Highlights Developed a 3-DoF cutting mechanism as a robotic bud thinning end-effector. Integrated and validated a cylindrical manipulator with the end-effector for apple bud thinning. Simulations recommended a manipulability region (500–1000 mm) to guide effective selection of target buds. Field trials confirmed the precise end-effector positioning for bud removal across canopy spatial configurations. ABSTRACT. Apple bud thinning is an essential early-season operation for achieving optimal crop load and high fruit quality; however, it is a labor-intensive task dependent on skilled labor. A robotic system featuring a 3 DoF cutting end-effector integrated with a cylindrical manipulator was developed and validated to automate apple bud thinning in orchard environments. Model-based simulations evaluated the manipulator’s reachable workspace and dexterity, identifying recommended manipulability regions between 500 and 1000 mm from the system to the canopy that guided effective target selection. Field trials with 30 bud removal operations confirmed the system’s ability to reach and align with apple buds across varying canopy structures, enabling precise positioning for effective thinning. However, the large blade opening occasionally caused damage to adjacent buds in densely clustered areas, highlighting a design limitation. Observations indicated minimal contribution of yaw control to successful bud alignment, suggesting a simplified manipulator configuration could suffice. The robotic system demonstrated feasibility for automating apple bud thinning while providing insights into necessary design refinements, including enhanced end-effector precision and integration of real-time vision for improved cutter alignment. These advancements have the potential to enable selective and reliable thinning, thus, reduce dependence on manual labor in commercial orchard operations. Keywords: Apple bud thinning, Crop load management, Machine vision, Multi-DoF end-effector, Orchard automation.
Highlights Agricultural recordkeeping was made easy by integrating artificial intelligence. Generative AI enhances data collection with real-time validation. Metadata recorded by dynamically asked questions to the user. User experience indicates the effectiveness of recordkeeping using MetaAg 2.0. ABSTRACT. Generative Artificial Intelligence (Gen-AI) enables adaptive and context-aware human–computer interaction for structured data collection. To address limitations of manual farm recordkeeping, MetaAg, an Android-based application, was developed to automatically capture spatial and temporal information of farm operations using a rule-based chatbot for metadata entry. In this work, the system was extended through a retrieval-augmented generative AI framework to improve the accuracy and completeness of metadata recording. The framework leverages Gen-AI through API-based interaction to dynamically generate context-aware queries and validate user inputs using a backend knowledge base. MetaAg 2.0 was implemented as a prototype system and evaluated using multiple LLMs, including GPT-4o-mini (GPT), lfm-2.5-1.2b-thinking (Liquid), and Deepseek-R1-0528 (Deepseek). Results show that the proposed system supports efficient recordkeeping with reduced user interaction load and improved interaction performance. Comparative analysis indicates that GPT-4o-mini consistently outperforms Liquid and Deepseek in terms of lower interaction time, reduced wait burden, and smoother user interaction. User experience evaluation yielded a usability index (UIX) of 86.2 ± 6.1, indicating high user acceptance. Qualitative feedback further highlights strengths in ease of use and efficient data capture, while identifying limitations in interaction rigidity, interface layout, and error correction flexibility. These findings demonstrate that the proposed framework enables accurate and efficient metadata collection while maintaining strong usability, providing a scalable approach for intelligent agricultural data management. Keywords: Chatbot, Database, Digital records, Farm metadata, Gen-AI, Human-AI interaction, Human-computer interaction.
Highlights LLM agents transform dairy digital twins from prediction to reasoning. Agentic twins integrate physics, sensors, and management objectives. Edge–cloud agent twins enable robust long-term barn deployment. Explainable AI bridges engineering models and farm decisions. Keywords: Agentic AI, Computational fluid dynamics, Digital twins, Edge–cloud computing, Heat stress mitigation, Large language models, Precision livestock farming.
