Evaluation of soil water percolation under different irrigation practices, antecedent moisture and groundwater depths in paddy fields

Agricultural Water Management 192 (2017) 149–158

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Evaluation of soil water percolation under different irrigation practices, antecedent moisture and groundwater depths in paddy fields Baoli Xu a , Dongguo Shao a,∗ , Xuezhi Tan a,b , Xia Yang a , Wenquan Gu a , Haoxin Li a a b

State Key Laboratory of Water Resources and Hydropower Engineering Science, Wuhan University, Wuhan 430072, PR China Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, T6G2W2, Canada

a r t i c l e

i n f o

Article history: Received 23 December 2016 Received in revised form 28 April 2017 Accepted 2 June 2017 Keywords: Deep percolation Antecedent soil moisture Irrigation amount Irrigation duration Groundwater depth HYDRUS-1D

a b s t r a c t Understanding soil water percolation in paddy fields is helpful to optimize irrigation schedule for rice production and improve water use efficiency under various irrigation practices and groundwater depths. Calibrated HYDRUS-1D model was used to simulate soil water movement and water balance in this study. We conducted scenario analyses based on the model to evaluate the combined effects of irrigation amount in an irrigation event (irrigation amount), irrigation duration, antecedent soil moisture, and groundwater depth on deep percolation (DP) in paddy fields. Results showed that during an irrigation event, there would be higher DP in paddy fields with higher antecedent soil moisture (≥−10 kPa), larger irrigation amount (7 cm) and/or free drainage in the bottom of rice root zones. We also used a classification and regression tree model to analyze the relative contribution of different factors to DP. Results indicated that antecedent soil moisture was the primary factor that contributed 46.3% of DP variation. Groundwater depth contributed 32.5% of DP variation, while irrigation amount (18.7%) and irrigation duration (2.5%) contributed least for DP variation. Furthermore, effects of these factors on DP interacted with each other. In scenario analysis, the contribution of antecedent soil moisture increased from 16.1% to 65.2% as the groundwater depth increased. When irrigation amount rose from 1 cm to 5 cm, the contributions of antecedent soil moisture increased to 77.6% from 57.1%; when irrigation amount was 7 cm, the contributions of antecedent soil moisture decreased to 46.4%. The contribution of irrigation amount rose to 55.7% from 28.4% with the increase of antecedent soil moisture, while the contributions of groundwater depth to DP showed opposite variation to irrigation amount as antecedent soil moisture altered. Based on relative contribution of these factors, optimal combinations of irrigation practices, antecedent soil moisture and groundwater depth were screened out to control DP for promoting rice growth and improving water use efficiency. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Percolation is a vital component of water balance in hydrologic processes in which water moves downward to groundwater. As a pathway for water losses from the rice root zone, deep percolation (DP) reduces water use efficiency in paddy field and accounts for 50–80% of water input (Belder et al., 2007; Cesari de Maria et al., 2016). Moreover, nitrogen leaching which is a threat to the groundwater environment is closely linked with DP (Refsgaard et al., 1999;

∗ Corresponding author. E-mail addresses: [email protected] (B. Xu), [email protected] (D. Shao), [email protected] (X. Tan), [email protected] (X. Yang), [email protected] (W. Gu), [email protected] (H. Li). http://dx.doi.org/10.1016/j.agwat.2017.06.002 0378-3774/© 2017 Elsevier B.V. All rights reserved.

Bouman et al., 2007). Therefore, it is necessary to understand how percolation is generated and its influential factors in paddy fields. Even though lysimeter experiments (e.g., Bethune et al., 2008; Hatiye et al., 2016) could measure the percolation of very small paddy soil profiles, it is difficult to directly measure on-site percolation in paddy fields. Thus, field percolation is often estimated as the residual of field water balance (Wang et al., 2012). However, the estimated percolation resulting from water balance analysis was not always reliable because of uncertainties in measuring other water balance components such as evapotranspiration. Although lysimeter experiments can precisely measure percolation, it is expensive to set up instruments and lysimeter is limited to some standard paddy fields. Thus, lysimeters are rarely used to measure paddy fields percolation. Consequently, a process-based model with numerical solution to water movement has been a widely-

150

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

used and efficient approach to estimate percolation on field scale, due to its low cost and flexibility. In addition, the percolation in paddy fields under a much wider range of scenarios including those in extreme conditions can be assessed in numerical models with calibrated parameters (Bah et al., 2009; Jyotiprava Dash et al., 2014; Lai et al., 2016). DP was closely related to water input, and DP represented the largest water losses in rice fields, especially for irrigated fields (Li et al., 2014). Peng et al. (2011) and Tan et al. (2013) found that DP decreased in paddy fields under alternate wetting and drying (AWD) irrigation, because AWD irrigation significantly decreased irrigation frequency and water inputs compared to continuously flooded irrigation. Ochoa et al. (2007) indicated that DP was significantly correlated with amount of irrigation water applied in flooded irrigation. Nevertheless, Janssen and Lennartz (2009) reported that ponded water depth in paddy fields had little influence on percolation rates. Besides, water input (irrigation or rainfall) patterns were highly associated with DP and solute leaching from paddy fields during rice growth season (Wang et al., 2010; Schwen et al., 2012). The duration of each irrigation event eliminated the differences in the vertical component of the wetting front between pulse irrigation and continuous irrigation (Elmaloglou and Diamantopoulos, 2008). However, the combined effects of these factors on DP have not been studied for paddy fields under various irrigation practices and surface storage capacity. The antecedent soil moisture is a criterion of implementing irrigation in some irrigation methods (Kukal et al., 2005; Tuong et al., 2005; Luo et al., 2009). The relationships between DP and antecedent soil moisture have been explored in some studies. Low antecedent moisture allows the soil to store more water and DP will not occur until soil water content is higher than the field capacity (Hatiye et al., 2016), even though smaller antecedent pressure head of dry land may increase percolation and seepage rates in flooded fields (Chen et al., 2002). Lai et al. (2016) found that the antecedent soil moisture showed more significantly positive correlation with DP than water input and its characteristics. Nevertheless, DP was not sensitive to the antecedent soil moisture in the research of Ochoa et al. (2007) in which flood irrigation was applied. Bethune et al. (2008) found no significant correlation between DP and the antecedent soil moisture before irrigation events in paddy fields under surface irrigation. The discrepancies illustrate the necessity to further clarify the complexity of the relationship between DP and antecedent soil moisture, and the potential processes of soil water movement that affect the relationship between them. DP also changes along with the groundwater depths (GWD). In lowland paddy fields with a shallow depth, the vertical soil water movement can switch from DP into capillary rise, so the groundwater could contribute to rice-use water (Boling et al., 2007). The capillary rise is weakened with the higher groundwater depth (Luo and Sophocleous, 2010; Hatiye et al., 2016). Studies on the contribution of groundwater depth to DP or capillary rise have been reported. However, the contribution of groundwater depth to the exchange of soil water and groundwater needs to be quantified to our knowledge. In paddy fields, different combinations of water management methods instead of a single method are usually adopted, but

