Spatiotemporal Simulation of Soil Moisture in Typical Ecosystems of Northern China: A Methodological Exploration Using HYDRUS-1D

Abstract

Global climate change has intensified the frequency and severity of drought events, posing significant threats to agricultural sustainability, particularly for water-sensitive crops such as tea. In northern China, where precipitation is unevenly distributed and evapotranspiration rates are high, tea plantations frequently experience water stress, leading to reduced yields and declining quality. Therefore, accurately simulating soil water content (SWC) is essential for drought forecasting, soil moisture management, and the development of precision irrigation strategies. However, due to the high complexity of soil–vegetation–atmosphere interactions in field conditions, the practical application of the HYDRUS-1D model in northern China remains relatively limited. To address this issue, a three-year continuous monitoring campaign (2021–2023) was conducted in a coastal area of northern China, covering both young tea plantations and adjacent grasslands. Based on the measured meteorological and soil data, the HYDRUS-1D model was used to simulate SWC dynamics across 10 soil layers (0–100 cm). The model was calibrated and validated against observed SWC data to evaluate its accuracy and applicability. The simulation results showed that the model performed reasonably well, achieving an R² of 0.739 for the tea plantation and 0.878 for the grassland, indicating good agreement with the measured values. These findings demonstrate the potential of physics-based modeling for understanding vertical soil water processes under different land cover types and provide a scientific basis for improving irrigation strategies and water use efficiency in tea-growing regions.

1. Introduction

Global climate change is intensifying, accompanied by an increased frequency and intensity of extreme weather events such as droughts, which have profound impacts on soil moisture (SM) and the soil environment. Meanwhile, rising atmospheric carbon dioxide (CO₂) concentrations may alter plant transpiration and strengthen the coupling between SM and evapotranspiration [1], directly affecting crop growth. Studies have shown that total global SM storage and actual evapotranspiration are projected to decrease in the future [2]. Low temperatures and droughts, two major environmental factors affecting global horticultural crop yields and geographic distribution, slow down the growth rate of tea plants and negatively affect the crop by influencing plant physiological processes such as photosynthesis [3]. Therefore, with global climate change, tea plantation areas need to adapt to more variable climatic conditions [4,5].

As a high-value-added cash crop, tea tree (Camellia sinensis) is best suited to cultivation in low-latitude regions with warm and humid climates [6,7]. The optimal global-scale environmental conditions for tea cultivation are an annual precipitation of 1000–2000 mm, a mean temperature of the driest season ranging from 12–20 °C, soil pH between 4.5 and 5.5, small diurnal temperature variation, and humid mountainous areas at high altitudes [8]. Rizhao City (Shandong Province, China) is a typical mid-latitude oceanic climate region with a large temperature difference between day and night, which is recognized by the world’s tea scientists as one of the three major coastal green tea cities [9]. Rizhao green tea is one of the most renowned teas in northern China and has gained significant recognition globally. However, due to the dry climate and low rainfall in Rizhao, coupled with climate warming and water shortages, the constraints of soil drought on tea production have become increasingly prominent. Especially for the cultivation of young Rizhao green tea, soil drought brings more obvious challenges and requires long-term, fine-tuned management during cultivation. Drought-resistant cultivation has become a key research focus in Rizhao green tea production.

Currently, drought-resistant cultivation in tea gardens focuses more on physiological and biochemical response mechanisms and drought-resistant quality selection and breeding [10] and lacks in-depth studies on SM conditions in the root system of field-grown tea trees. Soil water content (SWC), as a key link in the soil–plant–atmosphere continuum, is a key variable in regulating and predicting hydrological processes, influencing the process of energy and water exchange at the surface, and is a key factor in the growth of tea plants and ecosystem health [11]. Changes in tea yield are mainly driven by the daily dynamics of SM [12]. The process of water movement in the unsaturated zone is extremely complex due to soil heterogeneity and short-term variations in atmospheric boundary conditions at the soil surface. At present, high-precision simulation of water dynamics at different soil depths for young tea trees at the field scale still needs to be explored in depth.

Recent studies have increasingly focused on simplified or data-driven approaches to improve the applicability and efficiency of SM simulation at large or regional scales. For instance, Wang et al. (2025) developed a conceptual model based on simplified mathematical relationships and empirical formulas, which significantly enhanced the efficiency of simulating soil water–heat–salinity dynamics and crop growth processes while maintaining reasonable accuracy [13]. Wu et al. (2024) estimated soil water content using classification and regression tree (CART) methods, identifying precipitation (P) and accumulated precipitation (AP) as critical factors for estimating shallow-layer (≤50 cm) soil water content [14]. Du et al. (2024) constructed estimation models for SM and salinity using six machine learning algorithms—including backpropagation neural networks, support vector regression, partial least squares regression, random forest, gradient boosting regression trees, and XGBoost—and demonstrated satisfactory accuracy and application potential [15]. Although these methods offer clear advantages in large-scale estimation and modeling efficiency, they still present certain limitations in capturing water dynamics at the local scale within crop root zones, simulating unsaturated zone physical processes, and resolving vertical soil heterogeneity [16]. In particular, for drought-sensitive and vertically stratified young tea plantations in northern regions, detailed stratified simulation of SM still relies on physically based models with clear mechanisms and controllable parameters.

