灌溉优化中的数字孪生：优势、机遇与挑战

• Review

• 

22 October 2026

25 Pages

,

,

,

and

1

Ningbo Global Innovation Center, Zhejiang University, Ningbo 315100, China

2

College of Control Science and Engineering, Zhejiang University, Hangzhou 310027, China

3

Polytechnic Institute, Zhejiang University, Hangzhou 310022, China

4

NingboTech University, Ningbo 315100, China

Abstract

Efficient irrigation management is important to increase agriculture water use efficiency and achieve sustainable crop production under increasing water scarcity and climatic variability. Based on advanced devices and artificial intelligence (AI) algorithms, the irrigation system can be digitalized and transformed from experience-driven decision-making toward data-driven, predictive, and adaptive control. Digital Twins (DTs) enable smart irrigation management through continuous bi-directional data flows, dynamic model updating, intelligent prediction, and closed-loop execution. This paper provides a comprehensive review of the applications of DTs in irrigation management and optimization, covering Internet of Things (IoT), Remote Sensing (RS), and advanced simulation-inspired irrigation systems. Furthermore, we discuss the opportunities and challenges in DTs for irrigation optimization with the development of new techniques, and identify the major technological trends, existing bottlenecks, and future research directions for DT-enabled irrigation.

1. Introduction

Environmental scientists predict that water scarcity will be a serious problem in the coming decades because of the growing population and increasing water pollution [1,2]. Agriculture accounts for 70% of freshwater use. Soil water loss, inefficient water management, climate change, and inefficient irrigation systems are the main causes of agricultural water shortages [3,4,5]. Irrigation consumes approximately 75% of agricultural freshwater. However, most irrigation water is lost through evaporation, percolation, and deep infiltration [6]. Improving the fraction of available water that contributes to productive crop transpiration therefore remains an important objective of agricultural water management [7]. Inaccurate irrigation scheduling is one of the major reasons for the waste, which can easily cause over-irrigation or under-irrigation [8]. Both over-irrigation and under-irrigation can reduce crop yield and decrease product quality and, in worse cases, cause soil salinization and soil structure degradation [9].

Generally, irrigation scheduling depends on operators’ experience and simple meteorological data [10]. Researchers are always trying to develop new methods to improve irrigation efficiency [11]. One way is to simultaneously optimize irrigation scheduling and yield, and most common methods include evapotranspiration and water balance (ET-WB), soil moisture status, plant water status, and model-based methods [12]. The ET-WB method is widely used but highly depends on the accuracy of the estimated parameters including the reference evapotranspiration (ET 0) and the crop coefficient (K c) [13]. Both ET-WB-based and soil moisture-based methods apply soil water balance equations to schedule irrigation, but the soil moisture-based method employs varying thresholds through experiments, rather than a default value [14]. Plant-based methods are generally based on the relationship between plant status (e.g., water stress, growth stage) and soil moisture status [15,16]. The model-based methods include process-based models [17], regression models [18], and artificial intelligence models [19]. Another way is to enhance crop water use efficiency (WUE) using smart monitoring and irrigation control strategies [20]. Most studies monitor soil, weather, and plant information to improve WUE [21,22]. The plant information mentioned here mostly mean device measurements including electronic sensor data and images [23], not physiological function. Open-loop and closed-loop control are two mostly used irrigation strategies, and intelligent/smart irrigation control usually belongs to closed-loop systems [24]. It is suggested to enhance real-time monitoring to capture dynamic nature of weather, soil characteristics, and plant water use for open field application.

Agricultural data shortage is always a big challenge in irrigation management. Internet of Things (IoT) technology enables timely acquisition of environmental data and is a good tool to collect agricultural data [25]. When data are available, it depends how real the model is to provide decision-making support. The development of Digital Twins (DTs) brings opportunities for irrigation management [26]. The emergence of DTs leverages the real-time data acquisition capabilities of IoT, and by analyzing the potential relationships between various data, it establishes a virtual−physical mapping model [27]. DTs in agriculture are under development and the corresponding research is at a premature stage [28]. DTs enable precise irrigation management by integrating real-time data from sensors and using virtual models to simulate and optimize irrigation practices [29]. It is expected that DTs can comprehensively consider both information of the plant and the environment to formulate corresponding irrigation strategies and provide farmers with a platform on which to evaluate these irrigation strategies in parallel with their current farm practices and validate whether the solution meets their needs without risking their crops [30]. In addition, integrating DTs with other smart technologies can enhance their potential in irrigation management [31].

Despite the growing literature on agricultural DTs, irrigation-specific applications remain fragmented across sensing, crop modeling, hydraulic simulation, and intelligent control. Few reviews have compared these systems according to their physical–virtual synchronization, dynamic updating, decision feedback, and field validation. This study provides a comprehensive review of the application of DTs to precision irrigation management under increasing water scarcity. We have discussed the advantages, limitations, and prospective research directions of DT implementations in irrigation management. Particular attention was given to studies involving physical–virtual representation, real-time/periodically updated sensing data, model-based simulation, dynamic model updating, irrigation decision-making, and feedback control. Studies involving only conventional irrigation monitoring and standalone prediction without an explicit physical–virtual relationship were not considered core DT applications, although selected DT-inspired studies were retained where they provided useful insights into the transition toward operational irrigation DTs.

The remainder of this paper is organized as follows. Section 2 reviews the hierarchical architecture of DTs and their representative applications in agriculture. Section 3 examines three categories of DT applications in irrigation management: IoT-enabled, remote-sensing-enabled, and advanced simulation/model-driven approaches. Section 4 discusses the major limitations, technical challenges, and potential solutions associated with DT-enabled irrigation management. Section 5 summarizes the main findings and outlines future research directions.

2. Background

2.1. Definition and Functional Characteristics of Digital Twins

The concept of DTs was initially introduced as a part of Product Life Cycle Management (PLM) [32]. It was described as a virtual digital representation of a physical product, with bi-directional data flow between the physical object and its virtual counterpart [33], and has been applied in various industries including agriculture [34]. The development of DTs depends highly on integrating cyber-physical systems (CPS), IoT, and artificial intelligence (AI) [35]. According to the CIRP Encyclopedia of Production Engineering, DTs are defined as a representation of an active and unique product, which can range from physical devices, objects, and machines to intangible assets and systems consisting of products and services [36]. The key enabling technologies relevant to agricultural and irrigation DTs are summarized in Table 1. As the concept has matured, the terminology around DTs has become more domain-specific, adapting to the context in which it operates, with researchers continually refining its definition to accommodate emerging roles and unforeseen applications. Although the term of DTs has been broadly used in agriculture, not every digital model or IoT-enabled prediction system constitutes a full DT. A conventional digital model may operate independently of the physical system, while an IoT-based monitoring system mainly collects and displays physical observations. A digital shadow enables automatic data flow from the physical system to its virtual representation but does not establish a corresponding feedback mechanism from the virtual system to the physical system. In this review, a DT should include an identifiable physical entity or process together with a corresponding virtual representation. Observations from the physical system should update the state or parameters of the virtual model through real-time or periodically refreshed data. The virtual model should also provide simulation, prediction, diagnosis, or optimization functions to support decision-making. More advanced DT implementations further establish feedback from the virtual environment to the physical system through decision recommendations or automated control. Continuous physical–virtual synchronization and bidirectional interaction are therefore regarded as important characteristics that distinguish operational DTs from conventional digital models or IoT-based monitoring systems.

Accordingly, the studies reviewed in this paper should not be interpreted as representing the same level of DT maturity. Some systems achieve real-time physical–virtual synchronization and closed-loop control, whereas others remain DT-inspired frameworks that primarily provide simulation, prediction, or decision support. These differences are explicitly considered in the assessment of representative applications in Section 3.

2.2. Digital Twins in Agriculture

DTs have gained increasing attention, highlighted by the rapid growth in available examples. Currently, there are not many large-scale applications, with most of the existing ones primarily focusing on the integration of computational intelligence and remote sensing devices to achieve targeted system optimization and task decision-making (Table 2). The benefits of DTs in agriculture consist of addressing local system variability [37], creating automated operational pipelines [38], integrating data from multiple sources [39], and improving transparency and decision-making [40]. DTs have become a transformative technology with broad applications across agriculture to address complex challenges in farm management and resource optimization. Generally, agricultural DTs consist of several interconnected layers, including physical sensing, data connectivity, data processing, virtual modeling, and application-level decision or control (Figure 1). Data collected from the physical system are transmitted to the virtual environment for state estimation, simulation, prediction, and optimization, while the resulting decisions and control commands can be fed back to the physical system to support closed-loop management. Artificial intelligence can enhance agricultural DTs by supporting state estimation, parameter updating, prediction, and decision optimization. Skobelev et al. [41] developed a plant-oriented DT framework to improve the prediction of crop growth and dry matter accumulation. Multi-agent technology was used to represent successive developmental stages, with stage-related model parameters dynamically updated according to environmental observations and expert knowledge. This approach illustrates how continuously updated environmental information can be incorporated into a virtual crop representation to support adaptive decision-making.

