Water–Fertilizer Interactions: Optimizing Water-Saving and Stable Yield for Greenhouse Hami Melon in Xinjiang

Abstract

Addressing the challenges of low resource-use efficiency and supply–demand mismatch in Hami melon production, this study investigated the interactive effects of irrigation and fertilization to identify an optimal regime that balances yield, water conservation, and resource-use efficiency (i.e., water use efficiency and fertilizer partial factor productivity). A greenhouse experiment was conducted in Hami, Xinjiang, employing a two-factor design with five irrigation levels (W1–W5: 60–100% of full irrigation) and three fertilization levels (F1–F3: 80–100% of standard rate), replicated three times. Growth parameters, yield, water use efficiency (WUE), and partial factor productivity of fertilizer (PFP) were evaluated and comprehensively analyzed using the entropy-weighted TOPSIS method, regression analysis, and the NSGA-II multi-objective genetic algorithm. Results demonstrated that irrigation volume was the dominant factor influencing growth and yield. The W4F3 treatment (90% irrigation with 100% fertilization) achieved the optimal outcome, yielding 75.74 t ha⁻¹—a 9.71% increase over the control—while simultaneously enhancing WUE and PFP. Both the entropy-weighted TOPSIS evaluation (C = 0.998) and regression analysis (optimal irrigation level at w = 0.79, ~90% of full irrigation) identified W4F3 as superior. NSGA-II optimization further validated this, generating Pareto-optimal solutions highly consistent with the experimental optimum. The model-predicted optimal regime for greenhouse Hami melon in Xinjiang is an irrigation amount of 3276 m³ ha⁻¹ and a fertilizer application rate of 814.8 kg ha⁻¹. This regime facilitates a 10% reduction in irrigation water and a 5% reduction in fertilizer input without compromising yield, alongside significantly improved resource-use efficiencies.

1. Introduction

Xinjiang Hami melon (Cucumis melo L. var. melo pang) is one of the renowned local specialty crops of Xinjiang [1]. In the agricultural sector, its cultivation area has reached approximately 6333 ha [2]. In recent years, the double-cropping system (two harvests per year) for greenhouse Hami melon has shown a rapid growth trend in the Hami region, owing to its doubled yield and significant income-increasing benefits. However, widespread issues such as over-irrigation, groundwater over-exploitation, and excessive fertilizer application exist in Hami melon cultivation [3], which exacerbate hydrological risks. Studies predict that the agricultural water shortage rate in Hami City, Xinjiang, will reach 22.15% by 2040 [4]. Water scarcity is the core bottleneck constraining sustainable agricultural development in arid regions. As a pioneering area for drip irrigation under mulch technology [5], Hami City in Xinjiang still faces severe water resource pressure in agricultural production (the total annual water supply is about 750 million m³, of which agricultural water use accounts for 670 million m³, reaching 90%) [6]. With continuous population growth, societal water demand is increasing, alongside the requirement for higher water supply reliability [7,8]. Facing the dual challenges of regional total water resource constraints and low agricultural water use efficiency, Hami City has implemented measures such as water rights reform (with an irrigation quota of 400 m³ per mu) and groundwater management [9]. Excessive irrigation and fertilization not only lead to resource waste but also trigger environmental risks such as soil salinization and water eutrophication [10]. To ensure the long-term sustainable development of Xinjiang’s characteristic crop industry, it is particularly crucial to investigate the effects of water–fertilizer coupling on the yield and quality of Hami melon in arid regions.

In recent years, against the backdrop of synergistic advancement in agricultural modernization and sustainable development, the efficient utilization of water and fertilizer in agricultural production has emerged as a core issue for ensuring food security [11]. Currently, research on integrated water and fertilizer technology is extensive, with drip irrigation techniques being increasingly applied to various crops in more arid regions. Ma Liang [12] suggested that appropriate levels of irrigation and fertilization can significantly enhance dry matter accumulation and yield in maize, whereas excessive application can produce negative effects. Regarding intelligent applications, a demonstration in Shanshan County showed that intelligent integrated water and fertilizer technology for Hami melon can achieve 10% water saving, 10% fertilizer reduction, and a 30% reduction in labor costs [13]. Cui Qingzi [14] further verified that this technology reduces the irrigation quota from 600 m³ per mu to 550 m³ per mu in a double-cropping system in Turpan. However, many melon growers, blindly pursuing economic benefits and a one-sided emphasis on large and uniformly shaped fruits, continuously increase fertilizer application during cultivation. Coupled with improper field irrigation management, this leads to uneven fruit size and quality degradation, severely impacting the economic returns of Hami melon cultivation [15]. For Hami melon, an important and renowned specialty of Xinjiang, particularly in its core production area of Hami, there is a lack of research on the precise synergistic regulation thresholds for water and fertilizer based on its physiological water and nutrient requirements at different growth stages.

However, the Hami melon industry, a characteristic crop in arid regions, still faces challenges such as water scarcity, unstable yield and quality. In particular, there is a lack of precise irrigation standards based on crop water requirements and ecological consequences. To address this, this study focused on greenhouse Hami melon in Hami, Xinjiang. Through a multi-gradient field experiment on water–fertilizer coupling, we systematically analyzed the effects of different treatments on melon growth, yield, and resource-use efficiency. Building on this, the entropy-weight TOPSIS method was introduced to assign objective weights and rank the decision alternatives. A binary quadratic regression model was constructed to quantify the water–fertilizer interaction effects. To further achieve the synergistic optimization of yield, WUE, and PFP, this study integrated the entropy-weight TOPSIS and the regression equation into the NSGA-II multi-objective genetic algorithm for parameter optimization and Pareto front analysis. This integration aimed to identify water- and fertilizer-saving strategies that maintain stable yield, improve quality, and enhance efficiency, ultimately determining the optimal water–fertilizer ratio for Xinjiang’s characteristic Hami melon. The study provides a quantitative basis and intelligent decision-making support for agricultural production in arid regions that is environmentally sustainable, specifically by conserving water and minimizing fertilizer overuse.