Previous researchers have conducted in-depth studies on SM transport at the field scale. These studies typically focus on specific farmlands or experimental plots, requiring the use of physical-based models with high spatial resolution to investigate the spatiotemporal dynamics of SM under diverse climatic conditions and complex scenarios, while fully accounting for soil heterogeneity and local hydrological processes. By incorporating variables such as rainfall, temperature, humidity, soil type, and soil temperature, the HYDRUS-1D model is well-suited to quantify soil water composition, transport, and balance under different cropping systems, and to optimize irrigation strategies based on water supply and demand dynamics [17,18,19]. The HYDRUS-1D model was used to simulate long-term SM changes in the 0–1 m and 1–4 m soil horizons on the Loess Plateau of China from 1970 to 2060, and it was found that the model simulated deeper SM better than shallower soil, and that the model’s simulation accuracy in shallower soils (especially in the upper 0.6 m soil horizon) was low [20]. The HYDRUS-1D model incorporates soil hydraulic properties to simulate the movement of water through the soil profile, including water infiltration and evaporation [21]. Alternatively, regional monolayer soil hydraulic properties (SHPs) can be optimized using inverse modeling of the HYDRUS-1D model, which also highlights the spatial variability of SHPs [22,23]. The refinement of the vertical level of SM simulation using the HYDRUS-1D model by previous authors is still insufficient, and there are relatively few studies on the refinement of SM simulation in young tea plantations. Existing studies have confirmed the applicability of the HYDRUS-1D model—based on the Richards equation—in a variety of crops and ecosystems. The Richards equation is widely regarded as a numerical “truth” for simulating infiltration processes, and its solution is often used to validate more simplified models [24]. However, the HYDRUS-1D model still faces several critical challenges in practical applications, such as its complex model structure, high parameterization requirements, and susceptibility to convergence issues [25]. In addition, under arid or rapidly drying conditions in the unsaturated zone, the HYDRUS-1D model tends to significantly underestimate soil evaporation [26]. Currently, research on the application of HYDRUS-1D to simulate vertically stratified SM dynamics in typical northern coastal young tea plantations remains scarce. Although the HYDRUS-1D model is an effective tool for simulating vertically stratified soil water dynamics and provides valuable support for designing irrigation schedules and fertilization regimes in young tea plantations, issues related to parameter localization and boundary condition adaptation have not yet been systematically addressed. In the drought-sensitive and vertically heterogeneous soil environment of young tea plantations in Rizhao, further efforts are required to perform localized parameter inversion and refine boundary condition settings to ensure reliable simulation results.

The HYDRUS-1D model is an effective tool for irrigation and fertilization management in tea plantations, but its application in young tea plantations has been limited, and the accurate simulation of long-term stratified SM dynamics under field conditions remains insufficient. In this study, the natural rainfall method was used to determine the field water-holding capacity of different soil layers. Combined with the physical properties of the soil, the HYDRUS-1D model was employed to simulate the dynamic changes in soil water content and surface water fluxes in the 0–100 cm soil profile with high vertical resolution. The simulation results provide important data support and theoretical guidance for the development of precise irrigation schedules and for improving water-saving and drought mitigation strategies in tea plantations.

2. Materials and Methods

The proposed SWC simulation framework, combining a physics-based model and a data assimilation method, consists of two key steps (Figure 1).

STEP 1: Data Preparation and Processing

This step involves collecting and processing essential data, ensuring accurate model inputs. The required data typically include (1) meteorological data, comprising daily precipitation, air temperature (mean, minimum, and maximum), relative humidity, solar radiation, and wind speed, which were obtained from the local meteorological station (Rizhao Meteorological Bureau); (2) soil physical properties, including soil texture, bulk density, and saturated water content (θs), were determined through laboratory analysis. The remaining parameters, including residual water content (θr), saturated hydraulic conductivity (Ks), and van Genuchten parameters (α and n), were estimated via inverse modeling using HYDRUS-1D, constrained by the measured θs and observed SWC data; (3) vegetation root system characteristics (root depth, root density, and transpiration coefficient).

In this research, three years of continuous field monitoring (2021–2023) in a northern coastal tea plantation and grassland were conducted to collect real-time SWC and meteorological data. Additionally, soil hydraulic parameters were estimated via inverse modeling, using measured field capacity (FC) and soil texture analysis as constraints.

STEP 2: Construction of the Physics-Based Model

At this stage, a mechanistic model is selected and constructed to simulate SWC dynamics under different conditions. This process involves defining boundary conditions such as precipitation, evaporation, and root water uptake, optimizing and calibrating model parameters using observed SWC data, and executing model simulations to predict soil water transport.

For this research, the HYDRUS-1D model was chosen to construct the SWC simulation model. The model was calibrated and validated using measured SWC data from the tea plantation and grassland, with accuracy assessed using R² (coefficient of determination), RMSE (root mean square error), and MAE (mean absolute error) [27].

2.1. Overview of the Experimental Area

The experiment was conducted at the National Water-Saving Irrigation Production Training Base of Shandong Water Conservancy Vocational College, located in Rizhao City, Shandong Province (N 119.560577°, E 35.462755°). The area has a continental climate with a warm temperate zone and semi-moist monsoon characteristics [28], as illustrated in Figure 2. The area receives an average annual precipitation of 869.3 mm, with more than 70% concentrated between June and September, exhibiting uneven spatiotemporal distribution and frequent alternation between drought and flood events [29,30]. The average runoff depth in Rizhao City is 243.5 mm. The average water surface evaporation is 985 mm, while land evaporation is 570 mm, with the highest evaporation typically occurring in May [31].