Figure 1. The hierarchical architecture of a DT system in agriculture, which contains physical sensors (a), data connectivity (b), data processing (c), DT models (d), and applications (e). Agricultural data are acquired in real time via sensors in layer (a) and transmitted to the cloud/edge based on layer (b) for data processing in layer (c). In layer (d), virtual models are developed to simulate plant development, optimize resource use, predict yield, etc. In layer (e), the generated predictions and strategies are applied for precise irrigation, automated machinery scheduling, decision support, etc. Then, the control strategy is reflected to the real world.

Parewai and Köppen [42] proposed a detailed DT-based soil monitoring framework to collaboratively optimize both water use and productivity. This work gave a good example of applying machine learning in DT modeling, in which advanced statistical and textural features from multi-spectral soil images were extracted to enable data-driven refinement of root-zone moisture assessment and support precise irrigation management. In smart control, Zhao et al. [43] developed a DT simulation platform for autonomous agricultural vehicles, in which high-fidelity field environments were reconstructed using UAV-based oblique photogrammetry to generate dense 3D point clouds and HD maps at a real-world scale. At the system level, DTs deploy agent-based modeling and game-theoretic algorithms to resolve conflicts in resource allocation across irrigation, machinery, and labor schedules. This holistic integration of robotic control, predictive analytics, and multi-agent optimization elevates farm-wide energy efficiency, demonstrating how DTs unify disjointed agricultural processes into an adaptive, closed-loop ecosystem.

These applications demonstrate the multi-scale and interdisciplinary nature of agricultural DTs. In irrigation management, DTs require closer integration of real-time environmental sensing, crop–water process modeling, dynamic model updating, and decision or control feedback to capture the continuously evolving interactions among soil, crops, weather, and irrigation systems.

Table 1. Key enabling technologies for agricultural and irrigation DTs.

Table 2. Representative applications of DTs relevant to agricultural water management and irrigation.

3. Assessment of Exemplary Digital Twin Use Cases in Irrigation

In general, the application of DTs in agricultural irrigation is implemented with some specific techniques: high computing ability to process large amounts of environmental data in real time; effective integration of hardware infrastructure and data processing platforms to support real-time decision-making and system scalability; crop growth-based model dynamic adaptation for decision-making; and a closed-loop control system formation for fast feedback. This work discussed several applications of DTs in irrigation, including three classic categories: IoT-DT-inspired irrigation enabling remote monitoring and automation; RS-DT-inspired irrigation leveraging remote sensing for large-scale environmental adaptation; and advanced simulation-DT-inspired irrigation.

3.1. IoT–Digital Twin-Inspired Irrigation Systems

The integration of IoT and DTs has progressed from basic data logging to the sophisticated orchestration of physical and virtual entities. An IoT–DT irrigation system mostly focuses on real-time monitoring and closed-loop control (Figure 2). Escribà-Gelonch et al. [40] addressed the interoperability challenges in agricultural cyber-physical systems by proposing a comprehensive DT orchestration framework. The authors emphasize that DTs, by integrating real-time data, IoT devices, and AI, enable precise modeling and optimization of agricultural processes. The core of their study involves a standardized protocol designed to synchronize heterogeneous IoT sensors with cloud-based computational models (Figure 3).

Figure 2. Architecture of a IoT-DT-inspired irrigation system, in which weather, soil, and crop information are collected using distributed sensors (a) and transmitted through an IoT gateway (b) to a cloud-based big data platform to be pre-processed and stored in a distributed management system (c). The processed data are fed into a DT model to simulate, predict, and optimize agricultural processes (d). The model results are back to the IoT gateway to actuate operations (e) or are delivered through a human–machine interface (f).

Figure 3. Orchestration architecture of an IoT-driven DT system, in which AI algorithms and simulation modules process real-time data alongside historical records (data library) to generate predictions in the virtual space. These insights are converted into commands for IoT actuators to establish a closed-loop control mechanism.

Moving from theoretical orchestration to field implementation, Wangere [72] developed a tangible DT prototype for surface drip irrigation monitoring in the Arusha region, Tanzania. The study used ZigBee wireless networks and a Node-RED dashboard to map physical parameters (e.g., soil moisture) to a virtual interface in real time. This prototype validated the feasibility of low-cost hardware (e.g., solar-powered) in off-grid environments. The study also revealed deployment challenges, including the lack of remote firmware updates (OTA) and safety risks associated with high-voltage pumps control. To go beyond simple data visualization and automate irrigation commands, Manocha et al. [29] proposed a smart irrigation framework based on IoT and DTs to optimize water usage through algorithms. Unlike previous studies that mostly focused on data transmission, this research integrates an Adaptive Neuro-Fuzzy Inference System optimized by Genetic Algorithms (ANFIS-GA) to dynamically calculate optimal irrigation weights using non-linear soil and weather data (Figure 4a). Quantitative results showed that the ANFIS-GA model outperformed the Artificial Neural Network (ANN), Fuzzy Inference System (FIS), and ANFIS-GA benchmarks, achieving a testing accuracy of 95.85% for soil-condition classification. The original study also evaluated predictive performance using RMSE, R 2, VAF, and MEDAE. Following these advancements in algorithmic precision, Merizig [73] focuses on the irrigation management in arid areas, where water scarcity is critical. In this work, a Deep Learning-based DT was developed specifically for high-evaporation conditions to validate the viability of DTs in extreme environments. One key innovation is the utilization of an immersive 3D visualization interface (Figure 4b), which allows managers to intuitively monitor pump status and crop water needs in a virtual space. The experimental results show that the proposed Deep Neural Network (DNN) model outperformed traditional FAO mathematical models with a Mean Absolute Error (MAE) of only 0.3 (<1.17) in predicting irrigation needs. This demonstrates that data-driven DTs can achieve precise water conservation compared with conventional agronomic formulas in severe climatic conditions.

Figure 4. (a) The optimization-focused framework reported by Manocha et al. [29], highlighting the ANFIS-GA algorithm in the cloud layer for precise irrigation scheduling. (b) The visualization-focused framework reported by Merizig et al. [73] for arid areas, emphasizing the 3D modeling module in the DT layer for immersive monitoring.

Recently, Adamo [74] moved away from traditional monolithic architectures to propose an IoT framework based on micro-services for greenhouse management. The core of this research is modeling water and nutrient management as a data-driven sequential decision problem. Real-time IoT observations were used as inputs to an Integer Linear Programming (ILP) model to optimize the proportion of the nutrients and water. In greenhouse field experiments, the results showed this solution can reduce water usage by 30%, and save 47.5% of nitric acid and 67.7% of macronutrients compared with the conventional systems using a PID controller. It shows that DT systems can handle multi-variable and multi-objective optimization problems by decoupling complex agricultural logic through micro-services.

As service nodes increase, standardizing communication between heterogeneous components becomes a new challenge. Addressing this, Alves et al. [30] adopted an Industrial Internet perspective and introduced a standardized IoT platform based on FIWARE (Figure 5) to solve interoperability issues in agricultural DTs. A unique feature of this work is the combination of the Open Platform Communications Unified Architecture (OPC UA) industrial protocol with Siemens Plant Simulation software to build a DT capable of discrete event simulation. This architecture enables a complete data connection from soil probes to the cloud. Crucially, it allows managers to evaluate the behavior and risks of irrigation strategies in a virtual environment before deploying them in the field.

Figure 5. The standardized DT data flow: (a) soil probes and weather data are connected via (b) the IoT Agent JSON, and the industrial irrigation system and simulation use the IoT Agent OPC UA protocol; (c) the architecture centers on the Orion context broker, which manages real-time context information; (d) the Cygnus connector persists historical data into MongoDB and MySQL databases, which are used for (f) the irrigation recommendation engine and the Grafana visualization dashboard.

Most aforementioned systems focus on the physical state of irrigation infrastructure, and few include the crops’ dynamic physiological responses. Xu et al. [75] constructed a DT model to simulate the growth process of winter wheat by integrating multi-source sensing data (IoT and drone multispectral data) with a mechanistic crop growth model (DSSAT) (Figure 6). The SUBPLEX algorithm was used for model calibration and dynamic parameter adjustment. After calibration, the model achieved an R 2 of 0.98 for simulations of Leaf Area Index (LAI) and Above-Ground Biomass (AGB). The simulation errors for flowering and maturity stages were 0.6 and 1.1 days, respectively, while the yield simulation error remained below 0.6/hm 2. Based on the simulated crop growth status, the framework was further used to support management strategies for different winter wheat varieties.

Figure 6. The bidirectional data mapping between (a) the real field (left) and (c) the twin field (right) in a winter wheat growth simulation framework.

