Advancing Open Science

Individual cattle identification underpins traceability, genetic management and disease surveillance in precision livestock farming. Ear tags and RFID remain the dominant practice but are subject to loss, tampering and welfare concerns, which motivates contactless biometric alternatives. Most deep learning systems used for this task optimise recognition accuracy under a closed-set assumption, provide little quantitative evidence for their explanations, and do not separate threshold selection from final evaluation. We present an offline, explainable cattle identification framework that integrates a ConvNeXt-Tiny backbone trained with ArcFace, cosine-similarity open-set decision making, Grad-CAM attribution and a locally hosted vision–language model. We evaluate it under a protocol that separates the training, enrolment, threshold selection and test roles, so that the conditions under which each reported number holds are explicit. Of the 300 identities in a publicly available cattle face dataset, 99 are withheld from training entirely, the enrolment gallery is disjoint from the training images, and the decision threshold is selected on a development split and frozen before the test split is scored. Working from 128 × 128 pixel images, and on 343 known and 478 unknown test queries, the framework attains a Top-1 accuracy of 98.25%, a known-versus-unknown ROC-AUC of 0.9887 and a detection equal error rate of 4.59%. At the high-security operating point the false accept rate is 2.72%, and the misidentification rate is zero at every threshold examined. Confidence intervals are obtained by identity-level rather than image-level resampling, which widens the interval on the false accept rate by a factor of 2.2. An ablation over sixteen configurations retrained on the identical split shows that the muzzle-focused cropping usually assumed necessary in the literature is in fact counterproductive, and quantitative explanation metrics show that the model’s evidence is distributed across the face rather than concentrated in the muzzle region.

Future streamflow projections are critical for water management. In this study, we coupled the Geomorphology-Based Ecohydrological Model (GBEHM) with the Physics-aware Hybrid Learning and eXtreme Gradient Boosting models to provide a preliminary assessment of streamflow variations in the Minjiang River basin (MRB) over the period of 2020–2099 under the emission scenarios of five CMIP6 models, using data from the 2010s as the baseline. We considered both climatic and anthropogenic influences, assuming that the current anthropogenic disturbances and river network configuration will remain unchanged. The performance of the GBEHM is acceptable, with error metrics exceeding 0.80 and 0.60 before and after the impoundment of the Zipingpu Reservoir, respectively. The cascade of data-driven models demonstrates good performance, with error metrics exceeding 0.90 over the whole simulation period. Under the influence of climate change, the decadal mean streamflow at Zipingpu station will decrease by 2.73–12.16% before 2069 and increase thereafter, while at Gaochang station, it will generally increase by 1.44–13.67% after 2020. Moreover, the decadal mean streamflow at Pengshan station will increase by 13.22–36.41% over the coming decades. However, the combined effects of anthropogenic disturbances and climate change will significantly decrease future streamflow by 61.91–112.16 m³/s on average, corresponding to a reduction of 14.08–25.51% from the baseline. We also suggest strategies to mitigate future water risks and enhance basin management in the MRB.

Protein content is one of the most important indicators for evaluating the nutritional value and quality grade of royal jelly. To achieve rapid and non-destructive quantification of protein content in royal jelly, this study employed near-infrared hyperspectral imaging (NIR-HSI) to acquire hyperspectral images of royal jelly samples, while the Kjeldahl method was simultaneously used as the reference method for protein determination. During data processing, seven spectral preprocessing methods—including the first derivative (1-Der), the second derivative (2-Der), Savitzky–Golay smoothing (SG), normalization (Normalize), baseline correction (Baseline), standard normal variate (SNV), and multiplicative scatter correction (MSC)—were comparatively evaluated. After outlier samples were identified and removed using the Mahalanobis distance method, Principal Component Regression (PCR) and Partial Least Squares Regression (PLSR) models were established for quantitative prediction of protein content in royal jelly. The PLSR models consistently outperformed the PCR models in predicting protein content in royal jelly. Among all preprocessing methods, the PLSR model developed using 1-Der spectra processing exhibited the best predictive performance, with a determination coefficient of calibration set () of 0.94, a determination coefficient of cross-validation set () of 0.90, a determination coefficient of prediction set () of 0.97, and a root mean square error of prediction (RMSEP) of 0.31%. These results support the feasibility of rapid and non-destructive laboratory-scale screening of protein content in royal jelly. The proposed method circumvents the drawbacks of conventional physicochemical analyses, which are labor-intensive, time-consuming and sample-destructive, and provides a basis for rapid and non-destructive laboratory-scale screening of protein content in royal jelly.

