Development and Validation of an IoF-Based Sensing Architecture for Microclimate Monitoring: Quantifying Spatial ET₀ Bias in Mediterranean Viticulture

1. Introduction

The depletion of freshwater resources and the increasing unpredictability of precipitation patterns pose critical challenges to sustainable agricultural production worldwide. Among the strategies to address water scarcity, precision irrigation management stands out as one of the most effective approaches, as it directly links water application to actual crop requirements rather than relying on generalized schedules [1,2]. Central to precision irrigation is the accurate estimation of reference evapotranspiration (ET 0), which serves as the primary input for calculating crop water demand (ET C) through crop coefficient (K C) relationships [3]. Widely recognized as the benchmark for estimation, the FAO Penman–Monteith method [4] necessitates a comprehensive set of inputs, including air temperature, relative humidity, wind speed, and solar radiation [5].

In practice, these meteorological inputs are typically sourced from government-operated weather stations, which serve broad geographic areas. Specifically, in the context of Türkiye, the climatic parameters utilized in the study are traditionally analyzed based on official records obtained from the national meteorological network, managed by the Turkish State Meteorological Service (TSMS). However, considerable evidence suggests that spatial heterogeneity in microclimatic conditions, driven by differences in topography, land cover, proximity to urban heat islands, and local vegetation patterns, can introduce significant deviations between regional meteorological observations and actual field-level conditions [6].

Achieving accurate results in ET C parameter optimization through various computational and statistical ET 0 estimation methodologies [7,8,9,10,11,12,13,14,15] remains essential for the effective implementation of precision irrigation strategies. However, there remains a need for site-specific benchmark comparison of these models using localized, real-time data to overcome the inherent limitations of distant TSMS stations.

The integration of Internet of Farming (IoF) frameworks and low-cost sensor platforms has recently become instrumental in the high-fidelity acquisition of environmental variables that govern evapotranspiration dynamics. Technological advances have made it both technically and economically feasible to deploy field-level monitoring systems tailored to individual agricultural plots [16,17,18]. Several studies have demonstrated the potential of these architectures; for instance, Abba et al. [19] reported satisfactory performance for a smart IoT-based irrigation system, while Rodríguez-Robles et al. [20] developed an autonomous sensor network for rural environments. Similarly, Rajak et al. [21] emphasized the role of smart sensors in enhancing crop productivity. In a complementary approach, Ikram et al. [22] demonstrated IoT-based decision systems for yield maximization. Studies in the Thessaly plain [23] further integrated surface measurements with remote-sensing data to evaluate irrigation needs against the FAO framework. Furthermore, the deployment of Machine Learning (ML) algorithms alongside these IoT frameworks has shown superior efficacy in predictive modeling of ET values [24,25,26,27]. Such precision in forecasting serves as a cornerstone for optimizing irrigation scheduling and enhancing overall agricultural productivity [28]. Complementing these efforts, recent systematic reviews have synthesized trends and challenges in IoT-based irrigation sensing more broadly [29], while advances in computationally efficient ET estimation approaches continue to support real-time smart-agriculture applications [30].

While the broader literature predominantly employs the term Internet of Things (IoT) to describe sensor-based connectivity in agricultural contexts, the present study adopts the more domain-specific concept of IoF. IoF extends the generic IoT paradigm by explicitly embedding agricultural operational logic, including crop-specific data acquisition, agrometeorological processing, and irrigation decision-support, into the system architecture [17]. Unlike general-purpose IoT deployments, IoF frameworks are designed from the outset to capture, transmit, and interpret the biological and physical variables that govern plant-water relations under field conditions, making the term more precise and scientifically appropriate for the system described herein. This terminological distinction carries substantive scientific implications that directly shape the present study’s design and contribution. At the system design level, the IoF paradigm mandates the co-integration of crop-specific sensing priorities, agrometeorological processing logic, and irrigation decision-support within a unified architecture, commitments directly reflected in the sensor selection, canopy-level placement, and duty-cycle management strategy employed here; a generic IoT deployment, being agnostic to plant-water relations, would not by design require the simultaneous acquisition of microclimate, soil moisture, and irrigation water quality parameters. At the scientific contribution level, the IoF framing shifts the core research question from whether sensors can transmit data reliably (an IoT-level concern) to whether field-embedded agrometeorological sensing can resolve the spatial ET 0 bias introduced by reliance on distant regional stations; this reframing positions the agronomic consequences of spatial data mismatch, rather than sensor engineering performance alone, as the primary scientific output of the study. To better position the proposed system within the current state of the art, representative low-cost agricultural sensing platforms reported in the recent literature are compared in Table 1.