Highlights This study presents a seven-year operational review of a large-scale cable-suspended phenotyping system. Standardized operation and data protocols enable the delivery of high-resolution, high-frequency phenotypic datasets. The platform supports a wide range of phenotyping applications, from morphological to physiological trait analysis. The system plays a critical role in developing advanced sensing technology for phenotyping and precision agriculture. ABSTRACT. High-throughput plant phenotyping (HTPP) significantly improves plant phenotyping efficiency by integrating advanced sensing technologies, data processing, and modeling techniques. Over the last two decades, field-based HTPP systems have evolved from handheld sensors to sophisticated robotic platforms. Large-scale, ground-based phenotyping facilities have made significant contributions to advancing technology through their high sensor payloads, proximal measurement capabilities, and unmatched spatial and temporal resolution. This review paper presents a detailed and quantitative analysis of the operational performance of the cable-suspended NU-Spidercam phenotyping facility at the University of Nebraska–Lincoln from 2017 to 2023, focusing on daily operations, data management strategies, and system maintenance. Additionally, the paper systematically summarizes representative studies performed at the facility across morphological, spectral, and physiological phenotyping domains. Finally, the discussion highlights future directions, emphasizing the NU-Spidercam’s role in validating mobile phenotyping platforms, enabling precision agriculture research, supporting fundamental remote sensing studies, and facilitating the transfer of advanced sensing techniques to affordable, mobile platforms such as drones. Keywords: Artificial intelligence, Computer vision, Deep learning, Field plant phenotyping, Large-scale facility, Machine learning, Operational review, Physiological phenotyping.
Highlights DEM simulation was applied in the loss proportion modeling, overcoming limitations of the collection box tests. The multi-factor nonlinear compensation models for entrainment loss monitoring were established and validated. By integrating the models with an embedded system, high-precision online correction of loss monitoring was realized. ABSTRACT. As an essential component of total grain loss, accurate tracking of entrainment loss is crucial for reducing economic losses. To obtain precise information on actual entrainment loss, it is necessary not only to have reliable sensor hardware with low measurement error and good stability, but also to establish an accurate mathematical relationship between the sensor output and the true entrainment loss. Based on the entrainment loss monitoring sensor previously developed by our research group, this study further investigates the proportional relationship between the monitored and actual entrainment loss under various harvesting conditions. Specifically, for representative crops such as rice, wheat, and rapeseed, discrete element method (DEM) simulations were conducted using the Hertz-Mindlin contact model to build multi-component particle models including grains, short and long stems, and light impurities, along with their mechanical properties. Full spike tooth and striated bar-spike tooth threshing device models were developed to reflect actual harvesting configurations. Single-factor and multi-factor DEM simulations were carried out for feeding rate, threshing gap, and drum speed to analyze their effects on the monitoring proportion and its variation trends. Field sampling tests were conducted to verify and calibrate the simulation results. The results show that the threshing gap and drum speed are the main factors affecting the monitoring proportion, while the feeding rate has a negligible influence within a reasonable range. The developed multi-variable nonlinear fitting models exhibited high prediction accuracy with an adjusted R-squared greater than 0.95, and the relative deviations between model prediction and experimental data did not exceed 1.54%. By integrating the models with an embedded system, real-time compensation of the monitoring data can be achieved, significantly improving the accuracy, stability, and adaptability of entrainment loss monitoring, while reducing dependence on field sampling and saving considerable time and labor costs. Keywords: Combine harvester, Discrete element method, Entrainment loss, Proportion model, Real-time monitoring.
Highlights A case study validation of an LLM-driven dairy farm decision support framework. Integrated time-series IoT data and industry-standard management advice to ground system output in truth. Open-weight and locally viable LLMs demonstrate strong performance next to closed frontier models. ABSTRACT. Precision agriculture has revolutionized the dairy industry by providing farmers with high-resolution data streams and real-time production analytics. This study presents an integrated precision livestock farming framework that combines modern hardware, machine learning, and artificial intelligence to deliver deployable end-to-end dairy farming decision support. We evaluated an agentic large language model (LLM) framework that employs a hybrid retrieval-augmented generation (RAG) system grounded in real-time dairy farm data, industry-standard management guidelines, and peer-reviewed literature. Real-time cow-specific behavior data were gathered using machine-learning-driven behavior classification based on an ultra-wideband localization sensor network with wearable ear-tag devices. A diverse suite of open-weight LLMs (3–24B) running locally on realistic in-barn hardware was evaluated in this framework alongside proprietary frontier models from OpenAI and Google. The performance of each model was evaluated in each part of the framework: structured query language (SQL) generation and data-retrieval capability, literature and management guidance summarization, and end-to-end dataset contextualization and management decision support. Results indicate that both proprietary models and open-weight alternatives perform well in this framework. The trade-offs between performance, latency, and privacy provide valuable insights into the design of LLM-assisted dairy barn management frameworks. Specifically, models in the 4–20B parameter range were observed to match enterprise-scale counterparts in end-to-end contextualization performance, though often with increased latency due to offload to system memory. These findings validate a decentralized approach to an LLM-driven dairy farming assistance framework that integrates sophisticated wireless wearable sensor technology to deliver site-specific decision support without reliance on external connectivity. Keywords: Artificial intelligence, Domain adaptation, Edge computing, Large language models, Precision livestock farming, Small language model.
