Dynamic Modeling of Soil Water Dynamics and Nitrogen Species Transport with Multi-Crop Rotations Under Variable-Saturated Conditions

Abstract

Excessive application of nitrogen (N) fertilizers in agriculture poses significant environmental risks, notably nitrate leaching into groundwater. This study evaluates soil water dynamics and the transport of urea, ammonium, and nitrate under variable-saturated conditions in a long-term experimental field in Croatia, Europe. Utilizing HYDRUS-1D and HYDRUS-2D models, we simulated water flow and nitrogen transformation and transport across six lysimeter-monitored locations over four years (2019–2023), incorporating diverse crop rotations and N addition. Key modeled processes included nitrification, urea hydrolysis, and nitrate leaching, integrating field-measured parameters and climatic conditions. The models achieved high reliability, with R² values for water flow ranging from 0.58 to 0.97 and for nitrate transport from 0.13 to 0.97; however, some cases reported lower reliability. Results revealed that nitrate leaching was influenced by precipitation patterns, soil moisture, crop growth stages, and fertilization timing. Peak nitrate losses were observed during early crop growth and post-harvest periods when elevated soil moisture and reduced plant uptake coincided. The findings highlight the importance of optimizing nitrogen application strategies to balance crop productivity and environmental protection. This research demonstrates the effectiveness of numerical modeling as a tool for sustainable nitrogen management and groundwater quality preservation in agricultural systems. It also indicates the need for further development by capturing some of the processes such as identification in the N cycle.

1. Introduction

Nitrogen (N) fertilizers are indispensable for modern agriculture, serving as critical inputs to sustain crop growth and meet the increasing global food demand. However, the excessive application of these fertilizers has significant environmental implications, notably nitrate leaching into groundwater and the emission of greenhouse gases [1,2,3,4]. Nitrate leaching, driven by its high mobility in soil, poses a particular risk to groundwater quality, as agricultural activities are responsible for up to 60% of global groundwater nitrate contamination [5]. This contamination threatens both human health—through conditions such as methemoglobinemia—and ecosystems, necessitating urgent interventions to optimize nitrogen management [6,7,8]. Despite extensive research on these processes, gaps remain in fully capturing the complexity of nitrogen transport and transformation in soils, indicating the need for the development and application of predictive modeling tools to improve management strategies and mitigate environmental risks.

Effective nitrogen management is challenging due to the complex interactions between soil properties, crop needs, and hydrological processes [9,10]. Current practices often fail to synchronize nitrogen availability with crop uptake, resulting in significant nitrogen losses and environmental degradation in agroecosystems [11]. This is further exacerbated in irrigated agriculture, where high water application rates accelerate nitrate leaching beyond the root zone [12]. Thus, understanding the dynamics of nitrogen transport in various soils under varying agricultural practices and environmental conditions is critical for developing strategies to mitigate its environmental impact while sustaining crop productivity.

In addition to extensive field trials and controlled conditions experiments, numerical modeling has emerged as a pivotal tool for addressing the challenges associated with nitrogen management in agriculture. Among the variety of available models, the HYDRUS suite (1D/2D/3D) [13] stands out as the most frequently used for simulating nitrate transport in agricultural settings. This widespread adoption is attributed to its versatility and accuracy in capturing the complex interactions of water and solute transport within the soil-plant-atmosphere system. Another notable model, LEACHM [14], has also proven effective, as highlighted in a comprehensive review by Rawat et al. [15], which underscores the continued relevance of these tools in understanding and predicting agricultural processes. Recent advancements in modeling have highlighted the importance of integrating diverse variables to understand nitrogen dynamics and mitigate environmental impacts. Liu et al. [16] developed the WHCNS (Water Heat Carbon Nitrogen Simulator) model to explore water and nitrogen management strategies in North China, demonstrating the utility of coupling soil water and nitrogen transport with crop-specific demands to optimize productivity and sustainability. While WHCNS provides valuable insights and captures relevant processes, including soil organic carbon (SOC) turnover and crop growth, the HYDRUS model used in this study offers superior versatility in simulating variably saturated conditions and complex soil-plant-atmosphere interactions, which are critical for nitrate transport studies. By focusing on these processes, HYDRUS allows for a more detailed understanding of the interactions between soil hydraulic properties, climatic conditions, and nitrogen transformations, making it particularly suitable for evaluating nitrate leaching in diverse agricultural systems.

The HYDRUS models are particularly valued for their robust capabilities in simulating water flow, solute transport, and nitrate movement under various field and climatic conditions [17,18,19,20,21]. These models have demonstrated utility in evaluating and optimizing nitrogen management strategies, including fertilizer application timing, irrigation scheduling, and drainage practices [22,23,24]. By extending the experimental results or as an optimization and predictive tool, HYDRUS supports the development of efficient nitrogen use practices, which can aid in minimizing nitrate leaching while maintaining crop productivity. Its integration into agricultural research underscores its importance as a decision-support tool for sustainable agriculture.

In the study by Azad et al. [25], the effective use of HYDRUS-2D as a simulation tool for optimizing fertigation management was presented, focusing on minimizing nitrate losses and aligning nitrogen availability with crop requirements. By coupling the simulation outputs with a convolutional neural network (CNN), the researchers developed a computationally efficient emulator to predict nitrate uptake and losses under varying irrigation and fertilizer regimes. The study evaluated daily potential evapotranspiration, irrigation water, and fertilizer inputs over recent days as parameters influencing nitrate uptake. While Azad et al. [25] primarily focused on fertigation optimization within controlled irrigation systems, our study expands the application of HYDRUS-1D and HYDRUS-2D to investigate nitrate transport under variably saturated field conditions with multiple crop rotations, including spatial redistribution of nitrate.

In the extensive field study, Chen et al. [26] applied the modified HYDRUS (2D/3D) model to investigate soil nitrogen dynamics and interspecies nitrogen competition in tomato-corn intercropping systems with varying spatial arrangements. Results indicated that the HYDRUS model accurately captured soil nitrogen dynamics, with normalized root mean square errors (nRMSE) between 2.8% and 12.4%. Results showed that IC2–2 (two rows of tomatoes intercropped with two rows of corn) achieved the highest nitrogen uptake and corn yield, making it the most sustainable system. The findings highlight HYDRUS’s precision in modeling nitrogen dynamics and offer practical insights for improving intercropping practices. Similarly, He et al. [27] investigated the interactions between irrigation practices and nitrogen fertilization on soil water dynamics and nitrate leaching. Their study revealed that irrigation scheduling significantly influences nitrogen mobility, with over-irrigation exacerbating nitrate leaching beyond the root zone. This aligns with the findings of our current research, which identified the critical role of precipitation patterns and soil water content in driving nitrate losses. While He et al.’s [27] work focused on managed irrigation systems, our study extends this understanding to natural rainfall scenarios, further underscoring the importance of timing and amount of nitrogen application. Together, these studies demonstrate the necessity of context-specific modeling approaches to address the multifaceted challenges of nitrogen management in agriculture.

