Evaluating Maize Drought and Wet Stress in a Converted Japanese Paddy Field Using a SWAP Model

: Japanese government recommend farmers to cultivate upland crops in paddy ﬁelds (“converted ﬁelds”) to suppress the overproduction of rice. Converted ﬁelds are subject to excessively wet and dry conditions that reduce the yield of non-rice crops. Drought and wet stresses are critical to crop growth within speciﬁc growth periods. To provide data for use in mitigating crop yield reduction, we evaluated drought and wet stresses in maize ( Zea mays L.). A SWAP (soil–water–atmosphere–plant) model was applied to a converted maize ﬁeld. Observations were carried out in 2019 and 2018 for model calibration and validation. Thereafter, we evaluated the water stress of maize in 2019 (actual conditions) and at a tillage depth 11 cm deeper (scenario conditions). We found that (1) drought and wet stresses occurred within the relevant critical growth periods under actual conditions; (2) in the critical periods, the drought and wet stresses under scenario conditions were 33%–75% and 10%–82%, respectively, of those under actual conditions; (3) water stress at depths of 10 and 20 cm was lower under the scenario conditions than under the actual conditions. These results indicate that deeper tillage may mitigate both drought and wet stresses and can be used to reduce water stress damage in converted ﬁelds. under actual conditions, respectively; (3) water stress between soil depths of 10 and 20 cm was lower under the scenario conditions than under the actual conditions. Those results indicate that deeper tillage may mitigate both drought and wet stresses. We hypothesize that by evaluating drought and wet stresses at di ﬀ erent depths, we can identify the critical layer or depth for crop growth; this will promote further reduction in water-related crop damage in converted ﬁelds. The method presented here can be applied to ﬁelds with suboptimal water conditions, and may enhance yield. Further, we believe that our study will support precise stress control management, and boost crop yield and quality, when combined with precision farming applications in converted ﬁelds.


Introduction
In order to regulate rice production in relation to demand for it and other agricultural products, maize and other "upland" crops are cultivated in temporarily drained paddy fields, known as "converted fields", in Japan [1]. The drainage in paddy fields is poor, mainly due to hardpans that exist at approximately 20 cm below the soil surface. Converted fields are therefore at risk of waterlogging. Moreover, we must consider the interaction of weather conditions during the cultivation period. This period in Japan often includes the rainy season (from mid-June to mid-July, after mid-September), the dry season (from mid-July to beginning of September). Therefore, both drought and wet stresses often reduce upland crop yields, including those from converted fields [2].
Maize (Zea mays L.) cultivation for feed has been attempted in converted fields in southwestern Japan. Previous studies have shown that drought and wet stresses must be considered in relation to specific growth periods for maize cultivation. Çakir [3] demonstrated that drought stress at the 2 of 15 beginning of the vegetative or tasselling growth stages reduced crop height; irrigation conducted approximately 40-65 days after emergence of maize could eliminate this effect (tasselling is the stage at which stamens appear, following maximum growth). Moreover, Kanwar et al. [4] showed that maize yield was most significantly affected by wet stress at the flowering stage: 60-80 days after planting.
To quantify drought and wet stresses on the crop separately, we used a SWAP (soil-wateratmosphere-plant) model [5,6] in this study. This model quantifies those stresses using crops' critical pressure heads and has been applied in several studies. Anan et al. [7] evaluated the stresses in potato plants from precipitation and continuous drought in a reclaimed crop field, and Yuge et al. [8] conducted a scenario analysis to suggest an optimal irrigation regime in a reclaimed field with sorghum and potato. We hypothesized that we would be able to develop strategies to reduce water-related stress damage to crops by applying the SWAP model. This study focused on a converted maize field to evaluate (1) the water stress under actual field conditions and (2) the effect of tillage depth on water stress-we assumed this is one of the stress reduction strategies. We mainly evaluated draught and wet stresses in their critical growth periods [3,4].

Soil Water Movement
A SWAP model is a vertical one-dimensional model which uses Richards' equation to describe soil water movement: where θ is the volumetric water content (cm 3 cm −3 ), t is time (d), K(h) is the unsaturated hydraulic conductivity (cm d −1 ), h is the matric potential head (−cm), z is the vertical coordinate (cm), and S a (z,h) is the soil water extraction rate by plant roots (cm 3 cm −3 d −1 ). The SWAP model can function with a time-step varying from 10 −6 to 0.2 days to obtain calculation convergence.
To describe unsaturated hydraulic conductivity K(h), the SWAP model uses the Mualem-van Genuchten model [9,10]. The analytical θ(h) is described by the following equation: where θ sat is the saturated water content (cm 3 cm −3 ), θ res is the residual water content (cm 3 cm −3 ), α (cm −1 ), n, and m (−) are empirical shape factors; and m is 1−1/n. K(h) is calculated using the following equations: where K sat is the saturated hydraulic conductivity (cm d −1 ) and λ is a shape parameter (=0.5).