Highlights The GANs have been applied on the data modalities of RGB image, wearable sensor data, ultrasound image, thermal image, and hyperspectral image in animal farming. The GANs can increase accuracy by over 14% in animal management via augmenting development datasets. The GANs could generate unrealistic images, and more advanced model architectures and training strategies are needed. ABSTRACT. In modern artificial intelligence (AI) modeling, achieving optimal classification and/or regression performance is keenly pursued when applying these models to improve animal production systems. Large-scale, balanced, and high-quality datasets are tremendously beneficial for developing robust and generalizable AI models that learn data representations directly from the training process. Livestock and poultry are complex, individually different, time-varying, and dynamic living organisms, creating significant challenges due to biological variability and unstructured data collection environments. Limited laboratory capacities and restrictions on animal usage also hinder the creation of large datasets. Data augmentation can automatically generate synthetic data to expand datasets, boosting model performance while reducing manual efforts in data collection and annotations. Besides classical image processing-based data augmentation techniques, generative adversarial networks (GANs), consisting of two competing neural networks (a ‘generator’ and a ‘discriminator’), provide a novel approach that can learn to generate realistic data samples by having the networks iteratively challenge each other. There is an increasing awareness of utilizing GANs and their variants for data augmentation and synthesis in precision agriculture to improve AI model performance. This review presents an overview of the evolution of GAN architectures along with the applications in animal farming domains, covering a wide range of animal species, farming systems, and production phases. Applications as well as the challenges and opportunities of GANs are discussed for future research. Keywords: Animal management, Artificial intelligence, Data augmentation, Deep learning, Precision agriculture.
Highlights Machine learning was applied to recognize the operational status of a rice harvester from GPS data. The latitude, longitude, and time information were collected, and combined with geospatial information. SVM, RM, and kNN were used to recognize the operational status, achieving a maximum accuracy of 93.4%. ABSTRACT. This study examined the feasibility of recognizing the operational status of a rice harvester using GPS data combined with geospatial information and machine learning (ML) algorithms. A GPS receiver was installed on two harvesters in South Korea to collect spatial and temporal data. The collected GPS data were combined with land-use information using GIS tools to construct two datasets: the first containing four features and the second additional features from temporally neighboring records. Three supervised ML algorithms—support vector machine (SVM), random forest (RF), and k-nearest neighbor (kNN)—were applied to classify the operational status into four categories: operating, moving, in storage, and other status. Among the algorithms, RF showed the highest performance, achieving 93.4% accuracy, particularly when trained with the second dataset. The study demonstrated the potential of leveraging GPS and geospatial data with ML algorithms to produce efficient and accurate statistics on agricultural machinery usage. Keywords: GPS, Machine learning, Operational status, Rice harvester.
