Homogeneous Dataset Research Articles

Gathering new knowledge from existing monitoring dataset of Campania marine coastal area (Southern Italy).

In this study we propose, a multi-step strategy of selection and characterization of long-term dataset of contaminant concentrations in different environmental matrices (i.e., water and sediment). Starting from a high quality and homogeneous dataset of chemical parameters, a selection of a usable refined dataset followed by statistical characterization and hazard assessment was performed. The database of chemical contamination data from monitoring activities in the coastal marine water area of the Campania Region (Italy) produced by the Regional Agency for Environmental Protection of Campania (ARPAC) between 2013 and 2019 was utilized. Descriptive and multivariate statistics were applied to the extracted data subset to describe spatial variability and to investigate the relationships between matrices and contaminants. In addition, the impact on the sediment matrix was considered using the contamination factor (Cf), the sediment quality guidelines (SQG), and the probability of exceedance (PoE) of given thresholds for metals and organic compounds. The results highlighted the main anthropogenic pressures between the Gulf of Pozzuoli and the Gulf of Napoli, and the potential hazard posed in particular by metals (i.e., Pb, C, and Hg), TBT, and PAHs. A wide range of As concentration along the Campania coastline and Ni occurrence in the southern part of the region, mainly attributable to geogenic origin, was also evidenced. This approach allows extracting new knowledge from large dataset systematically collected by ARPAC monitoring activity, to support possible actions of contamination control and mitigation.

Homogenization of Norwegian monthly precipitation series for the period 1961–2018

Abstract. The primary goal of the analysis was to establish a high-quality precipitation reference dataset, which is both consistent and homogeneous, for calculation of the new standard climate normals (1991–2020). Climatol homogenization method was applied to detect inhomogeneities in 325 Norwegian precipitation series, during the period 1961–2018. Results from homogeneity testing found inhomogeneities in 29 % of the 325 series, however, only 25 % were classified as inhomogeneous after conferring with metadata and therefore adjusted. Relocation of the precipitation gauge and automation were the main causes of all the inhomogeneities in the Norwegian series, explaining 71 % and 12 % respectively of all detected breaks. Results further showed benefits of incorporating metadata to the automatically detected inhomogeneities. Linear trend analysis showed increasing trends in the period 1961–2018 except in autumn where a decreasing trend was observed. The homogeneity analysis produced a 58-year long homogenous dataset for 325 monthly precipitation sums with regional temporal variability and spatial coherence that is better than that of non-homogenized series. The dataset is more reliable in explaining the large-scale climate variations and was used to calculate the new climate normals in Norway.

The Ariel Target List: The Impact of TESS and the Potential for Characterizing Multiple Planets within a System

The ESA Ariel mission has been adopted for launch in 2029 and will conduct a survey of around 1000 exoplanetary atmospheres during its primary mission life. By providing homogeneous data sets with a high signal-to-noise ratio and wide wavelength coverage, Ariel will unveil the atmospheric demographics of these faraway worlds, helping to constrain planet formation and evolution processes on a galactic scale. Ariel seeks to undertake a statistical survey of a diverse population of planets; therefore, the sample of planets from which this selection can be made is of the utmost importance. While many suitable targets have already been found, hundreds more will be discovered before the mission is operational. Previous studies have used predictions of exoplanet detections to forecast the available planet population by the launch date of Ariel, with the most recent noting that the Transiting Exoplanet Survey Satellite (TESS) alone should provide over 1000 potential targets. In this work, we consider the planet candidates found to date by TESS to show that, with the addition of already confirmed planets, Ariel will already have a more than sufficient sample to choose its target list from once these candidates are validated. We showcase the breadth of this population, as well as exploring, for the first time, the ability of Ariel to characterize multiple planets within a single system. Comparative planetology of worlds orbiting the same star, as well as across the wider population, will undoubtedly revolutionize our understanding of planet formation and evolution.

Ensemble Graph Attention Networks

Graph neural networks have demonstrated its success in many applications on graph-structured data. Many efforts have been devoted to elaborating new network architectures and learning algorithms over the past decade. The exploration of applying ensemble learning techniques to enhance existing graph algorithms have been overlooked. In this work, we propose a simple generic bagging-based ensemble learning strategy which is applicable to any backbone graph models. We then propose two ensemble graph neural network models – Ensemble-GAT and Ensemble-HetGAT by applying the ensemble strategy to the graph attention network (GAT), and a heterogeneous graph attention network (HetGAT). We demonstrate the effectiveness of the proposed ensemble strategy on GAT and HetGAT through comprehensive experiments with four real-world homogeneous graph datasets and three real-world heterogeneous graph datasets on node classification tasks. The proposed Ensemble-GAT and Ensemble-HetGAT outperform the state-of-the-art graph neural network and heterogeneous graph neural network models on most of the benchmark datasets. The proposed ensemble strategy also alleviates the over-smoothing problem in GAT and HetGAT.

