2.1 Location and experimental setup

We carried out a field experiment at two rhizotron facilities in Selhausen, North Rhine-Westphalia, Germany (50°52^′ N, 6°27^′ E). The field is slightly inclined with a maximum slope of around 4°. One rhizotron facility was located upslope (F1) with around 60 % gravel by weight in the 10 cm topsoil, while the second rhizotron facility was downslope (F2) with silty soil (stone content is around 4 % by weight).

Each rhizotron facility was divided into three subplots of 7.25 m by 3.25 m: two rainfed plots (P1, P2), and one irrigated plot (P3). In rainfed plot P1, other sowing densities and dates were used than in the other plots, and we therefore excluded these plots. Silage maize Zea mays L. `Zoey' was sown on 4 and 8 May in 2017 and 2018, respectively, with a plant density of 10.66 seeds m⁻² (Fig. 1a; Table 1). Detailed information on crop management practices is provided in Table 1.

Figure 1Daily maximum air temperature (T_max) (°C), daily maximum air vapor pressure deficit (VPD) (kPa) in the two growing seasons in (a) 2017 and (b) 2018, and cumulative (sum) of rainfall and irrigation from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) in the two growing seasons in (c) 2017 and (d) 2018. The vertical dashed black lines in (a) and (b) indicate silking time. Vertical grey lines in (a) and (b) indicate the measured days for leaf gas exchange and leaf water potential. Two lines for 2017F2P2 and 2017F2P3 are overlapped by the lines from 2017F1P2 and 2017F1P3, respectively.

Download

Table 1Crop phenology and management information for different treatments in 2017 and 2018. Dates follow the format mm/dd.

^a From sowing to harvest. ^b For rainfall for the whole growing season.

Download Print Version | Download XLSX

2.2 Water application

The irrigation systems (T-Tape 520-20-500 drip lines (Wurzelwasser GbR, Müzenberg, Germany)) were installed parallel to the crop rows with 0.3 m intervals. A nearby weather station (approx. 100 m from the experiment) recorded weather variables (global radiation, temperature, relative humidity, precipitation, and wind speed) every 10 min. In addition, the precipitation amount was manually collected by a plastic rain gauge next to each rhizotron facility. The Penman–Monteith equation was employed to estimate reference evapotranspiration. Daily crop evapotranspiration was calculated based on the single crop coefficient and the reference evapotranspiration (Allen et al., 1998). Irrigation amounts were estimated as the weekly sum of the calculated crop evapotranspiration. A total amount of 230 mm precipitation was recorded during the growing period (136 d), while average, minimum, and maximum daily air temperatures were 17.6, 8.3, and 25.3 °C, respectively (Fig. 1b). The crop on the irrigated plots (2017F1P3 and 2017F2P3) received in total 130 mm (10 times, every 5–7 d, using 13 mm of irrigation water per event) between mid-June to the end of August (Fig. 1b). Average, minimum, and maximum daily air temperatures in 2018 were higher than in 2017 with 19.2, 10.85, and 27.3 °C, respectively (Fig. 1b). The summer season in 2018 could be considered an extreme year with respect to plant growth at our experimental location due to exceptionally hot and dry weather conditions. The crop received only 91.3 mm of rain during the growing period of 2018 (107 d). The crop on the irrigated plots 2018F1P3 and 2018F2P3 was irrigated every 5–7 d (in total 13 times), with a total amount of irrigation of 257 and 239 mm between mid-June and mid-August, respectively (Fig. 1d). To avoid crop failure due to severe drought in 2018, the rainfed plot in the stony soil (2018F1P2) had to be irrigated (in total 66 mm) four times (using 13, 22, 13, and 18 mm, respectively) (Fig. 1d). Detailed estimates of irrigation amount and intervals could be found in Nguyen et al. (2022a).

2.3 Measurements

2.4 Calculation of total root length, root system conductance, stem, and whole-plant hydraulic conductance

To estimate the total root length from minirhizotubes, we adopted option 2, which was described in G. Cai et al. (2017). Total root length per square meter soil surface area within each soil layer (m m⁻²) was computed by multiplying the root length density with the corresponding soil layer thickness. The root length density was determined in each depth by dividing the measured root length per minirhizotron image by the assumed volume the roots would have occupied in absence of the tube, i.e., W ⋅ L ⋅ tube radius (see G. Cai et al., 2017).

Following Nguyen et al. (2020), the effective soil water potential was calculated based on hourly measured soil water potential (ψ_i) and normalized root length density at six depths (10, 20, 40, 60, 80, and 120 cm) (NRLD_i) and soil layer thickness (Δz_i) in the soil profile (Eq. 1).