REBA, RULA, and OWAS are widely used observational ergonomic risk assessment methods based on predefined ordinal scoring rules. The final risk scores generated by these methods play a critical role in determining and prioritizing ergonomic interventions under limited resources to reduce musculoskeletal risks and related losses. Because the corresponding scores are deterministic functions of their component inputs, the present study focuses on machine-learning (ML) models as surrogate-analysis tools rather than replacements for exact scoring procedures. Ordinal Logistic Regression (AT, IT, SE), Random Forest (RF), Ordinal XGBoost, and Ordinal LightGBM were evaluated on full-factorial rule-space datasets using 5 × 3 nested cross-validation, 95% confidence intervals, and Friedman tests with Nemenyi post-hoc comparisons. Local transition fidelity and controlled adjacent-category and Gaussian perturbations were additionally examined against the exact scoring rules, while Random-Forest SHAP analysis was used to characterize the fitted surrogate structures. RF provided the highest overall clean-rule-space fidelity for REBA (MAE = 0.287, QWK = 0.982), RULA (MAE = 0.011, QWK = 0.991), and OWAS (MAE = 0.244, QWK = 0.929). However, under controlled input perturbations, the exact scoring procedures generally retained lower error than the clean-trained RF surrogates, indicating that the results do not support replacing the original rules with ML. Weight-sensitivity analysis further showed that the SRP-based method ranking was conditional on criterion weights: RULA ranked first in 52.23% of 100,000 sampled weight vectors, followed by OWAS (37.43%) and REBA (10.35%). Overall, the framework provides a reproducible way to examine surrogate fidelity, local rule-space transitions, model interpretation, and the weight sensitivity of ergonomic method selection.

Fast charging is critical to the wider adoption of electric vehicles, but simply increasing the charging current intensifies polarization and degradation, creating an inherent conflict between the charging speed and battery lifetime. Multi-stage charging can alleviate this conflict by redistributing the current throughout charging, although its performance depends jointly on the protocol structure and stage parameters. To address this problem, this work proposes a hierarchical multi-objective optimization method that coordinates these two design levels. A reduced-order electrochemical–thermal–aging model, developed and validated using systematic degradation experiments, is employed to predict electrothermal responses and capacity loss throughout the battery lifetime. For each candidate stage number, the charging rates, switching voltages, and exit-current ratios are jointly optimized, while the resulting Pareto performance and implementation complexity are compared at the structure level. The results showed that a three-stage structure captures most of the attainable performance gains without unnecessary control complexity. Full-lifetime cycling experiments demonstrated that the representative protocols achieved different balances between charging speeds and cycle lives. The 4C-referenced protocols extended cycle lives by 25.2–35.9% with only 2.7–4.1% longer initial charging times. The 3C-referenced protocols shortened initial charging times by 5.6–8.6%, while their cycle lives remained broadly comparable to 3C CCCV, ranging from a 9.5% decrease to a 6.9% increase. Multi-level degradation analyses further associated the lifetime improvements with a slower accumulation of anode-related capacity loss and interfacial deposits, together with reduced impedance growth.