Role of radiomics in predicting immunotherapy response.

SummaryImmunotherapies have revolutionised cancer management. Despite their success, durable responses are limited to a subset of patients. Prediction of immunotherapy response in patients has proven to be difficult due to a lack of robust biomarkers. Routinely collected imaging may offer an additional information source to personalise patient treatment, with advantages over tissue‐based biomarkers. Quantitative image analysis or radiomics, which involves the high‐throughput extraction of imaging features, has the potential to non‐invasively predict cancer histology, outcomes and prognosis. This review evaluates the value of radiomics in patients undergoing immunotherapy, with a summary provided of the performance of radiomics models in predicting immunotherapy response and toxicity, as well as immune correlates. Much of the literature focussed on clinical endpoints and correlates to tissue biomarkers, particularly in lung cancer, while few studies investigated association with immune‐related adverse events. Strengths of the studies included more frequent use of clinical trial datasets, homogenous patient cohorts and high‐quality diagnostic scans. Limitations of the studies include heterogeneity in study methodology, lack of well‐defined homogenous imaging datasets, limited open publishing of imaging datasets, coding and parameters used for radiomics signature development and limited use of external validation datasets. Future research should address the above limitations, as well as further explore the relationship between radiomics and immune‐related adverse effects and less well‐studied biological correlates such tumour mutational burden, and incorporate known clinical prognostic scores into radiomics models.

The financial and green effects of cultural values on mission drifts in European social enterprises

AbstractThe sustainable social enterprises (SEs) literature shows that SEs have to simultaneously pursue economic, social, and environmental aims. However, tensions between these objectives can make this a challenging task and lead towards mission drift. This work investigates if the cultural dimension can forecast the mission drift. We empirically analyze this relationship in three stages. In the first stage, we identify a homogeneous dataset of 287 sustainable SEs from seven EU countries from 2011 to 2020. Then, in the second stage, we apply the data envelopment analysis (DEA) methodology to calculate the efficiency of the SEs. An efficient SE has to simultaneously achieve social, environmental, and economic aims. We calculate a proxy of the mission drift and generate a dichotomous category variable that assigns value 1 to the SE not affected by mission drift, 0 otherwise. In the last stage, we implement the 2SLS logistic regression between the variable that identifies the SEs affected by mission drift and three cultural dimensions: avoidance of uncertainty, masculinity, and short‐term orientation.

Point-of-care ultrasonography: Downstream utilization of and diagnostic (dis)agreements with additional cross-sectional imaging

ObjectivesPoint-of-care ultrasonography (POCUS), defined as ultrasonography (US) performed and interpreted by the clinician, is increasingly performed. This study aimed to determine the frequency of and reasons why clinicians of the emergency department request cross-sectional imaging after POCUS and how often radiologists experience diagnostic (dis)agreements. MethodsThis retrospective study included a consecutive series of 503 patients who underwent POCUS at the emergency department of a tertiary care center. ResultsDownstream cross-sectional imaging was performed in 77 (15.3%) of 503 POCUS examinations. Reasons for additional cross-sectional imaging were, in order of decreasing frequency: suspicion of pathology that was not assessed with POCUS in 46 cases (59.7%), confirmation of conclusive POCUS findings in 21 cases (27.3%), inconclusive POCUS (i.e. insufficient visualization of the structure of interest to make a diagnosis, despite an attempt of the POCUS operator) in 7 cases (9.6%), a combination of inconclusive POCUS and suspicion of pathology that was not assessed with POCUS in 2 cases (2.6%), and clarification of incidental findings on POCUS in 1 case (1.3%). In the 21 cases that underwent additional cross-sectional imaging to confirm POCUS findings, POCUS agreed with additional cross-sectional imaging in 19 (90.5%) and disagreed in 2 (9.5%) cases. ConclusionsThe use of POCUS appears to not cause any considerable downstream overutilization of cross-sectional imaging. In addition, radiologists experience few diagnostic disagreements when asked to perform second opinion cross-sectional imaging. Future studies with more homogeneous datasets in terms of POCUS operators are required to confirm our results.