We followed Ohm's law analogy by dividing the hourly sap flow by the difference between effective soil water potential and shaded leaf water potential to estimate root system conductance (K_{soil_root} – Eq. 2), between shaded leaf water potential and sunlit leaf water potential to estimate stem hydraulic conductance (K_stem – Eq. 3), and between effective soil water potential and sunlit leaf water potential to estimate whole-plant hydraulic conductance (K_{soil_plant} – Eq. 4).

During 1 measurement day, four values of the K_{soil_root}, K_stem, and K_{soil_plant} were obtained from measurements between 11:00 and 14:00 CEST. The average and standard deviation of these hourly measurements were calculated for each measurement day in order to present the seasonal dynamics of those variables. To capture the diurnal and seasonal variations of sap flow and sunlit leaf water potential, in addition, we plotted the hourly sap flow and hourly difference in effective soil water potential and sunlit leaf water potential for 3 measurement days starting from predawn and whole seasons, respectively, to derive the slope, which is also K_{soil_plant}.

2.5 Statistical analysis

Regression analysis was performed to understand the relationship between the sap flow volume and the difference in effective soil water potential and sunlit leaf water potential as well as the relationship between the total aboveground biomass and cumulated water transpired (sap flow volume). These analyses allow one to derive the slope as a proxy of K_{soil_plant} and transpiration use efficiency, respectively. Since all measured data have their own measurement errors, the generalized Deming regression was employed. We performed relationships (via correlation coefficient and statistical significant levels) of midday leaf An, Gs, and E with midday K_stem, K_{soil_plant}, K_{soil_root}, sunlit leaf potential, ψ_{soil_effec}, and the difference in ψ_{soil_effec} and sunlit leaf water potential (ψ_difference). All data processing and analyses were conducted using the R statistical software (R Core Team, 2022).

3.1 Root growth under different water treatments, soil types, and climatic conditions

Observed root lengths (cm cm⁻²) from the minirhizotubes at different soil depths in the first week of June (stem elongation), around silking, and at harvest in two growing seasons are shown in Fig. 2. Root length was similar among water treatments at the start of stem elongation in both years (Fig. 2a and d). The difference in root length was pronounced at silking and harvest between the soil types. More root growth was observed in the silty soil than in the stony soil with the same water treatment (i.e., 2.5–6 times higher at a depth of 40 cm). This indicated the strong negative effects of stone content on root development. In the stony soil, root length in the irrigated plot (F1P3) was slightly higher than in the rainfed plot (F1P2). In contrast, the rainfed treatment (F2P2) in the silty soil showed much higher root length, especially from 40 to 120 cm depth, than the irrigated plot (F2P3) in both growing seasons. Much lower stone content and deep soil cracks in the silty soil (Morandage et al., 2021) allow root extension to the subsoil, particularly in the rainfed plot F2P2. Root length in the rainfed treatment (F2P2) in 2018 was higher than in 2017, which implies that root further developed to exploit the water in the soil under rainfed conditions to meet the higher evaporative demand.

Figure 2Observed root length from minirhizotubes (cm cm⁻²) from 10, 20, 40, 60, 80, and 120 cm soil depth from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) in the two growing seasons in 2017 (a – 8 June, b – at silking on 13 July, c – at harvest on 12 September) and in 2018 (d – 7 June, e – 1 week after silking – 18 July, f – 1 week before harvest – 16 August).

Download

Total root length (m m⁻²) estimated from minirhizotubes and its ratio to shoot dry matter (m kg⁻¹) on three measured dates (as in Fig. 2) are shown in Fig. 3. Total root length was much higher for the silty plots than for stony plots. In 2017, the highest total root length was observed in the rainfed plot of the silty soil (F2P2) with approximately 9166 and 9878 m m⁻² around silking and harvest, respectively, which was almost 2 times higher than in the irrigated plot (F2P3). These figures were higher in 2018 than in 2017, where total root length of F2P2 was 10 188 m m⁻² and 13 750 m m⁻² at silking and harvest time, respectively. For the rainfed stony soil (F1P2), soil water depletion around the beginning of June 2017 (Fig. S1a in the Supplement) and from the first 2 weeks of June to harvest in 2018 (Fig. S2a) caused the strong reduction of shoot biomass. In the stony soil, the shoot dry matter of the irrigated plot (F1P3) and the rainfed plot (F1P2) was 1275 and 536 g m⁻² at silking time (e.g., 19 July 2018 – DOY 200; Fig. S3a and b). However, there was a minor difference between F1P2 and F1P3 in terms of the ratio of root length to shoot dry matter. In the silty soil, a decrease in soil water potential was not pronounced (compared to stony soil) in either 2017 or 2018 (Figs. S1b and S2b). In 2018, shoot biomass in the irrigated stony soil (F1P3) and silt soil (F2P3) was similar (1275 and 1299 g m⁻², respectively on 19 July 2018 – DOY 200), while the shoot biomass of the rainfed silty soil (F2P2) was 876 g m⁻² (Fig. S3a and b). However, the ratios of root length to shoot biomass in the rainfed plot of the silty soil (F2P2) were 3 and 6 times higher than those in the irrigated silty soil (F2P3) and stony soil (F1P3), respectively (e.g., 18 July, DOY 199). Moreover, total root length was relatively equal among treatments at the start of set elongation (8 June – DOY 159) in both years, while this was the opposite for the ratio of root length to shoot dry matter. This firstly illustrated that the finer soil texture without stones and with soil cracks could favor the root growth, which indicates strong interactions of root and soil conditions. Secondly, the larger root length and higher atmospheric evaporative demand in 2018 than in 2017 also indicate the interaction of root growth and climatic conditions.

