whether a new result falls within the of O show a decline in Rubisco activity (Pell et al.
low, high or average range of previous observations.
1994). If the effects of elevated CO and elevated O on Moreover, it may allow the detection of contrasting re- Rubisco were additive, then the decrease in activity sponses between (groups of) species or environments, would result in a reduction of photosynthetic capacity.
before such differences have been explicitly tested in aspecifically designed experiment. Finally, because thestrength of the interaction is prone to random variation (Fig. 2), average values across experiments may give abetter estimate of the strength of the interaction under Experiments with CO ×UV-B interactions are scarce study. However, when interpreting the results of a meta- (8 observations). As with other interactions, data are analysis, one should keep in mind that this approach has variable, and the average SLB does not deviate signifi- some limitations. First, unnoticed mistakes may have oc- cantly from 0 (Fig. 5; P>0.5). Thus, elevated CO may curred in the experimental phase or during calculation of not compensate for the harmful effect of UV-B. The rea- the data on which the compilation is based. Second, re- son for this could be that UV-B primarily affects photo- searchers may have chosen to refrain from publishing system II, whereas CO influences carboxylation and data that were found to be statistically non-significant, stomatal conductance. On the other hand, elevated CO which may bias the overall picture (Gurevitch and Hedges 1999). Third, the available studies are not necessarily aweighted random sample of global vegetation, implyingthat estimates of the response of the 'average' C plant or vegetation are extrapolations with unknown confi-dence margins. Fourth, we can never exclude that an ob-served class difference in SLB (e.g. woody plants versusherbs) is confounded with another difference across spe-cies (e.g. sun versus shade species), or a difference in ex-perimental conditions (cf. Lloyd and Farquhar 2000).
Such a risk is particularly evident when only a few stud-ies have been carried out, as in the case of CO ×light in- teractions. A last point to consider, especially in the con-text of the present review, is that we assumed that inter-actions would be similar for CO concentrations ranging between 550 and 1100 µmol mol-1, and that the BERvalues change linearly between the assumed optimal and Fig. 6 Summary of the average growth response of plants for an
interaction between elevated CO and other environmental factors.
Responses are calculated using a biomass enhancement ratio of Given these considerations we face a dilemma. Ideal- 1.47 for plants grown under optimal conditions. The average slope ly, conclusions would be based on large-scale experi- was calculated from the data of Figs. 4, 5 and 6, and the average ments that study CO ×environment interactions for a reduction in growth at 350 µl l-1 CO as calculated in the compiled wide range (say >15) of ecologically contrasting species.
literature. The dashed line indicates the biomass enhancement byelevated CO that would compensate for biomass reduction under Even in this case, true generality is only achieved if re- stress conditions, not only in a proportional but also in an absolute searchers at different laboratories independently arrive at similar conclusions. As such large-scale screenings arerare, and the vast majority of experiments is restricted toone to four species, we have to accept that most of the (e.g. Idso and Idso 1994) do not hold. The growth en- generalisations will come from combining information hancement by elevated CO is severely reduced at low from a variety of experiments. To minimise the chance temperatures or poor nutrient supply. This is not only ex- effect alluded to in Fig. 2, we suggest using an experi- plained by the more negative SLB values, but also by the mental design with more than two levels of the interact- generally strong growth reduction in those experiments ing factor, giving more degrees of freedom to estimate (GRS >0.5). The average growth enhancement by elevat- the overall response. Moreover, if plant-to-plant varia- ed CO at optimal conditions is not significantly altered tion is not of prime interest, all precautions possible by high UV-B, high salinity or low irradiance, mainly should be taken to minimise and control plant-to-plant because the average SLB values were only marginally variability within the experimental population (Poorter different from zero. The interaction with water was sig- and Garnier 1996), which