2.2. Experimental Design

To ensure the normal growth of young tea trees in the experimental tea garden, the soil of the experimental tea garden was excavated and backfilled in situ from March to July 2021, with a backfilling depth of 1 m [32]. The soil in the study area was classified as Haplic Luvisol according to the World Reference Base for Soil Resources [33]. The measurement depth was set to 0–100 cm, based on both field surveys with local tea farmers in Rizhao and previous studies. According to these sources, this depth encompasses the main distribution zone of tea tree feeder roots, where most water and nutrient uptake occurs. This depth also represents the typical range in northern Chinese tea gardens where SM and organic matter are most concentrated, making it highly relevant for studying the growth conditions of tea trees. The average bulk density of the tea garden soil at a depth of 0–100 cm was 1.55 g/cm³, determined using the ring knife method according to NY/T 1121.4–2006 [34]. The average organic matter content was 17.93 g/kg, measured by the potassium dichromate oxidation–external heating method in accordance with HJ 615–2011 using a visible spectrophotometer (722G, Shanghai Spectrum Instruments Co., China) [35]. High-quality clonal tea seedlings propagated by cuttings, which were strong and free of pests and diseases, were selected for planting with a spacing of 50 cm between rows. Tea gardens with uniform growth and level topography, along with adjacent grasslands, were selected for monitoring SWC and temperature. Both the tea plantation and grassland plots were arranged in a north–south orientation, with each treatment replicated three times. Each plot was equipped with one soil moisture and temperature sensor, installed at the center, which recorded data at 10 soil depths (10, 20,..., 100 cm) (Figure 2). Measurements were taken once per hour throughout the monitoring period. The grass, as a herbaceous plant, is the crop that covers the largest area in the construction of ecological tea gardens, belongs to the same perennial crop as the tea tree, and the growth process has a similar seasonal growth cycle, which is suitable for long-term monitoring and research, and the grass near the tea plantation was selected as the control group, with an average bulk density of the grass at 0–100 cm depth of 1.68 g/cm³, and an average organic matter content of 20.12 g/kg. Drip irrigation was used in the tea plantation, and sprinkler irrigation was used in the grassland, which were irrigated and fertilized according to the real production management experience to ensure the growth of the crops, and the amount of irrigation water was measured in each irrigation. During the overwintering period, a plastic covering was used in the tea plantation to protect the young tea plants.

2.3. Definition and Measurement of Indicators

2.3.1. Soil Water Content (SWC)

In this study, soil water content (SWC) is used as the quantitative indicator of SM. It refers to the volumetric ratio of water volume to total soil volume, typically expressed in cm³/cm³ or as a percentage. SWC is the primary variable both measured and simulated in the modeling process. It serves as the quantitative indicator of SM and is the variable measured and simulated in the modeling process. A multi-depth soil moisture and temperature sensor (Insentek, Beijing Oriental Ecological Technology Ltd., Co., Beijing, China) with 10 integrated measurement nodes was used in each plot. The sensor was vertically inserted into a 1-meter-deep borehole at the plot center, enabling simultaneous hourly monitoring at 10 soil depths (10 to 100 cm). The Insentek sensor was capable of measuring surface temperature, soil water content (SWC), electrical conductivity, and temperature at multiple depths on an hourly basis. All sensors were calibrated in situ using the gravimetric (drying) method after installation to ensure measurement accuracy.

2.3.2. Field Water Holding Capacity (FC)

Field capacity (FC) is the amount of water the soil retains after excess gravitational water has drained, and the downward movement slows significantly. It represents the upper limit of plant-available water and is typically measured 24–48 h after heavy rainfall or irrigation under natural drainage conditions, expressed as volumetric soil water content (e.g., cm³/cm³ or %). FC was determined using the natural rainfall method. The time period with natural rainfall exceeding 80mm was selected, and measurements were taken after excess gravity water had drained, and the soil reached FC within 24–48 h following the rainfall [36]. FC was then calculated by summing and averaging the daily 24 h SWC data, as measured by the Insentek sensor.

2.3.3. Meteorological Data

Evapotranspiration was calculated using the Penman equation, and meteorological data such as precipitation were provided by the micro-meteorological station established at the training base [37].

2.4. Construction of Physics-Based Model

2.4.1. Unsaturated Seepage Modeling

In the context of a vertical coordinate system (with positive values in the downward direction), the motion of unsaturated water flow in soil can be represented by the one-dimensional Darcy–Richards equation [38]:

$$C{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }t}}={\displaystyle \frac{\mathsf{\partial }}{\mathsf{\partial }z}}\left[K\left({\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }z}}-1\right)-S(z,t)\right]$$

(1)

where C (1/m) is the SM difference (=dθ/dh), h (m) is the soil water potential, z (m) is the vertical distance from the surface, K (m/s) is the unsaturated hydraulic conductivity, S (z, t) (m/s) is the source-sink term, which is mainly considered as the amount of root water uptake, and t (s) is the time.