the combined effects of water management measures on DP are not well understood. The interaction of different influential factors on DP in paddy fields should been considered to enhance our understanding of DP process. Besides, the influences of water management measures of irrigation practices on water balance components were rarely studied using HYDRUS-1D. Furthermore, it is essential in establishing an optimal irrigation scheme or controlled irrigation to improve water use efficiency by controlling DP based on scenario analysis. In this study, using a calibrated HYDRUS-1D, factors including irrigation amount in an irrigation event (irrigation amount, IA), irrigation duration (ID), antecedent soil moisture (AM) before an irrigation event, and groundwater depth (GWD) were selected for scenarios analysis of DP in paddy fields. These factors are easy to be controlled in rice production. The objectives of this study are (i) to quantify relative contributions of different factors to DP, (ii) to analyze the interactive impacts of different factors on DP, and (iii) to explore an optimal field water management strategy to control DP. 2. Materials and methods 2.1. Field experiments The study site (112◦ 10 , 30◦ 49 ; elevation of 72 m) is located in Zhanghe Irrigation District (ZID), Jingmen City, China. The study site has a typical subtropical monsoon climate and receives average annual rainfall of 915.0 mm. 56.1% of the yearly rainfall occurs between May and September and rainfall has a high seasonal and annual variability. In ZID, annual 20 cm pan evaporation ranges from 1300 to 1800 mm and annual mean air temperature is 16 ◦ C. The soil is a typical lowland paddy soil. Rice (Oryza sativa) is the main crop planted. Irrigation water is supplied by the Zhanghe Reservoir and water saving irrigation is widely adopted in paddy fields for rice planting. Field experiments were conducted in 2010 and 2011 during the rice growing season by a split-plot design. Details of the experimental layout were described by Tan et al. (2014). The soil profile in paddy fields consisted of three layers, i.e., the cultivated horizon layer (CHL), the plow pan layer (PPL) and the illuvial horizon layer (IHL) (Table 1). Three replicates of 250 cm3 undisturbed soil cores were sampled in each layer. Bulk density and particle size distribution were determined with stove-drying method and soil particle size and shape measurement system (AZ-S0300), respectively. Soil saturated hydraulic conductivity was measured with constant head method and soil water retention was estimated with simplified evaporation method in Hyprop System, based on which van Genuchten’s –h relationships were optimized. Soil properties of each soil layer were shown in Table 1. During the rice growing season, alternate wetting and drying (AWD) irrigation was applied and the amount of irrigation volume for each irrigation event was 3 cm. The upper limit of ponded water depth was 10 cm throughout rice growing season. The lower limit of pressure head were 0 cm in turning green period, −50 cm in early tillering period and −150 cm in late tillering period at the depth of 18 cm, and in booting and heading period and milk ripening period the lower limit were −100 cm and −50 cm, respectively, at the depth of 33 cm. Drainage

Table 1 Soil particle classification and hydraulic parameters of soil profile in the study site. Soil layer (cm)

Sand (%)

Silt (%)

Clay (%)

b (g cm−1 )

 r (cm3 cm−3 )

 s (cm3 cm−3 )

␣ (cm−1 )

n

Ks (cm d−1 )

CHL (0–18) PPL(18–33) IHL (33–100)

20.2 16.1 36.4

45.5 44.7 37.2

34.3 39.2 26.4

1.33 1.56 1.43

0.098 0.069 0.062

0.43 0.38 0.41

0.021 0.011 0.034

1.31 1.23 1.41

7.43(7.89) 0.48(0.45) 18.2

CHL, cultivated horizon layer; PPL, plow pan layer; IHL, illuvial horizon layer; b , bulk density;  r , residual volumetric water contents;  s , saturated volumetric water contents; ␣ and n, fitting parameters of soil water characteristic curve; Ks , saturated hydraulic conductivity and the calibrated values were shown in parentheses.

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

and naturally drying were applied in yellow ripening period. After rice transplanting, groundwater depth varied between 67.5 and 85 cm in 2010, and between 66–92.5 cm in 2011. Daily soil pressure head at the depth of 18, 33 and 72 cm was measured by piezometers and tensiometers during the rice growing season.

