Field Data Sets Research Articles

Dual-consistency constraints network for noisy facial expression recognition

Although existing facial expression recognition (FER) methods have achieved great success, their performance degrades significantly under noisy labels caused by low-quality images, ambiguous expressions, and subjective and incorrect labeling. Recent studies have shown that deep neural networks (DNNs) can easily overfit noisy labels, which poses a great challenge to FER task in real-world scenarios. To address this issue, we propose a novel Dual-consistency Constraints Network (DC-Net) to automatically suppress noisy samples during training. Specifically, we first propose a Class Activation Mapping (CAM) Attention Consistency (CAC), which makes the model focus on partially important feature information. As a result, we obtain more robust local feature representations and reduce excessive attention to noisy labels. Then, a Class Feature Consistency (CFC) is designed to encourage the model to focus on the global semantic information of the image. Finally, with the collaboration of the CAC and the CFC, DC-Net can learn robust local and global feature information to prevent the model from learning biased information with noisy labels. We conducted extensive experiments on three field datasets, including RAF-DB, AffectNet, and FERPlus2013. Experimental results show that DC-Net significantly outperforms state-of-the-art noisy labeling methods at different noise rates and generalizes well to other tasks with a large number of classes, such as CIFAR100 and Tiny- ImageNet.

Adapt-cMolGPT: A Conditional Generative Pre-Trained Transformer with Adapter-Based Fine-Tuning for Target-Specific Molecular Generation.

Small-molecule drug design aims to generate compounds that target specific proteins, playing a crucial role in the early stages of drug discovery. Recently, research has emerged that utilizes the GPT model, which has achieved significant success in various fields to generate molecular compounds. However, due to the persistent challenge of small datasets in the pharmaceutical field, there has been some degradation in the performance of generating target-specific compounds. To address this issue, we propose an enhanced target-specific drug generation model, Adapt-cMolGPT, which modifies molecular representation and optimizes the fine-tuning process. In particular, we introduce a new fine-tuning method that incorporates an adapter module into a pre-trained base model and alternates weight updates by sections. We evaluated the proposed model through multiple experiments and demonstrated performance improvements compared to previous models. In the experimental results, Adapt-cMolGPT generated a greater number of novel and valid compounds compared to other models, with these generated compounds exhibiting properties similar to those of real molecular data. These results indicate that our proposed method is highly effective in designing drugs targeting specific proteins.

Generation of reverse time migration dip gathers with the stabilized Poynting vector and their application in the improvement of subsalt images

Subsalt imaging is challenging with coherent noise prevalent beneath complex salt structures. The coherent noise degrades the imaging quality, making image enhancement and interpretation difficult and potentially erroneous. The dip gathers of reverse time migration (RTM) serve as an ideal domain for separating signals from coherent noise due to their distinct distributions in the dip domain. Methods utilizing the Poynting vector offer an efficient and cost-effective means to produce dip gathers. However, the presence of zero points in the Poynting vector causes instabilities in direction or angle estimation, leading to the leakage of reflection energy into false dip angles. This issue complicates the separation of desired signals from coherent noise. We address this instability by employing a stabilized Poynting vector to produce high-quality RTM dip gathers. The stabilized Poynting vector does not contain zero points within the range of wave propagation, thus mitigating the instability problem. The dip gathers generated using the stabilized Poynting vector provide a clearer and more precise depiction of signal and noise distributions, allowing us to identify the boundaries of desired signals and mute all noise outside these boundaries. Two numerical examples with a synthetic dataset and a field dataset are used to demonstrate our method's effectiveness in reducing coherent noise and improving quality of subsalt images.

