Accurate mapping of Soil moisture (SM) is essential for effectively monitoring agricultural drought. However, the coarse spatial resolution of passive microwave products, including the 9 km Soil Moisture Active Passive (SMAP) retrievals, limits their effectiveness at regional and local scales. To address this limitation, three machine learning-based downscaling frameworks were compared to improve SMAP SM resolution from 9 km to 1 km over Békés County, Hungary. The study period covered the growing seasons (April to October) from 2020 to 2023. A set of multi-temporal MODIS-derived variables, including vegetation indices (NDVI, EVI), daytime and night-time land surface temperature, and evapotranspiration, along with land cover classification and topographic elevation, were combined as auxiliary predictor variables. Three machine learning algorithms, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Gradient Boosting Machine (GBM), were trained and evaluated. The results showed that (1) the RF model had the highest accuracy during the testing (R2 = 0.71, RMSE = 0.0295 m3/m3) phase and validation against four in situ monitoring stations with confirmed reliable SM estimation at the local scale; (2) daytime LST was the most important predictor in all models, underscoring the strong thermal–moisture coupling that governs surface SM dynamics; and (3) the validated RF model produced 1 km Standardized Soil Moisture Index (SSI) maps that effectively captured inter-annual drought variability, identifying the severe drought of July 2022. Overall, this study presents a downscaling approach for generating high-resolution SM data suitable for Central European agricultural environments. The resulting 1 km SM and SSI products provide valuable tools for decision-makers to enhance planning during drought periods and reduce agricultural losses through improved irrigation scheduling.
Forest microclimate strongly influences medicinal plant growth and habitat suitability; however, year-round characterization of forest microclimate for medicinal plant cultivation remains limited. This study presents a year-long, multi-station microclimate dataset and evaluates its potential for preliminary suitability assessment. Three monitoring stations (Pt1–Pt3) were deployed within the Plant Genetic Conservation Project under the Royal Initiative (RSPG), Thailand. Air temperature and relative humidity were continuously monitored at 30-min intervals from January to December 2025 using an Internet of Things (IoT)-based system. We used descriptive statistics, coefficients of variation (CV), repeated-measures analyses, and adjusted pairwise comparisons to evaluate year-round, seasonal, and site-specific microclimatic variabilities. Significant spatial differences were observed among the monitoring stations (p < 0.05). Pt3 exhibited the lowest annual mean temperature (26.05 °C), the highest relative humidity (80.60%), and the lowest environmental variability (CV = 16.41%), whereas Pt2 showed the highest temperature (30.72 °C), the lowest humidity (60.83%), and greater variability. A relative microclimatic comparison showed that Pt3 was cooler, more humid, and more stable; Pt1 exhibited intermediate conditions; and Pt2 was warmer, drier, and more variable. These findings characterize relative temperature–humidity conditions among the monitored sites and provide baseline information relevant to future species-specific assessment of medicinal plant cultivation.
Automated monitoring of welfare-relevant behavior in laying hens (Gallus gallus domesticus) has advanced with developments in computer vision, deep learning, and precision livestock farming (PLF). This narrative review integrates the ethological basis of welfare indicators with the development and readiness of monitoring technologies. It distinguishes routinely expressed diagnostic behaviors, including preening, locomotion, dustbathing, feeding, drinking, and nesting, from high-priority welfare risks such as aggression, piling, feather pecking, and inactivity or prostration. The review traces the progression from manual ethograms to semi-automated tools and recent artificial intelligence (AI) applications. These include You Only Look Once (YOLO)-based detection, multi-object tracking with BoT-SORT (a robust association-based tracking algorithm), pose estimation, and multimodal sensor fusion. Application-specific Technology Readiness Levels (TRL 1–9) indicate that most behavior-analysis systems remain at TRL 4–6. Several technologies extend into TRL 6–8, whereas few established systems reach TRL 9. Persistent barriers include domain shift under production conditions, annotation costs, inconsistent validation protocols, and limited economic accessibility. By linking welfare relevance, evaluation level, and deployment evidence, this review identifies priorities for scalable and actionable monitoring in commercial laying-hen production.
