Assessment of Evapotranspiration–Yield Relationships in Northern China Tea Plantations: A Basis for Crop Water Productivity Improvement

Abstract

Global climate warming and freshwater scarcity are intensifying water stress in agricultural fields, severely constraining sustainable agricultural development. As a typical C3 perennial cash crop, tea (Camellia sinensis) is naturally suited to low-latitude regions with abundant heat and evenly distributed precipitation, and it is highly sensitive to environmental factors such as temperature and moisture. In northern hilly tea-producing areas, tea plantations often encounter multiple challenges including uneven rainfall distribution and poor soil water retention, resulting in prominent water supply–demand imbalances that critically limit stable and efficient tea production. To explore efficient water-saving irrigation strategies adapted to such ecological conditions, this study was conducted in the Yushan Tea Plantation, Rizhao City, Shandong Province, China. Based on field monitoring data across three growing seasons (spring, summer, and autumn) from 2021 to 2023, five irrigation treatments were evaluated: conventional sprinkler irrigation (CK), drip irrigation (D), micro-sprinkler irrigation (W), drip irrigation with straw mulching (SD), and micro-sprinkler irrigation with straw mulching (SW). Actual crop evapotranspiration (ET_c act) was estimated using the soil water balance method, and actual fresh tea leaf yield (FTLY) and crop water productivity (CWP) were measured. Results showed that the SW treatment significantly improved both FTLY and CWP across all three seasons, with summer FTLY in 2022 increasing by 56.58% compared to CK and maximum CWP in spring and autumn reaching 0.916 kg/m³, demonstrating excellent stability and adaptability. Among all irrigation strategies, the SW treatment also exhibited the best regression fitting and yield prediction accuracy. The regression model validated by leave-one-out cross-validation (LOOCV) for the SW treatment demonstrated strong robustness and reliability (R² = 0.734; RMSE = 208.12 kg/ha; MAE = 183.31 kg/ha). Notably, the samples with the largest prediction errors across all treatments were nearly all associated with the highest or near-highest ET_c act values, indicating that model accuracy tends to decrease under extreme evapotranspiration conditions. The results show the synergistic effect of irrigation–mulching integration on enhancing CWP in northern perennial tea systems, providing empirical evidence and theoretical support for developing efficient irrigation strategies in hilly tea-growing regions of Northern China.

1. Introduction

The global food crisis, driven by freshwater scarcity and coupled with greenhouse gas emissions leading to climate warming, poses a serious threat to sustainable human development [1,2,3]. Climate warming increases reference evapotranspiration (ET₀) during the crop growing season, reducing the availability of effective soil moisture and intensifying agriculture’s dependence on irrigation [4,5]. Water use efficiency (WUE) and crop water productivity (CWP) are similar in that they both measure crop yield per unit of water consumed, and precision irrigation, along with improving the efficiency of both, is a key strategy to address these challenges [6,7]. As a typical C3 perennial economic crop, tea (Camellia sinensis) is highly sensitive to water availability and environmental fluctuations. With the intensification of global climate change, the increasing frequency of drought events has exacerbated water stress, which, in combination with temperature and humidity fluctuations, imposes significant constraints on both tea yield and quality [8]. In the future, climate change will pose a threat to global tea production, and the suitability for tea cultivation may decline significantly [9].

Drought, characterized by an imbalance of precipitation relative to the long-term average and potentially mitigated through irrigation, is among the most critical environmental stressors affecting the yield and geographical distribution of horticultural crops worldwide. It adversely affects tea plants by disrupting key physiological processes such as photosynthesis, ultimately slowing down their growth and reducing productivity [10]. Recent studies have demonstrated that the impacts of climatic variables such as temperature and precipitation on tea production vary significantly across seasons [11]. In northern tea-producing regions like Rizhao, Shandong Province of China, despite advantages such as ample solar radiation and large diurnal temperature variation, challenges such as uneven spatiotemporal precipitation, frequent seasonal droughts, and poor soil water retention in mountainous tea plantations severely limit soil water availability. These limitations directly affect tea growth, yield formation, and leaf quality [12]. Under these conditions, intensified ET_c act and persistent water deficits coexist, highlighting the urgent need for efficient irrigation and water management strategies to enhance both water resource utilization and crop productivity in tea-growing systems.

Extensive research has demonstrated that water-saving cultivation techniques—such as drip irrigation, micro-sprinkler irrigation, and straw mulching—have performed well in farmland water management by effectively reducing non-productive water loss, improving soil moisture conditions, and ultimately enhancing crop yield and either CWP or WUE [13,14,15,16]. For example, studies on spring maize have shown that ET_c act significantly decreases with reduced irrigation input, indicating a direct regulatory effect of water input on ET_c act processes [17]. In the tea production system, a field experiment on Anxi Tieguanyin revealed that a treatment applying 3.5 L of water irrigated at 10-day intervals led to improvements in both yield and quality, highlighting the critical role of optimized irrigation schedules in perennial crop systems [18]. Although the individual effects of these techniques have been validated across various crops, comprehensive studies on yield, ET_c act, and CWP under different irrigation regimes in tea plantations remain relatively limited. In particular, there is a lack of in-depth investigation into the combined effects of drip irrigation, micro-sprinkler irrigation, and straw mulching. Existing research has primarily focused on single irrigation methods or short-term trials, which fail to capture the long-term dynamics of soil water use in perennial tea systems under complex ecological conditions. Therefore, it is urgently necessary to conduct systematic assessments of ET_c act dynamics and CWP response mechanisms under integrated irrigation–mulching treatments, based on multi-season field observations, in order to develop efficient and site-specific water-saving irrigation strategies for hilly tea-growing regions.