Automated knee cartilage segmentation for heterogeneous clinical MRI using generative adversarial networks with transfer learning.

This study aimed to build a deep learning model to automatically segment heterogeneous clinical MRI scans by optimizing a pre-trained model built from a homogeneous research dataset with transfer learning. Conditional generative adversarial networks pretrained on the Osteoarthritis Initiative MR images was transferred to 30 sets of heterogenous MR images collected from clinical routines. Two trained radiologists manually segmented the 30 sets of clinical MR images for model training, validation and test. The model performance was compared to models trained from scratch with different datasets, as well as two radiologists. A 5-fold cross validation was performed. The transfer learning model obtained an overall averaged Dice coefficient of 0.819, an averaged 95 percentile Hausdorff distance of 1.463 mm, and an averaged average symmetric surface distance of 0.350 mm on the 5 random holdout test sets. A 5-fold cross validation had a mean Dice coefficient of 0.801, mean 95 percentile Hausdorff distance of 1.746 mm, and mean average symmetric surface distance of 0.364 mm. It outperformed other models and performed similarly as the radiologists. A transfer learning model was able to automatically segment knee cartilage, with performance comparable to human, using heterogeneous clinical MR images with a small training data size. In addition, the model proved robust when tested through cross validation and on images from a different vendor. We found it feasible to perform fully automated cartilage segmentation of clinical knee MR images, which would facilitate the clinical application of quantitative MRI techniques and other prediction models for improved patient treatment planning.

Stratospheric Moistening After 2000

AbstractThe significant climate feedback of stratospheric water vapor (SWV) necessitates quantitative estimates of SWV budget changes. Model simulations driven by the newest European Centre for Medium‐Range Weather Forecast reanalysis ERA5, satellite observations from the Stratospheric Water and OzOne Satellite Homogenized data set, Microwave Limb Sounder, and in situ frost point hygrometer observations from Boulder all show substantial and persistent stratospheric moistening after a sharp drop in water vapor at the turn of the millennium. This moistening occurred mainly during 2000–2006 and SWV abundances then remained high over the last decade. We find strong positive trends in the Northern Hemisphere and weak negative trends over the South Pole, mainly during austral winter. Moistening of the tropical stratosphere after 2000 occurred during late boreal winter/spring, reached values of ∼0.2 ppm/decade, was well correlated with a warming of the cold point tropopause by ∼0.4 K/decade and can only be partially attributed to El Nino‐Southern Oscillation and volcanic eruptions.

Ensemble machine learning approach for electronic nose signal processing

Electronic nose (e-nose) systems have been reported to be used in many areas as rapid, low-cost, and non-invasive instruments. Especially in meat production and processing, e-nose system is a powerful tool to process volatile compounds as a unique ‘fingerprint’. The ability of the pattern recognition algorithm to analyze e-nose signals is the key to the success of the e-nose system in many applications. On the other hand, ensemble methods have been reported for favorable performances in various data sets. This research proposes an ensemble learning approach for e-nose signal processing, especially in beef quality assessment. Ensemble methods are not only used for learning algorithms but also sensor array optimization. For sensor array optimization, three filter-based feature selection algorithms (FSAs) are used to build ensemble FSA such as reliefF, chi-square, and gini index. Ensemble FSA is developed to deal with different or unstable outputs of a single FSA on homogeneous e-nose data sets in beef quality monitoring. Moreover, ensemble learning algorithms are employed to deal with multi-class classification and regression tasks. Random forest and Adaboost are used that represent bagging and boosting algorithms, respectively. The results are also compared with support vector machine and decision tree as single learners. According to the experimental results, our ensemble approach has good performance and generalization in e-nose signal processing. Optimized sensor combination based on filter-based FSA shows stable results both in classification and regression tasks. Furthermore, Adaboost as a boosting algorithm produces the best prediction even though using a smaller number of sensors.

PD-LGD correlation for the banking lending segment: Empirical evidence from Russia

The Bank of Russia is one of the unique banking regulators in the world as it discloses granular reporting information per the existing credit institutions with the available historical track record. Same time the number of banks dramatically declined from above two and a half thousands in 1990s to one thousand in 2010 and to around 350 in 2021. Such information stimulates designing default probability (PD) models for the Russian banks. There is a separate stream of research that studies the amount of negative capital revealed when the Russian bank got its license withdrawn. However, the existing papers have several shortcomings. First, most of them do not account for the structural breaks in data. Second, there is no search for the best fitting model, just a model is offered and the coefficients of interest are interpreted. Third, the best model is poorly interpretable. Forth, the existing models make short-term forecasts. Fifth, there is no a LGD model for Russian banks, though the amount of negative capital upon license withdrawal was considered. Thus, our research objective is to study PD-LGD correlation (PLC) for the Russian banks. To do so, we improve the existing Russian banks PD model and create a respective novel LGD model. We use the homogenous dataset from 2016 to 2021. We find that PLC for Russian banks equals to +22%.

