Modeling Comprehensive Deficit Irrigation Strategies for Drip-Irrigated Cotton Using AquaCrop

Abstract

The development of a crop production strategy through the use of a crop model represents a crucial method for the assurance of a stable agricultural yield and the subsequent enhancement thereof. There are currently no studies evaluating the suitability of the AquaCrop model for the drip irrigation of Gossypium barbadense in Southern Xinjiang, which is the primary planting region for Gossypium barbadense in China. In order to investigate the performance of the AquaCrop model in simulating the growth of cotton under mulched drip irrigation, the model was locally calibrated and validated according to different irrigation thresholds during a key growth period of two years. The results of the simulation for total soil water (TSW), crop evapotranspiration (ET_c), canopy coverage (CC), aboveground biomass (Bio), and seed cotton yield demonstrated a high degree of correlation with the observed data, with a root mean square error (RMSE) of <11.58%. The Bio and yield simulations demonstrated a high degree of concordance with the corresponding measured values, with root mean square error (RMSE) values of 1.23 t ha⁻¹ and 0.15 t ha⁻¹, respectively. However, the predicted yield declined in the verification year, though the prediction error remained below 15%. Furthermore, the estimated evapotranspiration (ET_c) value demonstrated a slight degree of overestimation. Generally, the middle and late stages of cotton growth led to an overestimation of the TSW content. However, the prediction error was less than 13.99%. Through the calculation of each performance index of the AquaCrop model, it is found that they are in the acceptable range. In conclusion, the AquaCrop model can be employed as a viable tool for predicting the water response of cotton to drip irrigation under mulched film in Southern Xinjiang. Based on 64 years of historical meteorological data, three years were selected as scenarios for simulation. Principal component analysis (PCA) showed that, in a local wet year in Southern Xinjiang, the irrigation quota was 520 mm, and the irrigation cycle was 6 days/time. In normal years, the irrigation quota was 520 mm, with an irrigation cycle of 6 days/time. In dry years, the irrigation quota was 595 mm, with an irrigation cycle of 10 days/time. This allowed for higher seed cotton yields and irrigation water productivity, as well as the maximization of cotton yields and net revenue in the arid oasis area of Southern Xinjiang.

1. Introduction

Global warming has led to a rise in the frequency of meteorological disasters, such as abnormally high temperatures, droughts, and heavy rainfall. These events create instability in agricultural production, which, in turn, affects crop yields [1,2]. Among the various abiotic stresses, drought is one of the most significant threats to agricultural production. Anomalous temperature increases and precipitation uncertainty directly affect soil moisture evaporation and plant transpiration in agricultural fields, which, in turn, affect crop water consumption. Furthermore, fluctuations in regional precipitation have implications for both local irrigated agriculture and water management [3,4]. Agriculture is a significant contributor to the economy of Xinjiang, and ensuring the continued production of food is crucial for maintaining social stability in the region. As a typical irrigated agricultural area in the arid oasis zone of inland Northwest China, Xinjiang tends to apply excessive amounts of water to ensure a suitable soil microenvironment for crop growth in arid or semi-arid regions due to high temperatures, and the high evaporative demand during the crop-growing season [5] further exacerbates water shortages in Xinjiang. Consequently, the efficient utilization of water resources in Xinjiang is crucial for the sustainable development of agriculture.

Gossypium barbadense, which is also known as long-staple cotton, is a high-end raw material for textiles. Due to its susceptibility to light and heat conditions, only the United States, Egypt, and a few other countries are able to cultivate it. With the advent of molecular genetics and biotechnology, new varieties of Gossypium barbadense with enhanced tolerance to the environment and superior fiber length have been successfully developed [6,7], paving the way for cultivation in arid regions. The demand for Gossypium barbadense has been on the rise in recent years due to its exceptional fiber quality [8]. It is easy to grow Gossypium barbadense in Southern Xinjiang because the weather is right for it. This plant has very good fibers and can be spun easily, so Southern Xinjiang has become China’s main source of high-quality Gossypium barbadense (cotton). The cultivation of cotton is susceptible to environmental fluctuations, with climatic conditions and agronomic practices exerting a significant influence on the crop’s yield and quality [9]. However, in the context of climate change and water scarcity [10], there is a dearth of research on irrigation strategies for cotton with low water consumption and high water use efficiency. Therefore, the effective reduction of irrigation water consumption for cotton represents a crucial approach to alleviating the shortage of water resources in the Southern Xinjiang region, improving the efficiency of water resources, optimizing the structure of the cotton industry in Xinjiang, and facilitating the upgrading of the cotton industry there [11].

When employed in conjunction with experimental studies, crop simulation models have proven to be invaluable tools for integrating a vast array of environmental and management factors that influence crop production. The most common crop models include WOFOST [12,13], APSIM [14,15], DSSAT [16,17], and AquaCrop, a water-driven model based on a conserved relationship between biomass and transpiration. This model is widely used for crop yield simulation due to its fewer input parameters, simplicity of operation, and accuracy of simulation [18,19,20]. To date, the model has been tested in a variety of soils, crops, and climates around the world. The objective of these tests was to determine crops’ response to soil moisture and to facilitate on-farm irrigation management under climate change [17,21,22]. Researchers from around the world have simulated a wide range of crops and have determined the growth parameters for them.

Research has shown that the AquaCrop model can reasonably replicate the growth processes of popular crops, such as cotton [23,24], potato [25], wheat [22], and maize [18], as well as those of vegetables, such as saffron [26], moss [27], and oilseed rape [28]. Furthermore, the model can simulate grain yields at the end of a growing season while accounting for biomass buildup and salt and fertility impacts. In addition, models may be used to determine the best fertility and irrigation schedules for a range of climatic circumstances by analyzing long-term meteorological data series, frequencies, and a variety of scenarios and field management techniques [29,30,31]. Usually, AquaCrop has to be calibrated and validated against regional crop types and environments. The application of AquaCrop to the modeling of deficit irrigation optimization for Gossypium hirsutum L. under long-term climate conditions has been investigated in the dry oasis area of Southern Xinjiang and the humid region of the North Plain in China (NPC) [23,24]. It has been discovered that variations in the model parameters for the same crop can also have an impact on the model’s simulated mistakes in yield and biomass [32]. Because of differences in plant height, leaf shape, and other geographical and temporal variables of plant structure and morphology, cotton displays several genotypes. These traits set it apart from Gossypium hirsutum L., which influences the traits of the canopy of cotton. As a result, the yield of cotton may not be precisely predicted by past studies on Gossypium hirsutum L. variants. Thus, it is essential to look into the characteristics of the cotton model in the dry oasis area near the southern border. This is necessary to close the gap in the knowledge of the AquaCrop simulation parameters for cotton in the vicinity of the desert oasis.

This study set out to achieve three goals: (1) determining irrigation strategies for years with varying precipitation categories; (2) assessing the total soil water, canopy cover, biomass, and crop yields in AquaCrop with different simulated water stresses; (3) using field measurements to calibrate and validate the AquaCrop model for cotton under various irrigation regimes.

2. Materials and Methods

2.1. Site Description

The research was conducted on an experimental farm (76°56′51″ E, 39°12′ 28″ N) in Yuepuhu County, Kashgar, Xinjiang, China (2021–2022). A map of the study area is shown in Figure 1. The soil in the cotton field was classified as silt loam with 0.40% organic matter, 0.30 g kg⁻¹ available nitrogen, 11.5 mg kg⁻¹ available phosphorus, 86 mg kg⁻¹ available potassium, 17.3 g kg⁻¹ total nitrogen, and 10.1 g kg⁻¹ nitrogen. Other relevant soil characteristics (0–100 cm) before seeding are shown in Table S1. In order to ensure uniformity between test plots, soil and fertilizer management was not carried out at the experimental sites prior to the study. The climate of the region is characterized by a temperate, arid continental climate with hot summers and cold winters. The annual sunshine time of the test site is 2825 h, the annual average precipitation is 66.4 mm, and the evaporation is 1600 mm. Detailed information on the temperature, precipitation, and reference evapotranspiration (ET₀) in the different growth stages of cotton can be found in Figure S1.