Accurate estimation of ET_c act is critical for irrigation planning and improving CWP. Traditional methods for estimating ET_c act and assessing crop water stress primarily rely on empirical models based on meteorological parameters (e.g., the Penman–Monteith and Blaney–Criddle equations) and conventional remote sensing techniques. Although these approaches are convenient for large-scale applications, they often lack precision in heterogeneous small-scale ecosystems due to spatial variability. To improve estimation accuracy, researchers have proposed various refined models in recent years. The Shuttleworth–Wallace (S-W), Domingo–Kite (D-K), and Eva-f models, all based on the dual-source energy balance theory, are capable of separately identifying canopy transpiration and bare-soil evaporation. These models have been widely applied in croplands, grasslands, and forest ecosystems, thereby enhancing the physical representation of ET_c act components [19]. In parallel, data-driven approaches such as deep learning have been increasingly adopted to enhance prediction accuracy. Machine learning models like Random Forest (RF) have shown promising performance in predicting daily ET_c act, particularly in modeling complex nonlinear water dynamics [20]. In contrast, the soil water balance method, based on in situ measurements of volumetric soil water content across different layers, offers a practical and direct means to dynamically estimate water deficit and ET_c act throughout the crop growth period [21]). It is especially suitable for long-term, stratified monitoring and irrigation management in perennial systems. Notably, tea plantations are commonly located in hilly regions with complex terrain and heterogeneous soil types, leading to significant spatial variability in root-zone soil moisture [22]. This amplifies the need for high-resolution, field-based ET_c act estimation methods. Under such conditions, the soil water balance method presents a viable solution for refined water regulation in tea plantations. However, studies investigating the joint response of ET_c act, yield, and WUE under integrated irrigation and mulching treatments—particularly in perennial crop systems like tea—remain scarce. Most existing research has focused on annual crops or single management strategies, which are inadequate for addressing the fine-scale water management challenges posed by complex topography and heterogeneous soils in hilly tea plantations. Therefore, there is an urgent need to conduct comprehensive analyses of ET_c act dynamics and CWP response mechanisms under combined irrigation–mulching regimes using multilayer soil moisture monitoring and water balance techniques, aiming to develop efficient water-saving management strategies tailored to tea cultivation.

Considering the existing limitations in water assessment and regulation in northern tea-growing regions, it is imperative to establish a comparative research framework integrating field data and process-based mechanisms. This study was conducted in Rizhao, Shandong Province—a representative high-latitude tea-producing region in Northern China—and tested five water management treatments: conventional sprinkler irrigation (CK), drip irrigation (D), micro-sprinkler irrigation (W), drip irrigation with straw mulch (SD), and micro-sprinkler irrigation with straw mulch (SW). The estimation of ET_c act was conducted using the soil water balance method, in conjunction with FTLY and CWP measurements. The objectives of this study were as follows: (1) identify the effects of irrigation–mulch coupling on tea plantation ET_c act regulation; (2) determine the optimal water management strategy balancing high CWP and high yield; (3) provide theoretical and practical support for designing precision irrigation systems for tea plantations under frequent drought conditions in Northern China.

2. Materials and Methods

2.1. Study Site

The field experiment was conducted at Yushan Tea Plantation in Rizhao City, Shandong Province, China (35.454035° N, 119.419334° E), a representative hilly tea-producing area in Northern China (Figure 1). The region features a warm temperate, semi-humid monsoon climate with abundant sunlight, significant diurnal temperature variation, and moderate precipitation. The average annual temperature is 13.2 °C, with an average of −1.0 °C in January and 25.4 °C in August. Annual precipitation is approximately 840.3 mm, over 70% of which occurs between June and September, leading to pronounced seasonal rainfall variability and periodic drought conditions. The tea plantation is located in a coastal mountainous area, where irrigation water is primarily obtained by pumping from nearby reservoirs. This area is suitable for studying soil moisture dynamics and ET_c act estimation under varying irrigation and mulching practices in northern tea gardens.

2.2. Experimental Design

This experiment employed five irrigation and mulching treatments, including traditional sprinkler irrigation under extensive management (CK), high-yield management with drip irrigation (D), micro-sprinkler irrigation (W), straw mulching combined with drip irrigation (SD), and straw mulching combined with micro-sprinkler irrigation (SW) (

Table 1. Experimental treatment description.

Treatment Method	Code	Description
Flood Irrigation	CK	The control treatment employed a conventional sprinkler irrigation method without mulching or water regulation, in which water was pumped by a tractor-driven system into high-pressure hoses and directly sprayed onto the tea plants.
Drip Irrigation	D	Drip irrigation lines were installed between tea rows using pressure-compensated emitters with a rated flow of 4 L/h and emitter spacing of 30 cm. Emitters were placed on both sides of tea plant roots to ensure uniform water distribution and localized wetting for precise irrigation under low-pressure conditions.
Micro-spray Irrigation	W	Low-pressure micro-spray irrigation was applied using micro-sprinklers with a rated flow of 40 L/h and a spraying radius of 1.5 m, mounted at approximately 50 cm height in the middle of tea rows. The system maintained a working pressure of about 0.15 MPa, producing a fine mist irrigation pattern.
Straw Mulching + Drip Irrigation	SD	Straw mulching (0.6 kg/m2 of maize straw) was applied between rows in addition to drip irrigation to reduce surface evaporation and stabilize soil moisture and temperature.
Straw Mulching + Micro-spray Irrigation	SW	Straw mulching (0.6 kg/m2 of maize straw) was added to the micro-spray irrigation system, combining mist irrigation with surface coverage for dual regulation of soil water conditions.

). Each treatment was replicated three times in a randomized block design. The experimental tea cultivar was Huangshan Chu Ye, a widely cultivated local green tea variety in Rizhao known for its strong cold resistance, high yield, and quality. Tea plants were uniformly 12 years old and free of pests and diseases. The plots were established on terraced slopes at Yushan Tea Plantation with each terrace equipped with concrete retaining walls and protective vegetation strips to minimize surface runoff. Tea bushes were planted with a spacing of 0.3 m within rows and 1 m between rows, and each plot measured 6 m × 8 m (48 m²). Except for the CK treatment, all plots were managed following standardized high-yield cultivation practices, with consistent fertilization and field management throughout the growing period. Details of the treatment setups are shown in Table 1.

2.3. Sampling, Measurements, and Calculations

2.3.1. Soil Volumetric Water Content

Soil moisture in the root zone (0–50 cm) of the tea plantation was manually measured using a ProCheck portable soil moisture reader (METER Group, Pullman, WA, USA) in conjunction with soil moisture sensors (TEROS 12; VWC accuracy ± 0.03 m³/m³ with general calibration), and was calibrated using the oven-drying method. The experimental site is located in a hilly mountainous tea-growing region, where the soil below 50 cm is predominantly compacted and unfavorable for root development and water movement. In Rizhao’s mountainous areas, particularly in the low hill regions surrounding reservoirs, the groundwater table typically ranges from 1 to 5 m below the surface [23]. Given the limited hydraulic connectivity between the groundwater and the shallow root zone, soil moisture in tea plantations primarily depends on precipitation and surface irrigation. Based on these site conditions, soil moisture measurements in this study focused on the effective root zone (0–50 cm). Monitoring points were established at five soil depths: 10 cm, 20 cm, 30 cm, 40 cm, and 50 cm. For each measurement, the sensor was manually inserted vertically into the specified depth, and the volumetric water content (VWC, %) was recorded in real time using the ProCheck device. Soil moisture at each depth layer was measured three times at different locations, and the average value was used to represent the soil moisture at that layer. The monitoring period covered the months from March to November (2021–2023), corresponding to the spring, summer, and autumn seasons in the Northern Hemisphere. Measurement frequency was increased during key phenological stages and before and after irrigation events. Simultaneously with sensor readings, soil samples were collected and oven-dried to determine gravimetric water content. These values were then converted to volumetric water content based on soil bulk density and used to calibrate sensor readings via linear regression fitting.

