Moister soils at elevated CO₂ stimulate root biomass but suppress aboveground biomass production in Lolium perenne

Introduction

There are convincing ecophysiological arguments to support the notion that the growth of C₃ plants is limited by the current atmospheric concentration of carbon dioxide ([CO₂]). Firstly, the net rate of photosynthetic carbon assimilation rises as the [CO₂] increases in relation to the oxygen concentration because the carboxylation rate of ribulose-1,5-bisphosphate (RuBP) by Rubisco, the primary carboxylating enzyme, increases while the oxygenation rate of RuBP lessens (von Caemmerer and Farquhar 1981; Long 1991; Long and Drake 1992; Farquhar 1997; Long et al. 2004). Further, stomatal aperture reduces as [CO₂] rises (Gifford 1979; Drake et al. 1997; Farquhar 1997; Long et al. 2004), reducing leaf and canopy-level transpiration and leading to water savings that prolong soil moisture benefits from precipitation or irrigation (Gifford 1979; Long et al. 2004; Morgan et al. 2004). Increases of growth by elevated CO₂ concentration (eCO₂), particularly of leaf area, have the potential to increase plant water use thereby counteracting the effects of reduced stomatal aperture (Fatichi et al. 2016; Gray et al. 2016), but many field studies have nonetheless demonstrated that eCO₂ leads to a reduction in canopy transpiration (e.g. Hussain et al. 2013; Kellner et al. 2019), with a ~6% average eCO₂-induced reduction of evapotranspiration across ecosystems (Bernacchi and VanLoocke 2015). These two physiological effects – increased photosynthesis and greater water use efficiency – lead to the widely held conclusion that the effect of elevated CO₂ concentrations (eCO₂) should be strongest when moisture is limited, most particularly in grasslands (Gifford 1979; Morgan et al. 2004; Reich et al. 2006a; Körner et al. 2007; McMurtrie et al. 2008; Morgan et al. 2011). This generalisation is supported by experimental results in which the biomass response to eCO₂ was strongest in dry years (Morgan et al. 2004, 2011) and modelling shows stronger biomass responses to eCO₂ in drier sites (Fatichi et al. 2016). However, other studies find no relationship between water supply and the strength of the eCO₂ effect (Reich et al. 2006a; Norby and Zak 2011; Reich et al. 2014), limiting generalisations.

Both water availability and plant demand for water vary seasonally in grasslands (Knapp et al. 2001) meaning that any influences of eCO₂ on improved water use efficiency would also vary seasonally. In addition, the eCO₂ effect strengthens with increasing nitrogen (N) availability (Reich et al. 2006a, 2006b; McMurtrie et al. 2008), and N availability and N demand by plants are both strongly affected by the seasonal distribution of rainfall (Borken and Matzner 2009). Thus, it is logical that any relationships between the eCO₂ effect on grassland biomass and water supply should vary seasonally. Indeed, results from the TasFACE experiment demonstrated that eCO₂ stimulated grassland biomass production by an average of 13.9%, but this effect was heightened in years with more rainfall in summer and suppressed in years with more rainfall in autumn or spring (Hovenden et al. 2014). In addition, the differences among experiments in average grassland biomass response to eCO₂ are strongly related to seasonal precipitation (Hovenden et al. 2019), indicating that the season in which precipitation occurs interacts with the mechanisms whereby eCO₂ increases grassland biomass production.