Boundary Conditions
The Penman-Monteith direct method was applied as a top boundary [11,12]: e sat − e a r a, can ∆ v + γ a 1 + r s, min r a, can LAI e f f , e sat − e a r a, soil ∆ v + γ a 1 + r soil r a, soil , (6) where T p is the potential transpiration of a dry canopy (mm d −1 ), W frac is the fraction of the day in which the canopy is wet (−), V c is the vegetation cover (−), ∆ v is the slope of the vapor pressure curve (kPa C −1 ), L w is the latent heat of vaporization (J kg −1 ), R n is the net radiation flux at the canopy surface (J m −2 d −1 ), G is the soil heat flux (J m −2 d −1 ), p 1 accounts for unit conversion (=86,400 s d −1 ), ρ a is the air density (kg m −3 ), C a is the heat capacity of moist air (J kg −1 C −1 ), e sat is the saturation vapor pressure (kPa), e a is the actual vapor pressure (kPa), r a,can is the aerodynamic resistance of the crop (s m −1 ), γ a is the psychrometric constant (kPa C −1 ), r s,min is the minimal stomatal resistance (s m −1 ), LAI eff is the effective leaf area index (−), E p is the potential evaporation (mm d −1 ), r soil is the soil resistance of wet soil (s m −1 ), and r a,soil is the aerodynamic resistance of the soil surface (s m −1 ). On a daily basis, the soil heat flux G is assumed to be negligible. The vegetation cover V c is obtained by the following equation: where κ dir and κ dif are the extinction coefficients for direct (0.80) and diffuse (0.72) solar light, respectively, and LAI is the actual leaf area index (−). LAI eff is derived from LAI: The aerodynamic resistance of the crop r a,can is determined using the resistance for uniform crop r a,can,0 : where z m is the height of wind speed measurements (m), z h is the height of temperature and humidity measurements (m), d is the zero plane displacement of wind profile (m), z om is the roughness parameter for momentum (m), z oh is the roughness parameter for heat and vapor (m), κ vk is the von Karman constant (0.41), and u is the wind speed measurement at z m (m s −1 ). The parameters d, z om , and z oh are defined as: where h crop is the crop height (cm). The SWAP model has several options for its bottom boundary condition, such as a prescribed groundwater level, a prescribed bottom flux, a bottom flux of zero, or a free drainage of soil profile. In this study, we decided to use a prescribed groundwater level, because relatively high groundwater levels were present during the field observations.

Crop Water Stress
The maximum possible root water extraction rate per day at a certain depth S p (z) is calculated by the following equation using the potential transpiration T p (cm d −1 ): where l root (z) is the root length density at a certain depth (cm cm −3 ) and D root is the root layer thickness (cm). The root water extraction rate is reduced due to suboptimal soil conditions, namely excessively wet and dry conditions; the soil water extraction rate by plant roots S a (z,h) is calculated considering the water stress described by the function proposed by Feddes et al. [13]: where α rw is the reduction factor for excessive wet and dry conditions (−). The reduction factor α rw changes from 0 to 1.0 depending on the critical pressure heads h 1 , h 2 , h 3 , and h 4 ( Figure 1). Root water extraction is optimal between h 2 and h 3 ; the reduction under wet conditions occurs when wetter than h 2 ; under dry conditions, the reduction occurs when drier than h 3 . The value of h 3 depends on the water demand of the atmosphere and varies with T p ; it is expressed by following equations: where h 3l and h 3h are the critical pressure heads for low and high transpiration rates, T low and T high , respectively (−cm). We used the default values for T low and T high : 0.1 and 0.5 cm d −1 . The actual transpiration rate T a (cm d −1 ) is yielded by integrating S a (z,h) over the root layer. Drought and wet stresses are evaluated by subtracting T a from T p . where αrw is the reduction factor for excessive wet and dry conditions (−). The reduction factor αrw changes from 0 to 1.0 depending on the critical pressure heads h1, h2, h3, and h4 ( Figure 1). Root water extraction is optimal between h2 and h3; the reduction under wet conditions occurs when wetter than h2; under dry conditions, the reduction occurs when drier than h3. The value of h3 depends on the water demand of the atmosphere and varies with Tp; it is expressed by following equations: where h3l and h3h are the critical pressure heads for low and high transpiration rates, Tlow and Thigh, respectively (−cm). We used the default values for Tlow and Thigh: 0.1 and 0.5 cm d −1 .
The actual transpiration rate Ta (cm d −1 ) is yielded by integrating Sa(z,h) over the root layer. Drought and wet stresses are evaluated by subtracting Ta from Tp. Figure 1. The reduction factor for excessive wet and dry conditions, αrw, as a function of matric potential head; h1, h2, h3, and h4 are critical pressure heads (−cm); h3 is between h3h and h3l values for high and low transpiration rates, respectively; h3 depends on the water demand of the atmosphere (Tlow and Thigh) and varies with the potential transpiration rate Tp. Wet stress occurs when wetter than h2, whereas drought stress occurs when drier than h3.

Field Experiment
Field observations were conducted in 2018 and 2019 at a converted field (100 × 43 m) in Okayama, Japan (37°34ʹ53.4ʺ N; 113°54ʹ11.8ʺ E). The soil in the field was lowland paddy soil according to the Japan soil inventory, NARO (http://soil-inventory.dc.affrc.go.jp/) [14]. The field was divided into two plots for other research purposes; for our purposes, we defined these as calibration and validation plots (Figures 2 and 3). Rotary tillage was carried out to a depth of 12 cm in both plots before sowing maize for whole crop silage. The reduction factor for excessive wet and dry conditions, α rw , as a function of matric potential head; h 1 , h 2 , h 3 , and h 4 are critical pressure heads (−cm); h 3 is between h 3h and h 3l values for high and low transpiration rates, respectively; h 3 depends on the water demand of the atmosphere (T low and T high ) and varies with the potential transpiration rate T p . Wet stress occurs when wetter than h 2 , whereas drought stress occurs when drier than h 3 . Japan soil inventory, NARO (http://soil-inventory.dc.affrc.go.jp/) [14]. The field was divided into two plots for other research purposes; for our purposes, we defined these as calibration and validation plots (Figures 2 and 3). Rotary tillage was carried out to a depth of 12 cm in both plots before sowing maize for whole crop silage. Figure 1. The reduction factor for excessive wet and dry conditions, αrw, as a function of matric potential head; h1, h2, h3, and h4 are critical pressure heads (−cm); h3 is between h3h and h3l values for high and low transpiration rates, respectively; h3 depends on the water demand of the atmosphere (Tlow and Thigh) and varies with the potential transpiration rate Tp. Wet stress occurs when wetter than h2, whereas drought stress occurs when drier than h3.

