Scaling-up From Leaf to Whole-plant Level for Water Use Efficiency Estimates Based on Stomatal and Mesophyll Behavior


 AimsPrediction of whole-plant short-term water use efficiency (WUEs,P) is essential to indicate plant performance and facilitates comparison across different temporal and spatial scales. Here, the isotope model for WUEs,P was scaled-up from the leaf to the whole-plant level.MethodsFor WUEs,P modelling, leaf gas exchange information, plant respiration and “unproductive” water loss were taken into account. Specifically, in shaping the expression of the WUEs,P, we emphasized the role of both stomatal (gsw) and mesophyll conductance (gm). ResultsThe verification showed that estimates of gsw from the coupled photosynthesis (Pn,L)-gsw model accounting for the effect of soil water stress slightly outperformed the model neglecting the soil water status effect, and the established coupled Pn,L-gm model proved more effective in the estimation of gm than the previously proposed model. Introducing the two diffusion control functions into the whole-plant model, the developed model for WUEs,P effectively captured its response pattern to different CO2 concentration (Ca) and soil water content (SWC) conditions. ConclusionsOverall, this study confirmed that accurate estimation of WUEs,P requires an improved predictive accuracy of gsw and gm. These results have important implications for predicting how plants respond to climate change.

respond to climate change.

26
The ratio of carbon assimilation to water loss, water use efficiency (WUE), is essential to optimize 27 plant water use (Medrano et al., 2015) and can be defined at different temporal and spatial scales. At the 28 leaf level, WUE describes the leaf net photosynthetic rate (Pn,L) relative to the leaf transpiration rate (EL).

29
Both processes are controlled by stomatal conductance (gsw). The Pn,L is also controlled by mesophyll

36
Investigating whole-plant WUE at smaller temporal scales not only facilitates our understanding of 37 whole-plant long-term WUE and the underlying mechanism, but also allows us to compare across 38 different temporal and spatial scales. There have, however, been a few attempts to relate gsw (and or gm) 39 to whole-plant WUE at smaller temporal scales, or models to predict response pattern of whole-plant 40 short-term WUE (WUEs,P) to environmental changes. Estimation of WUEs,P is frequently conducted on 41 the assumption that leaf short-term WUE (WUEs,L) is representative of WUEs,P (Hu et al., 2010). (Farquhar and Richards, 1984;Farquhar et al., 1989;Hubick and Farquhar, 1989). Built upon this concept, 4 remains unclear which one is the most useful approach. In general, the WUE model scaling from leaf to 67 whole-plant level needs to be revised and improved based on well-modeled stomatal and mesophyll 68 behavior.

69
In this study, we developed a model to estimate gm based on the empirical relationship between gm  Previous studies found that leaf stomatal conductance (gsw, mol H2O·m -2 ·s -1 ) is highly correlated 80 with photosynthesis (Pn,L, µmol·m -2 ·s -1 ). Based on this, a series of models on the basis of the linear 81 relationship between gsw and Pn,L has been proposed (Ball et al., 1987;Jarvis, 1976;Leuning et al., 1990).

102
proposed a function to predict the linkage between gm and soil water status: ( where f(θm) is the mesophyll conductance limitation function, which depends on soil water stress; gm,p is 105 the potential (unstressed) gm. This model has been used to represent the feedback of gm to soil water stress 6 2011), which prompted us to establish a coupled Pn,L-gm function to model gm by imposing similar 111 limitation functions to mesophyll behavior as that to stomatal behavior. Based on the empirical 112 relationship between Pn,L and gm, the proposed model is as follows:

119
where θ is the actual soil water content (%); θc and θw are soil water content levels at field capacity 120

137
Considering respiration and water loss from non-photosynthetic organs, the ratio of instantaneous net 138 photosynthesis to transpiration can be scaled from the leaf to the whole-plant level: The WUEi,P inferred from Eq. (10) with the tunable parameter qs = 0.25 is model configuration 1, while 8 CO2 assimilation to water loss. At the diel time scale, not only the role of respiration and water loss from 151 non-photosynthetic parts (twigs and stem) during the daytime need to be included, but also respiration 152 and water loss from whole parts (leaf, twigs and stem) during the nighttime contribute substantially to 153 WUEs,P. When all these processes are taken into account, the time-integrated WUEs,P is as follows: 154 c,s n,P n,L n,L c,s s,P w,s w,s sw p L (1 ) 1.6

162
where Ci is the leaf intercellular CO2 concentration (µmol·mol -1 ). The photosynthetic 13 C discrimination 167 where a is the fractionation associated with the atmospheric CO2 diffusion at the boundary layer (2.9‰);

168
am is the fractionation of CO2 diffusion and dissolution in the liquid phase (1.8‰); b is the fractionation 9 atmospheric CO2 and water-soluble organic materials (WSOM, fast-turn-over carbohydrates) in leaves, respectively.

235
Measurements were made every 3 days.

236
The measured WUEi,P was the ratio between Pn,p to Ep (Pn,p / Ep), and the measured WUEs,P was the 237 ratio between accumulative carbon gain and cumulative water loss, that is, WUEs,P = (Pn,P -RP) / (EP + 238 Ed). 244 made on at least three different leaves in each canopy layer at 9:00, 13:00, and 17:00. No significant 245 differences (p > 0.05) in these measurements among different canopy layers were observed. Almost all 246 leaves were exposed to similar light intensities, and thus, the effect of internal leaves was not considered.

247
In this study, we assumed that a period of 30 days was long enough for saplings to be subjected to the    Changes in SWC and Ca led to significant effects on gsw (p < 0.05), with a maximum of 0.0963 280 mmol H2O·m -2 ·s -1 at C400 × 19.65% of SWC and a minimum of 0.0155 mol H2O·m -2 ·s -1 (Fig. 1).