151

and h4 = −16,000 cm). Root growth rate calculation was based on the assumption in HYDRUS-1D that root depth is the product of the maximum rooting depth and a root growth coefficient. And root growth was calculated as input during the growing season. 2.4. Initial and boundary conditions

2.2. Model description HYDRUS-1D software packages (Version 4.16) was used to model the vertical water flow in this study. The model was developed to simulate water, heat and solute movement in onedimensional variably saturated media (Simunek et al., 2008). Water flow is described by the Richards’ equation:



∂ ∂ = K (h) ∂t ∂z





∂h +1 ∂z

-S

(1)

where h is the soil water pressure head (cm),  is the volumetric water content (cm3 cm−3 ), t is time (T), Z is the spatial coordinate (cm, positive upward), K(h) is the unsaturated hydraulic conductivity function (cm d−1 ) and S is the sink term referring to root water uptake (cm d−1 ). The unsaturated soil hydraulic K–h and –h relationships are general highly nonlinear. According to van Genuchten (1980), the soil water retention and hydraulic conductivity functions are given as:



(h) =

r +

s −r m (1+|˛h|n )

, h<0

(2)

s , h≥0



K(h) = Ks Se l 1 − (1 − Se 1/m )



m 2

(3)

In the study, soil water flow was simulated in soil profile with the thickness of 100 cm. The initial conditions were defined using soil pressure head distribution measured in field experiments during the model calibrating and validating period. Initial soil moisture in CHL (0–18 cm under soil surface) was saturated due to ponded water in paddy fields before rice transplanting. Initial soil moisture between the bottom of CHL and groundwater level was linearly interpolated in model. The upper boundary condition was the atmospheric boundary condition with a surface water layer (with a maximum of 10 cm), represented by the values of rainfall, irrigation, Ep and Tp fluxes. Irrigation in model simulations was treated as rainfall in paddy fields regardless of spatial heterogeneity. The lower boundary condition was pressure head derived from groundwater table. The left and right side boundaries were set as no-flux boundaries. 2.5. Model performance criteria Simulated soil pressure head of the model was compared with the observed values using two statistical parameters:



n

1 S − O 2 i i RE =  n

n

where:  − r 1 Se = m=1− n s − r

(4)

The above equations contain five independent parameters:  r and s denote the residual and saturated volumetric water contents (cm3 cm−3 ), respectively; ␣ (cm−1 ) and n (–) are empirical coefficients of soil water characteristic curve; l (–) is the pore connectivity parameter (=0.5); and Se (–) is the effective saturation. 2.3. Rice root water uptake functions Daily reference evapotranspiration (ET0 ) during rice growing season was computed by Penman-Monteith equation (Allen et al., 1998) based on available climate data from Tuanlin Irrigation Experimental Station. Rice potential evapotranspiration (ETC ) was derived from multiplying ET0 by rice coefficients (KC ) in different development stage (Allen et al., 1998). ETC was partitioned into potential soil evaporation (Tp ) and potential rice transpiration (Ep ) based on leaf area index (LAI) as a function of the growth stage (Pachepsky et al., 2004). Ep = ETC × exp−ˇLAI

(5)

Tp = ETC − Ep

(6)

Where ˇ is an extension coefficient for global solar radiation and the value of ˇ was taken as 0.3 for rice (Singh et al., 2003). LAI at different rice growth stages was collected from Tuanlin Irrigation Experimental Station. Actual root water uptake, S(z, t), was calculated in terms of Tp , soil water moisture and root growth and distribution using the general function of Feddes et al. (1978) in HYDRUS-1D package. The threshed parameters were used to parameterize the water stress response function were calibrated by Singh et al. (2003) for rice (h1 = 100 cm, h2 = 55 cm, h3 (high) = −160 cm, h3 (low) = −250 cm,

NSE = 1 −

(7)

Oi

i=1

(Oi − Si )2

i=1 n

(Oi − Om )

(8) 2

i=1

where RE and NSE are the root mean square error and Nash-Sutcliffe modeling efficiency (Nash and Sutcliffe, 1970), respectively, n is the number of observations, Oi is the observed values, Si is the simulated values, Om is the mean values of observed data. 2.6. Scenarios In order to determine the impacts of irrigation practices (irrigation amount and duration), antecedent soil moisture and groundwater depths on DP, scenario analysis of DP in irrigation events was conducted in the study: four antecedent soil moisture (AM) conditions, four irrigation amount (IA) gradients, three irrigation duration (ID) modes, and four groundwater depth (GWD) control methods. Irrigation is carried out depending on the rice growth status and soil moisture. Four IA gradients were selected derived from differTable 2 Details of antecedent soil moisture, irrigation amount, irrigation duration, and groundwater depths of scenario analysis. Influential factors

Scenario values

Antecedent Soil Moisture (AM, kPa) Irrigation Amount (IA, cm) Irrigation Duration (ID, hour) Groundwater Depth (GWD, cm)

−30 1 1 30

−20 3 3 60

−10 5 5 90

−2 7 – FR

Note: The form of influential factor scenario value below means an influential factor with a certain value. For example, IA 1 means amount of water per irrigation (irrigation amount) is 1 cm. FR referred to free drainage.

152

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

Fig. 1. Observed and simulated soil pressure head in three soil layers during calibrating (a, 2010) and validating (b, 2011) periods with HYDRUS-1D simulation.

ent degrees of alternate wetting and drying irrigation (Mao, 2002) and local practice. Under different IA gradients, three ID modes were considered. Short ID usually means larger irrigation intensity with identified irrigation amount. ID was determined by field area, water supply capacity of the channel or pump and IA in a paddy field. In ZID, ID ranged within several hours in a field, and it was set according to field experiments and investigation. Soil water content in CHL has been used as a trigger for irrigation (Kukal et al., 2005; Tuong et al., 2005; Luo et al., 2009). Antecedent soil moisture of CHL before irrigation in scenario analysis ranged from −30 kPa to nearly saturated (Table 2). The antecedent soil moisture was set as the initial condition of model simulation in scenario analysis. Similarly, four groundwater depths were applied in scenario analysis as Table 2 showed. Correspondingly, the boundary condition was set as pressure head with groundwater depth of 30 cm, 60 cm and