2.3.2. Irrigation Amount

Due to the high dependence of tea plant growth on adequate water supply, an empirical irrigation strategy was implemented based primarily on the growth rate of tea plants and local farmers’ management practices. Irrigation was generally avoided during tea-picking periods to maintain leaf quality and was initiated when the growth rate slowed due to insufficient water supply. The irrigation amount was estimated by multiplying the discharge rate of the irrigation device (per unit time) by the duration of each irrigation event. The total irrigation volume was then accumulated on a 10-day basis to obtain the periodic irrigation amount. Irrigation records for each treatment were converted to equivalent water depths (cm or mm) and summarized based on the cumulative amount from the previous 10 days.

2.3.3. Rainfall Distribution Analysis

To analyze the temporal distribution of rainfall, monthly 10-day rainfall totals (i.e., three periods per month) were calculated using daily precipitation data from 2021 to 2023, obtained from the local meteorological station. The Concentration Index (CI) was used to quantify rainfall concentration within each month. CI ranges from 0 to 1, with higher values indicating that rainfall is concentrated in a single 10-day period, and lower values suggesting a more uniform distribution across the month. The CI for each month was calculated using the formula:

$$CI={\displaystyle \frac{P_{max}}{P_{month}}}$$

(1)

where $P_{max}$ is the precipitation in the wettest 10-day period of the month, and $P_{month}$ is the total precipitation for that month. This method allows identification of anomalous concentration events and supports comparison of interannual rainfall patterns.

2.3.4. Fresh Tea Leaf Yield (FTLY)

The FTLY was determined periodically for each treatment. For each plot, uniform tea plants (excluding border rows) were randomly selected for yield measurement. At least three sampling points with consistent growth were identified per plot, and 1 m of tea plants was harvested at each point. The harvested leaves were weighed after drying to determine dry weight, which was then converted to yield per hectare. All measurements were conducted using precision weighing equipment with an accuracy of at least 0.01 g.

2.3.5. Actual Crop Evapotranspiration (ET_c Act)

The actual crop evapotranspiration (ET_c act) of tea plantation was estimated using the classical soil water balance method. In most northern tea-growing regions, such as the Yushan Tea Plantation in Rizhao, Shandong Province, tea gardens are typically located on slopes or hilly terrain where the groundwater table is deep. The active root zone of tea plants (generally 10–50 cm) is far from the groundwater table, and capillary rise is negligible. Therefore, the contribution of groundwater to root zone soil moisture can be ignored. Additionally, surface runoff caused by precipitation was considered negligible due to the installation of perimeter barriers around the experimental tea garden. According to the water balance equation [24],

$$E{T}_{c,act}=\Delta S+P+I-D$$

(2)

where $E{T}_{c,act}$ is the evapotranspiration (mm), $\Delta S$ is the change in soil water storage (mm), $P$ is the effective precipitation during the growing period (mm), provided by the nearby meteorological station, $I$ is the irrigation amount (mm), and $D$ denotes deep percolation, which was considered negligible in this study.

The change in soil water storage (ΔS) was calculated as follows:

$$\Delta S={\displaystyle \sum (\Delta {\theta }_i\times Z_i)}$$

(3)

where $\Delta {\theta }_i$ represents the change in volumetric water content in soil layer i (%), and $Z_i$ is the thickness of the corresponding soil layer (mm). Five layers (10, 20, 30, 40, and 50 cm) were considered in the calculation, corresponding to the effective root zone of tea plants in this hilly area.

2.3.6. Crop Water Productivity (CWP)

Crop water productivity (CWP) was calculated as the ratio of FTLY to ET_c act over the corresponding growth period, reflecting the crop’s ability to convert consumed water into economic yield [17], and is expressed as follows:

$$CWP=FTLY/E{T}_{c,act}$$

(4)

where $CWP$ is crop water productivity (kg/m³), and $E{T}_{c,act}$ is the evapotranspiration during the same period (mm).

2.3.7. Precipitation Data

Rainfall data were collected from the tea plantation meteorological observation station.

2.3.8. Quadratic Regression Model with a Single Independent Variable

To investigate the influence mechanism of water-related factors on tea yield under different irrigation practices, quadratic regression models were developed for each of the five treatments: conventional surface irrigation (CK), drip irrigation (D), micro-sprinkler irrigation (W), straw mulching combined with micro-sprinkler irrigation (SD), and straw mulching combined with drip irrigation (SW). In each model, ET_c were used as independent variables, while FTLY (Y) was taken as the dependent variable. The general form of the model is as follows:

$$FTLY=a\cdot E{T_{c,act}}^2+b\cdot E{T}_{c,act}+c$$

(5)

where $FTLY$ is the fresh leaf yield per unit area (kg/ha), and $E{T}_{c,act}$ is the accumulated evapotranspiration (mm).

2.3.9. Leave-One-out Cross-Validation (LOOCV) for Model Performance Evaluation

To evaluate the generalization performance of the regression model under limited sample size conditions, a leave-one-out cross-validation (LOOCV) approach was employed. This method is particularly suitable for small datasets, as it maximizes the use of available data for both training and validation, thereby reducing the risk of overfitting and improving the robustness of the model evaluation. In this method, each sample was iteratively excluded as the test set while the remaining samples were used for model training, ensuring robust assessment of prediction accuracy across all irrigation treatments. To comprehensively evaluate the prediction accuracy and stability of the regression model under different irrigation treatments, three commonly used statistical metrics were employed: the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and mean bias error (MBE). Specifically, R² measures the proportion of variance in yield explained by the model. RMSE quantifies the standard deviation of prediction errors, while MAE captures the mean absolute deviation between predicted and observed values. MBE represents the mean bias, where positive values indicate systematic overestimation and negative values indicate underestimation. The corresponding formulae are as follows:

$$R^2=1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^n}{(y_i-{\widehat{y}}_i)}^2}{{\displaystyle \sum _{i=1}^n}{(y_i-\overline{y})}^2}}$$

(6)

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}\sum _{i=1}^n{(y_i-{\widehat{y}}_i)}^2}$$

(7)

$$MAE={\displaystyle \frac{1}{n}}\sum _{i=1}^n|y_i-{\widehat{y}}_i|$$

(8)

$$MBE={\displaystyle \frac{1}{n}}\sum _{i=1}^n({\widehat{y}}_i-y_i)$$

(9)

where $y_i$ is the observed value, ${\widehat{y}}_i$ is the predicted value, and $\overline{y}$ is the mean of the observed values.

