Model-Based Optimization of Design Parameters of Subsurface Drain in Cotton Field under Mulch Drip Irrigation

Abstract

In this study, the influence of the relevant parameters of a subsurface pipe layout on soil water and salt transport in a cotton field under mulched drip irrigation is explored. Based on the measured data of the changes in the groundwater level and salt in the field, the DRAINMOD numerical model has been used for simulating the water and salt dynamics of a salinized cotton field under subsurface pipe drainage. The results of the investigation show that the DRAINMOD model can accurately simulate the changes in the hydrological conditions and the salt-leaching process in the study area. The average deviation between the simulated and measured values of the groundwater depth in 2013 and 2014 was −1.72 cm and 2.43 cm, the average absolute deviation was 3.84 cm and 2.43 cm, the root mean square error was 5.14 cm and 3.63 cm, and the correlation coefficient was 0.87 and 0.94, respectively. The average deviation between the simulated and measured values of soil salinity in 2013 and 2014 was −0.68 g/kg and −1.86 g/kg, the average absolute deviation was 1.60 g/kg and 1.99 g/kg, the root mean square error was 1.95 g/kg and 2.99 g/kg, and the correlation coefficient was 0.82 and 0.86, respectively, which are all within the acceptable error range. After validation, the model was used to simulate and analyze the desalination process of a cotton field in the study area for 27 different subsurface pipe layout modes. The projection pursuit classification model has been combined with the accelerated genetic algorithm based on real-number coding. The comprehensive benefits of the subsurface pipe layout were evaluated using the construction cost, average desalination rate, and relative yield of cotton as the evaluation indices. The results show that C₁₁ (buried depth 2.1 m, spacing 30 m) is the optimal layout of the subsurface pipe. The results of this study can provide theoretical support and scientific guidance for the popularization and application of subsurface pipe salt discharge technology and drip irrigation under film in the arid inland areas of northwest China.

1. Introduction

Soil salinization is a global environmental problem [1,2]. The formation of salinized soil is affected by various factors such as climate, topography, and tillage methods, which can directly or indirectly affect the absorption of soil nutrients by crops, reduce the utilization efficiency of nutrients in the soil, deteriorate the physical, chemical, and biological properties of the soil, and cause adverse effects on crop growth and development [3,4,5]. Xinjiang, located in the northwest region of China, has a wide distribution area and different types of salt compositions owing to the high salt content in the soil, strong surface evaporation, and low precipitation. It is one of the most concentrated areas of saline soil distribution in China and one of the main factors restricting the sustainable development of agriculture and resource utilization in Xinjiang [6,7].

Soil improvement [8], subsoiling soil [9], straw mulching [10], burying [11], paddy-upland rotation [12], chemical modifiers [13,14,15], and other measures can reduce the damage caused by the salt content in the soil by varying degrees. Planting Tamarix Chinensis, Puccinellia Tenuiflora, Suaeda Glauca, etc., to improve the saline-alkali wastelands has obvious soil improvement benefits [16,17,18,19,20]. Ahmad et al. [21] found that soil salinity can be effectively reduced by cultivating salt-tolerant plants. Roberto et al. [22] found that the salt content and pH value of 0–30 cm soil decreased significantly under the experimental treatment conditions of subsoiler combined with planting wheat straw. Bajpai et al. [23] found in a field experiment study that the yield, Nuptake rate, and Nuptake of Egyptian clover inoculated with relatively salt-tolerant rhizobium strain KB3 under the improvement of soil gypsum were significantly higher than those without inoculation. Although there are numerous methods available for improving the soil from saline-alkali land, salt is not separated from the soil in these methods. If agricultural irrigation measures are unreasonable, secondary salinization of soil will occur. Pipe drainage salt implies following the ‘salt with water, salt with water’ water movement law. It fully dissolves the soil salt and penetrates the underground water through pipeline discharge, thus effectively reducing the salt content of the soil, controlling the groundwater level and improving the physical and chemical properties of the soil [24]. Azhar et al. [25] revealed that the installation of underground drainage systems could effectively control the salinity levels of irrigated land in the study area. Koji et al. [26] found that the simultaneous action of irrigation and subsurface drainage significantly increased the rate of soil desalination. I.Ben Aissa et al. [27] set up a subsurface pipe drainage facility. After two years of experimental research, the test results showed that the desalination could be 30% in the first year and 15% in the second year, which proved that subsurface pipe drainage could effectively reduce soil salt content. Liu [28], Yang [29], and Shi et al. [30] carried out salt drainage field experiments with different subsurface pipe spacings or buried depths in Xinjiang. The results from these studies showed that using a subsurface pipe could reduce soil salinity and increase crop yield. Fully understanding the soil water and salt dynamics in the subsurface drainage process is the basis and key to formulating a subsurface drainage and salt discharge system. Numerical simulation has been proved to be one of the most convenient and reliable means for obtaining this information, as done in various research studies. Li [31] and Shi [32] used the HYDRUS software to verify the water and salt movement parameters of subsurface drainage. The results of these studies showed that the simulated values were in good agreement with the measured values and could describe the soil water and salt dynamics in the process of subsurface drainage and salt discharge. The DRAINMOD model was developed by Skaggs of the Department of Biological and Agricultural Engineering at North Carolina State University Inter-water balance model [33]. With continuous promotion and changes in demand, the function of the DRAINMOD model has been further expanded and improved. Luo et al. [34] modified the DRAINMOD under limited conditions; Kandil [35], Breve [36], Moursi [37], and Youssef et al. [38] extended the DRAINMOD model and established DRAINMOD-S, DRAINMOD-N, and DRAINMOD-NII models. The applicability was extended from humid areas to arid and semi-arid areas, and the large slope and freeze–thaw conditions in winter were corrected. Li [39] and Dou [40] used the DRAINMOD model to simulate the migration of water and salt, and their simulation results showed that this model, which simulates the hydrological changes in the soil and the process of salt leaching, exhibits high precision. In view of the successful application of the DRAINMOD software in a simulation of the subsurface pipe drainage and salt discharge process, a comprehensive evaluation of this software has been performed in this study for its application prospects in subsurface pipe drainage in a mulched drip irrigation cotton field in Xinjiang. In this study, based on the field leaching test (ground leaching pressure salt, subsurface pipe drainage, and salt discharge), DRAINMOD software has been used to simulate the soil water and salt transport process in the subsurface pipe drainage area under the condition of leaching pressure salt, and the numerical model of the subsurface pipe drainage and salt discharge and the related water and salt movement parameters have been verified. In addition, a preliminary exploration of the effect of the subsurface pipe drainage has been executed. This study lays a foundation for further analysis and comparison of the effect of salt discharge in different underground pipe drainage methods and for determining the optimal buried depth and spacing of underground pipes in the study area.