Ultimately, one of the major objectives of investigating eCO₂ effects on biomass production is to provide improved understanding and data to refine and constrain land–atmosphere and global climate models (De Kauwe et al. 2013, 2014, 2017; Zaehle et al. 2014; Medlyn et al. 2015; Ryan et al. 2017). However, the estimation of carbon balances to the rising [CO₂] requires a detailed understanding of belowground as well as aboveground responses. This is particularly the case with grasslands, in which more than 50% of the biomass typically occurs belowground (Canadell et al. 1995; Carol Adair et al. 2009). Unfortunately, the difficulties inherent in estimating belowground biomass mean that there are far fewer reports of belowground responses to eCO₂ (e.g. Owensby et al. 1993; Pineiro et al. 2017) than of aboveground responses (Ainsworth and Long 2005; Lee et al. 2010) and many estimations of vegetation responses to eCO₂ rely solely on aboveground measurements. As biomass responses are often used to infer the impacts of rising [CO₂] on carbon balance and sequestration, this could be problematic if belowground responses differ to those observed aboveground. Indeed, there is evidence that root responses to eCO₂ are very variable (Arnone et al. 2000) and are dependent upon prevailing soil nutrient and moisture conditions. Assumptions that root responses to eCO₂ are stimulated by limiting water and nutrient supply need to be tested directly (Madhu and Hatfield 2013). Pineiro et al. (2017) demonstrated that eCO₂ effects on fine root growth decline with reducing water availability (Pineiro et al. 2017), in direct opposition to aboveground responses to eCO₂, which are believed to increase as water availability declines (Fatichi et al. 2016). Therefore, relying solely on assessments of eCO₂ effects on aboveground productivity could be misleading if total net primary productivity or carbon balance are of interest.

In an effort to reduce the impact of harvesting on experimental results, many manipulative experiments in grasslands collect an annual productivity estimate at peak biomass based on samples collected at one or very few times. Results from other experiments are often analysed and presented in terms of annual production, even if biomass is sampled many times per year. Therefore, despite evidence that the CO₂ effect on grassland biomass production might be strongly seasonal and may be obscured by annual figures (Hovenden et al. 2014; Hovenden et al. 2019), few analyses have been designed specifically to determine how seasonal precipitation or irrigation influences the overall biomass response to eCO₂ in grasslands.

Here, we use an experiment designed to investigate how alterations of seasonal water supply influence the biomass response of an agriculturally important grassland to eCO₂. The TasFACE2 experiment was established in 2015 as the world’s only free air CO₂ enrichment (FACE) experiment in which soil water supply was completely manipulated with no influence of precipitation on soil water content (Brinkhoff et al. 2019). By collecting relatively high-frequency estimates of above- and belowground biomass production, we aim to test whether limitation of water supply or the provision of excess water have different impacts on the biomass response to eCO₂ in different seasons. By doing so in a system that is fertilised regularly, we reduce the impact of nutrient limitation on the growth response to eCO₂, thus enabling the impacts of seasonal water availability itself to be determined. We also test whether the growth response to eCO₂ is linear in this system by testing at three, rather than the usual two, CO₂ concentrations. Specifically, here we test the following hypotheses: (1) that the stimulation of aboveground biomass production by eCO₂ will be greater when water supply is restricted; (2) that the stimulation of belowground biomass production by eCO₂ will be lower when water supply is restricted; (3) that the extent to which (1) and (2) occur will depend upon the season in which water supply is restricted; and (4) because of the divergent impacts of water restriction on the eCO₂ effect on above- and belowground biomass production, that the total (i.e. above- + belowground) biomass response to eCO₂ will not be predictable from aboveground responses.

Materials and methods

Results

Aboveground biomass and soil water content

Elevated CO₂ did not stimulate daily aboveground biomass production of perennial ryegrass in any watering treatment at any time in the TasFACE2 experiment (Fig. 1a). In fact, daily aboveground biomass production was reduced by eCO₂ in 6 of the 12 harvest periods with there being no impact of CO₂ treatments in the remainder of sampling period (Fig. 1a). Importantly, irrigation treatment had no influence on the CO₂ effect on aboveground biomass production (CO₂ × Irrigation P = 0.99). Reductions in aboveground biomass production by eCO₂ seemed to be limited to summer, autumn and winter, with no differences in aboveground biomass among FACE treatment levels in spring (Fig. 1a). The impact of a 75 μmol CO₂ mol⁻¹ increase in [CO₂] was mostly similar to the impact of an increase of 150 μmol CO₂ mol⁻¹, indicating that the CO₂-response of aboveground biomass production was not dose-related (Fig. 1a). When considered cumulatively over the entire 19 month sampling period, plots exposed to 400 μmol CO₂ mol⁻¹ produced 2868 ± 95.8 g DM m⁻² aboveground biomass whereas those exposed to 475 μmol mol⁻¹ produced 16.0% less (2409 ± 52.9 g DM m⁻²) and those exposed to 550 μmol CO₂ mol⁻¹ produced an essentially identical 2420 ± 90.7 g DM m⁻², a reduction of 15.6% from the controls.