The selected study site in our research is a long-term experimental trial (20 years) with continuous crop rotation focusing on environmental effects on soil and groundwater quality originating from conventional agricultural practices. In one relevant study Filipovic et al. [28] evaluated the performance of zero-tension plate lysimeters and the HYDRUS-2D model in simulating water flow and nitrate dynamics in silty-clay soils with a high groundwater table over four years (2007–2010) at the same site. Results showed that lysimeter efficiency was limited during high plant water demand and low groundwater conditions, causing water diversion. HYDRUS-2D effectively reproduced water and nitrate outflows, highlighting the impact of lysimeter plates on water and solute transport. The findings demonstrate the utility of HYDRUS-2D for understanding nitrate behavior and improving lysimeter experiments under field conditions. In a recent study at the same site, Krevh et al. [10] studied the long-term soil water and nitrate dynamics (2009–2020) with the aid of HYDRUS-1D. The HYDRUS-1D model successfully reproduced lysimeter outflows and nitrate dynamics, which was confirmed with high R² values (water: 93% above 0.7 and nitrate: 73% above 0.7), indicating the good performance of the model simulating nitrification chain reactions. Principal component analysis (PCA) was performed to identify the relationships among all soil properties and environmental characteristics. The results showed the complex interaction of soil hydraulic properties, precipitation patterns, plant uptake, and N application. At all locations, a decreasing trend of nitrate leaching over the investigation period was observed. The current study integrates HYDRUS-1D and HYDRUS-2D simulations with multi-year field data across diverse conditions and cropping systems. This approach captures nitrate leaching under variably saturated conditions while accounting for seasonal and interannual variability driven by agricultural practices and climate.

The outcome of the current study leverages the HYDRUS model to evaluate nitrogen transport and nitrate leaching under unsaturated soil conditions influenced by conventional agricultural practices. The study was performed during the period from 2019 to 2023 with multi-crop rotation and extensive monitoring of soil water dynamics and quality. The aim of the study was to combine numerical water flow and nitrification chain models with laboratory- and field-obtained data to assess model performance and reveal the most dominant process governing nitrate transport under variable saturated conditions. The research aims to provide actionable insights into nitrogen management strategies that minimize environmental risks while maintaining agricultural productivity. By integrating field data with simulation modeling, this study contributes to the growing body of knowledge on sustainable agriculture and environmental conservation.

2. Materials and Methods

2.1. In Situ Experimental Set-Up and Associated Properties

The study area is situated in the Biđ field region of eastern Croatia, spanning coordinates 18°25′ to 18°33′ E and 45°07′ to 45°11′ N. The experimental research was conducted between 2019 and 2023. Daily meteorological data were obtained from a nearby weather station located at 45°09′ N and 18°42′ E. The long-term averages (1981–2023) for annual precipitation and temperature in this region are 682.4 mm and 12.0 °C, respectively. The experimental study was carried out during a four-year period (2019–2023).

The study was performed at six selected locations, representing the dominant soil types in that area (gley soils). Soils were classified according to the World Reference Base for Soil Resources [29]: Luvic Stagnic Phaeozem Siltic (Horizons: Ap-Bt-Bg-C) locations L1, L2, and L6, Haplic Fluvisol Eutric Siltic (Horizons: Ap-A/Bw-Cg-Cr) location L3, and Haplic Gleysol Calcaric Eutric Siltic (Horizons: Ap-Bg-Cr-Cg) location L4 and L5. For the soil particle size distribution analysis, disturbed soil samples were sampled from the six locations in three repetitions and multiple depths (as specified in

Table 1. Particle size distribution, saturated water content (θ_s), bulk density, hydraulic conductivity (K_s), water retention at selected pressures, and soil texture at six locations (L1–L6) at the Biđ experimental site in Croatia (n = 3).

Location	Depth (cm)	Sand (%)	Silt (%)	Clay (%)	Texture	θs (cm3 cm−3)	Soil Bulk Density (g cm−3)	Ks (cm day−1)	Water Retention at Pressure (kPa)
									33	625	1500
L1	0–40	13	65	22	Silt Loam	0.38	1.59	11	0.34	0.22	0.20
	40–75	4	63	33	Silty Clay Loam	0.37	1.57	15	0.34	0.22	0.20
	75–105	14	54	32	Silty Clay Loam	Not measured
	105–150	5	69	26	Silt Loam
L2	0–30	9	67	24	Silt Loam	0.36	1.56	17	0.33	0.17	0.16
	30–75	2	61	37	Silty Clay Loam	0.37	1.55	12	0.39	0.31	0.28
	75–130	8	72	20	Silt Loam	Not measured
	130–200	13	69	18	Silt Loam
L3	0–40	6	60	34	Silty Clay Loam	0.37	1.49	14	0.35	0.22	0.19
	40–90	6	60	34	Silty Clay Loam	0.38	1.55	9	0.34	0.22	0.19
	90–130	6	63	31	Silty Clay Loam	Not measured
	130–170	10	67	23	Silt Loam
L4	0–25	3	56	41	Silty Clay	0.40	1.47	12	0.41	0.32	0.29
	25–80	2	57	41	Silty Clay	0.41	1.46	10	0.35	0.22	0.20
	80–110	4	64	32	Silty Clay Loam	Not measured
	110–150	5	69	26	Silt Loam
L5	0–30	5	54	41	Silty Clay	0.42	1.37	12	0.39	0.28	0.22
	30–70	3	54	43	Silty Clay	0.41	1.55	14	0.37	0.27	0.21
	70–100	3	54	43	Silty Clay	Not measured
L6	0–30	5	75	20	Silt Loam	0.43	1.56	16	0.29	0.22	0.17
	30–70	7	73	20	Silt Loam	0.44	1.39	12	0.29	0.22	0.17
	70–100	9	60	31	Silty Clay Loam	Not measured
	100–150	12	72	16	Silt Loam

). The particle size distribution was determined using the combination of sieving and sedimentation procedure [30]. For the measurements of bulk density and the soil hydraulic properties at L1–L6 locations, undisturbed soil samples (100 cm³) were taken from the first two soil horizons (Table 1). The saturated water content (θ_s) was measured using a saturation pan, and the points of the soil water content of the soil water retention curve (SWRC) were measured using a pressure plate apparatus [31] with applied pressure heads of 33 (field capacity), 625, and 1500 (wilting point) kPa. The saturated hydraulic conductivities (Ks) were measured using the constant head method [32]. The basic physical soil properties are presented in Table 1.

