Uncertainty In Estimates Research Articles

From Envelope Spectra to Bearing Remaining Useful Life: An Intelligent Vibration-Based Prediction Model with Quantified Uncertainty

Bearings are pivotal components of rotating machines where any defects could propagate and trigger systematic failures. Once faults are detected, accurately predicting remaining useful life (RUL) is essential for optimizing predictive maintenance. Although data-driven methods demonstrate promising performance in direct RUL prediction, their robustness and practicability need further improvement regarding physical interpretation and uncertainty quantification. This work leverages variational neural networks to model bearing degradation behind envelope spectra. A convolutional variational autoencoder for regression (CVAER) is developed to probabilistically predict RUL distributions with confidence measures. Enhanced average envelope spectra (AES) are used as network input for its physical robustness in bearing condition assessment and fault detection. The use of the envelope spectrum ensures that it contains only bearing-related information by removing other rotor-related frequencies, hence it improves the RUL prediction. Unlike traditional variational autoencoders, the probabilistic regressor and latent generator are formulated to quantify uncertainty in RUL estimates and learn meaningful latent representations conditioned on specific RUL. Experimental validations are conducted on vibration data collected using multiple accelerometers whose natural frequencies cover bearing resonance ranges to ensure fault detection reliability. Beyond conventional bearing diagnosis, envelope spectra are extended for statistical RUL prediction integrating physical knowledge of actual defect conditions. Comparative and ablation studies are conducted against benchmark models to demonstrate their effectiveness.

Estimating trauma prevalence from incomplete human skeletal remains

AbstractTraumatic lesions on human skeletal remains are widely used for reconstructing past accidents or violent encounters and for comparing trauma prevalence across samples over time and space. However, uncertainties in trauma prevalence estimates increase proportionally with decreasing skeletal completeness, as once-present trauma might have gone missing. To account for this bias, samples are typically restricted to skeletal remains meeting a predefined minimum completeness threshold. However, the effect of this common practice on resulting estimates remains unexplored. Here, we test the performance of the conventional frequency approach, which considers only specimens with ≥ 75% completeness, against a recent alternative based on generalized linear models (GLMs), integrating specimen completeness as a covariate. Using a simulation framework grounded on empirical forensic, clinical, and archaeological data, we evaluated how closely frequency- and GLM-based estimates conform to the known trauma prevalence of once-complete cranial samples after introducing increasing levels of missing values. We show that GLM-based estimates were consistently more precise than frequencies across all levels of incompleteness and regardless of sample size. Unlike GLMs, frequencies increasingly produced incorrect relative patterns between samples and occasionally failed to produce estimates as incompleteness increased, particularly in smaller samples. Consequently, we generally recommend using GLMs and their extensions over frequencies, although neither approach is fully reliable when applied to largely incomplete samples.

Factors influencing the cost-effectiveness of radiofrequency ablation for Barrett's esophagus with low-grade dysplasia in Australia.

Endoscopic eradication therapy using radiofrequency ablation (RFA) is considered an acceptable alternative to surveillance monitoring for Barrett's esophagus with low-grade dysplasia (LGD). This study aimed to estimate whether RFA for LGD is cost-effective and to determine which factors influence cost-effectiveness. A Markov model was developed to estimate the incremental cost per quality-adjusted life year (QALY) gained for RFA compared with endoscopic surveillance. An Australian longitudinal cohort study (PROBE-NET) provides the basis of the model. Replacing surveillance with RFA yields 10 fewer cases of HGD and 9 fewer esophageal adenocarcinoma (EAC)-related deaths per 1000 patients' treatment, given on average 0.192 QALYs at an additional cost of AU$9211 (€5689; US$6262) per patient (incremental cost-effectiveness ratio AU$47,815 per QALY). The model is sensitive to the rate of EAC from LGD health state, the utility values, and the number of RFA sessions. Hence, the incremental benefit ranges from 0.080 QALYs to 0.198 QALYs leading to uncertainty in the cost-effectiveness estimates. When the cancerous progression rate of LGD falls <0.47% per annum, the cost-effectiveness of RFA becomes questionable. RFA treatment of LGD provides significantly better clinical outcomes than surveillance. The additional cost of RFA is acceptable if the LGD to EAC rate is >0.47% per annum and no more than three RFA treatment sessions are provided. Accurate estimates of the risk of developing EAC in patients with LGD are needed to validate the analyses.