Fig. 1.

The impact of (a) [CO₂] and (b) irrigation treatments on aboveground biomass production in the TasFACE2 experiment. Main panels display daily dry matter production; insets show total dry matter production over the sample period. Those plots receiving limited water supply received 40% less water than those in adequate plots and those receiving excess received 25% more. Values shown are means ± s.e. (n = 4 in (a), n = 12 in (b)).

CO₂ treatments and irrigation both influenced soil water content (SWC) but this effect varied among seasons (Fig. 2). The seasonal restriction of irrigation had obvious impacts on soil water content (SWC) in the top 10 cm of the soil at each [CO₂], demonstrating the effectiveness of the irrigation treatments (Fig. 2a), with plots receiving limited irrigation having ~6% lower SWC than the other plots (Fig. 2a). SWC did not differ between sectors receiving excess irrigation and those receiving adequate irrigation in any season (Fig. 2a). Although SWC varied substantially over time, eCO₂ tended to increase SWC; but this effect varied seasonally (Fig. 2b). Overall, the soils was moister in plots at 475 and 550 μmol CO₂ mol⁻¹ than in control plots in autumn and winter but in spring only plots at 550 μmol CO₂ mol⁻¹ had higher SWC than control plots, with plots at 475 μmol CO₂ mol⁻¹ lying in between. Interestingly, there was no effect of eCO₂ on SWC during summer (Fig. 2b).

Fig. 2.

The impact of (a) irrigation and (b) [CO₂] on mean seasonal soil water content in the uppermost 10 cm in the TasFACE2 experiment. Those plots receiving limited water supply received 40% less water than those in adequate plots and those receiving excess received 25% more. Values shown are means ± s.e. (n = 12 in (a), n = 4 in (b)).

Irrigation treatments also did not influence the daily aboveground biomass production in the TasFACE2 experiment (Fig. 1b), nor did they interact with CO₂ treatment. In general, the moister the soil, the lower the production rate of aboveground biomass (Fig. 3). In autumn and winter, the reduction in aboveground biomass production by increasing SWC was consistent across CO₂ levels (Fig. 3). In spring, however, aboveground biomass production was not affected by SWC at any [CO₂] (Fig. 3). The only substantial influence of [CO₂] on the relationship between SWC and aboveground biomass production occurred during summer, at which time increasing SWC tended to reduce aboveground biomass production in eCO₂ plots but not in control plots, as indicated by the negative slope of the relationship between SWC and aboveground biomass (Fig. 3). Therefore, increasing soil water content tended to reduce aboveground biomass production in autumn and winter in all plots and during summer in plots exposed to eCO₂.

Fig. 3.

The influence of soil water content (SWC) on mean daily aboveground biomass production as a function of [CO₂] in the TasFACE2 experiment. Values shown are average slopes with 95% confidence limits of the relationship between aboveground biomass production and SWC. The dashed line shows a slope of zero. Positive values indicate a stimulatory effect; negative values indicate a suppressive effect of increasing SWC.

Root biomass

The lack of any stimulation of aboveground biomass by eCO₂ was not reflected in root biomass production (Fig. 4). Indeed, there was a strong impact of FACE treatment on root biomass production (P = 0.01), but this effect was mediated both by time and the irrigation treatment (P < 0.02). Root biomass production tended to increase with increasing [CO₂] more or less linearly and rather consistently over time (Fig. 4a) in plots provided with adequate irrigation, whereas limiting irrigation in some seasons and providing excess in others had complex impacts on root growth (Fig. 4b–d).

Fig. 4.

The impact of [CO₂] on root growth in the TasFACE2 experiment, as a function of irrigation treatments. Main panels display daily dry matter production; insets show total dry matter production over the sample period. Those plots receiving limited water supply received 40% less water than those in adequate plots and those receiving excess received 25% more. Values shown are means ± s.e. (n = 4).