281
Irrespective of Ca, gsw increased sharply as water stress was alleviated, and this effect was less evident 282 when SWC exceeded 17.03% and even decreased when gsw peaked at 19.65% of SWC. In all cases, gsw 283 increased with elevated Ca (Fig. 1

287
In the absence of additional parameterization associated with soil water stress, the gsw simulated by

301
The regression analysis between measured and modeled gsw is presented in Figure 3. When gsw was 302 calculated from Eq. (2), the modeled values were significantly related with measurements (p < 0.01), and 303 the correlation coefficient R 2 decreased from 0.88 to 0.68 as qs increased from 0.25 to 1.50 (Fig. 3). The 304 gsw calculated from Eq. (1) was also significantly correlated with measurements (p < 0.01), with R 2 being 305 equal to 0.87. However, estimates of gsw from Eq. (2), using qs = 0.25, showed slightly better agreements 306 (higher R 2 and slope closer to 1) with measurements than estimates from Eq. (1), and the former caused Eq.

313
The gm was significantly influenced by SWC and Ca treatments (p < 0.05), with a pattern similar to 314 gsw ( Figs. 1 and 4). The gm ranged between 0.0131 and 0.0571 mol·m -2 ·s -1 , which was significantly lower 315 than gsw (p < 0.05). In all cases, elevation of Ca decreased gm. The gm increased rapidly with SWC; 316 however, the amplification was substantially less evident as SWC exceeded 17.03% and even decreased 317 as SWC increased from 19.65% to 25.55% (Fig. 4).

326
(4), using various qm values, produced a more complicated tendency to SWC. At any given Ca, the 327 amplification was more obvious due to the rapid reduction in qm at SWC levels below 17.03%, while the 328 opposite pattern was observed when SWC exceeded 19.65%. When the SWC increased from 17.03 to 329 19.65%, the reduction in qm induced a decrease in the amplification of the simulated gm at ambient Ca 330 (Fig. 5b)

335
The relationships between measured and modeled gm, calculated by Eqs.
(3) and (4) using various gm was calculated from Eq. (4), the R 2 between measured and modeled results deceased as qm increased 6a and 6b).

349
SWC also significantly influenced the measured WUEi,L (p < 0.05), which increased as the severe drought 350 was alleviated (SWC increased from 10.48% to 14.41%), followed by a decline with increasing SWC that the simulated WUEi,L increased as the SWC improved from 14.41 to 17.03% at C400 and C600, 354 departing from the observed decreasing trend (Fig. 7a).

355
At the whole-plant level, a significant effect of Ca and SWC on the measured instantaneous WUE

357
When the SWC increased from 10.48% to 14.41%, the percentage increase in the measured WUEi,P was 358 more pronounced at C800 than at C400 and C600. In response to further increases in SWC, the measured 359 WUEi,P generally decreased sharply, but this impact was less evident when the soil water status was more 360 than 19.65% of SWC. In both model configurations, the measured and simulated WUEi,P values were 361 similar in their response patterns to SWC × Ca, except when the SWC increased from 14.41% to 17.03% 362 at C400 and C600 (Fig. 7b).  results, which was slightly less than that caused in C2 with a deviation of 3.14 ± 2.52 (2.62 ± 1.90) 372 mmol·mol -1 (Fig. 8). This indicates that the C1 for WUEi,L and WUEi,P behaved slightly better than the   The measured and simulated WUEs,P values are shown in Figure 9. At severe drought (10.48% of 381 SWC), the measured WUEs,P peaked at C600 and was lowest at C800, while the simulated WUEs,P, in both 382 model configurations, reached its maximum at C600 and its minimum at C400 (Fig. 9)

394
In both model configurations, the modeled and measured WUE values were significantly correlated 395 (p < 0.01); however, the R 2 values in C1 were higher than those in model C2 (Fig. 10). Estimates of 396 WUEs,P in C1 led to a deviation of 2.77 ± 2.23 mmol·mol -1 from the measured values in comparison with 397 that caused by C2, with a deviation of 2.91 ± 2.95 mmol·mol -1 (Fig. 10). Therefore, the observed WUEs,P

409
was slightly more capable to represent the observed pattern of gsw than Eq. (1) (Fig. 3). These results 410 align with previous study highlighting the importance of the inclusion of the water stress function in the 411 coupled Pn,L-gsw model (Sala and Tenhunen, 1996). However, the addition of the soil water stress- Ca and was therefore less suitable to simulate gm (Fig. 6). In contrast, the predictive accuracy improved 418 considerably when estimating gm using the coupled Pn,L-gm model (Eq. (4)) (Fig. 6). Therefore, the

465
WUEs,P estimated from C1 caused average uncertainties of 24.09% (21.61%). The relative small 466 discrepancies between mean value and standard deviation in uncertainties of gsw, gm, and WUEs,P indicate 467 that these estimation methods were not stable, although model performance was improved. In addition, 468 the WUEs,P was more sensitive to gsw than to gm. That is, 10% error in gsw introduced 6.17% error in

474
Overall, the explored model for WUEs,P, scaling from leaf to whole-plant level, is based on well-475 characterized coupled Pn,L-gsw (Eq. (2)) and Pn,L-gm models (Eq. (4)). However, we recognize that using 476 only data of pot-grown Platycladus orientalis saplings acclimated in growth chambers, with relatively 477 similar canopy components (canopy structure, light interception), is not convincing enough for a general 478 verification of the developed modeling approach. It is therefore important to collect data of more plant exchange measurements. We found the coupled Pn,L-gsw model incorporating the water stress-dependent 484 function with well parameterized qs (Eq. (2)) agreed slightly better with the measured gsw values than the