2.2. Design of the Test Pit Experiment and Crop Management

The cotton (Gossypium barbadense. Xin78) was sown on 15 April 2021 and 15 April 2022, and it was harvested on 15 October 2021 and 28 September 2022. A local machine-picked cotton planting mode is adopted. The irrigation method in this experiment was mulched drip irrigation using a single-wing labyrinth irrigator produced by Xinjiang Tianye Water Saving Company (Shihezi, China). The distance between adjacent emitters of drip irrigation tape was 20 cm, the rated working pressure of the emitters was 0.1 MPa, and the design flow rate of the emitters was 3.2 L h⁻¹. Three drip irrigation tapes were arranged under a 2.05 m plastic film, and each tape was used to irrigate two rows of cotton (Figure S1c). The irrigation water source was well water; the buried depth of the well was below 15 m, the salinity of the water was about 0.954 g L⁻¹, and the irrigation water was controlled with a water meter. The irrigation water quotas at the squaring stage (SQS) and the blooming stage (BLS) were based on different proportions of FC (field capacity) and were used as the upper limit of irrigation water and the lower limit of irrigation water. Each irrigation quota was based on formula 1. The crop water consumption (ET) was estimated with the following water balance formula (Formula (2)). Based on the different water requirements at the SQS and the BLS, the drip irrigation test schemes in 2021 and 2022 are shown in Figure S1a,b. The irrigation test plan and actual irrigation conditions are shown in

Table 1. Irrigation test scheme and actual irrigation situation.

Year	Treatment	Irrigation Threshold		Irrigation Quotas and Frequency of Irrigation			Irrigation during Critical Fertility Period (mm)	Whole Fertility Cycle
		Squaring Stage	Blooming Stage	Seeding Stage (mm)	Squaring Stage (mm)	Blooming Stage (mm)		Total Irrigation (mm)
2021	W1	50–100%	50–100%	70	42 × 2	66 × 4	346	416
	W2	60–80%	60–80%	70	17 × 4	26 × 11	356	426
	W3	60–90%	60–90%	70	25 × 3	39 × 8	390	460
	W4	60–100%	60–100%	70	34 × 3	52 × 6	415	485
2022	WD1	55–90%	60–90%	70	29 × 2	39 × 6	295	365
	WD2	70–100%	70–100%	70	29 × 2	26 × 10	321	391
	WD3	55–90%	60–90%	70	29 × 2	13 × 21	334	404
	WD4	55–90%	70–90%	70	21 × 3	39 × 6	299	369
	WD5	55–90%	80–90%	70	21 × 3	26 × 10	325	395
	WD6	65–90%	80–90%	70	21 × 3	13 × 21	338	408
	WD7	65–90%	70–90%	70	13 × 5	39 × 7	338	408
	WD8	65–90%	80–90%	70	13 × 5	26 × 11	351	421
	WD9	75–90%	60–90%	70	13 × 7	13 × 20	350	420
	CK	75–90%	70–90%	70	-	-	575	645

Note: The irrigation schemes for 2021 and 2022 were not the same; there were upper and lower irrigation restrictions during the square period and flowering period. The treatments selected in 2021 were only for model calibration. The irrigation amount and irrigation cycle refer to the squaring stage and blooming stage; the squaring stage and blooming stage are the key periods of fertility, and the total irrigation is the irrigation amount for the whole fertility cycle of cotton.

. Fertilizer dripped into the soil at the root of the cotton along with water. The amount of fertilizer applied was determined according to the local planting method; the fertilizer applied at the SQS was applied at 30% of the total amount, the fertilizer applied at the BLS was applied at 70% of the total amount, and the total fertilizer was applied at 510 kg ha⁻¹ N, 454 kg ha⁻¹ P₂O₅, and 163 kg ha⁻¹ K₂O.

The quota for each drip irrigation event was determined with the following formula:

1000 M = h p (θ_max − θ_min)

(1)

where M is the individual irrigation quota (mm) under drip irrigation; h is the designed wetting depth (cm), which was 40 cm at the cotton budding stage and 60 cm from the flowering stage to the boll opening stage (BOS); p is the moisture ratio, which was 0.7 under drip irrigation; θ_max and θ_min are the upper and lower limits of soil moisture (volume moisture content) in the designed wetting layers, respectively.

$$ET =P +U +IW- R - D -10 \sum _{i=1}^n[H_i ( {\theta }_{i1} - {\theta }_{i2}\left)\right]$$

(2)

where ET, P, and U represent the amounts of crop water consumption, precipitation, and groundwater recharge, respectively (mm); IW, R, and D represent the amounts of irrigation, surface runoff, and deep leakage, respectively (mm); i is the number of soil layers; n is the total number of soil layers; Hi is the thickness of soil layer i (cm); θ_i1 and θ_i2 are the volumes of soil moisture content at the beginning and end of the calculation interval, respectively (%). Due to the fact that there was no surface runoff and deep leakage in the experimental area because of the limited drip irrigation and precipitation, R = 0 and D = 0.

2.3. Measurement Indicators and Methods

According to the characteristics of cotton growth and canopy development, in the seedling stage (SS), squaring stage (SQS), flowering stage (FS), boll stage (BS), and boll opening stage (BOS), the soil moisture content, above-ground biomass, and leaf area were monitored. The soil moisture content was measured with a Diviner 2000 soil moisture meter the day before irrigation and the day after irrigation; the instrument was calibrated with the drying method. The calibration formula is shown in Formula (3). The depth was measured at 0–100 cm, the 0–40 cm plowing layer was measured in 10 cm intervals, and the 40–100 cm range was measured in 20 cm intervals. The layout of the probe is shown in the planting model diagram in Figure S1b. The soil moisture content was determined by using the average soil moisture content at the point where a common water probe was placed [33]. On each measurement date, five representative plants were taken from each plot to obtain the seasonal aboveground biomass. The samples were killed at 105 °C for 30 min and then dried at 75 °C to a constant weight. In addition, 10 cotton plants were randomly selected in each plot. The leaf area was measured using the punching method [34], where the leaves of individual plants were punched into 40 pieces with a 15 mm punch. The LAI data were then converted into canopy cover (CC) using the LAI and CC relationship formula established by Hsiao et al. [21].

SF = 0.06623 × θ ^0.1093

(3)

CC = 1.005 (1 − exp (−0.6 LAI))^1.2 × 100%

(4)

$$\mathit{LAI}={\displaystyle \frac{\mathit{total} \mathit{area} \mathit{of} \mathit{the} \mathit{leaf}}{\mathit{occupied} \mathit{land} \mathit{area}}}$$

(5)

Here, SF is the capacitance–frequency reading ratio, and θ is the moisture content in the soil volume (cm³ cm⁻³).

HI = Ya/B

(6)

At the time of harvest, a 6.90 m² (2.3 m × 3 m) sample of cotton was collected from the center of each plot. The samples were then dried and weighed in order to determine the seed cotton yield. Here, Ya is the seed cotton yield of cotton (t ha⁻¹), B is the aboveground biomass of cotton (t ha⁻¹), and HI (harvest index) is the ratio of the seed cotton yield to aboveground biomass (%).

The yields, amount of irrigation water applied, and consumptive water use were used to compute and analyze WP_Irrig and WP. The current prices and production costs were collected under the conditions of a free market, excluding any crop subsidies received, based on prevailing market rates and actual expenses [35].

WP_Irrig = Ya/IW

(7)

WP = Ya/ET

(8)

Here, WP_Irrig is the water productivity per amount of irrigation water applied (kg m⁻³), IW is the amount of irrigation water applied (m³ ha⁻¹), ET is the actual crop evapotranspiration (crop consumptive water use) calculated for the plots (m³ ha⁻¹), WP is the water productivity with respect to the crop consumptive water use (kg m⁻³), NR (net revenue) was calculated by subtracting the total costs from the total revenue (CNY ha⁻¹), and TC (total cost) included all costs of growing cotton, including labor costs, material costs, water costs, and electricity costs (CNY ha⁻¹).