When treatment effects on SWC were taken into consideration, there was a consistent and positive effect of eCO₂ on root biomass production (Fig. 5a). When compared at overall mean SWC, root biomass production at 550 μmol mol⁻¹ was 63% greater than that at 400 μmol CO₂ mol⁻¹ (Fig. 5a). Further, there was a positive relationship between SWC and root biomass production (F_1,274 = 12.9, P < 0.001) in all seasons (Fig. 5b). Overall, therefore, moister soils led to greater root biomass production, particularly in spring, but there was a strong, independent stimulatory effect of eCO₂ (Fig. 5a).

Fig. 5.

(a) Mean daily root production at overall mean soil water content (SWC) as a function of [CO₂] in the TasFACE2 experiment. Values shown are marginal means with 95% confidence limits. (b) The influence of increasing SWC on mean daily root production as a function of season in the TasFACE2 experiment. Values shown are the slopes with 95% confidence limits of the relationship between the natural logarithm of daily root biomass production and SWC in each season.

Root:shoot and total biomass

Plants growing in eCO₂ plots had a considerably increased root-to-shoot ratio compared to plants growing in control plots across all harvest periods and all irrigation treatments (P < 0.01; Fig. S2), although the degree of stimulation of root-to-shoot ratio by eCO₂ varied both seasonally (P < 0.001) and with irrigation treatment (P = 0.01). The mean root-to-shoot ratio of plants supplied with adequate irrigation was 187% greater at 550 μmol CO₂ mol⁻¹ than at 400 μmol CO₂ mol⁻¹ but in plots receiving less than adequate irrigation in any season, the root-to-shoot ratio at 550 μmol CO₂ mol⁻¹ was approximately double that at 400 μmol CO₂ mol⁻¹ (Fig. S2). Similarly, although eCO₂ increased the root-to-shoot ratio across the year, the stimulation was greatest in spring (Fig. S2).

Limiting irrigation or supplying excess water in any season altered the relationship between [CO₂] and total biomass production (Fig. 6b–d), which was remarkably linear in plots supplied with adequate water at all times (Fig. 6a). Thus, elevating the [CO₂] to 475 μmol CO₂ mol⁻¹ had virtually no impact on total biomass production in any plots in which irrigation was restricted at any time of year (Fig. 6b–d). The situation at 550 μmol CO₂ mol⁻¹, however, was quite different with some increase in total biomass production occurring in all plots and substantial increases occurring in two of the four irrigation treatments (Fig. 6). Thus, CO₂-response of total biomass production in this ecosystem was strongly non-linear when water supply was limited or excessive in any season.

Fig. 6.

The impact of [CO₂] on cumulative total biomass (i.e. above- + belowground biomass) production in the TasFACE2 experiment, as a function of irrigation treatments. Those plots receiving limited water supply received 40% less water than those in adequate plots and those receiving excess received 25% more. Values shown are means ± s.e. (n = 4).

When compared at a common SWC, eCO₂ tended to increase total biomass production, most consistently during spring (Fig. 7a), with the response being strongly variable in both winter and summer. There was a striking effect of season on the influence of [CO₂] on the slope of the relationship between total biomass production and SWC (Fig. 7b). In autumn, winter and spring, increasing [CO₂] tended to increase the slope of the relationship of SWC with biomass, but this effect was most pronounced in the spring (Fig. 7b). During summer, eCO₂ had little impact on the relationship between SWC and biomass, but what influence there was tended to be negative (Fig. 7b). Importantly, it was only during spring that total biomass production increased with increasing SWC, as in all other seasons, total biomass production tended to be reduced by increases in SWC (Fig. 7b).

Fig. 7.

(a) Daily total biomass (i.e. above- + belowground biomass) production at overall mean soil water content (SWC) as a function of [CO₂] and season in the TasFACE2 experiment. Values shown are marginal means with 95% confidence limits. (b) The influence of increasing SWC on daily total biomass production as a function of season and [CO₂] in the TasFACE2 experiment. Values shown are the slopes with 95% confidence limits of the relationship between the natural logarithm of daily total biomass production and SWC in each season at each [CO₂]. The dashed line shows a slope of zero.