To determine the best combination of the subsurface pipe layout, it is necessary to analyze the desalination effect, the actual needs of the project, and the investment budget. The traditional methods are mostly based on the experience of experts, and the evaluation results are arbitrary, to some extent. The study by Zhang et al. [41] involves the multi-scheme selection of water-saving irrigation and is based on the projection pursuit classification (PPC) model test of the real-coded accelerated genetic algorithm (RAGA). Satisfactory results have been achieved by their model, and human disturbance of the weight matrix in the fuzzy comprehensive evaluation has been avoided to the greatest extent. Taking this result into consideration, the PPC model [42] based on RAGA [43] has been applied in this study to perform a comprehensive evaluation of the optimal subsurface pipe layout. The optimal projection direction of the state variable index in the PPC model was optimized by RAGA, and the projection values of each sample were obtained. These projection values were sorted according to their size to determine the optimal combination of the subsurface drainage layout. This study provides a theoretical basis for the improvement of saline-alkali soil and the restoration and reconstruction of abandoned farmlands in arid and semi-arid regions.

2. Materials and Methods

2.1. Details of the Experimental Site

The study area is located in the 122nd Regiment of the Eighth Agricultural Division of Xinjiang Production and Construction Corps (44°37′ N to 44°48′ N, 85°27′ E to 85°41′ E), the specific location is shown in Figure 1. It has a flat terrain and is located in the southern margin of the Junggar Basin in Xinjiang. The total area of arable land is 8.79 × 10 hm² (8.79 × 10 ha). The average annual rainfall in the test area is 141.8 mm, the average annual potential evaporation is 1826.2 mm, and the average annual sunshine hours are 2861.6 h. The soil in the test area is sandy loam, and the initial salt content of the soil is more than 10 g/kg.

2.2. Experimental Design and Data Acquisition

The study area was improved by subsurface drainage and irrigation leaching. The inner diameter of the subsurface pipe in the test area was 70 mm, the opening gap was less than 1 mm, the buried depth was 2.2 m, the spacing was 48 m, the slope was 3%, and the type of drip irrigation-supporting equipment used included a single-wing labyrinth drip irrigation belt with a dripper flow rate of 3.2 l/h. The planting crop in the study area was cotton, and the crop variety was Chuangza 100. The drip irrigation mode comprised one film, two pipes, and six rows, as shown in Figure 2. The details of the irrigation system are given in

Table 1. Irrigation details of the study area in the year 2013–2014.

Irrigation Date	Irrigation Quota (m3/hm2)
20 April	1800
1 May	600
1 June	600
21 June	600
30 June	600
11 July	600
21 July	600
31 July	600
11 August	600
21 August	600

. The standard crop coefficient for cotton recommended by the Food and Agriculture Organization of the United Nations was used as the crop coefficient. The experimental period of this study was two years, and the planting dates are given in

Table 2. Monthly groundwater levels for the irrigation season in the year 2013–2014.

