Estimation Of Value Research Articles

Dimension-independent weak value estimation via controlled SWAP operations

Weak values of quantum observables are a powerful tool for investigating quantum phenomena. Some methods for measuring weak values in the laboratory require weak interactions and postselection, while others are deterministic, but require statistics over a number of experiments that grows linearly with the dimension of the measured system in the worst case over all possible observables. Here we propose a deterministic dimension-independent scheme for estimating weak values of arbitrary observables. The scheme is based on controlled operations and associates states and observables in the mathematical expression of the weak value to preparations devices and measurement devices in the experimental setup, respectively. Thanks to this feature, it provides insights into the relation between states of two identical quantum systems at a single moment of time and states of a single quantum system at two moments of time, also known as two-time states. Specifically, our scheme provides an alternative expression for two-time states and establishes a link between two-time states accessible through the controlled- scheme and bipartite quantum states with positive partial transpose. Published by the American Physical Society 2024

U-Net Semantic Segmentation-Based Calorific Value Estimation of Straw Multifuels for Combined Heat and Power Generation Processes

This paper proposes a system for real-time estimation of the calorific value of mixed straw fuels based on an improved U-Net semantic segmentation model. This system aims to address the uncertainty in heat and power generation per unit time in combined heat and power generation (CHPG) systems caused by fluctuations in the calorific value of straw fuels. The system integrates an industrial camera, moisture detector, and quality sensors to capture images of the multi-fuel straw. It applies the improved U-Net segmentation network for semantic segmentation of the images, accurately calculating the proportion of each type of straw. The improved U-Net network introduces a self-attention mechanism in the skip connections of the final layer of the encoder, replacing traditional convolutions by depthwise separable convolutions, as well as replacing the traditional convolutional bottleneck layers with Transformer encoder. These changes ensure that the model achieves high segmentation accuracy and strong generalization capability while maintaining good real-time performance. The semantic segmentation results of the straw images are used to calculate the proportions of different types of straw and, combined with moisture content and quality data, the calorific value of the mixed fuel is estimated in real time based on the elemental composition of each straw type. Validation using images captured from an actual thermal power plant shows that, under the same conditions, the proposed model has only a 0.2% decrease in accuracy compared to the traditional U-Net segmentation network, while the number of parameters is significantly reduced by 74%, and inference speed is improved 23%.

Estimation of liver standardized uptake value in F18-FDG PET/CT scanning: impact of different malignancies, blood glucose level, body weight normalization, and imaging systems.

The aim of this work was to investigate homogeneity and stability of liver SUV in terms of different malignancies considering different body normalization schemes and blood glucose concentrations as well as PET/CT imaging systems. The study included 207 patients with four different types of cancers namely breast, lymphoma, lung, and bone-metastasis. Data acquisition was performed with GE Discovery IQ, Biograph mCT, uMI 550, and Ingenuity TF64 after a single intravenous injection of 194 ± 67.5 MBq of 18F-FDG. In body weight normalization, SUVmax and SUVmean in bone-mets as well as SUVmean in lung patients were not statistically different among scanners especially for data corrected for glucose levels (p = 0.062, 0.121, and 0.150, respectively). In SUVlbm derived from lung patients, there was no significant differences in Philips in comparison to GE and Siemens (both, p > 0.05) for data corrected and not corrected for glucose levels. In SUVbsa, the only non-significant difference revealed among scanners was in the measurements of SUVmean obtained from lung and bone-mets (p = 0.107 and 0.114) both corrected for glucose levels. In SUVbmi, SUVmean of lung and bone-mets as well as SUVmax of bone-mets showed a non-significant differences among the four different scanning systems (p = 0.303, 0.091, and 0.222, respectively) for data corrected for glucose levels. Liver glucose correction needs further investigations in individual tumors but could be potentially affected by whether measurements are made on SUVmean versus SUVmax, body weight normalization, as well as the imaging system. As such, selection of normalization to body weight method should be carefully selected before clinical adoption and clinically adopted and body surface area would provide the highest correlation. As such, normalization of body weight should be carefully made before clinical adoption. SUVmean proves to be useful and stable metric when liver is corrected for blood glucose levels.