Field Experiment
Field observations were conducted in 2018 and 2019 at a converted field (100 × 43 m) in Okayama, Japan (37°34ʹ53.4ʺ N; 113°54ʹ11.8ʺ E). The soil in the field was lowland paddy soil according to the Japan soil inventory, NARO (http://soil-inventory.dc.affrc.go.jp/) [14]. The field was divided into two plots for other research purposes; for our purposes, we defined these as calibration and validation plots (Figures 2 and 3). Rotary tillage was carried out to a depth of 12 cm in both plots before sowing maize for whole crop silage.    Figure  3). We could not measure matric potential in our environment (i.e., low budget, labor deficit for sensor maintenance); we measured volumetric water content using capacitive soil moisture sensors (EC-5; METER Group) because they measure value accurately and are inexpensive. The sensors were set up in three locations at depths of 10, 20, and 30 cm ( Figure 3). The daily change in volumetric water content for calibration was acquired by averaging the readings of the three soil moisture sensors at each depth. Volumetric water content and groundwater level were recorded every 30 min.
Prior to the field observations, we collected soil samples from depths of 10, 20, and 30 cm in the center of the field using 100 mL soil samplers (DIK-1801; Daiki Rika Kogyo Co., Kounosu, Japan) to obtain soil hydraulic data for the Mualem-van Genuchten model. Three samples were collected from each depth. Saturated hydraulic conductivity was measured by the falling head method [15]. The soil water retention curve was generated by the soil column (matric potential head range: −1 to −31.6 cm) [16] and pressure plate methods (matric potential head range: −100 to −16,000 cm) [17]. The soil hydraulic properties at each depth were determined by averaging the values obtained from the three samples.
The maize crop height was measured on 9 May, 23 May, 12 June, 25 June, and 25 July 2019. Crop height on each occasion was determined by averaging the values obtained from 10 samples. Rooting depth was measured on 17 July 2019, from a single sample, to avoid field destruction.

Field Experiment for Model Validation
Field data for model validation were collected in the validation plot during the 2018 maize growing season. In this plot, maize was sown on 2 August 2018, and observations were made between 10 August and 20 November 2018. Sensor installation and use were as for 2019, except that the soil moisture sensors were used only at a depth of 10 cm (Figure 3). The crop height was measured on 14 August and 5 November 2018. The rooting depth was not obtained during these observations.  Figure 3). We could not measure matric potential in our environment (i.e., low budget, labor deficit for sensor maintenance); we measured volumetric water content using capacitive soil moisture sensors (EC-5; METER Group) because they measure value accurately and are inexpensive. The sensors were set up in three locations at depths of 10, 20, and 30 cm ( Figure 3). The daily change in volumetric water content for calibration was acquired by averaging the readings of the three soil moisture sensors at each depth. Volumetric water content and groundwater level were recorded every 30 min.
Prior to the field observations, we collected soil samples from depths of 10, 20, and 30 cm in the center of the field using 100 mL soil samplers (DIK-1801; Daiki Rika Kogyo Co., Kounosu, Japan) to obtain soil hydraulic data for the Mualem-van Genuchten model. Three samples were collected from each depth. Saturated hydraulic conductivity was measured by the falling head method [15]. The soil water retention curve was generated by the soil column (matric potential head range: −1 to −31.6 cm) [16] and pressure plate methods (matric potential head range: −100 to −16,000 cm) [17]. The soil hydraulic properties at each depth were determined by averaging the values obtained from the three samples.
The maize crop height was measured on 9 May, 23 May, 12 June, 25 June, and 25 July 2019. Crop height on each occasion was determined by averaging the values obtained from 10 samples. Rooting depth was measured on 17 July 2019, from a single sample, to avoid field destruction.

Field Experiment for Model Validation
Field data for model validation were collected in the validation plot during the 2018 maize growing season. In this plot, maize was sown on 2 August 2018, and observations were made between 10 August and 20 November 2018. Sensor installation and use were as for 2019, except that the soil moisture sensors were used only at a depth of 10 cm (Figure 3). The crop height was measured on 14 August and 5 November 2018. The rooting depth was not obtained during these observations.

Data for SWAP Model Simulation
The SWAP model requires various input parameters to conduct a simulation, as specified in Section 2.1 above. Table 1 shows the sources used for these data.