Figure 3Observed root length from minirhizotubes (m  m⁻²) and ratio of root length per shoot dry matter (m kg⁻¹) from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) in the two growing seasons (left panel) in 2017 and in 2018 (right panel). More specifically, 8 June (DOY 159), 13 July (DOY194), and 12 September (DOY 255) in 2017 are stem elongation, silking, and harvest, respectively; 7 June (DOY 158), 18 July (DOY 199), and 16 August (DOY 228) in 2018 are stem elongation, 1 week after silking, and 1 week before harvest, respectively (see also Fig. 2).

Download

3.2 Stomatal conductance, photosynthesis, transpiration, and K_{soil_plant}

3.2.1 Diurnal course of stomatal conductance, photosynthesis, transpiration, and water potential at leaf level

After a long period with high temperatures and no rainfall, soil water reduction in the rainfed plot of the stony soil (F1P2) on 17 July 2018 (Fig. S2) resulted in 3 times lower net photosynthesis (An), stomatal conductance (Gs), transpiration (E), and leaf water potential (ψ_leaf) compared to the remaining treatments (Fig. S4). This indicates that the soil water content strongly affected the stomatal conductance. Stomatal closure was very pronounced around midday in F1P2, while this was not the case in F2P2, indicating the soil type strongly affected the stomatal conductance and leaf gas exchange. Leaf gas exchange and leaf water potential in F1P2 were still much lower than in other plots (Fig. 4). On 18 July 2018, after application of 22.75 mm of irrigation water (at 16:00), photosynthesis, stomatal conductance, transpiration, and leaf water potential were slightly increased in F1P2. However, these were still smaller than in F2P2 and the two irrigated plots.

Figure 4Diurnal course of (a) photosynthetically active radiation (PAR) and vapor pressure deficit (VPD), (b–e) leaf net photosynthesis (An), (f–i) leaf stomatal conductance (Gs), (j–m) leaf transpiration (E), and (n–q) leaf water potential (LWP) on 18 July in maize in 2018 before irrigation at the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2). Measurement was carried out from a shaded leaf (plus symbol with line) and two sunlit leaves (solid dot – lines and solid square – lines). Crop was irrigated at 13:00, 13:00, and 16:00 for F1P3, F2P3, and F1P2, respectively (22.75 mm for each plot) (Fig. S2). Black arrows indicate time of irrigation.

Download

On the next day after irrigation, leaf gas exchange and water potential were considerably increased in F1P2 (Fig. S5). Leaf curling was also less pronounced than the 2 previous days. Predawn and midday leaf water potentials were around −0.4 and −1.6 MPa for all plots, respectively. Leaf transpiration rate was around 3.1 mmol m⁻² s⁻¹ for all water treatments and soil types at 00:00. This indicated the recovery of the plant after watering at the rainfed plot with stony soil (F1P2).

3.2.2 Seasonal course of stomatal conductance, photosynthesis, transpiration, water potential, and plant hydraulic conductance at the leaf level

Seasonal stomatal conductance (Gs) and leaf water potential (ψ_leaf) are described in Fig. 5. The relationship between two variables was rather noisy and nonlinear. The leaf water potential showed distinct patterns among treatments in one growing season. Minimum ψ_leaf was maintained at around −1.5 MPa in the irrigated plot in the stony soil (F1P3) and two plots in the silty soil (F2P2 and F2P3). Lower minimum ψ_leaf could be observed in the rainfed plot with stony soil (F1P2), but it did not go beyond −2 MPa. Minor leaf curling was observed only in the second week of June in F1P2 in 2017. In 2018, the higher temperature and vapor pressure deficit resulted in lower minimum ψ_leaf in all treatments and soil types compared to 2017. The minimum ψ_leaf was around −2 MPa in F1P3, F2P2, and F2P3, while ψ_leaf could drop below −2 MPa in F1P2, which was due to the severe soil water deficit. The low Gs and ψ_leaf were associated with measurement dates when the substantial leaf curling was observed from mid-July to the end of growing season in F1P2 in 2018 (Figs. S3c, d and S6c, d).