will also improve the precision nificant, but the effect was small. The interaction be- tween elevated CO and O was strong. This is the only type of stress where biomass is stimulated more thantwofold under elevated CO (BER values at high O are often larger than 2). The average value is above the dot-ted line, indicating that the loss of biomass at elevated The effect of an interaction between CO and any envi- O is more than compensated by the presence of elevated ronmental factor will not only depend on the slopes of CO. However, the biomass of high-CO plants at high the lines (Figs. 3, 4 and 5), but also on the magnitude of O concentrations is not as large as that of high-CO the growth reduction due to the stress factor at ambient CO. This is taken into account in Fig. 6, where we plot the average BER values against the average GRS, as ex-plained in Fig. 1. At optimal conditions (GRS=0), we as- sumed a BER value of 1.47 (average from the compila-tion by Poorter et al. 1996). The BER values at non-opti- The responses in Fig. 6 are average values of literature mal conditions were then derived from the average SLB data for both herbaceous and woody species. Some time and GRS values in the present compilation. The dashed ago, Curtis and Wang (1998) reviewed the growth re- line in the figure indicates the extent to which the en- sponse of woody plants to elevated CO. To the extent hancement in plant biomass by elevated CO should in- crease in order to compensate for growth losses at non- conclusions and ours are in agreement. This can be ex- optimal conditions, not only in a proportional but also in plained by the fact that we did not find systematic differ- an absolute way. Clearly, propositions that plants under ences between woody seedlings and herbaceous species stress will always respond relatively more strongly to for any of the environmental factors, although some (ir- enrichment than those under optimal conditions radiance, water) are on the verge of significance. Con- clusions deviate strongly for the factor ozone, where we calculated much stronger responses both for herbaceousand woody species. The fact that Curtis and Wang Plant-to-plant variability in biomass within treatments is (1998) had only two data points for this factor may ex- one of the factors that explains contrasting CO ×environ- plain the different results. We were not able to find sys- ment interactions published in the literature. On average, tematic differences in the compiled literature between re- the growth stimulation by elevated CO is smaller at low sponses of gymnosperms and hardwood seedlings. This nutrient availability and low temperature, increases may imply that the differential response of stomatal con- somewhat at low water supply, and is substantially high- ductance with respect to increased CO does not neces- er at high ozone concentrations. There is a strong paucity sarily lead to a strongly different CO ×environment in- of data on the interaction with light, salt, UV-B, nitroge- nous air pollutants and SO , but, with the exception of We have not paid attention to C and Crassulacean ac- SO , average responses are small. No systematic differ- id metabolism species, because far less information is ences were found between woody and herbaceous spe- available for the response of these species under sub- or supra-optimal conditions. However, as their response toelevated CO is generally smaller than that of C species Acknowledgements We thank Ep Heuvelink, Eric Garnier, Gina
(Poorter et al. 1996), we expect the CO ×environment Adams and Manuela Chaves for trustfully providing us with (par- tially unpublished) data for incorporation in our analyses. InekeStulen, Jan Goudriaan, Marcel van Oijen and an anonymous re-viewer thoughtfully commented on a previous version of themanuscript.
×environment interaction. Data are for herbaceous species and listed in alphabetical order ences are given as first author and year of the publication olin (1996) Physiol Plant 97:674-684; V illiams (1981) Plant Physiol 68:1406-1409; ard (1999) Global Change Biol 5:857-867; fects of climate change on marine coastal ecosystems, 1; Nicolas (1993) Aust J Plant Physiol 20:349-360; Oberbauer 17-129; Reinert (1997) New Phytol 137:41 eed Sci 36:751-757; Radoglou (1992) Ann Bot 70:245-256; Reddy (1998) En- Crop Sci 38:122-128; Nakamura (1999) Photosynthetica 37:61-72; Newbery (1996) (1986) Can J Bot 64:2993-2998; Olszyk (1997) Agric Ecosyst Environ 66:1-10; Paez (1984) J Agric Sci 102:687-693; Patterson (1982) W (1997) Plant Ecol 128:182-191; Rufty (1994) Physiol Plant 91:503-509; Sa (1998) J.