2.4. Data Processing and Statistical Analysis

Quality assurance and quality control (QA/QC) procedures were applied to all datasets used in this study. All data were collected, processed, and analyzed following the relevant calibration and validation methods described in the corresponding sections to ensure accuracy and reliability. Data processing and statistical analyses were conducted using Microsoft Excel 2025 (Microsoft Corporation, Redmond, WA, USA), SPSS 21.0 (IBM Corp., Armonk, NY, USA), and Python 3.10 (Python Software Foundation, https://www.python.org/, accessed on 10 August 2025). A two-way analysis of variance (ANOVA) was performed to evaluate the effects of irrigation treatments and seasons on fresh tea leaf yield and crop water productivity (CWP). When significant differences were detected at the p < 0.05 level, the Least Significant Difference (LSD) test was used to compare treatment means. Graphs were primarily generated using Origin 2023 (OriginLab Corporation, Northampton, MA, USA), with additional data processing and visualization performed using Python libraries such as pandas (v2.2.1; https://pandas.pydata.org/, accessed on 10 August 2025), numpy (v1.26.4; https://numpy.org/, accessed on 10 August 2025), and matplotlib (v3.8.3; https://matplotlib.org/, accessed on 10 August 2025). Regression analysis and leave-one-out cross-validation (LOOCV) were implemented using the scikit-learn library (v1.4.1; https://scikit-learn.org/, accessed on 10 August 2025).

3. Result

3.1. Rainfall Analysis

To reveal the temporal distribution characteristics of rainfall in the study area, monthly 10-day rainfall totals and the corresponding CI were calculated for the years 2021 to 2023. The CI ranges from 0 to 1, with higher values indicating that rainfall was concentrated in a single 10-day period, while lower values suggest a more even distribution across the month (Figure 2).

From an interannual perspective, 2021 recorded the highest total rainfall (1026.1 mm), with a relatively balanced distribution and a moderate average CI of 0.76. For example, rainfall in May (CI = 0.95) and October (CI = 0.87) was mainly concentrated in a single 10-day interval, showing typical episodic precipitation. In 2022, highly concentrated rainfall events were more prominent. Although the total annual rainfall was slightly lower than in 2021, extreme precipitation in June increased the seasonal total, and the average CI for the three growing seasons reached 0.89. Rainfall was particularly concentrated in spring (March–May) and autumn (September–November), with April and October both reaching a CI of 1.00, indicating that rainfall occurred almost entirely within a single 10-day period—anomalous concentration events. In contrast, 2023 had the lowest annual rainfall (763.1 mm) and a generally lower rainfall concentration (average CI = 0.71). It is noteworthy that summer rainfall in 2023 exhibited a relatively uniform temporal distribution (CI = 0.67), suggesting consistent but moderate precipitation events without significant intensity peaks.

During the spring season (March to May), rainfall demonstrated significant fluctuations and a marked tendency toward temporal concentration. The three-year average spring rainfall was 82.77 mm, but with substantial interannual fluctuation, showing a typical “dry spell followed by sudden downpour” pattern. Spring 2022 was extremely dry, with only 42.2 mm of rainfall, resulting in severe water stress during the spring tea growing period. In contrast, spring 2023 recorded the highest rainfall (124.6 mm), with distinct peaks in early April (36 mm) and late May (32 mm). Spring 2021, although lacking extreme events, also exhibited staged concentration, with 49.6 mm in late April and 49 mm in mid-May. The average CI during spring was 0.80, indicating significant concentration. The highest concentration occurred in spring 2022 (CI = 0.88), with April reaching CI = 1.00, indicating that rainfall was entirely concentrated within a single 10-day window. Although such concentration can rapidly replenish soil moisture, it often results in unstable water availability, posing challenges for spring tea bud development and pest and disease control.

Summer (June–August) is the wettest season in the tea-growing region, with a three-year average rainfall of 537.83 mm—significantly higher than that in spring and autumn—and showing considerable interannual variability. In 2022, extreme rainfall events occurred, with total summer precipitation reaching 741.5 mm. Notably, late June recorded a 10-day rainfall of 338.5 mm, the highest decadal value over the study period, indicating an intense rainfall concentration. In comparison, July and August of 2021 received 246.5 mm and 213 mm, respectively, with relatively even rainfall distribution. Although no extreme peaks occurred in 2023, each of the three summer months still exceeded 140 mm in total rainfall, reflecting a prolonged high-intensity rainfall pattern. The average summer Concentration Index (CI) was 0.63, the lowest among all seasons, suggesting that rainfall events were more continuous and characterized by steady moderate-to-heavy precipitation. In 2023, the CI value was 0.67, further confirming the absence of abrupt rainfall concentration. These summer rainfall characteristics are generally favorable for rapid tea plant growth and nutrient uptake. However, the persistently moist conditions may exacerbate the risk of pests and diseases, highlighting the need for improved drainage and disease management.

Autumn (September–November) showed substantial interannual rainfall variation and a generally high concentration tendency. The average autumn rainfall over the three years was 212.13 mm. In 2021, autumn rainfall was the highest (326 mm), with early and mid-September recording 94 mm and 93 mm, respectively, forming two clear rainfall peaks. In 2022, a ‘wet September–dry October’ pattern was observed, with episodic heavy rainfall in mid-September and mid-November. Conversely, 2023 experienced the driest autumn (128 mm), resulting in increased irrigation pressure for autumn tea. The average CI in autumn was 0.79, similar to that of spring, indicating a greater likelihood of abrupt, high-intensity rainfall events. In 2022, the CI peaked at 0.91, with nearly all rainfall in September (CI = 0.99) and October (CI = 1.00) concentrated within single 10-day periods. Such concentrated events elevate the risk of surface runoff and waterlogging in tea fields. In contrast, September 2021 had a CI of only 0.40, indicating a more dispersed rainfall pattern conducive to gradual soil moisture recharge and improved root respiration. Autumn rainfall variability is closely related to typhoons and cold air outbreaks, warranting greater attention to the mitigation of extreme weather impacts during the late growing season.

