Seasonal variability in evapotranspiration partitioning and its 1 relationship with crop development and water use efficiency 2 of winter wheat

15 The partitioning of evapotranspiration (ET) into soil evaporation (E) and crop transpiration (T) 16 is fundamental for accurately monitoring agro-hydrological processes, assessing crop 17 productivity, and optimizing water management practices. In this study, the isotope tracing 18 technique was used to partition ET and quantify the root water uptake sources of winter 19 wheat during the 2014 and 2015 growing seasons in Beijing, China. The correlations between 20 seasonal ET partitioning and the leaf area index (LAI), grain yield, and water use efficiency 21 (WUE) were investigated. The fraction of T in ET (FT) between the greening and harvest 22 Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2018-234 Manuscript under review for journal Hydrol. Earth Syst. Sci. Discussion started: 14 May 2018 c © Author(s) 2018. CC BY 4.0 License.


Introduction
Evapotranspiration (ET) represents a critical component of the water cycle in the soil-plant-atmosphere continuum (SPAC), which is fundamental for crop development and for determining water use efficiency (WUE).Partitioning ET into soil evaporation (E) and plant transpiration (T) can provide deep insight into an evaluation of the water saving potential and optimization of agro-management practices (Newman et al., 2006;Guan and Wilson, 2009;Agam et al., 2012).The majority of previous studies referred to T as the productive component of the crop yield, while E was described as the non-productive water loss (Agam et al., 2012;Van Halsema and Vincent, 2012;Ding et al., 2017).A positive linear function has generally been used to describe the relationship between the yield and T provided that water was the main limiting factor on the yield (Hanks, 1983;Ben-Gal and Shani, 2002;Tolk and Howell, 2009).Nevertheless, some studies claimed that E might indirectly benefit the crop yield by creating a microclimate that is more favorable for crop growth and productivity (Stanhill, 1973;Tolk et al., 1995;Burt et al., 2005;Kustas and Agam, 2013).Therefore, further research is necessary to separate the E and T components of ET and investigate their interrelationships with crop development and the WUE.
The partitioning of ET has been studied using several methods and techniques (Kool et al., 2014;Sutanto et al., 2014;Sprenger et al., 2016).Conventional hydrometric methods employed various techniques to directly measure ET (e.g., the eddy covariance technique, micro Bowen ratio energy balance method, and the weighing lysimeter approach) in addition to E (via a micro-lysimeter) or T (using sap flow sensors) (Mitchell et al., 2009;Cavanaugh et al., 2011;Liu et al., 2002;Zhang et al., 2003;Sun et al., 2006).Simulation models such as the Shuttleworth-Wallace model, the Food and Agriculture Organization (FAO) dual crop coefficient model, the HYDRUS-1D model, and the hybrid dual-source model (TVET) have also been used to simultaneously calculate E and T (Li et al., 2010;Zhang et al., 2013;Ding et al., 2017;Sutanto et al, 2012;Guan and Wilson, 2009).Since various fractionation processes such as condensation and evaporation leave characteristic imprint on the isotopic composition of water, the stable water isotopes of 18 O and D are considered ideal (natural)     tracers for separating E and T from ET and tracking water through the soil based on distinct Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.isotopic signatures of water fluxes (Brunel et al., 1997;Sprenger et al., 2016).Water isotopes are highly fractionated during E processes, causing the remaining soil water to become enriched in heavy isotopes; meanwhile, T does not modify the isotopic composition, since there is typically no isotopic fractionation during water uptake or transport through roots and stems (Wang et al., 2010a;Sutanto et al, 2014).
Due to recent advances in the techniques and instruments used to collect measurements of δ 18 O and δD for both liquid water and water vapor, isotope-based methods have been increasingly applied to agricultural systems to precisely partition ET at different time scales (Wang et al., 2012;Wang and Yamanaka, 2014;Zhang et al., 2011;Wang et al., 2016;Lu et al., 2017;Wei et al., 2018).It was reported that T accounts for 20-80% of the total seasonal ET in sparse canopies and row crops, especially under arid and semi-arid conditions (Agam et al., 2012;Coenders-Gerrits et al., 2014;Kool et al., 2014;Lu et al., 2017).At the daily scale, the ratio of T within ET (F T ) varied over a wide range of 0.2-1 within a rice paddy field during a complete growing season in Mase, Tsukuba (Wei et al., 2015).The daily F T also changed greatly (0.52-0.96) throughout the growing season of maize in northwestern China (Wen et al., 2016;Wu et al., 2017).Substantial differences in F T were discovered between the late filling stage (0.83) and the stage of wax ripeness (0.6) in an irrigated field of winter wheat in the North China Plain (NCP) (Zhang et al., 2011).The values of F T changed from 0.46 to 0.74 after an irrigation event during the early growth stage of winter wheat in the NCP and in central Morocco (Wang et al., 2012;Aouade et al., 2016).These studies revealed very distinct changes in F T throughout the crop-growing season and the significant influence The seasonal variations in ET partitioning are strongly associated with crop development (Sprenger et al., 2016).The leaf area index (LAI) is often regarded as an effective crop parameter for explaining the variabilities in the E/ET ratio (F E ) and F T .It is commonly believed that F E decreases exponentially with the LAI for most crops and that the F T increases logarithmically with an increase in the LAI in the absence of water stresses (Villalobos and Fereres, 1990;Liu et al., 2002;Yu et al., 2009;Kato et al., 2004).However, Kang et al. (2003) proposed that F T and the LAI exhibited a saturation relationship for wheat and maize in a semi-arid region of Northwest China.Several recent studies identified a power-law correlation between F T and the LAI for agricultural systems at both the global scale and in certain croplands (Wang et al., 2014;Wei et al., 2015;Wu et al., 2017;Lu et al., 2017;Zhao et al., 2018).In addition, numerous possibilities were suggested for high F T even under low LAI conditions.To illustrate the global variability in the partitioning of ET, Wang et al. (2014) further developed a function relating F T to the growth stage relative to the timing of the peak LAI.It was evident that the LAI within different growth stages should be utilized to evaluate the variability in ET partitioning and crop water use capabilities.
The ability of a crop to access water resources from different soil horizons can be estimated via the root water uptake (Asbjornsen et al., 2007;Wang et al., 2010b;Zhang et al., 2011;Yang et al., 2015).Common methods applied to assess water uptake patterns include the IsoSource model in addition to less than three-layer linear mixing models and Bayesian mixing models (Phillips and Gregg, 2003;McCole and Stern, 2007;Moore and Semmens, Hydrol. Earth Syst. Sci. Discuss., https 2008;Stock and Semmens, 2013).The MixSIAR framework is the latest Bayesian stable isotope analysis mixing model in R that considers multiple sources of uncertainty and provides definite proportions of source contributions.It has been employed successfully to determine the contributions of soil water at different layers to the water uptake of summer maize (Ma and Song, 2016).The root water uptake also indicates the availability of soil water resources to crops, and it varies with different agricultural management practices.Therefore, combining the seasonal partitioning of ET with the development of the LAI and root water uptake patterns can provide a comprehensive understanding of E and T processes.It also help design a reasonable irrigation depth, which is vital for improving the crop yield and WUE in regions with a high food demand and limited water resources such as in the NCP.
The NCP constitutes one of the major food production regions in which winter wheat represents the main water-consuming crop.In addition, the NCP provides approximately 69% of the wheat production for all of China.However, irrigated agriculture practices throughout the NCP are facing critical challenges (i.e., very limited water supplies) to the provision of sufficient quantities of food.To optimize the irrigation strategies for winter wheat, considerable research has been conducted to determine the relationships among the seasonal ET with the crop yield and WUE (Li et al., 2005;Sun et al., 2006;Shang et al., 2006;Liu et al., 2013).E and T were also partitioned, and an exponential relationship between F E and the LAI was established (Liu et al., 2002;Yu et al., 2009).Furthermore, E was reported to be an unproductive water loss, and thus, it should be reduced in regions with a severe water deficit.
Recently, Zhang et al. (2011) simultaneously addressed the partitioning of ET and the characterization of root water uptake depths for winter wheat during the growing season.In this study, the isotope mass balance approach was utilized in conjunction with the soil water balance method to partition the ET of winter wheat during the 2014 and 2015 growing seasons in Beijing, China.The three primary objectives of this study were to (1) detect the variations of ET partitioning during the different growth stages of winter wheat, (2) quantify the seasonal root water uptake patterns of winter wheat, and (3) determine the relationships between ET partitioning and the LAI, grain yield and WUE.The results were applied to establish optimal agricultural management practices and design the irrigation depth.The climate in this area is sub-humid with a mean annual precipitation of 540 mm, but only 20-30% of the precipitation occurs during the winter wheat season (Cai et al., 2009).The mean annual temperature is 12.1°C and mean seasonal reference evapotranspiration (ET 0 ) of winter wheat is 610 mm.Soils in the 2-m profile were sampled every 20 cm depth to measure their physical and chemical properties.The depths with similar soil particle size and organic carbon were merged into one layer.The soil profile was finally divided into four layers and their main properties are shown in Table 1.