Precision and uncertainty in natural gas calorific value estimation: advanced combinatorial predictive models for complex pipeline systems

Abstract In recent years, natural gas pipelines have been characterized by multiple intake points, the mixing of different gas sources, and significant variations in gas quality. The existing volumetric measurement system is no longer suitable for the development and operation of natural gas pipelines in China. To ensure accurate and equitable implementation of natural gas measurement, there is a gradual shift towards energy measurement. However, due to numerous measurement interfaces, installing chromatographs at all measurement stations would lead to substantial investment costs. Therefore, predicting the calorific value of natural gas is a key technology for energy measurement. In view of the issues existing in the current methods of obtaining natural gas calorific value, different prediction models of natural gas calorific value are constructed. According to the influencing factors of natural gas calorific value, the uncertainty evaluation is carried out, and the accuracy of different calorific value prediction models is analyzed by case data. The maximum error of regression equation (RE), response surface analysis (RSA), BP neural network and genetic algorithm (GA) prediction model is 2.29%, 0.43%, 0.59% and 0.45% respectively. However, the combined calorific value (CCV)prediction model is 0.41%. The results indicate that the predicted value calculated by the CCV prediction model is closer to the actual value and has higher prediction performance, which provides a new method for the calorific value prediction and measurement management of natural gas.

Digitally Controllable Multifrequency Impedance Emulator for Bioimpedance Hardware Validation

ABSTRACTThe accurate emulation of the electrical impedance of biological tissues is crucial for the development and validation of bioimpedance measurement devices and algorithms. This paper describes a digitally controllable impedance emulator capable of reproducing values representative of tissue bioimpedance in user‐specified resistance, reactance, and frequency ranges up to 1 MHz. The presented solution uses a 2R‐1C impedance model to emulate the impedance characteristics of a biological tissue. Specific selection of each element value in this model is achieved using analog multiplexers with low resistance. A MATLAB algorithm was developed for value estimation using target impedance requirements. An example design to emulate impedance from 1 kHz to 1 MHz with 10 to 400 resistance and maximum reactance is provided. The nonideal behavior of this design was evaluated and compared against experimentally collected impedance measurements. Deviations of <1% were observed between experimental and theoretical resistances for values up to 100 kHz (with approximately 5% deviations up to 1 MHz) and reactance deviations were also <1% up to 10 kHz. High frequency deviations are attributed to the parasitic capacitance in the realization of the design. The experimental results validate the design approach and realization for low frequencies. Overall, the innovation of the proposed approach is the control of both resistance and reactance for emulating electrical impedance representative of biological tissues.

A new Bayesian method for estimation of value at risk and conditional value at risk

AbstractValue at Risk (VaR) and Conditional Value at Risk (CVaR) have become the most popular measures of market risk in Financial and Insurance fields. However, the estimation of both risk measures is challenging, because it requires the knowledge of the tail of the distribution. Therefore, Extreme Value Theory initially seemed to be one of the best tools for this kind of problems, because using peaks-over-threshold method, we can assume the tail data approximately follow a Generalized Pareto distribution (GPD). The main objection to its use is that it only employs observations over the threshold, which are usually scarce. With the aim of improving the inference process, we propose a new Bayesian method that computes estimates built with all the information available. Informative prior Bayesian (IPB) method employs the existing relations between the parameters of the loss distribution and the parameters of the GPD that models the tail data to define informative priors in order to perform Metropolis–Hastings algorithm. We show how to apply IPB when the distribution of the observations is Exponential, stable or Gamma, to make inference and predictions. .Afterwards, we perform a thorough simulation study to compare the accuracy and precision of the estimates computed by IPB and the most employed methods to estimate VaR and CVaR. Results show that IPB provides the most accurate, precise and least biased estimates, especially when there are very few tail data. Finally, data from two real examples are analysed to show the practical application of the method.

Assessing extreme significant wave height in China’s coastal waters under climate change

Accurately estimating the return values of significant wave height is essential for marine and coastal infrastructure, particularly as climate change intensifies the frequency and intensity of extreme wave events. Traditional models, which assume stationarity in wave data, often underestimate future risks by neglecting the impacts of climate change on wave dynamics. Combining time series decomposition and recurrence analysis, the research develops a nonstationary framework to predict significant wave height. The stochastic component is modelled using a stationary probability distribution, while the deterministic component is predicted based on sea surface temperature projections from CMIP6 climate scenarios. The model evaluation demonstrates strong predictive capability for both stochastic and deterministic components. Application of the model to China’s coastal waters reveals significant discrepancies between stationary and nonstationary return value estimates. Compared to conventional distribution models, the nonstationary model predicts substantial increases in extreme wave heights. These findings underscore the importance of adopting nonstationary models to more accurately assess future risks posed by extreme wave events in a changing climate.