3.2. Effects of Different Water-Saving Irrigation Techniques on Irrigation Amount

A statistical analysis was conducted on the irrigation volumes of five treatments (CK, D, W, SD, SW) across three harvesting stages—spring, summer, and autumn—from 2021 to 2023 (Figure 3). Among all treatments, the control (CK) consistently exhibited the highest irrigation volumes in each season, reaching 215.97 mm in spring, 168.75 mm in summer, and 91.67 mm in autumn. These results reflect the typical water demand pattern of tea plants under conventional management. All other treatments effectively reduced irrigation amounts to varying extents. Specifically, during the spring tea season, irrigation amounts were reduced from 215.97 mm in CK to 104.86 mm (D), 114.58 mm (W), 118.06 mm (SD), and 111.81 mm (SW). In the summer season, irrigation amounts under the water-saving treatments ranged from 85.42 mm (D) to 79.17 mm (SW), compared to 168.75 mm in CK. During the autumn season, irrigation further declined across all treatments, with CK at 91.67 mm and the lowest value observed in SW (52.78 mm). Overall, seasonal irrigation followed the trend of spring > summer > autumn, which aligns with the phenological water requirements of tea plants under northern climate conditions.

To assess the interannual stability of each irrigation strategy, the coefficient of variation (CV) was calculated. Results showed that CK maintained relatively stable irrigation patterns, with CVs ranging from 10% to 20% across all seasons. In contrast, SW exhibited the highest variability in spring (CV = 41.09%), and all water-saving treatments had CVs exceeding 30%, indicating that spring irrigation is more susceptible to climatic variability. In summer, the W treatment had the highest CV (24.71%), while SD exhibited the lowest (15.37%), suggesting that SD offered better irrigation stability during this period. In autumn, SW showed the lowest CV at 8.22%, indicating both high water-saving potential and strong interannual stability.

Water-saving efficiency was calculated relative to CK. All water-saving treatments achieved substantial reductions in irrigation volume across all growth stages. The highest water-saving rate in spring was observed under D treatment (51.45%), while W, SD, and SW maintained rates between 45% and 49%. In summer, SW achieved the highest water-saving rate of 53.09%, showing a significant advantage. Although irrigation volumes were already low in autumn, SD and SW treatments still achieved reductions exceeding 40%. Overall, both D and SW treatments demonstrated excellent water-saving potential and adaptability across multiple growth stages, making them promising options for broader application.

3.3. Analysis of Tea Garden ET_c Act Under Different Treatments

From the perspective of multi-year average cumulative ET_c act, the SW treatment exhibited the lowest value (1153.84 mm), followed by D (1173.92 mm) and W (1176.79 mm) (Figure 4). Across all treatments, spring ET_c act was relatively low, and the CK treatment showed significantly higher ET_c act than other treatments, with an average daily ET_c act of 3.7 mm/day. In contrast, differences among the water-saving irrigation treatments were not statistically significant. The overall variation in ET_c act among treatments during spring was small, with limited standard error, indicating similar soil moisture conditions across treatments in this season.

Summer represented the peak ET_c act season due to elevated temperatures and higher evaporative demand. All treatments showed daily ET_c act values exceeding 7 mm/day, while water-saving treatments (ranging from 7.1 to 7.3 mm/day) exhibited significantly lower ET_c act than the CK treatment (8.1 mm/day), reflecting the effectiveness of improved irrigation strategies in mitigating water loss. In autumn, ET_c act gradually decreased with the decline in temperature. No significant differences were observed in cumulative ET_c act among treatments; the CK treatment had an average daily ET_c act of 3.5 mm/day, while the other treatments ranged from 3.2 to 3.3 mm/day.

In terms of interannual variation, the ET_c act process varied considerably across years. In 2021, total ET_c act reached its highest level, with CK peaking at 1567.9 mm during the growing season, significantly surpassing all water-saving treatments. The spring and autumn seasons in 2021 featured higher average daily ET_c act across treatments compared to other years; the CK treatment averaged 4.7 mm/day, while the W and SD treatments had the lowest values, both at 3.1 mm/day. In 2022, the CK treatment recorded an ET_c act of 1440.5 mm during the growing season. Notably, daily ET_c act values in spring were lower than those in other years, with CK at 3.1 mm/day and water-saving treatments ranging between 1.1 and 1.8 mm/day. In summer, daily ET_c act peaked across all treatments, with CK reaching 9.5 mm/day and water-saving treatments ranging from 8.3 to 9.1 mm/day. Autumn daily ET_c act values for all treatments were between 2.6 and 3.1 mm/day. In 2023, ET_c act levels across all treatments were the lowest of the three years. CK had an annual ET_c act of 1206.03 mm, representing a decrease of nearly 20% compared to 2021 and 2022. In this year, average daily ET_c act ranged from 2.1 to 3.4 mm/day in spring, 5.8 to 7.1 mm/day in summer, and 2.3 to 2.7 mm/day in autumn.

3.4. Effects of Different Water-Saving Irrigation Techniques on FTLY

To investigate the impact of different irrigation treatments on FTLY, yield data from spring, summer, and autumn tea seasons in a Rizhao tea plantation were analyzed for the period 2021–2023 (Figure 5). Distinct seasonal variations in FTLY were observed, with summer tea generally producing higher FTLY than spring and autumn tea. This pattern can be attributed to the physiological characteristics of tea plants, including enhanced photosynthesis and accelerated bud development under high-temperature and high-humidity conditions. Yields of autumn and spring tea were relatively similar, following a general order of summer > autumn > spring, indicating that summer tea is most sensitive to water availability and offers the greatest potential for yield enhancement.