Field experiments
The winter wheat (variety: Zhongmai-175) was planted on October 9, 2013, and harvested on June 8, 2014, during the 2014 growing season.The sowing and harvest dates during the 2015 growing season were October 11, 2014, and June 8, 2015, respectively.Five irrigation and fertilization treatments (T1, T2, T3, T4, and T5) were applied from winter greening to the harvest season.Here, treatment T5 refers to the agricultural management practices employed by local farmers with a total irrigation of 240 mm and a nitrogen supply of 210 kg N ha -1 (as urea) (Table 2).In comparison with the reference treatment (T5), the T1 and T2 treatments had reduced irrigation of 60 mm from greening to jointing and 80 mm from the filling to the harvest period, respectively.The T3 and T4 treatments both reduced the irrigation during jointing-heading or heading-filling stage by 80 mm compared with treatment T5.The nitrogen (N) application rates for the T1 and T3 treatments were both 0.5-fold of that for treatment T5, while 1.5-fold of the nitrogen in treatment T5 was applied to both the T2 treatment and the T4 treatment.All the irrigation was provided in a single application per stage.The application date was 27 Mar, 22 Apr, and 22 May in 2014, while it was 29 Mar, 9 May, and 21 May in 2015, respectively.The detailed irrigation and fertilization schedules for these five treatments are shown in Table 2. Three replicates were conducted for every treatment in the plots with each area of 6 × 5 m.Basin irrigation with groundwater was implemented for all of the treatments.Precision-leveled basins were used to prevent runoff.
The soil water contents in the 2-m soil profile were measured at a 20-cm interval every 5-7 days in each plot using a TRIME-IPH probe based on the Time domain Reflectometry with Intelligent MicroElements technique (IMKO GmbH, Ettlingen, German).Additional measurements were conducted when soil water samples were collected for isotope analysis as well as before and after each irrigation or heavy rainfall event.Meanwhile, three plants in each plot were selected to manually observe their leaf areas (obtained by multiplying the leaf length and width), which were then calibrated using a leaf scanner (F915900, Canon, Canada).
The LAI was calculated as the product of the calibrated leaf area per plant and the number of plants per unit area.The grain was air-dried, and the crop yield was recorded separately for each plot after harvesting.
Meteorological data including the precipitation, maximum and minimum air temperatures, solar radiation, wind speed and relative humidity were recorded every 30 min by the automatic weather station (Monitor Sensors, Caboolture QLD, Australia).The rainfall amounts were 77.0 mm and 74.7 mm between the greening and harvest seasons in 2014 and 2015, respectively.Both seasons were dry at 75% precipitation exceedance probabilities (PEPs) in terms of the rainfall frequencies during the last five decades in the Beijing area.
However, there was an additional 34.8 mm of precipitation during the greening-jointing period and 26.9 mm less precipitation during the jointing-heading period in 2015 relative to 2014.