Cost-effectiveness of an online supervised group physical and mental health rehabilitation programme for adults with post-COVID-19 condition after hospitalisation for COVID-19: the REGAIN RCT

BackgroundFollowing the COVID-19 pandemic, millions of people continue to experience ongoing physical and mental health sequelae after recovery from acute infection. There is currently no specific treatment for the diverse symptoms associated with post-COVID-19 condition. Physical and mental health rehabilitation may help improve quality of life in such patients. This study reports the cost-effectiveness of a programme of physical and mental health rehabilitation compared to best practice usual care in people with post-COVID-19 condition who were previously hospitalised.MethodsWe conducted an economic evaluation within a randomised controlled trial from the perspective of the UK national health service (NHS) and personnel social services perspective (PSS). Resource used and health-related quality of life were collected using bespoke questionnaire and the EQ-5D-5 L questionnaire at three, six, and 12 months. Incremental costs and quality adjusted life years accrued over the follow-up period were estimated and reported as the incremental cost-effectiveness ratio. Estimate uncertainty was managed by multiple imputation and bootstrapping cost-effectiveness estimates; and displayed graphically on the cost-effectiveness plane.ResultsOver a 12-month time horizon, incremental costs and QALYs were £305 (95% CI: -123 to 732) and 0.026 (95% CI: -0.005 to 0.052) respectively. The ICER was £11,941 per QALY indicating cost-effective care. Sensitivity analyses supported the base case findings. The probability of the intervention being cost-effective at a £30,000 per QALY willingness-to-pay threshold was 84%.ConclusionThe within-trial economic evaluation suggested that people with post-COVID-19 condition after hospitalisation should be offered a programme of physical and mental health rehabilitation as it likely reflects a cost-effective use of NHS resources. Hospitalisation for COVID-19 has become less commonplace: further evaluation in non-hospitalised patients may be worthwhile.Trial registrationISRCTN registry ISRCTN11466448 23rd November 2020.

Uncertainty in model estimates of global groundwater depth

Abstract Knowing the depth at which groundwater can be found below the land surface is critical for understanding its potential accessibility by ecosystems and society. Uncertainty in global scale water table depth (WTD) limits our ability to assess groundwater’s role in a water cycle altered by changing climate, land cover, and human water use. Global groundwater models offer a top–down pathway to gain this knowledge, but their uncertainty is currently poorly quantified. Here, we investigate four global groundwater models and reveal steady-state WTD disagreements of more than 100 m for one-third of the global land area. We find that model estimates of land areas with shallow groundwater at &lt;10 m depth vary from 10% to 71% (mean of 23%). This uncertainty directly translates into subsequent assessments, as land areas with potential groundwater accessibility for forests, population, and areas equipped for irrigation, differ substantially depending on the chosen model. We explore reasons for these differences and find that contrary to observations, 3 out of 4 models show deeper water tables in humid than in arid climates and greatly overestimate how strongly topographic slope controls WTD. These results highlight substantial uncertainty associated with any global-scale groundwater analysis, which should be considered and ultimately reduced.