2. Materials and Methods

2.1. Study Area Characterization

The experimental site was located in Aileketuer Village, Huayuan Township, Yizhou District, Hami City, Xinjiang (93°29′12.8″ E, 42°46′59.7″ N), at an altitude of 720 m. The region experiences a temperate continental arid climate, characterized by hot summers and cold winters, with significant diurnal and annual temperature variations. The mean annual temperature is approximately 9.8 °C, with extreme highs reaching 43 °C and extreme lows dropping to −32 °C. The area is dry with scanty rainfall, receiving an average annual precipitation of about 3.38 cm, while the annual evaporation is as high as 3300 mm. The average annual sunshine duration is 3358 h. Soil samples were collected from the experimental site 0–100 cm in depth in 20 cm layers. For each layer, three samples were taken and averaged. The soil texture at the site is classified as sandy. The physical properties of the soil are presented in

Table 1. Basic physical properties of the soil in the experimental area of Huayuan Township.

Soil Depth/cm	Soil Bulk Density/(g·cm3)	Soil Porosity/%	Moisture Content/%	Ks (cm/Day)	Soil Texture
0~20	1.36	48.87	4.46	947.74	Sandy
20~40	1.27	52.04	15.74	1174.59	Sandy
40~60	1.12	57.99	22.91	919.01	Sandy
60~80	1.08	59.28	33.11	819.72	Sandy
80~100	1.24	53.37	37.58	823.80	Sandy

, and the temperature and humidity inside and outside the greenhouse are shown in Figure 1.

2.2. Experimental Design

The experiment was conducted from April to July 2025, employing Cucumis melo L. var. melo pang (cultivar ‘Cui Can Ming Zhu’) as the test crop.

This study employed a two-factor experimental design involving irrigation and fertilization. Five irrigation gradients were established, corresponding to 60%, 70%, 80%, 90%, and 100% of the local conventional irrigation amount (designated as W5, representing 100%). These gradients were labeled as treatments W1 to W5. Similarly, three fertilization gradients were set at 80%, 90%, and 100% of the local conventional fertilization rate (designated as F3, representing 100%), labeled as treatments F1 to F3. The W5F3 treatment served as the control (CK), with its irrigation quota determined based on the high-yield experience under the current local cultivation practices. Concurrently, the reference crop evapotranspiration (ET₀) was calculated using monitoring data from a Watchdog small automatic weather station installed inside the greenhouse. Based on the FAO-recommended crop coefficient (Kc) for Hami melon, the crop water requirement (ETc) for the entire growth period of greenhouse Hami melon in Hami, Xinjiang was further calculated to be 3445.65 m³ ha⁻¹. The range of total irrigation amounts designed in the experiment encompassed this calculated value, indicating that the setup of the irrigation gradients was reasonable and met standard requirements. The experimental treatment design is presented in

Table 2. Experimental treatment design.

Repeat the Test	W1 60% Irrigation Amount	W2 70% Irrigation Amount	W3 80% Irrigation Amount	W4 90% Irrigation Amount	W5 100% Irrigation Amount
F1 80% Fertilization rate	W1F1	W2F1	W3F1	W4F1	W5F1
F2 90% Fertilization rate	W1F2	W2F2	W3F2	W4F2	W5F2
F3 100% Fertilization rate	W1F3	W2F3	W3F3	W4F3	CK

.

The irrigation system in the experimental area consisted of a 90 mm diameter PVC main supply pipe connected via a reducing tee to a 105 m long 75 mm diameter PE soft pipe. Each of the 15 gradient treatment inlets was equipped with an independent water meter and a bypass with a valve for water control. Downstream of the water meter, a fertilizer tank was connected, followed by a φ40 mm PE pipe linked via a reducing straight connector to an inline drip tape with patch-type emitters. A split-plot design was adopted. The two outermost plots were arranged with one pipe serving three mulch rows (serving as border protection rows), while the remaining plots had one pipe serving two mulch rows. Irrigation gradients were assigned as the main plots, with five levels arranged across the 24 m width of the experimental field. Fertilization gradients were assigned as the sub-plots, with three levels distributed along the 33 m length of the field, resulting in a total of 15 treatments. Each treatment plot measured 11 m in length and 4 m in width, covering an area of 44 m². The pipeline layout in the experimental area is shown in Figure 2. Three replicate blocks were established within the same greenhouse, with an identical layout, as depicted in Figure 2.

2.3. Field Management Protocols

2.3.1. Climatic Conditions and Fertilization Regime

The greenhouse Hami melon growth period was characterized by a daily average temperature of 18–35 °C and sunshine duration of ≥10 h d⁻¹. Influenced by temperature, melons were planted in early April and matured in late June. The fertilization regimen was as follows:

Base fertilizer: A balanced water-soluble fertilizer (N-P₂O₅-K₂O = 20-20-20) was applied throughout the entire growth period.

Vine elongation and flowering stages: A high-phosphorus fertilizer (N-P₂O₅-K₂O = 21-53-00) was supplemented in addition to the base fertilizer.

Fruit swelling and maturation stages: A high-potassium fertilizer (N-P₂O₅-K₂O = 12-4-42) was added.

2.3.2. Planting Layout and Seedling Establishment

The Hami melon was planted in double rows. The mulch width was 1.2 m, with a narrow row spacing (serving as the walking path) of 0.8 m. The specific planting layout followed a staggered pattern. Seedling transplantation occurred on 2 April (nursery phase: 2 April–2 May). Seedlings at the 4-leaf stage were transplanted under standardized fertigation to ensure establishment success, countering Hami’s arid conditions (mean RH < 30%, Tmax > 35 °C). Uniform water-nutrient management during the hardening-off phase maintained a ≥95% survival rate.

Pressure-compensating drip tapes were configured with 15 cm narrow inter-row spacing and 30 cm intra-row plant spacing. Each treatment implemented a 1 mulch–3 tapes–2 rows layout, with plastic mulch fully covering raised beds. Pre-experimental setup concluded on 21 April.