Water sampling and isotopic analyses
Different waters including the precipitation, irrigation water, soil water, and stem water were sampled to analyze the isotopic composition of 18 O and D. The precipitation was collected after each rainfall event via a rain collector coupled with a polyethylene bottle and funnel.A ping-pong ball was positioned at the funnel mouth (Wang et al., 2012).The ping-pong ball Evaporation was then prevented during the rainfall process.The collected rainwater was transferred to a bottle immediately, sealed and stored.Irrigation water was sampled in each irrigation event.
Three stems of each treatment were sampled at an interval of about one week.Each stem was taken from the part between the soil surface and the first node of one representative plant.
It was cut into pieces in 2-3 cm length, then put into a vial and sealed with parafilm.All the epidermises of the stems were removed to eliminate the effect of the isotopically depleted atmospheric water vapor on the stem water isotopic compositions (Brunel, 1997).
The soil water at depths of 10, 20, 30, 50, 70, 90, 110, 150, and 200 cm was sampled on and after the day of collecting stem water, and after each irrigation or heavy rainfall event.A suction lysimeter made of a Teflon pipe and porous ceramic cup was installed and used to abstract the soil water at each depth (Wang et al., 2012).If the soil water content was too low to collect soil water by the suction lysimeter, soil sample instead was collected using a hand auger.
All of the stem and soil samples were kept refrigerated (-15°C to -20°C) prior to measuring the isotopic compositions.The cryogenic vacuum distillation system (LI-2000, LICA, Beijing, China) was applied to extract water in the soil and stem samples (West et al., 2006).The ratio of 2 H/ 1 H and 18 O/ 16 O of different water samples were measured on a Los Gatos Research (LGR) DLT-100 liquid water isotope analyzer (San Jose, CA, America).They were calibrated against the VSMOW international standards and converted to δD and δ 18 O values.The measuring precision for δD and δ 18 O was ±1‰ and ±0.1‰, respectively.

Evapotranspiration partitioning methods
Transpiration changes soil water content but keep soil water isotopic composition constant because water uptake from soil by plants does not result in isotopic fractionation (Zimmerman et al., 1967).On the contrary, both soil water content and soil water isotopic composition are changed in evaporation process (Allison and Barnes, 1983).Many previous studies reported that the water balance and isotope mass balance equations were robust to partition ET into E and T when sampling intervals were short (Hsieh et al., 1998;Robertson and Gazis, 2006;Wenninger et al., 2010;Wang et al., 2012).In this study, the ET in the day of the stem water sampling was partitioned into E and T using the following soil water balance and isotope mass balance equations in the 0-200 cm profile: where m and δ represent the water flux and isotopic composition of δ 18 O in different waters, respectively, f and i denote the final and initial state of the soil water storage in one sampling day of stem water, respectively, P is the precipitation, I is the irrigation, D is the drainage out of the soil profile, and R is the surface runoff.There were two or three times of stem water sampling during each growth period.The average value of the partitioned E or T during one growth period was used to represent the ET partitioning result in this period.
The final and initial soil water storage (m f and m i ) in Eq. ( 1) was calculated using the measured depth-weighted volumetric soil water content.Meanwhile, the precipitation (m P ) was obtained from meteorological observations, while the irrigation (m I ) was artificially The values of δ i and δ f are the depth-weighted δ 18 O averages for the whole soil profile collected on and after the day of stem water sampling, respectively, while δ P and δ I are the measured δ 18 O values of the precipitation and irrigation, respectively.The δ 18 O value of evaporation (δ E ) is estimated using the fraction factor α liquid-vapor = (δ l +1000)/ (δ v +1000) (Wang et al., 2012).The evaporated water δ v (δ E in Eq. ( 3)) is assumed to be in isotopic equilibrium with the soil water δ l (δ i in Eq. ( 3)).The value of α liquid-vapor is given as 1.0102 at an air temperature of 15º C following Clark and Fritz (1997).As there is no fractionation in the T processes of winter wheat (Wang and Yakir, 2000), the value of δ T is determined using the measured δ 18 O of stem water.