Plant factories with artificial light (PFAL) enable precise environmental control for vertical indoor agricultural production systems. However, their multi-layer configuration often creates stagnant air zones with significant temperature and humidity gradients. While ventilation systems are essential for mitigating these issues, their effective design requires accurate and distributed climate monitoring. The deployment of distributed microclimate sensors in PFAL environments remains challenging when multiple identical I2C sensors are required, particularly in low-cost monitoring architectures. Because of their fixed I2C addresses, these devices cannot be connected directly to the same bus without conflicts, prompting the need for multiple controllers and thereby increasing system costs. This study evaluates the implementation of a low-cost, multiplexed IoT sensor network for PFAL microclimate monitoring based on an ESP32 microcontroller and an I2C multiplexer (TCA9548A). This network enables simultaneous operation of five SHT20 temperature and humidity sensors at different levels within a PFAL structure. The multiplexing architecture generated coherent multipoint measurements, demonstrating its practical suitability for microclimate monitoring in PFAL environments. However, the system’s performance was compromised by its reliance on the local Wi-Fi network, resulting in intermittent connectivity failures and significant data loss during internet outages. The presented findings and observed limitations underscore the need for complementary strategies, such as local data buffering or a dedicated private network, to ensure reliable long-term monitoring in smart agricultural environments.
Tomato is a major global crop, yet foliar diseases seriously affect yield and quality. Accurate and efficient disease identification techniques are of great significance for ensuring agricultural production safety. To address the limitations of existing methods in feature extraction capability, model complexity, and dependence on large-scale labeled data, as well as the difficulty of recognizing early-stage diseases whose visual symptoms are not yet fully developed, this paper proposes SAEFormer, a lightweight and robust disease recognition model. The model integrates a Multi-scale Selective Fusion Attention Block to enhance the ability to model multi-scale semantic information in lesion areas. In addition, by leveraging a self-supervised loss derived from dense relative localization as an auxiliary regularization term, the model’s generalization ability under limited training data is notably enhanced. To optimize normalization and improve inference efficiency, the RepBN normalization strategy is further adopted, significantly reducing computational cost while maintaining model performance. Experimental results on Dataset A show that SAEFormer achieves a Top-1 accuracy of 87.86% with 24.14 M parameters and 5.35 GFLOPs, demonstrating a favorable balance between recognition accuracy and model complexity. The training curves further indicate stable convergence during model optimization. Ablation experiments validate the complementary contributions of the proposed modules. Moreover, SAEFormer achieves competitive performance in cross-dataset evaluation on Dataset B, indicating its potential adaptability to different data distributions. Overall, SAEFormer provides an efficient approach to tomato leaf disease recognition and shows potential for deployment in precision agriculture applications.
Autonomous driving relying on visual navigation plays a vital role in promoting automation within the jujube industry. Conventional visual navigation strategies fail to satisfy the demands of straddle-type jujube harvesters owing to their unique row-straddling configuration and complex orchard environments. Accordingly, this paper proposes a novel binocular vision-based path detection algorithm for autonomous jujube harvesters. In the proposed method, target trunks detected from binocular images are used to generate a navigation path that better aligns with the operating trajectory of harvester. A Single Shot MultiBox Detector (SSD) deep learning model is employed to detect trunk bounding boxes. To mitigate interference induced by false detections from the deep learning model, a curve-fitting-based path calibration strategy is implemented. Experimental results demonstrate that the proposed algorithm achieves a detection speed of 14.38 fps with a false detection rate of 3.64%, satisfying the operational demands for autonomous driving of jujube harvesters. Furthermore, this algorithm can be extended to other orchard mobile robots that execute row-straddling operations similar to jujube harvesters.