Soil physicochemical properties are also critical factors influencing yield. Qin et al. [76] extended DT applications to water−salt regulation to address the secondary salinization caused by irrigation without drainage in arid regions. In this work, an intelligent Irrigation−Drainage collaborative system was constructed based on Long Short-Term Memory (LSTM) to manage drainage and high-salinity water treatment in a closed-loop system. The system integrated multi-layer IoT sensing with LSTM-based prediction and a digital twin platform for coordinated irrigation–drainage management. The LSTM model achieved R 2 values of 0.97 for soil Electrical Conductivity (EC) and 0.92 for water content prediction. This work provided a digital solution for the sustainable restoration of agroecosystems. Pavón et al. [77] digitalized the existing traditional pressurized irrigation system at the University of Alicante (UA) and achieved high-throughput real-time data synchronization by integrating Apache Kafka. In this work, an intelligent management platform was developed to fuse IoT with Geographic Information Systems (GISs). Notably, the study compared Building Information Modeling (BIM) with GISs for web-based rendering and found that GISs are superior to visualize large-scale networks. The network-level DTs provide a low-cost technical path for incidents management and the modernization of legacy irrigation facilities.

As the depth and breadth of interactions between IoT devices and the cloud expand, the cybersecurity boundary of these systems becomes increasingly fragile. As warned by Wangere et al. [72] during field deployment, remote control mechanisms are easy targets for attacks. Ramesh et al. [78] shifted focus from agronomic optimization to infrastructure defense and introduced a security simulation framework based on AWS TwinMaker to systematically simulate cyber threats, such as Man-in-the-Middle (MitM), Denial of Service (DoS), and data tampering. By integrating machine learning classifiers like Support Vector Machines (SVM) and Random Forest, the system can identify network traffic anomalies and trigger alerts in real time. This work highlights that an irrigation DT system needs both high-precision agricultural decision-making and robust digital immune capabilities.

Overall, IoT-enabled DTs provide a practical basis for physical–virtual synchronization through continuous or near-real-time field sensing. Current systems support monitoring, prediction, irrigation optimization and, in some cases, closed-loop control. However, their implementation maturity varies considerably. Many studies remain focused on sensing and prediction, while fewer achieve continuous model updating and feedback to physical irrigation processes. Further improvements in interoperability, communication reliability, sensor robustness, and system security are therefore needed for operational deployment.

3.2. RS–Digital Twin-Inspired Irrigation Systems

Discrete sensor nodes struggle to fully characterize large farmlands with high spatial heterogeneity, and the cost of dense hardware deployment is prohibitively high. Researchers have studied Remote Sensing (RS) technology in agricultural management. RS-based DT systems use spectral imagery from satellites, Unmanned Aerial Vehicles (UAVs) and subaerial cameras to retrieve Crop Water Stress Index (CWSI) and Vegetation Index (NDVI) on a large scale. Awais et al. [79] pioneered the deep integration of UAV remote sensing with DTs to solve the problem of limited coverage of ground-based sensors. This study proposed an intelligent framework using multi-spectral and thermal sensors to assess crop water status. Unlike systems based on ground data, this framework constructs a twin model architecture with a data fusion layer, in which critical features from multi-spectral imagery were extracted using Principal Component Analysis (PCA) and fused with heterogeneous data from ground weather stations using machine learning algorithms to assess crop water status and generate precise irrigation decisions. Field tests were implemented in a tea garden irrigation system to validate the feasibility of applying machine learning to process remote sensing images for irrigation decision support.

Monitoring the current growth status of crops is insufficient for proactive agricultural management. Pal et al. [80] further explored the potential of Unmanned Aerial Systems (UASs) in time-series forecasting to increase predictive accuracy. A DT framework was developed to predict cotton growth by combining real-time data with historical RGB imagery from 2020 to 2023 (Figure 7). This method not only extracts current Canopy Cover and Excess Greenness (EXG) but also precisely forecasts the crop’s growth trajectory up to 30 days in advance. The system achieved a relative RMSE of only 9% for canopy cover prediction in quantitative validation. The method optimizes irrigation scheduling during the growing season and provides critical decision support to determine the crop termination timing.

Figure 7. The predictive DT framework integrating (a,b) UAS data mining with (c) hybrid modeling to forecast crop growth for (d) optimized crop management, in which dotted line means optional operations.

UAVs are unsuitable for continuous monitoring of vast commercial plantations because of battery life and labor costs. To achieve large-scale automated management, Bellvert et al. [81] applied satellite images to automatically manage irrigation scheduling for commercial vineyards. It assimilated Sentinel-2 fraction of Absorbed Photosynthetically Active Radiation (fAPAR) data into a DT model to execute Regulated Deficit Irrigation (RDI) strategies in a decision system. Quantitative results show that the simulated evapotranspiration aligns well with remote sensing estimates, with a Root Mean Square Deviation (RMSD) ranging between 0.98 and 1.14 mm day−1 for potential evapotranspiration. In the absence of high-resolution thermal sensors, it is challenging to estimate water stress in complex row crops. It is also challenging to define precise irrigation Management Zones (MZs) with limited sampling points. Rodrigues et al. [82] proposed a multi-source data-fusion approach to balance mapping accuracy and field-sampling effort for irrigation management zoning. Landsat-8 and Sentinel-2 observations were integrated with sparse proximal apparent electrical conductivity (aEC) measurements, while candidate auxiliary covariates were selected using the Bayesian Information Criterion (BIC) and adjusted R 2. Among the evaluated geostatistical methods, Kriging with External Drift (KED) performed best under sparse sampling, achieving R 2 = 0.78 while using only approximately 4% of the original ground observations. These results demonstrate the potential of remote–proximal sensor fusion to delineate irrigation management zones with reduced field-sampling requirements. However, the resulting horizontal zoning may not fully capture vertical soil−water variability, particularly under severe drought conditions. Camps et al. [83] proposed the AI4WATER DTs to assess historic droughts in Catalonia. The model added a Ground Penetrating Radar (GPR) mobile rover to complement satellites and fixed probes. The data fusion method enables the DTs to dynamically measure non-linear soil moisture at various depths. In this study, LSTM networks were used to deal with long-term soil moisture lag effects and increase prediction accuracy. DTs have shown good performance in integrating macro-scale water budgets with micro-scale soil profiles.

Millán et al. [84] developed a DT model to evaluate the Irri_Desk system at a 15-hectare commercial tomato farm in Spain over two years. Soil sensors and weather data were integrated in a rigorous comparative experiment. The system benchmarked farmer experience, standard remote sensing platforms, and DT-based automated scheduling. Quantitative results show the DT-based method achieved the highest economic water efficiency (284.81 EUR/mm) with a high yield (140 t/ha). Furthermore, the study envisioned the potential of Generative AI and Blockchain technology for future irrigation traceability.

Overall, RS-enabled DTs provide an important means of extending spatial observation beyond point-based field sensors. They are particularly useful for field- and farm-scale irrigation management, where remote sensing can complement sparse ground measurements of crop and soil conditions. However, revisit intervals, UAV operational costs, spatial–temporal resolution mismatch, and indirect estimation of root-zone states can limit continuous synchronization. Further integration with ground sensing, dynamic model updating, and feedback control is therefore needed to support more complete operational DTs.

3.3. Advanced Simulation and Strategic Digital Twin-Inspired Irrigation Systems

In general, DTs require physics-based reasoning and simulation capabilities. Hybrid and simulation-driven DT systems mostly focus on simulating complex water infrastructure and analyzing energy efficiency, which can provide decision support when data are unmeasurable. Moreira et al. [85] proposed a systematic methodology to develop DTs for Pressurised Collective Irrigation Systems (CISs). This study went beyond simple data collection by combining telemetry data from physical assets (e.g., pumps and pipes) with the EPANET hydraulic simulation model (Figure 8). Through this bidirectional physical−virtual connection, the system can calculate water balance and energy efficiency indicators in real time. It can even identify metering errors, such as hydrometer faults. An application in the Odivelas irrigation district in Southern Portugal showed that the hydraulic simulation-based DTs can effectively diagnose energy losses in the system.

Figure 8. The DTs methodology for Pressurized Collective Irrigation Systems, which used the EPANET hydraulic model to optimize real-time water and energy balance.

The accuracy of hydraulic simulation models relies heavily on parameter calibration. In real open-channel irrigation systems, the critical parameter, Manning’s roughness coefficient (n), often changes over time and space, which makes it hard to capture with static settings. To address this, Liu et al. [86] proposed a segmented estimation (SE) framework for real-time DT calibration and innovatively embedded the Ensemble Kalman Filter (EnKF) algorithm into channel hydraulic simulations (Equation (1)):

S a = S f + K(y − HS f)

(1)

where S a is the updated/calibrated parameters (e.g., the corrected roughness coefficient), S f is the initial prediction from the hydraulic model, K is the Kalman gain matrix which determines the weight given to the observation data and the model prediction, y is the real-time measured data (e.g., water levels), and H is the observation operator that maps the model state space to the observation space. The framework can correct the roughness coefficient in near real time (Figure 9). Compared with empirical equations, it improved model prediction accuracy by 45–60% for long-distance (>10 km) canals. This work solves the high-fidelity challenge in hydraulic simulation and ensures the virtual model precisely mirrors the physical hydrodynamic processes in real time.