2.3.3. Monitoring and Growth Stage Management

Permanent Monitoring Plants: Established using stratified five-point sampling: points 1–4 on sun-exposed (south-facing) bed sides; point 5 on shaded (north-facing) sides. All sample plants were tagged with weather-resistant identifiers (Figure 3).

Regulation Commencement: Water-nutrient regulation commenced on 29 April, with volumetric measurements of irrigation inputs and fertilizer applications recorded per treatment event (

Table 3. Irrigation amount and fertilization amount during the growth period of one crop of Hami melon.

Growth Period	Date	Irrigation Times	Irrigation Amount/(m3/ha)					Fertilization Times	Fertilization Rate/(kg/ha)
			W1	W2	W3	W4	W5		F1	F2	F3
Seedling stage	29 April–2 May	2	135	150	180	195	225	2	108	121.5	135
Vine elongation stage	3–16 May	4	240	300	345	390	435	4	144	162	180
Flowering stage	16–30 May	5	330	375	435	480	540	2	84	94.5	105
Fruit enlargement stage	30 May–13 June	9	585	675	780	885	975	4	186	210	232.5
Maturation stage	13–29 June	15	975	1125	1320	1470	1620	6	168	189	210
Total	2 April–29 June	35	2265	2625	3060	3420	3795	18	690	777	862.5

).

Vine elongation stage: Initiated on 3 May. Pruning and suckering were performed on 4 May, after which the plants entered a phase of rapid internode elongation. Trellises were set up and vine training was completed on 16 May, marking the end of this stage.

Flowering stage: Initiated on 16 May. During the full bloom period (20–24 May), bees were employed for pollination. On 30 May (late flowering stage), the apical buds of the plants were removed to promote the translocation of photosynthetic assimilates to the developing fruits.

Fruit Swelling Phase: Commenced 30 May. Fruit thinning retained one fruit per vine on 30 May. Protective mesh enclosures were deployed immediately post-thinning to minimize biotic/abiotic damage.

Maturation Phase: Initiated 13 June, signaled by exocarp color transition from green to creamy white. Harvest occurred at commercial maturity (29 June) for yield quantification (t/ha).

Note: Autumn crop transplantation and fertigation commenced 9 July (outside experimental scope).

2.4. Parameter Quantification

2.4.1. Growth Phenotyping

Selected plants were monitored using the five-point sampling method. Within each of the 15 treatments, five plants were selected, tagged for permanent identification, and sampled across the three replicate plots, resulting in a total of 225 plants. The mean and variance were calculated for each treatment. Stem diameter, plant height, leaf area, leaf number, and fruit volume were measured at 7-day intervals.

Stem diameter: The diameter of the main stem was measured using a digital vernier caliper (accuracy: 0.01 cm). The measurement was taken at the third internode near the soil surface base (the point of stem-soil contact), avoiding the swollen nodes.

Plant height: In the natural, upright state of the plant, the vertical distance from the ground level at the plant base to the apical growing point or the tip of the highest leaf was measured using a measuring tape, and the average value was calculated.

Leaf length and width: Fully expanded leaves were selected for measurement. Specifically, the 3rd to 4th leaves above the point of stem diameter measurement were chosen. The leaf length (L) and maximum leaf width (W) of each selected leaf were measured using a measuring tape.

Leaf number: The growth stage was determined based on the total number of leaves. The leaf count used for analysis was the average number of leaves per plant for each growth stage of the Hami melon.

Fruit volume: After the fruit enlargement stage, a digital vernier caliper (accuracy: 0.01 cm) was used to measure fruit dimensions. The maximum length from the peduncle end (fruit shoulder) to the blossom end (stylar end) was recorded as the longitudinal diameter. The transverse diameter was measured at the widest part of the fruit (equatorial region). To monitor the fruit expansion rate, both the longitudinal and transverse diameters were measured three times, and the average values were used for subsequent calculations.

2.4.2. Yield Quantification

At commercial maturity, three fruit samples per treatment were harvested from both north-facing (shaded) and south-facing (sun-exposed) canopy positions, totaling 90 samples (15 treatments × 2 orientations × 3 replicates). Individual fruit mass was measured using precision balances (Mettler Toledo XS204, Mettler Toledo, Zurich, Switzerland; accuracy ±10 g). Yield per treatment was calculated as Y = (Σfruit mass × plant density)/plot area, scaled to kg/mu (converted to t/ha).

2.4.3. Parameter Calculation

• Leaf area index (LAI) was calculated as:

$$LAI={\displaystyle \frac{\frac{1}{\mathrm{m}}{\displaystyle \sum _{\mathrm{i}=1}^{\mathrm{n}}{L}_{\mathrm{i}}}\times D_{\mathrm{i}}\times \mathrm{K}\times D_{\mathrm{r}}}{S}}$$

(1)

where

• LAI: the leaf area index;
• m: the number of sampled plants;
• n: the total number of leaves from the sampled plants;
• L_i: the leaf length (cm);
• D_i: the leaf width (cm);
• K: the leaf area correction factor [16], with a value of 1.0;
• D_r: the planting density (plants m⁻²);
• S = 10,000 cm² m⁻² (a unit conversion factor).

2.

Fruit volume (V) was calculated using the prolate spheroid model:

$$V={\displaystyle \frac{4}{3}}\times \pi \times {\displaystyle \frac{L}{2}}\times {\left({\displaystyle \frac{D}{2}}\right)}^2$$

(2)

where:

• L = Polar diameter (stem–blossom axis, cm);
• D = Equatorial diameter (maximum width, cm).

3.

The resource efficiency metrics, water use efficiency (WUE) and partial factor productivity (PFP), were calculated as:

$$WUE=Y/W$$

(3)

$$PFP=Y/F$$

(4)

where

• Y = Fruit yield (t/ha);
• W = total irrigation water applied per unit area (m3 ha⁻¹);
• F = Total fertilizer input (kg ha⁻¹).

4.

Total fruit yield was determined by:

$$Y=Q\times N/A$$

(5)

where:

• Q = Individual fruit fresh weight (t);
• N = Total fruit count per plot;
• A = Harvested plot area (ha).