Month	Groundwater Table Depth/cm
	2013	2014
4	210	205
5	185	188
6	175	175
7	175	177
8	178	179
9	185	182
10	190	192

. The sampling time was 25 May 2013 (seedling stage), 20 July 2013 (flowering stage), 30 September 2013 (boll opening stage), 20 May 2014 (seedling stage), 25 July 2014 (flowering stage), and 20 September 2014 (boll opening stage). Three observation points, T1, T2, and T3, were set at 8, 16, and 24 m, respectively, from the dark tube. Three sampling points were selected for each observation point: the center of the wide line under the film, the center of the narrow line under the film, and the center of the bare land outside the film (nine boreholes, each with a depth of approximately 2 m, 10 layers of sampling in 0–20, 20–40, 40–60, 60–80, 80–100, 100–120, 120–140, 140–160, 160–180, and 180–200 cm). The water and salt content of the soil samples at the same depth of the nine boreholes were averaged to represent the water and salt content between the two subsurface pipes.

After the test, an average of three groups of test results was taken. The moisture content of the soil sample was determined by the drying method. A solution with a soil-to-water ratio of 1:5 was prepared, and the electrical conductivity (EC) of the soil sample extract was determined using a Rex DDS-11A digital conductivity meter (Shanghai Instruments Scientific Instrument Co., Ltd., Shanghai, China) The relationship between the conductivity of the soil extract and the soil salt content [28] is:

$$\mathrm{Q}=3{.72\mathrm{EC}}_{1:5}+0{.44, \mathrm{R}}^2=0.98$$

(1)

where Q is the soil salinity (in g/kg), and EC is the conductivity of the soil extract (in mS/cm). The soil desalination rate [31] is the ratio of the difference between the initial and final values of the soil salinity to the initial value of the soil salinity.

According to the observation wells on the south and north sides of the study area (the observation wells were arranged in the middle of two adjacent buried pipes), the fluctuation range of the groundwater depth in the study area during the irrigation season (April to October) was 1.75–2.10 m, and during the non-irrigation season, it was below 2.20 m, as shown in Table 2. In 2013 and 2014, the underground pipe drainage, the groundwater in the observation well, and the irrigation water in the West Bank Canal were sampled three times using three bottles, each with a capacity of 500 mL. The salinity of the samples was measured by the drying method, and an average value of the measurements was taken. The depth of the groundwater sampling depended on the height of the groundwater level. After sowing on 15 April 2013, the initial salt content of each sampling point was collected. In order to simplify the initial conditions of the model, the average salt content and the initial water content of each soil layer were calculated and measured, as shown in

Table 3. Initial salt content and initial moisture content of each soil layer.

Soil Depth/cm	Initial Soil Salt Content/(g·kg–1)	Initial Soil Volumetric Water Content/(m−3·m−3)
0~20	23	15.42
20~60	20	17.03
60~100	18	20.67
100~140	14	26.34
140~160	15	29.41
160~200	15	32.52

. The DRAINMOD model was established to simulate the hydrological changes and the salt transport process in the soil during the growth period of cotton. The simulated value of the hydrology was verified by the measured groundwater level in the study area, and the simulated value of salt was verified by the measured value of the soil profile salinity at the soil layer depths from 0–140 cm.

2.3. Basic Principles of the DRAINMOD Model

The DRAINMOD model is a quasi-two-dimensional field water balance calculation model [21]. Based on the meteorological, soil, and crop data of the input simulation area, the model calculates the water balance of the soil profile daily at the midpoint of the two parallel drainage ditches/pipes in the drainage farmland, including infiltration, surface runoff, evaporation and transpiration, underground drainage, deep leakage, and groundwater depth; model input parameters are shown in

Table 4. Key input parameters of the DRAINMOD model.

Drain depth (cm)		2200
Drain spacing (m)		48
Drain coefficient (cm/day)		1.4
Effective drain radius (cm)		1.1
Depth to restrictive layer (cm)		235
Cotton planting date		15 April 2013
		10 April 2014
Cotton harvest Date		30 September 2013
		20 September 2014
Growth period segmentation (Days after planting)/day	Seedling stage	1~31
	Growth period	32~81
	lowering Boll Period	82~135
	Maturity	136~180
growing up with deeper roots/cm	Seedling stage	3
	Growth period	4~60
	lowering Boll Period	60
	Maturity	60~20

. For a certain period of time, the surface water balance equation is:

$$P=F+\Delta S+\mathit{RO}$$

(2)

where P is the amount of rainfall or irrigation (in cm), S is the water storage variation in the field surface (in cm), RO is the internal surface runoff (in cm), and F is the infiltration amount (in cm).

The water balance of the soil profile can be expressed as:

$$\Delta V=D+\mathit{ET}+\mathit{DS} - F$$

(3)

$\Delta$V is the change in the water content in the soil (in cm), D is the lateral displacement (in cm), ET is the evapotranspiration (in cm), and DS is the deep leakage (in cm).