MixSIAR model
The MixSIAR Bayesian mixing model (v2.1.3)incorporating with dual stable water isotopes was used in the MixSIAR model for estimating the probability density functions of variables as the MCMC was advantageous to estimate the entire distribution for each variable.The MCMC parameter run length was set to "very long" to converge on the true posterior distribution for each variable.The model error was evaluated using the SIAR (process and residual).The estimated 5 th , 25 th , 50 th , 75 th , and 95 th percentiles of the posterior contributions of each source described the distribution associated with the proportional contribution of each source to winter wheat.The 50% percentile represented the median source contribution value for each source.

Data analysis
The statistical analyses of the variation in each isotopic composition, soil moisture distribution, ET component and associated fraction, and root water uptake pattern during each season and treatment were all performed using a Statistical Package for the Social Sciences (SPSS) 19.0 software package.The WUE was defined as the ratio between the crop yield and the total ET from the greening to harvest season of winter wheat (Hussain et al., 1995).In this study, the highest WUE value without an evident decrease in the grain yield was used as the primary criterion with which to evaluate the optimal agricultural management practice.
Regression analyses of either F T or T with the LAI, crop yield and WUE were all performed to investigate the relationships between the partitioning of ET and crop development.
The smaller slope of the LMWL in 2015 than in 2014 was ascribed to a faster evaporation rate of falling raindrops (Wang et al., 2010b).As shown in Fig. 1 and 4.0 mm d -1 in the greening-jointing, jointing-heading, heading-filling, and filling-harvest periods, respectively.The higher daily mean ET flux during the mid-season stage (i.e., the jointing-filling stage) was mainly due to a higher LAI and an increased biomass.

Seasonal variations in ET partitioning
The seasonal variations in the partitioning of ET are shown in Fig. 3.The daily mean T changed significantly among the different periods during the experimental seasons of both 2014 and 2015 (p < 0.01) (Fig. 3).The daily mean T was evidently small (2.0 mm d -1 ) during the early growth stage of greening-jointing and reached a high level during the jointing-heading and heading-filling periods (4.4 and 4.6 mm d -1 , respectively), after which it declined moderately to 3.4 mm d -1 until the winter wheat harvest.In contrast to T, a substantial seasonal variance in the daily mean E was detected only in 2014 with values of 1.1, 0.3, 0.8, and 0.6 mm d -1 during the greening-jointing, jointing-heading, heading-filling, and filling-harvest periods, respectively (Fig. 3).In 2015, the differences in the daily mean E among the four periods were small with an average value of 0.8 mm d -1 .A significant (Fig. 4).These results demonstrate that the average F T during the jointing-heading period of the 2015 season (0.82) was much lower than that of the 2014 season (0.94).In particular, the decreasing of F T in the jointing-heading period from 2014 to 2015 was evident under treatments T1, T2, and T5 with reductions of 0.25, 0.14 and 0.16, respectively.Moreover, the performance of F T in each growth period was notably distinct among the different treatments (Fig. 4).Compared with the mean level of F T for all of the treatments, F T was 16.9% larger during the greening-jointing period but significantly less (17.6%, p < 0. 05) during the filling-harvest period under treatment T1.The T5 under the reference agricultural management practices had the smallest F T during greening-jointing period in 2015.
The F T value during the whole season had an average value of 0.82, and it did not vary significantly among the seasons and treatments (with an SD of 0.03, p > 0. 05) (Table 4).The T during the jointing-heading and heading-filling periods (T jf ) accounted for approximately 50% of the seasonal ET, and T jf exhibited a significant positive linear correlation with the total ET (R 2 = 0.82, p < 0.01).Therefore, T jf played a critical role in determining the variations in ET throughout the growing season.Fig. 4 demonstrates that the value of T jf was greatly different between the two experimental seasons.The average T jf was 34.9 mm more in 2015 than in 2014, occupying 76.9% of the increment (45.3 mm) in the total ET from 2014 to 2015.Furthermore, both the largest T jf and the highest total ET were observed under the reference treatment (T5) in 2015.(0.7-2.0), while T was the predominant partition in the ET during the mid-growing season when the LAI exceeded 2.7 (Fig. 5).However, F T reached a maximum of 0.78 with a small LAI (1.11) under treatment T1 during the 2015 season.The seasonal changes in F T can be effectively described as a power-law function of the LAI (F T = 0.61 LAI 0.21 , R 2 = 0.66, p<0.01) for winter wheat (Fig. 5).This implied that crop development played a major role in driving the contribution of T to ET and that the LAI could provide insights into estimating the variability in F T throughout the growing season of winter wheat.