Localization for surface microseismic monitoring based on arrival time correction and VFOM

Abstract Unconventional resources have emerged as the primary source to meet the escalating demand for energy consumption, with hydraulic fracturing standing out as an effective means of boosting production. The utilization of microseismic monitoring is crucial for acquiring real-time or semi-real-time extension information of the fracture network to guide the fracturing process. The precise positioning of microseismic events is a fundamental aspect of microseismic monitoring. Traditional methods relying on (relative) arrival time significantly impact positioning accuracy due to picking errors. While waveform-based methods offer high accuracy, they require precise velocity models and are time-consuming. To overcome challenges associated with arrival time pickup and velocity accuracy, we introduce a virtual field optimization method (VFOM) based on arrival time correction. Initially, an equivalent velocity model is established, and the arrival time difference resulting from the model transformation of the master event is calculated to correct the observed arrival time of the target event. Subsequently, we match detector pairs, establish hyperboloids based on the corrected arrival time difference, and employ the intersection point of all hyperboloids as the positioning result. After that, we use the location results of the master event to enhance the accuracy of the target event. Finally, we apply the proposed method to both synthetic test and field datasets, demonstrating a significant improvement in the positioning accuracy and stability provided by the novel method. The robustness against arrival time error renders it a suitable choice for surface monitoring applications where signal quality is compromised. Furthermore, the simplified velocity model significantly diminishes the computational requirements in the positioning process, enhancing its efficiency, and consequently holds vast potential for application in real-time monitoring.

Integrating Satellite Images and Machine Learning for Flood Prediction and Susceptibility Mapping for the Case of Amibara, Awash Basin, Ethiopia

Flood is one of the most destructive natural hazards affecting the environment and the socioeconomic system of the world. The effects are higher in the developing countries due to their higher vulnerability to disaster and limited coping capacity. The Awash basin is one of the flood-prone basins in Ethiopia where the frequency and severity of flooding has been increasing. Amibara district is one of the flood-affected areas in the Awash basin. To minimize the effects of flooding, reliable and up-to-date information on flooding is highly required. However, flood monitoring and forecasting systems are lacking in most basins of Ethiopia including the Awash basin. Therefore, this study aimed to (i) identify important flood causative factors, (ii) evaluate the performance of random forest (RF), linear regression, support vector machine (SVM), and long short-term memory (LSTM) machine learning models for flood prediction and susceptibility mapping in the Amibara area. For developing flood prediction and susceptibility modeling, nine causative factors were considered, namely elevation, slope, aspect, curvature, topographic wetness index, soil texture, rainfall, land use/land cover, and curve number. The Pearson correlation coefficient and information gain ratio (InGR) techniques were used to evaluate the relative importance of the factors. The machine learning models were trained and tested using 400 historic flood points collected from the 10 September 2020 Sentinel 2 image, during which a flood event occurred in the area. Multiple metrics, namely precession, recall, F1-score, accuracy, and receiver operating characteristics (area under curve), were used to evaluate the performance of the models. The results showed that all the factors considered in this study were important; elevation, rainfall, topographic wetness index, aspect, and slope were more important while land use/land cover, curve number, curvature, and soil texture were less important. Furthermore, the results showed that random forest outperformed in predicting and mapping flooding for the study area whereas the linear regression model showed the next best performance to RF. However, SVM performed poorly in flood prediction and susceptibility mapping. The integration of satellite and field datasets coupled with state-of-the-art-machine learning models are novel approaches and thus improved the accuracy of flood prediction and susceptibility mapping. Such methodology improves the state-of-the-art knowledge in this field and fills the gaps of traditional flood mapping techniques. Thus, the results of the study can provide crucial information for informed decision-making in the processes of designing flood control strategies and risk management.