Initial and bottom boundary conditions
Matric potential head Converted from observed volumetric water content Groundwater level Measured

Data for calibration
Volumetric water content Measured

Model Calibration
The simulation period was 24 April to 17 July 2019; this is the same as the observation period for volumetric water content and groundwater level. The simulation domain was set at a depth of 100 cm. The domain was divided into three parts (layers 1, 2, and 3), based on soil sampling and soil moisture sensor depths. Initial Mualem-van Genuchten parameters from field observation were assigned to each layer. The nodal spacing increased from 0.2 cm near the soil surface to 5.0 cm with depth ( Table 2). The initial matric potential head was obtained by transforming the initial volumetric water content, and was applied accounting for depth; below a depth of 30 cm, the lowest measurement depth, the initial matric potential head was determined based on the initial groundwater level. Table 3 shows the crop data used in the simulation. We assumed that crop height and rooting depth were zero at the beginning of the simulation because the observed initial maize height was negligible. The rooting depth was assumed to reach its maximum depth by the end of simulation, and to increase linearly [8]. The rate of increase in crop height was based on observed values; the maximum crop height at the end of the simulation was interpolated using the observed crop heights on 25 June and 25 July 2019. We applied the default critical pressure heads for maize

. Model Calibration
The simulation period was 24 April to 17 July 2019; this is the same as the observation period for volumetric water content and groundwater level. The simulation domain was set at a depth of 100 cm. The domain was divided into three parts (layers 1, 2, and 3), based on soil sampling and soil moisture sensor depths. Initial Mualem-van Genuchten parameters from field observation were assigned to each layer. The nodal spacing increased from 0.2 cm near the soil surface to 5.0 cm with depth ( Table 2). The initial matric potential head was obtained by transforming the initial volumetric water content, and was applied accounting for depth; below a depth of 30 cm, the lowest measurement depth, the initial matric potential head was determined based on the initial groundwater level. Table 3 shows the crop data used in the simulation. We assumed that crop height and rooting depth were zero at the beginning of the simulation because the observed initial maize height was negligible. The rooting depth was assumed to reach its maximum depth by the end of simulation, and to increase linearly [8]. The rate of increase in crop height was based on observed values; the maximum crop height at the end of the simulation was interpolated using the observed crop heights on 25 June and 25 July 2019. We applied the default critical pressure heads for maize from Wesseling et al. [20] because it is difficult to measure. A trial and error process [21][22][23]

Model Calibration
The simulation period was 24 April to 17 July 2019; this is the same as the observation period for volumetric water content and groundwater level. The simulation domain was set at a depth of 100 cm. The domain was divided into three parts (layers 1, 2, and 3), based on soil sampling and soil moisture sensor depths. Initial Mualem-van Genuchten parameters from field observation were assigned to each layer. The nodal spacing increased from 0.2 cm near the soil surface to 5.0 cm with depth ( Table 2). The initial matric potential head was obtained by transforming the initial volumetric water content, and was applied accounting for depth; below a depth of 30 cm, the lowest measurement depth, the initial matric potential head was determined based on the initial groundwater level. Table 3 shows the crop data used in the simulation. We assumed that crop height and rooting depth were zero at the beginning of the simulation because the observed initial maize height was negligible. The rooting depth was assumed to reach its maximum depth by the end of simulation, and to increase linearly [8]. The rate of increase in crop height was based on observed values; the maximum crop height at the end of the simulation was interpolated using the observed crop heights on 25 June and 25 July 2019. We applied the default critical pressure heads for maize from Wesseling et al. [20] because it is difficult to measure. A trial and error process [21][22][23] was used to optimize soil hydraulic parameters.

Model Validation
Model validation was performed using the optimized soil hydraulic parameters from the calibration and observation data from 2018. The simulation period was from 10 August to 20 November 2018, to match the observation period. Calibration values were used for soil layer assignment and nodal spacing. We assumed the same maximum rooting depth, 50 cm, as for the calibration plot (Table 3). Linear interpolation, from 0 cm on the sowing date of 2 August 2018 to the maximum model depth at the end of the simulation, provided the estimate of initial rooting depth on 10 August 2018. We assumed a sowing-date crop height of 0 cm, changing in accordance with the measured height (Table 3); the initial crop height was estimated using this minimum value and the observed height on 14 August 2018. After 5 November 2018, we assumed that the height did not increase.

Evaluation of Calculation Error
During calibration and validation, we evaluated the agreement at each time point between the observed (θ obs ) and calculated (θ cal ) volumetric water content using the root mean square error (RMSE): where N is the number of time points, and obs,t and cal,t are the observed and calculated data at each time point. Moreover, we calculated RMSE between the observed and calculated matric potential head (h obs and h cal ) in the calibration plot to consider possible error in water stress estimation; the matric potential head was converted from θ obs and θ cal by using calibrated Mualem-van Genuchten parameters.

Evaluation of Water Stress under Actual and Scenario Conditions
We estimated the daily changes in T p , T a , drought stress, and wet stress using the 2019 observations as the actual conditions. We then evaluated the difference in water stress between the actual and a scenario in which tillage reached 11 cm deeper than under actual conditions; the soil hydraulic properties of layer 2 were therefore the same as those of layer 1. The tillage depth in the scenario conditions assumed to exceed a depth where hardpans usually exist.

