The energy flux of three-dimensional waves in the atmosphere: Exact expression for a basic model diagnosis with no equatorial gap

A model diagnosis for the energy flux of off-equatorial Rossby waves in the atmosphere has previously been done using quasi-geostrophic equations and is singular at the equator. The energy flux of equatorial waves has been separately investigated in previous studies using a space-time spectral analysis or a ray theory. A recent analytical study has derived an exact universal expression for the energy flux which can indicate the direction of the group velocity for linear shallow water waves at all latitudes. This analytical result is extended in the present study to a height-dependent framework for three-dimensional waves in the atmosphere. This is achieved by investigating the classical analytical solution of both equatorial and off-equatorial waves in a Boussinesq fluid. For the horizontal component of the energy flux, the same expression has been obtained between equatorial waves and off-equatorial waves in the height-dependent framework, which is linked to a scalar quantity inverted from the isentropic perturbation of Ertel’s potential vorticity. The expression of the vertical component of the energy flux requires computation of another scalar quantity that may be obtained from the meridional integral of geopotential anomaly in a wavenumber-frequency space. The exact version of the universal expression is explored and illustrated for three-dimensional waves induced by an idealized Madden-Julian Oscillation forcing in a basic model experiment. The zonal and vertical fluxes manifest the energy transfer of both equatorial Kelvin waves and off-equatorial Rossby waves with a smooth transition at around 10°S and around 10°N. The meridional flux of wave energy represents connection between off-equatorial divergence regions and equatorial convergence regions.

Explainable machine learning: A case study on impedance tube measurements

Machine learning (ML) techniques allow for finding hidden patterns and signatures in data. Currently, these methods are gaining increased interest in engineering in general and in vibroacoustics in particular. Although ML methods are successfully applied, it is hardly understood how these black box-type methods make their decisions. Explainable machine learning aims at overcoming this issue by deepening the understanding of the decision-making process through perturbation-based model diagnosis. This paper introduces machine learning methods and reviews recent techniques for explainability and interpretability. These methods are exemplified on sound absorption coefficient spectra of one sound absorbing foam material measured in an impedance tube. Variances of the absorption coefficient measurements as a function of the specimen thickness and the operator are modeled by univariate and multivariate machine learning models. In order to identify the driving patterns, i.e. how and in which frequency regime the measurements are affected by the setup specifications, Shapley additive explanations are derived for the ML models. It is demonstrated how explaining machine learning models can be used to discover and express complicated relations in experimental data, thereby paving the way to novel knowledge discovery strategies in evidence-based modeling.

35 Comparative Performance of Prediction Model, Non-Expert and Tele-Diagnosis of Common External and Middle Ear Disease Using A Patient Cohort from Cambodia That Included One Hundred and Thirty-Eight Ears

Introduction Globally 6% of the population suffers from disabling hearing loss and the majority resides in low- and middle-income countries, but diagnosis and treatment are hampered by poor availability of expert diagnosis. We compared the utility of tele-diagnosis, non-expert diagnosis, and prediction model diagnosis as a screening tool for common external and middle ear disorders. Method We recruited consecutive adult and paediatric patients presenting with ear or hearing symptoms to ENT outpatients at Children’s Surgical Centre, Cambodia. Each participant underwent sequential symptomatic and otoscopic assessment by a non-specialist and an ENT specialist. The non-specialist captured data using a novel automated symptom questionnaire loaded onto a smartphone otoscope. An ENT specialist in the UK subsequently reviewed these data. Results 138 ears were recruited. The prediction model performed poorly, but absence of otorrhoea was found to reliably exclude a diagnosis of chronic suppurative otitis media (negative predictive value=0.99). Both on-site non-expert and expert tele-diagnosis had high diagnostic specificity (90-99% and 86-99%), but low sensitivity (<43% and 32-100%). Conclusions We report the first study to directly compare approaches for non-specialist diagnosis of disorders of the middle/external ear, which shows suboptimal but comparable performance using an automated questionnaire, on site non-expert diagnosis, or remote expert diagnosis

Diagnosis Failure Cause of complex Pharmaceutical System by Bayes Learning for Decision Support

This work proposes a real application of diagnosis protocol for complex pharmaceutical process drifts. Main challenge is to identify and classify failure causes of production process. The model which we have proposed is structured in the causal graph form, named “Hierarchical Naïve Bayes” (HNB) formalism. Our contribution is the presentation of a methodology that allows developing flexibility in particular complex pharmaceutical production context. A data extraction and processing prototype is performed in this paper from real pharmacy company to build Bayesian model. Diagnosis results are decision support elements that built based on HNB probabilities. Furthermore, this work can be applied in order to improve production quality in businesses competition.

Transport model diagnosis of the mean age of air derived from stratospheric samples in the tropics

Stratospheric profiles of the mean age of air estimated from cryogenic air samples acquired during the CUBE/Biak field campaign over Indonesia are investigated with the aid of an atmospheric chemistry transport model nudged to ERA-Interim meteorological fields. Application of the boundary impulse response (BIR) method and Lagrangian backward trajectories to the transport field simulated by a single model prove useful in interpreting the observational results, which include discrepancies between CO2- and SF6-derived mean ages. This may be because the BIR method takes unresolved diffusive processes into account while the Lagrangian method distinguishes the pathways the air parcels have taken before reaching the sample site. The capability to estimate the vertical profiles of the clock tracer concentrations and the water vapor “tape recorder” is another advantage of the Lagrangian method, confirming the reality of the trajectory calculations. The profile of CO2-mean age is reproduced reasonably well by trajectory-derived mean age, while BIR-derived mean age is much greater than CO2 age at 28 and 29 km, possibly due to high diffusivity in the transport model. On the other hand, SF6 age is reproducible only in the lower stratosphere, but far exceeds the trajectory-derived mean age above 25 km. As air parcels of mesospheric origin are missing in the Lagrangian age estimation, this discrepancy, together with the fact that the observed SF6 concentrations are much lower than the trajectory-derived values in this height region, is consistent with the idea that the stratospheric air samples are mixed with SF6-depleted mesospheric air, leading to overestimation of the mean age.