A computational topology-based method for extracting fault surfaces

Fault surface extraction is a crucial step in seismic interpretation, which can help structural interpretation and structural modeling. A key focus of fault surface extraction research is to extract fault surfaces in their entirety as much as possible, rather than just in fault segments, which is more challenging in some complex fault situations. To address this challenge, we develop a fault surface extraction method based on computational topology to extract fault surfaces in their entirety as much as possible from a fault attribute and effectively handle some complex fault situations, such as intersecting faults. From a given seed point on the target fault, we use the idea of regional growth to search for high-confidence points on the target fault, called fault control points, under the constraints of the fault attribute and the calculated fault orientations. Through these fault control points, we extract the fault boundary and process the fault attribute so that only the target fault is included. Furthermore, we use an operation in computational topology called collapse to extract the target fault from the processed fault attribute using the fault boundary as a constraint. By incorporating fault orientation information and using a relatively large search distance during the control point search, our method enables the integration of segmented faults and facilitates the handling of complex fault situations such as intersecting faults. The collapse operation ensures that the extracted fault surfaces align with the fault attribute, correspond to the actual fault locations in seismic data, and enhance fault continuity. In addition, we develop an automatic method for picking seed points to realize the extraction of all the faults in the research data. We test our method on several field data sets and the experimental results demonstrate its effectiveness. In some complex fault situations, such as intersecting faults, our method performs well and indicates a significant improvement over the compared method.

A deep neural network and transfer learning combined method for cross-task classification of error-related potentials.

Error-related potentials (ErrPs) are electrophysiological responses that naturally occur when humans perceive wrongdoing or encounter unexpected events. It offers a distinctive means of comprehending the error-processing mechanisms within the brain. A method for detecting ErrPs with high accuracy holds significant importance for various ErrPs-based applications, such as human-in-the-loop Brain-Computer Interface (BCI) systems. Nevertheless, current methods fail to fulfill the generalization requirements for detecting such ErrPs due to the high non-stationarity of EEG signals across different tasks and the limited availability of ErrPs datasets. This study introduces a deep learning-based model that integrates convolutional layers and transformer encoders for the classification of ErrPs. Subsequently, a model training strategy, grounded in transfer learning, is proposed for the effective training of the model. The datasets utilized in this study are available for download from the publicly accessible databases. In cross-task classification, an average accuracy of about 78% was achieved, exceeding the baseline. Furthermore, in the leave-one-subject-out, within-session, and cross-session classification scenarios, the proposed model outperformed the existing techniques with an average accuracy of 71.81, 78.74, and 77.01%, respectively. Our approach contributes to mitigating the challenge posed by limited datasets in the ErrPs field, achieving this by reducing the requirement for extensive training data for specific target tasks. This may serve as inspiration for future studies that concentrate on ErrPs and their applications.

3D ground-penetrating radar data analysis and interpretation using attributes based on the gradient structure tensor

In near-surface geophysics, ground-penetrating radar (GPR) surveys are routinely used in a variety of applications including those from archaeology, civil engineering, hydrology, and soil science. Thanks to recent technical developments in GPR instrumentation and antenna design, 3D surveys comprising several hundred thousand traces can be performed daily. Especially in complex environments such as sedimentary systems, analyzing and interpreting the resulting GPR volumes is a time-consuming and laborious task that is still largely performed manually. In the past few decades, several data attributes have been developed to guide and improve such tasks and assure a higher degree of reproducibility in the resulting interpretations. Many of these attributes have been developed in image processing or computer vision and are routinely used, for example, in reflection seismic data interpretation. Especially in sedimentary systems, variations in the subsurface are accompanied by variations of GPR reflections in terms of the amplitudes, continuity, and geometry in view of the dip angle and direction. A promising tool to analyze such structural features is known as the gradient structure tensor (GST). To date, the application of the GST approach has been limited to a few 2D GPR examples. Thus, we take the basic idea of GST analysis and introduce and evaluate the corresponding attributes to analyze 3D GPR data. We apply our GST approach to one synthetic and two field data sets imaging diverse sedimentary structures. Our results demonstrate that our set of GST-based attributes can be efficiently computed in three dimensions and that these attributes represent versatile measures to address different typical interpretation tasks and, thus, help for an efficient, reproducible, and more objective interpretation of 3D GPR data.