Algorithm XXXX: The Delaunay Density Diagnostic

Accurate approximation of a real-valued function depends on two aspects of the available data: the density of inputs within the domain of interest and the variation of the outputs over that domain. There are few methods for assessing whether the density of inputs is sufficient to identify the relevant variations in outputs—i.e., the “geometric scale” of the function—despite the fact that sampling density is closely tied to the success or failure of an approximation method. In this paper, we introduce a general purpose, computational approach to detecting the geometric scale of real-valued functions over a fixed domain using a deterministic interpolation technique from computational geometry. The algorithm is intended to work on scalar data in moderate dimensions (2-10). Our algorithm is based on the observation that a sequence of piecewise linear interpolants will converge to a continuous function at a quadratic rate (in \(L^{2}\) norm) if and only if the data are sampled densely enough to distinguish the feature from noise (assuming sufficiently regular sampling). We present numerical experiments demonstrating how our method can identify feature scale, estimate uncertainty in feature scale, and assess the sampling density for fixed (i.e. static) datasets of input–output pairs. We include analytical results in support of our numerical findings and have released lightweight code that can be adapted for use in a variety of data science settings.

Modeling the population-level impact of a third dose of MMR vaccine on a mumps outbreak at the University of Iowa

Mumps outbreaks among fully vaccinated young adults have raised questions about potential waning of immunity over time and need for a third dose of the measles, mumps, rubella (MMR) vaccine. However, there are currently limited data on real-life effectiveness of the third-dose MMR vaccine in preventing mumps. Here, we used a deterministic compartmental model to infer the effectiveness of the third-dose MMR vaccine in preventing mumps cases by analyzing the mumps outbreak that occurred at the University of Iowa between August 24, 2015, and May 13, 2016. The modeling approach further allowed us to evaluate the population-level impact of vaccination by different timing in relation to the start of the outbreak and varied coverage levels, and to account for potential sources of bias in estimating vaccine effectiveness. We found large uncertainty in vaccine effectiveness estimates; however, our models showed that early introduction of a third dose of MMR vaccine during a mumps outbreak can be effective in preventing transmission. School holidays, such as the winter break, likely played important roles in preventing mumps transmission.

Large language model uncertainty proxies: discrimination and calibration for medical diagnosis and treatment.

The inability of large language models (LLMs) to communicate uncertainty is a significant barrier to their use in medicine. Before LLMs can be integrated into patient care, the field must assess methods to estimate uncertainty in ways that are useful to physician-users. Evaluate the ability for uncertainty proxies to quantify LLM confidence when performing diagnosis and treatment selection tasks by assessing the properties of discrimination and calibration. We examined confidence elicitation (CE), token-level probability (TLP), and sample consistency (SC) proxies across GPT3.5, GPT4, Llama2, and Llama3. Uncertainty proxies were evaluated against 3 datasets of open-ended patient scenarios. SC discrimination outperformed TLP and CE methods. SC by sentence embedding achieved the highest discriminative performance (ROC AUC 0.68-0.79), yet with poor calibration. SC by GPT annotation achieved the second-best discrimination (ROC AUC 0.66-0.74) with accurate calibration. Verbalized confidence (CE) was found to consistently overestimate model confidence. SC is the most effective method for estimating LLM uncertainty of the proxies evaluated. SC by sentence embedding can effectively estimate uncertainty if the user has a set of reference cases with which to re-calibrate their results, while SC by GPT annotation is the more effective method if the user does not have reference cases and requires accurate raw calibration. Our results confirm LLMs are consistently over-confident when verbalizing their confidence (CE).

Risk of death from hematological malignancies in the South Urals Population Exposed to Radiation cohort