In the DRAINMOD model, in order to simplify the calculation, it has been assumed that the salt is divided into the simplest inert and non-absorbent solute transport, and vertical migration is the main form of salt movement in the soil. According to Fick’s law and the mass conservation theorem, the soil salt migration equation can be expressed as:

$$\frac{\partial \left(\theta c\right)}{\partial t}=\frac{\partial {(D}_{\mathit{dh}}\frac{\partial c}{\partial z})}{{\partial }_{\mathrm{z}}}–\frac{\partial \left(\mathrm{qc}\right)}{\partial \mathrm{z}}$$

(4)

where D_dh is the hydrodynamic dispersion coefficient (in cm²/h^–1), c is the solution concentration (in g/l), θ is soil moisture content (in %), z is the soil depth coordinate (in cm), t is the washing time (in h), and q is the soil moisture flux (in mm/day). D_dh can be calculated using the following expression:

$$D_{dh}=\lambda \left|v\right|+D_{O\tau }$$

(5)

where λ is the longitudinal dispersion (in cm), v is the soil pore water flow rate (in cm/day), $D_O$ is the molecular diffusion coefficient of salt in free water (in cm²/day), τ is the pore curvature and is expressed by τ = ${\mathsf{\theta }}^{7/3}/{\mathsf{\theta }}_{\mathrm{s}}^2$, whereθ_s is the saturated water content of soil (in %).

The DRAINMOD model can reflect the crop yield of different drainage systems. The model calculates the relative yield of crops by cumulatively calculating the excessive water level (sum of excess water index in the top 30 cm of soil (SEW30)) and the number of drought days of crops, with a groundwater depth of less than 0.3 m at each growth stage. The relative yield of crops in the model is mainly affected by excessive soil moisture (waterlogging stress), insufficient water (drought stress), and planting delay. The relative crop yield of the model is expressed as:

$$\mathit{YR}=Y/Y_O {\times 100=\mathit{YR}}_P{ \times \mathit{YR}}_W{ \times \mathit{YRd} \times \mathit{YR}}_S$$

(6)

where $\mathit{YR}$ is the relative yield of crops (in %), $Y$ is the actual annual yield (in kg/hm²), $Y_O$ is the average annual maximum yield (in kg/hm²), $Y{R}_d$ is the relative yield considering only the drought factor (in %), $Y{R}_p$ is the relative yield considering only the sowing delay (in %), $Y{R}_S$ is the relative yield considering only the salt stress (in %), and $Y{R}_w$ is the relative yield obtained by only considering waterlogging factor (in %).

(1) Calculation of the relative yield of crops under waterlogging stress:

$$R_{yw}=109\times 0.51\times {\left({\displaystyle \sum }_{j=1}^{n}C{S}_{wj}\times \left(30-X_i\right)\right)}_w$$

(7)

where $C{S}_{wj}$ is the waterlogging sensitive factor on the jth day, n is the number of days of crop growth, and $X_i$ is the groundwater depth on the ith day (in cm).

(2) Calculation of the relative crop yield under drought stress:

$$R_{yd}=100\times 1.22\times \left({\displaystyle \sum }_{j=1}^{n}C{S}_{dj}\times {\displaystyle \sum }_{k=1}^m\left(1-\frac{AE{T}_K}{PE{T}_K}\right)\right)$$

(8)

where $C{S}_{dj}$ is the jth daylong drought sensitive factor, n is the number of days of crop growth, -$AE{T}_k$ is the actual evapotranspiration on the jth day of life (in cm/day), and $PE{T}_K$ is the potential evapotranspiration on the jth day of life (in cm/day).

2.4. The RAGA-PPC Model

The basic idea of the PPC model [41] is to project multiple sets of high-dimensional data to a low-dimensional subspace using the dimensionality reduction method, estimate the projection value that can reflect the structure or characteristics of the high-dimensional data, and analyze the specific structural characteristics of the high-dimensional data based on the size and direction of the projection value. RAGA [43] is an improved accelerated genetic algorithm based on real-number coding, which overcomes the shortcomings of binary coding, thus making the coding length of an individual equal to the number of decision variables. Consequently, it greatly enhances the optimization ability of the algorithm. Based on RAGA, the PPC model was used in this study for the RAGA operation to determine the projection direction corresponding to the maximum or the minimum index function values as the optimal projection direction.