Remote sensing estimation of dissolved organic carbon concentrations in Chinese lakes based on Landsat images

Long-term estimation of dissolved organic carbon (DOC) dynamics based on satellite data provides important information to comprehensively understand the lake carbon cycle, climate change, and sustainable water resource management. Here, we proposed a novel DOC concentration estimation model based on water classification using a large field dataset (n = 1929) collected from 123 Chinese lakes. We divided water into terrestrial-DOM-dominated and endogenous-DOM-dominated water bodies using the optimal spectral slope ratio (SR) threshold determined by setting the classification scenario. Then, we calibrated the optimal DOC estimation model using stepwise quadratic polynomial regression for the different types of water and determined the final DOC estimation model. The final DOC estimation model performed well in both the calibration and validation datasets, with R2 ≥ 0.78, RMSD ≤ 0.78 mg/L, and MAPD ≤ 16.51 %. Subsequently, we mapped the spatiotemporal dynamics of DOC concentrations in Chinese lakes with surface area greater than 1 km2 (n = 2621) between 1986 and 2021 using the Landsat data. Spatially, we observed obvious spatial heterogeneity in the DOC concentrations in the Chinese lakes, with the highest DOC concentration occurring in the Northeast Region (NER) (4.39 ± 0.21 mg/L) and the lowest DOC concentration occurring in the Qinghai-Tibet Plateau Region (QTR) (3.76 ± 1.12 mg/L). Temporally, the average DOC concentration in Chinese lakes generally decreased at a rate of − 0.09 mg/L decade−1 from 1986 to 2021, during which 56.5 % of the lakes experienced a decrease, 29.6 % of the lakes experienced no obvious change, and 13.9 % of the lakes experienced an increase. We also revealed that human activities are the main drivers of DOC increases in lakes, while vegetation and climate change are the main drivers of DOC decreases in lakes. Our results contribute to improving the accuracy of organic carbon estimations in lakes worldwide and provides meaningful insights into the lake carbon cycle..

Downscaling canopy photochemical reflectance index to leaf level by correcting for the soil effects

The photochemical reflectance index (PRI) is a promising remote sensing signal for monitoring vegetation physiology. Variations in leaf PRI are usually attributed to either the energy-dependent xanthophyll cycle or the carotenoid-chlorophyll ratio, both indicative of leaf physiology. However, canopy PRI is subject to soil, canopy structure, and incident and viewing angles, and thus has a weaker and more complicated relationship with vegetation photosynthetic activity or leaf pigment composition. Therefore, downscaling canopy PRI to the leaf level is essential for accurate remote sensing of vegetation physiology using PRI. An earlier investigation (P.Yang, RSE, 279, 113,133, 2022) illustrates that structural and angular variations in canopy PRI primarily result from varying degrees of soil interference. In this study, a soil correction method is proposed to mitigate the soil effects on top-of-canopy (TOC) reflectance at the PRI bands. The soil effects on TOC reflectance at 531 nm and 570 nm are respectively estimated as R531soil×R675/R675soil and R570soil×R675/R675soil, where R675 is TOC red reflectance, Rλsoil soil reflectance at wavelengthλ. R675/R675soil approximates the fraction of the observed sunlit soil, as leaves are nearly black at 675 nm due to strong absorption of chlorophyll, and R675 is mainly contributed from soil. To assess the effectiveness of the soil correction method, a wheat field dataset, a corn field dataset, and a simulated dataset, were utilized. The soil-adjusted and the original canopy PRI were compared with the leaf PRI for the real and synthetic scenarios that had various soil brightness, leaf chlorophyll content and canopy structure. Both the field and numerical experiments demonstrate that, for the canopies with low vegetation coverage and substantial soil contamination, the original canopy PRI was largely different from the leaf PRI, displaying substantial structural and angular dependence. In comparison, the soil-adjusted canopy PRI was more closely aligned with the PRI observed in the sunlit leaves in all three datasets. This study shows that accounting for the soil effects with TOC red reflectance allows downscaling canopy PRI to the leaf level. The soil-adjusted canopy PRI contributes to remote sensing of the xanthophyll cycle or the carotenoid-chlorophyll ratio from canopy PRI.