Soil hydraulic functions needed for water flow modeling were described using the van Genuchten–Mualem model [33], which is defined as follows:

$$\begin{gathered} \theta \left(h\right)={\theta }_r+{\displaystyle \frac{{\theta }_s-{\theta }_r}{{(1+{\left|\alpha h\right|}^n)}^m}}\mathrm{for}h<0 \\ \theta \left(h\right)={\theta }_s\mathrm{for}h\ge 0 \end{gathered}$$

(1)

$$K\left(h\right)={K_s{S}_e^l{(1-(1-S_e^{{\displaystyle \frac{1}{m}}})}^m)}^2$$

(2)

$$S_e={\displaystyle \frac{\theta -{\theta }_r}{{\theta }_s-{\theta }_r}}$$

(3)

$$m=1-{\displaystyle \frac{1}{n}};n>1$$

(4)

where θ(h) and K(h) are volumetric water contents [L³ L⁻³] and unsaturated hydraulic conductivities [L T⁻¹] at the soil water pressure heads of h (L), respectively; θ_r and θ_s denote residual and saturated water contents [L³ L⁻³], respectively; S_e is the effective saturation, K_s is the saturated hydraulic conductivity [L T⁻¹], α is the inverse of air-entry value or bubbling pressure [L⁻¹], n is the pore size distribution index, and l is the pore connectivity parameter. While the saturated water content, θ_s, was measured, the remaining parameters of the soil water retention curve (1) (θ_r, α, and n) were optimized using the RETC software Version 6.00 [34] (

Table 2. Optimized soil hydraulic parameters (RETC) used for numerical simulations and parameters obtained by laboratory methods (θ_s, K_s) for soils at selected locations (L1–L6) at Biđ experimental field in eastern Croatia.

	Depth (cm)	θr (cm3 cm−3)	θs (cm3 cm−3)	Ks (cm day−1)	α (cm−1)	n (-)
L1	0–40	0.0	0.38	11	0.00261	1.18
	40–75	0.0	0.37	15	0.00263	1.17
L2	0–30	0.0	0.36	17	0.0018	1.26
	30–75	0.0	0.37	12	0.00017	1.25
L3	0–40	0.0	0.37	14	0.00158	1.20
	40–90	0.0	0.38	9	0.00285	1.18
L4	0–25	0.0	0.40	12	0.00032	1.19
	25–80	0.0	0.41	10	0.0527	1.17
L5	0–30	0.0	0.42	12	0.00136	1.20
	30–70	0.0	0.41	14	0.00212	1.17
L6	0–30	0.0	0.43	16	0.01241	1.19
	30–70	0.0	0.44	12	0.05525	1.17

) by fitting measured data. Unsaturated hydraulic conductivities (2) were predicted using the θ_s, θ_r, and n values, the measured values of saturated hydraulic conductivities, K_s, and the pore connectivity parameter equaled to an average value for the majority of soils (l = 0.5) and used in numerous modeling studies using the HYDRUS suite [35].

2.2. Crop Rotation and Fertilizer Application During 2019–2023

During the study period, various arable crops were cultivated at the lysimeter locations using standard agronomic practices typical of conventional crop production in the broader region of Eastern Slavonia. Agricultural monitoring was meticulously conducted, with detailed records of sowing and harvest dates, as well as precise documentation of fertilizer applications, including type, timing, and quantity (

Table 3. Crop rotation, sowing, harvest, and fertilizer application dates at each of the lysimeter locations.