Drone Insights: Unveiling Beach Usage through AI-Powered People Counting

Ocean beaches are a major recreational attraction in many coastal cities, requiring accurate visitor counts for infrastructure planning and value estimation. We developed a novel method to assess beach usage on the Gold Coast, Australia, using 507 drone surveys across 24 beaches. The surveys covered 30 km of coastline, accounting for different seasons, times of day, and environmental conditions. Two AI models were employed: one for counting people on land and in water (91–95% accuracy), and another for identifying usage types (85–92% accuracy). Using drone data, we estimated annual beach usage at 34 million people in 2022/23, with 55% on land and 45% in water—approximately double the most recent estimate from lifeguard counts, which are spatially limited and prone to human error. When applying similar restrictions as lifeguard surveys, drone data estimated 15 million visits, aligning closely with lifeguard counts (within 9%). Temporal (time of day, day of the week, season) and spatial (beach location) factors were the strongest predictors of beach usage, with additional patterns explained by weather variables. Our method, combining drones with AI, enhances the coverage, accuracy, and granularity of beach monitoring, offering a scalable, cost-effective solution for long-term usage assessment.

Joint torque estimation method of industrial robot by combining error compensation model based on LSTM and iterative weighted parameter identification

It is essential for human-robot interaction to accurate joint torque value estimation of industrial robots without force/torque (F/T) sensors. The most common method to estimate the value of robot joint torque is based on dynamic model parameter identification. However, no matter how elaborate the modeling methods are used, there are always uncertainty errors when robot’s joint rotation direction changes. In this paper, an identification framework is proposed to estimate optimal model parameters, and a deep learning network is proposed to compensate for the uncertainty errors caused by the unmodeled dynamics part. At first, the dynamic parameters are estimated during the Least Squares (LS) identification process considering the noise influence and data outliers, and the nonlinear friction model is unified into the identification framework. Secondly, based on Long-Short Term Memory (LSTM) network, an error compensation model (ECM) is proposed to establish the mapping relationship between the joint motion and the identification errors. Finally, a 6-DOF robot is used for parameter identification and ECM validation. Experimental results show that the proposed identification method is better than the LS method. Compared with the torque value estimated by the identification method, the proposed ECM can compensate for the uncertainty errors and improve the torque estimation accuracy.

Interval prediction method for dynamic zoning of displacements in super-high arch dam under significant water level fluctuations

Displacement is one of the most intuitive indicators reflecting the influence of internal and external environments on dam structures. However, existing displacement monitoring models mostly rely on static fitting of single point and deterministic point value estimation, failing to deeply consider the similarity and uncertainty of displacement in dynamic changes under significant water level fluctuations. To address this issue, the displacements are first divided into three periods: ascending, descending, and stable periods. Second, based on the Ward criterion, the displacements of measuring points in different periods are clustered and partitioned, and the CRITIC method is applied to amalgamate the displacement of multiple points within each partition into comprehensive displacements, which reflect the displacement changes of the corresponding partition. Then, the comprehensive displacements are predicted utilizing deep learning model convolutional neural network-gated recurrent unit (CNN-GRU). Finally, to consider the uncertainty and randomness of displacement changes, prediction intervals based on Bootstrap method are employed to quantify the impact of uncertain factors. Engineering examples show that the proposed method can provide scientific basis and technical support for the safety monitoring and health diagnosis of super-high arch dams.

Efficient algorithm for stochastic rumor blocking problem in social networks during safety accident period

Rumor sources spread negative information throughout the network, which may cause unbelievable results in real society especially for social safety field. Propagating positive information from several “protector” users is an effective method for rumor blocking once the rumor is detected. Based on the probability of each user being a rumor, “protector” nodes need to be selected in order to prepare for rumor blocking. Given a social network G=(V,E,P,Q), where P(u,v) is the probability that v is activated by u after u is activated, and Q is the weight function on node set V, Qv is the probability that v will be a rumor source. Stochastic Rumor Blocking (SRB) problem is to select k nodes as “protector” such that the expected eventually influenced users by rumor is minimized. SRB will be proved to be NP-hard and the objective function is supermodular. We present a Compound Reverse Influence Set sampling method for estimation of the objective value which can be represented as a compound set function. A randomized greedy algorithm with theoretical analysis will be presented and other two different “protector” selection strategies will be proposed for comparison. Finally, we evaluate our algorithm on real world data sets and do comparison among different strategies.