Figure 9. In the segmented estimation (SE) framework, the Ensemble Kalman Filter (EnKF) is embedded into the DT model for real-time calibration of Manning’s roughness coefficient (n).

Beyond calibrating macro-channel parameters, the micro-level fluid behavior of emitters (e.g., drippers) also affects irrigation system performance. To investigate complex pressure and velocity distributions, Preite et al. [87] advanced simulation from the network level to the component level. In this study, a high-precision digital model based on Fluid Dynamics was developed by combining 1D system simulation (Flownex version 2022) with 3D Computational Fluid Dynamics (Ansys Fluent version 2022 R2) to simulate flow fields inside drippers and pipes. The model can predict pressure and velocity distributions with a determination coefficient exceeding 93%. In addition, it can precisely identify anomalies like dripper clogging or pipe leaks by monitoring deviations between simulated and measured values in real time and optimize network design based on physics-based deep simulation. El Hachimi et al. [88] extended DT applications from maintenance to agricultural scheduling, which addressed water supply−demand conflicts caused by concentrated sowing in traditional Open-Channel Irrigation Systems (OCISs). An intelligent agent framework based on Deep Reinforcement Learning (DRL) was developed with the AquaCrop crop model and an OCIS hydraulic model. The agent can learn through trial and error in the virtual environment to find the optimal sowing date distribution using the Deep Deterministic Policy Gradient (DDPG) algorithm. Results show that the DDPG agent can adhere to hydraulic constraints and increase total crop yield by 4.69% with a high WUE (=1.95). Yang et al. [89] reviewed the evolution of Irrigation Management Informatization (IMI) in China and discussed its transition toward AI-driven DTs to improve water use efficiency. The study highlights that irrigation systems are evolving from automation to cognitive intelligence with breakthroughs in Large Language Models (LLMs). China has launched DT pilots in 48 large and medium-sized irrigation districts since 2022 (Ban Nong Shui Han [2022] No. 1163).

Recently, Yang et al. [90] conducted Digital Plant in irrigation optimization based on a Functional−Structural Plant Model (FSPM) to simulate the biophysical processes of water transport, transpiration, and organ-level growth in a virtual space. Modern FSPMs, such as the 2024 updates to the GreenLab [91] and OpenSimRoot architectures [92], can be used for real-time quantification of water stress impacts on biomass allocation and topological evolution. This is helpful for irrigation precision by simulating the leaf water potential (ψ l) and stomatal conductance (g s) to optimize water transport in the crop. However, the complexity of solving non-linear differential equations for thousands of virtual organs often results in simulation latencies that are incompatible with real-time edge deployment. Xu et al. [93] proposed a digital-twin-oriented irrigation scheduling framework named IrriTwin to optimize irrigation scheduling in an organ level. A model-in-the-loop virtual soil-crop system was developed based on a soil water model and the GreenLab crop growth model, in which irrigation actions were generated using the Soft Actor-Critic (SAC) algorithm [94]. The soil water model is a simplified root-zone averaged soil water balance model that describes the changes in soil water content per unit ground surface area during each crop growth cycle. The GreenLab crop growth model is an organ-level FSPM that describes crop biomass production, organ-level allocation, and structural formation in a unified framework. Those two models were coupled to represent the effects of irrigation-induced soil moisture changes on crop growth. The irrigation scheduling task was modeled as a Markov Decision Process, in which the irrigation probability and irrigation amount for each growth cycle are selected as the action space, and the reward function is designed to encourage biomass accumulation. Under the same total water supply, the SAC-based strategy achieved higher biomass than both a uniform irrigation schedule and a beta-based schedule that distributed irrigation amounts across crop growth cycles according to a beta distribution. However, IrriTwin does not yet realize a full digital twin without real-time synchronization and continuous interactions between the virtual model and the physical model.

Simulation-driven DTs are valuable tools to represent hydraulic, crop, and soil processes that cannot be directly observed in the field. Mechanistic models, data assimilation, and reinforcement learning can support prediction, scenario testing, and irrigation optimization before decisions are applied to the physical system. However, high-fidelity simulation alone does not constitute a complete DT. Without continuous field-data updating and physical–virtual interaction, such systems remain digital models or DT-oriented simulation frameworks.

The representative studies discussed above are grouped according to their dominant technical characteristics into IoT-enabled, RS-enabled, and simulation-driven DTs. These categories are not mutually exclusive, because many systems combine multiple sensing and modeling technologies. The reviewed systems also differ in functional maturity, ranging from monitoring and prediction to decision support, optimization, and control (Table 3). Taken together, the reviewed studies show clear differences in data characteristics, synchronization, functional maturity, and validation level. IoT-enabled DTs generally provide stronger temporal connectivity with physical irrigation processes, RS-enabled DTs extend spatial observation, and simulation-driven DTs improve the representation of complex hydraulic, crop, and soil processes. However, only a limited number of studies demonstrate continuous physical–virtual synchronization together with operational feedback. Future development should therefore emphasize online calibration, continuous synchronization, field validation, and stronger links between model performance and practical irrigation outcomes.

Table 3. Comparison of representative DTs and DT-inspired applications in irrigation management.

4. Opportunities and Challenges in DTs for Irrigation Optimization

DTs have shown great potential in optimizing irrigation management through estimating current soil/crop water states, predicting future water demand, optimizing irrigation decisions, and updating decisions. However, their practical deployment is still limited by the challenges in multi-source data analysis, accurate twin modeling, and field validation.

4.1. Multi-Source Data Analysis

One advantage of DTs in irrigation is making irrigation decisions by combining multi-source data, including soil/weather/crop conditions, irrigation discharge, UAV imagery, satellite observations, etc. In general, multi-source data have different temporal and spatial resolutions. Keeping heterogeneous data streams is highly important for irrigation DTs. However, the value of multi-source data depends not only on data availability, but also on data quality, continuity, consistency, and interoperability. It is important to break through the existing bottleneck of real-time heterogeneous-data integration and interoperability for DT deployment. To ensure data quality and continuity of field observations, real-time monitoring usually faces the challenges of device errors, drift, missing data, and communication failures. These errors may propagate into state estimation, model calibration, and irrigation decisions if they are not detected in time. Measurement errors and sensor drift can be detected and corrected using automated quality-control methods, such as ANN [95] and PCA [79]. Intelligent data reconstructions methods such as LSTM [76,83] can be applied to estimate missing measurements, and edge computing can temporarily store sensor measurements to prevent temporary network outages [48].

Another important challenge is the spatiotemporal mismatch among heterogeneous observations. Point-based soil sensors may provide high-frequency measurements but limited spatial coverage, whereas UAV and satellite observations provide broader spatial information at lower or irregular temporal frequencies. Directly combining these data may therefore introduce scale inconsistencies. Data aggregation, interpolation, data assimilation or state-estimation methods are needed to map observations from different sources onto comparable spatial and temporal scales. In DTs, one does not need to fuse raw device data directly but use them to estimate common physical states. Here, DTs can become a common physical framework into which heterogeneous observations are fused through shared state variables. In addition, uncertainty from different data sources should be considered during state and parameter updating, because biased or low-quality observations may destabilize online calibration. Reliable multi-source data integration is therefore essential for dynamic model calibration, physical–virtual synchronization, and robust irrigation decision-making.

4.2. Accurate Twin Model

In irrigation, environment−plant−water relationships change spatially and temporally. If the twin model cannot accurately represent field conditions, the system may give poor irrigation decisions. The development of adaptive DTs using real-time observations and data assimilation remains a major challenge for irrigation optimization. One is the limited observability and identifiability of model parameters. Many soil and crop parameters cannot be directly and continuously measured. Machine learning and deep learning methods have been applied to estimate soil [42] and crop status [96]. Beyond repeated parameter fitting, dynamic calibration can be formulated as a joint state–parameter estimation problem. This enables the DTs to adapt to changing soil, crop, and environmental conditions without overfitting short-term measurement noise. In addition, slowly varying soil parameters and rapidly changing crop parameters should not be updated at the same frequency. Sensitivity and identifiability analyses can therefore be conducted before online adaptive calibration to select influential and observable parameters. Cross-scale integration presents another problem. For instance, the irrigation decision made at the organ, plant, irrigation-zone, and field scales can be totally different, and simply enlarging a field-level DT does not necessarily produce an irrigation-district DT. Hierarchical and multi-scale DT architectures are therefore required to connect processes and decisions across different spatial and organizational scales.

Model-performance indicators such as R 2, RMSE, and MAE are useful to evaluate agreement between simulated and observed variables, but they do not directly represent irrigation performance. A model with high predictive accuracy may still result in limited water savings or yield improvement if decision rules, control actions, or field conditions are not properly considered. Therefore, irrigation DTs should be evaluated using both model-level metrics and practical outcomes, such as irrigation water savings, crop yield, WUE, and energy consumption. Linking improvements in model accuracy to measurable operational outcomes is important for assessing the practical value of DTs.