The soil water retention curve and hydraulic conductivity function are commonly described by the Mualem–van Genuchten (MVG) model [39]:

$$C{\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }t}}={\displaystyle \frac{\mathsf{\partial }}{\mathsf{\partial }z}}\left[K\left({\displaystyle \frac{\mathsf{\partial }h}{\mathsf{\partial }z}}-1\right)-S(z,t)\right]$$

(2)

$$S_e={\displaystyle \frac{\theta -{\theta }_r}{{\theta }_s-{\theta }_r}}=\left\{\begin{array}{c}{\displaystyle \frac{1}{{\left(1+|\alpha h{|}^n\right)}^m}},h<0\\ 1,h\ge 0\end{array}\right.$$

(3)

$$C=\left\{\begin{array}{c}mn\alpha \left({\theta }_s-{\theta }_r\right)S_e^{1/m}{\left(1-S_e^{1/m}\right)}^m,h<0\\ S_s,h\ge 0\end{array}\right.$$

(4)

$$K=\left\{\begin{array}{c}{K}_s{S}_e^{1/2}{\left[1-{\left(1-S_e^{1/m}\right)}^m\right]}^2,h<0\\ K_s,h\ge 0\end{array}\right.$$

(5)

In the equation, $\theta$ (m³/m³) represents SWC with subscripts, $s$ and $r$ denote saturated and residual water contents, respectively, $K_s$ (m/s) is the saturated hydraulic conductivity, $\alpha$ (1/m) is a scaling parameter, which is inversely related to the air entry pressure, and n and m are dimensionless parameters related to the shape of the curve, with $m=1-{\displaystyle \frac{1}{n}}$. $S_e$ represents the degree of saturation, and $S_s$ (/m) is the specific storage of the saturated soil.

2.4.2. HYDRUS-1D Model

The computational processes and input conditions involved in the HYDRUS-1D model are as follows: The one-dimensional saturated–unsaturated soil water movement equation is calculated based on the Richards equation [40]. In HYDRUS-1D, the crop evapotranspiration (ET_c) is calculated based on the FAO-recommended Penman–Monteith method [41].

In this study, the calibration period for the grassland is from 26 March 2021 to 20 August 2023, and the validation period is from 21 August 2023 to 25 November 2023. The calibration period for the tea plantation is from 3 June 2022 to 20 August 2023, and the validation period is from 21 August 2023 to 31 December 2023. The model simulates the temporal variation of SWC at 10 soil depths (every 10 cm) below the surface down to 100 cm.

A variable time step method is used, with the time step adjusted based on the convergence iteration count [42]. The initial time step is set to 0.001 days, with a minimum time step of 1 × 10⁻⁵ days and a maximum of 5 days. The upper boundary condition for the simulation is an atmospheric boundary with runoff, while the lower boundary condition is a free drainage boundary.

Using the ‘Water Flow–Soil Hydraulic Parameters’ function of the HYDRUS-1D model, the soil hydraulic parameters were optimized and determined by obtaining daily average SWC from measured data and performing inverse modeling based on FC, soil particle distribution, and bulk density to achieve the best fit between the measured and simulated data. This approach allowed for the determination of the soil hydraulic parameters for 10 soil layers (0–100 cm) (

Table 1. Soil hydraulic parameters.

Vegetation	Soil Depth (cm)	θr	θs	Alpha	N	Ks (cm/d)	l
Young tea plants	0–10	0.105	0.350	0.0365	1.070	329	0.5
	10–20	0.095	0.298	0.0115	1.125	361	0.5
	20–30	0.095	0.362	0.0179	1.180	382	0.5
	30–40	0.075	0.365	0.0119	1.150	358	0.5
	40–50	0.085	0.356	0.0115	1.125	360	0.5
	50–60	0.095	0.361	0.0134	1.158	264	0.5
	60–70	0.085	0.370	0.0135	1.112	125	0.5
	70–80	0.085	0.375	0.0135	1.075	363	0.5
	80–90	0.057	0.388	0.0132	1.146	112	0.5
	90–100	0.060	0.530	0.0175	1.155	384	0.5
Grassland	0–10	0.025	0.298	0.0236	1.785	180	0.5
	10–20	0.048	0.300	0.0238	1.442	190	0.5
	20–30	0.040	0.358	0.0185	1.42	4.6	0.5
	30–40	0.040	0.326	0.0133	1.529	1.95	0.5
	40–50	0.061	0.340	0.0101	1.441	1.5	0.5
	50–60	0.050	0.3501	0.0124	1.381	7.1	0.5
	60–70	0.045	0.325	0.0048	1.400	4.7	0.5
	70–80	0.028	0.335	0.0085	1.232	4.2	0.5
	80–90	0.023	0.324	0.0021	1.468	4.0	0.5
	90–100	0.083	0.298	0.0085	1.231	4.5	0.5

Notes: θ_r represents the residual water content; θ_s represents the saturated water content; Alpha is the inverse of the air entry value; N is the pore size distribution coefficient; K_s is the saturated hydraulic conductivity; and l is the pore connectivity coefficient.

).

2.4.3. Model Validation

The accuracy of the model was evaluated based on R², MAE, and RMSE. R² represents the degree of fit between the simulated and observed values. The closer MAE and RMSE are to zero, the smaller the bias between the simulated and observed values.

$$R^2={\displaystyle \frac{{\left[\sum _{i=1}^n(X_i-\overline{X})\times (Y_i-\overline{Y})\right]}^2}{{\sum _{i=1}^n(X_i-\overline{X})}^2\times {\sum _{i=1}^n(Y_i-\overline{Y})}^2}}$$

(6)

$$MAE={\displaystyle \frac{\sum _{i=1}^n\left|\left.X_i-Y_i\right|\right.}{n}}$$

(7)

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

(8)

where $X_i$ represents the observed value at the i-th time period, $Y_i$ represents the simulated value at the i-th time period, $n$ is the number of sample points, $\overline{X}$ is the mean of the observed values, and $\overline{Y}$ is the mean of the simulated values.

2.5. Data Processing and Statistical Analysis

Data processing and statistical analyses were performed using Microsoft Excel 2023 and Python libraries (e.g., NumPy 1.26.4; https://numpy.org, accessed on 10 June 2025). Figures were generated using Origin 2023 (OriginLab Corporation, Northampton, MA, USA). One-way analysis of variance (ANOVA) was applied to assess differences among treatments, followed by post hoc comparison using the Least Significant Difference (LSD) method at the 0.05 significance level (p < 0.05).

