Benefits of increasing transpiration efficiency in wheat under elevated CO2 for rainfed regions

Higher transpiration efficiency (TE) has been proposed as a mechanism to increase crop yields in dry environments where water availability usually limits yield. The application of a coupled radiation and TE simulation model shows wheat yield advantage of a high‐TE cultivar (cv. Drysdale) over its almost identical low‐TE parent line (Hartog), from about −7 to 558 kg/ha (mean 187 kg/ha) over the rainfed cropping region in Australia (221–1,351 mm annual rainfall), under the present‐day climate. The smallest absolute yield response occurred in the more extreme drier and wetter areas of the wheat belt. However, under elevated CO2 conditions, the response of Drysdale was much greater overall, ranging from 51 to 886 kg/ha (mean 284 kg/ha) with the greatest response in the higher rainfall areas. Changes in simulated TE under elevated CO2 conditions are seen across Australia with notable increased areas of higher TE under a drier climate in Western Australia, Queensland and parts of New South Wales and Victoria. This improved efficiency is subtly deceptive, with highest yields not necessarily directly correlated with highest TE. Nevertheless, the advantage of Drysdale over Hartog is clear with the benefit of the trait advantage attributed to TE ranging from 102% to 118% (mean 109%). The potential annual cost‐benefits of this increased genetic TE trait across the wheat growing areas of Australia (5 year average of area planted to wheat) totaled AUD 631 MIL (5‐year average wheat price of AUD/260 t) with an average of 187 kg/ha under the present climate. The benefit to an individual farmer will depend on location but elevated CO2 raises this nation‐wide benefit to AUD 796 MIL in a 2°C warmer climate, slightly lower (AUD 715 MIL) if rainfall is also reduced by 20%.

efficiency). The nature of TE, comprising components of biomass and transpiration, contributes to it being a complex trait from a breeding point of view . This complexity is increased because of TE's dependence on environmental factors like vapor pressure deficit (VPD) of the atmosphere, whereby TE is reduced as VPD increases (Tanner & Sinclair, 1983). As such, it is difficult to separate the genetic components of TE from the environmental components, but as Sinclair (2012) showed, it is possible to do this by defining TE as an inverse function of VPD (TE = k d /VPD).
The resulting crop-dependent transpiration coefficient (k d ) offers a way to normalize TE against changing VPD. Under typical dryland field conditions, TE for wheat varies from around 3 to 9 g of biomass growth per kg water transpired with k d typically ranging from 4 to 6 Pa (Kemanian, St€ ockle, & Huggins, 2005). Older cultivars appear to have a lower potential ("unstressed") TE (i.e., when measured under ample water supply and low VPD) than more recently released cultivars, and that has been attributed to rising atmospheric CO 2 concentrations over recent years (Fletcher & Chenu, 2015). Crop TE itself can vary more than the assumed constant k d , particularly when water stressed under high VPD or under increasing atmospheric CO 2 . However, varied assumptions of total biomass accumulation (i.e., shoot + root or shoot only) and consistent VPD algorithms for the sampling period contribute to the observed variance for which more complex models of TE apply (Kemanian et al., 2005).
The stable isotope (d 13 C) signature in biomass has been used as a surrogate for leaf-level TE, to produce high-TE wheat cultivars (Condon, Richards, Rebetzke, & Farquhar, 2004;Rebetzke, Condon, Richards, & Farquhar, 2002). This was based on the original work relating carbon isotope discrimination (CID) to TE in wheat (Farquhar & Richards, 1984). This original work also delineated the important distinction between carbon isotope composition, d 13 C, and the biologically meaningful process of CID. Field evaluation of the high-TE cultivar (Quarrion) showed significant gains in TE (11%-21%) over a season measured against a low-TE cultivar (cv. Matong) but realizing final yield was more complex than just biomass gain . The complexity comes from disproportionate changes in transpiration and biomass, with changes in assimilate partitioning considered independent of those factors that primarily control TE (e.g., VPD). This partitioning is particularly important when considering grain yield, because grain growth occurs later in the season when crops are typically water-limited and experiencing terminal drought and crops have the capacity to shift varying amounts of C assimilated earlier in the season into the grain. This early work showed significant advantage of related high-TE lines over low-TE lines in drier locations (e.g., Rebetzke et al., 2002) but the benefits over a wider range of environmental conditions remain largely unknown. An independent assessment of the value of TE was undertaken separately by Australian Grains Technology â breeding company. They grew a high-TE cultivar, Drysdale, side-by-side with its closely related low-TE parent Hartog across 60 site-year combinations throughout the Australian wheat belt (Rebetzke et al., 2009).
Their studies confirmed the significant yield benefit of greater TE across a broader range of environments ranging in yield from 0.3 to 6 t/ha, and a particular benefit of high TE at yields of 4 t/ha and below.
Among changing environmental factors, increasing atmospheric CO 2 will increase leaf-level TE for virtually all plants because elevated CO 2 promotes C assimilation and at the same time decreases stomatal conductance and therefore transpiration.
Recent work from the Australian Grains Free-Air Carbon Dioxide Enrichment (AGFACE) facility provided some evidence that a high-TE trait might still be an advantage under higher atmospheric CO 2 concentrations (Tausz-Posch, Norton, Seneweera, Fitzgerald, & Tausz, 2013;Tausz-Posch, Seneweera, Norton, Fitzgerald, & Tausz, 2012). High-TE cultivar Drysdale was grown side-by-side with low-TE parent Hartog and while both cultivars in this analysis had improved TE under elevated CO 2 , differences between the two cultivars indicated greater TE for Drysdale with growth under elevated CO 2 potentially increasing the response of this trait. The exact mechanisms for this increasing advantage were not entirely clear (Tausz-Posch et al., 2012). Additionally, the AGFACE experiment does not fully represent a future climate even at its present location in Horsham, Australia. We therefore used crop simulation modeling to (1) better understand the genetic and environmental components of TE in these experiments, and (2) extrapolate experimental observations to other environments beyond this site.
Specifically, we explored potential benefits of increased TE in wheat across the wheat growing areas of Australia employing a validated model considered sufficiently mechanistic to model water-limited wheat crop growth and yield. We consider these effects under the present climate and likely future warmer and drier climate scenarios under elevated atmospheric CO 2 concentrations. This study region represents a large proportion of global wheat production areas and is typical of many rainfed cropping environments throughout the world experiencing significant changes in climate (e.g., CIMMYT Mega environment 1, 2, 4, and 8; CIMMYT 2014).