2.4. Description of the Model (AquaCrop 6.1)

This model establishes a functional connection between CC and ET by disaggregating ET_c into components related to transpiration (Tr) and evapotranspiration (E). Transpiration or evapotranspiration coefficients are employed to estimate ET₀. The aboveground biomass (B) is estimated using the computed Tr and the standard crop water productivity (WP*). The harvest index value is then utilized to convert B into the final yield (Ya) using the following formulas:

Tr = Kc_Tr,x × CC × ET₀

(9)

E = K_r (1 − CC) K_ex × ET₀

(10)

$$B=\mathit{WP}* \times \left({\displaystyle \frac{\mathit{Tr}}{{\mathit{ET}}_0}}\right)$$

(11)

Ya = f_HI HI₀ B

(12)

The standard crop transpiration coefficient (Kc_Tr,x) is calculated using the following variables: canopy cover in CC (%), the maximum standard crop transpiration coefficient, and the evapotranspiration reduction coefficient. The K_r coefficient is employed to regulate the impact of surface water deficits. The maximum soil evapotranspiration coefficient is represented by K_ex, while the water stress regulating factor is represented by f_HI. Finally, the reference harvest index is represented by HI₀.

Crop leaf and canopy development, stomatal opening and transpiration, canopy senescence, and the yield index are the main factors affected by water stress. The susceptibility of a crop to water stress is indicated by the water stress coefficient (K_s). The AquaCrop model employs four processes to explain how water stress affects crops: the yield index, stomatal closure, fast canopy senescence, and restricted canopy development. The implications of these processes depend on the sensitivity of the root zone to the total decrease in available soil water. The user may select their desired watering schedule, irrigation strategies, and irrigation schedule using the AquaCrop model. A film module within the site management documentation of the model allows the user to choose various film materials and coverage to replicate film circumstances. In addition to the effects of field surface features and weed control on soil moisture, the model also considers the impact of temperature, salinity stress, and soil fertility on crops. A comprehensive account of the underlying principles, assumptions, and hypotheses of these simulations is provided [19,21,36].

2.5. Model Input, Calibration, and Verification

The experimental field data from the 2021 and 2022 seasons were employed for calibration and verification purposes. A total of 14 groups of treatments were utilized for calibration in 2021, while 10 treatments were employed for verification in 2022. The three principal irrigation growth stages (SQS, FS, and BS) were each associated with 14 distinct irrigation thresholds. The meteorological data required for the model, comprising solar radiation (MJ m⁻²), maximum temperature (°C), minimum temperature (°C), and rainfall (mm), were provided by a HOBO weather station installed at the test site with an RX3000 data logger (Onset, Bourne, MA, USA). The ET₀ value is estimated using the Penman–Monteith formula via AquaCrop’s ET₀ calculator. Carbon dioxide concentrations were calculated using the Mauna Loa Observatory’s annual average.

Table 2. Crop parameters of the AquaCrop model.

Crop Parameter	Value	Remarks
Base temperature (Tbase)/(°C)	12	Measured
Upper temperature (Tupper)/(°C)	35	Measured
Crop transpiration coefficient KcTR,x	1.30	Calibrated
Initial canopy cover (CC0)/(%)	1.22	Calibrated
Canopy growth coefficient (CGC)/(% d−1)	6.9	Calibrated
Maximum canopy cover (CCX)/(%)	88	Calibrated
Canopy decline coefficient (CDC)/(% d−1)	8.8	Calibrated
Normalized crop water productivity (WP)/(g m−2)	24	Calibrated
Reduction coefficient of WP* during yield formation (f yield)	74	Calibrated
Maximum effective rooting depth/(m)	0.70	Recommended
Reference harvest index (HI0)/(%)	35	Calibrated
Upper threshold for canopy expansion (Pexp,upper)	0.2	Calibrated
Lower threshold for canopy expansion (Pexp,lower)	0.64	Calibrated
Upper threshold for stomatal closure (Pclo,upper)	0.32	Calibrated
Upper threshold for canopy senescence (Psen,upper)	0.6	Calibrated
Lower threshold of the impact of salt on crop growth (Ece,lower)/(dS m−1)	8	Calibrated
Upper limit of impact threshold of salt on crop growth (Ece,upper)/(dS m−1)	27	Calibrated

presents the remaining parameters of the model.

2.6. Model Evaluation Indicators

To assess the model’s performance during calibration and validation, five statistical metrics were utilized, as listed in Formulas (13) to (17). These were the prediction error (Pe), coefficient of determination (R²), root mean square error (RMSE), Nash coefficient (EF), and Willmott’s consistency index (d).

$$\mathit{Pe}={\displaystyle \frac{\left(S_i - M_i\right)}{M_i}}$$

(13)

$$R^2={\left({\displaystyle \frac{{\sum }_{i=1}^n (S_{i }- \stackrel{-}{S}) (O_i-\stackrel{-}{O})}{\sqrt{{{\sum }_{i=1}^n (S_i-\stackrel{-}{S})}^2{{\sum }_{i=1}^n (O_i-\stackrel{-}{O})}^2}}}\right)}^2$$

(14)

$$\mathit{RMSE}=\sqrt{{\displaystyle \frac{1}{n}} {\sum }_{i=1}^n{ (S_i -{ O}_i)}^2}$$

(15)

$$\mathit{EF}={\displaystyle \frac{{\sum }_{i=1}^n{ (O_i-O)}^2-{\sum }_{i=1}^n{ (S_i -{ O}_i)}^2}{{\sum }_{i=1}^n{ (O_i-O)}^2}}$$

(16)

$$d=1-{\displaystyle \frac{{\sum }_{i=1}^n{ (S_{i }-{ O}_i)}^2}{{\sum }_{i=1}^n{ (\left|{ S}_i-O\right|+\left|O_i-O\right|)}^2}}$$

(17)

In this context, n represents the number of observations, Si denotes the simulated value, Oi signifies the measured value, S stands for the simulated mean, and O represents the measured mean. A superior fit is indicated by lower Pe and RMSE values, as well as higher R² and d values. If EF ≥ 0.80, the model performs well; if EF ranges from 0.60–0.79, the model performs well; if EF ranges from 0.40–0.59, the model performs acceptably; if EF ≤ 0.39, the model performs poorly.

2.7. Irrigation Scenario Design

Following calibration and validation, the model was employed to evaluate cotton growth on Lower Island in diverse field management and historical climate scenarios (Figure S1). The irrigation amount at the seedling stage was 70 mm, and the irrigation quotas at the key growth stage were 360, 405, 450, and 525 mm, respectively. The comprehensive implementation of historical climate scenarios in the AquaCrop model is shown in

Table 3. Implementation of comprehensive historical climate scenarios in the AquaCrop model.

Simulation Scenario	Irrigation Amount (mm)	Irrigation Cycle (d)	Irrigation Frequency	Simulation Scenario	Irrigation Amount (mm)	Irrigation Cycle (d)	Irrigation Frequency
T1	360	6	16	T9	450	6	16
T2	360	8	12	T10	450	8	12
T3	360	10	10	T11	450	10	10
T4	360	12	8	T12	450	12	8
T5	405	6	16	T13	525	6	16
T6	405	8	12	T14	525	8	12
T7	405	10	10	T15	525	10	10
T8	405	12	8	T16	525	12	8

Note: The irrigation amount in the table refers to cotton irrigation during the critical growth period (mm); the irrigation cycle refers to the critical growth irrigation cycle (d). Irrigation frequency refers to the number of times the water was filled during the critical growth period. The key growth periods were the SQS and the BLS. In addition, the seedling stage was irrigated four times according to the irrigation volume of 70 mm, and the boll opening stage (BOS) was not irrigated according to the local irrigation habits.