3. Results and Analysis

3.1. Calibration and Validation of the DRAINMOD Model

Based on the hydrological and salt data of the monitoring points in the study area from 2013 to 2014, the hydrological process and the salt-leaching process predicted by the model were calibrated and verified. The data from 2013 were used for calibration, whereas those from 2014 were used for verification. The statistical indices used for evaluating the simulation results were the mean deviation (A.Mean), the mean absolute deviation (A.D), the root mean square error (RMSE), and the correlation coefficient (R²), which are mathematically expressed in Equations (9) to (12):

$$A.Mean=\frac{1}{n}{\displaystyle \sum }_{i=1}^n\left(O_i-P_i\right)$$

(9)

$$A.D.=\frac{1}{n}{\displaystyle \sum }_{i=1}^n\left|O_i-P_i\right|$$

(10)

$$RMSE=\sqrt{\frac{1}{n}{\displaystyle \sum }_{i=1}^n{\left(O_i-P_i\right)}^2}$$

(11)

$$R^2=\frac{{{\displaystyle \sum }}_{i=1}^N\left(O_i-\overline{O}\right)\left(P_i-\overline{P}\right)}{\sqrt{{{\displaystyle \sum }}_{i=1}^N{\left(O_i-\overline{O}\right)}^2}\sqrt{{{\displaystyle \sum }}_{i=1}^N{\left(P_i-\overline{P}\right)}^2}}$$

(12)

where n is the numerical comparison quantity, O_i is the measured value, P_i is the simulation value, and $\overline{O}$ and $\overline{P}$ the mean observed and simulated values.

The results of the experimental measurements and the simulation for the groundwater level are shown in Figure 3. From the figure, it can be seen that the simulated and measured values of soil salinity in the calibration plot are A.Mean = −1.72 cm, A.D = 3.84 cm, RMSE = 5.14 cm, and R² = 0.87. The A.Mean, A.D, RMSE, and R² of the simulated and measured groundwater levels in the verification area are 2.43 cm, 2.43 cm, 3.64 cm, and 0.94, respectively. The range of variation and the trend of the predicted and the measured values of the groundwater level are almost identical, and the error, which is closely related to rainfall and irrigation, is within the acceptable range. When the rainfall and irrigation are large, the water level rises, and when there is no rainfall and irrigation, the water level gradually decreases. There is no significant difference in the groundwater level between 2013 and 2014. The results of the desalination process predicted by the DRAINMOD simulation are shown in Figure 4. The simulated and measured values of soil salinity obtained are A. Mean = −0.68 g·kg^–1, A.D = 1.60 g·kg^–1, RMSE = 1.95 g·kg^–1, and R² = 0.82. The simulated and measured values of the groundwater level in the verification area are A.Mean = −1.86 g·kg^–1, A.D = 1.99 g·kg^–1, RMSE = 2.99 g·kg^–1, and R² = 0.86. The range of variation and the trend of the salt value predicted by the model and the corresponding measured values are almost identical, and the error is within the acceptable range. Some differences are observed between the simulated and measured values of the verification plot and the calibration plot, which could be due to the fact that the soil is obviously affected by the actual external factors, and there are subtle differences between the boundary conditions in the model construction process and the actual boundary conditions. The boundary conditions in the actual situation are more complex than those constructed in the model, and these differences will have a definite impact on the simulation results of the model.

In summary, the DRAINMOD model provides an accurate simulation of the hydrological changes in the soil and the salt-leaching process in the study area. The degree of model deviation is within the acceptable range.

3.2. Analysis and Results of the Simulation of the Salt Content in the Soil

Based on the commonly used values of the spacing and buried depth of the underground pipe layout [39], a buried depth of 1.9, 2.1, and 2.3 m and a spacing of 20, 30, 40, 50, 60, 70, 80, 90, and 100 m were selected. From these, 27 groups of drainage spacing and drainage pipe depth combinations were made. The specific combination method is shown in Figure 5. In order to study the continuous dynamic changes in the soil salinity as a function of time during the cotton growth period for the different subsurface pipe combinations, the soil water and salt transport process in the study area for the 27 groups of subsurface pipe combinations was simulated and calculated by the calibrated subsurface pipe model. The study period was from 20 April 2013 to 10 October 2014.

Soil salinity directionally migrates with soil water flow under drip irrigation [44]. Figure 5 shows the change in desalination rate in 0–20 cm, 20–40 cm, 40–60 cm, 60–80 cm, 80–100 cm, 100–120 cm, and 120–140 cm soil layers for the 27 different subsurface pipe combinations. It can be seen from the figure that the desalination effect is the best in the 0–80 cm soil layer, and the salt accumulation is obvious in the 80–140 cm soil layer. The desalination rate (compared to the initial salt content in 2013) decreases with the increase in the soil depth value. The desalination rate of each combination in C₁–C₂₇ in Figure 5b is higher than that in Figure 5a in the 0–80 cm soil layer, indicating that during the cotton growth period in 2013, the 0–80 cm soil layer increased with the increase in the irrigation times. The soil salinity moves downward with water and gradually decreases. Further, the desalination rate in the early growth period is greater than that in the later period. In Figure 5d, the salt accumulation rate of each combination of C₁–C₂₇ is greater than that in Figure 5c, indicating that during the cotton growth period in 2014, the change in the salt content in the soil was similar to that in 2013, and the total salt content gradually moved from shallow soil to deep soil.

Taking the buried pipe arrangement spacing of 48 m and buried depth of 2.2 m as an example, the soil salinity in the study area is observed to decrease gradually with time. Further, the average salinity concentration in the 0–140 cm soil layer decreases to about 6 g·kg^–1, and the average leaching rate is approximately 3.88 g·kg^–1. The average annual desalination rate of the 27 combinations proposed in this study is between 0.40–5.57 g/kg, and the desalination duration is between 2–30 a.