Research on extreme value estimations and dynamic coefficients of wheel load for high-speed railway based on generalized Pareto distribution theory

The design value of wheel load is one of the key design parameters for high-speed railway ballastless track structure, which essentially belongs to the extreme estimation of wheel load. Based on the generalized Pareto distribution theory, this article carries on the research on the extreme value estimations of wheel load excited by whole wavelength range irregularity, which provides a basis for the design of ballastless track structure. When using automatic threshold selection method, sample declustering can improve the independence between data. The shape parameter screening method is proposed for the extreme value estimations of wheel load, and the corresponding sample size and shape parameter range for each cluster are determined. A reasonable sample size is determined, and the mean of the extreme estimations of wheel load is used as the final extreme estimation of wheel load. The dynamic coefficients that match the design life of the ballastless track structure are obtained.

Estimation of gross calorific value of coal based on the cubist regression model.

The gross calorific value (GCV) of coal is an important parameter for evaluating coal quality, and regression analysis methods can be used to predict GCV. In this study, we proposed a GCV prediction model based on cubist regression. To develop a good regression model, feature selection of input variables was performed using a correlation analysis and a recursive feature elimination algorithm. Thus, in this study, we determined three sets of variables as the optimal combination for regression models: proximate analysis variables (Set 1: moisture, standard ash, and volatile matter), element analysis variables (Set 2: carbon, sulfur, and oxygen), and comprehensive index variables (Set 3: carbon, volatile matter, standard ash, sulfur, moisture, and hydrogen). Results for comparison with multiple linear regression, random forest regression, and numerous previous prediction models, such as gradient boosting regression tree, support vector regression (SVR), backpropagation neural networks, and particle swarm optimization-artificial neural network (PSO-ANN), indicate that these seven regression models have the best fitting effect on the comprehensive index variables among the three sets of input variables. The cubist model showed higher prediction accuracy and lower error than most other models (R2, mean absolute error, root mean square error, and average absolute relative deviation percentage values are 0.990, 0.476, 0.668, and 0.086% for the proximate analysis variables; 0.992, 0.381, 0.596, and 0.140% for element analysis variables; and 0.999, 0.161, 0.219, and 0.087% for comprehensive index variables, respectively). The cubist model combines the advantages of decision tree and linear regression, which not only enables it to perform well in terms of accuracy but also makes the model highly interpretable because it is based on multiple sublinear equations. In addition, the cubist model shows obvious advantages in terms of running speed, especially compared with SVR and PSO-ANN, which require complex parameter optimization. In summary, the cubist model considers the prediction accuracy, model interpretability, and computational efficiency as well as provides a new and effective method for GCV prediction.

Machine Learning Valuation in Dual Market Dynamics: A Case Study of the Formal and Informal Real Estate Market in Dar es Salaam

The housing market in Dar es Salaam, Tanzania, is expanding and with it a need for increased market transparency to guide investors and other stakeholders. The objective of this paper is to evaluate machine learning (ML) methods to appraise real estate in formal and informal housing markets in this nascent market sector. Various advanced ML models are applied with the aim of improving property value estimates in a market with limited access to information. The dataset used included detailed property characteristics and transaction data from both market types. Regression, decision trees, neural networks, and ensemble methods were employed to refine property appraisals across these settings. The findings indicate significant differences between formal and informal market valuations, demonstrating ML’s effectiveness in handling limited data and complex market dynamics. These results emphasise the potential of ML techniques in emerging markets where traditional valuation methods often fail due to the scarcity of transaction data.

A genomic analysis of twinning rate and its relationship with other reproductive traits in Holstein-Friesian cattle