Seasonal variations in root water uptake patterns
The contributions from soil water in different layers to root water uptake estimated using the MixSIAR model are shown in Fig. 6.The average contributions of soil water to winter wheat within the 0-20, 20-70, 70-150, and 150-200 cm layers during the 2014 growing season were 28.7%, 30.0%, 26.9%, and 14.4%, respectively.The root water uptake depth tended to become deeper with crop development (Fig. 6a).Winter wheat mainly acquired soil water from the 0-20 cm (63.6%), 20-70 cm (67.9%), 70-150 cm (54.4%), and 70-150 cm (39.8%) layers during the greening-jointing, jointing-heading, heading-filling, and filling-harvest growth periods, respectively.The 150-200 cm layer contributed a certain amount of soil water to winter wheat since the jointing-heading period and reached a maximum mean proportion of 27.2% in the filling-harvest period.
As shown in Fig. 6, higher quantities of shallow soil water were taken up by winter wheat in 2015, particularly within the top layer (0-20 cm) with average contributions of 28.7% and 42.6% in 2014 and 2015, respectively.The predominant water uptake depth was 0-20 cm in both the greening-jointing period (70.4%) and the heading-filling period (63.4%) in 2015.

Relationships between grain yield and WUE with seasonal partition of ET
The crop yield and WUE of winter wheat throughout the growing season for each treatment are shown in Table 4.The mean grain yield was 6759.9 kg ha -1 with an SD of 478.5 kg ha -1 .
Compared with the reference treatment, the T1, T2, and T3 treatments reduced the grain yield by more than 10%, while treatment T4 raised the yield by 0.9% for the 2014 season.
Treatment T5 exhibited the lowest grain yield in the 2015 season, while the other treatments showed a 15.6% increment on average.The mean WUE was 24.9 kg ha -1 mm -1 in the 2014 season and 21.9 kg ha -1 mm -1 in the 2015 season.The variability in the WUE among the different treatments was greater in 2015 than in 2014.The maximum WUE was observed under T1, whereas T5 showed the smallest WUE in each season.
The results demonstrate that both the grain yield and the WUE were not significantly correlated with the F T throughout the experimental season (p > 0.05).The observed grain yield positively increased with T jf in 2014, whereas it decreased with T jf in 2015.The grain yield reduced remarkably with excessive values of T jf (205.8 mm) under treatment T5 in 2015 (Fig. 7 and Table 4).A significant quadratic relationship was found between the grain yield and T jf (R 2 = 0.77, p < 0.01) (Fig. 7).The peak grain yield was 7062.6 kg ha -1 at a T jf value of 155.8 mm along this fitting curve.The WUE also had a significant quadratic correlation with T jf (R 2 = 0.87, p < 0.01) (Fig. 7).The peak WUE along the fitting curve was 24.9 kg ha -1 mm -1 with a T jf value of 117.5 mm.Most T jf values were larger than this critical value except for that under treatment T3 in 2014.As T jf exceeded 117.5 mm, the WUE declined to a minimum value of 16.0 kg ha -1 mm -1 with a continuous increase in T jf (Fig. 7).
This suggested that the magnitude of T jf controlled both the grain yield and the WUE for winter wheat in this region.