Year	Location	Crop	Sowing Date	Harvest Date	Fertilizer Application Date	Amount and Type of Fertilizer
2019	L1	Maize	10 May 2019.	30 October 2019.	15 May 2019.	250 kg/ha NPK (15:15:15)/120 kg/ha urea
					10 June 2019.	50 kg/ha urea
	L2	Triticale	15 November 2018.	10 July 2019.	10 March 2019.	200 kg/ha KAN
	L3	Alfalfa	2016.	Multiple times	/	/
		Wheat	20 October 2018.	20 July 2019.	15 March 2019.	300 kg/ha KAN
	L4	Sugar beet	17 March 2019.	18 September 2019.	10 March 2019.	400 kg/ha NPK (7:20:30)/250 kg/ha NPK (15:15:15)
		Winter Barley	10 October 2019.		20 May 2019.	100 kg/ha KAN
	L5	Wheat	25 October 2018.	14 July 2019.	15 March 2019.	300 kg/ha KAN
		Canola	15 September 2019.		10 September 2019.	300 kg/ha NPK (0:20:30)/150 kg/ha urea
	L6	Maize	10 May 2019.	30 October 2019.	15 May 2019.	250 kg/ha NPK (15:15:15)/120 kg/ha urea
					10 June 2019.	50 kg/ha urea
2020	L1	Soybean	25 April 2020.	01 October 2020.	24 April 2020.	300 kg/ha NPK (15:15:15)
	L2	Soybean	08 April 2020.	15 September 2020.	/	/
		Wheat	19 October 2020.		/	/
	L3	Alfalfa	2016.	Multiple times	/	/
	L4	Maize	15 April 2020.	03 October 2020.	13 April 2020.	150 kg/ha urea/400 kg/ha NPK (15:15:15)
						200 kg/ha KAN
		Wheat	20 October 2020.		10 May 2020.	300 kg/ha NPK 0:20:30
	L5	Winter Barley	25 October 2019.	28 June 2020.	10 March 2020.	100 kg/ha KAN
					15 July 2020.	40.000 l/ha slurry
					20 October 2020.	300 kg/ha NPK 7:20:30
	L6	Canola	20 September 2019.	28 June 2020.	10 April 2020.	150 kg/ha KAN
		Winter Barley	20 October 2020		30 September 2020.	300 kg/ha NPK 0:20:30
2021	L1	Soybean	15 April 2021.	07 September 2021.	10 January 2021.	250 kg/ha NPK (0:20:30)
		Wheat	29 October 2021.	-	18 October 2021.	400 kg/ha NPK (15:15:15)
	L2	Wheat	15 October 2020.	10 July 2021.	01 March 2021.	200 kg/ha KAN
					02 April 2021.	200 kg/ha KAN
					05 May 2021.	150 kg/ha KAN
	L3	Maize	25 April 2021.	9 September 2021.	25 April 2021.	120 kg/ha urea
					25 April 2021.	600 kg/ha NPK (15:15:15)
					01 June 2021.	150 kg/ha KAN
		Winter Barley	15 October 2021.	-	-	-
	L4	Maize	25 April 2021.	01 October 2021.	25 April 2021.	250 kg/ha urea
						500 kg/ha NPK (15:15:15)
						175 kg/ha KAN
	L5	Maize	20 April 2021.	20 August 2021.	18 April 2021.	350 kg/ha NPK (15:15:15)
					18 April 2021.	150 kg/ha urea
					20 May 2021.	100 kg/ha KAN
		Winter Barley	15 October 2021.	-	03 October 2021.	300 kg/ha NPK (15:15:15)
	L6	Winter Barley	15 October 2020.	20 June 2021.	26 March 2021.	300 kg/ha KAN
2022	L1	Wheat	20 October 2021.	03 July 2022.	20 February 2022.	180 kg/ha, KAN
					29 March 2022.	200 kg/ha, KAN
					10 November 2022.	300 kg/ha, NPK 0:15:15
	L2	Soybean	10 April 2022.	05 September 2022.	/	/
		Wheat	27 October 2022.	/	26 October 2022.	250 kg/ha NPK 0:20:30
	L3	Winter Barley	15 November 2021.	25 June 2022.	10 March 2022.	220 kg/ha KAN
						200 kg/ha NPK 15:15:15
	L4	Soybean	20 April 2022.	25 September 2022.	20 April 2022.	400 kg/ha NPK 15:15:15
						100 kg/ha urea
		Wheat	20 October 2022.	/	19 October 2022.	300 kg/ha NPK 0:20:30
						150 kg/ha urea
	L5	Winter Barley	10 October 2021.	05 July 2022.	20 February 2022.	100 kg/ha KAN
					10 August 2022.	Slurry 17.000 l/ha
	L6	Maize	15 April 2022.	01 October 2022.	15 April 2022.	150 kg/ha urea
						400 kg/ha NPK 15:15:15
						175 kg/ha KAN
2023	L1	Soybean	12 April 2023.	15.09.2023.	/	/
	L2	Wheat	25 October 2023.	/	/	/
	L3	Wheat	27 October 2023.	04 July 2023.	09 March 2023.	200 kg/ha KAN
					21 April 2023.	200 kg/ha KAN
	L4	Maize	24 April 2023.	05 September 2023.	15 April 2023.	175 kg/ha urea
					24 April.2023.	600 kg/ha NPK 15.15.15
					15 May 2023.	210 kg/ha KAN
		Winter Barley	25 October 2023.	/	/	/
	L5	Wheat	25 October 2023.	05 July 2023.	09 March 2023.	200 kg/ha KAN
					21 April 2023.	200 kg/ha KAN
	L6	Maize	22 April 2023.		18 April 2023.	Slurry, 15.000 l/ha
					22 April 2023.	300 kg/ha NPK 15:15:15
		Winter Barley	20 October 2023.		/	/

). In addition to mineral fertilizers, organic fertilizers (solid manure and slurry) were applied at some locations. However, due to substantial variability in their composition and concentrations, these nutrient sources were excluded from numerical modeling to ensure accuracy and consistency in simulation results. The rigorous monitoring approach ensured a high level of data reliability, enabling robust calibration and validation of the HYDRUS models.

2.3. In-Situ Lysimeter Experiments

Zero-tension lysimeters [36] (circular, ⌀ 0.5 m, height 0.05 m) were installed at six selected locations (L1–L6). Each location was equipped with a set of two lysimeters to allow for sample replication. The installation involved excavating a vertical trench to a depth of 2 m and creating a horizontal slot at a depth of 50 cm, corresponding to varying local groundwater levels, typical agricultural management practices (e.g., tillage, harvest), and the average depth of the main root mass within the crop rotation system (Figure 1).

A round lysimeter plate, filled with disturbed soil material from the installation horizon, was inserted into the slot, ensuring that the soil profile above the plate remained undisturbed. A PVC net was applied over the lysimeter plate for filtration, preventing small soil particles from washing out with the leachate. Outflow pipes connected the lysimeter plates to soil water collection containers positioned at the edge of the fields for easy access and sample retrieval. Leachate was collected according to significant precipitation events throughout the years. Soil water samples in replicates (2×) were taken from zero-tension lysimeters and monitoring wells, corresponding to the moisture conditions in the field. Concentrations of NO₃⁻ and NH₄⁺ were determined (e.g., [28]) by a continuous flow auto-analyzer (San++ Continuous Flow Analyzer, Skalar, Netherlands). All samples were filtered through 0.45 μm membrane filters and stored for up to 2 days at 4 °C without acid preservation to reduce the potential interference of the dissolved organic matter.

2.4. Modeling Water Flow and Nitrogen Dynamics

Both HYDRUS 1D and 2D models were utilized to simulate water flow and solute transport [13]. The two-dimensional model allowed representation of the unsaturated soil around the lysimeter plate and visualization of seasonal dynamics of water and nitrate dynamics. The Richards equation, which describes isothermal Darcian flow in a variably saturated rigid porous medium, is used in the model, which in general form may be written as follows:

$${\displaystyle \frac{\partial \theta }{\partial t}}=\nabla \left(K \nabla H\right)-{\mathrm{S}}_{\mathrm{w}}$$

(5)

where θ is the volumetric water content [L³ L⁻³], K is the unsaturated hydraulic conductivity [L T⁻¹], H is the hydraulic head [L], and S_w is a sink term, accounting for plant water uptake [T⁻¹], ∇ is the spatial gradient operator, and t is time (T). Analytical expressions proposed by [32] (Equations (1)–(4)) for the soil water retention curve, θ(h), and the hydraulic conductivity function, K(θ), are used in both models.