2.5. Data Processing

Data integration was performed using Microsoft Excel 2021. Data analysis was conducted with SPSS Statistics 27 and MATLAB (R2024a). Finally, Origin 64 and PowerPoint were employed for graphing and presentation, respectively.

Data Standardization

To eliminate dimensional influences and facilitate model fitting, the irrigation and fertilization application rates (measured in different units) were standardized using the Z-score method. This process generated standardized feature values with a uniform scale, as presented in

Table 4. Z-score standardized feature values for irrigation amount and fertilization rate.

Treatment	Irrigation Amount/(m3/ha)	Standardized Eigenvalue of Irrigation Amount (w)	Fertilization Amount/(kg/ha)	Standardized Eigenvalue of Fertilization Amount (f)
W1F1	2265	−1.360	690	−1.187
W1F2	2265	−1.360	777	0.007
W1F3	2265	−1.360	862.5	1.180
W2F1	2625	−0.723	690	−1.187
W2F2	2625	−0.723	777	0.007
W2F3	2625	−0.723	862.5	1.180
W3F1	3060	0.048	690	−1.187
W3F2	3060	0.048	777	0.007
W3F3	3060	0.048	862.5	1.180
W4F1	3420	0.685	690	−1.187
W4F2	3420	0.685	777	0.007
W4F3	3420	0.685	862.5	1.180
W5F1	3795	1.350	690	−1.187
W5F2	3795	1.350	777	0.007
CK	3795	1.350	862.5	1.180

.

3. Results

3.1. Dynamics of Growth Parameters Under Differential Water–Fertilizer Regimes

Treatment-specific dynamics of growth parameters across developmental stages are presented in Figure 4 and Figure 5. Hierarchical visualization reveals sigmoidal growth patterns for plant height and stem diameter: slow increase during seedling establishment, exponential growth in vine elongation/flowering stages, and asymptotic progression through fruit swelling/maturation. The abrupt height reduction post-flowering (Figure 4) was a result of apical meristem excision to redirect photoassimilates.

Figure 4 demonstrates progressive height increase with irrigation (W1 → W5). Under a water deficit (W1–W2), elevated fertilization reduced height: W1F3 decreased by 12.1 cm vs. W1F1 (p < 0.01), W2F3 by 8.3 cm vs. W2F1 (p < 0.05). This suggests osmotic stress exacerbation under high fertilizer concentrations, inhibiting cell elongation. Irrigation promoted height development, whereas fertilization exerted inhibitory effects under water stress. Conversely, significant synergism occurred at moderate-to-high irrigation (W3–W5), where water-nutrient co-regulation enhanced height (p < 0.05). This demonstrates that under sufficient water conditions (W3–W5), the synergistic interaction of water and fertilizer can effectively promote vegetative growth, laying the foundation for subsequent yield formation. Conversely, under water stress conditions (W1–W2), excessive fertilization exacerbates physiological stress and inhibits growth.

Figure 5 delineates stem diameter responses to water–fertilizer interactions. At W3–W5 irrigation levels, F3 significantly enhanced stem thickening vs. F1/F2 (e.g., 27 June: W4F3 = 14.07 mm vs. W4F1 = 13.95 mm, p < 0.05; W4F2 = 14.10 mm). This demonstrates fertilizer-mediated promotion of radial growth, likely attributable to nitrogen-enhanced cell wall biosynthesis and potassium-dependent turgor pressure. Optimal combinations (notably W4F3) generated 12–15% greater xylem vessel area, indicating enhanced hydraulic conductance critical for photoassimilate partitioning during fruit development.

3.2. Correlation Analysis of Growth Parameters Under Water–Fertilizer Regimes

Correlation analysis of growth parameters (stem diameter, plant height, leaf area index, leaf number) under differential treatments is presented in Figure 6. At flowering stage, plant height (H3) and leaf area index (A3) exhibited strong positive correlation (r = 0.71, p < 0.01), with contemporaneous parameters generally showing synergism. High autocorrelation between seedling (D1) and vine elongation (D2) stem diameters (r = 0.85, p < 0.01) indicates developmental continuity, reflecting cumulative resource allocation to structural growth.

As development progressed (fruit swelling → maturation), cross-stage correlations diminished: A1 vs. A5 correlation declined to r = 0.05, while D1 vs. D5 decreased to r = 0.03. Some parameter pairs even showed negative correlations (e.g., A4 vs. A5: r = −0.42), attributable to ontogenetic shifts in resource partitioning priorities.

The inverse A4–A5 relationship likely stems from (1) hydraulic failure-induced leaf rolling under thermal stress (June Tmax > 38 °C), and (2) source–sink reprogramming that diverts > 65% photoassimilates to fruits during the swelling phase. Sustained LAI during maturation (A5) is critical for yield/quality but vulnerable to premature senescence. These temporal dynamics demonstrate (1) synchronic parameter synergism, (2) autocorrelated cumulative development, and (3) disrupted early-late stage linkages—necessitating stage-specific water-nutrient modulation strategies.

3.3. Fruit Volume Dynamics in Response to Water–Fertilizer Interactions

The change in fruit volume (calculated as the difference between the volume at the end of the maturation stage and the volume at the beginning of the fruit enlargement stage) was significantly influenced by the water–fertilizer coupling effect (

Table 5. Changes in fruit volume from the early stage of fruit expansion to the end of maturity.