3.3. Model Prediction Results for Crop Yield under Natural Rainfall and Irrigation Conditions

The drainage strength of farmland is embodied in the buried depth and spacing arrangement of the drainage pipes. Timely elimination of the field water in the study area and reduction in the groundwater level can avoid the impact of waterlogging on crop yield. However, the characteristics of rainfall and irrigation distribution require that the drainage intensity should not be too large, in order to avoid a rapid decline in the groundwater level, or too low to affect the crop yield. Figure 6 shows the results of the average relative yield of cotton simulated by the DRAINMOD model as a function of the drainage spacing and buried depth when the surface drainage is good. In general, the relative yield of cotton decreases with increasing drainage spacing, especially when the spacing is more than 40 m. This result is similar to the results obtained by Wen et al. [45] using the DRAINMOD model to simulate the crop yield at different drainage spacings. In their study, under the same spacing conditions, the crop yield increased with increasing drainage depth. Jing [46] also drew the same conclusion in their study on the lime concretion black soil area multi-objective farmland drainage system optimization arrangement.

3.4. Analysis and Evaluation of the Comprehensive Benefits of the Subsurface Pipe Layout

Table 5. Comprehensive cost schedule of the subsurface pipe construction experiment.

Quota Number	Cost Name	Work Amount	Amount/Yuan
			Budget Estimate Quota Base Price	Footing
1	Trench earthwork/m3	211.2	6	950.4
2	Backfill/m3	211.2	4	633.6
3	Wells brick/m3	4	100	320
4	Ditch bottom artificial leveling/m2	111.15	0.2	22.2
5	Land machinery leveling/working hours	0.74	500	222.3
6	Gravel cushion/m3	97	50	3880
7	Processing bellows/m	180	10	900
8	Glass steel pipe/pipe for processing water-collecting wells	8	800	6400
9	Resin reinforced integrated collecting well/seat	4	800	2800
10	Bellows + non-woven fabric/m	84	4.6	302.4
11	250U—PVC pipe/m	160	80	9600
12	Drip tape/m	2198	0.3	549.5
14	Drip irrigation water fee/yuan		3000	3000
15	Manual management fee/yuan		4000	4000
16	Transport costs/yuan		3000	3000
2.2 m buried depth, 48 m spacing cost total/(yuan·hm−2)				36,580.4

gives a detailed list of the comprehensive project cost of the underground pipe construction test for a buried depth of 2.2 m and a spacing of 48 m. Based on Table 5, the comprehensive project cost of C₁–C₂₇ corresponding to the 27 types of subsurface pipe construction tests was budgeted, and the results thus obtained are given in

Table 6. Simulation statistics of the soil desalination effect for the 27 pipe drainage measures.

Processing Number	Depth, Spacing/(m)	Construction Cost/(Ten Thousand Yuan/hm2)	Average Desalination Rate/(g/kg)	Cotton Relative Yield (%)
C1	1.9, 20	2.90	−0.09	90.9
C2	1.9, 30	2.69	5.53	86.3
C3	1.9, 40	3.40	4.43	76.6
C4	1.9, 50	4.28	3.56	61.1
C5	1.9, 60	3.99	4.59	51.9
C6	1.9, 70	4.63	3.38	43
C7	1.9, 80	5.00	2.17	35.5
C8	1.9, 90	5.23	1.33	28.5
C9	1.9, 100	6.36	0.4	23.5
C10	2.1, 20	3.92	−1.56	90.9
C11	2.1, 30	3.61	5.57	86.6
C12	2.1, 40	3.42	4.12	77.9
C13	2.1, 50	4.30	4.35	62
C14	2.1, 60	4.01	4.59	53
C15	2.1, 70	4.64	3.65	44.1
C16	2.1, 80	5.01	2.47	36.7
C17	2.1, 90	5.23	1.51	29.9
C18	2.1, 100	6.36	0.59	24.1
C19	2.3, 20	3.96	−2.72	91
C20	2.3, 30	3.64	5.5	86.9
C21	2.3, 40	3.44	4.43	78.6
C22	2.3, 50	4.32	3.98	62.4
C23	2.3, 60	4.02	4.51	53.4
C24	2.3, 70	4.66	3.94	44.6
C25	2.3, 80	5.02	2.55	37.2
C26	2.3, 90	5.24	2.55	30.5
C27	2.3, 100	6.37	1.7	24.6

. In this evaluation, three indices, namely, construction cost, desalination rate, and relative yield of cotton, were selected (the smaller the construction cost, the better the desalination rate and the relative yield of cotton). The observation data of the different treatments are listed in Table 6.