Brown Rot (Monilinia spp.) and Leaf Curl (Taphrina deformans) are principal fungal diseases affecting peach (Prunus persica L. Batsch) production, lacking validated AI-based diagnostic tools in tropical highland orchards. This study presents a compact Convolutional Neural Network (3658 trainable parameters), applied identically to fruit and leaf classification, integrated with a background-removal preprocessing pipeline and evaluated through stratified 5-fold cross-validation on 800 in-situ images from four orchards in Cómbita and Choachí, Colombia. The proposed architecture achieved mean accuracies of 82.8% (fruit) and 95.3% (leaf), with AUC values of 0.87 and 0.98, and a trained model footprint of approximately 100 KB, supporting storage- and bandwidth-efficient deployment. Benchmarked against ImageNet-pretrained MobileNetV3-Small and MobileNetV2 under an identical protocol, the proposed architecture matched or exceeded MobileNetV3-Small on leaf classification despite a 257-fold smaller parameter count, and achieved comparable or lower inference latency than both larger backbones. To our knowledge, this is the first validated system for simultaneous detection of both pathogens in Prunus persica under real field conditions, combining a compact, deployment-ready architecture with an ablation-verified preprocessing pipeline. The proposed model was deployed in the DurAPP web platform, giving peach growers in tropical highland regions a practical, low-footprint diagnostic tool suited to smallholder farming conditions.
Reliable maturity detection of Sichuan pepper is critical for intelligent harvesting, precision agriculture, and automated quality assessment. However, existing vision-based methods often suffer from performance degradation in natural orchard conditions due to illumination variation, fruit occlusion, background interference, and substantial appearance diversity. In addition, the lack of dedicated Sichuan pepper maturity datasets limits the development and reliable evaluation of deep learning-based detection models. To address these challenges, this study proposes C-MorphYOLO (Circular-Morphological YO-LO), a morphology-aware detection framework for Sichuan pepper maturity recognition under field conditions. By incorporating circular morphological information into feature representation learning, the proposed framework enhances the perception of fruit structural characteristics and improves fine-grained maturity discrimination. Specifically, a Circular Depthwise Convolution Module (CDCM) is developed to strengthen boundary and shape-aware feature extraction, while a Feature Enhancement-based Upsampling Module (FEUM) and a Multi-Scale Adaptive Spatial Attention Gate (MASAG) are introduced to improve small-object representation and detection robustness. Furthermore, a natural-scene Sichuan pepper maturity dataset containing 2763 images and 2962 annotated instances was established for model training and comprehensive evaluation. Experimental results on the held-out validation set demonstrate that C-MorphYOLO achieves a Precision of 96.1% and an mAP50:95 of 89.1%, representing absolute improvements of 2.2% and 2.7%, respectively, compared with the YOLOv12s baseline. These results demonstrate that the proposed framework effectively improves maturity recognition and localization performance, providing a practical vision-based solution for intelligent Sichuan pepper harvesting and automated agricultural management.
Radio frequency identification (RFID) tags are now included in the plastic wrap used to protect seed cotton formed into cylindrical or “round” modules on modern cotton harvesters. In this paper, the development of a new work tool system for handling round modules with articulated wheel loaders or telehandlers is described. The work tool system reads the module-specific identification number from the RFID tags in the wrap and associates the module’s weight, seed cotton moisture content, GPS location, cotton ownership, and load information with the module serial number. Finite element analysis of critical components indicated that the system was capable of processing modules weighing 3178 kg (7000 lb.) Module weight was determined on the loader using measurements of the hydraulic pressure in the lift arm circuit. Seed cotton moisture content was measured using a custom-designed resistance-based probe. To help reduce the potential for lint bale contamination from module wrap plastic, the work tool system was designed to rotate modules so that the wrap can be cut within the manufacturer-recommended cut zone before the wrap is removed at the gin. The total cost for the system configured for fully automated data collection and module rotation control was $28,909.