The results of studies of long-term effects of population exposure in the South Urals in the 1950s were based for a long time on studies of the effects of exposure in the Techa River cohort and later in the East Urals Radioactive Trace cohort. After the creation of the South Urals Population Exposed to Radiation cohort, combining all persons exposed in the South Urals in the period from January 1, 1950, to December 31, 1960 the size of the cohort doubled, follow-up period reached 71 years, and the number of person-years at risk increased to 1,964,333. The average dose to red bone marrow for all cohort members was 231mGy. Regression analysis using a simple parametric excess relative risk model was performed using the EPICURE statistical package. The analysis resulted in confirmation of a statistically significant (p &lt;0.001) linear doseresponse relationship of mortality for all hemoblastoses, and leukemia. The excess relative risk and 95% confidence intervals of death at 2-year latency period from all hemoblastoses were 0.71/Gy (0.28;1.31); from all leukemias - 1.28/Gy (0.55-2.39) and from leukemias excluding chronic lymphocytic leukemia -1.52/Gy (0.64-2.94). The quadratic model also significantly described the dose dependence; (the differences between the models were not significant). Increasing the follow-up period to 71 years resulted in a slight decrease in risk estimates. The width of the confidence intervals of the risk estimates decreased by more than a factor of 3 compared to earlier studies in the Techa River cohort, which indicates a decrease in uncertainties of risk estimates. Increasing the size of the cohort made it possible to obtain significant risk values for individual population groups (by sex, age, etc.). Analysis of risk values modification did not reveal significant differences in dose dependence on the studied factors, including attained age and age at exposure. The study will be continued to investigate in more detail the influence of time-dependent factors on dose dependence, as well as to assess the risk of death from different cellular forms of leukemia.

In-situ online investigation of biogenic volatile organic compounds emissions from tropical rainforests in Hainan, China

Biogenic volatile organic compounds (BVOCs) emitted by tropical plants represent a significant proportion of global emissions, but the in-situ BVOC measurements in tropical rainforests are extremely sparse. Herein, a vehicle-mounted mobile monitoring system was developed for in-situ online investigations of BVOC emissions from thirty representative tree species in the tropical rainforests of Hainan Island, southern China. The results showed that monoterpenes were the primary BVOCs emitted from most broadleaf trees. Isoprene, sabinene, γ-terpinene and β-ocimene were the most abundant BVOCs. Localized canopy-scale emission factors (EFs) exhibited notable discrepancies with the defaults in the Model of Emissions of Gases and Aerosols from Nature (MEGAN), specifically with isoprene EFs being slightly lower than the model, while the EFs for monoterpenes and sesquiterpenes were 1 to 27 times higher than those in MEGAN. The BVOC emission inventory for the predominant mature forest species in Hainan Island in 2023 was further estimated to be 244.43 Gg C·yr−1, with isoprene and monoterpenes accounting for 74 % and 16 %, respectively. Additionally, unimodal monthly variation patterns were revealed, with BVOC emissions peaked in July (30.08 Gg C·yr−1) and bottomed in January (8.84 Gg C·yr−1). This study demonstrates the potential and versatility of the applied mobile platform for in-situ online measurements of plant volatiles. Our findings provide important data support for reducing uncertainties in BVOC emission estimations in tropical rainforests and for evaluating their health benefits in the context of forest therapy.

The ozone radiative forcing of nitrogen oxide emissions from aviation can be estimated using a probabilistic approach

Reliable prediction of aviation’s environmental impact, including the effect of nitrogen oxides on ozone, is vital for effective mitigation against its contribution to global warming. Estimating this climate impact however, in terms of the short-term ozone instantaneous radiative forcing, requires computationally-expensive chemistry-climate model simulations that limit practical applications such as climate-optimised planning. Existing surrogates neglect the large uncertainties in their predictions due to unknown environmental conditions and missing features. Relative to these surrogates, we propose a high-accuracy probabilistic surrogate that not only provides mean predictions but also quantifies heteroscedastic uncertainties in climate impact estimates. Our model is trained on one of the most comprehensive chemistry-climate model datasets for aviation-induced nitrogen oxide impacts on ozone. Leveraging feature selection techniques, we identify essential predictors that are readily available from weather forecasts to facilitate the implementation therein. We show that our surrogate model is more accurate than homoscedastic models and easily outperforms existing linear surrogates. We then predict the climate impact of a frequently-flown flight in the European Union, and discuss limitations of our approach.