Highlights Regenerative agriculture needs to be defined within the context of a region, and the associated agricultural practices must be regionally optimized. The economic viability of regenerative practices must be a primary consideration for the success of production agricultural systems. Many opportunities exist for successfully implementing regenerative practices in cotton production systems across the semi-arid Texas High Plains. ABSTRACT. For the Texas High Plains (THP), regenerative agriculture represents a natural progression in conservation and system resiliency. Regenerative agriculture encompasses a broader theme of conservation that includes community-level longevity and economic viability, all while promoting concurrent environmental stewardship. The THP represents a vast region of agricultural production, representing a range of agricultural commodities across semi-arid environments with varying levels of precipitation. Including some of the most profitable cotton-producing counties in the United States, the THP produces approximately 30% of the country’s annual cotton (Gossypium hirsutum) crop and 30% of its fed cattle (Johnson et al., 2013). However, the future agricultural viability and subsequent economic longevity of the region are threatened by two concurrent anthropogenic disasters, global climate change and groundwater withdrawal. With the region’s future at risk, opportunities exist to implement regenerative agricultural practices to increase the sustainability of cotton agroecosystems as producers transition into deficit-irrigated or dryland production systems. Regenerative practices, such as no-tillage, cover cropping, crop rotations, and integrated crop-animal grazing systems, provide prospective solutions that increase carbon sequestration, increase soil water conservation, improve soil health, and reduce net emissions of greenhouse gases. This review encompasses a range of regenerative agricultural practices that may be beneficial to cotton production across the THP region, including the Southern High Plains, Northern High Plains, and Rolling Plains. Keywords: Cotton, Cover cropping, Crop rotation, Crop-livestock integration, Regenerative agriculture, Semi-arid, Texas high plains.
Highlights A hybrid CNN–WE model markedly improves the accuracy of barley disease classification. Adaptive weighting enhances decision stability by reducing the influence of weaker classifiers. The proposed ensemble achieves 98.72% accuracy, outperforming individual models. The framework supports accurate barley disease detection and contributes to the next generation of smart agriculture. ABSTRACT. Rhynchosporium commune and Pyrenophora teres are two of the most prevalent barley diseases, and their highly similar visual symptoms make accurate diagnosis and differentiation particularly challenging. Artificial intelligence (AI) and machine learning (ML) methods have been widely applied to plant disease detection; however, their performance often degrades under diverse field conditions and noisy image data. Moreover, reliance on a single model increases vulnerability to its specific limitations, especially when diseases exhibit overlapping visual characteristics. To address these challenges, this study proposes a hybrid model that integrates deep feature extraction using a Convolutional Neural Network (CNN) with a weighted ensemble (WE) of ML classifiers. The CNN-extracted features are fed into four classifiers: Support Vector Machine (SVM), k-Nearest Neighbors (k-NN), Random Forest (RF), and Decision Tree (DT), each leveraging distinct strengths in handling complex, nonlinear, and interpretable data. A dynamic weighting strategy is then applied, assigning higher weights to classifiers with superior validation accuracy, and the final prediction is obtained through weighted majority voting. Unlike conventional ensemble methods that treat all classifiers equally, this adaptive weighting enhances decision stability, reduces the influence of weaker models, and significantly improves overall accuracy. Experimental results demonstrate classification accuracies of 82.5% for CNN, 87.3% for k-NN, 83.7% for DT, 91.6% for SVM, and 93.5% for RF, while the proposed WE achieves 98.72% accuracy. Furthermore, the ensemble model attains class-wise precision of 98.5% for R. commune, 97.5% for P. teres, and 98.5% for healthy samples, marking a substantial improvement over individual classifiers. These findings highlight the effectiveness of combining automatic deep feature extraction with adaptive ensemble learning, offering a robust framework for accurate barley disease detection and contributing to the advancement of intelligent plant health monitoring and precision agriculture systems. Keywords: Barley disease classification, Convolutional neural networks, Deep feature extraction, Ensemble classifiers, Machine learning, Weighted ensemble.