Model Calibration
During 2019, climatic conditions were typical: the weather became drier after sowing, followed by the rainy season (Figure 4a). During the study period, precipitation was highest, at 51.7 mm d −1 , on 15 June 2019. Thereafter, precipitation occurred more frequently. The highest groundwater level, −14.6 cm, was recorded on 2 May 2019, after which it gradually decreased to its lowest value, approximately −85 cm; we note that this was the limit of observation for the water level monitoring sensor. The groundwater level changed frequently after 15 June 2019, and high groundwater levels of approximately −20 cm appeared on 16 June and 15 July 2019.
The model reproduced the volumetric water content well using the optimized Mualem-van Genuchten parameters, whereas it tended to under-or overestimate this value when using the initial parameters from the soil samples. In Figure 4, panels b-d show θ obs , θ cal , and the volumetric water content with initial parameter θ ini at depths of 10, 20, and 30 cm, respectively. The initial parameters and soil dry bulk density are shown in Table 4; Table 5 shows calibrated parameters. At a depth of 10 cm, θ ini was higher than θ obs and θ cal during the simulation period, except between 2 and 12 June 2019, when precipitation was less frequent. The values of θ obs and θ cal at a depth of 10 cm varied between 0.3 and 0.5 cm 3 cm −3 , and the change in these two values was more frequent than in their equivalents at the other two depths. At a depth of 10 cm, θ obs and θ cal increased with rising precipitation (Figure 4a,b). Before 14 June 2019, θ obs and θ cal changed moderately and had three peaks, namely, 30 April, 20 May, and 7 June 2019, when precipitation was more than 20 mm d −1 . After 15 June 2019, when precipitation occurred frequently, θ obs and θ cal increased, and θ cal changed more frequently than θ obs . At a depth of 20 cm, θ ini was higher than θ obs and θ cal throughout the simulation period: θ obs and θ cal were between 0.4 and 0.5 cm 3 cm −3 . At a depth of 30 cm, θ obs was between 0.4 and 0.6 cm 3 cm −3 , whereas θ cal was around 0.5 cm 3 cm −3 : θ ini was almost the same as θ cal . Between 27 May and 14 June 2019, θ obs at a depth of 30 cm gradually decreased; in contrast, θ cal changed minimally. During this period, the lowest groundwater level was applied in the simulation. However, the actual ground water level at this time might have exceeded the sensor's measurement limit, causing the difference between θ obs and θ cal . properties of layer 2 were therefore the same as those of layer 1. The tillage depth in the scenario conditions assumed to exceed a depth where hardpans usually exist.

Model Calibration
During 2019, climatic conditions were typical: the weather became drier after sowing, followed by the rainy season (Figure 4a). During the study period, precipitation was highest, at 51.7 mm d −1 , on 15 June 2019. Thereafter, precipitation occurred more frequently. The highest groundwater level, −14.6 cm, was recorded on 2 May 2019, after which it gradually decreased to its lowest value, approximately −85 cm; we note that this was the limit of observation for the water level monitoring sensor. The groundwater level changed frequently after 15 June 2019, and high groundwater levels of approximately −20 cm appeared on 16 June and 15 July 2019.
The model reproduced the volumetric water content well using the optimized Mualem-van Genuchten parameters, whereas it tended to under-or overestimate this value when using the initial parameters from the soil samples. In Figure 4, panels b-d show θobs, θcal, and the volumetric water content with initial parameter θini at depths of 10, 20, and 30 cm, respectively. The initial parameters and soil dry bulk density are shown in Table 4; Table 5 shows calibrated parameters. At a depth of 10 cm, θini was higher than θobs and θcal during the simulation period, except between 2 and 12 June 2019, when precipitation was less frequent. The values of θobs and θcal at a depth of 10 cm varied between 0.3 and 0.5 cm 3 cm −3 , and the change in these two values was more frequent than in their equivalents at the other two depths. At a depth of 10 cm, θobs and θcal increased with rising precipitation (Figure 4a,b). Before 14 June 2019, θobs and θcal changed moderately and had three peaks, namely, 30 April, 20 May, and 7 June 2019, when precipitation was more than 20 mm d −1 . After 15 June 2019, when precipitation occurred frequently, θobs and θcal increased, and θcal changed more frequently than θobs. At a depth of 20 cm, θini was higher than θobs and θcal throughout the simulation period: θobs and θcal were between 0.4 and 0.5 cm 3 cm −3 . At a depth of 30 cm, θobs was between 0.4 and 0.6 cm 3 cm −3 , whereas θcal was around 0.5 cm 3 cm −3 : θini was almost the same as θcal. Between 27 May and 14 June 2019, θobs at a depth of 30 cm gradually decreased; in contrast, θcal changed minimally. During this period, the lowest groundwater level was applied in the simulation. However, the actual ground water level at this time might have exceeded the sensor's measurement limit, causing the difference between θobs and θcal.   θ res : residual water content; θ sat : saturated water content; α and n: empirical shape factors; K sat : saturated hydraulic conductivity; λ: a shape parameter. The RMSEs between θ obs and θ cal at depths of 10, 20, and 30 cm were 0.030, 0.012, and 0.027 cm 3 cm −3 , respectively. Crescimanno and Garofalo [24], also using a SWAP model, reported that their largest RMSE (0.037 cm 3 cm −3 ) related to volumetric water content was low enough to indicate good model accuracy. Similarly, our largest RMSE (0.030 cm 3 cm −3 ) was sufficiently low for this purpose.
The initial and calibrated soil water retention curves were almost the same for layer 3; they differed for layers 1 and 2 ( Figure 5). We sampled the soil after conducting rotary tillage, but calibration was conducted for the observation period. We consider this to be the cause of the differences between the soil water retention curves of the relevant layers. The effect of tillage diminishes with time and as a consequence of the wet/dry cycle [25,26]. Or et al. [26] indicated that the wet/dry cycle reduced total porosity (reflected in the saturated water content) and structural porosity, whereas textural porosity remained constant. The calibrated soil water retention curve of layer 1 may reflect those changes. For layer 2, the calibrated soil water retention curve shows reduced saturated water content, i.e., reduced total porosity. As the soil in the field was clayey and precipitation occurred frequently, clay eluviation might occur in this layer [27]. We consider the calibrated soil water retention curves to be reasonable for use in the simulation.   The RMSEs between θobs and θcal at depths of 10, 20, and 30 cm were 0.030, 0.012, and 0.027 cm 3 cm −3 , respectively. Crescimanno and Garofalo [24], also using a SWAP model, reported that their largest RMSE (0.037 cm 3 cm −3 ) related to volumetric water content was low enough to indicate good model accuracy. Similarly, our largest RMSE (0.030 cm 3 cm −3 ) was sufficiently low for this purpose.
The initial and calibrated soil water retention curves were almost the same for layer 3; they differed for layers 1 and 2 ( Figure 5). We sampled the soil after conducting rotary tillage, but calibration was conducted for the observation period. We consider this to be the cause of the differences between the soil water retention curves of the relevant layers. The effect of tillage diminishes with time and as a consequence of the wet/dry cycle [25,26]. Or et al. [26] indicated that the wet/dry cycle reduced total porosity (reflected in the saturated water content) and structural porosity, whereas textural porosity remained constant. The calibrated soil water retention curve of layer 1 may reflect those changes. For layer 2, the calibrated soil water retention curve shows reduced saturated water content, i.e., reduced total porosity. As the soil in the field was clayey and precipitation occurred frequently, clay eluviation might occur in this layer [27]. We consider the calibrated soil water retention curves to be reasonable for use in the simulation. Figure 5. The analytical soil water retention curves generated by initial and calibrated Mualem-van Genuchten parameters. (a-c) Results for layers 1, 2, and 3, respectively. The grey area represents the zero-water-stress zone; drought or wet stresses occur outside this area. In the scenario condition, layers 1 and 2 used the calibrated soil water retention curve of layer 1. Figure 5. The analytical soil water retention curves generated by initial and calibrated Mualem-van Genuchten parameters. (a-c) Results for layers 1, 2, and 3, respectively. The grey area represents the zero-water-stress zone; drought or wet stresses occur outside this area. In the scenario condition, layers 1 and 2 used the calibrated soil water retention curve of layer 1.