GAMMA: A General Agent Motion Model for Autonomous Driving

This letter presents GAMMA, a general motion prediction model that enables large-scale real-time simulation and planning for autonomous driving. GAMMA models <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">heterogeneous</i> , <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">interactive</i> traffic agents that operate under diverse road conditions, with various geometric and kinematic constraints. GAMMA treats the prediction task as constrained optimization in traffic agents’ velocity space. The objective is to optimize an agent’s driving performance, while obeying all the constraints resulting from the agent’s kinematics, collision avoidance with other agents, and the environmental context. Further, GAMMA explicitly conditions the prediction on human behavioral states as parameters of the optimization model, in order to account for versatile human behaviors. We evaluated GAMMA on a set of real-world benchmark datasets. The results show that GAMMA achieves high prediction accuracy on both homogeneous and heterogeneous traffic datasets, with sub-millisecond execution time. Further, the computational efficiency and the flexibility of GAMMA enable (i) simulation of mixed urban traffic at many locations worldwide and (ii) planning for autonomous driving in dense traffic with uncertain driver behaviors, both in real-time. The open-source code of GAMMA is available <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">online</uri> .

Sources of bias in artificial intelligence that perpetuate healthcare disparities-A global review.

While artificial intelligence (AI) offers possibilities of advanced clinical prediction and decision-making in healthcare, models trained on relatively homogeneous datasets, and populations poorly-representative of underlying diversity, limits generalisability and risks biased AI-based decisions. Here, we describe the landscape of AI in clinical medicine to delineate population and data-source disparities. We performed a scoping review of clinical papers published in PubMed in 2019 using AI techniques. We assessed differences in dataset country source, clinical specialty, and author nationality, sex, and expertise. A manually tagged subsample of PubMed articles was used to train a model, leveraging transfer-learning techniques (building upon an existing BioBERT model) to predict eligibility for inclusion (original, human, clinical AI literature). Of all eligible articles, database country source and clinical specialty were manually labelled. A BioBERT-based model predicted first/last author expertise. Author nationality was determined using corresponding affiliated institution information using Entrez Direct. And first/last author sex was evaluated using the Gendarize.io API. Our search yielded 30,576 articles, of which 7,314 (23.9%) were eligible for further analysis. Most databases came from the US (40.8%) and China (13.7%). Radiology was the most represented clinical specialty (40.4%), followed by pathology (9.1%). Authors were primarily from either China (24.0%) or the US (18.4%). First and last authors were predominately data experts (i.e., statisticians) (59.6% and 53.9% respectively) rather than clinicians. And the majority of first/last authors were male (74.1%). U.S. and Chinese datasets and authors were disproportionately overrepresented in clinical AI, and almost all of the top 10 databases and author nationalities were from high income countries (HICs). AI techniques were most commonly employed for image-rich specialties, and authors were predominantly male, with non-clinical backgrounds. Development of technological infrastructure in data-poor regions, and diligence in external validation and model re-calibration prior to clinical implementation in the short-term, are crucial in ensuring clinical AI is meaningful for broader populations, and to avoid perpetuating global health inequity.

INGe: Intensity-ground motion data set for Italy

In this paper we present an updated and homogeneous earthquake dataset for Italy compiled by joining the intensities available in the Italian Macroseismic Database DBMI15 and the peak ground motion (PGM) parameters present in the Engineering Strong-Motion (ESM) accelerometric data bank. The database has been compiled through an extensive procedure of evaluation and revision based on two main steps: 1) the selection of the earthquakes in DBMI15 with homogeneous macroseismic intensities in terms of data sources and 2) the extraction of all the localities reporting intensity data which are located within 3 km from the accelerograph stations that recorded the data. The final dataset includes 519 intensity-PGM data pairs from 65 earthquakes and 227 stations in the time span 1972–2016. The reported intensities are expressed either in the Mercalli-Cancani- Sieberg (MCS) or the European macroseismic (EMS-98) scales. The events are characterized by magnitudes in the range 4.1–6.8 and depths in the range 0–55 km. Here, we illustrate the data collection and the properties of the database in terms of recording, event and station distributions as well as macroseismic intensity points. Furthermore, we discuss the most relevant features of engineering interest showing several statistics with reference to the most significant metadata (such as moment magnitude, several distance metrics, style of faulting etc). The dataset is expected to be useful for benchmarking existing and for developing new ground motion intensity conversion equations offering a common basis, and sparing the time and effort required for assembling to the interested researchers. The dataset is available at https://zenodo.org/record/4623732#.YNX-AZMzbdc.