Influencing factors of seasonal variations in ET partitioning
The daily T flux estimated in this study (ranging from 2.0 to 4.6 mm d -1 ) was similar to those of 1.02-4.91mm d -1 (Zhang et al., 2011) and 0.8-4.5 mm d -1 (Liu et al. 2002) under surface irrigation in the NCP determined via the isotope/eddy covariance and weighing/micro-lysimeters methods, respectively.E is a significant component of ET, especially when the LAI is low.The seasonal F E (0.18) was also in accordance with the value of 0.23 reported by Liu et al. (2002).The F E value calculated in our study reached up to 0.35 during the greening-jointing period, which is consistent with the estimation of 0.30 from Zhang et al. (2011).This study indicated that the seasonal changes in F T could be effectively described through a power-law function of the LAI.This relationship was similar to those obtained in recent studies at both the global and the field scale (Wang et al., 2014;Wei et al., 2015;Wu et al., 2017;Lu et al., 2017).The strong correlation between F T and LAI confirmed that F T was controlled by LAI at seasonal timescale (Wang and Yamanaka, 2014;Wang et al., 2014;Wei et al., 2015;Wei et al., 2018).When LAI was less than 2.7, F T increaseed significantly with crop development in the early growing season and then it converged towards a stable value beyond LAI of 2.7.This threshold of LAI (2.7) to distinguish the two different changing trends of F T agreed well with the values of 2.5 and 3.0 reported in Wei et al. (2018) and Kang et al. (2003), respectively.F T has been shown to reach a high level (0.90 for agricultural systems at the global scale and 0.58 for a paddy field), even under low LAI conditions (Wang et al., 2014;Wei et al., 2015).In this study, the estimated F T reached up to 0.78 with a small LAI (1.11) under treatment T1 during the 2015 growing season.The above comparisons indicate that the ET partitioning results in this study are reliable.
Besides LAI, F T was influenced greatly by soil moisture, especially the topsoil moisture in 0-20 cm depth.Previous studies indicated that F T generally decreased with increasing topsoil moisture due to increase of E under the same LAI conditions (Liu et al., 2002;Yu et al., 2009;Wei et al., 2018).A negative linear correlation was found between F T and surface soil water content (θ v ) when LAI was about 1.8 (F T =-1.38θ v +1.0, R 2 =0.98, p<0.01) during greening-jointing period in our experiments.It was suggested that keeping surface soil dry without affecting the crop ET was an important way to reduce E in the early growing season (Liu et al., 2002).However, increasing θ v remarkably increased F T at LAI of about 4.0 (F T =1.83θ v +0.6, R 2 =0.74, p<0.01) during filling-harvest period.
Factors controlling E and T were coupled in ways to affect F T under dry climate condition particularly during jointing-heading period in 2015.Adequate rainfall falling during greening-jointing period (35 mm) led to larger θ v at the early stage in jointing-heading period (mean of 0.19).Great availability of soil moisture in the topsoil increased water contribution to E. Furthermore, the strong atmosphere demand remarkably promoted E at late stage of the jointing-heading period (Zhao et al., 2018).This resulted in the significant increase of the E into the upper dry layer via hydraulic lift through the process of root water uptake (Jha et al, 2017;Li et al., 2010).High water uptake from deep layers (32.2% and 23.5% in the 70-150 cm and 150-200 cm layers, respectively) may improve the plant leaf water content and maintain T rates and dry matter production.
Distributions of soil moisture and root water uptake patterns were significantly influenced by different irrigation and fertilization treatments especially in dry seasons.More frequently irrigated treatments have previously been reported to have more roots in the surface layer than less irrigated treatments (Zhang et al., 2004).Meanwhile, nitrogen fertilizers stimulated root growth near the soil surface and abundant soil nitrogen content might increase the drought resistance of the root system under water limited condition (Kmoch et al., 1957;Carvalho and Foulkes, 2013).Ma and Song (2016) showed that the soil water contribution had a significantly positive and linear relationship with the proportion of root length.Therefore, plant primarily took up soil water from the top layer (0-20 cm) under the T4 and T5 treatments even though the climate was dry during jointing-heading period in  et al., 2004;Jha et al., 2017).This confirmed that over 80% of plant water took up soil water from the 70-200 cm layer under the less irrigation treatments of T1 and T2.When soil water near the surface was replenished by irrigation, the extraction depth returned to the surface layer (such as in heading-filling period) and subsequently moved downward again until harvest of winter wheat.

Application for optimizing water management practices
With the abovementioned ET partitioning results and the fitted WUE-T jf and Yield-T jf curves, the irrigation and fertilization schedules were optimized.As shown in Fig. 7, the value of T jf should be controlled between 117.5 and 155.8 mm to obtain both a high grain yield and a high WUE.The T jf under treatments T1, T2, T4, and T5 in 2014 and that under treatment T1 in 2015 acquired in this study were within this range.An additional irrigation of 140 mm was required for treatment T5 compared with the T1.Although the T1 treatment in 2014 had a larger WUE, its grain yield was diminished by 8.5% relative to 2015.Therefore, the T1 treatment in 2015 optimally improved the WUE and maintained a high grain yield.The optimal irrigation and fertilization schedules can be determined as two irrigations during the greening-jointing (20 mm) and heading-filling (80 mm) periods and one fertilization (105 kg ha -1 N) during the greening-jointing period.The designed wetting layer should be controlled at depths of 0-70 cm because wheat primarily sourced soil water from the 0-70 cm layer during the experimental seasons.This practice could make better use of the deep soil water storage and avoid deep percolation compared with a traditional wetting depth of 100 cm (Zhang et al., 2011).
The obtained optimal agricultural management practice is supported by previous studies in the NCP.The first small irrigation was applied mainly to the top soil to ensure that the fertilization was distributed evenly throughout the plot.The irrigation in the early growth stage reduced the grain yield because it enhanced the development of non-functional tillers, which consume the reserved nutrients (Sun et al., 2006).Wang et al. (2014) and Lu et al. (2017) reported that water loss via E could be much higher during the vegetative stage than during later growth stages.Reducing the irrigation amount during the greening-jointing period could increase the depletion of deep soil water, and it was definitely necessary to improve the WUE of winter wheat.Meanwhile, the heading growth period was extremely sensitive to water stresses, and irrigation is strongly recommended during this period (Zhang et al., 2003;Li et al., 2005;Shang and Mao, 2006).Therefore, the obtained optimal schedule in this study is appropriate, as it could conserve 140 mm of irrigation water and 105 kg ha -1 N of fertilizer with respect to the reference practices.

Further scopes of this study
The ET of winter wheat was partitioned effectively into E and T using the isotope mass ET besides E and T and need further estimation.Second, the water flux at the bottom boundary of the soil profile was generally neglected in the estimation of ET due to small changes in soil moisture during winter wheat growing season under limited irrigation in the NCP (Zhang et al., 2003;Li et al., 2005;Li et al., 2010).However, drainage should be accurately evaluated by the Darcy's law when soil moisture at the bottom boundary is above field capacity.Third, as calculated from the MixSIAR model, each soil layer had a different contribution to the root water uptake.Incorporating these contributions into the isotopic mass balance equation can reflect the variation in the gradient of the isotopic profile.Finally, high-frequency measurements of isotopic composition of soil, stem and gas water will improve understanding the seasonal variation in ET partitioning.