Plant Nutr 21:2207-2218; Samarakoon (1995) Aust J Plant Physiol 22:33-44; Schwarz (1984) J Exp Bot 35:193-196; Serraj (1999) Global Change Biol 5:283-291; Sionit (1981) Agron J 73:1023-1027; Sionit (1982) Agron J 74:721-725; Sionit (1983) Crop Sci 23:329-333; Sionit (1987) Can J Plant Sci 67:59-67; Sritharan (1992) Garten- bauwissenschaft 57:246-251; Stirling (1998) New Phytol 140:343-354; T Wheeler (1991) Crop Sci 31:1209-1213; W (2000) Plant Physiol 124:767-779; Zerihun (2000) Ann Bot 86:723-730; Ziska (1997) 18-123; Imai (1979) Jpn J Crop Sci 48:409-417; Israel ageningen, pp 64-67; Lenssen (1993a) PhD thesis, Free 126; Carlson (1982) Oecologia 54:50-54; Chen (1997) Gartenbauwis- 1; Barett (1995) J Biogeogr 22:331-339; Barnes (1992) New Phytol 121:403-412; field; Almeida (1999) Plant Soil 210:159-166; Arp (1998) Plant Cell Environ Aben (1999) Aust J Plant Physiol 26:759-766; Adams (1996) PhD thesis, University of Barnes (1995) Global Change Biol 1:129-142; Bernacci (2000) Global Change Biol 6:855-863; Bowler (1993) New Phytol 144:515-522; Bowler (1996) New Phytol 132:391-401; Campbell (1993) Proc 17th Grassland Conf, Palmerston North, New Zea- senschaft 62:30-37; Coleman (1992) Ecology 73:1244-1259; Coleman (1993) Oecologia 93:195-200; Cure (1988) Crop Sci 28:671-677; De Luis (1999) Physiol Plant 107:84-89; Ferris (1995) New Phytol 131:491-501; Ghanoum (1998) Aust J Plant Physiol 25:627-636; Goudriaan (1983) Neth J Agric Sci 31:157-169; Greer (2000) Aust J Plant Physiol 27:301-310; Harmens (2000) Ann Bot 86:833-839; Heagle (1993) New Phytol 123:751-762; Heagle (1999) Crop Sci 39:731-744; Hocking (1985) Ann Bot 55:835-844; Hughes (1971) Ann Bot 35:933-945; Hunt (1995) Ann Bot 75:207-216; (1990) J Plant Nutr 13:1419-1433; Jablonski (1997) Can J Bot 75:533-545; Jansen (1986) Biological control of photosynthesis, Nijhof (1996) J Environ Qual 25:908-916; Larigauderie (1988) Oecologia 77:544-549; Lenssen (1990) In: Goudriaan et al (eds) The greenhouse ef Oecologia 84 207-214; McConnaughay (1993) Oecologia 94:550-557; Miller (1998) ×environment interaction. Data are for woody species listed in alphabetical order given as first author and year of the publication Arnone (1990) New Phytol 116:55-66; Arp (1998) Plant Cell Garnier E, Freijsen AHJ (1994) On ecological inference from lab- Environ 21:1-11; Ball (1997) Plant Cell Environ 20:1158-1160; oratory experiments conducted under optimum conditions. A Bassow (1994) Ecol Appl 4:593-603; Bazzaz (1993) Ecology Whole Plant Perspective on Carbon-Nitrogen Interactions.