The partial differential equations governing nonequilibrium chemical transport of solutes involved in a sequential first-order decay chain during transient water flow in a variably saturated rigid porous medium are included and solved in the HYDRUS program. In our case (e.g., simulation of urea, ammonium, and nitrate transport), the solute transport equations, which separately solve the transport of each chemical species, are simplified as follows:

For urea:

$${\displaystyle \frac{\partial \theta c_1}{\partial t}}=\nabla \left(\theta D\nabla c_1\right)-\nabla \left(q{c}_1\right)-{\mu }_a\theta c_1-S_w{c}_1$$

(6)

For ammonium:

$${\displaystyle \frac{\partial \theta c_2}{\partial t}}+\rho {\displaystyle \frac{\partial s_2}{\partial t}}=\nabla \left(\theta D\nabla c_2\right)-\nabla \left(q{c}_2\right)-{\mu }_v\theta c_2-{\mu }_n\theta c_2+{\mu }_a\theta c_1-S_w{c}_2$$

(7)

For nitrate:

$${\displaystyle \frac{\partial \theta c_3}{\partial t}}=\nabla \left(\theta D\nabla c_3\right)-\nabla \left(q{c}_3\right)+{\mu }_n\theta c_2-S_w{c}_3$$

(8)

where c_i is the liquid phase concentration of the chemical species i (subscripts 1, 2, and 3 represent urea, ammonium, and nitrate, respectively), [M L⁻³], D is the dispersion coefficient tensor [L² T⁻¹], q is the volumetric flux density (cm/day) [L T⁻¹], ρ is the bulk density of the soil [M L⁻³], s₂ is the adsorbed concentration of ammonium [M M⁻¹], μ_a is the first-order reaction rate constant [T⁻¹] representing nitrification of urea to ammonium, μ_v is the first-order reaction rate constant [T⁻¹] representing volatilization of ammonium to ammonia, and μ_n is the first-order reaction rate constant [T⁻¹] representing nitrification of ammonium to nitrate. The relationship between ammonium in solution (c₂) and adsorbed (s₂) is described as follows:

$$s_2=K_d{c}_2$$

(9)

where K_d is the distribution coefficient for ammonium [L³ M⁻¹]. The dispersion tensor has the following components:

$$\theta D_{xx}={\epsilon }_L{\displaystyle \frac{q_x^2}{\left|q\right|}}+{\epsilon }_T{\displaystyle \frac{q_z^2}{\left|q\right|}}+\theta \tau D_0$$

$$\theta D_{zz}={\epsilon }_L{\displaystyle \frac{q_z^2}{\left|q\right|}}+{\epsilon }_T{\displaystyle \frac{q_x^2}{\left|q\right|}}+\theta \tau D_0$$

(10)

$$\theta D_{xz}={\epsilon }_L-{\epsilon }_T{\displaystyle \frac{q_x{q}_z}{\left|q\right|}}+\theta \tau D_0$$

where |q| is the absolute value of the volumetric flux density [L T⁻¹], q_x and q_z are the volumetric flux densities in x and z directions [L T⁻¹], ε_L and ε_T are the longitudinal and transverse dispersivities [L], respectively, D₀ is the coefficient of the molecular diffusion for solute in free water [L² T⁻¹], and τ is the tortuosity factor.

2.5. Initial and Boundary Conditions

Simulations were conducted for six previously described soil profiles (lysimeters 1, 2, 3, 4, 5, and 6) using the HYDRUS-1D and HYDRUS-2D programs. HYDRUS-1D was applied to model water flow within the zero-tension lysimeters. The soil profiles were modeled to a depth of 50 cm, with two distinct horizons present in each profile, whose boundaries varied by location. The depth of the surface soil layer and the hydraulic parameters are detailed in Table 1 and Table 2. For all scenarios, the initial soil water potential conditions were defined based on groundwater level (limnigraph; LIM-1 and LIM-4) observations, capturing the onset of the partially saturated zone. The readings from the limnigraphs were recorded daily on a cm scale using an automated system. Boundary conditions were carefully assigned to simulate realistic soil water dynamics. The seepage conditions were applied to model the percolating water collected by the lysimeter. The top boundary was set to atmospheric conditions, incorporating natural precipitation and evapotranspiration processes while accounting for surface runoff, while the lysimeter plate had been assigned with seepage face conditions. In addition to that, the 2D model had bottom boundary conditions corresponding to variable pressure head conditions based on the groundwater table oscillations. This approach ensured that the simulations reflected the hydrological interactions and environmental conditions of the study area accurately.

The simulations were performed separately for each of 4 years with particular crop rotation included as presented in Table 3. Evapotranspiration was calculated using the climatic data measured at the Gradište meteorological station. Both evaporation from the soil surface and transpiration by plants were calculated on a daily basis using the Penman-Monteith equation [37]:

$$E{T}_o={\displaystyle \frac{0.408∆\left(R_n-G\right)+\gamma \frac{900}{T+23}{u}_2(e_s-e_a)}{∆+\gamma (1+0.34u_2)}}$$

(11)

where ET_o is the reference crop evapotranspiration [mm day⁻¹], R_n is the net radiation at the crop surface [MJ m⁻² day⁻¹], G is the soil heat flux density [MJ m⁻² day⁻¹], T is the air temperature at 2 m height [°C], u₂ is the wind speed at 2 m height [m s⁻¹], e_s is the saturation vapor pressure [kPa], e_a is the actual vapor pressure [kPa], e_s − e_a is the saturation vapor pressure deficit [kPa], Δ is the slope vapor pressure curve [kPa °C⁻¹], and γ is the psychrometric constant [kPa °C⁻¹]. Root water uptake was simulated using the approach of Feddes et al. [38]. The calculated potential transpiration flux was then converted into actual root water uptake by combining the piecewise linear water stress response model proposed by Feddes et al. [38] (parameters: P0 [L], POpt [L], P2H [L], P2L [L], P3 [L]) and a root density function that accounts for the root density and growth (e.g., [39,40]).

The parameters needed for nitrogen species modeling were set as follows. The bulk densities were set according to Table 1. The longitudinal dispersivity along the direction of flow was taken as 5 cm, and the transverse dispersivity was taken as 0.5 cm. Molecular diffusion was neglected. The first-order reaction term representing nitrification of urea to ammonium (μ_a), which acts as a sink in Equation (6) and as a source in Equation (7), was 0.38 per day [16]. The first-order reaction term representing nitrification of ammonium to nitrate (μ_n), which acts as a sink in Equation (7) and as a source in Equation (8), was 0.2 per day [16]. The first-order reaction term for volatilization of ammonium to ammonia (μ_v) was 0.0552 per day [41]. The distribution coefficient for ammonium (K_d) is assumed to be 3.5 cm³g⁻¹ [17]. The unrestricted passive root uptake of urea, ammonium, and nitrate were assumed.