Model Validation
The SWAP model with optimized parameters was able to simulate the volumetric water content under significantly different weather conditions. Figure 6a shows the groundwater level and precipitation in the validation plot in 2018; precipitation intensity and groundwater level were higher in 2018 than in 2019 (Figure 4a). Before 29 September 2018, the groundwater level was relatively high, and precipitation occurred frequently. The maximum daily precipitation (82.7 mm d −1 ) occurred on 29 September 2018. Thereafter, the groundwater level gradually decreased to its lowest value (−90 cm), and precipitation became less frequent. Figure 6b shows the temporal changes in θ obs and θ cal at a depth of 10 cm: the trends in θ cal and θ obs were matched; both values changed with the precipitation. The RMSE between θ cal and θ obs was 0.029 cm 3 cm −3 . This result indicates that, even with high levels of precipitation and groundwater, the model was able to reproduce actual volumetric water content using our calibrated parameters.

Model Validation
The SWAP model with optimized parameters was able to simulate the volumetric water content under significantly different weather conditions. Figure 6a shows the groundwater level and precipitation in the validation plot in 2018; precipitation intensity and groundwater level were higher in 2018 than in 2019 (Figure 4a). Before 29 September 2018, the groundwater level was relatively high, and precipitation occurred frequently. The maximum daily precipitation (82.7 mm d −1 ) occurred on 29 September 2018. Thereafter, the groundwater level gradually decreased to its lowest value (−90 cm), and precipitation became less frequent. Figure 6b shows the temporal changes in θobs and θcal at a depth of 10 cm: the trends in θcal and θobs were matched; both values changed with the precipitation. The RMSE between θcal and θobs was 0.029 cm 3 cm −3 . This result indicates that, even with high levels of precipitation and groundwater, the model was able to reproduce actual volumetric water content using our calibrated parameters.

Change in Matric Potential in the Calibration Plot
Calculated matric potential head (hcal) in the calibration plot captured the trend of change in observed value (hobs) except after 15 June 2019, when precipitation was more frequent (Figures 4a and  7). We excluded positive pressure head in Figure 6. Before 15 June 2019, hcal at a depth of 10 cm changed similarly to hobs except between 30 April and 19 May 2019. In this period, hcal was almost 10 times higher than hobs; this reflected lower θcal than θobs in the same period (Figures 4b and 7a). At depths of 20 and 30 cm, the changes in hobs and hcal were similar. After 27 May 2019, hobs at a depth of 30 cm became larger than hcal due to the decrease in θobs in this period (Figures 4d and 7c). Except at a depth of 30 cm, hcal differed from hobs after 15 June 2019. At a depth of 10 cm, hcal tended to be smaller than hobs; hcal was larger than hobs at a depth of 20 cm. The RMSEs between hobs and hcal at depths of 10, 20, and 30 cm were 1.0, 0.7, and 0.6 log10(−cm), respectively. These results indicate that the calculated water stress after 15 June 2019 might contain larger error than that before the day.