| MATERIALS AND METHODS
We reanalyzed the published data from Tausz-Posch et al. (2012) using the CAT-wheat model (O'Leary et al., 2015). The model (CAT-Wheat) is a landscape-scale model that has recently been successfully tested against other AGFACE data (O'Leary et al., 2015). The advantage of the CAT model is its unique feature of analyzing crop performance across diverse landscapes (Christy et al., 2013).

| Experimental site and growing conditions
A detailed description of the site set-up is given in Mollah, Norton, and Huzzey(2009). Briefly, the Australian Grains Free Air CO 2 Enrichment (AGFACE) facility is located at Horsham, Victoria, Australia (36°45 0 07″S, 142°06 0 52″E, 127 m above mean sea level). The site is a clay vertosol (Isbell, 1996), which has~35% clay at the soil surface increasing to 60% at 1.4 m depth. Four ambient CO 2 (a  Table 1). In addition, each of the three experimental series was run under rainfed conditions and supplemental irrigation. This resulted in three additional sets of environmental growing conditions (Table 1). Closely related cultivars "Drysdale" and "Hartog" were sown into flat beds at 0.27 m row spacing on the plots. The high-TE Drysdale was selected from a backcross-2 population (Hartog 9 3/Quarrion) using Hartog as the low-TE recurrent parent. Summarizing, a total of six different environments were tested under both ambient CO 2 (375 mol/mol) and elevated CO 2 (550 mol/mol) concentrations, using differing combinations of water supply, seasonal, and sowing date variations (Table 1).

| Biomass, grain yield, and morphological measurements
For each experimental series, total above-ground biomass (leaves, stems, spikes) was measured at flowering (growth stage DC65; Zadoks, Chang, & Konzak, 1974) and at physiological maturity (DC90). As phenological development was similar for both cultivars and in both CO 2 treatments, both cultivars grown under ambient and elevated CO 2 were sampled on the same dates. At each sampling date, plants from 0.5 m 2 of each subplot ("whole sample") were handharvested and then dried for 72 hr at 70°C (DC65) or 40°C for 72 hr (DC90). The DC65 samples were weighed while DC90 samples processed for grain yield, above-ground biomass and 1,000 kernel weight. All parameters were expressed on an area basis (m 2 ).