90 cm, while free drainage bottom was adopted when groundwater depth was very deep (free drainage) in the other scenarios. Moreover, the average daily ETC was 3.58 mm d−1 during rice growing season (from June 1st to August 31st) with meteorological data in 2000–2016 from Tuanlin Irrigation Experimental Station. Variation of ETC within a day was estimated according to the assumption of Fayer (2000), in which ETC between 18:00 and 6:00 in the next day constitutes 1% of the total daily ETC and the rest ETC of other 12 h follows a sinusoidal shape distribution. And then Ep and Tp in scenario analysis were estimated by Eq. (5) and Eq. (6). According to field survey in ZID, paddy fields are usually irrigated in the morning to reduce the evaporation loss. Therefore, irrigation in scenario analysis was carried out at 6:00. Based on the alternate wetting and drying irrigation practice (Mutero et al., 2000; Mao, 2002), simulated cycle (an irrigation event) was set as 7 days. Cor-

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158 Table 3 Model performance statistics of simulated soil pressure head in calibration (2010) and validation (2011) periods. Layer depth

2010 2011

18 cm

33 cm

153

2

72 cm

RMSE(cm)

NSE

RMSE(cm)

NSE

RMSE(cm)

NSE

0

14.66 104.42

0.93 0.91

5.92 10.27

0.98 0.93

1.70 1.88

0.98 0.93

-2

RMSE and NSE were root mean square error and Nash-Sutcliffe modeling efficiency, respectively.

-4 -6

respondingly, the cumulative DP in each scenario was estimated with the calibrated HYDRUS-1D model.

IA_1

IA_3

IA_5

IA_7

-8 AM_-30

2.7. Data analysis

AM_-20

AM_-10

AM_-2

3

The correlation between DP and studied factors was analyzed with Spearman correlation coefficient, and the differences of DP among scenarios of one single factor under various gradients with other factors’ scenarios grouped were assessed by the analysis of variance (ANOVA) in SPSS 22.0 (SPSS 22.0, SPSS Inc., IL, USA). Relative contribution of antecedent soil moisture, irrigation amount, irrigation duration, and groundwater depth to DP was quantified by the analysis of classification and regression tree (CART) model in Clementine 12.0. Additionally, the integrated influences of scenario analysis factors were also evaluated with the CART model.

1 -1 -3 -5

GWD_30

GWD_60

GWD_90

GWD_FR

-7 -9

3. Results and discussion

AM_-30

3.1. HYDRUS-1D model calibration and validation

2

Many studies have proven that HYDRUS-1D is a reliable tool to describe water movement in saturated-unsaturated soil profile, such as paddy field soil. In this study, water flow model based on HYDRUS-1D was calibrated and validated using field experimental soil pressure head in 2010 and 2011(Fig. 1), respectively. Simulated results of HYDRUS-1D were in excellent agreement with the observed values of the pressure head at soil depths of 18, 33 and 72 cm. The NSEs were larger than 0.91 in calibration and validation periods (Table 3). The simulation performed well and soil hydraulic parameters can represent the soil profile features, although the simulation overestimated pressure head in late growing season in 2011. Moreover, the simulation also accurately described soil pressure head before and after irrigation events triggered by low pressure head (Fig. 1). Based on the simulation, water balance of the soil profile was analyzed (Table 4) and the closure errors were relatively small indicating that HYDRUS-1D can produce relative exact water balance components. Thus, the model in the study can be used to simulate bottom flux in the following scenario analysis.

0

3.2. Bottom flux in scenarios 192 cumulative bottom fluxes (deep percolation or groundwater capillary rise) at the depth of 1 m in corresponding irrigation events were obtained by HYDRUS-1D simulation. Deep percolation was detected in 59.4% of irrigation events, and most (90) of deep percolation occurred when higher antecedent soil moisture (AM, ≥−10 kPa), larger irrigation amount (IA, 7 cm) or free drainage

AM_-20

AM_-10

AM_-2

-2 ID_1

-4

ID_3

ID_5

-6

Fig. 2. Mean cumulative bottom fluxes under four antecedent moisture (AM) conditions with other factor identified. The negative values represent DP while the positive values represent capillary rise, and it is the same as below. Note: IA, irrigation amount; ID, irrigation duration; GWD, groundwater depth. The form of influential factor scenario value below means an influential factor with a certain value. For example, IA 1 means amount of water per irrigation (irrigation amount) is 1 cm.

was adopted. Specially, DP was higher than 5 mm d−1 in 22.9% of the irrigation events and only 3 irrigation events with DP were under AM 20 and AM −30. Groundwater capillary rise happened in the rest irrigation events. Lower antecedent soil moistures make it more possible for groundwater capillary rise to occur (Fig. 2). Fig. 2 showed mean bottom fluxes (deep percolation or groundwater capillary rise) of scenario analysis. As antecedent soil moistures increased, groundwater capillary rise turned into deep percolation at the bottom boundary with other factors identified

Table 4 Water balance analysis (mm) of simulation in calibration (2010) and validation (2011) periods.

2010 2011

Rainfall

Irrigation

Ec

Tc

Percolation

SWC

Closure error

235.60 366.5

390 210

149.2 152.3

255.8 274.4

284.3 224.4

64.2 61.1

0.5 −13.5

Ec , Tc were actual evaporation and actual transpiration, respectively; SWC was the simulated soil water consumption during rice growing season.