Change in Matric Potential in the Calibration Plot
Calculated matric potential head (h cal ) in the calibration plot captured the trend of change in observed value (h obs ) except after 15 June 2019, when precipitation was more frequent (Figures 4a  and 7). We excluded positive pressure head in Figure 6. Before 15 June 2019, h cal at a depth of 10 cm changed similarly to h obs except between 30 April and 19 May 2019. In this period, h cal was almost 10 times higher than h obs ; this reflected lower θ cal than θ obs in the same period (Figures 4b and 7a). At depths of 20 and 30 cm, the changes in h obs and h cal were similar. After 27 May 2019, h obs at a depth of 30 cm became larger than h cal due to the decrease in θ obs in this period (Figures 4d and 7c). Except at a depth of 30 cm, h cal differed from h obs after 15 June 2019. At a depth of 10 cm, h cal tended to be smaller than h obs ; h cal was larger than h obs at a depth of 20 cm. The RMSEs between h obs and h cal at depths of 10, 20, and 30 cm were 1.0, 0.7, and 0.6 log 10 (−cm), respectively. These results indicate that the calculated water stress after 15 June 2019 might contain larger error than that before the day.  (Table 5). We excluded data that showed positive values.

Water Stress under Actual Conditions
Our simulation revealed that maize cultivation under actual conditions would be highly likely to suffer from water stress. Figure 8a shows Tp and Ta in the calibration plot in 2019 as estimated by the SWAP model. Ta was significantly lower than Tp during most of the simulation period, reaching an approximate extreme of only 20% of Tp. From 28 May to 15 June 2019, less precipitation occurred, and the groundwater level was low (Figure 4a). Thus, the difference between Tp and Ta might have been induced by dry conditions. After 15 June 2019, groundwater level and precipitation increased; Ta during this period may have decreased due to the wet conditions. However, several zeroprecipitation days (namely, 17-20 June, 23-25 June, and 4-8 July 2019) occurred during this period. Although the comparison between Tp and Ta indicates that water stress occurred, we could not distinguish whether it was drought or wet stress that affected Ta.
Separate evaluations of drought and wet stresses are likely to be necessary when determining appropriate adaptation measures. Figure 8b shows the drought and wet stresses separately, highlighting the critical growth periods for drought and wet stresses [3,4]. As shown in Figures 4a  and 8b, little rain fell until 15 June 2019, and drought stress increased during this period. The maximum drought stress level occurred on 13 June 2019. Wet stress predominated after 15 June 2019, when precipitation was more frequent. Moreover, the highest groundwater levels appeared on 16 June, 12 July, and 15 July 2019, which could have induced wet stress. The two highest levels of wet stress occurred on 16 June and 16 July 2019. Figure 8b indicates that under the actual conditions, drought and wet stresses appeared between 3 and 13 June 2019 (40-50 days after emergence) and between 23 June and 13 July 2019 (60-80 days after emergence), respectively. This implies that the actual conditions were disadvantageous for maize growth in terms of water stress. We note that the second largest peak of wet stress on 16 June 2019 may affect crop growth adversely even the stress occurred in the critical period for drought stress (Figure 8b).  (Table 5). We excluded data that showed positive values.

Water Stress under Actual Conditions
Our simulation revealed that maize cultivation under actual conditions would be highly likely to suffer from water stress. Figure 8a shows T p and T a in the calibration plot in 2019 as estimated by the SWAP model. T a was significantly lower than T p during most of the simulation period, reaching an approximate extreme of only 20% of T p . From 28 May to 15 June 2019, less precipitation occurred, and the groundwater level was low (Figure 4a). Thus, the difference between T p and T a might have been induced by dry conditions. After 15 June 2019, groundwater level and precipitation increased; T a during this period may have decreased due to the wet conditions. However, several zero-precipitation days (namely, 17-20 June, 23-25 June, and 4-8 July 2019) occurred during this period. Although the comparison between T p and T a indicates that water stress occurred, we could not distinguish whether it was drought or wet stress that affected T a .
Separate evaluations of drought and wet stresses are likely to be necessary when determining appropriate adaptation measures. Figure 8b shows the drought and wet stresses separately, highlighting the critical growth periods for drought and wet stresses [3,4]. As shown in Figures 4a and 8b, little rain fell until 15 June 2019, and drought stress increased during this period. The maximum drought stress level occurred on 13 June 2019. Wet stress predominated after 15 June 2019, when precipitation was more frequent. Moreover, the highest groundwater levels appeared on 16 June, 12 July, and 15 July 2019, which could have induced wet stress. The two highest levels of wet stress occurred on 16 June and 16 July 2019. Figure 8b indicates that under the actual conditions, drought and wet stresses appeared between 3 and 13 June 2019 (40-50 days after emergence) and between 23 June and 13 July 2019 (60-80 days after emergence), respectively. This implies that the actual conditions were disadvantageous for maize growth in terms of water stress. We note that the second largest peak of wet stress on 16 June 2019 may affect crop growth adversely even the stress occurred in the critical period for drought stress (Figure 8b).  Figure 9, panels a and b, show the drought and wet stresses for the actual and scenario conditions, respectively. In the scenario, drought stress was lower than under actual conditions for most of the simulation period. From 3 to 15 June 2019, when drought stress under actual conditions was relatively high in relation to the relevant critical growth period (defined as the drought-dominant period), the stress under the scenario conditions was 33%-75% of that under actual conditions ( Figure  9a). Wet stress under the scenario conditions was also lower than, or the same as, that under actual conditions. In the growth period during which wet stress was critical, stress under scenario conditions was 10%-82% of that under actual conditions (Figure 9b). These results suggest that making the tillage depth 11 cm deeper tends to reduce both forms of stress during the maize critical growth periods.