. A total of 16 irrigation schemes based on different stages were established for irrigation cycles of varying lengths, including 6, 8, 10, and 12 days. Based on the meteorological data from 1958 to 2022 collected by the Meteorological Data Center of the China Meteorological Administration (http://data.cma.cn (accessed on 20 March 2023)) in Yuephu County, Kashgar, the cumulative rainfall and accumulated ET₀ during the cotton growing period were calculated (Figure S2a,b). The Pearson III distribution method was employed for the classification of three precipitation year types [37]. Tsakiris, G. et al. [38] calculated the RDI_st value based on cumulative rainfall and cumulative ET₀ and used the Survey drought index (RDI_st) from 1958 to 2022 to categorize the severity of drought (Figure S2c). The analysis of precipitation categories and drought severity from 1958 to 2022 indicated that 2021 and 2022 were normal and wet years, respectively. Based on the Pearson III precipitation year type and RDI_st value, 2004, 2013, 2015, 2016, and 2005 were selected as the wet years for the simulation. The normal years were 1962, 1992, 1963, 1984, and 1988. The simulated dry years were 1961, 1979, 1975, 1970, and 1969.

$${a_k}^{\left(i\right)}={\displaystyle \frac{\sum _{j=1}^k P_{\mathit{ij}}}{\sum _{j=1}^k P{\mathit{ET}}_{\mathit{ij}}}}$$

(18)

$${{\mathit{RDI}}_{\mathit{st}}}^{\left(i\right)}={\displaystyle \frac{y^{\left(i\right)}-\overline{y}}{{\mathsf{\sigma }}_y}}$$

(19)

Here, P and PET are the precipitation and potential evapotranspiration of the j-th month of the i-th year, yⁱ is lna_k, ${\overline{y}}^i$ is its arithmetic mean, and σ_y is its standard deviation; k is the number of growing seasons: 64.

2.8. Data Analysis

The data obtained in this experiment were processed using Excel 2021 (Microsoft Corp, Redmond, WA, USA). A planting pattern diagram was created using Microsoft PowerPoint 2021 (Microsoft Corp, Redmond, WA, USA). The results of each treatment were analyzed using SPSS statistics 25.0 (SPSS Inc. IBM Corp., Armonk, NY, USA). Graphs were generated using Origin 9.0 (Origin Lab Corp., Northampton, MA, USA).

3. Results

3.1. Total Soil Water And ET_c

The simulated and measured values of TSW in 2021 were employed for parameter calibration (Figure 2), and validation was carried out using the simulated and measured values of TSW for different treatments in 2022 (Figure 3). Overall, the simulated and observed TSW values for all treatments in 2021 demonstrated a strong correlation with precipitation and irrigation (Pe = −12.37%–−6.82%, R² = 0.27–0.72, RMSE = 7.88–27.75, EF = −0.34–−0.64, d = 0.60–0.91). In the model validation using the data from 2022, the model overestimated the TSW content in the middle and later stages of growth, and the indicators of goodness of fit (i.e., Pe, R², RMSE, EF, d) were −22.75–13.99, 0.27–0.80, 3.98–21.95 mm, 0.11–0.75, and 0.55–0.94, respectively. During the early reproductive phase, there was an overall trend of the simulated values of TSW being underestimated. The discrepancy between the simulated and actual values of TSW gradually increased as the lower irrigation limit was approached when the top irrigation limit was 90% FC. For the extremely water-stressed WD1 and overwatered CK treatments, the discrepancies between the simulated and measured TSW values were even more pronounced, with WD1 exhibiting Pe, R², RMSE, EF, and d values of −5.14–−7.97, 46.49%, 10.26 mm, 0.32, and 82.7, respectively. The Pe, R², RMSE, EF, and d values for CK were −22.75–−12.43, 27.03%, and 21.95%, respectively. The Pe, R², RMSE, EF, and d values for CK were 27.03%, 21.95 mm, 0.24, and 0.55, respectively. A comparison of the observed and simulated values of total ET_c for the two years of the growing season of cotton is presented in

Table 4. Simulated and measured crop evapotranspiration (ET_c).

Year	Treatment	ETc (mm)
		Simulated	Measured	Pe (%)	Dif
2021	W1	441	445	−0.70	−3
	W2	490	481	1.72	8
	W3	492	480	2.43	12
	W4	499	512	−2.70	−14
2022	WD1	421	436	−3.53	−15
	WD2	427	448	−4.80	−22
	WD3	429	456	−6.11	−28
	WD4	419	423	−1.17	−5
	WD5	433	447	−2.97	−13
	WD6	432	462	−6.47	−30
	WD7	432	468	−7.67	−36
	WD8	434	474	−8.50	−40
	WD9	436	485	−10.06	−49

Note: Dif: difference between the measured and simulated ET_c; Pe: prediction error of the simulated ET_c with respect to the measured values.

. The simulated values of ET were greater than the observed values, and Pe was greater for all treatments, with a range of −10.06–2.47%. The simulated values of total ET_c for these two years displayed a range of −10.06–2.47%, with an R² value of 0.54, RMSE of 25.30 mm, EF of −0.22, and d of 0.77. These results indicate that the simulation of ET_c was not satisfactory.

3.2. Canopy Cover

Figure 4 and Figure 5 illustrate the growth process of cotton. The AquaCrop model accurately simulated the development of the calibrated CC for each treatment. As indicated by the high R² (≥0.97), the high d value (≥0.99), the high EF value (≥0.95), and the low estimation error (−6.86% ≤ Pe ≤ 26.17%, 3.78% ≤ RMSE ≤ 7.08%), the model demonstrated a high degree of accuracy. The maximum canopy cover (CC) that could be achieved with the 14 irrigation treatments for cotton decreased with the increase in water stress applied. Across all treatments, the impact of water deficiency stress on cotton growth was greater in 2022 than in 2021. The validation phase in 2022 revealed higher estimation errors than those observed during calibration, particularly in treatments with significant water deficits. In the 2022 validation, the ranges for Pe, R², RMSE, EF, and d for canopy cover across treatments were 15.90%–56.80%, 0.93–1.00, 2.50–11.98%, 0.82–0.99, and 0.96–1.00, respectively. The experimental region was located in an arid inland zone at the southern border, where the WD1, WD4, and WD7 treatments demonstrated an inadequate lower limit of irrigation at the BLS, resulting in an acceleration of leaf senescence. In the validation phase, the model demonstrated a tendency to overestimate the development and senescence of CC. The larger error observed in the WD1 treatment was primarily attributed to a low lower limit of soil moisture, resulting in the model’s overestimation of canopy cover in the mid-to-late growth stage of cotton. Conversely, the irrigation floors of 65% FC and 75% FC were found to be highly accurate. Excessive water stress at the BLS was identified as a factor contributing to the decreased accuracy of the simulation of the canopy because the temperatures observed in the southern region of the border were higher than the upper limit of temperature for the growth of cotton (T_upper).

3.3. Aboveground Biomass

Figure 6 illustrates the results of the calibration of the biomass accumulation process for cotton. Figure 7 presents the results of the validation of the biomass accumulation process for cotton. The AquaCrop model estimated the progress of aboveground biomass with satisfactory accuracy under the different irrigation treatments. The difference between the simulated and measured values decreased with the lower irrigation limit in the treatment with severe water stress. In the calibration phase, the values of Pe, R², RMSE, EF, and d for all treatments were within the ranges of −39.74–21.41%, 0.97–1.00, 1.53 t ha⁻¹–3.34 t ha⁻¹, and 0.94–1.00, respectively. The measured values were found to be closer to the simulated values in the validation phase, with the simulation effect becoming more credible. The corresponding value of the goodness-of-fit index was found to be −24.86%–−24.41%, with a value of 0.97. The model displayed a tendency to underestimate the observed values, particularly after BS, with the following values: −0.98, 1.91 t ha⁻¹–4.36 t ha⁻¹, 0.87–0.98, and 0.97–1.00, respectively. This overestimation was particularly evident in the water-stressed treatments.