The distribution of velocity and sediment concentration under currents in coastal seas

A power law velocity profile for engineering application was obtained based on the field data set of flow velocity and sediment concentration in coastal areas of Jinzhou, Nantong, and Zhoushan. Using theoretical analyses, van Rijn's formula for sediment concentration profile was modified. Tests of the new power-law velocity profile and the modified formula of suspended sediment concentration against a large data set were performed. The results show a good fit to the data and that these two formulae could demonstrate the basic principles of vertical distribution of flow velocity and suspended sediment concentration, respectively.

Parameter Estimation of Cellular Communication Systems Models in Computational MATLAB Environment: A Systematic Solver-based Numerical Optimization Approaches

Model-based parameter estimation, identification, and optimisation play a dominant role in many aspects of physical and operational processes in applied sciences, engineering, and other related disciplines. The intricate task involves engaging and fitting the most appropriate parametric model with nonlinear or linear features to experimental field datasets priori to selecting the best optimisation algorithm with the best configuration. Thus, the task is usually geared towards solving a clear optimsation problem. In this paper, a systematic-stepwise approach has been employed to review and benchmark six numerical-based optimization algorithms in MATLAB computational Environment. The algorithms include the Gradient Descent (GRA), Levenberg-Marguardt (LEM), Quasi-Newton (QAN), Gauss-Newton (GUN), Nelda-Meald (NEM), and Trust-Region-Dogleg (TRD). This has been accomplished by engaging them to solve an intricate radio frequency propagation modelling and parametric estimation in connection with practical spatial signal data. The spatial signal data were obtained via real-time field drive test conducted around six eNodeBs transmitters, with case studies taken from different terrains where 4G LTE transmitters are operational. Accordingly, three criteria in connection with rate of convergence Results show that the approximate hessian-based QAN algorithm, followed by the LEM algorithm yielded the best results in optimizing and estimating the RF propagation models parameters. The resultant approach and output of this paper will be of countless assets in assisting the end-users to select the most preferable optimization algorithm to handle their respective intricate problems.

Systems genomics of salinity stress response in rice.

Populations can adapt to stressful environments through changes in gene expression. However, the role of gene regulation in mediating stress response and adaptation remains largely unexplored. Here, we use an integrative field dataset obtained from 780 plants of Oryza sativa ssp. indica (rice) grown in a field experiment under normal or moderate salt stress conditions to examine selection and evolution of gene expression variation under salinity stress conditions. We find that salinity stress induces increased selective pressure on gene expression. Further, we show that trans-eQTLs rather than cis-eQTLs are primarily associated with rice's gene expression under salinity stress, potentially via a few master-regulators. Importantly, and contrary to the expectations, we find that cis-trans reinforcement is more common than cis-trans compensation which may be reflective of rice diversification subsequent to domestication. We further identify genetic fixation as the likely mechanism underlying this compensation/reinforcement. Additionally, we show that cis- and trans-eQTLs are under different selection regimes, giving us insights into the evolutionary dynamics of gene expression variation. By examining genomic, transcriptomic, and phenotypic variation across a rice population, we gain insights into the molecular and genetic landscape underlying adaptive salinity stress responses, which is relevant for other crops and other stresses.