Highlights AI can enhance agricultural safety through predictive models, LLMs, computer vision, and wearables. Machine learning is a tool that has the potential to predict hazards using historical data on injury incidents, weather, and worker behavior. Computer vision and remote sensing applications could be used to detect unsafe conditions for real-time risk mitigation. Wearable AI devices present opportunities to monitor worker health and prevent injuries in agricultural environments. ABSTRACT. Artificial intelligence (AI) has emerged as a transformative tool in various industries, including agriculture, where it has the potential to enhance safety and reduce injury risks. This review explores the application of AI techniques in occupational safety with a special focus on agricultural safety, focusing on predictive modeling, large language models (LLMs), computer vision, and wearable technologies. Due to the limited number of studies that address AI in agricultural safety, research from related fields, such as construction safety, was also considered for its potential applicability. The findings indicate that (1) predictive models that leverage machine learning (ML) algorithms can assess historical data and forecast hazards, enabling proactive safety measures. (2) LLMs can improve injury report analysis by extracting key terms and patterns to identify recurring risks. (3) Computer vision and remote sensing technologies enhance environmental monitoring by detecting real-time unsafe conditions. At the same time, (4) AI-powered wearable devices can track worker health indicators such as heart rate, potentially preventing injuries. A total of 85 studies were analyzed, providing insights into the diverse applications of AI in mitigating occupational hazards, specifically agricultural hazards. This review highlights the current advancements and future research opportunities for AI-driven safety interventions in high-risk occupational environments. Keywords: Agricultural safety, Artificial intelligence, Computer vision, Health, Injury prevention, Large language models, Machine learning, Occupational safety, Predictive modeling, Wearable technology.
Predicting crop nitrogen content (CNC) of maize in real-time is crucial for effective nitrogen (N) management and the overall fertilizer application plan. The objective of this research was to identify a suitable machine learning (ML) model for assessing the CNC of maize. This approach used vegetation indices (VIs) derived from unmanned aerial vehicles (UAVs) and satellite (Landsat-8 and Sentinel-2) imagery to obtain real-time spatio-temporal information on CNC at field and regional scales, thereby developing in-season N management tools. Various spectral vegetative indices (VIs) were extracted from UAV images. These indices are used as input parameters for distinct ML models such as the Random Forest algorithm (RF), Support Vector Machine (SVM), Long Short-Term Memory (LSTM), Extreme Gradient Boosting (XGBoost), and a linear regression model. Statistical parameters such as mean absolute error (MAE), coefficient of determination (R2), Root Mean Square Error (RMSE), Nash-Sutcliffe efficiency (NSE), percent bias (PBIAS), Kling-Gupta efficiency (KGE), and mean absolute percentage error (MAPE) were used in a randomized 10-fold cross-validation technique to evaluate the performance of the models. All these models were developed to predict CNC at the field scale, and, further, the applicability of the best model was tested at the regional scale across different growth stages. The findings of the study revealed that the RF algorithm exhibited superior performance, with RMSE, MAE, R2, NSE, PBIAS (%), MAPE (%), and KGE of 0.58, 0.41, 0.64, 0.63, 1.37, 17.24, and 0.67 for the testing of CNC prediction. Furthermore, it was shown that the performance of the model was more significantly influenced by the inclusion of VIs compared to individual spectral bands. This study observed that the RF model is effective in predicting CNC with the VIs combination of ENDVI, NDVI, and SAVI for Landsat-8 and Sentinel-2 images. The comparison of actual and predicted CNC using satellite imagery showed a similar trend for the different months, except during the initial growth stage. While these results offer valuable insights, our study indicated that the spatial resolution of satellite imagery can significantly influence the accuracy of CNC prediction. During the early growth stages of maize, limited canopy coverage and increased soil background exposure further reduce model reliability. Overall, this indicates the potential to assist farmers by precise prediction of spatio-temporal variation in CNC during different growth stages utilizing ML models.
. Peanut nodule phenotyping provides an important basis for assessing symbiotic N fixation potential and guiding cultivation management. However, root images often exhibit high texture similarity between nodules and roots, with smallscale targets co-existing with dense occlusion clusters, which allows segmentation errors to accumulate along the counting pipeline and manifest as systematic bias. To address key bottlenecks, insufficient separability in homogeneous backgrounds, missed detections of small-scale nodules, and under-segmentation in dense scenes, this study proposed a Root Nodule Mor(DSFE) was introduced to enhance discriminative representations between nodules and roots under texture-homogeneous conditions; Multi-scale Feature Aggregation Gating (MFAG) was incorporated to strengthen small-target responses and aggregate multi-scale features; and Geometry-Aware Distance-Field Regression (GADR) was employed to provide more stable splitting cues in dense occlusion regions. Experiments demonstrated that NMP-Mamba achieved consistent advantages in both segmentation and counting (Dice 95.35%, Nodule IoU 91.11%, MAE 5.33) while maintaining 51.6 FPS; ablation analyses further verified the complementary contributions of the components to counting-regression agreement and error-variance control. In addition, an end-cloud collaborative nodule diagnosis APP was built upon the model, with the predicted masks and phenotypic indicators packaged as structured inputs to an agriculture-oriented Expert Agent, enabling an end-to-end workflow from image segmentation and phenotypic quantification to diagnostic text and management recommendations and providing intuitive reference and assistance for nodule counting and phenotyping.