Discussion

In this study, we sought to determine whether experimentally-applied seasonal shortages or excesses in water supply altered the growth responses of a widely-sown C₃ pasture grass to elevated CO₂ concentrations. As hypothesised, we found that the growth responses to eCO₂ above ground differed markedly to those below ground, but the details of the responses themselves were quite unexpected. Our first hypothesis, that eCO₂ will stimulate aboveground biomass production more when water is limited was not proven, as this aspect of production was never stimulated by eCO₂ at TasFACE2. We did, however, find strong support for our second hypothesis, namely that eCO₂ will stimulate below ground biomass production less when water supply is limited.

In our experiment, aboveground biomass production was never increased by eCO₂, regardless of seasonal water supply, completely contrary to our first hypothesis. In contrast, there were substantial interactions between [CO₂] and irrigation on root growth. Generally, root growth increased with increasing [CO₂] but the degree of growth response depended strongly upon the irrigation treatment. Root growth in perennial ryegrass is strongly dependent upon the supply of C from the shoot (Robin et al. 2018), so it is logical that root growth will respond strongly to eCO₂ in this species. Our analyses revealed complex relationships between season, soil water content and biomass production that influenced the CO₂ response, meaning the results did not comply with our second or third hypotheses either. Finally, the results confirmed that total biomass responses could not be inferred from aboveground responses, complying with our fourth hypothesis, but not for the reasons we suggested.

Overall, the results from the TasFACE2 experiment delivered several important outcomes. First, total biomass (i.e. the sum of above and belowground biomass) production increased almost perfectly linearly with [CO₂] when plots were supplied with adequate irrigation in all seasons. In fact, total biomass increased by 14.0% when the [CO₂] was increased by 18.75% from 400 to 475 μmol CO₂ mol⁻¹. A further increase in the [CO₂] from 475 to 550 μmol CO₂ mol⁻¹, led to an additional 13.6% increase in total biomass, giving an overall productivity increase of 29.5% for a 150 μmol CO₂ mol⁻¹, or 37.5%, increase in [CO₂]. This is almost exactly the figure (31%) reported by Ainsworth and Long (2005) as the mean stimulation of light-saturated photosynthetic carbon assimilation of C₃ plants in FACE experiments. Moreover, a biomass increase of 29.5% is close to the theoretically calculated value for the direct growth stimulation for Lolium perenne, given the photosynthetic response to eCO₂ of ~40% in this species (Ainsworth et al. 2003; Ainsworth and Long 2005). Thus, in conditions of continuously adequate water supply and non-limiting nutrient availability, L. perenne plants in the TasFACE2 experiment increased their whole-plant growth rate in line with the eCO₂ stimulation of photosynthetic carbon assimilation, demonstrating that substantial increases in grassland biomass production are possible at elevated [CO₂], but also that this does not necessarily result in increased herbage production.

Second, the unique design of TasFACE2, permitting complete control of water supply, allowed us to determine that withholding water in one season, and supplying it in excess in another season, fundamentally changed the shape of the relationship between total biomass production and [CO₂]. In fact, increasing the [CO₂] from 400 to 475 μmol CO₂ mol⁻¹ resulted in no change in total biomass production if water supply was anything other than adequate in all seasons. This was unexpected because the water-savings effects of reduced stomatal conductance at eCO₂ were expected to lead to greater effects of eCO₂ on biomass production in periods during which water supply was less than adequate. It was clear from the results shown here that this was not the case, even when water was withheld in the warmest months. Our analysis of the relationships between soil water content and biomass production are key to understanding these findings.

Root biomass production was obviously strongly seasonal in L. perenne at this site, with highest growth rates occurring during spring. As previously suggested (Arnone et al. 2000; Pineiro et al. 2017), limiting water supply during this period of peak root growth reduced the belowground biomass response to eCO₂, because of the strong sensitivity of root growth to soil water content during spring (Fig. 5). Hence, the reaction of root growth to [CO₂] during the spring dominated the biomass responses. Root growth was increased by 81% (475 μmol CO₂ mol⁻¹) and 141% (550 μmol CO₂ mol⁻¹) in plots supplied with adequate irrigation in the spring, but CO₂ treatments had much less impact in plots receiving less than adequate irrigation, because the reduction in soil water content restricted root biomass production (Fig. 5b). What is also apparent is that under ‘ideal’ conditions, the root growth response to [CO₂] in L. perenne was linear up to 550 μmol CO₂ mol⁻¹, as has been demonstrated before (Anderson et al. 2010), and varying irrigation from adequate in any season alters this relationship. This influence of water supply on the response to [CO₂] is supported by other studies demonstrating that interactions among environmental factors can lead to substantial non-linearity of responses (Ackerly and Bazzaz 1995; Gill et al. 2002; Zhou et al. 2008; Zhu et al. 2016).