74:104-114; Broadmeadow (2000) New Phytol 146:437-451; Roy J, Garnier E (Eds) SPB Academic Publishing, The Hague, Brown (1986) Tree Physiol 2:223-232; Bruhn (2000) New Phytol 146:415-425; Catovsky (1999) Global Change Biol 5:507-518; Gifford RM (1992) Interaction of carbon dioxide with growth-lim- Centritto (1999) New Phytol 141:119-140; Conroy (1986) Ann iting environmental factors in vegetation productivity: impli- Bot 57:165-177; Conroy (1988) Plant Cell Environ 11:91-98; cations for the global carbon cycle. Adv Bioclimatol 1: 24-58 Conroy (1990) Plant Cell Environ 13:329-337; Conroy (1992) Greer DH, Laing WA, Campbell BD, Halligan EA (2000) The ef- Plant Cell Environ 15:843-847; Curtis (1995) New Phytol fect of perturbations in temperature and photon flux density on 129:253-263; Dickson (1998) Can. J For Res 28:1706-1716; El- the growth and photosynthetic response of five pasture species Kohen (1992) Ann Sci For 49:83-90; El-Kohen (1994) Tree to elevated CO. Aust J Plant Physiol 27: 301-310 Physiol 14:679-690; Gaucher (1998) In: De Kok and Stulen (eds) Gurevitch J, Hedges LV (1999) Statistical issues in ecological me- Responses of plant metabolism to air polution and global change, Backhuys, Leiden, pp 305-308; Gebauer (1996) New Phytol Hättenschwiler S, Körner C (2000) Tree seedling responses to in 134:85-93; Gleadow (1998) Plant Cell Environ 21:12-22; situ CO -enrichment differ among species and depend on un- Goudriaan (1983) Neth J Agric Sci 31:157-169; Griffin (1993) derstorey light availability. Glob Change Biol 6: 213-226 Oecologia 95:575-580; Griffin (1995) New Phytol 129:547-556; Heath J, Kerstiens G (1997) Effects of elevated CO on leaf gas Griffin (1997) Plant Soil 190:11-18; Guehl (1994) Tree Physiol exchange in beech and oak at two levels of nutrient supply: 14:707-724; Heath (1997) Plant Cell Environ 20:57-67; Hibbs Consequences for sensitivity to drought in beech. Plant Cell (1995) New Phytol 129:569-577; Hoffmann (2000) Oecologia 123:312-317; Johnsen (1993) Can J For Res 23:1033-1042; John- Hedges LV, Gurevitch J, Curtis PS (1999) The meta analysis of re- son (1995) Plant Soil 168/169:535-545; Kerstiens (1994) New sponse ratios in experimental ecology. Ecology 80: 1150-1156 Phytol 148:607-614; Koike (1993) Proc IGBP Symp 1992, pp Hodinott J, Scott R (1996) The influence of light and quality and 425-430; Lavola (2000) Physiol Plant 109:260-267; Lewis (1996) carbon dioxide enrichment on the growth and physiology of New Phytol 133: 431-443; Lovelock (1996) Funct Ecol seedlings of three conifer species. I. Growth responses. Can J 10:662-667; Miao (1992) Oecologia 90:300-304; Midgley (1995) J Biogeogr 22:185-191; Mortensen (1995) Environ Pollut Idso KE, Idso SB (1994) Plant responses to atmospheric CO en- 87:337-343; Murray (2000) Tree Physiol 20:421-434; Norby richment in the face of environmental constraints: a review of (1991) New Phytol 117:515-528; Oberbauer (1986) Can J Bot the past 10 years' research. Agric For Meteor 69: 153-203 64:2993-2998; Picon (1996) Ann Sci For 53:431-446; Pregitzer Idso SB, Kimball BA, Anderson MG Mauney JR (1987) Effects of (1995) New Phytol 129:579-585; Prior (1997) Tree Physiol atmospheric CO enrichment on plant growth: the interactive 17:397-405; Runion (1999) Tree Physiol 19:329-335; Schorte- role of air temperature. Agric Ecosys Environ 20: 1-10 meyer (1999) Aust J Plant Physiol 26:737-747; Sheu (1999) Envi- Jasienski M, Bazzaz FA (1999) The fallacy of ratios and the test- ron Exp Bot 41:57-65; Silvola (1992) Oecologia 91:208-213; ability of models in biology. Oikos 84: 321-326 Silvola (1993) Oikos 67:227-234; Silvola (1995) Plant Soil Jordan DB, Ogren WL (1984). The CO /O specifity of ribulose 168/169:547-553; Stewart (1993) Physiol Plant 88:493-500; 1,5-bisphosphate concentration, pH and temperature. Planta Sullivan (1994) Plant Cell Environ 17:311-317; Syvertsen (1999) Plant Soil 208:209-219; Tjoelker (1998) New Phytol Kerstiens G (1998) Shade-tolerance as a predictor of responses to 140:197-210; Tolley (1984a) Can J For Res 14:343-350; Tolley elevated CO in trees. Physiol Plant 102: 472-480 (1984b) Can J Bot 62:2135-2139; Townend (1993) Tree Physiol Kimball BA (1986a) Influence of elevated CO on crop yield. In: 13:389-400; Townend (1995) New Phytol 130:193-206; Tschap- Enoch HZ, Kimball BA (eds) Carbon Dioxide Enrichment of linski (1993) Tree Physiol 13:283-296; Tschaplinski (1995) New Greenhouse Crops Vol. II. Physiology, Yield and Economics.