Calibration of VISSIM for an Urban Corridor under Heterogeneous Traffic Conditions

The replication of heterogeneous traffic conditions in simulation tools is a challenging task due to involvement of various vehicle classes and their associated driving behaviours. Despite considerable efforts have been made in literature, further attention is still needed due to the complexity of heterogeneous traffic mixes, particularly for urban corridors with closely spaced intersections. In the present study, a systematic calibration procedure for VISSIM has been proposed by considering the vehicle-class specific driver behaviour parameters. The proposed procedure involved four key steps including; identification of sensitive parameters via sensitivity analysis, generation of random samples by using Latin Hypercube Design (LHD), development of regression model based on LHD samples and VISSIM output, and the use of regression model in Microsoft Excel Solver program to determine candidate parameter set producing measure of effectiveness (MOE) values closer to field values. After calibration, the model validation was also conducted which was done by assessing the statistical similarity between field and simulation data sets via paired t-test. The study concluded that the vehicles in heterogeneous traffic conditions tend to adopt smaller longitudinal and lateral gaps during standing and driving conditions. The proposed methodology can help the practitioners in developing simulation models for similar traffic conditions.

Exploring the relationship between pavement mean texture depth and mean profile depth: from theoretical derivation to field results

Pavement mean texture depth (MTD) and mean profile depth (MPD) are common evaluation indicators for characterizing the pavement surface. Clarifying the relationship between the two contributes to coordinate various measurement methods and evaluation criterions. The calculation methods of MTD and MPD were first introduced to trace the similarity between the indicators. Subsequently, the tortuosity area of surface texture was observed using x-ray Computer Tomography (CT). To further determine the model parameter, over 3000 sets of field data in Research Institute of Highway Ministry of Transport (RIOH) track were measured. Meanwhile, the field results of another track and existing recommended model were also used to validate the model. The results show that the shift relationship between MTD and MPD is theoretically unified into a linear model with a slope of 1 and an intercept related to the surface tortuosity. These tortuous areas, the main difference source between the two indicators, would lead to a greater equivalent texture depth under the coarser gradation or lower the compactness of the asphalt mixture. Even for the different track roads, the determined model (MTD = MPD + 0.03) has a lower mean relative estimation error, indicating good applicability. This study provides theoretical explanation and empirical references for the automated laser-based texture detection.

Boosting field data using synthetic SCADA datasets for wind turbine condition monitoring

State-of-the-art Deep Learning (DL) methods based on Supervisory Control and Data Acquisition (SCADA) system data for the detection and prognosis of wind turbine faults require large amounts of failure data for successful training and generalisation, which are generally not available. This limitation prevents benefiting from the superior performance of these methods, especially in SCADA-based failure prognosis. Data augmentation approaches have been proposed in the literature for generating failure data instances within a SCADA sequence to reduce the imbalance between healthy and faulty state data points, which is relevant to fault detection tasks. However, the successful implementation of DL-based failure prognosis methods requires the availability of multiple run-to-failure SCADA sequences. This paper proposes a data-driven method for generating synthetic run-to-failure SCADA sequences with custom operational and environmental conditions and progression of degradation. An Artificial Neural Network (ANN) is trained with signals that represent these factors to reconstruct the SCADA signals. Then, it is used to generate synthetic SCADA datasets based on data available from a wind turbine that experienced a gearbox failure. Synthetic data sets generated are evaluated on the basis of the similarity of their signal distributions, the temporal dynamics within each signal, and the temporal dynamics among different SCADA signals with those in similar field datasets. The results show that the generated synthetic datasets are consistent with their field counterparts, with a comparatively lower diversity in their dynamic behaviour in time.

Contribution to the distribution of mosquitoes in Great Britain, with northernmost records

Abstract We report occurrence data of mosquitoes in Scotland which were routinely collated by the British Mosquito Group for the period 1886 to 1993 (232 records), and by the Mosquito Recording Scheme since 2005 (11 records). We also report a set of new field data mainly based on mosquito immature sampling in aquatic habitats during the summers 1998 and 2012, in both Scotland (20 data) and England (11 data). The occurrences in Scotland (263 data in total) include a total mosquito fauna of 18 species. They include, for the British Isles, the northernmost record of Anopheles claviger sensu stricto, collected on Orkney, and the northernmost record of Culex pipiens, collected on Shetland, both in 2012. The latter is the northernmost mosquito record of Great Britain to date.