The pursuit of greater efficiency in agricultural operations, particularly in input application, has led producers to adopt strategies aimed at minimizing production costs. Aerial fertilizer application has emerged as a viable alternative due to its high operational efficiency and its ability to operate in conditions where ground-based application is not feasible. However, few studies have evaluated its efficiency, resulting in limited technical guidelines for calibration and adjustment. In this context, the present study aimed to evaluate the quality of broadcast application of solid potassium fertilizer via aircraft. The experiment was conducted using a split-plot design in a factorial arrangement with three flight altitudes (10, 15 and 20 m) and four application rates (50, 75, 100 and 125 kg ha−1), each with three replications. Longitudinal and transverse distributions were evaluated, as well as the correlation between wind speed and applied doses. The longitudinal analysis showed that flight altitude influenced both the uniformity of distribution and the effective dose applied, with a significant interaction between factors. In the transverse analysis, the overall transverse deposition pattern was predominantly governed by the 4.0–2.0 mm particle-size fraction, which represented approximately 90% of the recovered fertilizer mass across all flight heights. Although the granulometric composition remained consistent, the spatial distribution of individual particle-size classes varied with flight height, particularly for the finer fractions. Overall performance was achieved and the best results were observed at a 25 m swath width with flight heights between altitudes of 10 and 15 m. While the 10 m flight height resulted in lower eccentricity and greater fertilizer deposition, the 15 m flight height provided lower coefficients of variation after overlap simulation across most application rates, indicating more uniform transverse distribution.
Greenhouses in subtropical and tropical regions must be designed as agricultural engineering systems adapted to local climates, rather than simply replicating the “insulation” models of temperate areas. Under extreme climatic conditions such as persistent high temperatures, intense solar radiation, high humidity, heavy rainfall, and frequent extreme winds, greenhouses transform from enclosed insulation layers into selective climate filters, mitigating crop stress while maintaining close contact with the outdoor environment. This paper summarizes how these climate drivers are reshaping the use, structure, and control concepts of greenhouses, emphasizing that the performance of warm-zone greenhouses depends primarily on heat dissipation, humidity management, and biohazard control, rather than heating and insulation. In this review, we analyze the climatic boundary conditions that define warm-climate conservation cultivation, including long-term overheating risk, high UV radiation, vapor pressure deficit, and suppressed condensation tendency, as well as storm-induced uplift and dynamic loads. These constraints necessitate unique structural forms: tall, lightweight, well-ventilated building types with large roof and side openings, roof geometries that facilitate rainwater runoff, sophisticated drainage systems, and corrosion-resistant materials suitable for humid and coastal environments. Because insect netting significantly reduces ventilation, pest control and temperature regulation become co-design issues, requiring oversized vents, optimized airflow paths, and hybrid roof–mesh structures. Ventilation is considered the primary climate-control mechanism, supplemented by passive cooling measures such as shading and radiation/optical management (e.g., diffuse films and near-infrared-selective films). Active evaporative cooling is considered a conditional measure due to humidity limitations and disease risks. This paper also integrates the impacts on specific crops (fruits and vegetables, leafy greens, and orchids). It highlights emerging trends: typhoon-resistant and adaptive geometries, computational fluid dynamics (CFD)-based design, and sensor-rich IoT/digital twin control frameworks. These principles collectively establish a coherent design framework for achieving resilient, resource-efficient greenhouse production in warm climates.
Acerola is a tropical fruit rich in vitamin C and other bioactive compounds, but its high perishability limits storage and distribution. This study evaluated the hygroscopic behavior and storage stability of acerola pulp powder produced by foam-mat drying. The pulp was foamed with Emustab and dried at 60 °C. Fresh pulp and powder were analyzed for water activity, total titratable acidity, reducing sugars, vitamin C, carotenoids, total phenolic compounds, and antioxidant activity. Powder adsorption isotherms, hygroscopicity, caking, and solubility were evaluated at 10 and 25 °C under relative humidities of 20, 50, 70, and 90%. Drying increased the reducing sugar concentration from 5.24% to 45.49%. The Oswin model best described the Type II sigmoidal isotherms, with R2 values of 0.9921 and 0.9974 and mean relative errors of 4.61% and 3.27% at 10 and 25 °C, respectively. Equilibrium moisture content was higher at 25 °C, an atypical behavior associated with the high concentration of low-molecular-weight sugars. Hygroscopicity ranged from 8.77 ± 0.72% to 53.39 ± 0.51%. Although the powder has potential as a functional ingredient, its high hygroscopicity and limited physical stability require moisture-barrier packaging and formulation strategies to improve storage and industrial applicability.