A high-fidelity DT must continuously analyze heterogeneous data, simulate complex processes, update states and parameters, and optimize irrigation decisions in real time, which can impose a substantial computational burden. Computationally efficient approximations can therefore be used to reduce the cost of high-fidelity simulation. Physics-Informed Neural Networks (PINNs) [97] incorporate physical constraints, such as conservation relationships, into neural-network training and can improve physically consistent prediction when observational data are limited. Reduced Order Models (ROMs) [98] simplify high-dimensional models into lower-dimensional representations while preserving dominant system dynamics, thereby reducing simulation time. Figure 10 presents a good example of the Digital Plant framework that integrates field sensing, mechanistic crop models, PINNs, ROMs, and Reinforcement Learning (RL). A virtual sandbox provides a simulated environment to test irrigation strategies before field implementation, after which selected decisions can be returned to the physical system through a feedback loop.

Figure 10. Comprehensive framework of Digital Plant-driven irrigation optimization, which integrates multi-scale data acquisition, hybrid physics−AI modeling (PINNs/ROMs), and a decision-making Virtual Sandbox to improve plant-specific irrigation accuracy within a fault-tolerant closed-loop system.

4.3. Field Validation and Practical Deployment

A genuinely operational agricultural DT requires long-term and multi-site validation. Validation should go beyond model accuracy. It should also assess physical–virtual synchronization, decision quality, robustness, transferability, and the reliability of operational feedback. However, agricultural systems are highly heterogeneous. Soil conditions, microclimate, and crop traits can vary considerably even within the same field [99]. Even in a single plot, individual plants exhibit varying water use efficiencies due to genetic diversity and micro-climatic nuances. It is a challenge to achieve personalized model calibration in real time. Practical deployment introduces additional challenges. Sensor drift, missing data, communication delays, and actuator failures can degrade DT performance in the field. Computational efficiency and interoperability also become important when multiple sensing, modeling, and control components operate together. Therefore, field validation should examine not only prediction accuracy, but also the stability of online updating and the agronomic consequences of DT-based decisions.

The development of LLMs shows great potential in automatically tuning the physiological parameters of DTs. The LLMs can function as both a conversational interface and a proactive orchestrator, which can narrow the gap between the complex physiological outputs of the DT and the underspecified constraints of real-world farm operations. The bidirectional feedback loop between the physiological engine and the orchestrator can enable real-time dynamic parameter calibration and transform numerical simulations into semantic reasoning chains. The framework combines multi-source field observations, a physiological modeling engine, and an LLM-based orchestrator supported by a domain-specific knowledge base and Retrieval-Augmented Generation (RAG) (Figure 11). This architecture is conceptual and has not yet been validated as an operational irrigation DT.

Figure 11. Full-stack architecture of the LLM-integrated cognitive digital twin for irrigation optimization, including (A) a physiological modeling engine and (B) an LLM-based orchestrator.

5. Summary and Outlook

Irrigation methods and scheduling have been optimized in multiple ways across various disciplines to improve agricultural water use efficiency. The measurements of irrigation systems are becoming precise, multitudinous, and multidimensional. Intelligent data analysis can improve data utilization but mostly provides discrete perception. AI-driven modelling shows good performance but can cause prediction lag if data are not presented in time. DTs become a valuable solution to provide a physical-to-virtual platform for real-world application benefit from bi-directional data flow, intelligent modelling and closed-loop execution. In addition, one can develop different DTs in multiple scenarios, such as DTs of irrigation equipment or plant itself. If those DTs share the same environment, the hybrid combination system can also be developed. The reviewed studies show the following results: IoT-enabled DTs are suitable for real-time monitoring and system connectivity; remote-sensing-enabled DTs extend spatial coverage for crop and soil observation; simulation-driven DTs provide detailed representations of crop, soil, and hydraulic processes and support virtual testing of irrigation strategies. However, the degree of physical–virtual synchronization and feedback differs considerably among existing studies, and not all reported systems have achieved fully operational closed-loop DTs.

Several limitations still restrict practical deployment. These include heterogeneous data quality, sensor uncertainty, interoperability, model calibration, computational burden, and limited field validation. High model accuracy does not always translate directly into water savings, yield improvement, or higher water use efficiency. Many studies are still validated at a single site or over a short period, which limits their transferability across crops, climates, soils, and management conditions. In addition, agricultural operators are more likely to trust their own experiences than digital technologies. Including human interactions in DTs can help the user to accept the method, while helping the digital system more closely reflect the real-world system.

Future research should therefore focus on improving multi-source data integration, online model calibration, computational efficiency, and hierarchical multi-scale DT architectures. Long-term and multi-site field validation is also needed to connect model-level performance with practical outcomes such as water savings, crop yield, WUE, energy consumption, and operational reliability. Emerging technologies such as LLMs and reinforcement learning may further improve human–machine interaction and adaptive decision-making, but their use in irrigation DTs remains largely prospective and requires careful field validation before autonomous deployment.

Author Contributions

Conceptualization, F.C.; methodology, F.C.; software, F.C., Y.Y. and X.J.; validation, F.C.; formal analysis, F.C.; investigation, F.C.; resources, F.C.; data curation, F.C. and Y.Y.; writing—original draft preparation, F.C., Y.Y. and X.J.; writing—review and editing, F.C., L.M. and H.S.; visualization, Y.Y.; supervision, F.C.; project administration, F.C.; funding acquisition, F.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (grant number: 62476243) and Ningbo Science and Technology Innovation Yongjiang 2035 Key Technology Project (grant number: 2024Z265).

Data Availability Statement

Data are available from the corresponding author upon reasonable request due to privacy.