Our analyses demonstrate that the strong impact of eCO₂ on root biomass production at TasFACE2 was due to (1) a direct stimulatory effect of eCO₂, (2) the positive relationship between root production and increasing soil water content and (3) eCO₂-driven increases in soil water content. It is also evident that the large stimulation of root biomass production by eCO₂ in spring coincided with a substantially increased sensitivity of root biomass production to soil water content. Importantly, we found no evidence that eCO₂ altered the relationship between root production and soil water content. Instead, eCO₂ drove increases in root growth directly and by increasing soil water content.

Third, whereas both total biomass and root growth responses to eCO₂ were strong, at least at 550 μmol CO₂ mol⁻¹ and in some irrigation treatments, biomass allocation to aboveground plant parts was either unaffected or reduced by eCO₂. This is a surprising result, as it would be expected that shoot growth would be stimulated by eCO₂ in this fast-growing pasture grass, particularly under the conditions of relatively high nutrient availability. The reduction in aboveground growth occurred in all seasons but was least pronounced during the spring. Previous studies that have shown an overall suppression of aboveground production by eCO₂ have linked this to low productivity and overall growth (Song et al. 2019), but this is not the case here where total biomass was stimulated and aboveground biomass production rate was still high. The situation at TasFACE2 appears to be more a matter of altered biomass allocation, as the total biomass response to eCO₂ was strong. An increase in biomass allocation to roots at eCO₂ has been observed in many studies (Sage 1994; Anderson et al. 2010; Arndal et al. 2014, 2018), but this generally does not occur to the extent that aboveground growth is reduced. Further, shoot growth rate was high in all treatments, with annual herbage yield being well over 20 t ha⁻¹ in all plots, which is comparable to local production in optimal conditions (Rawnsley et al. 2014).

The reduced allocation to aboveground growth was unlikely to have been driven by P availability (Conroy et al. 1992) because the experiment was supplied with more than adequate P fertiliser (Rawnsley et al. 2014). Soil tests at the experimental site prior to establishment indicated that nutrients other than N and P were sufficient to support substantial growth, so it is unlikely that the results were due to any form of micronutrient deficiency.

Rather, it appears that the eCO₂-derived suppression of aboveground biomass production was due to a fundamental relationship between soil water content and aboveground growth and eCO₂ effects on soil water content. In autumn, winter and summer, aboveground growth at TasFACE2 tended to reduce as soil water content increased, probably because of increased allocation belowground. Thus, increases in soil water content driven by eCO₂ would lead to reductions in aboveground growth. Further, eCO₂ increased the sensitivity of aboveground biomass production to soil water content during summer, meaning that eCO₂ reduced aboveground biomass production independent of treatment effects on soil water content.

It is unclear why eCO₂ led to the reduction in shoot growth in this experiment. It is possible that eCO₂-induced alteration of shoot morphology drove the shoot growth response (Brinkhoff et al. 2019), possibly due to an increase in self-shading associated with the eCO₂-driven increase in tiller branching. Many long-term eCO₂ experiments in herbaceous vegetation have found that eCO₂ suppresses aboveground growth in particular years (Larsen et al. 2011; Hovenden et al. 2014; Andresen et al. 2018), including in sites with high productivity. Factors such as insufficient N supply and water deficit can prevent eCO₂ from stimulating biomass production (Reich et al. 2014), and eCO₂-driven reductions in soil N availability (Luo et al. 2004; Hovenden et al. 2008, 2017; Newton et al. 2010) could potentially lead to suppression of growth. Altering the seasonal water supply at TasFACE2 did change seasonal aboveground production, but this response was not influenced by [CO₂]. Thus, eCO₂ suppressed aboveground biomass production in the TasFACE2 experiment irrespective of water supply. The mechanisms leading to the reductions in shoot growth are still unknown and must surely be a matter of future research concentration, particularly as this could lead to a diminution of some of the supposed benefits from the rising [CO₂] (Hovenden and Newton 2018).