Accurate body weight monitoring is essential for efficient sheep production, yet conventional weighing methods are labor-intensive and require frequent animal handling. Computer vision provides a promising non-invasive alternative; however, many published studies rely on validation strategies that may overestimate predictive performance because repeated observations from the same animals are simultaneously included in training and testing datasets. This study developed and evaluated an integrated analytical framework for image-based body weight estimation in ewes that combines standardized image processing, robust frame-level quality control, longitudinal statistical modeling, and strict leakage-controlled validation. A longitudinal dataset comprising 20 Dorper ewes monitored over six sampling periods was acquired using top-view depth imaging synchronized with body weight measurements under semi-arid conditions. The analytical framework integrated three complementary modules: longitudinal mixed-effects modeling, early prediction of final body weight, and contemporaneous body weight estimation. Predictive analyses were evaluated using nested leave-one-animal-out cross-validation, in which all preprocessing, correlation filtering, feature selection, model optimization, and model selection were performed exclusively within the training animals of each outer fold. The longitudinal mixed-effects model accurately characterized individual growth trajectories (conditional R2 = 0.98). Initial body weight remained the strongest predictor of final body weight (R2 = 0.770), whereas image-derived morphometric descriptors alone showed limited predictive performance under strict animal-level validation (R2 = −0.899 to −0.031). Combining baseline body weight with selected morphometric descriptors produced modest but biologically informative improvements, achieving a maximum R2 of 0.839. Contemporaneous body weight estimation achieved moderate predictive performance (maximum R2 = 0.332) and revealed temporal changes in the importance of morphometric descriptors throughout growth. Overall, the proposed framework provides a reproducible methodology for evaluating image-derived phenotypes under rigorous animal-level validation, contributing to the development of more robust, interpretable, and biologically grounded computer vision systems for Precision Livestock Farming.
Unmanned aerial vehicle (UAV)-based multispectral remote sensing has emerged as a promising tool for monitoring crop physiological status and supporting precision disease management. However, the capacity of multispectral vegetation indices to detect foliar diseases under commercial field conditions remains insufficiently understood, particularly in perennial crops grown in humid tropical environments. This study evaluated the potential of UAV-derived multispectral vegetation indices to assess the physiological response of avocado (Persea americana Mill.) canopies affected by anthracnose caused by Colletotrichum fructicola in Amazonas, Peru. Disease incidence and severity were assessed through field evaluations, while multispectral imagery was acquired using a UAV equipped with a MicaSense RedEdge-MX Dual sensor. Vegetation indices including the Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Soil-Adjusted Vegetation Index (SAVI), Normalized Difference Red Edge Index (NDRE), Red Edge Chlorophyll Index (CIred-edge), Plant Senescence Reflectance Index (PSRI), and Visible Atmospherically Resistant Index (VARI) were calculated, and their relationships with anthracnose incidence were analyzed using Spearman’s rank correlation. Field observations confirmed a high incidence of foliar anthracnose with a heterogeneous spatial distribution across the orchard. Multispectral imagery successfully characterized spatial variability in the canopy physiological condition, revealing differences in vegetation vigor, chlorophyll-related reflectance, and senescence among experimental blocks. Nevertheless, none of the evaluated vegetation indices showed statistically significant correlations with anthracnose incidence (p > 0.05), indicating that spectral variability primarily reflected the general canopy physiological status rather than a disease-specific spectral response. These findings demonstrate that UAV-derived multispectral vegetation indices are valuable for monitoring spatial variability in the avocado canopy condition but have limited capability for independently detecting anthracnose under humid tropical field conditions. Future studies integrating multi-temporal UAV acquisitions, hyperspectral and thermal imagery, LiDAR, environmental variables, and machine-learning approaches are expected to improve the early detection and spatial prediction of anthracnose in avocado production systems.