| The CAT-Wheat model
The CAT-Wheat model originally calculated crop growth by a radiation use efficiency (RUE) approach whereby reduced water supply, nutrient stress or photoperiod factors reduced RUE by relative differences (O'Leary et al., 2015). The minimum value of these factors (i.e., the most limiting) was used to reduce RUE. While that version of the model performed satisfactorily in testing the response to elevated CO 2 , it did not simulate transpiration reduction directly as a consequence of high CO 2 . The model was modified here to increase its utility in simulating water dynamics similar to other contemporary models. Initial parameter adjustments were made using the OLEARY-CONNOR model (O'Leary & Connor, 1996) utilizing parameters driven by TE for determining growth, and software coding modifications to establish realistic amendments that were subsequently transferred to the landscape CAT-Wheat model. These two models are not identical and differ in assimilate partitioning and water and nitrogen simulation calculations.
Transpiration reduction and efficiency changes due to elevated CO 2 were added to CAT based on a formulation adapted from St€ ockle, Williams, Rosenberg, and Jones (1992) as incorporated into the CROP-SYST (St€ ockle, Donatelli, & Nelson, 2003) and STICS (Brisson et al., 2003) models. A correction to RUE is applied and the associated correction to TE, without double accounting, was made after O'Leary T A B L E 1 Summary of data underlying the results reported in Tausz-Posch et al. (2012). Three times of sowing, growing season rainfall, and irrigation applied. Observed above-ground biomass (at DC65 and DC90) and grain yield for both Hartog and Drysdale at ambient and elevated CO 2 concentrations (averages from four replicates each) (1) ( 2) where, CO 2 is the atmospheric CO 2 concentration (lmol/mol) and c Previous studies have demonstrated a yield advantage between 2% and 15% for low-CID lines (Drysdale) (Rebetzke et al., 2002(Rebetzke et al., , 2009) when compared with high discrimination lines (Hartog). The yield advantage of low-CID selected lines was associated with increases in aerial biomass and greater partitioning of dry matter to grain. Therefore, to simulate these increases, the crop transpiration coefficient (k c ) of Drysdale was increased from the CAT-Wheat default of 5.8-6 Pa while that of Hartog was decreased to 5.2 Pa.
These changes reflected the biomass growth response of the experimental observed ambient CO 2 data. An increase in TE is typically simulated with an approximately equal weighting to growth increase and transpiration decrease. We further increased the weighing on transpiration decrease for Drysdale by increasing the canopy resistance (r s ) from 36 (for Hartog) to 44 s/m (for Drysdale), thereby further altering the calculation of potential evapotranspiration by CAT and the TECC response in Equation (6) where, r s is the crop canopy resistance (s/m).

| The CAT-Wheat model ̶ Long-term analysis at Horsham
The CAT-Wheat model was applied to consider the productivity change resulting from increasing TE in wheat at the FACE experi- To reduce the confounding effect of "carry-over" stored soil water from the previous year's crop, the stored soil water status was reset 75 days before sowing to 10% plant available water for each soil depth increment to a total depth of 1 m, and a full plant available water profile below that depth.
2.6 | The CAT-Wheat model ̶ Spatial analysis across

Australia
The long-term analysis at Horsham was extended across Australia over a 54-year period using historic climate and the two additional climate sequences using the same method applied for the long-term analysis at Horsham. Model simulation was conducted on all privately owned, arable agricultural land (defined as slopes <5%) within the spatial region identified in Figure 1. The spatial area was divided into 1-km 2 grid cells for modeling. For each grid cell within this region, the CAT-Wheat model was applied for the two wheat cultivars sown each year after the "autumn-break," with crop yield response demonstrating intraseasonal variability associated with climate patterns and soil water availability.
For the long-term, single site at Horsham, 648 annual simulations were conducted to represent all scenarios. Upscaling to a total area evaluated of 71.3 9 10 6 ha ( Figure 1), the CAT-Wheat model simulated 462,085,560 years of wheat growth, by being applied to each 1 km 9 1 km grid cell on a daily basis using the 54 years of historic climate data , scaled to each grid pixel from the meteorological climate station of nearest proximity.   (Table 2), the TE coefficient and crop canopy resistance. The exception was the grain yield CO 2 response for Drysdale ( Figure 6f) with its observed slope of grain response to elevated CO 2 being 27.5% above unity compared to the modeled 18.3%.

| Long-term responses at Horsham
We applied the simulated trait differences between Drysdale and Hartog over 54 years of present and future climate scenarios at Horsham ( Figure 4). As sowing data were based on an amount of rainfall falling at an "autumn-break," there was no impact on the average sowing date by raising temperature by 2°C. However, the third climate sequence of reduced rainfall delayed the average sowing date by 16 days (Table 3). The increase in temperature did decrease the length of growing season which resulted in reduced growing season rainfall for the two climate sequences which were elevated by 2°C (Table 3).
Irrespective of the negative and positive effects of temperature and rainfall, the increased TE trait appears of consistent benefit at Horsham. The yield advantage ranged from around 127 to 152 kg/ ha under present-day CO 2 concentration and a likely future 550 ppm scenario, respectively (Figure 4a,b).