Conclusions
In this study, the isotope mass balance were coupled with water balance methods for the Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.
Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License. of irrigation on the partitioning of ET.It is therefore necessary to thoroughly clarify the seasonal variability in the partitioning of ET in association with its major influencing factors.
Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.However, the definite correlations between the magnitude and fraction of seasonal ET partitioning with the grain yield and WUE are still unclear.Further investigations are therefore required to demonstrate seasonal variations of ET partitioning and root water uptake pattern and quantify their relationships with the LAI, grain yield and WUE under different agricultural management practices.

Field
experiments with winter wheat were conducted from 2013 to 2015 at the Irrigation Experiment Station of the China Institute of Water Resources and Hydropower Research (IWHR) at Daxing, Beijing (39°37′N latitude, 116°26′E longitude, 40.1 m a.s.l.elevation).
Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.floated up when rain fell at the funnel mouth and enabled the rainfall to move into the bottle.
Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.controlled and therefore measurable.The soil moisture near the bottom boundary remained steady and generally below the field capacity throughout the experimental seasons.Therefore, the amount of drainage (m D ) was neglected in this study.In addition, no runoff (m R ) was observed during the field experiments.

(
δD and δ 18 O) was used to identify the water uptake sources of winter wheat.In field experiments, precipitation or irrigation water infiltrated and finally mixed into the old soil water.Groundwater could hardly contribute to crop water use (the average maximum rooting depth was 2 m for winter wheat) under the condition of the deep water table depth (mean of 16 m below the soil surface).It can be supposed that soil water at different depths was proportionally sourced by winter wheat.Four layers was divided as 0-20, 20-70, 70-150, and 150-200 cm depth along the 2-m soil profile in terms of their water isotopic compositions, Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.soil moisture contents and root distributions.The dual stable isotopes of the soil water in each layer (raw source data) and of the stem water (mixture data) were input to the MixSIAR model to quantify the main root water uptake depth.The Markov chain Monte Carlo (MCMC) , the soil water isotopes mainly fell below the LMWL, especially in 2015.The slope of the fitting line between δD and δ 18 O in soil water was lower in 2015 (2.8) than in 2014 (4.0).It indicated that the soil water was more strongly evaporated in 2015.According to two-way Analysis of variance (ANOVA), the isotopic profiles of the soil water showed significant differences among the different layers and growth stages (p < 0.05).The δD and δ 18 O values of the soil water in the surface layer (0-20 cm) were remarkably enriched and indicated that the soil water isotopes had been subjected to extremely evaporative fractionation.The soil water isotope values in the 0-20 cm layer were significantly different from those in the other layers throughout the growing seasons (p < 0.05).The soil water isotopes in the 20-70 and 70-150 cm layers were intensively fractionated since the jointing stage.No significant seasonal changes were detected in the isotopic compositions of the soil water in the 150-200 cm layer, and they were similar to those values of irrigation water (Fig.1).The stem water isotopes were mainly concentrated along the fitting line of the δD-δ 18 O relationship in soil water (Fig.1).The majority of the stem water isotopes in 2014 matched well with the soil water isotopes in the 0-150 cm layer; nevertheless, they were more enriched in 2015 and fell in the upper soil layer (0-70 cm).Therefore, the maximum root water uptake depth of winter wheat probably approached 150 Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.cm during the experimental seasons.3.2Seasonal changes in soil water storage and ETApproximately 127.9 mm of the soil water storage in the profile of 0-200 cm was consumed on average throughout the whole season from wheat greening to harvest.Approximately 92% of this consumption occurred in the 0-150 cm layer (Fig.2).The slight change in the soil moisture within the 150-200 cm layer was consistent with the small variation in the soil water isotopic compositions in the same layer (Figs.1-2).A greater amount of soil water storage (with a mean value of 35.2 mm) was consumed in 2015 than in 2014, primarily within the 0-70 cm layer (Fig.2).The largest reduction in the soil water storage during the 2015 season occurred during the jointing-heading period (98.1 mm), and this reduction accounted for 67.4% of the total loss.Among the five treatments, T4 showed the highest consumption of soil water storage during the 2014 (151.8 mm) and 2015 (174.5 mm) seasons.Sufficient irrigation during treatment T5 in 2014 led to the smallest observed reduction in the soil water storage (80.3 mm).However, the reduction in the soil water storage under treatment T5 notably increased to 143.4 mm in 2015.This was primarily caused by severe reductions in the soil water storage during the jointing-heading and filling-harvest periods without irrigation under dry climatic conditions.The total ET throughout the season from wheat greening to harvest was a mean of 292.8 mm with a standard deviation (SD) of 38.2 mm (Table4).The total ET increased on average by 45.3 mm in 2015 relative to 2014, and this was in general agreement with the observed increment of soil water consumption in 2015.The reference agricultural management practice (T5) remarkably raised the crop water consumption in terms of the largest ET value in the Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.growing seasons of both 2014 (304.0 mm) and 2015 (377.3 mm).The daily mean ET was significantly different (p < 0.01) among the four growth periods with values of 3.0, 5.0, 5.4, difference in the daily mean E between 2014 and 2015 occurred in the jointing-heading period, as it increased to 1.0 mm d -1 in 2015 due to severe drought induced by little precipitation and the lack of irrigation.The values of F T varied widely from 0.51 to 0.98 during the individual growth periods under different treatments.The mean values of F T were 0.65, 0.88, 0.84, and 0.85 during the greening-jointing, jointing-heading, heading-filling, and filling-harvest periods, respectively Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.