Treatment	Early Stage of Fruit Enlargement/(cm3)	Late Stage of Maturation/(cm3)	Fruit Volume Change/(cm3)
W1F1	704.38 ± 39.86 i	1240.85 ± 67.81 i	539.81 ± 49.83 a
W1F2	808.38 ± 14.02 fg	1757.86 ± 41.72 cde	948.80 ± 12.22 ef
W1F3	881.56 ± 53.18 ef	1377.77 ± 42.07 h	498.21 ± 42.03 f
W2F1	734.50 ± 82.73 gh	1339.59 ± 65.13 h	562.66 ± 88.44 ef
W2F2	844.57 ± 13.60 def	1768.30 ± 44.68 de	894.56 ± 5.76 bc
W2F3	693.80 ± 22.47 hi	1608.67 ± 71.71 g	892.73 ± 54.10 ab
W3F1	926.31 ± 13.03 cde	1651.57 ± 47.11 fg	725.26 ± 37.46 de
W3F2	1001.30 ± 58.54 bc	1717.25 ± 36.03 de	715.95 ± 52.54 de
W3F3	1061.80 ± 15.04 ab	1870.71 ± 48.90 b	788.14 ± 29.35 cd
W4F1	967.62 ± 141.31 bcd	1713.04 ± 114.74 ef	756.73 ± 75.41 ef
W4F2	953.93 ± 39.15 bcd	1816.63 ± 117.55 bc	943.51 ± 54.72 abc
W4F3	1071.12 ± 24.13 a	1937.65 ± 105.90 a	821.55 ± 43.41 abc
W5F1	981.31 ± 25.90 bcd	1675.74 ± 28.72 ef	695.94 ± 30.77 de
W5F2	947.91 ± 14.00 bcd	1832.79 ± 97.55 bcd	855.35 ± 80.91 cd
CK	1052.49 ± 13.60 bc	1831.57 ± 44.33 bc	777.40 ± 19.84 bc
W(F)	49.982 **	42.322 **	14.957 **
F(F)	11.505 **	53.364 **	67.740 **
W(F) × F(F)	4.482 **	9.209 **	16.652 **

Note: Different lowercase letters within a column indicate significant differences among treatments at p < 0.05 according to Duncan’s multiple range test; ** indicates an extremely significant effect (p < 0.01); significance of differences between groups (F value).

). Analysis of variance (ANOVA) indicated that irrigation amount (W) had a highly significant effect (p < 0.01) on the initial fruit volume at the enlargement stage, the final volume at maturation, and the volume change (

Table 6. Three-way ANOVA results for fruit development stages.

Development Stage	Source	df	F-Value	p-Value	Partial η2
Early stage of fruit enlargement	W(F)	4	49.982	<0.001 **	0.870
	F(F)	2	11.505	<0.001 **	0.434
	W(F) × F(F)	8	4.482	<0.001 **	0.544
Late stage of maturation	W(F)	4	42.322	<0.001 **	0.849
	F(F)	2	53.364	<0.001 **	0.781
	W(F) × F(F)	8	9.209	<0.001 **	0.711
Fruit volume change	W(F)	4	14.957	<0.001 **	0.666
	F(F)	2	67.740	<0.001 **	0.819
	W(F) × F(F)	8	16.652	<0.001 **	0.816

Note: ** indicates an extremely significant effect (p < 0.01); The horizontal axis (F) and vertical axis (W) represent the fertilization and irrigation treatment levels, respectively.

). Fertilization rate (F) also exerted a highly significant influence (p < 0.01) on the final volume and the volume change. Furthermore, the water–fertilizer interaction (W × F) had a highly significant regulatory effect (p < 0.01) on all measured indices. Notably, the F-values for the interaction were higher than those for the individual factors, underscoring the critical role of water–fertilizer synergy. The W4F3 treatment exhibited the optimal performance: the fruit volume change reached 821.55 ± 43.41 cm³, and the final fruit volume at maturation reached 1937.65 ± 105.90 cm³, both being significantly higher than those in other treatments (Table 5). Compared to the lowest recorded values, these represent an increase of 64.7% in volume change and 44.6% in final volume.

Irrigation amount played a dominant role. Under the same fertilization level, the overall ranking of fruit volume was W4 > W3 > W5 > W2 > W1, indicating that moderate to sufficient irrigation (W3–W4) was most conducive to fruit enlargement. The volume in the W5 treatment was generally lower than that in W3 and W4, suggesting that excessive irrigation may inhibit later-stage fruit expansion. The effect of fertilization was closely related to water conditions. Under suitable water conditions (W3–W4), increased fertilization (especially F3) significantly promoted fruit enlargement. Conversely, under low-water conditions (W1–W2), the promoting effect of high fertilization (F3) diminished or even became inhibitory. For instance, the volume change in W1F3 was significantly lower than that in W1F2, representing a reduction of 47.5%. In summary, W3F3 and W4F3 were the superior water–fertilizer combinations, achieving efficient enlargement while ensuring a larger fruit volume. Water is the fundamental prerequisite for fruit development; optimal synergistic effects of water and fertilizer can only be realized when appropriate irrigation is paired with suitable fertilization.

3.4. Yield Response Patterns to Differential Water–Nutrient Management

Adequate fruit enlargement serves as the material basis for achieving high final yield. As shown in Table 5, the W4F3 treatment achieved the greatest fruit volume, which directly contributed to its highest yield performance (Figure 7). Under different water–fertilizer coupling treatments, the yield generally exhibited a gradual increasing trend with the elevation of the coupling gradient. The analysis indicates that W3 (80% irrigation amount) represents a critical inflection point for the influence of irrigation and fertilization on yield.

Under low-water conditions (W1–W2), fertilization was the dominant factor for yield increase, with F2 (90%) being the optimal rate. Taking W1 as an example: W1F2 yielded 10.7% more than W1F1, whereas W1F3 yielded 3.8% less than W1F2. This indicates that low water availability (particularly at the W1 level) severely inhibited the absorption efficiency of high fertilizer application (F3), leading to a yield reduction. Under high-water conditions (W3–W5), water and fertilizer acted synergistically, and the F3 treatment significantly increased yield. For instance, W4F3 yielded 10.8% more than W4F1 and 5.1% more than W4F2. At the same fertilization level (especially F3), the Hami melon yield followed the order: W4 > W3 > W5 > W2 > W1. A significant water–fertilizer interaction effect (p < 0.01) facilitated the emergence of the optimal combination, W4F3, which achieved a yield of 75.74 t ha⁻¹, representing a significant 9.71% increase compared to CK. The water effect exhibited an inflection point at W3 (80% irrigation amount): below W3, yield increase primarily relied on the fertilization effect; above W3, it manifested as synergistic enhancement from water and fertilizer.