Acknowledgments

During the preparation of this manuscript, the authors used Google Gemini 3.6 flash for the purposes of drawing figures. The authors have reviewed and edited the output and take full responsibility for the content of this publication.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. Schweitzer, L.; Noblet, J. Water contamination and pollution. In Green Chemistry; Elsevier: Amsterdam, The Netherlands, 2018; pp. 261–290. [Google Scholar] [CrossRef] [Scilit]
2. Wallace, J.S.; Gregory, P.J. Water resources and their use in food production systems. Aquat. Sci.2002, 64, 363–375. [Google Scholar] [CrossRef] [Scilit]
3. Gosling, S.N.; Arnell, N.W. A global assessment of the impact of climate change on water scarcity. Clim. Change2016, 134, 371–385. [Google Scholar] [CrossRef] [Scilit]
4. Prosdocimi, M.; Tarolli, P.; Cerdà, A. Mulching practices for reducing soil water erosion: A review. Earth-Sci. Rev.2016, 161, 191–203. [Google Scholar] [CrossRef] [Scilit]
5. Xiang, X.; Li, Q.; Khan, S.; Khalaf, O.I. Urban water resource management for sustainable environment planning using artificial intelligence techniques. Environ. Impact Assess. Rev.2021, 86, 106515. [Google Scholar] [CrossRef] [Scilit]
6. Bwambale, E.; Naangmenyele, Z.; Iradukunda, P.; Agboka, K.M.; Houessou-Dossou, E.A.Y.; Akansake, D.A.; Bisa, M.E.; Hamadou, A.-A.H.; Hakizayezu, J.; Onofua, O.E.; et al. Towards precision irrigation management: A review of GIS, remote sensing and emerging technologies. Cogent Eng.2022, 9, 2100573. [Google Scholar] [CrossRef] [Scilit]
7. Wallace, J.S. Increasing agricultural water use efficiency to meet future food production. Agric. Ecosyst. Environ.2000, 82, 105–119. [Google Scholar] [CrossRef] [Scilit]
8. Jones, H.G. Irrigation scheduling: Advantages and pitfalls of plant-based methods. J. Exp. Bot.2004, 55, 2427–2436. [Google Scholar] [CrossRef] [Scilit] [PubMed]
9. Ebrahimian, E.; Seyyedi, S.M.; Bybordi, A.; Damalas, C.A. Seed yield and oil quality of sunflower, safflower, and sesame under different levels of irrigation water availability. Agric. Water Manag.2019, 218, 149–157. [Google Scholar] [CrossRef] [Scilit]
10. Levidow, L.; Zaccaria, D.; Maia, R.; Vivas, E.; Todorovic, M.; Scardigno, A. Improving water-efficient irrigation: Prospects and difficulties of innovative practices. Agric. Water Manag.2014, 146, 84–94. [Google Scholar] [CrossRef] [Scilit]
11. Grafton, R.Q.; Williams, J.; Perry, C.J.; Molle, F.; Ringler, C.; Steduto, P.; Udall, B.; Wheeler, S.A.; Wang, Y.; Garrick, D.; et al. The paradox of irrigation efficiency. Science2018, 361, 748–750. [Google Scholar] [CrossRef] [Scilit] [PubMed]
12. Gu, Z.; Qi, Z.; Burghate, R.; Yuan, S.; Jiao, X.; Xu, J. Irrigation scheduling approaches and applications: A review. J. Irrig. Drain. Eng.2020, 146, 04020007. [Google Scholar] [CrossRef] [Scilit]
13. Davis, S.L.; Dukes, M.D. Irrigation scheduling performance by evapotranspiration-based controllers. Agric. Water Manag.2010, 98, 19–28. [Google Scholar] [CrossRef] [Scilit]
14. Thompson, R.B.; Gallardo, M.; Valdez, L.C.; Fernandez, M.D. Determination of lower limits for irrigation management using in situ assessments of apparent crop water uptake made with volumetric soil water content sensors. Agric. Water Manag.2007, 92, 13–28. [Google Scholar] [CrossRef] [Scilit]
15. Veysi, S.; Naseri, A.A.; Hamzeh, S.; Bartholomeus, H. A satellite based crop water stress index for irrigation scheduling in sugarcane fields. Agric. Water Manag.2017, 189, 70–86. [Google Scholar] [CrossRef] [Scilit]
16. Himanshu, S.K.; Ale, S.; Bordovsky, J.; Darapuneni, M. Evaluation of crop-growth-stage-based deficit irrigation strategies for cotton production in the Southern High Plains. Agric. Water Manag.2019, 225, 105782. [Google Scholar] [CrossRef] [Scilit]
17. Thorp, K.R.; Hunsaker, D.J.; Bronson, K.F.; Andrade-Sanchez, P.; Barnes, E.M. Cotton irrigation scheduling using a crop growth model and FAO-56 methods: Field and simulation studies. Trans. ASABE2017, 60, 2023–2039. [Google Scholar] [CrossRef] [Scilit]
18. Linker, R.; Ioslovich, I.; Sylaios, G.; Plauborg, F.; Battilani, A. Optimal model-based deficit irrigation scheduling using AquaCrop: A simulation study with cotton, potato and tomato. Agric. Water Manag.2016, 163, 236–243. [Google Scholar] [CrossRef] [Scilit]
19. Liu, J.; Chang, F.; Yang, J.; Jie, X.; Lu, C.; Wang, C.; Xie, L.; Ma, L.; Su, H. Deep Reinforcement Learning for irrigation optimization: Advantages, opportunities, and challenges. Agric. Water Manag.2025, 322, 110030. [Google Scholar] [CrossRef] [Scilit]
20. Bwambale, E.; Abagale, F.K.; Anornu, G.K. Smart irrigation monitoring and control strategies for improving water use efficiency in precision agriculture: A review. Agric. Water Manag.2022, 260, 107324. [Google Scholar] [CrossRef] [Scilit]
21. Abioye, E.A.; Abidin, M.S.Z.; Mahmud, M.S.A.; Buyamin, S.; Ishak, M.H.I.; Abd Rahman, M.K.I.; Otuoze, A.O.; Onotu, P.; Ramli, M.S.A. A review on monitoring and advanced control strategies for precision irrigation. Comput. Electron. Agric.2020, 173, 105441. [Google Scholar] [CrossRef] [Scilit]
22. Hamami, L.; Nassereddine, B. Application of wireless sensor networks in the field of irrigation: A review. Comput. Electron. Agric.2020, 179, 105782. [Google Scholar] [CrossRef] [Scilit]
23. Harun, A.N.; Mohamed, N.; Ahmad, R.; Rahim, A.R.A.; Ani, N.N. Improved Internet of Things (IoT) monitoring system for growth optimization of Brassica chinensis. Comput. Electron. Agric.2019, 164, 104836. [Google Scholar] [CrossRef] [Scilit]
24. Klein, L.J.; Hamann, H.F.; Hinds, N.; Guha, S.; Sanchez, L.; Sams, B.; Dokoozlian, N. Closed loop controlled precision irrigation sensor network. IEEE Internet Things J.2018, 5, 4580–4588. [Google Scholar] [CrossRef] [Scilit]
25. Farooq, H.; Rehman, H.U.; Javed, A.; Shoukat, M.; Dudely, S. A Review on Smart IoT Based Farming. Ann. Emerg. Technol. Comput.2020, 4, 17–28. [Google Scholar] [CrossRef] [Scilit]
26. Aghaabbasi, M.; Ganjouei, F.A. Prospects of Digital Twin Systems for Smart Irrigation: A Comprehensive Review. Water Conserv. Sci. Eng.2026, 11, 27. [Google Scholar] [CrossRef] [Scilit]
27. Uhlemann, T.H.J.; Schock, C.; Lehmann, C.; Freiberger, S.; Steinhilper, R. The digital twin: Demonstrating the potential of real time data acquisition in production systems. Procedia Manuf.2017, 9, 113–120. [Google Scholar] [CrossRef] [Scilit]
28. Pylianidis, C.; Osinga, S.; Athanasiadis, I.N. Introducing digital twins to agriculture. Comput. Electron. Agric.2021, 184, 105942. [Google Scholar] [CrossRef] [Scilit]
29. Manocha, A.; Sood, S.K.; Bhatia, M. IoT-digital twin-inspired smart irrigation approach for optimal water utilization. Sustain. Comput. Inform. Syst.2024, 41, 100947. [Google Scholar] [CrossRef] [Scilit]
30. Alves, R.G.; Maia, R.F.; Lima, F. Development of a Digital Twin for smart farming: Irrigation management system for water saving. J. Clean. Prod.2023, 388, 135920. [Google Scholar] [CrossRef] [Scilit]
31. Purcell, W.; Neubauer, T. Digital Twins in Agriculture: A State-of-the-art review. Smart Agric. Technol.2023, 3, 100094. [Google Scholar] [CrossRef] [Scilit]
32. Grieves, M. Digital twin: Manufacturing excellence through virtual factory replication. White Pap.2014, 1, 1–7. [Google Scholar]
33. Grieves, M.; Vickers, J. Digital twin: Mitigating unpredictable, undesirable emergent behavior in complex systems. In Transdisciplinary Perspectives on Complex Systems: New Findings and Approaches; Springer International Publishing: Cham, Switzerland, 2016; pp. 85–113. [Google Scholar]
34. Angin, P.; Anisi, M.H.; Göksel, F.; Gürsoy, C.; Büyükgülcü, A. Agrilora: A digital twin framework for smart agriculture. J. Wirel. Mob. Netw. Ubiquitous Comput. Dependable Appl.2020, 11, 77–96. [Google Scholar] [CrossRef]
35. Tao, F.; Qi, Q.; Wang, L.; Nee, A.Y.C. Digital Twins and Cyber–Physical Systems toward Smart Manufacturing and Industry 4.0: Correlation and Comparison. Engineering2019, 5, 653–661. [Google Scholar] [CrossRef] [Scilit]
36. Stark, R.; Damerau, T. Digital Twin. In CIRP Encyclopedia of Production Engineering; Springer: Berlin/Heidelberg, Germany, 2019; pp. 1–8. [Google Scholar] [CrossRef] [Scilit]