Table 1. Comparison of representative low-cost agricultural IoT sensing platforms with the proposed IoF framework.

As summarized in Table 1, previous low-cost agricultural IoT sensing platforms have primarily focused on individual functionalities such as microclimate monitoring, smart irrigation, greenhouse automation, or ET 0 estimation. The novelty of the proposed system does not arise from the use of individual hardware components, which are mature and commercially available, but from their functional integration into an application-specific IoF framework for precision irrigation management. In contrast, the proposed IoF framework combines autonomous solar-powered operation, real-time vineyard microclimate monitoring, FAO-56 Penman–Monteith-based ET 0 estimation, validation against TSMS, and irrigation-oriented decision support within a unified three-tier architecture, thereby providing a more comprehensive framework for precision irrigation management.

To further situate the proposed framework within domestic and international practice, recent reviews and field deployments illustrate the diversity of application-specific IoF/IoT sub-implementations that have emerged internationally, including wireless energy-balance sensor networks for crop ET determination [31], IoT-based smart reference-ET estimation combined with ensemble machine learning [32], IoT-based microclimate monitoring for irrigation-water optimization [33], IoT-driven fertigation scheduling and greenhouse microclimate monitoring [34], UAV- and ground-based canopy transpiration mapping integrated with weather data [35], and IoT-integrated digital-twin frameworks for controlled-environment microclimate control [36] and an electronic-atmometer-based instrumentation design for evapotranspiration measurement [37], alongside broader systematic reviews of IoT sensing trends in irrigation management [29] and in precision agriculture more generally [38]. Within the Turkish agricultural engineering literature, comparable IoT-based prototypes have similarly been reported [17], though generally without the crop-specific agrometeorological integration that defines the IoF paradigm adopted here. This diversity of deployed systems underscores that IoF, unlike generic IoT, is not defined by a single hardware configuration but by the functional co-integration of agrometeorological sensing and irrigation decision logic, precisely the criterion against which the present framework is positioned in Table 1.

Despite these advances, existing low-cost agricultural sensing platforms generally focus on individual functionalities such as environmental monitoring, irrigation automation, or ET 0 estimation. Consequently, comprehensive IoF frameworks capable of integrating autonomous field operation, continuous microclimate monitoring, FAO-56 Penman–Monteith-based ET 0 estimation, validation against regional meteorological observations, and irrigation-oriented decision support remain scarce, particularly under commercial vineyard conditions. Much of the current literature focuses on system design, technical validation, and algorithmic efficiency rather than quantifying the agronomic impact of localized versus regional data. Few studies have systematically quantified the magnitude and seasonal dynamics of ET 0 deviations arising from spatial mismatch or translated these deviations into practical irrigation management implications. These questions are particularly vital for high-value crops like the grapevine (Vitis vinifera L.). Türkiye ranks among the top grape-producing countries globally, with a production of approximately 2.52 million tons in 2025 [39,40]. In the Mediterranean climate of the Antalya region, characterized by hot, dry summers, efficient water management is essential, as both over-irrigation and under-irrigation adversely affect yield and quality [41,42].

To address this discrepancy, the present study introduces an IoF-based weather station (Field) designed to capture microclimatic variables, providing a robust framework for the systematic benchmark comparison of localized ET 0 and ETc estimations against official regional observations. To this end, the following research questions are addressed:

• What is the statistical magnitude and temporal pattern of ET 0 deviations between TSMS-derived regional data and localized sources?

• Which meteorological variables function as the primary drivers of these discrepancies?

• What are the practical implications of these variances for crop water requirement estimation and irrigation scheduling?

To achieve this, the study focuses on the following four objectives:

• System development: Design and field deployment of a low-cost, modular IoF-based sensing architecture for real-time acquisition of climatic, soil, and water quality data in a commercial vineyard, providing the localized data infrastructure necessary for all subsequent analyses.

• Statistical quantification: Systematic comparison of ET 0 estimates from localized IoF data versus regional TSMS data to answer: What is the statistical magnitude of ET 0 deviations between TSMS-derived regional data and localized sources?