154

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

Table 5 The mean value and standard error (SE) of cumulative DP with the factor identified and the results of ANOVA (p < 0.05).

d

IA (cm)

DP (cm) Meand

SEe

1 3 5 7

0.67a −0.22a −1.34b −2.78c

0.42 0.42 0.45 0.42

AM (kPa) −30 −20 −10 −2

DP (cm) Meand

SEe

1.44a 0.49a −1.60b −4.00c

0.33 0.31 0.36 0.43

GWD (cm) 30 60 90 FR

DP (cm) Meand

SEe

−0.63a 0.46a −0.14a −3.65b

0.32 0.38 0.52 0.39

ID (h) 1 3 5 –

DP (cm) Meand

SEf

−1.14a −0.88a −0.74a –

0.40 0.41 0.42 –

Different letters (a, b, c) within the individual columns indicated a significant difference (p < 0.05) with a post hoc LSD test; e , n = 48; f , n = 64.

(Fig. 2). For example, in IA 1 mode, the average of cumulative bottom fluxes changed from 2.50 cm to −1.96 cm as the antecedent soil moisture increasing from −30 kPa to −2 kPa (Fig. 2a). When antecedent soil moisture was identified, larger IA resulted in

less groundwater capillary rise and higher deep percolation as expected. Threshold of IA that switched bottom flux from groundwater capillary rise to deep percolation decreased with the increase of antecedent soil moisture (Fig. 2a). The mean cumulative bottom

Fig. 3. Cloud chart of DP under different antecedent moisture conditions (AM), irrigation amount (IA) and groundwater depths (GWD). ID 1, ID 3, ID 5: irrigation duration is 1 h, 3 h and 5 h, respectively.

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

155

Fig. 4. Relative contribution rates of different factors under different antecedent moisture conditions (AM), groundwater depths (GWD) and irrigation practices (IA, irrigation amount; ID, irrigation duration).

fluxes were almost negative with antecedent soil moisture higher than −10 kPa despite of IA. ANOVA analysis results (Table 5) showed that GWD had a significant effect on bottom fluxes. The trend of bottom fluxes under different GWDs was similar to that of IA with the increase of antecedent soil moisture. The mean cumulative bottom fluxes were positive when the combinations of GWD 60, GWD 90 and AM –30 kPa, AM −20 kPa were adopted (Fig. 2b). Higher antecedent soil moisture saw less groundwater capillary rise and larger increasing gradient of deep percolation. GWD FR induced larger deep percolation compared to other groundwater depth controlling modes in all of the simulated irrigation events (Fig. 2b and Fig. 3), indicating that controlled groundwater level modes could enhance groundwater capillary rise and decrease deep percolation. However, in GWD 30 mode, capillary rise was only found when IA 1 and IA 3 modes were applied with antecedent soil moisture lower than −10 kPa (Fig. 3), indicating that higher groundwater level might contribute to deep percolation even though it could potentially increase the chance of groundwater capillary rise occurring (Liu et al., 2006; Luo and Sophocleous, 2010). Therefore, groundwater management should be adjusted to corresponding antecedent moisture and irrigation water input. Deep percolation or groundwater capillary rise did not obviously respond to irrigation duration (Fig. 2c and Table 5), reasons for which can be attributed to the followings. Firstly, the ranges of irrigation duration adopted in scenario analysis were small so that influence on DP might not be fully represented. Secondly, the soil profile features affected water infiltration. Wang et al. (2010) found that deep percolation dramatically increased as the irrigation rate increased at a field site with relative high hydraulic conductivity (12.36–19.35 cm d−1 at the depth of 0–150 cm of soil profile). However, hydraulic conductivity of soil profile in this study was low (Table 1), especially for the PPL. Infiltration was retarded by the compact PPL, thereby weakening the impact of irrigation duration. Even though irrigation duration showed limited influence on bottom fluxes, the longer irrigation duration potentially decreased deep percolation and increased groundwater capillary rise. Shal-

lower groundwater depths and higher irrigation amount enhanced the effect of irrigation duration on bottom flux (Fig. 3). 3.3. Relative contribution of influential variables As the analysis of CART, antecedent soil moisture has the highest relative contribution (46.3%) among four factors in the scenario analysis. The relative contributions of GWD and IA were 32.5% and 18.7%, respectively. The ID was only responsible for 2.5% of DP variation, further implying its limited effects. DP has been reported to be related to antecedent soil moisture and irrigation or rainfall (Chen et al., 2002; Hatiye et al., 2016). However, the contributions of factors above to DP were not consistent. Higher relative contributions of the AM in this study were found compared to the result of Ochoa et al. (2007) in which IA dominated DP process with the flood irrigation adopted, while antecedent soil moisture contributes more (85.7%) to DP under various rainfall patterns as Lai et al. (2016) reported. The difference indicated that the contribution of a certain factor to DP was influenced by other factors. The interaction analysis among the factors showed similar results between Spearman correlation analysis and the CART, and only the results of CART were demonstrated and discussed below. The relative contributions of irrigation amount, irrigation duration and groundwater depth to DP altered under different antecedent soil moisture conditions. In the study, GWD was the second important factor to DP variation. Almost half or more of DP variation was attributed to GWD when antecedent was identified (Fig. 4). The relative contribution of GWD decreased as antecedent moisture rose, while the relative contributions of irrigation amount and irrigation duration showed contrast trend. With antecedent soil moisture lower than −10 kPa, capillary rise dominated the bottom flux in some scenarios (Figs. 2 and 3). Soil suction reduced with the increase of soil moisture which weakened groundwater capillary rise (Liu et al., 2006) so as to reduce the contribution of GWD and enhance the influence of irrigation practice. When soil water moisture gradually increased, groundwater capillary rise was transformed into deep percolation so that the role of irriga-

156

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

Table 6 Optimal combinations of irrigation practice, antecedent moisture and groundwater depth for an irrigation event. (For interpretation of the references to colour in this table, the reader is referred to the web version of this article.)