3. Results

3.1. Rainfall and Temperature Variations

The analysis of rainfall, temperature, and soil surface temperature from 2021 to 2023 revealed significant seasonal variations in both rainfall and temperature (Figure 3). Specifically, over 60% of the annual precipitation occurred during the summer months (June to September) in each year, indicating a clear pattern of rainfall concentration during the growing season. Temperature and rainfall peaked in the summer, with the highest temperatures and maximum rainfall occurring between June and August each year. During this period, temperatures remained above 20 °C, reaching over 30 °C at their highest. The lowest temperatures were recorded from December to February of the following year, with temperatures often dropping below 0 °C, reaching a minimum close to −10 °C. The daily average temperature ranged from −9.46 to 30.92 °C, the daily maximum temperature ranged from 18.04 to 35.0 °C, and the daily minimum temperature ranged from −15.49 to 30.0 °C.

From 2021 to 2023, the maximum daily rainfall was 26.7 mm, recorded in June 2022, which was classified as an extreme rainfall event based on regional thresholds. Rainfall was highest in June, July, and August, measuring 66.35 mm, 65.35 mm, and 48.30 mm, respectively. The high rainfall in June and July was primarily caused by monsoon influences and typhoon-induced heavy precipitation. Rainfall gradually decreased in autumn, with monthly totals of 44.75 mm in September, 8.94 mm in October, and 9.84 mm in November.

3.2. FC at Different Depths

As illustrated in Figure 4, FC of the tea garden soil was higher than that of the grassland soil at most depths, particularly in deeper layers. In the tea garden, FC ranged from 29.00% to 40.65% across the 20–100 cm profile, while in the grassland, FC varied between 22.50% and 28.55%. Significant differences in FC were observed among soil layers, as indicated by the different lowercase letters (p < 0.05). In the tea garden, a similar trend was observed, with the FC peaking at the 100 cm layer (40.65a) and reaching a minimum at 30 cm (29.00g). Significant differences occurred across multiple depths, such as 34.96b at 80 cm and 31.91d at 70 cm, indicating spatial variability likely influenced by root distribution and mulching practices. In the grassland, FC increased significantly with depth, with the 90 cm layer exhibiting the highest value (28.55a) and the 20 cm layer the lowest (22.50g). Notably, significant stratification was evident from 40 to 100 cm, reflecting the impact of soil texture and water-holding capacity in deeper profiles. These results demonstrate that both land use type and soil depth have a significant impact on FC, with tea garden soils generally exhibiting stronger water retention capacity than grassland soils. This stratified distribution provides a theoretical basis for developing depth-specific irrigation strategies under different vegetation types.

3.3. HYDRUS-1D Model Validation

The HYDRUS-1D model demonstrated strong performance in simulating SWC for both the tea plantation and grassland during the calibration period, achieving an R² value greater than 0.75, which is generally considered a good fit for SWC modeling. For the validation period, the R² value for the tea plantation was 0.739, indicating a marginal decline in model performance compared to calibration. In contrast, the R² value for the grassland during the validation period exceeded that of the tea plantation, indicating better model performance for the grassland (Figure 5). Overall, the model showed a high degree of explanatory power in capturing SWC dynamics in both ecosystems, with marginally better performance for grassland SWC. In terms of RMSE and MAE values, the grassland showed slightly larger RMSE and MAE than the tea plantation, but the prediction accuracy of the model for both vegetation types was high, with minimal error, indicating that the model was able to closely simulate the measured data.

The simulated SWC values for the tea plantation at different depths showed a strong correlation with the observed values, particularly at the shallow and deep layers, suggesting high simulation accuracy at these depths. However, in the middle layers (e.g., 30 cm, 40 cm, and 50 cm), the correlation between the simulated and observed values was relatively lower. This is likely due to the comprehensive covering of the tea plantation with plastic sheeting during the winter dormancy of the young tea plants, which effectively suppressed SWC consumption.

3.4. SWC Simulation at Different Depths

Figure 6 illustrates the comparison between observed and simulated SWC at various depths in the tea plantation and grassland. The results indicate that the SWC in the tea plantation is consistently higher than that in the grassland across all depths, with the most pronounced differences observed at the 10 cm and 40 cm soil layers. Overall, the SWC in the grassland exhibits a larger fluctuation range within the 0–80 cm depth compared to the tea plantation, whereas the 90–100 cm soil layers show a greater fluctuation range in the tea plantation than in the grassland. For the grassland, SWC fluctuations decrease with increasing soil depth, stabilizing in deeper layers. Conversely, in the tea plantation, significant SWC fluctuations are still observed in the 90–100 cm layers, where the average SWC is notably higher than in the 10–80 cm layers. These findings highlight the distinct SWC dynamics in the tea plantation and grassland, likely influenced by vegetation type, root distribution, and water management practices.

In the grassland, SWC is relatively low during winter and spring, primarily due to grass wilting in winter and the absence of surface vegetation cover, which leads to the formation of frozen soil layers. Additionally, low spring rainfall exacerbates spring drought conditions. At the 10 cm soil depth, which is characterized by a high organic matter content and root systems, the soil structure is more complex and highly susceptible to external environmental influences [43,44]. The observed and simulated SWC at this depth display similar fluctuation patterns in both the tea plantation and grassland, although deviations are evident, particularly at peak and trough values. The overall SWC values in the grassland are lower than in the tea plantation, likely due to differences in soil properties, vegetation cover, and irrigation practices. In the tea plantation, the observed SWC ranges from 26.51% to 34.82%, while the simulated SWC ranges from 28.51% to 33.97%. In the grassland, the observed SWC ranges from 2.09% to 30.58%, whereas the simulated SWC ranges from 4.98% to 29.80%.