Exploring syntactical features for anomaly detection in application logs

Abstract In this research, we analyze the effect of lightweight syntactical feature extraction techniques from the field of information retrieval for log abstraction in information security. To this end, we evaluate three feature extraction techniques and three clustering algorithms on four different security datasets for anomaly detection. Results demonstrate that these techniques have a role to play for log abstraction in the form of extracting syntactic features which improves the identification of anomalous minority classes, specifically in homogeneous security datasets.

MIGHTEE-H i: the H i size–mass relation over the last billion years

ABSTRACT We present the observed H i size–mass relation of 204 galaxies from the MIGHTEE Survey Early Science data. The high sensitivity of MeerKAT allows us to detect galaxies spanning more than 4 orders of magnitude in H i mass, ranging from dwarf galaxies to massive spirals, and including all morphological types. This is the first time the relation has been explored on a blind homogeneous data set that extends over a previously unexplored redshift range of 0 &amp;lt; z &amp;lt; 0.084, i.e. a period of around one billion years in cosmic time. The sample follows the same tight logarithmic relation derived from previous work, between the diameter ($D_{\rm H\, \small {\rm I}}$) and the mass ($M_{\rm H\, \small {\rm I}}$) of H i discs. We measure a slope of 0.501 ± 0.008, an intercept of $-3.252^{+0.073}_{-0.074}$, and an observed scatter of 0.057 dex. For the first time, we quantify the intrinsic scatter of 0.054 ± 0.003 dex (${\sim } 10 {{\ \rm per\ cent}}$), which provides a constraint for cosmological simulations of galaxy formation and evolution. We derive the relation as a function of galaxy type and find that their intrinsic scatters and slopes are consistent within the errors. We also calculate the $D_{\rm H\, \small {\rm I}}{ \!-\! }M_{\rm H\, \small {\rm I}}$ relation for two redshift bins and do not find any evidence for evolution with redshift. These results suggest that over a period of one billion years in look-back time, galaxy discs have not undergone significant evolution in their gas distribution and mean surface mass density, indicating a lack of dependence on both morphological type and redshift.

Accuracy Improvement Method Based on Characteristic Database Classification for IMRT Dose Prediction in Cervical Cancer: Scientifically Training Data Selection.

PurposeConsistent training and testing datasets can lead to good performance for deep learning (DL) models. However, a large high-quality training dataset for unusual clinical scenarios is usually not easy to collect. The work aims to find optimal training data collection strategies for DL-based dose prediction models.Materials and MethodsA total of 325 clinically approved cervical IMRT plans were utilized. We designed comparison experiments to investigate the impact of (1) beam angles, (2) the number of beams, and (3) patient position for DL dose prediction models. In addition, a novel geometry-based beam mask generation method was proposed to provide beam setting information in the model training process. What is more, we proposed a new training strategy named “full-database pre-trained strategy”.ResultsThe model trained with a homogeneous dataset with the same beam settings achieved the best performance [mean prediction errors of planning target volume (PTV), bladder, and rectum: 0.29 ± 0.15%, 3.1 ± 2.55%, and 3.15 ± 1.69%] compared with that trained with large mixed beam setting plans (mean errors of PTV, bladder, and rectum: 0.8 ± 0.14%, 5.03 ± 2.2%, and 4.45 ± 1.4%). A homogeneous dataset is more accessible to train an accurate dose prediction model (mean errors of PTV, bladder and rectum: 2.2 ± 0.15%, 5 ± 2.1%, and 3.23 ± 1.53%) than a non-homogeneous one (mean errors of PTV, bladder and rectum: 2.55 ± 0.12%, 6.33 ± 2.46%, and 4.76 ± 2.91%) without other processing approaches. The added beam mask can constantly improve the model performance, especially for datasets with different beam settings (mean errors of PTV, bladder, and rectum improved from 0.8 ± 0.14%, 5.03 ± 2.2%, and 4.45 ± 1.4% to 0.29 ± 0.15%, 3.1 ± 2.55%, and 3.15 ± 1.69%).ConclusionsA consistent dataset is recommended to form a patient-specific IMRT dose prediction model. When a consistent dataset is not accessible to collect, a large dataset with different beam angles and a training model with beam information can also get a relatively good model. The full-database pre-trained strategies can rapidly form an accuracy model from a pre-trained model. The proposed beam mask can effectively improve the model performance. Our study may be helpful for further dose prediction studies in terms of training strategies or database establishment.