The RAGA-PPC model was established using the evaluation indices listed in Table 6, and the model was programmed using MATLAB R2019b(A commercial math software from The MathWorks, USA). While processing the data by RAGA, the number of selected populations was 400, the crossover probability was 0.8, the mutation rate was 0.8, 0.05, and the acceleration cycle was 20 times. The maximum index function value was 0.3656. The optimum projective direction a* = (0.0058,0.8841,0.1102), and to obtain projection values for various treatments, z*(j) = (0.5663, 1.2869, 1.1702,1.0806, 1.1616, 1.0262, 0.8717, 0.7554, 0.7166, 0.4616, 1.3773, 1.1352, 1.1917, 1.1667, 1.0676, 0.9171, 0.7853, 0.7451, 0.3099, 1.3715, 1.1815, 1.1451, 1.1586, 1.1097, 0.9305, 0.9282, 0.8970). For the best projection direction of each evaluation index, the contribution rate and the degree of influence of each evaluation index on the comprehensive evaluation can be seen. The desalination rate has the largest impact on the overall evaluation results, followed by the relative yield of cotton and the construction cost. The projection value, Z(i)*, is arranged from large to small, as shown in Figure 7. According to the criterion, namely, the larger the projection value, the better the comprehensive benefit of the subsurface pipe layout, C₁₁ (buried depth 2.1 m, spacing 30 m) was found to be the optimal subsurface pipe layout scheme.

4. Discussion

The DRAINMOD model is a proprietary model for farmland water and salt management. Previous research and applications have mainly focused on the prediction, design, and management of farmland water and salt [47,48,49,50]. Although the DRAINMOD model has the advantages of few parameters and high simulation accuracy [51,52,53], it still needs to obtain the necessary data such as meteorology, soil composition, water level, and salt concentration. Under the influence of some objective factors, there still are some difficulties in continuous observations. Therefore, Yu et al. [49] studied the sensitivity of the parameters of the DRAINMOD model. The results of their study showed that the lateral saturated hydraulic conductivity was a sensitive parameter when simulating the salt content of the soil profile, which provided a reference for improving the model simulation accuracy and expanding the applicability of the model. Zhang et al. [50] simulated and verified the reliability and practicability of the DRAINMOD model by observing and predicting the groundwater depth and soil profile salinity based on the research data of the subsurface pipe drainage in coastal saline-alkali land in Laizhou, Shandong Province, which provided an important reference for analyzing the law of water and salt transport in saline-alkali farmland under subsurface pipe drainage. Chen [54] used the DRAINMOD model to simulate the hydrological effects of different drainage measures in the study area and increased the average relative yield of cotton by increasing the irrigation amount. In this study, based on the collection of meteorological, soil composition, crop characteristics, groundwater level, soil profile salinity, and other data, the DRAINMOD model was used to predict and analyze the desalination process for the engineering measures undertaken in the study area, and good simulation results were obtained by calibration and verification. This shows that the DRAINMOD model has good applicability [49] and can be used to study the water and salt management of soil in the study area [50]. Kiran et al. [55] used the DRAINMOD model to simulate the change process of soil salinity under the condition of subsurface drainage in the Golewala basin, which proved that the DRAINMOD model was reliable for soil salinity prediction under subsurface drainage in arid and semi-arid areas of Punjab, India. Pourgholam et al. [56] used DRAINMOD-S and AquaCrop models to simulate the salt concentration of paddy soil profiles under shallow and saline water conditions. By comparing the measured and simulated soil salinity, DRAINMOD-S and AquaCrop can be used as useful tools to predict and simulate the change trend of soil salinity. Moreover, DRAINMOD-S has higher accuracy and performance than AquaCrop at higher salinity levels, because the DRAINMOD-S model is a more specific salinity problem model. Foda et al. [57] used the DRAINMOD model to predict the soil salinity in the Egyptian plain. The simulation results are in good agreement with the data collected on site. The research results can provide a theoretical basis for the local subsurface drainage technology. Compared to the statistical analysis method, the DRAINMOD model has the characteristics of strong prediction ability, high work efficiency, and high precision. It overcomes the shortcomings of the statistical analysis method, such as large data demand, discontinuous data, large timespan, and inability to reveal the natural desalination mechanism. Therefore, the DRAINMOD model can provide support for the preliminary prediction and analysis of underground pipe project planning and design, land use planning and management, and the reconstruction of ecological environments.