Figure 5Seasonal stomatal conductance to water vapor (Gs) versus leaf water potential (ψ_leaf) in 2017 (top panel) and in 2018 (bottom panel) at the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2). Vertically continuous and dashed lines indicated ψ_leaf at −1.5 and −2 MPa, respectively. Measurement was carried out from a shaded leaf (plus symbol) and two sunlit leaves (solid dots).

Download

The effective soil water potential (ψ_{soil_effect MD}), sunlit leaf water potential (ψ_{sunlitleaf MD}), stomatal conductance (Gs_MD), and whole-plant hydraulic conductance (K_{soil_plant MD}) at midday at several times during the growing season are presented in Figs. 6 and 7 for 2017 and 2018, respectively. As expected, there was not much difference in terms of ψ_{soil_effec MD} among F1P3, F2P2, and F2P3 from 2 August to 1 week before harvest in 2017. The lowest ψ_{soil_effec MD} was observed in F1P2. Leaf water potential dropped drastically, but K_{soil_plant MD} also increased strongly, whereas ψ_{soil_effec MD} remained quite similar (e.g., 18 July). This is because sap flow increased substantially on this day (e.g., from 2.34 mm d⁻¹ on 17 July to 6.97 mm d⁻¹ on 18 July for the F1P2). The stomatal conductance decreased a lot on this day, which could be explained by the fact that the atmospheric demand increased (e.g., global radiation was 13.6 MJ m⁻² on 17 July compared to 23.9 MJ on 18 July while daily VPD was 0.7 and 1.2 kPa, respectively) even more than the sap flow. Midday sunlit leaf water potential was not distinctively different among treatments with the lowest ψ_{sunlitleaf MD} around −1.6 MPa throughout season. Also, Gs_MD was rather similar among plots. The K_{soil_plant MD} ranged from 0.125 to 0.96 mm h⁻¹ MPa⁻¹ with a sharp reduction before harvest. In general, the lowest values of K_{soil_plant MD} were found in F1P2, which was consistent with the smaller overall seasonal K_{soil_plant} (as the slope of the linear relationship between sap flow and the difference in effective soil water potential and sunlit leaf water potential) (see Fig. S7).

Figure 6Dynamic of around midday (MD) of (a) the effective soil water potential (ψ_{soil_effec MD}), (b) sunlit leaf water potential (ψ_{sunlitleaf MD}), (c) stomatal conductance (Gs MD), and (d) whole-soil–plant hydraulic conductance (K_{soil_plant MD}) in the growing season 2017 from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2). Error bars indicate the standard deviation of the different values taken around midday (11:00, 12:00, 13:00, and 14:00) of different sunlit leaves. Whole-soil–plant hydraulic conductance is shown from 17 July when sap flow was measured. The black arrows indicate the irrigation events for the irrigated treatments F1P3 and F2P3 in the showing period.

Download

Figure 7Dynamic of around midday (MD) of (a) the effective soil water potential (ψ_{soil_effec MD}), (b) sunlit leaf water potential (ψ_{sunlitleaf MD}), (c) stomatal conductance (Gs MD), and (d) whole-soil–plant hydraulic conductance (K_{soil_plant MD}) in the growing season 2018 from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2). Error bars indicate the standard deviation of the different values taken around midday (11:00, 12:00, 13:00, and 14:00). Leaf water potential and stomatal conductance were two sunlit leaves and one shaded leaf at each measured hour. Whole-soil–plant hydraulic conductance is shown from 3 July when sap flow was measured. The black arrows indicate the irrigation events for the irrigated treatments F1P3 and F2P3, while the orange arrow indicates the irrigation application for the rainfed plot in the stony soil (F1P2).

Download

The ψ_{soil_effec MD} was substantially different in the two soil types and water treatments in 2018 (Fig. 7a). Both F1P2 and F1P3 showed a gradual drop in ψ_{soil_effec MD} from 15 June until the third week of July and then increased after irrigation events on 18 July (Fig. S2b). However, ψ_{soil_effec MD} of F1P2 was much lower than F1P3 toward the harvest. The ψ_{soil_effec MD} of F2P2 and F2P3 only decreased progressively from around 10 July till harvest even though there was water supply from the irrigation (Fig. S2b). The water applied by irrigation and coming in by rainfall was insufficient to wet up the deeper soil layers, which remained dry. The low Gs_MD corresponded to the lowest ψ_{sunlitleaf MD} and K_{soil_plant MD} from the F1P2 (Fig. 7c and d). The K_{soil_plant MD} from all plots was ranging from 0.12 to 0.91 mm h⁻¹ MPa⁻¹. There was the drop in K_{soil_plant MD} (i.e., 3–9 or 17–18 July) before irrigation in this plot. However, it increased after the irrigation (i.e., 10 and 19 July). This suggests that K_{soil_plant} depends strongly on the soil water content and the conductivity of the rhizosphere.