Among the five irrigation treatments, the four improved methods (D, W, SD, SW)—all exhibited varying degrees of positive effects on tea yield compared to the conventional flood irrigation control (CK). The SW treatment consistently produced the highest yields. For example, in the summer of 2022, the CK yield was 1524.98 kg/ha, while the D, W, SD, and SW treatments achieved 1963.16, 1954.83, 1950.66, and 2387.46 kg/ha, representing yield increases of 28.76%, 28.21%, 27.92%, and 56.58%, respectively. The SW treatment showed the most significant advantage. In the same year’s spring tea season, the CK yield was 959.01 kg/ha, which increased to 1295.81 kg/ha under the SW treatment, a 35.12% improvement; autumn tea yield increased from 1197.20 kg/ha to 1679.14 kg/ha, representing a 40.25% increase. Similar trends were observed in 2021 and 2023, strongly supporting the conclusion that integrated water–fertilizer saving irrigation techniques are effective in enhancing tea production.

Additionally, the W and D treatments also demonstrated individual yield enhancement potential in certain years. For instance, in the spring of 2023, the W treatment achieved a yield of 1440.95 kg/ha, a 37.99% increase over CK. In the same year’s autumn season, the D treatment reached 1643.72 kg/ha, clearly outperforming the control. The relative ranking of yield performance across treatments remained stable, with SW consistently achieving the best results in both absolute yield and interannual yield stability, while W and SD treatments showed slightly higher yield variability than other treatments.

3.5. Effects of Different Water-Saving Irrigation Techniques on CWP

Significant differences in crop water productivity (CWP) were observed among the three tea harvesting seasons (spring, summer, and autumn) under different treatments (Figure 6). The SW treatment significantly improved CWP across all three seasons, with the maximum CWP in spring and autumn reaching 0.916 kg/m³. On a three-year average, all water-saving irrigation treatments (D, W, SD, SW) recorded the highest CWP in the summer tea season, followed by autumn, with the lowest values in spring. However, seasonal comparisons indicated that CWP in the spring and autumn seasons was significantly higher than in summer for all water-saving irrigation treatments.

Interannual variations in seasonal CWP were also evident. Higher overall CWP levels were observed in the spring of 2022 and the autumn of 2023. In 2022, significant differences in CWP among the three seasons were detected within the same water-saving irrigation treatment, whereas, in 2021 and 2023, CWP in the summer season was significantly lower than in the corresponding spring and autumn seasons.

Across all three seasons, CWP under water-saving irrigation treatments was generally higher than that of CK (p < 0.05). In most cases, the combined mulching and water-saving irrigation treatments (SD and SW) outperformed the single irrigation treatments (D and W), although the differences were not statistically significant.

3.6. Tea Yield Modeling Under Different Irrigation Strategies: A Regression-Based Approach

To investigate the overall relationship between ET_c act and seasonal FTLY under different irrigation treatments, this study integrated spring, summer, and autumn yield and ET_c act data and performed a quadratic regression analysis across the five treatments (Figure 7). The results revealed a significant nonlinear relationship between seasonal yield and ET_c act. Under the micro-irrigation treatments (D, W, SD, and SW), the yield exhibited an increasing-then-decreasing trend in response to ET_c act. The fitted regression curves and corresponding coefficients of determination (R² ranging from 0.7232 to 0.877) indicated that ET_c act is a key driver of seasonal yield variability in tea plantations.

For treatments D, W, SD, and SW, the relationship between yield and ET_c act followed a concave-down parabolic pattern, reflecting a nonlinear response. Initially, yield increased with ET_c act due to improved water availability; however, beyond a certain threshold, further increases in ET_c act led to yield declines. This inflection point represents the optimal ET_c act level for achieving maximum yield. Among the treatments, the SW treatment exhibited the best fit (R² = 0.877), suggesting that the combination of straw mulching and micro-sprinkler irrigation enhances water supply stability, improves plant water status, and promotes CWP by regulating the partitioning between evaporation and transpiration.

In contrast, the CK treatment showed an upward-opening curve, with a positive quadratic term coefficient, indicating that, within the observed range, yield continued to increase with rising ET_c act. No yield suppression due to excessive irrigation or water saturation was observed. However, the CK treatment had the lowest R² value (0.7232), suggesting greater yield variability and weaker predictive capability, likely due to uncontrolled water supply and higher susceptibility to environmental disturbances.

Overall, a moderate level of ET_c act is beneficial for enhancing FTLY, while excessive or insufficient ET_c act can constrain yield potential. Among the five treatments, SW and SD demonstrated more favorable water–yield trade-offs and are thus recommended as precision irrigation strategies under water-limited conditions.

3.7. Comparison of FTLY Prediction Accuracy Under Five Irrigation Treatments Using LOOCV

To evaluate the performance of the regression model in predicting FTLY under different irrigation treatments, an LOOCV approach was employed for five irrigation strategies (Figure 8). Among all treatments, the SW treatment (straw mulching combined with drip irrigation) exhibited the best performance, with an R² of 0.87, and RMSE and MAE values of 289.28 and 260.34 kg/ha, respectively. This indicates a high level of agreement between predicted and observed values, reflecting strong model stability and reliability. In contrast, the SD treatment demonstrated the lowest model accuracy, with a negative R² (−0.24) and a high RMSE of 339.77 kg/ha, suggesting that the model failed to adequately capture yield variations under this irrigation regime and performed worse than simple mean-based predictions.

Based on the LOOCV results, there are significant differences in the predictive performance of the models for each irrigation treatment. The CK treatment model has a moderate fit but exhibits some underestimation bias. The D treatment model shows large prediction errors and demonstrates significant overestimation bias. The W treatment, although showing a noticeable overestimation bias, performs better than the D treatment and holds some reference value. The SD treatment model has a poor fit with large prediction errors, indicating the need for model improvement. On the other hand, the SW treatment model performs the best, with a good fit that explains about 73% of the data variability. Despite some overestimation bias, the overall prediction is quite accurate and suitable for practical application. Therefore, the SW treatment is the most reliable model, while the other treatments need further optimization.