Laser weeding offers significant advantages over conventional weed control methods; however, its practical deployment in maize fields remains challenging due to dynamic target variations, imprecise weed localization, and inefficient laser execution. In this study, a closed-loop laser weeding framework was developed for maize fields by integrating visual perception, multi-target tracking, safety-constrained decision making, and priority-based laser scheduling to achieve accurate and efficient weed treatment under dynamic field conditions. The system consists of visual perception, galvanometer control, and laser emission modules. ByteTrack was introduced to achieve stable ID assignment and continuous position feedback for weed targets, thereby reducing repeated ineffective irradiation and energy consumption. A target scheduling strategy constrained by a maize safety zone was further developed. An adaptive elliptical safety zone was constructed to screen candidate weed targets and optimize their priorities. By incorporating safety-zone modeling, galvanometer transition cost, and target urgency, the proposed strategy optimizes the laser striking sequence while reducing crop-injury risk and improving target-selection efficiency. The method was deployed on the developed platform and evaluated through field experiments under different travel speeds and illumination conditions. The results showed that the average weeding rate, maize seedling injury rate, and weed regrowth rate were 83.67%, 2.23%, and 5.23% under three travel speeds, and 82.57%, 2.63%, and 4.87% under three illumination levels, respectively. These results demonstrate that the proposed method enables stable weed tracking and efficient laser weeding while improving real-time performance, operational safety, and intelligent decision making. This study provides a deployable technical solution for precision laser weeding in field applications.

In recent years, China’s Central No. 1 Document has repeatedly highlighted the need to advance agricultural digitalization and green transformation, with the aim of improving production efficiency while reducing carbon emissions. For the livestock sector, emission reduction is an essential component of achieving the agricultural “dual carbon” goals, and digital transformation is becoming a functional mechanism for achieving a lower carbon footprint. Using balanced panel data, this study evaluates the development level of the digital economy and livestock-related carbon emissions, and further investigates how the former affects the latter. In addition, the paper examines whether regional economic development moderates this relationship and whether the effects extend across neighboring regions through spatial spillovers. According to the derived data: (1) During the sample period, livestock carbon emissions in China generally declined with fluctuations, although clear regional disparities remained; (2) The digital economy contributes to lower carbon emissions in livestock production; such emissions tend to be lower in regions with more advanced digital development; (3) The carbon reduction effect of the digital economy is further enhanced by regional economic development, indicating that its impact is more significant in more developed regions; (4) Spatial spillovers are also evident, as digital development in neighboring regions influences local livestock emissions. Collectively, the results highlight the importance of coordinating digitalization with low-carbon transitions in the livestock sector.

Background/Objectives: Hypovolaemia and hypotension in the postoperative neonatal period are life-threatening conditions difficult to diagnose with standard clinical parameters. The inferior vena cava collapsibility index (IVCCI) is a validated non-invasive tool in adults; its application in mechanically ventilated surgical neonates remains poorly defined. This study aimed to evaluate the association of the IVCCI with echocardiographic haemodynamic indices and its changes following volume expansion in surgically treated neonates. Methods: This monocentric observational study enrolled 36 surgical neonates (mean gestational age 38.1 weeks; mean weight 2799 g) with postoperative hypovolaemia and oliguria at the UOC NICU-Neonatology, AOU Policlinico G. Rodolico–San Marco, Catania (July 2020–September 2023). All patients were on controlled invasive mechanical ventilation. The IVCCI, cardiac output (CO), stroke volume (SV), and peak Doppler velocities (Vmax) of the pulmonary artery and aorta were measured before and 4–6 h after colloid infusion (10–20 mL/kg). Results: The IVCCI decreased significantly from 44.05 ± 7.8% to 15.14 ± 4.3% (p < 0.001). MAP improved from 36.89 ± 5.12 to 55.99 ± 4.87 mmHg (p < 0.001) and pH from 7.32 ± 0.04 to 7.36 ± 0.03 (p < 0.01). The Vmax, SV, and CO of both ventricles increased significantly (all p < 0.001). Significant inverse correlations were found between the IVCCI and MAP, Vmax, CO, and SV (all p < 0.001). No correlation was found with gestational age, birth weight, or pain scores. Conclusions: The IVCCI is a feasible and clinically useful non-invasive echocardiographic parameter for haemodynamic monitoring in critically ill surgical neonates. Its significant correlation with cardiac function indices supports its potential role in the bedside assessment of suspected hypovolaemia. Larger prospective studies are warranted.