Ta is the time that soil moisture was just below or reached the critical antecedent moisture of irrigation operation. BTTa and DBTTa is the cumulative bottom flux and averaged daily bottom flux at the time of Ta, respectively. Positive values refer to capillary rise and negative values refer to deep percolation.

tion amount and duration in DP variation became more important. The increase of irrigation amount strengthened the relative contribution of irrigation duration indicating that irrigation duration played more important role in water transformation under larger irrigation amount. As irrigation amount rose from 1 cm to 5 cm, the relative contribution of antecedent moisture increased and showed higher increase rate with lower irrigation amount (≤3 cm) applied. And in the same condition the contributions of GWD weakened. The increasing irrigation amount replaced or diluted the role of groundwater as a bottom flux source, thus led to a reduction of the relative contribution of GWD while increased the contribution of AM. It indicated that capillary rise might subside when irrigation amount was larger for each irrigation event (Fig. 2). In order to reduce DP, the combinations of high irrigation amount and antecedent moisture should be avoided. When irrigation amount was 7 cm, antecedent soil moisture and groundwater depth were two main reasons for DP variation and played almost comparable roles. Groundwater depth conditions also affected the relative contribution of other factors. The relative contribution of IA and AM showed contrast trend as groundwater depth increased. With extremely shallow water depth (30 cm), 80.8% of DP variation was attributed to irrigation amount, while the contribution of antecedent soil moisture exceeded 52.8% under other groundwater depth conditions. When groundwater depth was shallow, there was less room for soil profile to store water so that the relative contribution of AM was small. When groundwater depth was larger, however, the contribution of antecedent soil moisture decreased with the increase of groundwater depth. Contribution of ID was limited as shown in Fig. 4. However, the relative contribution of antecedent soil moisture, irrigation amount and groundwater depth differed under different irrigation duration conditions. The contribution of groundwater depth declined as irrigation duration became longer, while irrigation amount became more important, although antecedent soil moisture was still the most influential factor.

3.4. Optimization of water management combination Statistics of DP in scenario analyses were shown above and DP varied with different combinations of antecedent soil moisture, irrigation amount, irrigation duration, and groundwater depth. Thus the appropriate water management methods could be selected based on irrigation implementation and relevant requirements of DP. Among the 192 combinations, the time at which soil moisture of the CHL reached the antecedent soil moisture was from the 21st h to 168th h, meaning that irrigation intervals ranged between 21 h and 7 days. Shorter irrigation interval implies that farmers have to devote more time or even economic inputs because of frequent irrigation application, which is often avoided in irrigation scheduling. 5 days was determined in this study as the critical irrigation interval in accordance with the practice local farmers and some scholars adopted (Belder et al., 2004; Tan et al., 2013); thereby, 120 reasonable combinations were screened out and DP ranged from −12.29 to 6.21 mm d−1 , in which only irrigation amounts of 3 cm, 5 cm and 7 cm were then detected with antecedent soil moisture below −2 kPa. 47 combinations were found that groundwater capillary contributed to soil moisture in an irrigation event among 120 reasonable combinations, in which the combination of (3, ID, −30, 60/90) and (5, ID, −30, 90) demonstrated the highest capacity of groundwater capillary on average (>5 mm d−1 ). The increase of ID and/or the decrease of antecedent soil moisture both increased groundwater capillary, and so is the cooperation of less irrigation amount and shallower groundwater depth as Table 6 showed. These 47 combinations, especially for the combination of (3, ID, −30, 60/90) and (5, ID, −30, 90), eliminated the risk of water loss and leaching pollution by enhancing the groundwater capillary. In order to improve water use efficiency, these combinations of irrigation practices and groundwater depths were preferential for water management in paddy fields, especially when fertilizer is applied. However, DP of 9–15 mm d−1 might be necessary to rice growth during certain growth stages in the mid subtropical region (Xu et al.,

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

1998). Soil with low leaching rate often has oxidation-reduction potential too low for rice growth (Shan et al., 2005). Proper percolation is also beneficial to rice growth by regulating temperature and reducing salinity in paddy soil. Considering the threshold of DP, 6 of the 120 reasonable combinations (7, ID,−2, FR; 7, ID, −2, 90) met the requirements with irrigation amount of 7 cm and the antecedent soil moisture of −2 kPa (Table 6), in which DP was less with groundwater depth of 90 cm. As analyzed above, the appropriate combinations of the antecedent soil moisture, irrigation amount and groundwater depth as well as irrigation duration for an irrigation event were determined based on simulated irrigation interval and DP. However, the combinations are dynamically changing during rice growing period. Irrigation practices and groundwater depths should be adjusted synthetically considering rice growth and water use efficiency. Furthermore, DP may be also influenced by preferential flow through soil cracks as a result of AWD (Garg et al., 2009). Therefore, further study of irrigation practices and groundwater depths on DP during the whole rice growing stage was still needed. 4. Conclusions This study assessed the effects of irrigation practices and groundwater depths on DP with scenario analyses using HYDRUS1D in a paddy field site. The calibrated model based on HYDRUS-1D presented an excellent simulation of soil water dynamics. Among the 192 scenarios, both groundwater capillary rise and deep percolation (DP) were detected. The thresholds of the transformation between groundwater capillary rise and DP for antecedent soil moisture and irrigation amount were −10 ∼ −2 kPa and 5–7 cm, respectively. Too shallow (30 cm) or too deep groundwater depth (free drainage) contributed to DP, and it was more possible for capillary rise to occur with groundwater depth between 60 cm and 90 cm. According to the results of classification and regression tree model, antecedent soil moisture was the greatest factors influencing DP, and followed by groundwater depth and irrigation amount, while irrigation duration showed limited effects. This study also found the interaction among irrigation amount and duration, antecedent soil moisture and groundwater depth on DP. The contribution of antecedent soil moisture increased with the increase of groundwater depth and irrigation amount (from 1 cm to 5 cm) and decreased with irrigation amount of 7 cm. And the contribution of irrigation amount increased as antecedent soil moisture became moderate wet. The contribution of groundwater showed inverse variation compared to irrigation amount when antecedent soil moisture altered. After analyzing the influence of irrigation practices and groundwater depth on DP, 120 combinations were selected considering irrigation interval, and optimal combinations were furthermore screened out to control DP generation according to the rice growth and water use efficiency. Acknowledgements This work was jointly supported by the National Natural Science Foundation of China (grant number 51439006), the National Key Research and Development Program of China (no. 2016YFC0400101), and the National Natural Science Foundation of China (grant number 51379150). References Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration – guide-lines for computing crop water requirements. In: FAO Irrigation and Drainage Paper 56. Food and Agriculture Organization (FAO) of the United Nations, Rome, Italy.