Water scarcity has a significant impact on global agriculture, particularly in semi-arid regions, hindering economic development. The use of recycled urban wastewater in agriculture is a sustainable practice; however, it is essential to assess its impact on soil carbon stocks and microbial activity. This study hypothesized that the use of wastewater in soil cultivated with forage cactus and amended with 8 or 12 tons of sorghum straw as soil cover could increase carbon stocks, microbial biomass, and microbial activity compared to bare soil, even after only 8 months. The experiment was conducted in a tropical semi-arid region of Brazil, based on a factorial design with different cactus intercropping systems and soil cover treatments under wastewater irrigation. Overall, soil carbon stocks did not increase significantly compared to the control, although they increased by approximately 21% over the study period. However, soil cover increased C-CO2 emissions by 70% after 4 and 8 months. Microbial biomass carbon increased by 65% compared to the baseline (time 0), particularly in treatments with soil cover. Soil cover and consortium under wastewater irrigation improved microbial activity and biomass, even over a short experimental period, indicating a sustainable soil management strategy to enhance soil organic matter quality and microbial properties.
Weeds create a significant challenge in agricultural production by competing with crops for essential resources such as nutrients, sunlight, and water. This competition leads to reduced crop yields and quality, resulting in substantial economic losses. Consequently, there is a critical need for effective weed control strategies to mitigate the impact of unwanted plant growth and ensure sustainable agricultural practices. In precision agriculture, enabling autonomous navigation between crop rows during tasks such as weeding and harvesting presents a significant research challenge, particularly when leveraging cost-effective technological solutions. Effective robot navigation requires adaptive traffic management strategies and robust object recognition capabilities to distinguish between cultivated and uncultivated areas. In fact, the integration of computer vision techniques into these systems is essential for optimizing trafficability in cropping fields and enhancing robots’ dynamics for better working efficiency in agricultural environments. This review addresses the challenge of enhancing inter-row navigation in field crops and delivering reliable guidance for autonomous agricultural robots. A PRISMA-based systematic review methodology was adopted to identify, screen, and analyze 38 relevant studies selected from the Scopus and Web of Science databases. The selected studies are classified according to their target platform (Unmanned Ground Vehicles and Unmanned Aerial Vehicles) and grouped into three methodological categories: conventional computer vision, deep learning architectures, and hybrid approaches. The findings provide practical guidance for selecting appropriate vision-based crop row detection technologies according to the application requirements and highlight key research directions toward more robust, cost-effective, and adaptable autonomous navigation systems. This article presents an outline of artificial-intelligence-based row detection methods used in agricultural fields and a classification of related semantic segmentation approaches. Unlike previous surveys, it provides an overview of the technological progress in agricultural robots and navigation based on systems vision for crop row detection, with a focus on comparisons balancing technical performance with economic and practical constraints.
Achieving uniform agrochemical deposition in coffee is challenging because canopy structure, terrain, and wind conditions influence spray movement and retention. This study compared an unmanned aerial spray system (UASS), backpack sprayer, and tractor-mounted sprayers across three commercial coffee farms on Hawai‘i Island. Spray coverage, droplet density, droplet size metrics, and operational efficiency were evaluated using water-sensitive cards positioned throughout the canopy and analyzed using mixed-effects models. UASS produced significantly lower spray coverage and droplet density than the ground-based application systems, whereas backpack and tractor sprayers did not differ. Deposition patterns varied with canopy position, with application method effects depending on canopy height, depth, and aspect. Volume median diameter decreased in the upper canopy and with increasing wind speed, while relative span varied modestly among methods and was greater within the canopy interior. Canopy position and wind strongly shaped agrochemical deposition across spray platforms. Although UASS required less field labor and improved accessibility in terrain-limited systems, these operational advantages were accompanied by reduced deposition relative to ground-based sprayers. These findings demonstrate that canopy position and environmental conditions strongly influence agrochemical deposition and support UASS as a complementary application platform rather than a direct replacement for conventional sprayers under the conditions evaluated.
Modern viticulture is undergoing a paradigm shift, transitioning from a plant-centric view to the holobiont concept, which considers the grapevine (Vitis vinifera L.) and its associated microbiota as a single co-evolved functional unit. This review synthesizes current knowledge on the grapevine holobiont, identifying the drivers of microbial assembly (genotype, environment, management) and their functional implications for plant health and wine quality. We discuss the pillars of plant health, defining disease as a state of dysbiosis rather than merely the presence of a pathogen. Special attention is given to the spatial compartmentalization of the microbiome, from the gating mechanisms of the rhizosphere to the transient diversity of the anthosphere. Furthermore, we highlight the methodological evolution from culture-dependent techniques to Next-Generation Sequencing (NGS) and the emerging role of MALDI-TOF MS as a rapid, cost-effective tool for real-time monitoring. Finally, we propose a roadmap for microbiome-assisted viticulture that utilizes synthetic microbial communities (SynComs) and hologenomic breeding to enhance resilience against climate change.