Efficient approximate Bayesian inference for quantifying uncertainty in multiscale animal movement models

It is becoming increasingly important for wildlife managers and conservation ecologists to understand which resources are selected or avoided by an animal and how to best predict future spatial distributions of animal populations in the long term. However, inferring the patterns of space use by animals is a challenging multiscale inference problem, and formal uncertainty quantification of parameter estimates is an essential component of models that provide useful predictions across scales. In this study, we develop an approximate Bayesian inference framework for step selection models of animal movement which quantifies the uncertainty in estimates of resource selection and avoidance parameters within the Bayesian paradigm. The framework allows joint inference of movement and resource selection parameters of animals and is multiscale in that parameters inferred from fine scale movement steps scale to produce predictions of long-term patterns of space use. Our analysis focuses on simulated movement data in which we test the performance of our framework by altering movement parameters in the data-generating process. In our simulations, individuals respond to two environmental covariates and we employ all combinations of positive and negative selection coefficients corresponding to attraction to an environmental feature and avoidance of an environmental feature, respectively. In all scenarios, we recover the movement parameters used for the simulation of synthetic movement data using variational inference, an approximate Bayesian method, allowing us to formally quantify the uncertainty associated with each parameter for varying data set sizes. Our framework successfully recovered all combinations of movement parameters of the simulated data and accurately captured their posterior distributions given the available data suggesting that the framework is reliable and suitable for inferring how animals select resources and move on a landscape.Notably, our analysis shows that even for reasonably large data sets (circa 10,000 observations) there can still be considerable uncertainty associated with resource selection parameters which can in turn lead to inaccurate predictions of long term space use if not properly incorporated into the modelling approach. To further illustrate the utility of our approach, we also present a case study of its application to an example data set consisting of GPS locations of a fisher (Martes pennanti) Our approach will be of interest to ecologists looking to address conservation questions such as when and where animals are likely to spend most of their time. Furthermore, the approach could be used to predict new suitable areas for conservation based on how GPS collared animal use or avoid resources while including uncertainty around the predictions, thereby helping to make informed management decisions.

Dose Reconstruction for Epidemiological Studies among Ukrainian Chernobyl Cleanup Workers.

The present paper provides an overview of the methods and summarizes the results of estimating radiation doses and their uncertainties for Ukrainian-American epidemiological studies among the Chernobyl (Chornobyl) cleanup workers. After the Chernobyl accident occurred on April 26, 1986, more than 300,000 Ukrainian cleanup workers took part between 1986 and 1990 in decontamination and recovery activities at the site of the Chernobyl Nuclear Power Plant. The U.S. National Cancer Institute in collaboration with the Ukrainian National Research Center for Radiation Medicine conducted several epidemiological studies in this population. An important part of these studies was the reconstruction of the study participants' radiation doses and the assessment of uncertainties in doses. A method called realistic analytical dose reconstruction with uncertainty estimation (RADRUE) was used to calculate the doses from external irradiation during cleanup missions, which was the main exposure pathway for most study participants. At the initial phase of the accident during the atmospheric releases of radioactivity from the destroyed reactor, the cleanup workers also received doses from inhalation of radionuclides. In addition, study participants received doses at their places of residence, especially those who lived in highly contaminated areas. The radiation doses estimated for 2,048 male cleanup workers included in the Ukrainian-American epidemiological studies varied widely: (i) bone-marrow doses from external irradiation in the case-control study of leukemia of 1,000 cleanup workers ranged from 3.7 × 10-5 mGy to 3.3 Gy (mean = 92 mGy); (ii) thyroid doses in the case-control study of thyroid cancer in 607 persons from all exposure pathways combined were from 0.15 mGy to 9.0 Gy (mean = 199 mGy); (iii) gonadal doses in 183 cleanup workers from all exposure pathways combined in the study of germline mutations in the offspring after parental irradiation (trio study) ranged from 0.58 mGy to 4.1 Gy (mean = 392 mGy); (iv) thyroid doses in the human factor uncertainties study among 47 persons were from 20 mGy to 2.1 Gy (mean = 295 mGy); and (v) lung doses in the study of germline genetic variants associated with host susceptibility to COVID-19 estimated for 211 cleanup workers were from 0.024 mGy to 2.5 Gy (mean = 249 mGy). Doses of female cleanup workers were much lower than those of male cleanup workers: the mean doses for female cleanup workers were 27 mGy for 34 women included in the trio study and 56 mGy for 48 women participated in the study of germline genetic variants associated with host susceptibility to COVID-19. Uncertainties in dose estimates included two components: (i) inherent uncertainties arising from the stochastic random variability of the parameters used in exposure assessment and from a lack of knowledge about the true values of the parameters; and (ii) human factor uncertainties due to poor memory recall resulting in incomplete, inaccurate, or missing responses during personal interviews with cleanup workers conducted long after exposure. This paper also discusses possible developments and improvements in the methods to assess the radiation doses and associated uncertainties for cleanup workers.