At the 20 cm soil depth, the overall trends of observed and simulated SWC are consistent across both ecosystems, especially at peak values. However, the fluctuation range at this depth is significantly smaller than at 10 cm. In the tea plantation, the observed SWC ranges from 23.21% to 28.73%, while the simulated SWC ranges from 23.08% to 29.30%. In the grassland, the observed SWC ranges from 7.08% to 30.74%, whereas the simulated SWC ranges from 11.46% to 30.00%.

At the 30 cm soil depth, simulated SWC trends align well with observed peak values for both systems. In the tea plantation, the observed SWC ranges from 24.68% to 33.45%, while the simulated SWC ranges from 23.50% to 35.03%. For the grassland, the observed SWC ranges from 10.64% to 30.82%, and the simulated SWC ranges from 14.59% to 35.80%. Notably, simulated values at this depth are slightly higher than observed values, particularly later in the study period.

At the 40 cm soil depth, the SWC in the grassland remains generally lower than in the tea plantation, though the trends in peak values and their timing are comparable. In the tea plantation, the observed SWC ranges from 25.95% to 35.71%, while the simulated SWC ranges from 25.54% to 35.92%. In the grassland, the observed SWC ranges from 10.56% to 31.90%, whereas the simulated SWC ranges from 12.90% to 32.60%.

In deeper soil layers (50–60 cm), the simulated SWC for both the tea plantation and grassland exhibits consistent peak trends, with reduced frequency and amplitude of fluctuations compared to shallower layers. While the simulated and observed trends align relatively well at this depth, some discrepancies persist in regions with rapid fluctuations. In the tea plantation, observed SWC changes more gradually compared to simulated values. For the grassland, simulated SWC values are generally higher than observed values, especially later in the study period.

At the 70–80 cm soil depth, the observed and simulated SWC in the tea plantation show strong agreement, despite minor deviations in regions with greater fluctuations. Similar trends are observed in the grassland, with simulated values slightly exceeding observed values.

At the 90–100 cm soil depth, SWC in the tea plantation is higher than in the grassland. The simulated and observed SWC trends at this depth are generally consistent, but simulated values are slightly higher in both ecosystems. Notably, in the tea plantation, a significant deviation between simulated and observed SWC is observed at the 90 cm depth.

4. Discussion

4.1. Analysis of Factors Influencing SWC Dynamics

SM conditions are influenced by various factors, including precipitation, evapotranspiration, soil characteristics, and vegetation [45]. In this study, rainfall and temperature demonstrated clear seasonal variations, peaking in summer. This seasonal peak significantly increased SWC recharge; however, high temperatures led to increased evaporation, partially offsetting the recharging effect of rainfall. SWC in the topsoil layer (10 cm) of both the tea plantation and grassland showed substantial fluctuations in response to rainfall events. Regarding SWC at varying depths, fluctuations in the tea plantation SWC caused by rainfall and irrigation were smaller than those observed in the grassland, and the overall SWC in the tea plantation remained higher than in the grassland. Seasonal variations in SWC of tea garden were relatively minor at the 10–20 cm and 50–60 cm layers. This finding aligns with previous research, which observed that in tea plantation regions with annual rainfall exceeding 3000 mm, SWC content does not exhibit significant seasonal variation [12]. These findings provide valuable insights for improving irrigation strategies and soil water conservation practices in tea plantation regions with high annual rainfall.

Despite relatively low rainfall during the study period, the seasonal variation in SWC in the tea plantation was less pronounced compared to the grassland. This can be attributed to the tea plantation’s stronger water retention capacity, supported by the integrated effects of refined management practices. Previous studies have reported that soil moisture in tea plantations tends to exhibit greater spatial variability, with a more random and disordered moisture distribution compared to forests [46]. The soil moisture distribution in the tea plantation observed in this study also exhibited similar characteristics, which may further explain the unique pattern of minimal seasonal variation in SWC. Climatic variables such as temperature and precipitation have significantly different impacts on tea yield across seasons [47], and a comprehensive investigation into the impact of environmental factors on tea yield and quality in northern China requires further research.

FC is a critical factor for accurately determining plant-available water and plays a significant role in assessing soil suitability for crop growth and biomass production [48]. Due to soil heterogeneity, FC varies at different soil depths [49]. In agricultural production and soil management, targeted management is required based on soil characteristics at different spatial locations [50]. In this study, we precisely determined the variations in FC across different soil layers in both the tea plantation and the grassland. The results showed significant differences in FC at various depths, with the overall FC in the tea plantation being noticeably higher than that in the grassland. However, despite the higher FC in the tea plantation, its average organic matter content was lower than that of the grassland. Previous studies have indicated that soil organic matter typically enhances the soil’s water retention capacity [51], playing a crucial role in both SM retention and plant water utilization [52]. The higher FC observed in the tea plantation in this study is likely attributable to factors such as soil structure, texture, and vegetation cover, rather than just the influence of organic matter content. This indicates that, besides organic matter, other factors like soil mineral composition, particle distribution, and management practices also play crucial roles in soil water retention capacity. Estimating FC across different soil textures through large-scale regional SM monitoring networks is of great importance for drought warning and prediction, as well as for enhancing drought resilience in smart agriculture [53]. The monitoring and estimation methods used in this study provide a more accurate assessment of water retention capacity across different soil types, offering scientific support for agricultural management, particularly in the development of drought mitigation strategies. This also highlights the importance of integrating refined soil texture monitoring and management practices in the cultivation of young tea plantations, enabling a more effective response to the challenges posed by climate change.