Accurate estimation of grassland biomass is fundamental for designing sustainable grazing strategies and optimizing pasture management, yet conventional field methods remain labor-intensive, destructive, and difficult to scale. In this study, we exploit recent advances in computer vision to estimate multiple components of grassland biomass from overhead RGB imagery. The analysis is based on the Image2Biomass dataset, comprising 1162 annotated images of grasslands across Australia, each paired with laboratory-validated biomass measurements. A structured preprocessing pipeline was implemented, including exploratory data analysis, outlier mitigation, and logarithmic transformation of target variables, in accordance with the dataset evaluation protocol. Building on this foundation, we propose an encoder–decoder regression framework that integrates self-supervised visual representation learning with ensemble-based prediction. The encoder employs a DINOv2 Giant model as a feature extractor to capture detailed spatial and structural characteristics of the sward, while the decoder uses a stacking ensemble combining LightGBM, XGBoost, and Ridge Regression. Across 15 repetitions of shuffled four-fold cross-validation, the cross-fitted stacking ensemble achieved a weighted coefficient of determination of Rw2=0.7757±0.0171, a weighted mean absolute error of 8.3867±0.2736 g, and a weighted root mean squared error of 13.3994±0.4995 g. The ensemble significantly outperformed LightGBM, XGBoost, and Ridge Regression on the primary weighted R2 metric in paired comparisons (Holm-adjusted p<0.001 for all three comparisons). These results highlight the potential of computer vision methods as scalable, non-destructive tools for operational monitoring of grassland biomass, supporting more informed agronomic decision-making in pasture-based systems.
Cereal crops, including wheat and barley, are essential for global food security, but their productivity is strongly affected by nitrogen availability and water limitation. This study investigated the phenotypic responses of two commercially significant spring wheat cultivars, Videodur (DU) and Sensas (SW), and two spring barley cultivars, Tiroler Imperial (SG1) and Amidala (SG2), exposed to two nitrogen regimes, low nitrogen at 25 kg N/ha (N25) and high nitrogen at 130 kg N/ha (N130), under drought and well-watered conditions. Plants were monitored from the late vegetative stage through maturity under controlled multivariable climatic conditions similar to field settings. A high-throughput phenotyping workflow was applied, combining precision watering, RGB imaging, infrared thermography, and VNIR–SWIR hyperspectral imaging to quantify plant growth, projected digital biomass, plant temperature, water use efficiency, and spectral vegetation indices associated with pigment dynamics, water status, maturation, and senescence. The results revealed cultivar-specific responses to combined nitrogen and drought stress. Under drought conditions, the high nitrogen treatment (N130) increased plant temperature (Tplant) for barley (cv. SG1) and wheat (cv. SW) compared to N25, thereby accelerating early maturation. However, the decline in chlorophyll was not uniformly faster across all cultivars tested. The DU cultivar exhibited superior chlorophyll absorption and reflectance, indicating better drought adaptation compared to other tested species. The high nitrogen treatment (N130) reduced water use efficiency (WUE) in the SW and SG2 cultivars compared to N25, implying that these cultivars used more water. Enhanced nitrogen did not consistently improve water use efficiency but did accelerate the growth cycle. SG2 was particularly sensitive to drought, showing declines in vegetation indices, except for the Water Content Index, highlighting the need for precise water and nitrogen management. Overall, the integration of hyperspectral, thermal, RGB, and water use measurements enabled the identification of trait signatures linked to drought adaptation, nitrogen response, maturation, and senescence. These findings provide practical insights for optimizing nitrogen and irrigation management and for supporting breeding strategies aimed at improving cereal crop resilience under climate-change-associated stress conditions.