Fig. 5
Fig. 5 reveals that F T increased with the LAI and varied around an asymptotic value of Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.During the jointing-heading period, the limited water supply and high T rates remarkably promoted the average contribution of deep soil water with values of 32.2% and 23.5% in the 70-150 cm and 150-200 cm layers, respectively.Meanwhile, winter wheat took up significantly more soil water from the 20-70 cm layer (54.9%) during the filling-harvest period than during the other periods (p < 0.01).
Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.rate to 1.0 mm d -1 in the period.It was the topsoil moisture greatly influencing E, while the water used for T came from the whole root zone.Although continuous E caused the extreme consumption of surface soil moisture in the drought period, soil water storage in the subsurface layers could meet T requirement of crop.Soil water in the deep layers could move 2015.However, nitrogen deficiency promoted root growth in the deep soil layer(150-200 cm)      and increased water adsorption by 43.1% under the T1 and T3 treatments.Previous studies demonstrated that plants growing in drier environments with soil water deficit in the surface layer have deeper root systems and more branched seminal roots(Morita et al., 1997; Zhang      Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.
balance and water balance methods.The partitioning of ET changed between different irrigation and fertilization schedules and various crop development stages.The evaluation using isotopic data presented a quantitative correlation between seasonal change in the F T and the crop development of LAI.The relationships among the grain yield and WUE with the T jf were discovered.This isotope-based method provided insights into clarifying the hydrological processes in field ecosystem and optimizing water and nitrogen management practices.Nevertheless, several issues still need further investigation.First, although the interception flux is often neglected in many partitioning works, it indeed is a component of Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.
partitioning of evapotranspiration (ET) into crop transpiration (T) and soil evaporation (E) of winter wheat under different irrigation and fertilization treatment schemes during[2014][2015] in Beijing, China.The fraction of T in ET (F T ) showed averages of 65.4%, 87.7%, 83.8%, and 84.9% in the greening-jointing, jointing-heading, heading-filling, and filling-harvest periods, respectively.The performance of F T was notably distinct among the different treatments in each growing period.However, the value of F T throughout the season from greening to harvest did not vary significantly among the seasons and treatments (p > 0. 05) and had an average value of 0.82.The seasonal change in F T could be effectively described as a power-law function of the LAI (F T = 0.61 LAI 0.21 , R 2 = 0.66, p<0.01).Winter wheat mainly utilized soil water from the 0-20 cm (67.0%), 20-70 cm (42.0%), 0-20 cm (38.7%), and 20-70 Hydrol.Earth Syst.Sci.Discuss., https://doi.org/10.5194/hess-2018-234Manuscript under review for journal Hydrol.Earth Syst.Sci. Discussion started: 14 May 2018 c Author(s) 2018.CC BY 4.0 License.cm(34.9%) layers during the greening-jointing, jointing-heading, heading-filling, and filling-harvest periods, respectively.The main root water uptake depth increased with the crop development in 2014, whereas it was mostly concentrated within the 0-70 cm layer in 2015.F T was not significantly correlated with the grain yield and WUE (p > 0.05), and the total T during the jointing-heading and heading-filling periods (T jf ) had a significant quadratic relationship with the grain yield and WUE (p < 0.01).In order to obtain the optimal crop yield, 20 mm and 80 mm of irrigation water during the greening-jointing and heading-filling periods, and 105 kg ha -1 N of fertilization during the greening-jointing period were needed.The designed wetting layer should be controlled at depths of 0-70 cm.This study demonstrated the roles of seasonal ET partitioning obtained via isotope-based methods in determining the crop development and improving the WUE, and the findings acquired herein have important implications on irrigation and fertilization management.

Fig. 3 .
Fig. 3. Mean daily evaporation (E) and transpiration (T) rates of winter wheat during each

Fig. 4 .
Fig. 4. Seasonal variations in T, E, and the fraction of transpiration within evapotranspiration

Fig. 5 .
Fig. 5. Relationship between the fraction of transpiration within evapotranspiration (F T ) and

Fig. 6 .
Fig. 6.Proportions of the soil water contribution to winter wheat during each growth stage in

Fig. 6 .
Fig.6.Proportions of the soil water contribution to winter wheat during each growth stage in (a) 2014 and (b) 2015 (mean ± SD). ○ and * represent outliers with a 1.5× interquartile range (IQR) and a 3IQR, respectively.

Fig. 7 .
Fig.7.Relationships among the grain yield and water use efficiency (WUE) with the total transpiration during the jointing-heading and heading-filling periods (T jf ).The two vertical red dashed lines represented the T jf under the maximum WUE (T jf-WUEmax ) and Yield (T jf-Yieldmax ) conditions.

Table 1 .
Physical and chemical properties of the soil profile at the experimental site.