Plant roots, as the preferred pathway for water infiltration into the soil, have a significant impact on SM distribution, volumetric water content, and soil water pressure head [19]. In this study, clonal tea plants propagated by cuttings were selected, which primarily absorb water from the 0–15 cm soil layer [54]. Due to their shallow root system and high demand for surface SM, drip irrigation in the tea plantation is beneficial for more precise management of surface SM. This irrigation strategy ensures an adequate water supply during the tea plants’ critical growth stages. Moreover, the surface soil in both the tea plantation and grassland is strongly influenced by rainfall and irrigation, which play a key role in root water uptake. Deep SM can be transported to the surface through capillary rise, supplying water to the roots and playing a crucial role in crop growth [55,56]. Previous studies have shown that in typical semi-arid grassland regions, the surface soil above 60 cm is the main water consumption zone, while soil below 60 cm serves as the primary water storage zone [57]. In addition, studies have shown that the 5 cm rapid change layer and the 10 cm active layer exhibit quick fluctuations in moisture, while SM below 20 cm remains relatively stable, only fluctuating during heavy or continuous rainfall [58]. In this study, the fluctuation trends and range of SWC at different depths in the tea plantation and grassland showed differences, primarily due to the combined effects of root distribution, soil properties, and irrigation supply.

Due to the high demand for surface SWC in tea plants, a drip irrigation system was employed in the tea plantation, which helps to more precisely manage surface SWC conditions and ensure adequate water supply during critical growth stages. However, research indicates that within the optimal SWC range for tea seedlings, the average correlation coefficient between SWC and seedling biomass growth is −0.63, suggesting that moderately reducing SWC may promote the growth of tea seedlings [59]. Therefore, while drip irrigation allows for precise management of surface SWC, in practical application, moderate control of irrigation amounts to avoid over-saturation is also an important strategy for improving the growth efficiency of tea seedlings.

In addition, the surface soil in both the tea plantation and grassland is influenced by rainfall and irrigation, further affecting the root water uptake process. Meanwhile, deep SWC rises to the surface through capillary action, providing an important source of water for root absorption. These findings further highlight the importance of root distribution, soil properties, and water supply methods in regulating SWC dynamics across different soil layers.

4.2. SWC Dynamics Simulation Based on HYDRUS-1D

In this study, we utilized the HYDRUS-1D model to simulate soil water movement across 10 soil layers (0–100 cm) in the tea plantation and grassland, based on the measured FC. The overall results demonstrated a high degree of model fit. During the calibration period, the simulated SWC for the tea plantation achieved an R² value of 0.806, with an RMSE of 0.01824 and an MAE of 0.0144. For the grassland, the SWC simulation yielded an R² value of 0.809, with an RMSE of 0.03411 and an MAE of 0.0278. Although the R² value for the SWC simulation in the tea plantation decreased slightly to 0.739 during the validation period, with an RMSE of 0.0220 and an MAE of 0.0166, reflecting a reduction in accuracy, this could be attributed to the complexity of SWC supply in the tea plantation and external environmental changes. However, the performance of the model improved during the grassland validation period, with the SWC simulation achieving an R² value of 0.878, an RMSE of 0.0473, and an MAE of 0.0425, likely due to greater fluctuations in grassland SWC, which the model was able to capture effectively. The observed higher soil moisture values in the 0–60 cm depth under winter plastic mulch and localized irrigation compared to the HYDRUS-1D model simulations may be attributed to the fact that the plastic mulch effectively reduces evaporation, allowing for greater moisture retention at the soil surface [60]. Additionally, localized irrigation provides extra water directly to the root zone, which contributes to higher moisture levels than predicted by the model [61].

Compared with previous studies, the simulation performance in this study falls within the range reported in the existing literature. For example, Guo et al. (2024) conducted research on irrigated maize fields in northern China and found that the HYDRUS model achieved R² values ranging from 0.74 to 0.93 during the calibration period and from 0.79 to 0.91 during the validation period, with RMSE values consistently between 0.01 and 0.03 cm³/cm³ [62]. The results are comparable to the grassland simulations in this study, indicating that the HYDRUS-1D model can achieve satisfactory simulation accuracy even in non-agricultural systems. In the mangrove swamp rice production system of Guinea-Bissau in West Africa, the HYDRUS-1D model also exhibited excellent performance in simulating both soil water content and electrical conductivity, with R² values exceeding 0.97, demonstrating the model’s strong ability to capture the variability observed in the measured data [63]. Although the study by Garbanzo demonstrated higher statistical accuracy, the present research better highlights the applicability of the HYDRUS model under non-ideal conditions, considering the greater complexity of the simulation environment, differences in vegetation types, and the influence of vertical heterogeneity. In this study, a constant Ks was adopted due to the considerations of data availability and model simplification. Nevertheless, previous studies have indicated that the temporal variability of Ks may exert different levels of influence on water balance components under various land use types. Guo et al. (2025) used the dual-porosity module of HYDRUS-1D to simulate soil water content and water balance components in maize fields and forest ecosystems [64]. Their findings revealed that the temporal dynamics of Ks had a significant impact on water balance components in maize fields, while the effect was relatively minor in forest systems. Therefore, in perennial vegetation systems such as those considered in this study, the assumption of a constant Ks remains applicable to a certain extent. However, incorporating observed Ks variability in future simulations, particularly for land types subject to frequent cultivation disturbances (e.g., tea plantations), may help further improve the model’s predictive accuracy and applicability.