157

Bah, A.R., Kravchuk, O., Kirchhof, G., 2009. Sensitivity of drainage to rainfall, vegetation and soil characteristics. Comput. Electron. Agric. 68, 1–8, http://dx. doi.org/10.1016/j.compag.2009.03.005. Belder, P., Bouman, B.A.M., Cabangon, R., Guoan, L., Quilang, E.J.P., Yuanhua, L., Spiertz, J.H.J., Tuong, T.P., 2004. Effect of water-saving irrigation on rice yield and water use in typical lowland conditions in Asia. Agric. Water Manag. 65, 193–210, http://dx.doi.org/10.1016/j.agwat.2003.09.002. Belder, P., Bouman, B.A.M., Spiertz, J.H.J., 2007. Exploring options for water savings in lowland rice using a modelling approach. Agric. Syst. 92, 91–114, http://dx. doi.org/10.1016/j.agsy.2006.03.001. Bethune, M.G., Selle, B., Wang, Q.J., 2008. Understanding and predicting deep percolation under surface irrigation. Water Resour. Res. 44, 1–16, http://dx.doi. org/10.1029/2007WR006380. Boling, A.A., Bouman, B.A.M., Tuong, T.P., Murty, M.V.R., Jatmiko, S.Y., 2007. Modelling the effect of groundwater depth on yield-increasing interventions in rainfed lowland rice in Central Java, Indonesia. Agric. Syst. 92, 115–139, http:// dx.doi.org/10.1016/j.agsy.2006.05.003. Bouman, B.A.M., Humphreys, E., Tuong, T.P., Barker, R., 2007. Rice and water. Adv. Agron. 92, 187–237, http://dx.doi.org/10.1016/S0065-2113(04)92004-4. Cesari de Maria, S., Rienzner, M., Facchi, A., Chiaradia, E.A., Romani, M., Gandolfi, C., 2016. Water balance implications of switching from continuous submergence to flush irrigation in a rice-growing district. Agric. Water Manag. 171, 108–119, http://dx.doi.org/10.1016/j.agwat.2016.03.018. Chen, S.-K., Liu, C.W., Huang, H.-C., 2002. Analysis of water movement in paddy rice fields (II) simulation studies. J. Hydrol. 268, 259–271, http://dx.doi.org/10. 1016/S0022-1694(02)00180-4. Elmaloglou, S., Diamantopoulos, E., 2008. The effect of intermittent water application by surface point sources on the soil moisture dynamics and on deep percolation under the root zone. Comput. Electron. Agric. 62, 266–275, http://dx.doi.org/10.1016/j.compag.2008.01.008. Fayer, M.J., 2000. UNSAT-H Version 3.0: Unsaturated Soil Water and Heat Flow Model. Theory, User Manual, and Examples. Model Man., pp. 184. Feddes, R.A., Kowalik, P.J., Zaradny, H., 1978. Simulation of Field Water Use and CropYield. John Wiley & Sons, New York, NY [Online]. Garg, K.K., Das, B.S., Safeeq, M., Bhadoria, P.B.S., 2009. Measurement and modeling of soil water regime in a lowland paddy field showing preferential transport. Agric. Water Manag. 96 (12), 1705–1714, http://dx.doi.org/10.1016/j.agwat. 2009.06.018. Hatiye, S.D., Prasad, K.S.H., Ojha, C.S.P., Adeloye, A.J., 2016. Estimation and characterization of deep percolation from rice and berseem fields using lysimeter experiments on sandy loam soil. J. Hydrol. Eng. 21, 1943–5584, http://dx.doi.org/10.1061/(ASCE)HE0001365. Janssen, M., Lennartz, B., 2009. Water losses through paddy bunds: methods, experimental data, and simulation studies. J. Hydrol. 369, 142–153, http://dx. doi.org/10.1016/j.jhydrol.2009.02.038. Jyotiprava Dash, C., Sarangi, A., Singh, D.K., Singh, A.K., Adhikary, P.P., 2014. Prediction of root zone water and nitrogen balance in an irrigated rice field using a simulation model. Paddy Water Environ. 13, 281–290, http://dx.doi. org/10.1007/s10333-014-0439-x. Kukal, S.S., Hira, G.S., Sidhu, A.S., 2005. Soil matric potential-based irrigation scheduling to rice (Oryza sativa). Irrig. Sci. 23, 153–159, http://dx.doi.org/10. 1007/s00271-005-0103-8. Lai, X., Liao, K., Feng, H., Zhu, Q., 2016. Responses of soil water percolation to dynamic interactions among rainfall, antecedent moisture and season in a forest site. J. Hydrol. 540, 565–573, http://dx.doi.org/10.1016/j.jhydrol.2016. 06.038. Li, Y., Simunek, J., Jing, L.F., Zhang, Z.T., Ni, L.X., 2014. Evaluation of water movement and water losses in a direct-seeded-rice field experiment using Hydrus-1D. Agric. Water Manag. 142, 38–46, http://dx.doi.org/10.1016/j. agwat.2014.04.021. Liu, Y., Pereira, L.S., Fernando, R.M., 2006. Fluxes through the bottom boundary of the root zone in silty soils: parametric approaches to estimate groundwater contribution and percolation. Agric. Water Manag. 84, 27–40, http://dx.doi. org/10.1016/j.agwat.2006.01.018. Luo, Y., Sophocleous, M., 2010. Seasonal groundwater contribution to crop-water use assessed with lysimeter observations and model simulations. J. Hydrol. 389, 325–335, http://dx.doi.org/10.1016/j.jhydrol.2010.06.011. Luo, Y., Khan, S., Cui, Y., Peng, S., 2009. Application of system dynamics approach for time varying water balance in aerobic paddy fields. Paddy Water Environ. 7, 1–9, http://dx.doi.org/10.1007/s10333-008-0146-6. Mao, Z., 2002. Water efficient irrigation and environmentally sustainable irrigated rice production in China. Int. Comm. Irrig. Drain., 1–15. Mutero, C.M., Blank, H., Konradsen, F., Van Der Hoek, W., 2000. Water management for controlling the breeding of Anopheles mosquitoes in rice irrigation schemes in Kenya. Acta Trop. 76, 253–263, http://dx.doi.org/10.1016/S0001706X(00)00109-1. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I – a discussion of principles. J. Hydrol. 10, 282–290, http://dx.doi.org/10. 1016/0022-1694(70)90255-6. Ochoa, C.G., Fernald, A.G., Guldan, S.J., Shukla, M.K., 2007. Deep percolation and its effects on shallow groundwater level rise following flood irrigation. T. Asabe 50 (1), 73–81, http://dx.doi.org/10.13031/2013.22413. Pachepsky, Y.A., Smettem, K.R.J., Vanderborght, J., Herbst, M., Vereecken, H., Wosten, J.H.M., et al., 2004. Reality and fiction of models and data in soil hydrology. In: Fed-des, R.A. (Ed.), Unsaturated Zone Modeling. Kluwer Academic Publishers, Dordrecht.