4. Discussion

4.1. Effect of Rainfall on Tea Plantation Yield

This study revealed that, from 2021 to 2023, Rizhao exhibited a typical rainfall pattern characterized by “scarce and concentrated” rainfall in spring and autumn, and “abundant but dispersed” rainfall in summer. The interannual variation in the CI was substantial, providing a critical climatic backdrop for precision irrigation regulation. Driven by this climatic pattern, the annual yield of tea leaves under different treatments generally followed the order of summer tea > autumn tea > spring tea, which closely mirrored the temporal distribution of rainfall, indicating a sensitive yield response to precipitation. This finding aligns with previous research. For instance, analyses in several major tea-producing regions of India reported significant positive correlations between rainfall and tea yield, with annual rainfall showing R² values of 0.411 (p < 0.05) and 0.665 (p < 0.01), respectively [25]. However, the impact of rainfall variability on tea yield remains controversial across regions. Some studies found that years with extreme rainfall tend to be associated with higher yields. For example, Farukh et al. (2020) [26] observed that extreme rainfall events in northeastern Bangladesh were often accompanied by elevated tea production. Conversely, other research emphasizes the adverse effects of excessive or irregular rainfall. Gunathilaka et al. (2017) [27] reported that every 100 mm increase in annual precipitation corresponded to an approximate 1% reduction in tea yield. Mallik (2021) [28] also highlighted that, in the Dooars region of West Bengal, India, high temperature, humidity, and uneven rainfall during the monsoon season led to yield declines. Similarly, Spottiswoode (2009) [29] noted that excessive rainfall reduces solar radiation and photosynthesis and causes soil waterlogging, which impairs tea plant absorption capacity. In the present study, a significant extreme rainfall event occurred in Rizhao during late June 2022 (338.5 mm in a ten-day period), which negatively affected summer yields across most treatments. However, the SW treatment effectively alleviated water stress and achieved the highest yield recorded during the three-year period, demonstrating strong adaptability and regulatory capacity. These results suggest that appropriate water-saving irrigation technologies can enhance the resilience of tea plantations under rainfall extremization and help stabilize yields.

4.2. ET_c Act Dynamics Under Different Irrigation Regimes

ET_c act represents the primary pathway of water loss in tea growth and is influenced by climate, irrigation method, and ground cover management. Based on seasonal observations from 2021 to 2023, tea plantations in Rizhao exhibited distinct seasonal ET_c act trends—moderate in spring, peaking in summer, and declining in autumn—which is consistent with patterns observed in subtropical monsoon climates. For example, Geng et al. (2023) [20] reported that, in the Tianmu Lake tea region, the multi-year average daily ET_c act was 2.05 mm/day, with the highest daily value reaching 6.9 mm in July. The ET_c act levels were generally higher than those in some other subtropical regions. During summer, several treatments recorded average daily ET_c act exceeding 5 mm, and the annual totals surpassed those reported by Zheng et al. (2021) [30], who found maximum annual ET_c act values of 701.60 mm and 830.24 mm, and the highest monthly average ET_c act of 4.55 mm/day. This elevated ET_c act in Rizhao may be attributed to its coastal hilly topography, which leads to high solar radiation, wind speed, and temperatures in summer. Additionally, the proportion of uncovered or insufficiently mulched tea fields is relatively high, increasing surface evaporation. Yan et al. (2022) [19] similarly found that, in Anji white tea plantations, soil evaporation accounted for up to 23.79% of total ET_c act, indicating the significance of management practices in shaping ET_c act dynamics.

Straw mulching and drip irrigation have been widely recognized for their water-saving efficacy. In maize fields in Tongliao, China, drip irrigation combined with plastic mulch reduced total ET_c act by 11% compared to treatments without mulch, underscoring the effectiveness of ground cover in minimizing non-productive evaporation [31]. The findings further verify the applicability of this mechanism in northern tea plantation ecosystems. Specifically, treatments involving straw mulching and water-saving irrigation (D, W, SD, and SW) significantly reduced ET_c act compared to the control (CK), indicating stronger water regulation capabilities. On average over three years, SW treatment achieved the lowest ET_c act in summer and autumn, while SD was the most effective in spring. Despite these seasonal differences, the total annual ET_c act did not differ significantly among the four water-saving treatments, suggesting comparable efficacy in controlling water loss. Comparable results were reported by Wang and Juang (2024) [32], who observed that increased canopy and weed coverage in Taiwanese organic tea plantations led to a 33.8% higher ET_c act rate (8.38 mm/day) compared to conventionally managed sites. This discrepancy highlights that the effects of mulching on ET_c act depend not only on physical water retention and insulation mechanisms but also on regional climate, canopy architecture, and agronomic practices. Overall, the SD and SW treatments, which combine mulching and micro-irrigation, exhibit strong adaptability and regulatory potential for ET_c act control in tea plantations.

In this study, evapotranspiration was estimated using the soil water balance method, which assumes negligible deep percolation and capillary rise. However, under conditions of uneven rainfall distribution or extreme precipitation events, these assumptions may not hold, potentially introducing uncertainties into the accuracy of ET_c act estimation.

4.3. FTLY and CWP Characteristics Under Different Water-Saving Irrigation and Mulching Treatments

The results of this study indicate that straw mulching combined with water-saving irrigation techniques (D, W, SD, and SW) significantly improved tea yield—by 28.76%, 28.21%, 27.92%, and 56.58%, respectively—without substantially increasing water input. These treatments also effectively enhanced CWP, demonstrating a favorable water-saving and yield-enhancing synergistic effect. This finding is well aligned with previous studies. Zhang et al. (2022) [33] reported that plastic film mulching in a drip irrigation system increased tea yield by 20.2% and CWP by 29.3%, indicating the broad applicability of optimized combinations of irrigation and mulching technologies across various crops and ecological regions. Similarly, Madhu et al. (2011) [34] showed that the introduction of cover crops in tea plantations increased rainwater use efficiency by 19%, highlighting the strong water response potential of tea plants to mulching interventions. In a study on Camellia oleifera, Ye et al. (2021) [35] found that organic mulching significantly increased fruit yields under non-irrigated conditions, with yield improvements of 58%, 61%, and 124% for the cultivars ‘Hengdongdatao’, ‘Huashuo’, and ‘Xianglin 210’, respectively. Yang et al. (2022) [36] also confirmed that plastic film and straw mulch significantly enhanced both root yield and the accumulation of bioactive compounds in licorice by 26–34%. Similarly, Fang et al. (2024) [37] observed in a maize system that degradable biofilm and plastic film treatments significantly reduced ET by 65 mm during the dry season and improved CWP. These results collectively support the view that mulching improves soil hydrothermal conditions and contributes to more efficient water use. However, this study also revealed that, in micro-irrigation systems with already high water-use efficiency, the addition of straw mulch (i.e., SW compared to W) did not result in a further significant improvement in CWP. This suggests a potential marginal effect of mulching, wherein its benefits may diminish under highly efficient irrigation regimes. Further research has indicated that, in a subtropical tea plantation in Anxi, Chen et al. (2010) [18] achieved simultaneous improvements in tea yield and quality under water-saving conditions by optimizing irrigation frequency (e.g., irrigating every 10 days). This emphasizes that, instead of simply increasing water input or relying solely on mulching, adopting precise irrigation scheduling strategies is more effective in improving resource use efficiency. Therefore, future studies may consider introducing a dual-factor regulation approach involving both irrigation amount and frequency to explore precision irrigation models and management strategies better suited to the ecological conditions of northern tea plantations. In addition, Yu et al. (2023) [38] pointed out that significant differences in WUE exist among various crops, with high-WUE species such as cotton, oats, and cabbage demonstrating strong drought tolerance and stable productivity. In line with this, our study showed that CWP levels under the D, W, SD, and SW treatments were all significantly higher than those of the control group, indicating that tea plants also possess a strong physiological capacity for water regulation. With scientifically managed irrigation, tea plantations are expected to achieve efficient water use and stable yields even under drought or water-scarce conditions.