3.4. Biomass and Final Seed Yield

Figure 8 presents the outcomes of the model calibration and validation for the biomass and seed cotton yield. During the calibration process, the Pe, R², RMSE, EF, and d values for the simulated and measured aboveground biomass in all treatments were found to be −39.74–24.41%, 0.98, 1.23 t ha⁻¹, 0.99, and 0.99, respectively. During verification, the Pe, R², RMSE, EF, and d values were found to be −24.86%–24.41%, 0.96, 1.33 t ha⁻¹, 0.95, and 0.99, respectively. The simulation error for the biomass in the WD1, WD4, and WD7 treatments was considerable. The Pe, R², RMSE, EF, and d values between the simulated and measured values were −0.28–3.77%, 0.93, 0.15 t ha⁻¹, −0.91, and 0.75, respectively. The Pe, R², RMSE, EF, and d values were found to be in the ranges of −0.09–4.45%, 0.81–0.14 t ha⁻¹, 0.54, and 0.84. At the lower limit of 75% FC irrigation, the model was capable of accurately simulating biomass and yield. The yield of cotton displayed a positive correlation with the increase in the irrigation amount, reaching a peak at the full irrigation level. However, when the irrigation amount exceeded this threshold, the yield displayed a decline, which was potentially due to the adverse effects of over-irrigation.

3.5. Model Application

Following the calibration and validation of the model, 16 irrigation combinations were selected for simulation. These were imported from meteorological data from 1958 to 2022, with the basic parameters of the model remaining unchanged (Table 2). The three scenario years were based on five typical years and 16 irrigation treatments, resulting in a total of 240 scenarios. The simulations of seed cotton yield, final aboveground biomass, ET, WP, and WP_Irrig in different irrigation scenarios are presented in

Table 5. Comparison of the final yield, biomass, Bio, ET, WP, and WP_Irrig of cotton in different comprehensive scenarios in wet years, normal years, and dry years.

Treatment	Scenario Simulation
	Wet Years						Normal Years						Dry Years
	Ya	Bio	ET	WP	WPIrrig	HI	Ya	Bio	ET	WP	WPIrrig	HI	Ya	Bio	ET	WP	WPIrrig	HI
T1	6.25 e	16.10 f	509.52 fg	1.23 cd	1.46 a	38.86 b	5.16 fg	12.93 h	506.78 i	1.02 ef	1.20 a	39.96 cde	4.66 g	11.44 h	507.30 h	0.92 de	1.08 a	40.78 def
T2	6.18 f	15.70 g	496.24 hi	1.25 b	1.44 b	39.42 a	4.99 hi	12.37 i	488.16 j	1.02 de	1.16 bc	40.36 bcd	4.51 h	10.94 i	486.80 j	0.93 cd	1.05 b	41.28 cd
T3	6.20 ef	15.65 g	491.40 i	1.26 a	1.44 ab	39.62 a	5.09 gh	12.52 i	488.32 j	1.04 ab	1.18 ab	40.72 ab	4.62 g	11.02 i	486.74 j	0.95 b	1.07 a	41.90 ab
T4	6.05 g	15.24 h	478.56 j	1.27 a	1.41 c	39.74 a	4.91 i	11.96 j	467.64 k	1.05 a	1.14 cd	41.12 a	4.46 h	10.54 j	463.52 k	0.96 a	1.04 bc	42.38 a
T5	6.37 cd	16.67 e	525.48 de	1.21 ef	1.34 d	38.20 d	5.30 e	13.47 f	527.90 fg	1.01 fg	1.12 de	39.38 ef	4.85 e	12.02 f	532.24 f	0.91 ef	1.02 d	40.34 fg
T6	6.35 cd	16.55 e	521.00 de	1.22 de	1.34 d	38.42 cd	5.23 ef	13.15 g	515.06 hi	1.02 ef	1.10 e	39.86 de	4.84 e	11.73 g	517.54 g	0.94 c	1.02 d	41.25 cd
T7	6.32 d	16.30 f	510.36 f	1.24 bc	1.33 d	38.76 bc	5.28 ef	13.11 g	508.22 hi	1.04 abcd	1.11 e	40.32 bcd	4.87 e	11.63 g	510.24 h	0.96 ab	1.03 cd	41.90 ab
T8	6.23 ef	16.07 f	502.26 gh	1.24 b	1.31 e	38.82 b	5.15 fg	12.77 h	495.44 j	1.04 abc	1.09 e	40.48 bc	4.76 f	11.30 h	494.32 i	0.96 a	1.00 e	42.16 ab
T9	6.47 ab	17.33 c	543.80 bc	1.19 g	1.24 f	37.34 g	5.47 bc	14.14 d	549.96 cd	1.00 gh	1.05 f	38.78 gh	5.12 c	12.74 d	563.36 c	0.91 ef	0.98 f	40.18 g
T10	6.46 ab	17.23 c	539.28 bc	1.20 fg	1.24 f	37.46 fg	5.45 cd	13.92 e	542.70 de	1.01 fg	1.05 f	39.18 fg	5.11 c	12.47 e	546.82 d	0.94 c	0.98 f	40.98 de
T11	6.41 bc	16.95 d	528.88 d	1.21 ef	1.23 fg	37.80 ef	5.48 bc	13.85 e	535.00 ef	1.03 cde	1.05 f	39.62 ef	5.15 c	12.36 e	539.44 e	0.96 ab	0.99 ef	41.72 bc
T12	6.35 cd	16.65 e	518.46 e	1.23 cde	1.22 g	38.14 de	5.33 de	13.40 f	518.30 gh	1.03 bcde	1.03 f	39.84 de	4.97 d	11.94 f	519.66 g	0.96 ab	0.96 g	41.68 bc
T13	6.52 a	17.95 a	561.70 a	1.16 h	1.10 h	36.32 i	5.71 a	15.17 a	589.40 a	0.97 i	0.96 g	37.70 i	5.47 a	13.83 a	606.04 a	0.90 f	0.92 h	39.58 h
T14	6.51 a	17.82 a	556.76 a	1.17 h	1.09 h	36.54 hi	5.68 a	14.92 b	577.68 b	0.99 hi	0.96 g	38.36 h	5.42 a	13.52 b	588.96 b	0.92 de	0.91 h	40.18 g
T15	6.48 a	17.58 b	546.86 b	1.19 g	1.09 h	36.88 h	5.68 a	14.79 b	570.08 b	1.00 gh	0.95 g	38.50 h	5.47 a	13.45 b	583.94 b	0.94 c	0.92 h	40.70 efg
T16	6.45 ab	17.32 c	538.38 c	1.20 fg	1.09 h	37.24 g	5.59 ab	14.49 c	556.92 c	1.01 fg	0.94 g	38.66 gh	5.30 b	13.10 c	565.74 c	0.94 c	0.89 i	40.54 efg

Note: Ya is the seed cotton yield (t ha⁻¹), Bio is the final aboveground biomass of cotton (t ha⁻¹), ET is the evapotranspiration of cotton throughout its growing period (mm), WP is the water productivity with respect to the crop consumptive water use (kg m⁻³), WP_Irrig is the water productivity with respect to the amount of irrigation water applied (kg m⁻³), and HI is the cotton harvest index (%). Letters after numbers indicate a significant difference at p < 0.05, according to the Duncan test.

. A box diagram of the specific trends in the three scenario years is shown in Figure S3.

The average seed cotton yield in wet, normal, and dry years was 6.35 t ha⁻¹, 5.34 t ha⁻¹, and 4.98 t ha⁻¹, respectively. When the irrigation cycle was certain, the seed cotton yields in the three scenario years increased with the increase in irrigation water. When the amount of irrigation water was certain, an irrigation period of six days resulted in a yield increase of 0.01–5.02% compared with the other treatments. When the irrigation period was constant with 525 mm of irrigation water, the yield increased by 0.72–20.18% compared with the other treatments. In years with high precipitation, for instance, when the volume of irrigation water was fixed, the biomass, evapotranspiration (ET), water use efficiency (WP), and irrigation water use efficiency (WP_Irrig) of cotton displayed a tendency to decrease with the increase in the irrigation cycle. In the case of a six-day irrigation cycle, the biomass, ET, WP, and WP_Irrig were observed to be higher than those of the other treatments by 0.73–5.61%. The biomass reached a maximum of 17.95 t ha⁻¹ in the case of an irrigation water volume of 525 mm and an irrigation period of 12 d (T13). The maximum ET was 562 mm at 525 mm, with an irrigation period of 12 d (T13). The maximum WP was 1.27 kg m⁻³ at 525 mm, with an irrigation period of 12 d (T4). The maximum WP was 1.27 kg. At 525 mm, 6 d (T4), and 6 d (T4), the maximum WP was 0.73–5.61%, 0.86–6.47%, 0.32–3.28%, and 0.17–3.32%, respectively. The maximum WP_Irrig was observed at 525 mm and 6 d (T1), with a value of 1.46 kg m⁻³.

