Representativeness Of Datasets Research Articles

Parent attitudes towards data sharing in developmental science.

Data sharing in developmental science is increasingly encouraged, supported by funder and publisher mandates for open data access. Data sharing can accelerate discovery, link researchers with high quality analytic expertise to researchers with large datasets and democratise the research landscape to enable researchers with limited funding to access large sample sizes. However, there are also significant privacy and security concerns, in addition to conceptual and ethical considerations. These are particularly acute for developmental science, where child participants cannot consent themselves. As we move forward into a new era of data openness, it is essential that we adequately represent the views of stakeholder communities in designing data sharing efforts. We conducted a comprehensive survey of the opinions of 195 parents on data sharing in developmental science. Survey themes included how widely parents are willing to share their child's data, which type of organisations they would share the data with and the type of consent they would be comfortable providing. Results showed that parents were generally supportive of curated, but not open, data sharing. In addition to individual privacy and security concerns, more altruistic considerations around the purpose of research were important. Parents overwhelmingly supported nuanced consenting models in which preferences for particular types of data sharing could be changed over time. This model is different to that implemented in the vast majority of developmental science research and is contrary to many funder or publisher mandates. The field should look to create shared repositories that implement features such as dynamic consent and mechanisms for curated sharing that allow consideration of the scientific questions addressed. Better communication and outreach are required to build trust in data sharing, and advanced analytic methods will be required to understand the impact of selective sharing on reproducibility and representativeness of research datasets.

Personalized Stress Detection Using Biosignals from Wearables: A Scoping Review

Stress is a natural yet potentially harmful aspect of human life, necessitating effective management, particularly during overwhelming experiences. This paper presents a scoping review of personalized stress detection models using wearable technology. Employing the PRISMA-ScR framework for rigorous methodological structuring, we systematically analyzed literature from key databases including Scopus, IEEE Xplore, and PubMed. Our focus was on biosignals, AI methodologies, datasets, wearable devices, and real-world implementation challenges. The review presents an overview of stress and its biological mechanisms, details the methodology for the literature search, and synthesizes the findings. It shows that biosignals, especially EDA and PPG, are frequently utilized for stress detection and demonstrate potential reliability in multimodal settings. Evidence for a trend towards deep learning models was found, although the limited comparison with traditional methods calls for further research. Concerns arise regarding the representativeness of datasets and practical challenges in deploying wearable technologies, which include issues related to data quality and privacy. Future research should aim to develop comprehensive datasets and explore AI techniques that are not only accurate but also computationally efficient and user-centric, thereby closing the gap between theoretical models and practical applications to improve the effectiveness of stress detection systems in real scenarios.

AutoAugHAR: Automated Data Augmentation for Sensor-based Human Activity Recognition

Sensor-based HAR models face challenges in cross-subject generalization due to the complexities of data collection and annotation, impacting the size and representativeness of datasets. While data augmentation has been successfully employed in domains like natural language and image processing, its application in HAR remains underexplored. This study presents AutoAugHAR, an innovative two-stage gradient-based data augmentation optimization framework. AutoAugHAR is designed to take into account the unique attributes of candidate augmentation operations and the unique nature and challenges of HAR tasks. Notably, it optimizes the augmentation pipeline during HAR model training without substantially extending the training duration. In evaluations on eight inertial-measurement-units-based benchmark datasets using five HAR models, AutoAugHAR has demonstrated superior robustness and effectiveness compared to other leading data augmentation frameworks. A salient feature of AutoAugHAR is its model-agnostic design, allowing for its seamless integration with any HAR model without the need for structural modifications. Furthermore, we also demonstrate the generalizability and flexible extensibility of AutoAugHAR on four datasets from other adjacent domains. We strongly recommend its integration as a standard protocol in HAR model training and will release it as an open-source tool1.

Revolutionizing Medical Practice: The Impact of Artificial Intelligence (AI) on Healthcare

The twenty-first century has witnessed significant advancements in informatics, reshaping our understanding of data processing and accessibility. Artificial intelligence (AI), encompassing techniques such as machine learning (ML), deep learning (DP), and neural networks (NN), is poised to revolutionize medicine. AI holds the capability of analyzing vast amounts of data, extracting meaningful insights, and making accurate predictions, thereby empowering industries to make informed decisions, drive innovation, and enhance efficiency. The landscape of medical AI has evolved significantly, demonstrating expert-level disease detection from medical images and promising breakthroughs across various industries. AI revolutionizes medical practice by leveraging advanced algorithms and machine learning capabilities to improve diagnostics, treatment planning, and overall patient care. However, the deployment of medical AI systems in regular clinical practice still needs to be tapped, presenting complex ethical, technical, and human-centered challenges that must be addressed for successful implementation. While AI algorithms have shown efficacy in retrospective medical investigations, their translation into practical medical settings has been limited, raising concerns about their usability and interaction with healthcare professionals. Moreover, the representativeness of retrospective datasets in real-world medical practice is subject to filtering and cleaning biases. Integrating AI into clinical medicine holds great promise for transforming healthcare delivery, improving patient care, and revolutionizing aspects such as diagnosis, treatment planning, drug discovery, personalized treatment, and medical imaging. With advanced algorithms and machine learning capabilities, AI and robotics in Healthcare can analyze large volumes of medical data, extract meaningful insights, and provide accurate predictions, empowering healthcare professionals to make informed decisions and optimize resource allocation. The availability of extensive clinical, genomics, and digital imaging data, coupled with investments from healthcare institutions and technology giants, underscores the potential of AI in healthcare. This review article explores AI's powerful potential to revolutionize healthcare delivery across multiple domains, emphasizing the need to overcome challenges and harness its transformative capabilities in clinical practice.