37. Kim, S.; Heo, S. An agricultural digital twin for mandarins demonstrates the potential for individualized agriculture. Nat. Commun.2024, 15, 1561. [Google Scholar] [CrossRef] [Scilit] [PubMed]
38. Verdouw, C.; Tekinerdogan, B.; Beulens, A.; Wolfert, S. Digital twins in smart farming. Agric. Syst.2021, 189, 103046. [Google Scholar] [CrossRef] [Scilit]
39. Peladarinos, N.; Piromalis, D.; Cheimaras, V.; Tserepas, E.; Munteanu, R.A.; Papageorgas, P. Enhancing Smart Agriculture by Implementing Digital Twins: A Comprehensive Review. Sensors2023, 23, 7128. [Google Scholar] [CrossRef] [Scilit] [PubMed]
40. Escribà-Gelonch, M.; Liang, S.; Van Schalkwyk, P.; Fisk, I.; Long, N.V.D.; Hessel, V. Digital Twins in Agriculture: Orchestration and Applications. J. Agric. Food Chem.2024, 72, 10737–10752. [Google Scholar] [CrossRef] [Scilit] [PubMed]
41. Skobelev, P.O.; Mayorov, I.V.; Simonova, E.V.; Goryanin, O.I.; Zhilyaev, A.A.; Tabachinskiy, A.S.; Yalovenko, V.V. Development of models and methods for creating a digital twin of plant within the cyber-physical system for precision farming management. J. Phys. Conf. Ser.2020, 1703, 12022. [Google Scholar] [CrossRef] [Scilit]
42. Parewai, I.; Köppen, M. A digital twin approach for soil moisture measurement with physically based rendering simulations and machine learning. Electronics2025, 14, 395. [Google Scholar] [CrossRef] [Scilit]
43. Zhao, X.; Wang, W.; Wen, L.; Chen, Z.; Wu, S.; Zhou, K.; Sun, M.; Xu, L.; Hu, B.; Wu, C. Digital twins in smart farming: An autoware-based simulator for autonomous agricultural vehicles. Int. J. Agric. Biol. Eng.2023, 16, 184–189. [Google Scholar] [CrossRef] [Scilit]
44. Ait-Alla, A.; Kreutz, M.; Rippel, D.; Lütjen, M.; Freitag, M. Simulation-based Analysis of the Interaction of a Physical and a Digital Twin in a Cyber-Physical Production System. IFAC-PapersOnLine2019, 52, 1331–1336. [Google Scholar] [CrossRef] [Scilit]
45. Goap, A.; Sharma, D.; Shukla, A.K.; Rama Krishna, C. An IoT based smart irrigation management system using Machine learning and open source technologies. Comput. Electron. Agric.2018, 155, 41–49. [Google Scholar] [CrossRef] [Scilit]
46. Maguire, M.; Neale, C.M.U. Irrigation scheduling using hybrid remote sensing-based evapotranspiration model informed by unmanned aerial system acquired multispectral and thermal imagery. In The Autonomous Air and Ground Sensing Systems for Agricultural Optimization and Phenotyping VII; SPIE: Bellingham, WA, USA, 2022. [Google Scholar] [CrossRef] [Scilit]
47. Mogensen, R.S.; Rodriguez, I.; Schou, C.; Mortensen, S.; Sørensen, M.S. Evaluation of the impact of wireless communication in production via factory digital twins. Manuf. Lett.2021, 28, 1–5. [Google Scholar] [CrossRef] [Scilit]
48. Yun, S.; Park, J.H.; Kim, W.T. Data-centric middleware based digital twin platform for dependable cyber-physical systems. In 2017 Ninth International Conference on Ubiquitous and Future Networks (ICUFN); IEEE: Piscataway, NJ, USA, 2017; pp. 922–926. [Google Scholar] [CrossRef] [Scilit]
49. Tarneberg, W.; Skarin, P.; Gehrmann, C.; Kihl, M. Prototyping intrusion detection in an industrial cloud-native digital twin. In 2021 22nd IEEE International Conference on Industrial Technology (ICIT); IEEE: Piscataway, NJ, USA, 2021; pp. 749–755. [Google Scholar] [CrossRef] [Scilit]
50. Liu, Z.; Meyendorf, N.; Blasch, E.; Tsukada, K.; Liao, M.; Mrad, N. The role of data fusion in predictive maintenance using digital twins. In Handbook of Nondestructive Evaluation 4.0; Springer: Cham, Switzerland, 2025; pp. 1–23. [Google Scholar] [CrossRef] [Scilit]
51. Li, X.; Feng, M.; Ran, Y.; Su, Y.; Liu, F.; Huang, C.; Shen, H.; Xiao, Q.; Su, J.; Yuan, S.; et al. Big Data in Earth system science and progress towards a digital twin. Nat. Rev. Earth Environ.2023, 4, 319–332. [Google Scholar] [CrossRef] [Scilit]
52. Moser, A.; Appl, C.; Brüning, S.; Hass, V.C. Mechanistic Mathematical Models as a Basis for Digital Twins. In Digital Twins, Advances in Biochemical Engineering/Biotechnology; Springer International Publishing: Cham, Switzerland, 2020; pp. 133–180. [Google Scholar] [CrossRef] [Scilit] [PubMed]
53. Lohtander, M.; Ahonen, N.; Lanz, M.; Ratava, J.; Kaakkunen, J. Micro Manufacturing Unit and the Corresponding 3D-Model for the Digital Twin. Procedia Manuf.2018, 25, 55–61. [Google Scholar] [CrossRef] [Scilit]
54. Ren, Z.; Wan, J.; Deng, P. Machine-Learning-Driven Digital Twin for Lifecycle Management of Complex Equipment. IEEE Trans. Emerg. Top. Comput.2022, 10, 9–22. [Google Scholar] [CrossRef] [Scilit]
55. Ritto, T.G.; Rochinha, F.A. Digital twin, physics-based model, and machine learning applied to damage detection in structures. Mech. Syst. Signal Process.2021, 155, 107614. [Google Scholar] [CrossRef] [Scilit]
56. Mitsanis, C.; Hurst, W.; Tekinerdogan, B. A 3D functional plant modelling framework for agricultural digital twins. Comput. Electron. Agric.2024, 218, 108733. [Google Scholar] [CrossRef] [Scilit]
57. Chen, H.; Fang, C.; Xiao, X. Visualization of Environmental Sensing Data in the Lake-Oriented Digital Twin World: Poyang Lake as an Example. Remote Sens.2023, 15, 1193. [Google Scholar] [CrossRef] [Scilit]
58. Liu, S.; Lu, S.; Li, J.; Sun, X.; Lu, Y.; Bao, J. Machining process-oriented monitoring method based on digital twin via augmented reality. Int. J. Adv. Manuf. Technol.2021, 113, 3491–3508. [Google Scholar] [CrossRef] [Scilit]
59. Liu, L.; Guo, K.; Gao, Z.; Li, J.; Sun, J. Digital Twin-Driven Adaptive Scheduling for Flexible Job Shops. Sustainability2022, 14, 5340. [Google Scholar] [CrossRef] [Scilit]
60. Thürer, M.; Li, S.S.; Qu, T. Digital Twin Architecture for Production Logistics: The Critical Role of Programmable Logic Controllers (PLCs). Procedia Comput. Sci.2022, 200, 710–717. [Google Scholar] [CrossRef] [Scilit]
61. Sun, C.; Shi, V.G. PhysiNet: A combination of physics-based model and neural network model for digital twins. Int. J. Intell. Syst.2022, 37, 5443–5456. [Google Scholar] [CrossRef] [Scilit]
62. Croatti, A.; Gabellini, M.; Montagna, S.; Ricci, A. On the Integration of Agents and Digital Twins in Healthcare. J. Med. Syst.2020, 44, 161. [Google Scholar] [CrossRef] [Scilit] [PubMed]
63. Deng, M.; Yao, Z.; Li, X.; Wang, H.; Nallanathan, A.; Zhang, Z. Dynamic Multi-Objective AWPSO in DT-Assisted UAV Cooperative Task Assignment. IEEE J. Sel. Areas Commun.2023, 41, 3444–3460. [Google Scholar] [CrossRef] [Scilit]
64. Yan, Q.; Wang, H.; Wu, F. Digital twin-enabled dynamic scheduling with preventive maintenance using a double-layer Q-learning algorithm. Comput. Oper. Res.2022, 144, 105823. [Google Scholar] [CrossRef] [Scilit]
65. Putz, B.; Dietz, M.; Empl, P.; Pernul, G. EtherTwin: Blockchain-based Secure Digital Twin Information Management. Inf. Process. Manag.2021, 58, 102425. [Google Scholar] [CrossRef] [Scilit]
66. Madeira, R.N.; Santos, P.A.; Java, O.; Priebe, T.; Graça, E.; Sárközi, E.; Asprion, B.; Gómez, R.P.-B. Towards Digital Twins for Multi-Sensor Land and Plant Monitoring. Procedia Comput. Sci.2022, 210, 45–52. [Google Scholar] [CrossRef] [Scilit]
67. Skobelev, P.; Mayorov, I.; Simonova, E.; Goryanin, O.; Zhilyaev, A.; Tabachinskiy, A.; Yalovenko, V. Development of Digital Twin of Plant for Adaptive Calculation of Development Stage Duration and Forecasting Crop Yield in a Cyber-Physical System for Managing Precision Farming. In Cyber-Physical Systems, Studies in Systems, Decision and Control; Springer International Publishing: Cham, Switzerland, 2021; pp. 83–96. [Google Scholar] [CrossRef] [Scilit]
68. Ariesen-Verschuur, N.; Verdouw, C.; Tekinerdogan, B. Digital Twins in greenhouse horticulture: A review. Comput. Electron. Agric.2022, 199, 107183. [Google Scholar] [CrossRef] [Scilit]
69. Nemtinov, K.; Eruslanova, M.; Zazulya, A.; Nemtinova, Y.; Salih, H.S. Creating a digital twin of an agricultural machine. MATEC Web Conf.2020, 329, 05002. [Google Scholar] [CrossRef] [Scilit]
70. Dai, M.; Shen, Y.; Li, X.; Liu, J.; Zhang, S.; Miao, H. Digital Twin System of Pest Management Driven by Data and Model Fusion. Agriculture2024, 14, 1099. [Google Scholar] [CrossRef] [Scilit]