158

B. Xu et al. / Agricultural Water Management 192 (2017) 149–158

Peng, S.Z., Yang, S.H., Xu, J.Z., Luo, Y.F., Hou, H.J., 2011. Nitrogen and phosphorus leaching losses from paddy fields with different water and nitrogen managements. Paddy Water Environ. 9, 333–342, http://dx.doi.org/10.1007/ s10333-010-0246-y. Refsgaard, J.C., Thorsen, M., Jensen, J.B., Kleeschulte, S., Hansen, S., 1999. Large scale modelling of groundwater contamination from nitrate leaching. J. Hydrol. 221, 117–140, http://dx.doi.org/10.1016/S0022-1694(99)00081-5. Schwen, A., Yang, Y., Walton, R.J., 2012. A new experimental design reveals the impacts of land use and irrigation characteristics on bromide leaching. Vadose Zone J. 11 (4), http://dx.doi.org/10.2136/vzj2012.0077, 2344-2344. Shan, Y.H., Yang, L.Z., Yan, T.M., Wang, J.G., 2005. Downward movement of phosphorus in paddy soil installed in large-scale monolith lysimeters. Agric. Ecosyst. Environ. 111, 270–278, http://dx.doi.org/10.1016/j.agee.2005.05.011. Simunek, J., Sejna, M., van Genuchten, MTh., 2008. The HYDRUS-1D Software Package for Simulating the Onedimensional Movement of Water, Heat, and Multiple Solutes in Variably-Saturated Media. Version 4.08. Dep.of Environmental Sciences Univ.of California Riverside. Singh, R., van Dam, J., Jhorar, R.K., 2003. Water and salt balances at farmerfields. In: van Dam, J.C., Malik, R.S. (Eds.), Water Productivity of Irrigated Crops in Sirsa District, India. Integration of Remote Sensing, Crop and SoilModels and Geographical Information Systems. (http://www.futurewater.nl/downloads/2003VanDam WatPro.pdf).

Tan, X., Shao, D., Liu, H., Yang, F., Xiao, C., Yang, H., 2013. Effects of alternate wetting and drying irrigation on percolation and nitrogen leaching in paddy fields. Paddy Water Environ. 11, 381–395, http://dx.doi.org/10.1007/s10333012-0328-0. Tan, X., Shao, D., Liu, H., 2014. Simulating soil water regime in lowland paddy fields under different water managements using HYDRUS-1D. Agric. Water Manag. 132, 69–78, http://dx.doi.org/10.1016/j.agwat.2013.10.009. Tuong, T.P., Bouman, B.a.M., Mortimer, M., 2005. More rice, less water-integrated approaches for increasing water productivity in irrigated rice-based systems in Asia. Plant Prod. Sci. 8, 231–241, http://dx.doi.org/10.1626/pps.8.231. Wang, H., Ju, X., Wei, Y., Li, B., Zhao, L., Hu, K., 2010. Simulation of bromide and nitrate leaching under heavy rainfall and high-intensity irrigation rates in North China Plain. Agric. Water Manag. 97, 1646–1654, http://dx.doi.org/10. 1016/j.agwat.2010.05.022. Wang, P., Song, X., Han, D., Zhang, Y., Zhang, B., 2012. Determination of evaporation, transpiration and deep percolation of summer corn and winter wheat after irrigation. Agric. Water Manag. 105, 32–37, http://dx.doi.org/10. 1016/j.agwat.2011.12.024. Xu, Q., Yang, L.Z., Dong, Y.H., 1998. Rice Field Ecosystem in China. Chinese Agriculture Press, Beijing, pp. 150–156 (in Chinese). van Genuchten, M.T., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils1. Soil Sci. Soc. Am. J. 44, 892, http://dx.doi. org/10.2136/sssaj1980.03615995004400050002x.