Twin birth in dairy cattle is generally unfavorably associated with reproductive performance and calf survival in dairy cows. Genetic selection to reduce twinning rate in dairy cattle may be desirable, provided there are no undesirable correlated responses in other traits. The current study was undertaken to characterize the genomic architecture of twinning rate in the Irish Holstein-Friesian population, and to quantify the genetic relationship of twinning with other reproductive traits and milk yield. Calving records from the years 1996 to 2022 were used together with pedigree information to generate breeding value estimates for twinning rate. Genome-wide association analyses of twinning rate, calving interval, cow survival and age at first calving were conducted using de-regressed breeding values estimates for 2,656 Holstein-Friesian sires. Full genome sequence data imputed from approximately 50,000 single nucleotide polymorphisms were available for all sires. The heritability of twinning rate was 0.0118 ± 0.0010. Twinning rate was very weakly genetically correlated with both milk yield (0.13) and the reproductive traits (-0.26 to 0.14). Genomic analyses detected an association with twinning rate at 31.1 Mb on BTA11 in close proximity to genes for follicle-stimulating hormone receptor and luteinizing hormone-chorionic gonadotropin receptor, supporting previous studies. The most significant SNP in this region was not associated with milk yield, indicating the potential for selection to reduce twinning rate without detrimentally affecting milk yield. Novel SNP associations with age at first calving on BTA27 and from a meta-analysis of calving interval and age at first calving on BTA29 were also identified and are candidates for future validation and study.

Comprehensive genome-wide analysis of genetic loci and candidate genes associated with litter traits in purebred Berkshire pigs of Korea.

The objective of this study was to identify genomic regions and candidate genes associated with the total number of piglets born (TNB), number of piglets born alive (NBA), and total number of stillbirths (TNS) in Berkshire pigs. This study used a total of 11,228 records and 2,843 single-nucleotide polymorphism (SNP) data obtained from Illumina porcine 60 K and 80 K chips. The estimated genomic breeding values (GEBVs) and SNP effects were estimated using weighted single-step genomic BLUP (WssGBLUP). The heritabilities of the TNB, NBA, and TNS were determined using single-step genomic best linear unbiased prediction (ssGBLUP). The heritability estimates were 0.13, 0.12, and 0.015 for TNB, NBA, and TNS, respectively. When comparing the accuracy of breeding value estimates, the results using pedigree-based BLUP (PBLUP) were 0.58, 0.60, and 0.31 for TNB, NBA, and TNS, respectively. In contrast, the accuracy increased to 0.67, 0.66, and 0.42 for TNB, NBA, and TNS, respectively, when using WssGBLUP, specifically in the last three iterations. The results of weighted single-step genome-wide association studies (WssGWAS) showed that the highest variance explained for each trait was predominantly located in the Sus scrofa chromosome 5 (SSC5) region. Specifically, the variance exceeded 4% for TNB, 3% for NBA, and 6% for TNS. Within the SSC5 region (12.26 to 12.76 Mb), which exhibited the highest variance for TNB, 20 SNPs were identified, and five candidate genes were identified: TIMP3, SYN3, FBXO7, BPIFC, and RTCB. The identified SNP markers for TNB, NBA, and TNS were expected to provide valuable information for genetic improvement as an understanding of their expression and genetic architecture in Berkshire pigs. With the accumulation of more phenotype and SNP data in the future, it is anticipated that more effective SNP markers will be identified.

Research on Supply Chain Optimization and Management Based on Deep Reinforcement Learning

This research introduces a groundbreaking approach to supply chain optimization and management, termed as Deep Reinforcement Learning based Supply Chain Optimization and Management (DRL-SCOM). At the core of this approach is the utilization of advancements in Deep Reinforcement Learning (DRL), specifically through the integration of Randomized Ensembled Double Q-learning (REDQ) and Trust Region Policy Optimization (TRPO). DRL-SCOM is designed to effectively tackle the inherent complexities and dynamic challenges that are characteristic of supply chain management. One of the key strengths of DRL-SCOM lies in its use of REDQ, which plays a crucial role in mitigating the overestimation bias commonly associated with traditional Q-learning methods. This results in more accurate value estimation and policy improvement, a critical factor in the effective management of supply chains. Additionally, the integration of TRPO into the framework brings the advantage of safe and stable policy updates. Such stability is vital for maintaining the robustness required in the fluctuating environment of supply chain operations. The combination of REDQ and TRPO in DRL-SCOM creates a powerful synergy. REDQ’s ensembled learning approach, when fused with TRPO’s trust-region method, enables the framework to efficiently navigate the complex and high-dimensional decision space typical of supply chains. This allows for real-time optimization of decisions while staying within operational constraints. The DRL-SCOM methodology shows significant potential in addressing various aspects of supply chain management, from demand forecasting and inventory management to logistics, adeptly handling the nonlinearities and uncertainties that are prevalent in these areas. Thus, the DRL-SCOM framework emerges as an innovative solution, pushing the frontiers of traditional supply chain management. It paves the way for a more agile, responsive, and intelligent system, equipped to adapt to changing market demands and operational challenges. This approach represents a significant stride towards transforming supply chain management into a more advanced, data-driven, and adaptive field.