Phytol 129:63-71; Uselman (2000) Plant Soil 222:191-202; Utri- ainen (1998) In: De Kok and Stulen (eds) Responses of plant me- Kimball BA (1986b) CO stimulation of growth and yield under tabolism to air polution and global change, Backhuys, Leiden, pp environmental restraints. In: Enoch HZ, Kimball BA (eds) 467-469; Vivin (1997) Ann Sci For 54:597-610; Volin (1996) Carbon Dioxide Enrichment of Greenhouse Crops Vol. II.
Physiol Plant 97:674-684; Volin (1998) New Phytol 138:315-325; Physiology, Yield and Economics. CRC press, Boca Raton, pp Wayne (1998) Oecologia 114:335-342; Whitehead (1997) New Phytol 135:201-212; Wilkins (1994) Tree Physiol 14:769-779; Körner C (1991) Some often overlooked plant characteristics as Wong (1992) Aust J Bot 40:457-472; Yakimchuk (1993) Can J determinants of plant growth: a reconsideration. Funct Ecol 5: For Res 24:1-8; Zak (2000) Ecol Appl 10:34-46 Lloyd J, Farquhar GD (1996) The CO dependence of photosyn- thesis, plant growth responses to elevated atmospheric CO2 concentrations and their interaction with soil nutrient status. I.
General principles and forest ecosystems. Funct Ecol 10: 4-32 Lloyd J, Farquhar GD (2000) Do slow-growing species and nutri- Ball MC, Munns R (1992) Plant responses to salinity under elevat- ent-stressed plants consistently respond less to elevated CO ? ed atmospheric concentrations of CO. Aust J Bot 40: A clarification of some issues raised by Poorter (1998). Glob.
BassiriRad H, Gutschick VP, Lussenhop J (2001) Root system ad- Long SP (1994) The potential effects of concurrent increases in justments: regulation of plant nutrient uptake and growth re- temperature, CO and O on net photosynthesis, as mediated sponses to elevated CO. Oecologia 126: 305-320 by RubisCO. In: Alscher RG, Wellburn AR (eds) Plant Re- Curtis PS (1996) A meta-analysis of leaf gas exchange and nitro- sponses to the Gaseous Environment. Chapman & Hall, Lon- gen in trees grown under elevated carbon dioxide. Plant Cell Luo Y, Mooney HA (1999) Carbon Dioxide and Environmental Curtis PS, Wang X (1998) A meta-analysis of elevated CO effects on woody plant mass, form, and physiology. Oecologia 113: Makino A, Nakano H, Mae T, Shimada T, Yamamoto N (2000) Photosynthesis, plant growth and N-allocation in transgenic Drake BG, Gonzalez-Meler MA, Long SP (1997). More efficient rice plants with decreased Rubisco under CO enrichment. J plants: a consequence of elevated carbon dioxide? Annu Rev Plant Physiol Plant Molec Biol 48: 607-640 Medlyn BE (1996) The optimal allocation of nitrogen within the Rao MV, De Kok LJ (1994) Interactive effects of high CO and C3 photosynthetic system at elevated CO. Aust J Plant SO on growth and anti-oxidant levels in wheat. Phyton- Ann Morison JIL (1993) Responses of plants to CO under water limit- Rawson HM (1992) Plant response to temperature under condi- ed conditions. In: Rozema J, Lambers H, van de Geijn SC, tions of elevated CO. Aust J Bot 40: 473-490 Cambridge ML (eds) CO and the Biosphere. Kluwer