The total costs and net revenue corresponding to the simulated treatment were calculated based on the cotton prices and local prices of agricultural production materials in 2021 and 2022, as shown in Figure 9. The blue bar chart shows the total costs in wet years; the green bar chart shows the total costs in normal years; the red bar chart shows the total costs in dry years. Each subgraph’s colored bar chart depicts the net revenue of each simulation process. It was determined that the dry years had a discernible impact on the net yield of cotton. In dry years, the cotton yield was demonstrably lower than that in wet years or normal years. Furthermore, the difference in yield among all treatments in dry years was statistically significant. When the irrigation cycle was held constant, the net revenue of cotton gradually increased with the increase in the irrigation amount in dry years. When the amount of irrigation was held constant, the net revenue of cotton displayed a decreasing trend with increasing irrigation cycle length, reaching a maximum of 2.48 × 10⁴ CNY ha⁻¹ in the T15 treatment.

3.6. A Comprehensive Evaluation of the AquaCrop Simulation Results Based on Different Irrigation Scenarios

In order to elucidate the underlying drivers of changes in seed cotton yield, correlation analyses were conducted on various agronomic and economic variables, including WP_Irrig, WP, net revenue (NR), HI, total cost (TC), total irrigation (TI), irrigation frequency (IF), final aboveground biomass (Bio), and ET. The results of these analyses are presented in Figure 10. A significant correlation was found among Ya, Bio, TI, WP_Irrig, WP, ET, and TC for cotton in the three scenario years, as well as between wet and normal years. Additionally, a significant correlation was found between seed cotton yield and Ya, Bio, TI, WP_Irrig, WP, HI, ET, and TC. Furthermore, a strong correlation was observed among Ya, Bio, TI, ET, and TC (R² > 0.88) for cotton in wet and normal years. However, efficiency indicators such as WP_Irrig, WP, and HI demonstrated a negative correlation (R² > −0.76). Figure 10 illustrates the relative importance of each factor for the yield of cotton, with TC being identified as the most significant factor, followed by Bio, ET, TI, and IF. It is noteworthy that the TI in dry years was greater than that in normal and wet years, which was likely due to the higher evaporation rates in dry years, which are also characterized by higher temperatures. The correlation analysis revealed that TI and TC displayed a highly significant positive correlation with seed cotton yield during normal and dry years. Additionally, TI and Ya demonstrated a highly significant positive correlation, while NR displayed a highly significant positive correlation with TC. It can be concluded that cotton is more dependent on TI, and ensuring the amount of water allotted to cotton during normal and dry years is crucial for stabilizing yield. In addition, the primary drivers of seed cotton yield were ET, TI, and TC. This may be attributed to the fact that seed cotton yield is the result of a process that couples both irrigation management and anthropogenic management and is influenced by a combination of physiological and environmental factors.

Principal component analysis (PCA) was employed to conduct a comprehensive evaluation of various indexes pertaining to cotton [39]. The first two components were deemed valid for the three scenarios, as their sum exceeded two-thirds of the total variance. In wet years, for instance, the first two components accounted for 97.95% of the total variation. PC1 showed a positive correlation with yield, Bio, ET, and TI, as well as a negative correlation with WP_Irrig, WP, and HI. In PC2, which accounted for 12.12% of the total variance, IF, WP_Irrig, and NR displayed a robust positive correlation, whereas TI, WP_Irrig, and TC demonstrated a negative correlation (

Table 6. Factor load and variance contribution rate according to principal component analysis.

Eigenvector	Scenario
	Wet Years		Normal Years		Dry Years
	PC1	PC2	PC1	PC2	PC1	PC2
Ya	0.98	0.07	0.99	−0.08	0.97	0.19
Bio	0.99	−0.02	1.00	0.00	1.00	0.03
ET	0.99	0.12	1.00	0.08	1.00	−0.06
TI	0.94	−0.31	0.96	−0.27	0.95	0.29
IF	0.29	0.95	0.27	0.96	0.30	−0.94
WPIrrig	−0.92	0.37	−0.90	0.38	−0.88	−0.40
WP	−0.98	−0.17	−0.90	−0.40	−0.56	0.82
HI	−1.00	0.01	−0.98	−0.18	−0.82	0.55
TC	0.99	−0.11	0.99	−0.14	0.97	0.22
NR	0.96	0.15	0.99	−0.06	0.97	0.19
Eigenvalue	8.58	1.21	8.48	1.37	7.58	2.22
Percentage of variance	85.83	12.12	84.78	13.65	75.76	22.22
Cumulative variance	85.83	97.95	84.78	98.44	75.76	97.98

Note: Ya: seed cotton yield, Bio: biomass, ET: evapotranspiration, TI: total irrigation, IF: irrigation frequency, WP_Irrig: water productivity with respect to the amount of irrigation water applied, WP: water productivity with respect to crop consumptive water use, HI: harvest index, TC: total cost, NR: net revenue.

).

It was reported that PCA was used to obtain an important relationship between the yield and water amount [40,41]. This study used the first two components (PC1 and PC2) to create biplots of wet, normal, and dry years (Figure 11), explaining the influence of different scenario years on cotton yield, efficiency, moisture, and investment indexes. Upon evaluating the positive aspects of PC1 and PC2, a significant positive correlation was identified between Ya and Bio (Figure 11), particularly in years with low precipitation; here, maintaining a sufficient amount of final aboveground biomass played a pivotal role in the formation of seed cotton yield due to the high soil evaporation. This phenomenon may be explained by the strong correlation of TC, NR, and TI in dry seasons with seed cotton yield. Due to the high evapotranspiration of the crop itself, increasing the irrigation quota and decreasing the watering cycle can minimize the negative impacts on cotton yield due to climatic extremes. In contrast, Ya, TC, and NR were strongly correlated in wet and normal years. As long as the TI meets the normal growth and development of the crop and mild stress does not occur, the higher the investment, the higher the seed cotton yield. The results of the principal component analysis were consistent with those of the correlation analysis.

Aboveground biomass is indicative of the growth of cotton. Factors influencing agricultural irrigation management, including evapotranspiration (ET), total irrigation, and the number of irrigations, directly impact the seed cotton yield. Higher seed cotton yield necessitates human inputs, prompting concern regarding the investment and net revenue of economic factors for agricultural producers. Southern Xinjiang, a region where cotton is cultivated, is water-scarce. Therefore, improving water use efficiency and the harvest index is a crucial aspect of efficient water use in arid agriculture. In this study, a raw matrix was constructed based on simulated data from the 16 treatments in the three scenario years. The matrix was constructed using 10 indexes, namely, Bio (X₁), Ya (X₂), ET (X₃), TI (X₄), IF (X₅), TC (X₆), NR (X₇), WP_Irrig (X₈), WP (X₉), and HI (X₁₀). The raw data were initially standardized to eliminate the effect of magnitude (Figure 12). Subsequent to this, the comprehensive score and ranking of each treatment were obtained by substituting the standardized data into a linear function, as shown in

Table 7. The comprehensive score of each treatment based on Ya, Bio, ET, WP_Irrig, WP, HI, TI, IF, TC, and NR.