3.2.3 Relationships of stomatal conductance, transpiration, and photosynthesis with plant hydraulic variables at the plant canopy level

The slope of linear relationship between sap flow and difference in ψ_{soil_effec} and ψ_sunlitleaf is shown for 3consecutive days (leaf water potential measurements from the predawn) and before and after irrigation application (17, 18, and 19 July 2018) (Fig. 8). On both 17 and 18 July, the difference between ψ_{soil_effec} and ψ_sunlitleaf was around −1.6 MPa with very low transpiration rates in treatment F1P2, which was associated with very low plant hydraulic conductance and leaf curling. The whole-plant hydraulic conductance was disrupted on these 2 d (0.06 and 0.16 mm h⁻¹ MPa⁻¹ for 17 and 18 July, respectively). Water was supplied on 18 July at 13:00 for the irrigated plots (F1P3, F2P3) as well as for F1P2 at 16:00 (for saving the plant from death due to severe drought stress). K_{soil_plant} was slightly changed (0.43 and 0.57 mm h⁻¹ MPa⁻¹ for F1P3 on 18 and 19 July, respectively, and 0.5 and 0.58 mm h⁻¹ MPa⁻¹ for F2P3 on 18 and 19 July, respectively). However, the increase in K_{soil_plant} was substantial in the F1P2 after the irrigation. Soil water replenishment and an increase in the root–soil contact (Fig. 7a) allowed the K_{soil_plant} to recover overnight to 0.46 mm h⁻¹ MPa⁻¹. This resulted in a narrower water potential gradient between root zone and sunlit leaf and in a higher transpiration rate on 19 July.

Figure 8Relationship between sap flow and difference in effective soil water potential and sunlit leaf water potential (ψ_difference) from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) on 3 consecutive measurement days from predawn in 2018: (a) 17 July – DOY 198, (b) 18 July – DOY 199, and (c) 19 July – DOY 200. Crop was irrigated on 18 July (DOY 199) at 13:00, 13:00, and 16:00 for F1P3, F2P3, and F1P2, respectively (22.75 mm for each plot). The unit of slope in the linear regression (or soil–plant hydraulic conductance) is mm h⁻¹ MPa⁻¹. Regression was based on the DEMING approach. The asterisks next to the slopes indicate a significant correlation between two variables according to Pearson's method. ns denotes nonsignificant. ^* p<0.05. ${}^{**}$ p<0.01. ${}^{***}$ p<0.001.

Download

Seasonal averages of different midday hydraulic conductance components (root system hydraulic conductance – K_{soil_root}, stem hydraulic conductance – K_stem, and whole-plant hydraulic conductance – K_{soil_plant}) are shown in Fig. 9. In the same year, the K_stem was not much different among F1P3, F2P2, and F2P3 plots. The K_stem of those plots was slightly higher than in F1P2 in both years. In general, the K_{soil_root} was lower than the K_stem. Overall, the estimated K_{soil_plant} was around $\mathrm{1}/(\mathrm{1}/K_{\mathrm{soil}\mathit{\_}\mathrm{root}}+\mathrm{1}/K_{\mathrm{stem}}$) regardless of soil types, years, and water treatments. The K_{soil_root} and K_{soil_plant} in F1P2 in 2018 were much lower than the remaining plots, while the K_{soil_root} and K_{soil_plant} were not much different among plots in 2017. Our results indicated that there was an impact of the soil hydraulic conductance on K_{soil_root} and K_{soil_plant}. Although there is a large difference in total root length between the two soil types (e.g., F1P3 versus F2P2 or F2P3 versus F2P2), K_{soil_root} and K_{soil_plant} in those two plots were not much different. This could be explained by the fact that K_{soil_plant} was not only dependent on root length but also dependent on the variability in root segment hydraulic conductance.

Figure 9Comparison of different midday hydraulic components (mm h⁻¹ MPa⁻¹): soil–plant (grey bars), soil–root (yellow bars), and stem (blue bars) from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) in the two growing seasons (a) in 2017 and (b) in 2018. The error bars indicate the standard deviation from measurements around midday (11:00, 12:00, 13:00, and 14:00) on different measured days in 2017 (with $n=\mathrm{4}\times \mathrm{9}$ d; Figs. S10, S11, and 6) and in 2018 (with $n=\mathrm{4}\times \mathrm{10}$ d; Figs. S10, S12, and 7).

Download

3.3 Relative importance of root and leaf area growth to transpiration and crop performance at canopy level

Drought stress was observed in the rainfed plot (F2P2) in the second week of June 2017 with mild leaf rolling. The crop then recovered due to sufficient rainfall and lower evaporative demand. Drought stress occurring again at the stem elongation phase caused reduction of plant size (height and stem diameter) (Fig. S6) as well as a slight reduction of leaf area and biomass in this plot (Fig. S3a, c). Transpiration per unit of leaf area did not differ much among water treatments and soil types in 2017 (Fig. S8). The opposite was the case for the transpiration rate per unit of root length. The observed root length at different soil depths (Fig. 2) and total root length for two plots in the stony soil were much smaller than in the silty soil (Fig. 3). Therefore, transpiration per unit of root length in the stony soils (F1P2 and F1P3) was almost 3 times higher than transpiration in the silty soil. For the same soil, transpiration per unit root length of the irrigated treatment was slightly larger than in the rainfed plot.