Aca- Reekie EG, MacDougall G, Wong I, Hicklenton PR (1998) Effect demic Publishers, Dordrecht, pp 193-209 of sink size on the growth response to elevated atmospheric Munns R, Cramer GR, Ball MC (1999) Interactions between ris- CO within the genus Brassica. Can J Bot 76: 829-835 ing CO , soil salinity and plant growth. In: Luo Y, Mooney Rozema J (1993) Plant reponses to atmospheric carbon dioxide HA (eds) Carbon Dioxide and Environmental Stress. Academ- enrichment: interactions with some soil and atmospheric con- Pell EJ, Eckardt NA, Glick RE (1994) Biochemical and molecular Samarakoon AB, Gifford RM (1996) Water use and growth of cot- basis for impairment of photosynthetic potential. Photosynth ton in response to elevated CO in wet and drying soil. Aust J Peñuelas J, Estiarte M (1998) Can elevated CO affect secondary Saxe H, Ellsworth DS, Heath J (1998) Tree and forest functioning metabolism and ecosystem function? Trends Ecol Evol 13: in an enriched CO atmosphere. New Phytol 139: 395-436 Stitt M, Krapp A (1999) The interaction between elevated carbon Pérez-Soba M, Dueck TA, Puppi P, Kuiper PJC (1995) Interac- dioxide and nitrogen nutrition: the physiological and molecu- tions of elevated CO , NH and O on mycorrhizal infection, lar background. Plant Cell Environ 22: 583-622 gas exchange and N metabolism in saplings of Scots pine.
Stöcklin J, Schweizer K, Körner C (1998) Efects of elevated CO2 and phosphorus addition on productivity and community com- Poorter H (1998) Do slow-growing species and nutrient-stressed position of intact monoliths from calcareous grassland. Oeco- plants respond relatively strongly to elevated CO ? Glob Teramura AH, Murali NS. (1987) Intraspecific differences in Poorter H, Garnier E (1996) Plant growth analysis: evaluation of growth and yield of soybean exposed to ultraviolet-B radiation experimental design and computational methods. J Exp Bot under greenhouse and field conditions. Environ Exp Bot 26: Poorter H, Roumet C, Campbell BD (1996) Interspecific variation Turcsányi E, Cardoso-Vilhena J, Daymond J., Gillespie C, Balaguer in the growth response of plants to elevated CO : A search for L., Ollerenshaw JH, Barnes JD (2000). Impacts of tropospheric functional types. In: Körner C, Bazzaz FA (eds) Carbon Diox- ozone: past, present and likely future. In: Trace Gas Emissions ide, Populations, Communities. Academic Press, San Diego, and Plants (ed. by S.N. Singh), Kluwer Academic Publishers, Poorter H, Van Berkel Y, Baxter B, Bel M, Den Hertog J, Dijkstra Whitehead SJ, Caporn SJM Press MC (1997) Effects of elevated P, Gifford RM, Griffin KL, Roumet C, Wong SC (1997) The CO , nitrogen and phosphorus on the growth and photosynthe- effect of elevated CO on the chemical composition and con- sis of two upland perennials: Calluna vulgaris and Pteridium struction costs of leaves of 27 C species. Plant Cell Environ aquilinium. New Phytol 135: 201-211

Source: http://www.science.poorter.eu/2001_Poorter%26Perez-Soba_Oecologia.pdf