Treatment	Scenarios									Rank
	Wet Year			Normal Year			Dry Year			Wet Year	Normal Year	Dry Year
	Score	PC1	PC2	Score	PC1	PC2	Score	PC1	PC2
T1	−0.44	−0.75	1.76	−0.39	−0.74	1.74	−0.94	−0.71	−1.73	11	11	14
T2	−1.00	−1.21	0.51	−0.89	−1.14	0.69	−1.03	−1.08	−0.88	14	14	15
T3	−1.14	−1.28	−0.16	−1.04	−1.19	−0.16	−0.91	−1.20	0.04	15	15	13
T4	−1.65	−1.76	−0.86	−1.48	−1.60	−0.73	−1.06	−1.57	0.67	16	16	16
T5	0.16	−0.03	1.52	0.14	−0.08	1.50	−0.38	−0.03	−1.58	8	7	11
T6	−0.14	−0.22	0.44	−0.28	−0.38	0.35	−0.36	−0.38	−0.31	9	10	10
T7	−0.46	−0.48	−0.34	−0.55	−0.55	−0.55	−0.32	−0.59	0.60	12	12	9
T8	−0.79	−0.77	−0.92	−0.83	−0.80	−0.98	−0.41	−0.84	1.04	13	13	12
T9	0.80	0.72	1.33	0.69	0.60	1.24	0.23	0.67	−1.25	4	4	7
T10	0.52	0.56	0.26	0.36	0.39	0.16	0.25	0.32	0.03	6	6	6
T11	0.18	0.28	−0.53	0.09	0.22	−0.72	0.30	0.11	0.96	7	8	5
T12	−0.17	−0.04	−1.11	−0.26	−0.11	−1.18	0.14	−0.16	1.18	10	9	8
T13	1.47	1.56	0.84	1.61	1.70	1.06	1.14	1.74	−0.91	1	1	2
T14	1.19	1.39	−0.23	1.20	1.41	−0.14	1.12	1.41	0.14	2	2	3
T15	0.89	1.15	−0.99	0.96	1.26	−0.90	1.19	1.26	0.94	3	3	1
T16	0.59	0.89	−1.53	0.68	1.01	−1.37	1.04	1.03	1.07	5	5	4
Linear function	w = 0.876 w1 +0.124 w2			n = 0.862 n1 + 0.139 n2			d = 0.773 d1 + 0.227 d2			T13	T13	T15

.

In the wet year, the three treatments with the highest comprehensive evaluation scores were T13, T14, and T15. In the normal year, the three treatments with the highest comprehensive evaluation scores were T13, T14, and T15. In the dry year, the three treatments with the highest comprehensive evaluation scores were T15, T13, and T14. However, the implementation of water resource regulation measures in Southern Xinjiang reached the maximum allowable precipitation of 450 mm during the critical growth period. In light of the aforementioned considerations, the cotton irrigation quota of 520 mm was treated according to T9 in both the wet and dry years in order to satisfy the water regulations and ensure stable yield. The critical period of growth was 450 mm, the irrigation cycle was 6 days/time, and the number of irrigations was 16 as per the final plan. In consideration of the stabilization of the crop yield in dry years, the final plan was to treat the cotton in accordance with T15, with an irrigation quota of 595 mm, a critical growth period of 525 mm, an irrigation cycle of 10 days/time, and 10 irrigations.

Wet year:

The first principal component was w₁ = 0.115 X₁ + 0.114 X₂ + 0.116 X₃ + 0.11 X₄ + 0.034 X₅ + 0.116 X₆ + 0.111 X₇ − 0.107 X₈ − 0.114 X₉ − 0.116 X_10.

The second principal component was w₂ = −0.019 X₁ + 0.059 X₂ + 0.099 X₃ − 0.252 X₄ + 0.783 X₅ − 0.09 X₆ + 0.122 X₇ + 0.306 X₈ − 0.136 X₉ + 0.005 X_10.

Normal year:

The first principal component was n₁ = 0.118 X₁ + 0.116 X₂ + 0.117 X₃ + 0.113 X₄ + 0.032 X₅ + 0.116 X₆ + 0.116 X₇ − 0.107 X₈ −0.106 X₉ − 0.115 X_10.

The second principal component was n₂ = 0.003 X₁ − 0.058 X₂ + 0.057 X₃ − 0.1974 X₄ + 0.703 X₅ − 0.104 X₆ − 0.043 X₇ + 0.277 X₈ − 0.293 X₉ − 0.132 X_10.

Dry year:

The first principal component was d₁ = 0.132 X₁ + 0.129 X₂ + 0.131 X₃ + 0.126 X₄ + 0.039 X₅ + 0.128 X₆ + 0.128 X₇ − 0.116 X₈ − 0.074 X₉ − 0.108 X_10.

The second principal component was d₂ = 0.012 X₁ + 0.087 X₂ − 0.025 X₃ + 0.131 X₄ − 0.423 X₅ + 0.083 X₆ + 0.099 X₇ − 0.181 X₈ + 0.368 X₉ + 0.247 X_10.

4. Discussion

4.1. Applicability of the AquaCrop Model

In this study, AquaCrop was used to simulate the soil water dynamics in the cotton growing season in Southern Xinjiang. Similarly to this study, other studies have also reported that AquaCrop has satisfactory performance in estimating soil moisture content in Gossypium hirsutum L. with different types of irrigation management [20]. AquaCrop made good predictions of soil water dynamics and crop response in all treatments of cotton; for example, the TSW was accurately simulated for the treatment with a lower irrigation limit of 65% FC. In addition, some studies have shown that AquaCrop has limitations in simulating severe water stress [20,42,43], but it also underestimated the TSW (−12.37% ≤ Pe ≤ 6.82%, 0.27 ≤ R² ≤ 0.72, 7.88 mm ≤ RMSE ≤ 27.75 mm) for the treatment with a lower irrigation limit of 75% FC (−0.34 ≤ EF ≤ 0.64, 0.60 ≤ d ≤ 0.91). Because there is only so much water in a single irrigation cycle [36], the model may have trouble simulating a wet soil surface at a lower irrigation limit of 75% FC. This is because the model has to copy the complex water movement in the surface soil under drip irrigation. Secondly, the variations in the canopy of cotton may be different from those of Gossypium hirsutum L. due to differences in the variety, which further affected the variation rule for soil water evaporation in cotton [44,45]. Ultimately, the oversimplification of the model’s internal components, including soil water balance, root distribution pattern, canopy development and senescence, and biomass production and distribution, could be the cause [20,25,44].

The model poorly simulated ET (−10.06% ≤ Pe ≤ 2.43%), and it underestimated most of the ET in most of the treatments, which was related to the use of unadjusted or low evapotranspiration and evaporation coefficients [20]. On the one hand, a possible reason is that frequent irrigation with a small amount of water causes the evaporation of water from holes in membranes to increase, which increases the error of ET_c and causes it to be underestimated. Conversely, stress only affects the canopy cover curve during the vegetative stage [46], leading to a replacement of the canopy cover curve with the classical crop coefficient curve for calculating crop transpiration and soil evaporation, thereby reducing the simulation accuracy of ET [20,45,46].

For the CC simulation results, R² (≥0.97), EF (≥0.95), and d (≥0.99) were relatively high, while RMSE (3.78%–7.08%) and Pe (−6.86%–26.17%) were relatively low. In general, the AquaCrop model was able to more accurately simulate cotton’s CC growth process. However, for the treatment with a lower limit of 55% FC at the BLS, the CC simulation was poor, indicating that the model could not fully simulate the growth of CC under severe water stress. The main reason may be that canopy senescence is accelerated under water stress; senescence occurred in the simulated CC, but the growth rate of the measured CC was slowed down when water stress occurred in a short period of time. Zhang et al. [24] found that the model could effectively simulate the CC growth of cotton under brine irrigation (0.69 ≤ R² ≤ 0.99) (4.83% ≤ RMSE ≤ 23.42%). Both Feng et al. and Ran et al. reported a similar trend. The reason may be that the CC curve in AquaCrop is not sensitive to water stress throughout the growing season but is more sensitive to water stress in the vegetative stage, resulting in lower simulated CC values in the middle and late stages [42,45]. In 2012, Tan et al. [20] highlighted the model’s limitations in simulating CC under severe water stress in a study of Gossypium hirsutum L. The simulation of CC under the condition of a 55% irrigation limit in the T4 treatment was found to be underestimated, while the CC was overestimated in this experiment. This discrepancy may be attributed to the differences in plant types. The development process of the canopy in the early and middle periods is slower than that of Gossypium hirsutum L. Additionally, the time required to reach CC_max differed from that of Gossypium hirsutum L., which affected the evapotranspiration process of soil water.