The initial ammonium and nitrate contents in soils were set as nitrogen concentrations in soil water (expressed in mmol cm⁻³) for each species. Constant values within the entire soil profiles, which were measured at the beginning of each year in lysimeter water samples, were applied. The following boundary conditions were applied for the nitrogen species transport simulation. No flux at both the lateral boundaries and the lysimeter plate sides and bottom, and a third type of boundary condition at the remaining boundaries were assumed as the boundary conditions. The concentration fluxes of nitrogen for all 3 species were calculated as follows. The applied amount of urea in kg ha⁻¹ was transformed into the concentration of nitrogen in measured/applied precipitation (mmol cm⁻³) by dividing it by the molar mass (0.06), multiplying it by 2 (2 atoms of nitrogen exist for every molecule of urea), and then dividing by the volume of water per hectare. Application of fertilizer NPK (in the form 15:15:15), which was defined also in kg ha⁻¹, was similarly transformed into the concentration of nitrogen for both species (ammonium or nitrate) in measured/applied precipitation (mmol cm⁻³), assuming the percentage of each species (9% of NH₄, 6% of NO₃) in the fertilizer, the volume of water per hectare, and the molar mass of the species (0.018 for NH₄ and 0.062 for NO₃).

Model validation and performance assessment were carried out with the coefficient of determination (R²):

$$R^2=⌈{\displaystyle \frac{\sum _{i=1}^N\left(O_i-\overline{O}\right)\left(P_i-\overline{P}\right)}{{\left[\sum _{i=1}^N{\left(O_i-\overline{O}\right)}^2\right]}^{\mathrm{0,5}}{\left[\sum _{i=1}^N{\left(P_i-\overline{P}\right)}^2\right]}^{\mathrm{0,5}}}}⌉$$

(12)

where O_i is observation, P_i is prediction, Ō is average observation, and $\overline{P}$ is average prediction, while N is the sample size.

3. Results and Discussion

3.1. Modeling Water Dynamics

The cumulative water outflows measured during the 2019-2023 period using zero-tension lysimeters under field conditions, e.g., variable saturated conditions, along with the simulated values obtained using the HYDRUS-1D model, are shown in Figure 2 and Figure 3. The numerical simulations for each model run covered a 365-day period, starting on January 1 each year. The model successfully captured the measured outflow volumes during the study period, as evidenced by the coefficient of determination (R²) (

Table 4. Representation of model performance, i.e., coefficient of determination (R²) between measured and simulated water outflows during the period from 2019 to 2023 at lysimeter locations L1–L6.

	2019	2020	2021	2022	2023
L1	0.58	0.88	0.83	0.92	0.60
L2	0.74	0.84	0.78	0.94	0.59
L3	0.72	0.97	0.76	0.87	0.44
L4	0.77	0.88	0.82	0.82	0.37
L5	0.83	0.67	0.90	0.75	0.45
L6	0.70	0.85	0.75	0.83	0.45

). The R² values ranged from 0.58 to 0.83 in 2019; from 0.67 to 0.97 in 2020; from 0.75 to 0.90 in 2021; from 0.75 to 0.94 in 2022; and from 0.37 to 0.60 in 2023, respectively. The outflow volumes in 2019 were L1 = 177.2 mm; L2 = 186.6 mm; L3 = 192.7 mm; L4 = 162 mm; L5 = 201.5 mm; L6 = 179.4 mm. The average outflow volume (183.23 mm) accounted for 25% of the total annual precipitation (717.4 mm) in 2019. The outflow volumes in 2020 were L1 = 123.92 mm; L2 = 149.67 mm; L3 = 173.91 mm; L4 = 153.23 mm; L5 = 187.17 mm; L6 = 160.13 mm. The average outflow volume (158 mm) accounted for 24% of the total annual precipitation (658.1 mm) in 2020. The outflow volumes in 2021 were L1 = 133.6 mm; L2 = 155.5 mm; L3 = 156.5 mm; L4 = 176.1 mm; L5 = 179.7 mm; L6 = 122.5 mm. The average outflow volume (154 mm) accounted for 23% of the total annual precipitation (670.5 mm) in 2021. The outflow volumes in 2022 were L1 = 105.8 mm; L2 = 148.3 mm; L3 = 135.9 mm; L4 = 141.5 mm; L5 = 151.1 mm; L6 = 105.2 mm. The average outflow volume (131.3 mm) accounted for 22% of the total annual precipitation (604.8 mm) in 2022. The outflow volumes in 2023 were L1 = 107 mm; L2 = 112.5 mm; L3 = 135 mm; L4 = 126.4 mm; L5 = 145.6 mm; L6 = 127.8 mm. The average outflow volume (125.7 mm) accounted for 18% of the total annual precipitation (679.6 mm) in 2023.

These results align with previous studies that have demonstrated the accuracy of HYDRUS models in simulating water flow in variably saturated soils under diverse agricultural settings [15,26,28]. The findings confirm the model’s robustness in replicating natural processes, as observed in field-scale studies across different soil textures, climatic conditions, and management practices [10]. The ability of HYDRUS-1D and HYDRUS-2D to capture the temporal and spatial variability of soil water movement, including seasonal fluctuations and precipitation-driven percolation, highlights its suitability for evaluating long-term soil hydrological processes. Furthermore, the strong agreement between simulated and measured outflows reinforces the effectiveness of HYDRUS models as predictive tools for optimizing irrigation and fertilization strategies. These results are consistent with recent studies that emphasize the critical role of numerical modeling in understanding the interactions between soil hydraulic properties, crop uptake, and nitrate leaching dynamics, thereby supporting informed decision-making for sustainable agricultural water and nutrient management. The results indicate that outflow was measured during the periods when wetter soil conditions and increased precipitation occurred. During the early growth stages of the cultivated crop, plant water demand (ET) was reduced, allowing a portion of the water to percolate into deeper soil layers. Additionally, the distribution and amount of precipitation were the main factors influencing outflow dynamics. The annual variation in outflow volumes underscores the significant influence of precipitation distribution and crop-specific evapotranspiration rates on water dynamics. For instance, outflows were higher during wetter years (e.g., 2019) and lower in drier years (e.g., 2022), a pattern consistent with observations in similar studies on agricultural fields using lysimeters and modeling approaches [10,18]. The figure also shows that model-simulated values were sometimes lower and sometimes higher than the measured outflow volumes collected by the lysimeters. These discrepancies can be attributed to the lack of continuous sensor-based field measurements, the limited number of field data points, and the precision of the installed gravitational lysimeters. As indicated by Zhu et al. [42], zero tension lysimeters have some limitations in capturing the water fluxes precisely due to the potential for bypass flow around the plate. It is also worth considering that the number of observations directly influenced the goodness of fit between model simulations and measured data. Monthly lysimeter outflow monitoring does not provide sufficient data to determine an exact water balance alone; however, it can provide valuable insights into water dynamics under studied conditions and serve as a database for model calibration. Nonetheless, the results from the calibrated and validated model can be considered relevant and reliable. The cumulative outflow values simulated with HYDRUS-1D primarily depended on precipitation and evapotranspiration rates. The observed relationship between outflow dynamics and precipitation patterns aligns with findings by [10], who emphasized the dominance of rainfall events in driving percolation and groundwater recharge in fine-textured soils. Furthermore, the influence of evapotranspiration on cumulative outflows corroborates the work of Hanson et al. [17], who demonstrated the critical role of plant water uptake in modulating deep drainage during active growth periods.