In arid areas, subsurface drainage technology is important for improving the salinized soil by using the drainage of the farmland to wash the salt out of the soil [58]. The use of subsurface drainage technology is beneficial for leaching salt for achieving accelerated soil desalination [59], so that salt is separated from the soil with water, which provides good conditions for the growth and development of crops during the growth period and increases crop production. Studies have shown that the leaching effect of subsurface pipe drainage on soil salinity is closely related to the buried depth [60], spacing [61], irrigation amount, initial soil salinity [62], crop-planting patterns, and different outsourcing materials of the subsurface pipe [63]. This study shows that different drainage methods have different effects on soil desalination. Subsurface drainage provides a satisfactory desalination effect and can be used as the main technology for soil improvement. The soil desalination rate changes with the soil depth. In the range of 0–80 cm of the soil layer in this study, the soil desalinization rate decreases with the increase in the soil depth, and salt accumulation gradually appears after 80 cm, with the soil depth of 100–140 cm showing the most severe case of salt accumulation. This conclusion is similar to the conclusion of the study by Li [31], in which the desalination effect of 0–60 cm soil is obvious in the numerical simulation and the analysis of the subsurface pipe drainage in the drip irrigation cotton field under film in Xinjiang, but most of the salt is leached to the soil layer below 60 cm. However, the range of the desalinated soil layer is slightly different, which could be due to the difference in the buried depth and spacing of the subsurface in the test. At the same buried depth, the desalination rate decreases with the increase in the spacing of the buried pipes. This is similar to the conclusion drawn by Chen Mingyuan [63] when studying the law of water and salt migration under salt discharge by outsourcing geotextiles. The DRAINMOD model has been used to simulate the hydrological effect on the cotton yield for 27 types of subsurface pipe combinations in the study area. The results have shown that the relative yield of cotton increases with the increasing buried depth of the subsurface pipe and decreases with the increasing spacing of the subsurface pipe. The highest relative yield can reach 91%.

To date, the RAGA-PPC model has been widely used in different comprehensive evaluation schemes [41,42]. Based on the data characteristics of the evaluation indices, it can avoid the interference of human factors to the greatest extent. It can not only obtain the order of advantages and disadvantages of each scheme but also reflect the contribution rate of different indices to the overall evaluation using the best projection direction of evaluation indices, which can objectively and accurately evaluate the schemes. The DRAINMOD model has been used in this study to simulate 27 types of underground pipe layout schemes for a comprehensive evaluation. Based on the best projection direction of each evaluation index, the contribution rate and the degree of influence of each index on the comprehensive evaluation of underground pipe engineering have been determined. Based on the principle that the larger the projection value, the better the comprehensive evaluation effect, the projection values of different subsurface pipe combination schemes have been sorted to obtain the optimum combination of comprehensive evaluation.

5. Conclusions

Using the DRAINMOD model and taking the requirements of the cotton drainage and irrigation in the study area into consideration, the drainage layout and field hydrological effects of farmland subsurface pipes based on the requirements of waterlogging reduction were simulated and studied based on the long-term meteorological data and the common soil type parameters in the study area. The RAGA-PPC model was used to evaluate the comprehensive benefits of the 27 groups of concealed pipe layout combinations. The following conclusions were obtained by applying this model:

(1) The simulated values obtained from the DRAINMOD software were in good agreement with the measured values. The ranges of the RMSE, A.Mean, A.D, and R² values of the salt content were 1.95–5.14 g·kg⁻¹, −1.72–−0.68 g·kg⁻¹, 1.60–3.84 g·kg⁻¹, and 0.82–0.87, respectively. The ranges of the RMSE, A.Mean, A.D, and R² values of the simulated and measured groundwater level were 2.99–3.64 g·kg⁻¹, −1.86–2.43 g·kg⁻¹, 1.99–3.64 g·kg⁻¹, and 0.86–0.94, respectively, which were within the acceptable range. It is thus proved that the DRAINMOD model can simulate the hydrological changes in the soil and the salt-leaching process in the study area and can provide theoretical support and scientific guidance for the promotion and application of subsurface pipe salt discharge technology and drip irrigation under film in the inland arid area of Northwest China.

(2) The law of changes in the salt content in the soil for 27 types of subsurface pipe layout schemes was simulated by the model after validating the usage rate in the study area. The results showed that the desalination rate decreased with the increase in the soil depth. Under the same buried depth, the larger the spacing of the buried pipe, the smaller the desalination rate. The underground pipe engineering measures could greatly shorten the desalination process in the study area. Using the drainage system combination proposed in this study, the annual average desalination rate can be increased to 0.40–5.57 g/kg, and the desalination duration can be shortened to 2–30 a. Based on the simulation results, the relative yield of cotton was the maximum for a pipe spacing of 20 m and a buried depth of 1.9–2.1 m. The relative yield of cotton was negatively correlated with the pipe spacing and positively correlated with the buried depth.

(3) The RAGA-PPC model was used to establish the evaluation model of the comprehensive benefit of the subsurface pipe layout. The results showed that the indices affecting the comprehensive benefit of the subsurface pipe layout were the desalination rate, the relative yield of cotton, and the construction cost. The weight value of the desalination rate was much higher than the other indices, which should be analyzed emphatically in a comprehensive benefit evaluation. By evaluating the comprehensive benefits of 27 groups of buried pipe arrangement combinations, it was concluded that the best combination of buried pipes was C11 (buried depth spacing of 30 m). Based on the data characteristics of the evaluation indices, the RAGA-PPC model avoided the interference of human factors to the greatest extent and could objectively and accurately evaluate 27 types of underground pipe layout schemes.