In this study, the HYDRUS-1D model exhibited certain systematic biases during the dry winter–spring periods or phases of rapid fluctuation, indicating that its simulation accuracy is still influenced by climatic conditions, surface configuration, and the stability of soil hydraulic parameters. Kumar and Ojha (2025) noted that, during wheat and rice growth periods, the one-dimensional Richards model with constant soil hydraulic parameters (SHPs) often overestimates SWC under dry conditions and underestimates it during rainfall or irrigation events [65]. This is consistent with the slight overestimation of surface SWC observed in the tea plantation during the winter–spring season and its underestimation during the summer, suggesting that the model exhibits systematic deviations under unsaturated or rapidly drying conditions. Additionally, Ma et al. (2024) employed the HYDRUS model to simulate SWC dynamics in ridged paddy fields, confirming the model’s strong responsiveness to changes in cultivation structure [66]. Raj et al. (2019) found that the HYDRUS model performed better on south-facing slopes during wet periods and on north-facing slopes during dry periods, indicating its sensitivity to slope aspect under varying seasonal moisture conditions [67]. Although this study did not directly involve ridge tillage or slope aspect, the comparison between winter mulching in the tea plantation and natural exposure in the grassland also demonstrated the adaptability of the HYDRUS-1D model to different surface management practices. The R² value of 0.878 achieved during the grassland validation phase is comparable to the aforementioned studies, indicating that the model performs reasonably well under complex soil–vegetation–environment conditions. Previous studies have indicated that using numerical models and optimization algorithms, while accounting for the significant effects of terrain and soil texture on SWC dynamics, helps make more accurate irrigation decisions under complex field conditions [68]. In the future, SWC simulation results under different seasonal and surface management conditions can be integrated to optimize irrigation timing and water volume, thereby enhancing the practical applicability of the model in field conditions and promoting the effective implementation of precision irrigation management.

The application of the HYDRUS-1D model varies across different regions, vegetation types, and crops. For example, in arid inland delta oases, the SWC simulation results for groundwater-dependent vegetation showed good agreement with the measured values, with R² of 0.81, 0.89, and 0.78, respectively [69]. In the Lake Tana Basin in northwestern Ethiopia, the HYDRUS-1D model was used to simulate SWC dynamics for irrigated garlic and pepper crops, with R² values ranging from 0.64 to 0.77, RMSE from 0.021 to 0.063, and mean error (ME) between 0.0013 and 0.040 [70]. Similarly, the HYDRUS-1D model also demonstrated good performance in simulating SWC for sweet corn under tropical rainfed conditions [71]. At an agricultural experimental site in the Biđ region of eastern Croatia, the model performed excellently in most years, with R² values for water flow simulations exceeding 0.7 in 93% of the years [72].

The application of the HYDRUS-1D model in this research validated its adaptability and accuracy under different vegetation types and soil conditions. Although the model’s performance slightly declined during the validation period in the tea plantation, overall, it was able to effectively capture the dynamic changes in SWC, particularly in cases where grassland SWC exhibited significant fluctuations. This indicates that, under complex field conditions, the appropriate use of the HYDRUS-1D numerical model, combined with considerations of terrain and soil texture, can assist in making more precise irrigation decisions. Additionally, in some soil layers, the simulated values tended to be lower overall, suggesting certain limitations of the HYDRUS-1D model in accurately simulating water content at different soil depths. Future research should conduct model validation at multiple locations to expand its applicability, thereby further enhancing the adaptability of the SWC prediction model under different climatic conditions and soil types. This will provide a solid foundation for optimizing smart irrigation systems, improving water resource efficiency, and offering more reliable data support for agricultural production. Furthermore, the promotion and application of the model will contribute to the development of precision agriculture technologies, particularly in water-scarce areas, enabling more sustainable agricultural management.

5. Conclusions

This study employed the HYDRUS-1D model to simulate and validate the soil water dynamics across 10 soil layers (0–100 cm) in a young tea plantation and grassland in northern China from 2021 to 2023. The results demonstrated high simulation accuracy, with R² values of 0.806 and 0.809 for the tea plantation and grassland, respectively, during the calibration period, and 0.739 and 0.878 during the validation period. The RMSE and MAE for the tea plantation were 0.0182 and 0.0144 in the calibration period, and 0.0220 and 0.0166 in the validation period. For the grassland, RMSE and MAE in the validation period were 0.0473 and 0.0425, respectively, indicating the model’s strong capability to capture SWC variability.

FC in the tea plantation ranged from 29.05% to 40.92% at 20–100 cm depths, higher than that of the grassland (22.50% to 28.55%). Despite higher surface organic matter content in the grassland, the tea plantation exhibited superior water retention in deeper layers, suggesting the influence of soil structure, texture, and surface management in addition to organic matter.

The model accurately simulated shallow and deep SWC dynamics in the tea plantation, while showing slightly reduced performance in mid-soil layers, likely due to winter plastic mulch coverage. In contrast, the grassland exhibited greater variability in soil water content across 10–80 cm depths, which the model captured effectively. These findings confirm that HYDRUS-1D is a reliable tool for simulating soil water dynamics under different vegetation systems and provide scientific support for precision irrigation. Future improvements may incorporate data assimilation techniques such as particle filtering to enhance the model’s performance under more complex climatic and management scenarios.