The Determination of Capitalization Rate by the Remote Segments Approach: The Case of an Agricultural Land Appraisal

In the absence of comparative real estate data in the market segment of the property to be estimated, the appraiser may resort to income capitalization to estimate the market value. Often, however, the choice of which rate to apply is affected by subjective and arbitrary assessments. The estimation result can therefore be inaccurate and rather unclear. However, the Remote Segments Approach (RSA), through appropriate adjustments on the original values, prices, and incomes detected in the remote segments, makes it possible to arrive at an appraisal result consistent with estimative logic and real estate valuation standards. The proposed application illustrates the estimation of the market value of a specialized fruit orchard of avocado, which is to be considered new in relation to other fruit species already present in the reference area. The adjustments required by the RSA are solved with the General Appraisal System (GAS), defining the difference matrix based on relevant characters common to all segments considered. The application is carried out by comparing the segment in which the orchard being estimated falls (subject) with other remote market segments in which prices and incomes constituted by other tree crops are collected. The market value of the subject is derived by making adjustments to the prices and incomes observed in the remote segments of comparison with a comparison function constructed through relevant characters common to the segments considered. The comparison function makes it possible to arrive at the determination of the capitalization rate to be used in estimating the value of the fruit orchard by income approach. While it is based on the comparison of segments, the approach followed allows for a value judgment consistent with the estimation comparison and capable of providing a solution less conditioned by the appraiser’s expertise in the presence of particularly pronounced limiting conditions.

PPAT: An effective scheme ensuring privacy-preserving, accuracy, and trust for worker selection in mobile crowdsensing networks

The data content privacy protection and data accuracy are two important research issues in Mobile Crowdsensing (MCS). However, current researches have rarely been able to satisfy privacy protection as well as data accuracy at the same time, thus hindering the development of MCS. To solve the above issues, for the first time, we have proposed a Privacy Preserving, Accuracy, and Trust data collection scheme (PPAT) for MCS, which can protect the privacy of data content and maintain high accuracy at low-cost style. In PPAT scheme, First, we proposed a scrambled data privacy protection framework which can protect the data of each worker from being known to any third party, which can protect the data privacy of workers. The second, more importantly, we propose a truth value estimation method based on trust computing, which can obtain the truth value more accurately compared to the classic methods under privacy-preserving. In the proposed trust-based truth value calculation, the worker's trust is determined by comparing it with the weight of the trusted worker. Then, the truth value is calculated by the trust of the workers, so that the truth value obtained is more accurate. Through theoretical analysis, it is proved that the proposed PPAT scheme has good worker data content, worker trust, and truth value content privacy protection. Through a large number of simulation experiments, the strategy proposed in this paper has a good ability to protect data content privacy compared to the previous strategy, while improving data quality by 0.5%∼5.7%, and reducing data collection costs by 35.6%∼54.9%.

Practical non-stationary extreme value analysis of peaks over threshold using the generalised Pareto distribution: Estimating uncertainties in return values

Choice of tuning parameters influences the performance of non-stationary extreme value modelling for peaks over threshold using the generalised Pareto (GP) distribution. We examine the effect of tuning parameter choice on maximum roughness-penalised likelihood estimation of GP models, the shape and scale parameters of which are assumed to vary smoothly on a one-dimensional “directional” covariate domain, under a B-spline representation. We examine the effect of (a) extreme value model parameterisation, (b) relative roughness penalty of GP parameters as a function of covariate, (c) cross-validation strategy for roughness parameter tuning, and (d) estimator for return value, on the estimation of return values corresponding to return periods 1000× the period of a sample of size 1000. Bootstrap resampling is used for thorough uncertainty quantification. We also compare results with those from stationary inference.Results from a large simulation study of 16 cases broadly representative of North Sea conditions for significant wave height with direction, indicate that (i) multiple two-group cross-validation yields lower return value estimates that ten-group cross-validation (leading to negative bias on average, for the case studies considered), (ii) the quantile of the bootstrap predictive estimator yields larger values than the mean over bootstraps of the quantile estimate (leading to reduced omni-directional bias for the case studies considered). Further, (iii) the use of stationary models for non-stationary tails is only reasonable when a high extreme value threshold is set for the stationary analysis. However, (iv) the relative performance of different modelling strategies is sensitive to the specific characteristics of the case study.