3.2. Modeling Nitrate Dynamics

The cumulative nitrate concentrations measured in collected outflow from field lysimeters and the simulated nitrate values using the HYDRUS-1D model are shown in Figure 4 and Figure 5. The figures include locations where mineral fertilization was applied during the study period. Due to unknown properties of some of the organic types of fertilizer, the study was focusing on inorganic N sources alone. The agreement between measured and simulated values was good, as evidenced by coefficients of determination (R²). The R² values ranged from 0.13 to 0.75 in 2019; from 0.78 to 0.97 in 2020; from 0.68 to 0.90 in 2021; from 0.65 to 0.95 in 2022; and from 0.48 to 0.94 in 2023, respectively (

Table 5. Representation of model performance, i.e., coefficient of determination (R2) between measured and simulated nitrate outflows during the period from 2019 to 2023 at lysimeter locations L1–L6.

	2019	2020	2021	2022	2023
L1	0.75	0.84	0.68	0.90	0.94
L2	0.74	0.87	0.90	0.95	0.70
L3	0.13	0.97	0.80	0.81	0.79
L4	0.36	0.90	0.82	0.81	0.90
L5	0.62	0.78	0.69	0.72	0.48
L6	0.57	0.83	0.83	0.65	0.98

). Slightly lower R² values at some locations were attributed to processes that were not captured by the model (such as denitrification) and the limited number of samples collected during particular years.

Similar to the modeling of water dynamics and lysimeter outflows, the simulated nitrate concentrations in outflows showed some variability compared to measured values. However, the model effectively captured the overall trend of nitrate leaching in the soil. The simulations demonstrate that the occurrence of nitrate in the lysimeters aligns well with the directly measured concentrations, highlighting the model’s suitability for assessing nitrate transport in soil under the described conditions. HYDRUS enables the simulation of nitrogen species through first-order degradation reactions, modeling the hydrolysis of urea to ammonium and its subsequent nitrification to nitrite and nitrate [43]. Discrepancies between modeled and experimental nitrate transport data can be attributed to several factors, including soil heterogeneity, neglected soil structure effects in intensively managed agroecosystems, nonuniform rainfall and fertilizer distribution, and inaccuracies in crop uptake or nitrogen transformation parameters. Brunetti et al. [40] identified first-order degradation coefficients, organic nitrogen production rates, and Feddes’ parameters as key drivers of nitrate leaching in their global sensitivity analysis of HYDRUS-1D, though their study focused on organic nitrogen from wastewater application. In contrast, this study emphasizes inorganic nitrogen from fertilizers, which exhibit more consistent transformation rates due to controlled content. Rainfed open-field agriculture poses additional complexities, as precipitation can enhance nitrate leaching and soil moisture, influencing microbial processes such as nitrification and denitrification through aerobic-anaerobic cycles and interactions with dissolved organic carbon from plant residues [44].

Furthermore, the model, with the current level of measured parameters, provides a sufficiently accurate description and quantification of nitrate transport through the soil in conditions of intensive agricultural production. These findings reinforce the critical need for integrating seasonal dynamics and agricultural practices into nitrate management strategies. As Rawat et al. [15] highlighted, understanding the interplay of precipitation, crop phenology, and fertilizer application is essential for mitigating nitrate leaching risks. The results also show that the highest nitrate concentrations occurred during the early growth stages of crops and at the end of the year when the crop is absent. These periods correspond to increased soil moisture, bare soil without crop cover, and the application of fertilizers. This emphasizes the importance of considering seasonal dynamics and agricultural practices in managing nitrate leaching risks.

After conducting simulations in the one-dimensional HYDRUS-1D model, additional simulations were performed in the two-dimensional HYDRUS-2D model with the results presented here for 2023. Figure 6 presents results from lysimeter L6 to evaluate the influence of the surrounding soil zone on water flow in the immediate vicinity of the lysimeter. The two-dimensional modeling also accounted for the saturated zone, specifically the groundwater level, which affected the saturation of surface and subsurface soil layers at certain locations (e.g., LIM-4). This approach aligns with findings from other studies that emphasize the advantages of two-dimensional modeling in capturing complex hydrological processes near saturation zones [13,28].

Figure 6 illustrates the simulated soil water potential values at lysimeter L6 during 2023, as generated by the HYDRUS-2D model. The figure shows changes in soil water potential throughout the year within the soil profile, where red colors indicate wetter conditions, and blue colors represent drier soil states. These dynamics were primarily influenced by precipitation and evapotranspiration. Extremely low groundwater levels did not contribute to wetting of the surface soil layer during 2023. The results are presented on a monthly basis for the entire year.

The simulations revealed extremely dry soil conditions during periods of low precipitation and low groundwater levels, highlighting the limited contribution of groundwater to surface soil moisture under such conditions. However, during November and December 2023, interactions between elevated groundwater levels and precipitation significantly influenced the rhizosphere zone, particularly within 100 cm of the soil surface. These conditions facilitated the leaching of contaminants into the groundwater, emphasizing the need for integrated water and nutrient management strategies. The use of HYDRUS-2D in this context demonstrated its utility in capturing spatial variability and groundwater-surface interactions, as well as providing insights into the effects of local conditions on water dynamics and contaminant transport.