F I G U R E 1
The area evaluated above 220 mm/year rainfall within wheat belt for the two wheat cultivars and the location of the Australian Grains Free-Air Carbon Dioxide Enrichment experimental site used for CAT-Wheat parameterization. Graph shows frequency distribution of mean annual rainfall (mm/year, 1975-2005) across study region. Validation site symbols indicate the 60 locations of the validation data for Hartog and Drysdale shown in Figure 2d The variation in yield advantage of Drysdale under the likely future 550 ppm scenario had a wider distribution of response compared to current climate, with a greatly enhanced (>10%) yield in 22% of the years tested. However, in poorer seasons this yield advantage tended toward a yield loss (5% of years), reflecting a possible resource limitation during grain filling of Drysdale resulting from early enhanced vegetative growth during these seasons. For example, with the historic climate, 375 ppm scenario, our modeling predicted that Drysdale's yield was lower than Hartog's in the 1984 and 2014 seasons at Horsham. In both these years, the model predicted low plant available water at anthesis and the total rainfall between anthesis and harvest was only 20 mm. At anthesis, in these years, Drysdale's biomass was 5% higher which resulted in greater predicted water stress over the grain fill period. Overall, the range of yield advantage between the 25 and 75 percentiles was wider for the current climate compared to the future 550 ppm scenarios.

| Responses across Australian arable land
The application of the model across all arable land between 221 and  (Figures 1 and 5b). Under warmer, drier climates (Figure 5e,f) yield advantage of Drysdale declined more in the higher rainfall regions of the study area.
In these simulations, part of these differences was attributed to changes in TE with a median 8% greater TE of Drysdale over Hartog under the present climate (Figure 6a). This difference in TE increased marginally to 9% under elevated CO 2 (Figure 6b).   (2011)(2012)(2013)(2014)(2015) with and without supplementary irrigation (d) Grain yield CO 2 environment, while in a drier warmer climate the range was from 143 (ambient CO 2 ) to 212 (elevated CO 2 ) kg/ha (Table 4). The potential benefits of this increased genetic TE trait across the aver-    (Vadez, Kholova, Medina, Kakkera, & Anderberg, 2014). It provides results at an industry-relevant scale needed to assess the value of the TE trait and its value proposition in commercial breeding (Fischer, 2011;Rebetzke et al., 2013). This confirms the robust nature of the enhanced-TE trait for future climatic conditions and reflects the advantage of low-CID lines in buffering against intermittent water stress that may occur during the growing season (Rebetzke et al., 2002).
The cultivar Drysdale used here as a high-TE example was bred by selecting for low-CID in leaf tissue accumulated immediately prior to stem elongation when soil moisture conditions are favorable. A major advantage of selecting using CID is that TE is integrated over the lifetime that the plant material is sampled (Condon et al., 2004;Farquhar & Richards, 1984). Genetic variation in CID can arise from variation in both photosynthetic capacity and stomatal conductance and it is likely that the high-TE Drysdale has both a higher photosynthetic capacity at the leaf level and a reduced stomatal conductance (Condon et al., 2004). Both of these are likely to be advantageous at the crop level when there is insufficient soil water as there generally is in most of the Australian cropping regions. The additional assimilation should benefit biomass production whereas the reduced stomatal conductance may conserve soil water prior to anthesis, so that it can be used during the critical grain filling period where crop water use efficiency for grain production is high (Kirkegaard, Lilley, Howe, & Graham, 2007). The small increases in assimilation together with small reductions in stomatal conductance were confirmed in AGFACE field measurements of leaf physiology (Tausz-Posch et al., 2013). However, these physiology measurements also showed that the leaf-level TE advantages of Drysdale only manifest under certain conditions, and can be transient throughout the season. Nonetheless, CID measurements that "store" an assimilation-weighted, longer term signal of TE differences confirm that Drysdale has, on average, higher TE than Hartog.
The difference between Drysdale and Hartog in TE and canopy conductance shows the importance of saving some soil water early for later use during grain formation. By anthesis, our modeling showed that Drysdale had transpired 17 mm less water and conserved 60 mm more soil water than Hartog. This is in agreement with previous experimental evaluation of near-isogenic Hartog wheat lines varying in CID where it was determined that the yield advantage was achieved by a mixture of increased areal biomass growth together with greater partitioning of biomass to grain attributed to greater postanthesis water use (Rebetzke et al., 2002). In our model, increased biomass resulting from increased TE (increased crop transpiration coefficient and increased canopy resistance, Table 2) matched the experimental results in biomass (at both anthesis and maturity), and was sufficient to explain the grain yield advantage observed in the experiment.

| 1973
Elevated CO 2 is known to increase assimilation and decrease stomatal conductance and therefore increase leaf-level TE of all crops (Leakey et al., 2009). It is commonly assumed that the benefit from elevated CO 2 in growth and yield is greater under drier conditions.
Yet despite the greater TE in the drier areas ( Figure 6), the highest yield comes predominately from the wetter areas, where both high TE and high transpiration can occur ( Figure 5)