Peculiarities of justification of a representative set of requirements for intelligent vehicles

Introduction (problem statement and relevance). The key task of ensuring confidence in motor vehicles with elements of artificial intelligence (AI vehicles) is forming of a representative set of requirements compliance with which ensures the necessary guarantees of AI vehicle functionality and safety.Absence of generally accepted and statutory approaches to justification of the requirements content constrains application of artificial intelligence technologies in motor vehicles and, respectively, impedes creation and improvement of highly automated vehicles. The purpose of the study is to analyze the principles of forming of a representative set of requirements for AI vehicles. Methodology and research methods. The paper uses the principles of qualimetry adapted for the information systems based on machine learning algorithms, it also uses the methods of system analysis, mathematical statistics, combinatorics, set theory and propositional calculus principles. Scientific novelty and results. The approach is formulated to justify the statistically representative set of requirements for AI vehicles, and that allows adequate and accurate assessment of compliance of these systems with the expectations of the customers, developers, suppliers, regulatory authorities and other stakeholders. The structure of such expectations and priorities has been considered taking into account creation and application of AI systems. It is shown that stakeholder requirements apply to both the AI vehicle life cycle processes and the systems per se taking into account particular (given) conditions of their operation. Different significant factors, which variability sets the intended operating conditions, have been considered and conditions of representativeness of test data sets used for AI vehicle trials have been formulated. It is shown that, in order to reduce the dimensionality of space of the intended operating conditions, AI vehicles can be decomposed into separate functional subsystems and then into individual AI algorithms. The ways to perform such functional decomposition have been suggested. Practical significance.The principles of justification of the requirements content offered in the paper can be used in certification testing of artificial intelligence algorithms that are used in highly automated vehicles.

Do human mobility network analyses produced from different location-based data sources yield similar results across scales?

The burgeoning availability of sensing technology and location-based data is driving the expansion of analysis of human mobility networks in science and engineering research, as well as in epidemic forecasting and mitigation, urban planning, traffic engineering, emergency response, and business development. However, studies employ datasets provided by different location-based data providers, and the extent to which the human mobility measures and results obtained from different datasets are comparable is not known. To address this gap, in this study, we examined three prominent location-based data sources—Spectus, X-Mode, and Veraset—to analyze human mobility networks across metropolitan areas at different scales: global, sub-structure, and microscopic. Dissimilar results were obtained from the three datasets, suggesting the sensitivity of network models and measures to datasets. This finding has important implications for building generalized theories of human mobility and urban dynamics based on different datasets. The findings also highlighted the need for ground-truthed human movement datasets to serve as the benchmark for testing the representativeness of human mobility datasets. Researchers and decision-makers across different fields of science and technology should recognize the sensitivity of human mobility results to dataset choice and develop procedures for ground-truthing the selected datasets in terms of representativeness of data points and transferability of results.

Parent attitudes towards data sharing in developmental science

Background Data sharing in developmental science is increasingly encouraged, supported by funder and publisher mandates for open data access. Data sharing can accelerate discovery, link researchers with high quality analytic expertise to researchers with large datasets and democratise the research landscape to enable researchers with limited funding to access large sample sizes. However, there are also significant privacy and security concerns, in addition to conceptual and ethical considerations. These are particularly acute for developmental science, where child participants cannot consent themselves. As we move forward into a new era of data openness, it is essential that we adequately represent the views of stakeholder communities in designing data sharing efforts. Methods We conducted a comprehensive survey of the opinions of 195 parents on data sharing in developmental science. Survey themes included how widely parents are willing to share their child’s data, which type of organisations they would share the data with and the type of consent they would be comfortable providing. Results Results showed that parents were generally supportive of curated, but not open, data sharing. In addition to individual privacy and security concerns, more altruistic considerations around the purpose of research were important. Parents overwhelmingly supported nuanced consenting models in which preferences for particular types of data sharing could be changed over time. This model is different to that implemented in the vast majority of developmental science research and is contrary to many funder or publisher mandates. Conclusions The field should look to create shared repositories that implement features such as dynamic consent and mechanisms for curated sharing that allow consideration of the scientific questions addressed. Better communication and outreach are required to build trust in data sharing, and advanced analytic methods will be required to understand the impact of selective sharing on reproducibility and representativeness of research datasets.