Pen surface amendments are a tool that beef producers can use to reduce emissions and increase nutrient retention in manure. Previous studies demonstrated that aluminum sulfate (alum) could effectively lower ammonia (NH3) emissions but caused increased sulfide emissions. An alternative to alum is aluminum chloride (AlCl3). The objective of this study was to determine if the addition of AlCl3 to beef pen surface material (PSM) could lower NH3 losses while also lowering CO2, N2O and CH4 emissions. Aluminum chloride was added to pans containing 3 kg of PSM and 3 kg of water at a rate of 0%, 2.5%, 5%, and 10% (g g(-1) as-is) of the PSM mass, and emissions were measured for 21 days. The addition of AlCl3 to PSM lowered (p < 0.01) the pH to 5.8-7.2 compared to untreated PSM (8.2). Nitrogen and sulfur content of the PSM was not affected by AlCl3 treatment (p >= 0.14). Cumulative NH3 emissions decreased (p < 0.01) as AlCl3 inclusion increased and were 87.3, 60.5, 33.0, and 20.2 kg ha-1 for the 0%, 2.5%, 5%, and 10% AlCl3 treated PSM, respectively. Cumulative CO2, CH4, and N2O emissions also decreased linearly as the inclusion rate of AlCl3 increased. Aluminum chloride appears to be a promising option for beef feedlot producers to reduce emissions from the pen surface.
A comparison is run between deposition data collected downwind of UASS mosquito adulticide spraying and simulated values generated using AGDISPpro. AGDISPpro is a proprietary model. Therefore, to gain insight into UASS modeling, a detailed discussion of current numerical and simulation techniques described in the literature is given. UASS adulticide spraying offers a difficult case forAGDISPpro simulation, as very fine droplets are sprayed from relatively small UAVs at a greater height than in other UASS application scenarios. The AGDISPpro Lagrangian model often simulates within a factor of two (considered successful) but tends towards overestimation and tends to bring material down closer to the flightline than shown by the data. Data and model limitations are discussed.
Flow-paced composite sampling (FPCS) is widely used in stormwater quality monitoring for its practicality, cost-efficiency, and ability to reflect event concentrations and loads. Existing guidelines help configure FPCS protocols that minimize measurement uncertainty, and prior studies have evaluated its accuracy. However, these evaluations have been typically based on discrete sampling or modeled pollutographs, which lack the temporal resolution of hydraulic measurements and can fail to capture actual pollutant dynamics during storm events. This loss of resolution is largely driven by limitations of the automated sampling equipment's capacity. In this study, high-frequency (4-minute) water quality data from UV-Vis spectrophotometers were used to simulate thousands of FPCS scenarios and quantify uncertainty introduced by different sampling configurations. The simulations and literature review identified three primary sources of uncertainty: (1) percentage of event flow volume captured, (2) number of samples composited per event, and (3) timing of the first sample. Results show that to achieve +/- 10% accuracy with 90% confidence, of 100 samples composited in a jar, a minimum of 9 samples for nitrate-N, 13 for DOC, and 20 for TSS would be required. For a stricter +/- 5% accuracy target, 19 samples for nitrate-N, 22 for DOC, and similar to 29 for TSS would be needed. Sampling less than 85%-90% of the event flow volume resulted in systematic under- or overestimation depending on pollutant dynamics. Inclusion of a first flush sample taken prior to completion of the first flow interval, which therefore is not a flow-paced sample, introduced very large errors (>100%) and should be avoided especially for multi-peak events. This study provides an upgraded framework for configuring autosamplers based on expected runoff and capacity constraints, and introduces a post-event diagnostic tool to assess data quality, improving the reliability of FPCS-based stormwater pollutant monitoring.