Detection and Coil Embolisation of Pelvic Vein Incompetence for Chronic Pain in the Female Pelvis: A Development Phase Economic Analysis

Unexplained chronic pain in the female pelvis (CPFP) affects 7% of people indicating female sex in the UK. Evidence suggests that pelvic venous incompetence (PVI) could explain CPFP and that coil embolisation could provide relief. The aims of this study were to indicate (1) the cost effectiveness of detecting and treating PVI in people experiencing unexplained CPFP, and (2) the maximum value of further research, suggesting suitable areas. A decision tree compared standard care (regular prescribed analgesia) with an intervention comprising transvaginal duplex ultrasound to screen for PVI, venography to confirm the diagnosis, and coil embolisation treatment. The population was people experiencing unexplained CPFP. A UK National Health Service perspective and 2021 - 22 price year were used. Ten years of health costs and health related quality of life (HRQoL) effects for eligible 40 year olds were simulated. Evidence reviews informed diagnostic accuracy, health service usage, and unit costs. A single centre randomised controlled trial informed all other parameters. Probabilistic analysis incorporated parameter uncertainty in cost effectiveness estimates. Deterministic sensitivity analysis indicated drivers of uncertainty. Value of information methods measured the value of eliminating all relevant uncertainties, given uptake predictions. The main outcome measures were incremental cost and quality adjusted life years (QALYs) for the intervention compared with analgesia, the incremental cost effectiveness ratio (ICER), and expected value of perfect information. The mean ICER for the intervention was £4 558/QALY gained, and the probability that the ICER was within the UK cost effectiveness threshold (£20 000/QALY gained) was 90%. The expected value of perfect information about all model parameters was £46 M. All deterministic sensitivity analysis scenarios met the threshold, except the smallest plausible HRQoL effect of (resolving) CPFP. Detecting and treating PVI causing CPFP appears cost effective, but more primary research would be valuable to reduce decision uncertainty. Uncertainty in the HRQoL estimate for unexplained CPFP appeared to contribute most to decision uncertainty.

Vine copula based structural equation models

Gaussian linear structural equation models (SEMs) are often used as a statistical model associated with a directed acyclic graph (DAG) also known as a Bayesian network. However, such a model might not be able to represent the non-Gaussian dependence present in some data sets resulting in nonlinear, non-additive and non Gaussian conditional distributions. Therefore the use of the class of D-vine copula based regression models for the specification of the conditional distribution of a node given its parents is proposed. This class extends the class of standard linear regression models considerably. The approach also allows to create an importance order of the parents of each node and gives the potential to remove edges from the starting DAG not supported by the data. Further uncertainty of conditional estimates can be assessed and fast generative simulation using the D-vine copula based SEM is available. The improvement over a Gaussian linear SEM is shown using random specifications of the D-vine based SEM as well as its ability to correctly remove edges not present in the data generation using simulation. An engineering application showcases the usefulness of the proposals.