The differences in sap flow per plant between water treatments and soil types were more pronounced in 2018 (Fig. S9). The highest transpiration rate was observed in the irrigated plots (F1P3 and F2P3), followed by the rainfed plot of the silty soil (F2P2), and it was the lowest in the rainfed plot of the stony soil (F1P2). These observations were in line with the differences in biomass and leaf area index between the treatments (Fig. S3b, d) and plant size (Fig. S6b–d). In 2018, severe leaf rolling was observed in the rainfed plot (F1P2) from the beginning of June until the end of the growing period in 2018 (Fig. S3d). Similar to 2017, transpiration per unit of root length was much higher in the stony plots than in silty plots. Also, for the silty soil, transpiration per unit of root length of the irrigated plot (F2P3) was higher than in the rainfed plot (F2P2).

Figure 10Relationship between aboveground dry matter and cumulative sap flow from the rainfed (P2) and irrigated (P3) plots of the stony soil (F1) and silty soil (F2) in the two growing seasons in (a) 2017 and (b) 2018. The unit of slope linear relationship is g mm⁻¹. The lower number of data points in (b) in 2018 from the F2P2 and F2P3 plots was due to the missing values for measured sap flow because of sensor disconnection. For aboveground dry matter, each point represents the average of two sampling replicates, except the harvest with five sampling replicates.

Download

Higher cumulative transpiration in the irrigated plots did not result in higher transpiration use efficiency (TUE) in both soil types (Fig. 10). For instance, TUE was 16.87 and 15.59 g mm⁻¹ for F1P2 and F2P2, respectively, while it was 15.47 and 14.79 g mm⁻¹ for F1P3 and F2P3, respectively, in 2017 (Fig. 10a). For the same soil, the rainfed plot showed slightly higher TUE than the irrigated plot. When comparing the TUE of maize of the two soil types for the same water treatment, TUE in the stony soil was almost the same in the silty soil. The TUE was not much different among treatments and soil types in 2018. Overall, TUE in 2017 was higher than in 2018 (Fig. 10b).

4.1 Effects of soil types, water application, and climatic condition on root growth

Our root observations showed that soil type affected root growth more than water treatment (Fig. 2). Root growth was strongly inhibited by the stony soil where much lower root length was observed than in the silty soil, especially in the deeper soil layers. This was consistent with the findings reported in Morandage et al. (2021), where a linear increase in stone content resulted in a linear decrease in rooting depth across all stone contents and developmental stages. Also, both simulations and observations indicated that rooting depth increased due to the presence of cracks in the lower minirhizotron facility (Morandage et al., 2021), which could explain the high root length between 40 and 120 cm soil depths, which was observed in the silty soil in both years.

In terms of the ratios of root length to shoot biomass, Ordóñez et al. (2020) reported much larger figures of for instance 880 cm g⁻¹ in different locations and under different N application rates in maize growing in the Midwest United States. Jorda et al. (2022) reported a wide range of ratios of root length to shoot biomass from 200 to 1000 cm g⁻¹ around flowering time of maize, depending on the wild type and root hair mutant genotypes growing on either loamy or sandy soils. More roots and higher ratios of root length to shoot biomass were found in the sand than in the loam in both wild type and root hair mutant genotypes (Jorda et al., 2022; Vetterlein et al., 2022). Cai et al. (2018) observed much larger ratios of root length to shoot biomass in drought-stressed plots than in irrigated plots in both soil types in winter wheat, which indicated the alternation of sink–source relationships to cope with water stress. This study emphasized that more assimilates are used to promote root growth and extract more water under drought stress. However, this was not the case for the stony soil in our work where the drought stress was more pronounced, especially in 2018. A drop in soil water potential (Fig. S2b), and thus effective soil water potential (Fig. 6a), was substantial from 10 July 2018 to the harvest in the rainfed plot in the silty soil (F2P2), which was consistent with the reduction of leaf water potential (Fig. 6b), leaf area (Fig. S3c), total dry matter (Fig. S3d), and crop height (Fig. S6b) compared to the irrigated plot (F2P3). This indicates a mild water stress in 2018 in the rainfed plots in the silty soil. The larger ratios of root length to shoot biomass in this F2P2 plot in 2018 than in F2P3 could be explained by the change in source–sink relations where more assimilates were devoted to root growth, even at a later growth stage. Moreover, the low stone content and soil cracks (Morandage et al., 2021) might favor root growth in the deeper soil layers which are close to the shallow soil water table in the rhizotron facility with silty soil (Vanderborght et al., 2010). In conclusion, both soil texture and water conditions influenced the root growth; however, the effects of the former on root length were more pronounced than those of the latter.