Illustrative Example: Reduction of Water Stress by Changing Tillage Depth
The water stress at depths of 10 and 20 cm differed between the actual and scenario conditions ( Table 6). Throughout the drought-dominant period, drought stress occurred at a depth of 10 cm under both conditions. At a depth of 20 cm, however, although the average volumetric water content under scenario conditions was lower than under actual conditions, it was in the no-stress range, as indicated by the soil water retention curve (Figure 5a). During the wet stress critical period, stress tended to occur more often at a depth of 10 cm than at 20 cm (Table 6). Wet stress occurred on fewer days under scenario than actual conditions at both depths. These results indicate that the scenario conditions probably mitigate water stress at depths of 10 and 20 cm. Moreover, water stress varied with depth; by evaluating drought and wet stresses at different depths, future studies could identify the critical layer or depth for crop growth under actual and scenario conditions.  Figure 9, panels a and b, show the drought and wet stresses for the actual and scenario conditions, respectively. In the scenario, drought stress was lower than under actual conditions for most of the simulation period. From 3 to 15 June 2019, when drought stress under actual conditions was relatively high in relation to the relevant critical growth period (defined as the drought-dominant period), the stress under the scenario conditions was 33%-75% of that under actual conditions (Figure 9a). Wet stress under the scenario conditions was also lower than, or the same as, that under actual conditions. In the growth period during which wet stress was critical, stress under scenario conditions was 10%-82% of that under actual conditions (Figure 9b). These results suggest that making the tillage depth 11 cm deeper tends to reduce both forms of stress during the maize critical growth periods.

Illustrative Example: Reduction of Water Stress by Changing Tillage Depth
The water stress at depths of 10 and 20 cm differed between the actual and scenario conditions ( Table 6). Throughout the drought-dominant period, drought stress occurred at a depth of 10 cm under both conditions. At a depth of 20 cm, however, although the average volumetric water content under scenario conditions was lower than under actual conditions, it was in the no-stress range, as indicated by the soil water retention curve (Figure 5a). During the wet stress critical period, stress tended to occur more often at a depth of 10 cm than at 20 cm (Table 6). Wet stress occurred on fewer days under scenario than actual conditions at both depths. These results indicate that the scenario conditions probably mitigate water stress at depths of 10 and 20 cm. Moreover, water stress varied with depth; by evaluating drought and wet stresses at different depths, future studies could identify the critical layer or depth for crop growth under actual and scenario conditions.  The "drought-dominant period" is the period when drought stress was significant and the maize growth phase was drought-sensitive (3-28 June 2019). Stress occurrence days were calculated based on Figure 5 and volumetric water content under actual and scenario conditions (data not shown).

Conclusions
In this study, the drought and wet stresses in a converted maize field in Okayama, Japan, were evaluated using a SWAP model. The model was calibrated and validated using data from field observations. The water stress under actual and scenario (tillage depth 11 cm deeper) conditions were then estimated. Our study revealed that (1) drought and wet stresses appeared within the relevant critical growth stages of maize, under actual conditions; (2) water stress under the scenario condition tended to be lower than under actual conditions, with drought and wet stresses 33%-75% and 10%-82% of those under actual conditions, respectively; (3) water stress between soil depths of 10 and 20 cm was lower under the scenario conditions than under the actual conditions. Those results indicate that deeper tillage may mitigate both drought and wet stresses. We hypothesize that by evaluating drought and wet stresses at different depths, we can identify the critical layer or depth for crop growth; this will promote further reduction in water-related crop damage in converted fields. The method presented here can be applied to fields with suboptimal water conditions, and may enhance yield. Further, we believe that our study will support precise stress control management, and boost crop yield and quality, when combined with precision farming applications in converted fields. (b) wet stress. The critical growth periods for drought and wet stresses were defined based on Çakir [3] and Kanwar et al. [4]. The "drought-dominant period" is the period when drought stress was significant and the maize growth phase was drought-sensitive (3-28 June 2019). Stress occurrence days were calculated based on Figure 5 and volumetric water content under actual and scenario conditions (data not shown).

Conclusions
In this study, the drought and wet stresses in a converted maize field in Okayama, Japan, were evaluated using a SWAP model. The model was calibrated and validated using data from field observations. The water stress under actual and scenario (tillage depth 11 cm deeper) conditions were then estimated. Our study revealed that (1) drought and wet stresses appeared within the relevant critical growth stages of maize, under actual conditions; (2) water stress under the scenario condition tended to be lower than under actual conditions, with drought and wet stresses 33%-75% and 10%-82% of those under actual conditions, respectively; (3) water stress between soil depths of 10 and 20 cm was lower under the scenario conditions than under the actual conditions. Those results indicate that deeper tillage may mitigate both drought and wet stresses. We hypothesize that by evaluating drought and wet stresses at different depths, we can identify the critical layer or depth for crop growth; this will promote further reduction in water-related crop damage in converted fields. The method presented here can be applied to fields with suboptimal water conditions, and may enhance yield. Further, we believe that our study will support precise stress control management, and boost crop yield and quality, when combined with precision farming applications in converted fields. Funding: This research was supported by grants from the Project of the Bio-oriented Technology Research Advancement Institution, NARO (the special scheme project on vitalizing the management entities of agriculture, forestry and fisheries).