In the biomass simulation, crop transpiration was converted into biomass using a normalized water production efficiency (WP*) of 24 g m⁻², which differed from the WP* values found in prior research using AquaCrop in terrestrial cotton. This discrepancy was likely due to the potential influence of varietal differences in WP*. Zhang et al. and Wang et al. [24,47] simulated reductions in biomass production of 20 g m⁻² and 19 g m⁻² due to the WP* of salinity stress and a filmless mulching strategy, respectively. The model was used to simulate the final biomass of Gossypium hirsutum L. at the end of the fertility period with different salinity gradients and moisture stresses in filmless mulching. The results showed that Zhang overestimated the biomass, and Wang underestimated it. It is evident that different treatments can cause differences in model calibration.

In our study, the aboveground biomass in the treatment with a lower limit of irrigation of 65% FC was underestimated in the middle of the simulation and overestimated in the BS. This discrepancy may be attributed to the utilization of a constant WP* value throughout the growth period, though it should gradually decline as the plant matures. Another possible explanation is that the model underestimated the early leaf senescence due to water stress, leading to an overestimation of the canopy and transpiration. The leaves of cotton subjected to severe water stress were removed through the application of defoliants at the maturity stage. However, the treatment with a lower irrigation limit of 65% FC at the BLS extended the growth period of cotton to a certain extent, thereby mitigating the impact of defoliants. This finding is consistent with the results of Feng et al. and Steduto et al. [18,45]. According to previous studies, due to the physiological regulation and compensatory growth of cotton, mild water stress is beneficial for the growth of cotton, and the BLS is more sensitive to soil water stress than other stages [48,49]. Overall, AquaCrop is capable of adequately simulating the differences in yield and aboveground biomass quality due to varietal differences for cotton treatments under mild moisture stress.

4.2. Application of the Model to Different Irrigation Scenarios

Crop modeling and effective field irrigation decisions may address the impact of climate change on the cotton industry in Xinjiang [50,51]. In the scenario analysis, all of the common scenarios with different irrigation quotas were simulated and analyzed for different scenario years. In order to achieve optimal crop yield and irrigation water productivity, it is essential to utilize appropriate irrigation quotas in normal years to attain the highest possible yield. As evidenced by the field experiment conducted in the mulched cotton cultivation of Aksu Oasis in Southern Xinjiang, the 75% FC field irrigation method demonstrated the highest yield and water use efficiency, a result that aligned with the findings of our modeling analysis [52]. The application of AquaCrop also demonstrated that, under mulched drip irrigation, the irrigation water quota was 45 mm during the SQS, FS, and BS. When there were 11 irrigations, the yield was 6186.17 kg ha⁻¹, and the economic benefit was the largest [23].

In our scenario analysis, the irrigation system was optimized based on the critical period of cotton growth, and the simulation of different precipitation categories was realized. These appropriate water stress treatments may be effective measures to stabilize yield and improve water use efficiency. Cotton yield is related not only to varieties [32] but also to climatic factors [53,54]. Only two years of field trials served as the basis for the calibration and validation of AquaCrop due to data limitations. The model gave satisfactory estimates of the soil water dynamics, growth, and yield of cotton in normal and wet years. Nevertheless, the specific weather conditions in the single validation year may have affected the model performance and scenario analysis to some extent. In AquaCrop, too low of a soil moisture limit could seriously affect the simulation. In many previous studies, significant biases in the AquaCrop model were observed when simulating severe and extreme water stress [17,20,28,42,44]. In the scenario analysis, with the exception of 1978, 1990, and 1994, which were characterized by extreme drought, and 1959, 1973, and 1997, which were characterized by severe drought, the remaining years ranged from wet years to years with moderate drought conditions. Therefore, only under conditions of moderate aridity or humidity can we use the calibrated AquaCrop model with reliable accuracy. It is our contention that the AquaCrop model can be employed as an effective tool for providing information on the cultivation of island cotton under deficit irrigation, with the additional benefit of a supplementary study to verify the measured data of island cotton in drought years in Southern Xinjiang. Huang et al. [22] demonstrated that biochar has the potential to enhance soil texture, thereby augmenting cotton’s capacity to retain soil moisture. It is, therefore, necessary to adjust the soil texture by controlling the dosage and timing of biochar application in the future in order to improve the water use efficiency and fertilizer use efficiency of cotton in northwestern arid areas. In light of the water control and water restriction policies currently in place in Southern Xinjiang, as well as the associated economic costs, there is potential for improvement in the deficit irrigation system through the integration of an economic model.

4.3. Selection and Evaluation of Optimal Moisture Regulation for Machine-Picked Cotton in Southern Xinjiang

PCA is an unsupervised multivariate analysis method [55]. Many researchers have used PCA for data analysis [40,56,57]. For instance, Jarwar et al. [56] employed PCA to identify the Chinese Gossypium hirsutum L. varieties with the highest scores in terms of fiber and yield. The results of the principal component analysis (PCA) indicated that when combining various factors, including growth factors, yield factors, agricultural irrigation management factors, economic factors, efficiency factors, and local water resources in Southern Xinjiang, the irrigation quota for cotton in both wet and dry years was found to be 520 mm. Additionally, the critical growth period was determined to be 450 mm, the irrigation cycle was once every six days, and irrigation occurred 16 times as per the final plan. In consideration of the stabilization of crop yield in dry years, the final plan was to treat island cotton according to T15, with an irrigation quota of 595 mm, a critical growth period of 525 mm, an irrigation cycle of 10 days, and an irrigation frequency of 10 times. In addition, a strict method was used to choose a wide range of meteorological data based on the AquaCrop model in order to find the best irrigation plans for island cotton in several different scenario years. This study also underscored the significance of water regulation on the yield, benefit, and water use efficiency of island cotton in desert oasis areas of Xinjiang.

5. Conclusions

The AquaCrop model was used to simulate the growth performance of cotton (Gossypium barbadense) under mulched drip irrigation in Southern Xinjiang, China. The model was calibrated and validated at the local level based on two years of field data. When different water stress conditions were tested and compared, it was found that the model had a significant simulation effect on TSW, ET_c, CC, Bio, and Ya, with an RMSE of less than 11.58%. The TSW content was found to be overestimated in the middle and late periods of cotton growth, with the RMSE value ranging from 3.98 mm to 27.75 mm. The simulated evapotranspiration (ET_c) values were generally high, but the prediction error for each treatment was less than 11%. In the calibration year (2021), the model demonstrated satisfactory performance in simulating aboveground biomass and yield, with root mean square error (RMSE) values of 1.23 t ha⁻¹ and 0.15 t ha⁻¹, respectively. The simulation performance decreased in the verification year (2022), but the prediction error was less than 15%. AquaCrop can be a reliable tool for guiding the field management for irrigated cotton in Southern Xinjiang.

Following the verification of the dataset, which comprised 64 years of meteorological data, principal component analysis was employed to recommend cotton irrigation schemes in Southern Xinjiang, and three scenarios were presented. In wet years, the irrigation quota was 520 mm, and the critical period of growth irrigation was once every 6 days; in normal years, the irrigation quota was 520 mm, and the critical period of growth irrigation was once every 6 days. In dry years, an irrigation quota of 595 mm and irrigation once every 10 days in the critical growth period can provide higher seed cotton yield and irrigation water productivity so as to maximize cotton yield and net income in Southern Xinjiang. In addition, the model was only calibrated and validated on the basis of a two-year field trial under the weather conditions of normal and wet years. To confirm these conclusions, future confirmatory studies using long-term field data with large climatic variability will be necessary.