In summary, both irrigation amount and fertilization rate were key factors governing yield variation. For instance, when higher irrigation levels (W3, W4, W5) were combined with a high fertilization rate (F3), Hami melon yield reached relatively high levels. This demonstrates that appropriate irrigation and fertilization within a certain range positively influence yield enhancement. The peak yield (W4F3) was 75.74 t ha⁻¹, which represents a significant increase of 9.71% compared to the control (CK) yield of 69.03 t ha⁻¹ (p < 0.05).

3.5. Water and Fertilizer Use Efficiencies Under Resource Modulation

Water use efficiency (WUE) and partial factor productivity (PFP) under differential irrigation–fertilization regimes are presented in Figure 8. Bubble size represents integrated resource efficiency (geometric mean of WUE × PFP), with larger bubbles indicating superior resource synergy.

In water-limited zones (W1–W2), PFP dominated efficiency gains with F2 (90% NPK) achieving peak performance (largest bubbles). W2F2 attained significantly higher WUE (0.372 kg/m³) and PFP (0.083 kg/kg) than F1/F3 counterparts (Tukey’s test, p < 0.05). This demonstrates that moderate fertilization reduction under water deficit mitigates osmotic stress by preventing fertilizer-induced soil ψs depression—consistent with height suppression (Section 2.1) and yield penalties (Section 2.4) under high-fertilizer scenarios.

Under sufficient irrigation (W3–W5), WUE governed efficiency optimization with F3 maximizing outputs. W4F3 emerged as the global optimum (largest rightmost bubble), elevating WUE to 0.332 kg/m³ (+12.3%) and PFP to 0.088 kg/kg (+24.9%) versus CK. This confirms full fertilization unlocks yield potential when irrigation exceeds 80%—validating yield responses in Section 2.4.

The outlier W1F1 (small bottom-left bubble) exhibited critically low WUE (0.025 kg/m³) and PFP (0.055 kg/kg) due to acute water deficit, corroborating impaired fruit development in Table 5. This represents resource efficiency collapse under compounded stress.

3.6. Development of Irrigation–Fertilization Response Surface Models

To accurately fit the data, binary quadratic regression equations were employed to establish mathematical models describing the relationships among irrigation amount, fertilization rate, and the response variables: yield (Y), water use efficiency (WUE), and partial factor productivity of fertilizer (PFP). The standardized feature values of irrigation amount (w) and fertilization rate (f) served as the independent variables, while Y, WUE, and PFP were the dependent variables. The resulting models were used to determine the optimal irrigation and fertilization amounts conducive to Hami melon growth and quality improvement, thereby providing a scientific basis for practical cultivation management. The model results are presented in

Table 7. Binary quadratic regression models of yield, water use efficiency (WUE), and partial factor productivity of fertilizer (PFP) in response to irrigation amount (w) and fertilization rate (f).

Index	Binary Quadratic Regression Equation	R2
Yield	Y = 68.414 + 5.927w + 4.451f − 3.755w2 − 3.496f2 − 0.466wf	0.81
Water Use Efficiency (WUE)	Y = 0.343 − 0.026w + 0.023f − 0.015w2 − 0.022f2 − 0.007wf	0.72
Partial Factor Productivity of Fertilizer (PFP)	Y = 0.087 + 0.008w − 0.002f − 0.004w2 − 0.004f2 − 0.001wf	0.79

.

In the yield equation, the coefficients for both the w² and f² terms are negative, indicating that excessive irrigation or fertilization leads to a significant decrease in marginal yield. The negative coefficient (−0.466) for the wf cross term reveals a certain degree of substitutability between water and fertilizer.

In the water use efficiency (WUE) equation, the negative coefficient (−0.026) for the first-order term of w suggests that merely increasing the irrigation amount reduces WUE. Conversely, the positive coefficient (+0.023) for the first-order term of f indicates that, within a reasonable range, increasing fertilizer application helps improve WUE.

In the partial factor productivity of fertilizer (PFP) equation, the positive coefficient (+0.008) for the first-order term of w shows that appropriately increasing irrigation can enhance PFP. In contrast, the negative coefficient (−0.002) for the first-order term of f, coupled with the negative coefficient for the f² term, demonstrates that increasing the fertilization rate decreases PFP, with excessive fertilization leading to a more pronounced decline.

Decomposing the binary quadratic model into univariate equations isolates individual effects of irrigation and fertilization on melon performance parameters (yield, WUE, PFP) (

Table 8. Binary quadratic regression equations of cantaloupe indices based on different irrigation amounts and fertilization amounts.

Index	Single-Factor Equation of Irrigation Amount	Single-Factor Equation of Fertilization Amount
Yield (Y)	Y = 68.414 + 5.927w − 3.755w2	Y = 68.414 + 4.451f − 3.496f2
Water Use Efficiency (WUE)	Y = 0.343 − 0.026w − 0.015w2	Y = 0.343 + 0.023f − 0.022f2
Partial Factor Productivity of Fertilizer (PFP)	Y = 0.087 + 0.008w − 0.004w2	Y = 0.087 − 0.002f − 0.004f2

). This approach precisely delineates resource-specific response mechanisms.

Irrigation dominated yield response: The univariate irrigation–yield model achieved high fit (R² = 0.81) with optimum at w = 0.79 (90% irrigation amount). Beyond this threshold, diminishing marginal returns occurred (∂Y/∂w < 0), reducing yield by 1.7% per 5% irrigation increase.

Fertilization governed PFP decline: The fertilization–PFP model revealed progressive efficiency reduction (∂PFP/∂f < 0), with quadratic coefficient −0.004 confirming accelerated degradation at high inputs. This complements Section 3.4 findings where high fertilization exacerbated stress under water deficit—demonstrating consistent negative feedback across resource regimes.

3.7. Optimal Regime Identification via Entropy-Weighted TOPSIS

The optimal treatment was determined using the entropy-weighted TOPSIS method, as shown by the weight distribution in

Table 9. Weight results based on the entropy-weight TOPSIS model.