Fast Prediction and Optimization of Building Wind Environment Using CFD and Deep Learning Method

CFD offers advantages over wind tunnel experiments in the prediction and optimization of building wind environment; however, the computational costs associated with optimizing architectural wind environment remain a challenge. In this study, an approach that combines deep learning techniques with CFD simulations is proposed for the prediction and optimization of the architectural wind environment efficiently. A dataset of wind field is constructed using CFD simulation, considering various wind directions, wind speeds, and building spacing. Subsequently, a U-net deep learning model is trained as a surrogate model to rapidly predict the architectural wind field under different conditions. The results indicate that the model can accurately predict the wind field in buildings. The prediction time of building wind field is only 1/900 of that of CFD simulations, making it a viable surrogate model for wind environment optimization. Furthermore, considering all the building layouts and inflow conditions examined in this study, the maximum and minimum uniform wind speed area ratios Auni are 0.84 and 0.13, respectively. Under a single inflow speed, the maximum improvement in the Auni is 0.4, with an improvement rate of 48%. The results demonstrate the effectiveness of the proposed method as an efficient approach for optimizing architectural wind environment.

MicroSimACC: an open database for field experiments on the potential capacity impact of commercial Adaptive Cruise Control (ACC)

Commercial availability of vehicle automation has become mainstream. Most of today’s new vehicles can perform longitudinal car following autonomously via Adaptive Cruise Control (ACC). Field experiments demonstrate that today’s commercially available ACC vehicles provide similar headways and capacities as human-driven vehicles on freeways under steady-state and free-flow conditions. However, field tests also demonstrated that the design of today’s commercially available ACC vehicles can lead to further capacity reduction when operating in non-steady-state conditions where queues are present and speeds frequently fluctuate. These experiments generated MicroSimACC, a comprehensive set of field data that encompasses full speed range car following with interruptions from lane change manoeuvres. This will benefit the research community by providing benchmark data for developing models to be integrated into microscopic simulations for more prospective analyses and planning.

Estimating kill intervals for a specific prey species using location clusters from GPS-collared Eurasian lynx (Lynx lynx)

An increasing number of GPS telemetry studies have helped to gain important insights into predator-prey relationships in recent years. However, considerable time and effort is needed to evaluate whether GPS location clusters (GLCs) reflect predation events. To reduce field effort, predictive models are being developed to calculate predator kill intervals, but few studies have attempted to do this for a specific species of prey. Between 2013 and 2018, we studied predation by 13 GPS-collared Eurasian lynx (Lynx lynx) on Alpine chamois (Rupicapra rupicapra) in the northwestern Swiss Alps. Our objectives were to predict the total number of killed chamois, including potential kills in unchecked GLCs, and to evaluate if model predictions were sufficiently accurate. We built a set of generalized linear models (GLM) predicting the occurrence of GLCs containing lynx-killed chamois (1) versus GLCs containing other prey types or no prey (0) and compared their predictive performance by means of k-fold cross-validation. We found that model performance was very similar for all candidate models, with the full model yielding the best cross-validation result (accuracy = 0.83, sensitivity = 0.43, specificity = 0.94). Female lynx killed on average one chamois every 11.9 days (10.6–13.0 days, 95% CI); male lynx killed one chamois every 7.2 days (6.7–7.6 days, 95% CI). Our model showed high specificity for detecting non-chamois GLCs, but sensitivity for detection of GLCs with actual chamois kills was low. We conclude that the sensitivity of the models should be further improved, but the results can be sufficient for practical application. Predictive modelling approaches do not replace extensive fieldwork but require large sets of field data, high individual variability and thorough knowledge of a predator’s ecology and prey community. Our approach may provide useful results for binary classifications in rather simple predator-prey systems, but extrapolations from one study system to another might be difficult.