Climate‐Induced Uncertainty in Modeling Gross Primary Productivity From the Light Use Efficiency Approach

AbstractLight use efficiency (LUE) models, along with satellite‐based vegetation maps and climatic reanalysis products as drivers, are effective tools for estimating large‐scale gross primary productivity (GPP). However, the climate‐induced uncertainty in LUE‐based GPP remains poorly understood, particularly the temporal scaling uncertainty from ignored climatic fluctuations. Here, two 1‐hourly reanalysis products, along with site‐based observations, were employed to drive a two‐leaf LUE model at 194 eddy‐covariance (EC) sites. The observation‐driven and reanalysis‐driven GPP at the 1‐hourly resolution were evaluated against EC GPP to illustrate the uncertainty from reanalysis products, with mean absolute deviation (MAD) and Nash‐Sutcliffe efficiency (NSE) as criterion. Moreover, the climate‐induced temporal scaling uncertainty was characterized by comparing distributed GPP (modeled with 1‐hourly resolution climatic drivers) and lumped GPP (modeled with 6‐hourly resolution climatic drivers). At the 1‐hourly resolution, results demonstrated that the reanalysis‐driven GPP showed a weaker relationship with EC GPP (MAD = 0.14 gC m−2h−1, NSE = 0.48) than the observation‐driven GPP (MAD = 0.12 gC m−2h−1, NSE = 0.60), confirming the nonnegligible climate‐induced uncertainty from reanalysis products. Additionally, the climate‐induced uncertainty arising from gridded radiation was found to be significantly larger than that associated with temperature and vapor pressure deficit (VPD). At the 6‐hourly resolution, both the observation‐driven and reanalysis‐driven lumped GPP exhibited a lower relationship with EC GPP (MAD = 0.63 gC m−26h−1, NSE = 0.54) than the corresponding distributed GPP (MAD = 0.57 gC m−26h−1, NSE = 0.59), demonstrating that the climate‐induced temporal scaling uncertainty in 6‐hourly GPP estimates was significantly apparent. This study emphasizes the imperative to refine reanalysis products for more precise capture of short‐term fluctuations and to reduce scaling uncertainties in coarse temporal resolution GPP estimates.

Exponential or Unimodal Relationships Between Nighttime Ecosystem Respiration and Temperature at the Eddy Covariance Flux Tower Sites.

Ecosystem respiration is a key flux in the terrestrial carbon cycle and is affected substantially by temperature. This work analysed the time series data of nighttime net ecosystem exchange of carbon dioxide (NEEnight) from 196 FLUXNET2015 sites to re-evaluate the relationships between NEEnight and temperature. A total of 93 sites (48%) were identified to have a unimodal relationship between NEEnight and temperature. Site-specific apparent optimum temperature parameters were then estimated at these sites. We further assessed the impacts of using exponential or unimodal equations on NEEnight predictions. The predicted NEEnight values at high temperatures were substantially higher from the exponential-type equations (mean: ~200%) than from the unimodal equation (mean: ~30%), compared to the observed NEEnight. This study calls for using a unimodal equation to predict NEEnight (often considered as nighttime ecosystem respiration, ERnight), which could substantially improve the accuracy and reduce uncertainty in ER estimates, in particular under the scenario of global warming.

Distribution-free prediction intervals with conformal prediction for acoustical estimation.

Acoustical parameter estimation is a routine task in many domains. The performance of existing estimation methods is affected by external uncertainty, yet the methods provide no measure of confidence in the estimates. Hence, it is crucial to quantify estimate uncertainty before real-world deployment. Conformal prediction (CP) generates statistically valid prediction intervals for any estimation model using calibration data; a limitation is that calibration data needed by CP must come from the same distribution as the test-time data. In this work, we propose to use CP to obtain statistically valid uncertainty intervals for acoustical parameter estimation using a data-driven model or an analytical model without training data. We consider direction-of-arrival estimation and localization of sources. The performance is validated on plane wave data with different sources of uncertainty, including ambient noise, interference, and sensor location uncertainty. The application of CP for data-driven and traditional propagation models is demonstrated. Results show that CP can be used for statistically valid uncertainty quantification with proper calibration data.