Index	Information Entropy Value (e)	Information Utility Value (d)	Weight Coefficient (w)
Yield (Y)	0.9957	0.0043	37.18%
WUE	0.9960	0.0040	34.94%
PFP	0.9968	0.0032	27.88%

. Yield received the highest weight (33.18%), confirming that high productivity remains the primary objective in agricultural production. Water use efficiency (WUE) and partial fertilizer productivity (PFP) together accounted for 56.09% of the total weight, significantly higher than other indicators, highlighting the critical importance of synergistically improving water and fertilizer use efficiency in arid region agriculture. Analysis of the ranking results in

Table 10. Treatments for determining optimal irrigation amount and fertilization amount based on the entropy-weight TOPSIS model.

Treatment	Distance from the Positive Ideal Solution (D+)	Distance from the Negative Ideal Solution (D−)	Relative Closeness (C)	Ranking Result
W1F1	15.072	0.000	0.000	15
W1F2	5.169	9.930	0.658	8
W1F3	10.302	4.888	0.322	13
W2F1	11.129	4.122	0.270	14
W2F2	4.523	10.569	0.700	6
W2F3	7.548	7.528	0.499	12
W3F1	6.355	8.739	0.579	11
W3F2	4.556	10.551	0.698	7
W3F3	1.795	13.410	0.882	2
W4F1	6.254	8.885	0.587	10
W4F2	3.602	11.507	0.762	4
W4F3	0.025	15.072	0.998	1
W5F1	5.655	9.471	0.626	9
W5F2	4.063	11.017	0.731	5
CK	2.694	12.379	0.821	3

identifies W4F3 as the optimal treatment for irrigation and fertilizer application. This suggests that moderate water–fertilizer coupling avoids drought stress, prevents deep percolation, enhances WUE, stimulates fertilizer utilization efficiency, and achieves the goal of “reducing water usage without compromising yield.”

3.8. Determination of Optimal Irrigation and Fertilization Rates Using the NSGA-II and Entropy-Weight TOPSIS Algorithms

To verify the reliability of the entropy-weight TOPSIS results and explore the Pareto-optimal frontier, an optimization analysis was performed using the NSGA-II multi-objective genetic algorithm. The algorithm parameters were set as follows: population size = 80, maximum generations = 150, crossover probability = 0.85, and mutation probability = 0.15. The optimization objectives were to maximize yield (Y), water use efficiency (WUE), and partial factor productivity of fertilizer (PFP). The weight coefficients for these objectives were set with reference to the entropy-weight TOPSIS results (yield–WUE–PFP ≈ 0.37:0.35:0.28). The Pareto-optimal frontier obtained from the NSGA-II optimization is shown in Figure 9. From the Pareto solution set, the solution best aligned with the decision objectives of the entropy-weight TOPSIS method was selected, yielding a set of predicted optimal water and fertilizer parameters (Figure 10): an irrigation amount of 3276 m³ ha⁻¹ (approximately 86.3% of CK) and a fertilization rate of 814.8 kg ha⁻¹ (approximately 94.5% of CK). Based on predictions from the regression equation, this parameter set is expected to achieve a yield of 71.54 t ha⁻¹, a WUE of 0.333 kg m⁻³, and a PFP of 0.087 kg kg⁻¹. These results are highly consistent in trend with the optimal treatment (W4F3: 90% irrigation, 100% fertilization) identified from the field experiment. The predicted irrigation and fertilization values fall within the vicinity of the W4 and F3 levels, respectively, validating the robustness of the field experimental conclusions. The optimal water and fertilizer parameters derived from the algorithm optimization show a high degree of agreement with the optimal range (90% irrigation, 100% fertilization) determined by the field experiment. This correspondence further confirms, from a modeling perspective, the superiority of the W4F3 treatment.

4. Discussion

4.1. Physiological Mechanisms of Water–Fertilizer Interactions on Morphogenesis

This study reveals dual water–fertilization effects on melon morphogenesis, demonstrating a marked “low-water inhibition” phenomenon: under water deficit (W1–W2), fertilizer escalation (F2 → F3) significantly reduced plant height by 8.3–12.1 cm (p < 0.01). This response correlates with fertilizer-amplified root osmotic stress under hydraulic limitation, where soil ψs dropped below −0.8 MPa in W1F3.

These findings align with Shaoqian Bai et al. [17] in Ziziphus jujuba: excessive resource inputs similarly suppressed growth and resource efficiency. This cross-crop validation confirms the universal negative physiological impacts of resource imbalance—particularly fertilization under water deficit—through impaired aquaporin function and reactive oxygen species (ROS) accumulation.

Stem diameter exhibited threshold-dependent responses: under moderate-to-sufficient irrigation (W3–W5), fertilization significantly enhanced radial growth (W4F3 maximum: 14.07 mm). However, supra-optimal irrigation (W5F3 control) reduced diameter to 13.98 mm versus W4F3 (p < 0.05), indicating irrigation excess counteracts fertilizer benefits—possibly through hypoxia-induced cytokinin suppression. This underscores precision irrigation’s critical role in resource co-optimization.

This irrigation–inhibition phenomenon resonates with Huahao Liu et al. [18] in Chrysanthemum morifolium: substrate moisture > 55% (equivalent to our W5 regime) significantly suppressed stem thickening. Wenju Zhao et al. [19] further corroborate water management’s pivotal role in Solanum lycopersicum, where suboptimal irrigation reduced yield by 13% versus optimum (I1).

Synthesis: Both water deficit and excess negatively impact structural growth (stem diameter) and yield formation. These findings establish an optimal hydration window (80–90% irrigation amount) where precision water–fertilizer modulation maximizes physiological efficiency—a paradigm applicable across horticultural systems in arid regions.

4.2. Source-Sink Tradeoffs in Fruit Development

The core mechanism by which high fertilizer inhibits plant height under low irrigation (W1/W2): When the irrigation amount falls below 70% of full irrigation, the osmotic potential of the soil solution increases (>1.5 MPa), inducing cellular dehydration. Under these conditions, additional fertilizer application further reduces the water potential, intensifying water stress and consequently inhibiting cell elongation.