In the stony soil, which has a considerably smaller water-holding capacity than the silty soil, root length was considerably smaller than in the silty soil. Nevertheless, water uptake per unit root length was much larger than in the fine soil. This also means that the hydraulic conductance per unit root length must have been much larger in the stony soil than in the fine soil. Cai et al. (2018) observed a similar effect for winter wheat, but they found much smaller differences in the root-length-normalized root conductance. The higher root-length-normalized root conductance means that the anatomy of the root tissues must have been influenced by the soil texture and compensated the considerably smaller root length in the stony soil. Looking at the effect of water treatments in the silt soil, the nonirrigated plot had more roots than the irrigated one, and both had more roots in the year with high VPD. But the soil–root conductance was higher in the irrigated plot than in the rainfed plot. This means that in the irrigated plot, the soil–root conductance per unit root length was higher than in the rainfed plot. This could either be due to wetter soil conditions and higher soil conductance or be due to a larger conductance of the root tissues. Especially in 2017 when the silty soil was wetter, the slightly larger soil–root conductance in the irrigated plot is most likely the result of larger root tissue conductance in the irrigated plot. Thus, how root architecture (here represented simply by the total root length) and root tissue conductivities “respond” to drought stress might be opposite, depending on the comparisons that are made. When the stony soil and silt soil are compared, the higher “stress” due to lower water availability in the stony soil resulted in fewer roots with a higher root tissue conductance in the soil with more stress. When comparing the rainfed with the irrigated plot in the silty soil, the higher stress in the rainfed soil resulted in more roots with a lower root tissue conductance in the treatment with more stress. This indicates that the response to water stress can be different depending on soil conditions or water treatments.

4.2 Effects of soil types, water application, and climatic condition on stomatal conductance, photosynthesis, transpiration, leaf water potential, and plant hydraulic conductance

4.3 Relative contribution of water control by leaves and roots on transpiration and transpiration use efficiency

Responses of crops via stomatal control to reduce water loss at leaf scale while maintaining leaf photosynthesis and water use efficiency were reported earlier (Nguyen et al., 2022a; Vitale et al., 2007). In addition to that, in the maize experiments in 2017 and 2018, leaf rolling was observed in both rainfed plots in the stony and the silty soil in the second week of June 2017 and from the beginning of June until the end of the growing period in 2018. This indicates another dehydration avoidance mechanism resulting from morphological adjustments, which is an effective strategy for leaf senescence (Aparicio-Tejo and Boyer, 1983; Richards et al., 2002). Stomatal closure resulted in a larger reduction of transpiration and assimilation in the rainfed plots than in irrigated plots with the same soil type (Figs. 4, 5, S4, S5, and S9a). There was reduction of shoot biomass (also stem size and leaf size adjustments) in F1P2 compared to other plots. However, the TUE was not smaller in this plot than the remaining plots. These observations confirm that plant size adjustments through reduction of height, leaf width, and length are efficient responses to reduce water loss at canopy scale in addition to stomatal control at the leaf level.

Relative contribution of leaf area to transpiration has been highlighted in wheat where reduction of tiller number resulted in significantly lower leaf area index (LAI), thus lower canopy transpiration (Cai et al., 2018; Trillo and Fernández, 2005; Nguyen et al., 2022a). However, root system conductance per unit of leaf area and per unit root mass was strongly reduced and eventually more than reduction of leaf area under water stress (Trillo and Fernández, 2005). In our work, expressing the transpiration per unit of root length on the one hand allowed us to analyze the role of total root length to water uptake. However, on the other hand, the lower total root length did not necessarily result in a lower root water uptake and vice versa. For instance, the rainfed plot of the treatment F2P2 had the larger total root length, which could postpone the effect of soil water limitations in drying soils due to greater ability to extract water from subsoils. Therefore, transpiration was very similar between F2P2 and F2P3. Despite the much lower total root length in the stony soil, K_{soil_plant} in the irrigated plot (F1P3) was not much lower than in the same water treatment in the silty soil (F2P3; Figs. 6d, 7d, 8, and S7). This could be explained by the fact that the K_{soil_plant} variability not only depended on root architecture (here the root length and distribution) but also depended on the variability in root segment hydraulic properties, which has also been illustrated and discussed in Zwieniecki et al. (2002), Frensch and Steudle (1989), Meunier et al. (2018), Couvreur et al. (2014), and Ahmed et al. (2018). Moreover, the contribution of shoot hydraulic conductance could be large in plants (Gallardo et al., 1996; Trillo and Fernández, 2005; Sunita et al., 2014), which also confirmed in our work. In our work, K_{soil_plant} comprised root and shoot conductance, which are directly influenced by soil hydraulics. Our estimates of K_{soil_plant} varied with transpiration and gradients of ψ_sunlitleaf and ψ_{soil_effec}. Thus, any change in soil hydraulic conductance will change the root to shoot water potential. Consequently, it will affect the gradients between the shoot and root rhizosphere (Carminati and Javaux, 2020). Thus, our study reveals the importance of both soil texture characteristics and root phenotypic traits (here root length) in regulating plant transpiration (Cai et al., 2022a). Despite lower root length in the stony irrigated plot, transpiration rate was not much lower than in the silty irrigated plot in our work. This could be related to another property of the root, such as root segment conductance or other root traits (e.g., root hair). Further investigation with extensive measurements of roots including axial and radial root conductance at field scale will be required to better explain the observed results. Other traits like root hair density (Cai et al., 2022a) or higher root length density (Vadez, 2014) could contribute to the soil to root water potential and root-zone hydraulic conductance where dense root hairs delay soil water deficit in drying soils. However, contrasting results have shown that root hairs did not have an effect on root water uptake (see Jorda et al., 2022). The role of root hairs could not be analyzed in our work, which was based on root data from minirhizotron images.