The findings align with Karandish and Šimůnek [24], who showed that HYDRUS-2D effectively simulates soil water content, nitrate dynamics, and crop nitrogen uptake under different irrigation and nitrogen application scenarios. Their study highlighted that both optimal nitrogen application rates and irrigation levels (e.g., soil moisture state) are responsible and must be adjusted to achieve a balance between nitrogen use efficiency, economic returns, and environmental protection of groundwater. Partial root-zone drying (PRD) was identified as a superior strategy compared to deficit irrigation (DI), increasing nitrogen uptake and yield. Additionally, Chen et al. [26] demonstrated the precise capability of HYDRUS (2D/3D) in capturing nitrogen dynamics in tomato-corn intercropping systems. Their study revealed distinct nitrogen distributions, with NO₃⁻-N primarily accumulating in deeper soil layers (40–60 cm) due to its high mobility, while NH₄⁺-N remained concentrated in the upper 0–20 cm, similarly as found in our study (ammonium species transport was not presented due to low concentrations). The results additionally highlight the importance of spatial and temporal considerations in nitrogen management to reduce environmental risks and enhance agricultural productivity.

In summary, the combined insights from the present study and referenced works emphasize the value of HYDRUS-2D in designing efficient water and nutrient management strategies. By incorporating local precipitation patterns, soil properties, and crop-specific demands, this modeling approach offers practical tools for mitigating nitrate leaching, optimizing nitrogen application, and supporting sustainable agricultural practices.

Figure 7 illustrates the nitrate concentration values in the soil profile at lysimeter location 6 throughout 2023. At this location, mineral fertilization was applied as follows: NPK 15:15:15 (400 kg ha⁻¹) and urea (150 kg ha⁻¹) in April, and NPK 0:20:30 (300 kg ha⁻¹) in October. The initial nitrate concentration (day 0) was very low. Fertilization was applied on April 28th (day 119), which is evident in the profile by day 122. The fast transformation of ammonium to nitrate was observed (not shown due to very low concentrations), resulting an increase of the nitrate form of N subspecies leaching into deeper layers. The rapid transformation of ammonium to nitrate, leading to increased nitrate leaching into deeper soil layers, was observed (not shown due to very low ammonium concentrations). Similar findings were reported by Balkhi et al. [45], where the application of untreated wastewater for irrigation demonstrated significant nitrogen leaching, predominantly in nitrate form. Their study, using HYDRUS-2D, highlighted that nitrate leaching was more sensitive to irrigation volume than to fertilizer application, emphasizing the importance of optimized irrigation management to reduce nitrogen losses. The relatively slow movement of nitrates through the soil profile was observed throughout the year, primarily due to reduced precipitation and the low water permeability of the soil (Table 1, K_s = 16 cm day⁻¹). However, nitrate transport to the bottom of the soil profile (2 m depth), below the main root zone, was also evident.

Figure 8 illustrates the movement and leaching of nitrates through the soil profile during the study period, with observations at 60-day intervals throughout the year. The data show that nitrates leached through the soil profile for most of the year, albeit at lower concentrations. This behavior is largely due to the inability of nitrates to adsorb onto soil particles.

Differences between locations were observed, directly related to the amount and type of fertilizer applied (urea, KAN, and NPK) as well as the timing of nitrate distribution. At some locations, nitrate leaching occurred shortly after fertilization, leading to higher concentrations. At all locations, nitrates were detected, to varying degrees, at the bottom of the soil profile (50 cm depth). This suggests a potential risk of groundwater contamination. However, in a year with relatively low precipitation and low groundwater levels, nitrate transport was reduced. In the study by Gao et al. [46] the dual issue of nitrate leaching and storage in the vadose zone within intensive horticultural systems was highlighted. In kiwifruit production, more than 89% of nitrate in the 0–10 m vadose zone had leached beyond the root zone, representing three times the annual nitrogen input, leading to wasted resources and environmental concerns. While nitrate stored in the vadose zone can temporarily serve as a nutrient pool, it poses a significant risk of groundwater contamination [1], particularly in regions with shallow vadose zones, such as our study area.

It is important to emphasize that the dynamics of nitrate leaching are closely tied to the amount and distribution of precipitation and groundwater levels. This highlights a potential future risk of increased groundwater contamination, particularly in areas with surface (shallow) groundwater interaction. Preferential flow, where substances are transported through macropores during high-intensity rainfall events, could further exacerbate nitrate leaching and transport in structured soils [47,48].

These findings underline the importance of managing fertilization practices and monitoring precipitation patterns to mitigate risks of nitrate pollution in groundwater. The slow migration of nitrate to groundwater underscores the importance of including nitrate accumulation in nitrogen budgets to better understand the N cycle and improve agricultural management. Measures such as optimizing nitrogen application, adopting fertigation, and strengthening farmer education are crucial to mitigate nitrate losses and reduce environmental risks.

4. Conclusions

This study demonstrates the utility of numerical modeling (HYDRUS) for assessing soil water and nitrogen dynamics in agricultural fields under diverse crop rotations and fertilizer application. The results provide significant insights into the interplay of water flow, nitrogen species transport, and transformation processes under variably saturated conditions. Key findings include:

• Model Performance and Reliability: HYDRUS-1D and HYDRUS-2D models accurately simulated water dynamics and nitrate transport, validated by relatively high R² values across years (2019–2023). These models successfully captured nitrification and urea hydrolysis, highlighting their applicability for studying nitrogen cycling in agricultural systems; however, it was indicated that capturing the denitrification process might lead to improved model performance.
• Impact of Precipitation and Groundwater Levels: Seasonal dynamics revealed that nitrate leaching was strongly influenced by precipitation and shallow groundwater interactions. High soil moisture during early crop growth and post-harvest stages contributed to increased nitrate losses, emphasizing the need for precise timing of nitrogen applications.
• Optimization of Fertilization Practices: Simulations demonstrated that optimal nitrogen rates and application schedules could significantly reduce nitrate leaching while maintaining crop productivity. The study supports the adoption of integrated nitrogen management strategies, including fertigation and partial root-zone drying, to enhance nitrogen use efficiency and mitigate environmental risks.
• Environmental Implications: The study confirms that nitrate leaching poses a risk to groundwater quality, particularly under intensive agricultural practices. The slow migration of nitrates in deeper soil layers underscores the importance of including nitrate accumulation in nitrogen budgets to close the nitrogen cycle and inform sustainable agricultural practices.

By combining field data with robust numerical modeling, this research highlights the potential for using HYDRUS models as decision-support tools for sustainable agricultural management, contributing to improved nitrogen use efficiency and environmental conservation. Future studies should focus on integrating real-time data and expanding scenarios to refine management practices further and expand model capabilities by integrating relevant processes.