71. Appl, C.; Moser, A.; Baganz, F.; Hass, V.C. Digital Twins for Bioprocess Control Strategy Development and Realisation. In Digital Twins, Advances in Biochemical Engineering/Biotechnology; Herwig, C., Pörtner, R., Möller, J., Eds.; Springer International Publishing: Cham, Switzerland, 2020; pp. 63–94. [Google Scholar] [CrossRef] [Scilit] [PubMed]
72. Wangere, J.; Sinde, R.; Ally, M.; Mbwambo, O. A Digital Twin Approach and Challenges for Real-Time Automated Surface-Drip Irrigation Monitoring: A Case of Arusha Tanzania. In Smart and Secure Embedded and Mobile Systems. ICTA-EMOS; Springer: Cham, Switzerland, 2022. [Google Scholar] [CrossRef] [Scilit]
73. Merizig, A.; Mokhtari, B.; Harabi, Y.; Ayad, S.; Kassimi, D. Digital Twin-Based Smart Irrigation for Sustainable Water Management in Arid Areas. In 2025 7th International Conference on Pattern Analysis and Intelligent Systems (PAIS); IEEE: Piscataway, NJ, USA, 2025; pp. 1–6. [Google Scholar] [CrossRef] [Scilit]
74. Adamo, T.; Caivano, D.; Colizzi, L.; Dimauro, G.; Guerriero, E. Optimization of irrigation and fertigation in smart agriculture: An IoT-based micro-services framework. Smart Agric. Technol.2025, 11, 100885. [Google Scholar] [CrossRef] [Scilit]
75. Xu, X.; Gao, F.; Xiong, D.; Fan, Z.; Xiong, S.; Dong, P.; Qiao, H.; Ma, X. Digital twin-based winter wheat growth simulation and optimization. Field Crops Res.2025, 329, 109953. [Google Scholar] [CrossRef] [Scilit]
76. Qin, M.; Zhang, C.; Ma, G.; Ma, Y.; Feng, X.; Li, P.; Feng, K. Digital twin-enabled intelligent irrigation-drainage system for precision water-salt management in saline agroecosystems. Agric. Water Manag.2025, 322, 109957. [Google Scholar] [CrossRef] [Scilit]
77. Pavón, R.M.; Galao, O.; Pardo, M.A.; Alberti, M.G. Digital Twin development and implementation for the management of irrigation networks. Expert Syst. Appl.2026, 297, 129330. [Google Scholar] [CrossRef] [Scilit]
78. Ramesh, K.; Sasirekha, G.V.K.; Rao, M.; Bapat, J.; Das, D. Digital twin based what-if simulation of security attacks in smart irrigation systems. In 2024 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT); IEEE: Piscataway, NJ, USA, 2024; pp. 1–6. [Google Scholar] [CrossRef] [Scilit]
79. Awais, M.; Li, W.; Li, H.; Cheema, M.J.M.; Hussain, S.; Liu, C. Optimization of intelligent irrigation systems for smart farming using multi-spectral unmanned aerial vehicle and digital twins modeling. Environ. Sci. Proc.2022, 23, 13. [Google Scholar] [CrossRef] [Scilit]
80. Pal, P.; Landivar-Bowles, J.; Landivar-Scott, J.; Duffield, N.; Nowka, K.; Jung, J.; Chang, A.; Lee, K.; Zhao, L.; Bhandari, M. Unmanned aerial system and machine learning driven Digital-Twin framework for in-season cotton growth forecasting. Comput. Electron. Agric.2025, 228, 109589. [Google Scholar] [CrossRef] [Scilit]
81. Bellvert, J.; Pelechá, A.; Pamies-Sans, M.; Virgili, J.; Torres, M.; Casadesús, J. Assimilation of sentinel-2 biophysical variables into a digital twin for the automated irrigation scheduling of a vineyard. Water2023, 15, 2506. [Google Scholar] [CrossRef] [Scilit]
82. Rodrigues, H.; Ceddia, M.B.; Tassinari, W.; Vasques, G.M.; Brandão, Z.N.; Morais, J.P.; Oliveira, R.P.; Neves, M.L.; Tavares, S.R. Remote and Proximal Sensors Data Fusion: Digital Twins in Irrigation Management Zoning. Sensors2024, 24, 5742. [Google Scholar] [CrossRef] [Scilit] [PubMed]
83. Camps, A.; López-Martínez, C.; Gonga, A.; Gracia, G.; Perez-Portero, A.; Alonso-Gonzalez, A.; Vall-Llossera, M.; Park, H.; Perez, V.; Caselles, O.; et al. Ai4water: A digital twin for irrigated agriculture. In IGARSS 2024—2024 IEEE International Geoscience and Remote Sensing Symposium (IGARSS); IEEE: Piscataway, NJ, USA, 2024; pp. 3545–3548. [Google Scholar] [CrossRef] [Scilit]
84. Millán, S.; Montesinos, C.; Casadesus, J.; Vadillo, J.M.; Campillo, C. Use of a digital twin for water efficient management in a processing tomato commercial farm. Agronomy2025, 15, 2132. [Google Scholar] [CrossRef] [Scilit]
85. Moreira, M.; Mourato, S.; Rodrigues, C.; Silva, S.; Guimarães, R.; Chibeles, C. Building a Digital Twin for the Management of Pressurised Collective Irrigation Systems. In Proceedings of the 1st International Conference on Water Energy Food and Sustainability (ICoWEFS 2021); Springer: Cham, Switzerland, 2021. [Google Scholar] [CrossRef] [Scilit]
86. Liu, W.; Guan, G.; Tian, X.; Cao, Z.; Chen, X.; Shi, L. A real-time refined roughness estimation framework for the digital twin model calibration of irrigation canal systems. J. Irrig. Drain. Eng.2024, 150, 04023034. [Google Scholar] [CrossRef] [Scilit]
87. Preite, L.; Solari, F.; Vignali, G. Water management optimization in agriculture: A digital model development. Water Resour. Manag.2025, 39, 1261–1279. [Google Scholar] [CrossRef] [Scilit]
88. El Hachimi, C.; Belaqziz, S.; Khabba, S.; Karjoun, H.; Kharrou, M.H.; Ait Hssaine, B.; Er-Raki, S.; Daccache, A.; Chehbouni, A.; Weng, Q. Towards collective intelligence in agriculture: Deep reinforcement learning and digital twins for efficient management of collective irrigation water distribution systems. Energy Nexus2025, 20, 100599. [Google Scholar] [CrossRef] [Scilit]
89. Yang, X.; Hou, M.; Qi, F.; Dong, A.; Wang, Z. Policy Evolution and Frameworks in China’s Irrigation Management Informatization: Insights from China’s Experience from the Basics to AI-Driven Digital Twin Innovations. Irrig. Drain.2025, 74, 2197–2216. [Google Scholar] [CrossRef] [Scilit]
90. Yang, J.; Chang, F.; Ma, L.; Lei, X.; Su, H. Optimizing irrigation management based on digital plant using Image-based 3D reconstruction. In 2025 China Automation Congress (CAC); IEEE: Piscataway, NJ, USA, 2025; pp. 3147–3151. [Google Scholar] [CrossRef] [Scilit]
91. Hu, B.G.; De Reffye, P.; Zhao, X.; Yan, H.P.; Kang, M.Z. Greenlab: A new methodology towards plant functional-structural model–structural part. In Plant Growth Modelling and Applications; TsingHua University Press: Beijing, China; Springer: Berlin/Heidelberg, Germany, 2003; pp. 21–35. Available online: https://inria.hal.science/inria-00121235 (accessed on 1 July 2026).
92. Postma, J.A.; Kuppe, C.; Owen, M.R.; Mellor, N.; Griffiths, M.; Bennett, M.J.; Lynch, J.P.; Watt, M. OpenSimRoot: Widening the scope and application of root architectural models. New Phytol.2017, 215, 1274–1286. [Google Scholar] [CrossRef] [Scilit] [PubMed]
93. Xu, M.; Wang, X.; Kang, M.; Chang, F.; Hua, J.; Wang, H.; De Reffye, P. IrriTwin: Integrating reinforcement learning with soil water and crop models toward digital-twin-oriented irrigation scheduling. J. Hydrol.2026, 678, 136034. [Google Scholar] [CrossRef] [Scilit]
94. Liu, J.; Chang, F.; Wang, X.; Kang, M.; Lu, C.; Wang, C.; Hu, S.; Li, Y.; Ma, L.; Su, H. Smart irrigation scheduling for crop production using a crop model and improved deep reinforcement learning. Agriculture2025, 15, 2569. [Google Scholar] [CrossRef] [Scilit]
95. Teh, H.Y.; Kempa-Liehr, A.W.; Wang, K.I.K. Sensor data quality: A systematic review. J. Big Data2020, 7, 11. [Google Scholar] [CrossRef] [Scilit]
96. Huang, Y.; Chang, F.; Tao, Y.; Zhao, Y.; Ma, L.; Su, H. Few-shot learning based on Attn-CutMix and task-adaptive transformer for the recognition of cotton growth state. Comput. Electron. Agric.2022, 202, 107406. [Google Scholar] [CrossRef] [Scilit]
97. Rashid, S.; Abbas, H.; Ali, I.; Khalid, A.; Tesgera Agama, F.; Aldien, M.S. Evaluating physics informed neural networks for water contamination risk prediction and environmental sustainability. Sci. Rep.2025, 16, 603. [Google Scholar] [CrossRef] [Scilit] [PubMed]
98. Mahmood, F.; Ghiat, I.; Govindan, R.; Al-Ansari, T. Reduced-order modelling (ROM) approach for optimal microclimate control in agricultural greenhouses. Comput. Aided Chem. Eng.2020, 48, 1879–1884. [Google Scholar] [CrossRef] [Scilit]
99. Basso, B.; Antle, J. Digital agriculture to design sustainable agricultural systems. Nat. Sustain.2020, 3, 254–256. [Google Scholar] [CrossRef] [Scilit]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license.

Article Metrics

Citations

Multiple requests from the same IP address are counted as one view.