• Temporal pattern analysis: Regression-based identification of seasonal deviation dynamics and their meteorological drivers, with explicit attribution of the primary physical contributors to ET 0 discrepancy, addressing: Which meteorological variables function as the primary drivers of these discrepancies ?

• Irrigation impact assessment: Translation of ET 0 deviations into ET C differences for grapevine, directly addressing: What are the practical implications of these variances for crop water requirement estimation and irrigation scheduling?

The proposed layered architecture, comprising sensing, processing, and decision-support layers—was developed for deployment in commercial vineyards. Although its transferability to other crops, climatic regions, and management systems appears promising, further validation through multi-site studies is required.

2. Materials and Methods

2.1. Study Area and Experimental Setup

Field data collection was conducted in a vineyard located in the Antalya province of Türkiye (36°53′ N, 30°38′ E, altitude 39 m) covering the full annual cycle from January to December 2021. The region experiences a Mediterranean climate, characterized by hot, dry summers and mild, wet winters. To evaluate the spatial mismatch in meteorological data, observations from the national meteorological network, managed by the Turkish State Meteorological Service (TSMS), were utilized as a regional reference. Specifically, the datasets were retrieved from the TSMS Antalya Regional Station (Station No: 17302), the nearest automated weather station located approximately 15 km away in Antalya city center. The comparative analysis in this study specifically utilizes synchronized daily datasets from both the localized IoF field station and the TSMS records for the identical 2021 calendar year. This framework serves to benchmark in situ IoF sensing capabilities against a standardised regional reference and highlights the profound impact of microclimatic variations on irrigation requirement modeling. This regional baseline provides quality-controlled datasets, including air temperature, relative humidity, wind speed, and solar radiation, standardized according to international meteorological protocols.

2.2. Monitoring System Architecture

Signal acquisition and electrical interfaces. As summarized in Table 2, the proposed IoF platform integrates heterogeneous sensing devices employing different electrical output characteristics according to the measured parameter. Sensors such as air temperature, relative humidity, soil temperature, soil moisture, electrical conductivity, pH, barometric pressure, and solar radiation generate analog or conditioned analog measurement signals that are digitized by the DAQ subsystem using a 12-bit analog-to-digital converter (ADC) with 4096 discrete quantization levels. In contrast, pulse-based sensors, including the rainfall gauge and Hall-effect wind speed sensor, are acquired through digital pulse counting, while wind direction measurements are obtained from the corresponding directional sensing circuitry. Following acquisition, all sensor measurements are synchronized by the ESP32 controller, assigned timestamps, and assembled into a unified dataset prior to cloud transmission. The system utilizes a three-tier IoF architecture (perception, network, and application), as illustrated in Figure 1.

Figure 1. Overall architecture of the proposed Internet of Farming (IoF) platform based on the perception, network, and application layers. The perception layer acquires real-time environmental, soil, and irrigation-related measurements through distributed sensors connected to the data acquisition (DAQ) subsystem. The network layer manages data aggregation, wireless communication, and transmission through the GSM/GPRS cellular network using HTTP POST requests with JSON-formatted payloads. The application layer receives, stores, processes, and visualizes the transmitted data on the cloud server hosted at Akdeniz University, where measurements are archived in a MySQL relational database and made available to users through web- and mobile-based interfaces for irrigation monitoring and decision support.

Table 2. Technical specifications, measurement ranges, accuracy levels, and resolution of the sensors integrated into the IoF field station.

2.2.1. Perception Layer

This physical layer serves as the primary data collection unit, consisting of a modular multi-sensor field station and a Data Acquisition (DAQ) system. The field station integrates 12 sensors mounted on a telescopic mast extending up to 3.5 m, facilitating the real-time acquisition of microclimatic, edaphic, and hydrological variables directly from the vineyard (Figure 2). This layer is specifically responsible for capturing critical parameters such as air temperature, relative humidity, and solar radiation at the crop canopy level. Detailed technical specifications for the integrated sensors, including their respective measurement ranges and accuracy levels, are provided in Table 2.

Figure 2. Functional block diagram of the IoF field station illustrating the integration of the sensing units, DAQ subsystem, ESP32 controller, Raspberry Pi Zero Wireless module, GSM/GPRS communication unit, camera module, and power management system. Analog sensor signals are digitized by the 12-bit DAQ subsystem, while digital sensors communicate directly with the controller t