Corn stover is a widely available agricultural residue with significant potential as a feedstock for cellulosic biofuel and bioproduct production. However, challenges related to material heterogeneity and high pretreatment costs limit its widespread use. Anatomical fractions of corn stover-such as cobs, husks, stalks, and leaves-exhibit distinct chemical and physical properties that influence their suitability for conversion processes. Tailoring pretreatment strategies to specific fractions has previously been shown to reduce severity and increase conversion efficiency. This study evaluated systems for fractionating single-pass harvested corn stover into distinct anatomical groups, enabling targeted bioconversion pathways. Two final composite stover groups were targeted: cobs/leaves/fines and husks/stalks. Two primary processing approaches were investigated: a modified combine harvester separation and cleaning system and a rotary trommel screen followed by secondary air classification, size reduction mechanisms, and mechanical screening. The combine harvester system achieved grain separation with 98.7% purity and less than 1.5% stover contamination, but stover anatomical fractionation remained partial. After trommel screening, air classification and additional screening were required to achieve effective physical separation. Drying and crushing the large cob segments to alter their size and shape was necessary to facilitate separation from large stalk segments when screened. The final fractions, created by combining individual components into targeted composite streams of cobs/leaves/fines and husks/stalks, achieved purities of 90.2% and 72.7%, respectively. However, only 53.9% of the total cob/leaf/fines material was recovered in its corresponding final composite. While mechanical sieving and air classification proved effective, the process involved numerous operational steps whose costs must be carefully weighed against the potential pretreatment savings.
. This study investigated the pore structure, thermal conductivity, and thermal effusivity of carbon materials derived from six biomass types: eastern red cedar wood (CW), microalgal biomass (MB), digested sludge from municipal wastewater treatment facilities (DS), hazelnut (HN) and pecan shells (PS), and an equal weight mixture of CW-MB-DS (MIX). Pyrolysis process variables included temperature (500-700 degrees C), hold time (1-4 h), and catalyst to biomass ratio (0- 2). The study examined the complex effects of biomass chemical composition and production parameters on total pore volume, surface area, pore size distribution, and thermal properties of the resulting carbon. Among the samples, carbon derived from MB exhibited the lowest total pore volume (0.489 cm3/g) and surface area (1017.744 m2/g). It was predominantly microporous, with micropores accounting for approximately 83% of the total pore volume. The very low thermal conductivity (0.064 W/mK) and effusivity (197.9 Ws1/2/Km2) observed in the samples suggest their strong potential for thermal insulation applications. This is the first study evaluating correlations between pore size distribution and thermal properties across multiple biomass types. The findings of this study provide valuable insights for selecting suitable biomass sources and optimizing production methods to design carbon materials with tailored pore structures and thermal characteristics for specific applications.
Denitrifying woodchip bioreactors (DWBs) have been studied as an edge-of-field method of intercepting and removing nitrogen constituents in agricultural runoff; however, most research has been conducted in the U.S. Midwest, which has a significantly different climate from the Mediterranean climate found in the Willamette Valley of Oregon. A DWB unit was installed at the Oregon State University Dairy Center in Corvallis, Oregon, to evaluate performance under climate conditions found during the winter rain season (daily temperatures of 11.7 degrees C +/- 5.6 (mean +/- SD), monthly precipitation inputs of 154 +/- 30.8 mm/month, cumulative precipitation of 772 mm over 6 months). Frequent sample collection over three operating seasons demonstrated that good reductions in nitrate-N were achievable, although with rates at the low end of the range reported in literature (2.7 g N m(-3) d(-1) +/- SD 2.3). A moderate degree of "active management" to adjust for changing weather conditions proved to be a critical factor in improving nitrogen (N) removal. Internal measurements show that nitrogen attenuation is generally achieved at an estimated hydraulic retention time of 16 hours (at water temperatures of 9.5 degrees C +/- SD 1.8, mean inlet NO3-N concentrations of 4.0 to 4.7 mg N/L, and a total range of 0.0 - 24.0 mg N/L), at which point production of hydrogen sulfide routinely occurs. A discussion of factors and features to consider for DWB operation in a high winter rainfall climate regime is provided.