4.4. Comparative Analysis of FTLY–ET_c Act Relationship Under Contrasting Irrigation Strategies

Based on field-observed seasonal tea yield and corresponding ET_c act data from spring, summer, and autumn between 2021 and 2023, this study established quadratic regression models under five irrigation treatments (CK, D, W, SD, and SW) to investigate the response of yield to ET_c act. The results showed that, except for CK, all micro-irrigation and combined treatments exhibited downward-opening parabolic relationships, indicating a clear optimal ET_c act range. Among them, the SW treatment had the best model fit (R² = 0.877), suggesting that yield increased with ET_c act up to a certain threshold, beyond which excessive ET_c act led to a decline in yield—demonstrating a nonlinear ET_c act–FTLY response. The regression models not only revealed the nonlinear response mechanism between tea yield and ET_c act but also provided a theoretical basis for defining optimal water regulation thresholds under precision irrigation. Additionally, they offer methodological insights for parameterization and site-specific calibration of water–yield relationships in crop simulation models [39]. This nonlinear pattern aligns with the findings of Manna et al. (2024) [40], who reported that seasonal ET_c act and both dry matter accumulation and yield in potatoes followed quadratic trends, with R² values of 0.545 and 0.456, respectively. Similarly, they observed yield reductions when ET_c act exceeded optimal levels. In contrast, some studies highlight a linear positive correlation between ET_c act and crop yield. For instance, Irmak (2024) [41] found that increasing ET_c act consistently improved maize yield. Sandhu and Irmak (2022) [42] further demonstrated that, for soybean, every 25.4 mm increase in ET_c act could enhance yield by 0.01–0.52 t/ha, depending on planting density and cultivar characteristics. Moreover, Tadesse et al. (2015) [43] showed that ET_c act was significantly correlated with detrended grain yield across 76% of 41 cropping zones, with R² ranging from 0.4 to 0.82. These findings underscore the predictive value of ET_c act in crop yield modeling and highlight the broader applicability of ET_c act–yield relationships across regions and crops.

The present study demonstrated a significant quadratic relationship between ET_c act and FTLY, particularly under the SW treatment, which achieved the highest model performance and prediction accuracy. Considering the limited sample size, this study employed LOOCV to evaluate the generalization capability of the regression model. In this study, LOOCV played a crucial role in assessing the robustness and reliability of the model. This nonlinear pattern aligns with findings in previous modeling studies [44], where nonlinear models often outperformed linear ones in representing the complex response of crop yield to environmental variables such as ET_c act and temperature. The evaluation of the model using LOOCV revealed that, for all irrigation treatments, the largest prediction residuals occurred at the samples with the highest or near-highest ET_c act values. This indicates that, although the regression model performed well under average conditions, it exhibited a clear lack of robustness under extreme ET_c act scenarios. This finding is consistent with the study by Xu et al. (2025) [45], which also reported a significant decline in model prediction accuracy under extreme climatic conditions, suggesting the need to incorporate physiologically meaningful variables or adopt alternative modeling approaches to improve performance. Incorporating additional predictors such as vapor pressure deficit, the leaf area index, or soil texture may improve model responsiveness under extreme conditions.

4.5. Strengths and Limitations

This study offers a comprehensive evaluation of regression models for predicting FTLY under various irrigation treatments using leave-one-out cross-validation (LOOCV), ensuring robust and reliable model performance assessment. By investigating five irrigation strategies, this study provides valuable insights into how different irrigation practices impact tea yield predictions. The models are grounded in empirical data from actual field experiments, providing context-specific insights into tea plantation management.

However, the models developed are based on specific conditions at the Rizhao tea plantation, which limits their generalizability to other regions with different environmental conditions. The estimation of evapotranspiration using the soil water balance method involves assumptions (e.g., negligible deep percolation) that may not hold under extreme rainfall events, potentially introducing uncertainty. This study primarily relies on empirical data, without fully incorporating the underlying physical processes of evapotranspiration and yield, which could affect model accuracy. Furthermore, the regression models were based on data from a single site over a continuous three-year period, which may restrict their applicability to other geographic regions, soil types, or tea cultivars. Future research should incorporate multi-site and multi-year datasets, adopt more advanced modeling approaches (such as ensemble learning), and improve the measurement of soil hydrological processes to enhance the robustness, accuracy, and generalizability of the findings.

5. Conclusions

This study systematically analyzed the response characteristics and interrelationships among FTLY, ET_c act, and CWP under different irrigation and mulching treatments. The results revealed the following:

This study evaluated the predictive performance of FTLY in tea plantations under different irrigation treatments and conducted a comprehensive analysis of the regression models’ performance using leave-one-out cross-validation (LOOCV). The results showed that the SW treatment (straw mulching combined with drip irrigation) performed the best among all irrigation treatments, with the highest model fit (R² = 0.87) and relatively low prediction errors (RMSE = 289.28 kg/ha, MAE = 260.34 kg/ha), indicating its high reliability and stability in water management and yield prediction.

Although other water-saving irrigation treatments also demonstrated good yield enhancement potential in certain years, the SW treatment exhibited the most stable and optimal performance across all seasons, proving the significant advantages of integrated water and fertilizer-saving irrigation techniques in improving tea production. Future research should further explore the long-term impacts of different irrigation schemes on soil moisture dynamics and tea plant growth to optimize irrigation strategies.

However, there are certain limitations to this study. The research was conducted only in the Rizhao tea plantation, and the models developed may be limited in applicability to other regions with different climate and soil conditions. Future studies should validate the models in different locations. Additionally, this study primarily relies on empirical data and does not fully incorporate the physical processes between evapotranspiration and yield. Future research could improve model accuracy by incorporating more complex mechanistic models.