Classifying liganded states in heterogeneous single-particle cryo-EM datasets.

A powerful aspect of single-particle cryogenic electron microscopy is its ability to determine high-resolution structures from samples containing heterogeneous mixtures of the same macromolecule in different conformational or compositional states. Beyond determining structures at higher resolutions, one outstanding question is if macromolecules with only subtle conformation differences, such as the same protein bound with different ligands in the same binding pocket, can be separated reliably, and if information concerning binding kinetics can be derived from the particle distributions of different conformations obtained in classification. In this study, we address these questions by assessing the classification of synthetic heterogeneous datasets of Transient Receptor Potential Vanilloid 1 generated by combining different homogeneous experimental datasets. Our results indicate that classification can isolate highly homogeneous subsets of particle for calculating high-resolution structures containing individual ligands, but with limitations.

Hierarchical autoclassification of cryo-EM samples and macromolecular energy landscape determination

Background and objectiveCryo-electron microscopy using single particle analysis is a powerful technique for obtaining 3D reconstructions of macromolecules in near native conditions. One of its major advances is its capacity to reveal conformations of dynamic molecular complexes. Most popular and successful current approaches to analyzing heterogeneous complexes are founded on Bayesian inference. However, these 3D classification methods require the tuning of specific parameters by the user and the use of complicated 3D re-classification procedures for samples affected by extensive heterogeneity. Thus, the success of these approaches highly depends on the user experience. We introduce a robust approach to identify many different conformations presented in a cryo-EM dataset based on Bayesian inference through Relion classification methods that does not require tuning of parameters and reclassification strategies. MethodsThe algorithm allows both 2D and 3D classification and is based on a hierarchical clustering approach that runs automatically without requiring typical inputs, such as the number of conformations present in the dataset or the required classification iterations. This approach is applied to robustly determine the energy landscapes of macromolecules. ResultsWe tested the performance of the methods proposed here using four different datasets, comprising structurally homogeneous and highly heterogeneous cases. In all cases, the approach provided excellent results. The routines are publicly available as part of the CryoMethods plugin included in the Scipion package. ConclusionsOur results show that the proposed method can be used to align and classify homogeneous and heterogeneous datasets without requiring previous alignment information or any prior knowledge about the number of co-existing conformations. The approach can be used for both 2D and 3D autoclassification and only requires an initial volume. In addition, the approach is robust to the “attractor” problem providing many different conformations/views for samples affected by extensive heterogeneity. The obtained 3D classes can render high resolution 3D structures, while the obtained energy landscapes can be used to determine structural trajectories.

The New Radiosounding HARMonization (RHARM) Data Set of Homogenized Radiosounding Temperature, Humidity, and Wind Profiles With Uncertainties

AbstractObservational records are more often than not influenced by residual non‐climatic factors which must be detected and adjusted for prior to their usage. In this work, we present a novel approach, named Radiosounding HARMonization (RHARM), providing a homogenized data set of temperature, humidity and wind profiles along with an estimation of the measurement uncertainties for 697 radiosounding stations globally. The RHARM method has been used to adjust twice daily (0000 and 1200 UTC) radiosonde data holdings at 16 pressure levels in the range 1,000–10 hPa, from 1978 to present, provided by the Integrated Global Radiosonde Archive. Relative humidity (RH) data are limited to 250 hPa. The applied adjustments are interpolated to all reported levels. RHARM is the first data set to provide homogenized time series with an estimation of the observational uncertainty at each sounding pressure level. By construction, RHARM adjusted fields are not affected by cross‐contamination of biases across stations and are fully independent of reanalysis data. Analysis of trends for temperature, RH and winds highlights increased geographical coherency of trends over 1978–2000 globally, but especially in the Northern Hemisphere and South America. RHARM shows warming trends of 0.39 K/decade at 300 hPa in the Northern Hemisphere and of 0.25 K/decade in the tropics. The RHARM adjustments also reduce differences with the European Centre for Medium‐Range Weather Forecast ERA5 reanalysis, with the strongest effect in the Northern Hemisphere for temperature and relative humidity. For wind speed, the comparison indicates a good agreement with ERA5 in the troposphere.