The pasture at TasFACE2 had only been established for 2 years by the end of the study period, so it is possible that experimental results from a long-established pasture would differ. Although the sward appeared fully developed, the root systems might still have been establishing 2 years after sowing. Thus, the preferential allocation of biomass to the roots, which was increased by eCO₂, might not be as pronounced in long-established pastures (Luscher et al. 2000). However, high-input pastures are regularly re-sown, at approximately 10-year intervals in intensively-managed sites, to introduce newly released varieties, to manage species composition or invigorate productivity or as part of rotational practices, so the results presented here are relevant to the grazing industry. Moreover, longer-term experiments in similar ecosystems (Lüscher et al. 1997; Obermeier et al. 2017; Andresen et al. 2018) do not show substantial changes in eCO₂ effects after the initial years, so it is possible that the results shown here would persist over the longer term. It is also possible that the restriction in irrigation applied in this experiment was not sufficient to cause marked water stress. Greater restriction of irrigation might have contributed to a greater impact of eCO₂ on biomass production, as benefits from eCO₂-derived water savings would only be likely to occur at lower soil water contents.

The climatic conditions during the experiment did not deviate greatly from long-term means (Table S1), although the impact of climate change is evident in the slightly warmer conditions (+0.6°C at night and +0.8°C in the day) during the study period than the average from the previous 30 years. Therefore, it is unlikely that climatic peculiarities drove the patterns observed. The substantial site works involved in establishing the experiment could have caused some degree of soil compaction and this might have influenced soil hydraulic conductivity and root growth patterns compared to a less-disturbed site. However, intensively-managed pastures tend to have high levels of traffic from stock and vehicles, so potential alterations of soil bulk density caused by the establishment of the experiment would not be as extreme as might be the case in a less intensively-managed ecosystem. Finally, the rainfall exclosures themselves caused some shading, potentially reducing leaf temperature as well as light exposure. All experimental plots had the same infrastructure, but it is possible that the light exposure varied among sub-plots within each plot. The randomisation of irrigation treatments among the sub-plots should have prevented any confounding effects but it is possible that some degree of difference in light exposure among the replicate sub-plots could have obscured the effects of irrigation. However, the fact that the site is on a gently north-facing (i.e. sunward) slope in a region with naturally high insolation, means that the implications of shading are likely to be minor.

The linear relationship between [CO₂] and total biomass production in sectors supplied with adequate irrigation in all seasons indicates both that the experiment was operating effectively and that the direct stimulation of photosynthesis by eCO₂ was as expected. Few other studies have investigated grassland biomass production along a [CO₂] gradient (Johnson et al. 1993; Anderson et al. 2001; Fay et al. 2015; Polley et al. 2019). Our results show that underlying relationships between biomass production rates and soil water content can influence the outcome of eCO₂ treatments on root and shoot growth, and provide strong evidence that these relationships vary seasonally. Our analyses allowed us to compare biomass production rates at a common soil water content, thus separating the influence on growth of eCO₂-driven differences in soil water content from more direct eCO₂ influences. Disentangling the exact mechanisms leading to the strongly disparate and seasonal relationships between eCO₂, soil water content and above- versus belowground growth will improve our ability to predict future agricultural production and carbon cycle responses to the rising [CO₂].

Supplementary material

Supplementary material is available online.

Data availability

The data supporting the findings of this study are openly available via the University of Tasmania Research Data Portal (https://rdp.utas.edu.au).

Conflicts of interest

The authors declare no conflicts of interest.

Declaration of funding

This research was supported by the Australian Research Council Discovery Projects Scheme grant numbers DP150102426 and DP190102025.

Author contributions

MJH and RPR designed the study. REB, MP and AF maintained the experiment and ALS, REB, KS, ZAB, MP, AF, MN and MJH collected the data. MJH, ALS and ZAB analysed the data and all authors contributed to interpretation of the results. MJH, ALS and RPR wrote the manuscript with assistance from all authors. All authors gave their final approval for publication.