This study investigated soil–water–plant relations, more specifically the interactions between the root and shoot growth processes and water fluxes under variations of soil water status and atmospheric demands. To the best of our knowledge, the comprehensive data collected from soil to root, plant, and atmosphere under field conditions in this work were unique. However, we acknowledge the lack of treatment replicates, which was due to the complex and expensive construction of the rhizotron facilities. We also acknowledge the small size of plots that did not allow the extensive destructive sampling (i.e., leaf area, biomass, or determination of leaf water potential). Each rhizotron site originally contained the irrigated, rainfed, and rainout sheltered plots (Nguyen et al., 2022a; Cai et al., 2016). The overall aim of the experiments was to investigate the root and shoot responses and gas fluxes (CO₂ and H₂O) of wheat and maize to the variations of soil water and soil hydraulics. Note that the studies did not intend to investigate the impacts of similar irrigation strategies on plant water status among seasons (i.e., in 2017 and 2018) because the irrigation practices were less common in the regions. The collapse of manual rainout shelters due to strong wind after the 2016 growing season resulted in only two water treatments (rainfed and irrigated). Based on experiences from the previous seasons (wheat), we argued that such combinations of two water treatments and two soil types, to some extent, could still create a wide range of soil water conditions for the maize trial. For instance, the “rainfed” treatment on the stony soil in the upper rhizotron (F1P2) could lead to severe water stress compared to other treatments, especially in the summertime when the atmospheric evaporative demand is high. In fact, mild water stress was observed at F1P2 around mid-June in 2017. In 2018, the sites were slightly modified to induce more severe water stress (Nguyen et al., 2022a). One rainfed plot with stony soil had late sowing, while one rainfed plot with silty soil had a higher sowing density (data not shown in the study). Unprecedented weather (extremely hot and dry) in 2018 resulted in severe drought stress at the rainfed plots with stony soil. To compare the effects of soil types and water treatments on the crop, we present here only data from two plots (rainfed and irrigated) for two soil types. In spite of the experimental limitations, the relative differences among the treatments, soil types, and seasons, as well as measured dates, are clearly illustrated, which ultimately supported the overall aim of our study.

The simultaneous measurements of atmospheric conditions, leaf water potential, and transpiration rates (coupled with measurements of root, stem, and whole-soil–plant hydraulic conductance, root architecture (root length), and soil water potential distribution) illustrate the complex responses of the shoot and root growth and hydraulic conductance vulnerability to soil water availability. The different responses of crop processes to soil hydraulics and climatic conditions suggest further field investigations for other soil types, growing seasons, and water regimes. Future studies considering the effects of progressive soil drying or irrigation strategies on plant water status and crop growth at field conditions will be necessary. This is very relevant for those crop-growing regions that require irrigation. Our results show that the leaf water potential threshold can vary within the same genotype depending on soil types, climatic conditions, and water management. Large variability in minimum leaf water potential has been reported for maize genotypes under greenhouse conditions (Welcker et al., 2011). Field studies are required concerning the stomatal functions, water relations, hydraulic vulnerability traits, and root : shoot responses, especially of different maize cultivars in responding to drought stress. This will suggest implications for selecting agronomic cultivars and traits under changing climates. Results from this study show that soil–crop models should not only focus on simulating stomatal regulations to capture the response to drought stress but also require adequate representations of root and leaf growth and adjustments. The soil hydraulics strongly influenced soil water availability and crop growth. Regional application of soil–crop models for simulating gas fluxes and crop growth processes and for estimating irrigation amounts must account for the environmental heterogeneity within the spatial simulation unit, whereas the soil heterogeneity is the key variable.