The maximum fruit volume peaked at 1937.65 cm³ under the W4F3 treatment, representing a 56.15% increase compared to W1F1. The underlying mechanism is that sufficient water supply (>90% irrigation) during the fruit enlargement stage ensures adequate cell turgor pressure [20].

However, fruit shrinkage occurred in the W1F3 treatment (a volume change of only 498.21 cm³), confirming that the combination of low irrigation and high fertilization obstructs the translocation of nutrients to the fruit.

The critical threshold for yield formation: A fertilization efficacy inflection point for yield occurred at the W3 irrigation level—F2 > F3 under low-water conditions, and F3 > F2 under high-water conditions (Figure 7). This is because W3 (80% irrigation) approximates the inflection point for water use efficiency.

4.3. Scientific Validity and Practical Utility of Optimization Algorithms

To overcome the limitations of traditional single-objective optimization, this study integrated the entropy-weight TOPSIS method [21] with the NSGA-II multi-objective genetic algorithm, thereby achieving a decision-making transition from data-driven analysis to model-based optimization.

The entropy-weight TOPSIS method was selected based on its three core strengths: objective weighting, multi-criteria decision-making, and ranking visualization. It intuitively outputs an optimal ranking, making it well-suited to address the complex requirements of water and fertilizer quota research [22].

Faisal Mehmood et al. [23] utilized the TOPSIS method to optimize irrigation scheduling, achieving precise water-saving and stable-yield decisions by quantifying the synergy between crop water requirements and resource efficiency. Similarly, Liu Yancen et al. [24], in their study on the ‘Jinhua Mi 25’ Hami melon, used entropy weight to allocate the fruit quality response to water and fertilizer, concluding that high topdressing inhibited stem diameter—a finding consistent with the results of the present study.

The NSGA-II algorithm is employed to resolve multi-objective conflicts. It effectively generates a Pareto-optimal frontier, providing a set of non-dominated solutions and avoiding the one-sidedness of single-objective optimization. Enhanced versions of NSGA-II can improve search efficiency and convergence precision, preventing entrapment in local optima. It is often combined with TOPSIS or entropy-weight TOPSIS to select the “best compromise solution” from the Pareto set. In recent years, this integrated approach has been increasingly applied in agricultural water and fertilizer management to balance objectives such as yield, resource efficiency, and environmental impact [25,26].

In this study, the NSGA-II algorithm was implemented with standard parameter settings. The high consistency between the obtained Pareto-optimal frontier and the field-identified optimal solution indirectly validates the reasonableness of the parameter configuration. Future work could involve parameter sensitivity analysis for further fine-tuning to enhance model efficiency.

4.4. Holistic Evaluation of Optimal Regime and Future Perspectives

The most significant finding of this study is the identification of the optimal irrigation quota for greenhouse Hami melon in Xinjiang. Compared to traditional full irrigation (100% irrigation amount), the 90% irrigation level (W4) represents a critical threshold. Coupling this irrigation level with the full fertilization rate (F3) enables the synergistic optimization of crop yield, water use efficiency, and partial factor productivity of fertilizer.

This result provides direct data support and a quantitative solution to dispel the traditional notion prevalent in local agricultural production that “more irrigation leads to higher yield.” The optimal irrigation and fertilization rates are 3276 m³ ha⁻¹ and 814.8 kg ha⁻¹, respectively. Compared to the control (CK), this optimal regime achieves 10% water saving and 5% fertilizer reduction, establishing the ideal water and fertilizer parameters for Xinjiang’s characteristic Hami melon crop.

Currently, common research on Hami melon primarily focuses on aspects such as nutritional composition, flavor compounds, postharvest preservation, and juice processing [27,28,29]. There is a lack of studies investigating the variation patterns of quality indicators—such as external size, firmness, and sugar content—across different growth stages of Hami melon.

While the W4F3 treatment yielded the highest production, whether it also optimizes quality indicators like sugar content and firmness requires further testing and analysis. The ultimate optimal treatment must be determined by comprehensively considering both yield and quality to meet the market demands for Hami melon.

Future research could build upon the optimal water and fertilizer ranges identified in this study. Further investigation into the effects of slight water regulation (e.g., between W3 and W4) or fertilization formulas on comprehensive quality traits—such as soluble solids content, fruit firmness, moisture content, and edibility—is warranted. This would facilitate the development of a precise management model that synergistically optimizes the “yield-efficiency-quality” triad, thereby comprehensively enhancing the economic and productive benefits of the Hami melon industry.

These conclusions are based on a single-site, one-season experiment; further multi-year and multi-location studies are needed to validate their broader applicability.

5. Conclusions

This study investigated the interactive effects of irrigation and fertilization on greenhouse Hami melon in Xinjiang. The main findings are summarized as follows:

1.

Irrigation volume was identified as the dominant factor affecting melon growth and yield. Through integrated field experimentation and entropy-weight TOPSIS analysis, the treatment combining 90% of full irrigation with 100% standard fertilization (W4F3) was determined to be optimal. This regime supported a yield of 75.74 t ha⁻¹, which represents a 9.71% increase over conventional practice, alongside a 10% reduction in irrigation water. These results indicate the potential for synergistic improvement in water and fertilizer resource-use efficiency.

2.

Multi-objective optimization using the NSGA-II algorithm validated the experimental findings, generating a Pareto-optimal solution highly consistent with the W4F3 treatment. The model-predicted optimal regime (irrigation: 3276 m³ ha⁻¹, fertilization: 814.8 kg ha⁻¹) suggests that a 10% water saving and 5% fertilizer reduction compared to the control is achievable while maintaining yield stability and enhancing resource-use efficiencies. This provides a quantitative basis for decision-making in the green production of characteristic crops in arid regions.

3.

A critical threshold was observed at the W3 irrigation level (80% of full irrigation). Below this threshold, water becomes the primary limiting factor, and increased fertilization can exacerbate osmotic stress, as evidenced by the significant suppression of plant height under the W1F3 treatment. Above this threshold, water and fertilizer exhibit synergistic effects, although excessive irrigation (>100%) leads to diminishing returns.