A machine learning paradigm for necessary observations to reduce uncertainties in aerosol climate forcing.

Uncertainties in estimates of climate cooling by anthropogenic aerosols have not decreased significantly in the last two decades, partly because observational constraints on crucial aerosol properties simulated in Earth System Models are insufficient. To help address this insufficiency in aerosol observations, we describe a paradigm for deriving higher-level aerosol properties with machine learning algorithms that use only lidar observations and reanalysis data as predictors. Our paradigm employs high-accuracy suborbital lidar and collocated in situ measurements to train and test two fully-connected neural network algorithms. We use two lidar data sets as input to our machine learning algorithms. The first data set consists of suborbital lidar observations not previously used in the training of the machine learning algorithms. The second data set consists of simulated UV-only observations to preview the algorithms' predictive capabilities in anticipation of data from the ATmospheric LIDar system on the EarthCARE satellite, which was launched in May 2024. Here we show that our algorithms predict two crucial aerosol properties, aerosol light absorption and cloud condensation nuclei concentrations with unprecedented accuracy, yielding mean relative errors of 21% and 13%, respectively, when suborbital lidar data are used as predictors. These errors represent significant improvements over conventional aerosol retrievals. Applied to future satellite missions, the paradigm presented here has great potential for constraining Earth System Models and reducing uncertainties in their estimates of aerosol climate forcing and future global warming.

Harmonizing Campylobacter risk assessments across European countries – can the pooled process hygiene criteria data be used in the Danish risk assessment model?

This study investigated the possibility of harmonizing quantitative microbiological risk assessment (QMRA) for Campylobacter spp. across European (EU) countries. French Campylobacter data (2020–2021) from neck skin (NS) pools, sampled at slaughterhouses under the European surveillance component for Food Business Operators (FBOs), were adapted to inform a QMRA model that, among others, has been used within the Danish Action Plan against Campylobacter, on the basis of single leg skins (LS) data. Datasets included culture results (in colony forming unit per gram, CFU/g) from 1,284 broiler flocks slaughtered at 13 slaughterhouses representing broiler production in western France. Five pools (of 2–4 NS samples each) per flock were tested. One pool per tested flock was randomly chosen for the analysis. After conducting descriptive statistics (on flock prevalence and meat contaminations across months and years), three contamination transformation factors (CTFs) were estimated to translate NS pools contaminations into single LS contamination, based on data from French and Danish studies. A reference simulation scenario (ScenRef) was set with CTF = 3.2 (i.e. NS pool concentration divided by 3.2); while other 13 scenarios represented an alternative scenario analysis to investigate the impact of: the CTF value (ScenMin with CTF = 2 and ScenMax with CTF = 10), censored test results (ScenUncens) and random choice of pool per flock (ScenSampling-1 to 10), on the risk estimates. The average monthly/annual risk of human disease per poultry meal and the monthly/annual relative risk (RR) of 2021 compared to 2020, were estimated. In ScenRef, the annual RR was 1.22, suggesting an increase of risk of ≈ 22 % in 2021 compared to 2020. The impact of CTFs, censored data and randomized pool sampling per flock, on the annual and (most) monthly RRs, appeared limited. This study gives an overview of the strengths and limitations to be considered for adapting the French FBO data into the Danish model and to harmonize risk assessments across EU countries, accordingly. To reduce uncertainty in risk estimates, it could be considered increasing representativeness of NS tested flock populations and/or using LS rather than NS samples; because LS samples are more representative of actually retailed meat contaminations. If NS pools are maintained, the relationships between concentrations on NS pools and those on consumed meat requires further investigation.