The Wildfire Dataset: Enhancing Deep Learning-Based Forest Fire Detection with a Diverse Evolving Open-Source Dataset Focused on Data Representativeness and a Novel Multi-Task Learning Approach

This study explores the potential of RGB image data for forest fire detection using deep learning models, evaluating their advantages and limitations, and discussing potential integration within a multi-modal data context. The research introduces a uniquely comprehensive wildfire dataset, capturing a broad array of environmental conditions, forest types, geographical regions, and confounding elements, aiming to reduce high false alarm rates in fire detection systems. To ensure integrity, only public domain images were included, and a detailed description of the dataset’s attributes, URL sources, and image resolutions is provided. The study also introduces a novel multi-task learning approach, integrating multi-class confounding elements within the framework. A pioneering strategy in the field of forest fire detection, this method aims to enhance the model’s discriminatory ability and decrease false positives. When tested against the wildfire dataset, the multi-task learning approach demonstrated significantly superior performance in key metrics and lower false alarm rates compared to traditional binary classification methods. This emphasizes the effectiveness of the proposed methodology and the potential to address confounding elements. Recognizing the need for practical solutions, the study stresses the importance of future work to increase the representativeness of training and testing datasets. The evolving and publicly available wildfire dataset is anticipated to inspire innovative solutions, marking a substantial contribution to the field.

Longitudinal assessment of demographic representativeness in the Medical Imaging and Data Resource Center open data commons.

The Medical Imaging and Data Resource Center (MIDRC) open data commons was launched to accelerate the development of artificial intelligence (AI) algorithms to help address the COVID-19 pandemic. The purpose of this study was to quantify longitudinal representativeness of the demographic characteristics of the primary MIDRC dataset compared to the United States general population (US Census) and COVID-19 positive case counts from the Centers for Disease Control and Prevention (CDC). The Jensen-Shannon distance (JSD), a measure of similarity of two distributions, was used to longitudinally measure the representativeness of the distribution of (1)all unique patients in the MIDRC data to the 2020 US Census and (2)all unique COVID-19 positive patients in the MIDRC data to the case counts reported by the CDC. The distributions were evaluated in the demographic categories of age at index, sex, race, ethnicity, and the combination of race and ethnicity. Representativeness of the MIDRC data by ethnicity and the combination of race and ethnicity was impacted by the percentage of CDC case counts for which this was not reported. The distributions by sex and race have retained their level of representativeness over time. The representativeness of the open medical imaging datasets in the curated public data commons at MIDRC has evolved over time as the number of contributing institutions and overall number of subjects have grown. The use of metrics, such as the JSD support measurement of representativeness, is one step needed for fair and generalizable AI algorithm development.

RWD13 Representation of Sex at Birth in Real-World Evidence, Randomized Clinical Trial, and Surveillance, Epidemiology, and End Results (SEER) Cancer Patient Registry Data By Cancer Type

In randomized clinical trials (RCTs) for oncology, individuals assigned female at birth have historically been underrepresented. However, the significance of this variance and extent to which it impacts the representativeness of cancer research data sets have not been established. We evaluated 5 recently completed real-world (RWE) studies, corresponding RCTs, and SEER data from a parallel time period to assess female sex representation.

Interday Stability of Taxi Travel Flow in Urban Areas

Taxi travel flow patterns and their interday stability play an important role in the planning of urban transportation and public service facilities. Existing studies pay little attention to the stability of the travel flow patterns between days, and it is difficult to consider the impact of dynamic changes in daily travel demand analysis when supporting related decision making. Taxi trajectory data have been widely used in urban taxi travel-pattern analysis. This paper uses the taxi datasets of Shenzhen and New York to analyze and compare the interday stability of the taxi travel spatial structure and the flow volume based on the improved Levenshtein algorithm and geographic flow theory. The results show that (1) interday differences in taxi travel flow are obvious in both spatial structure and flow volume, high-frequency origin–destination (OD) trips are relatively stable; (2) the ODs between the central urban area and surrounding areas exhibit high traffic volume and high interday stability, and the ODs starting or ending at an airport exhibit high traffic stability; (3) one week’s data can describe 86% of the overall travel structure and 84% of travel flow in Shenzhen, and one week’s New York data can describe 73% of travel structure and 76% of travel flow. There are differences in the travel patterns of people in different cities, and the representativeness of datasets in different cities will be different. These findings can help to better understand the outcomes of taxi travel patterns derived from a relatively short period of data to avoid potential misuse in related decision making.

Landscape Analysis of the Application of Artificial Intelligence and Machine Learning in Regulatory Submissions for Drug Development From 2016 to 2021.

Landscape Analysis of the Application of Artificial Intelligence and Machine Learning in Regulatory Submissions for Drug Development From 2016 to 2021.

Perspectives on computing ethics: a multi-stakeholder analysis

PurposeComputing ethics represents a long established, yet rapidly evolving, discipline that grows in complexity and scope on a near-daily basis. Therefore, to help understand some of that scope it is essential to incorporate a range of perspectives, from a range of stakeholders, on current and emerging ethical challenges associated with computer technology. This study aims to achieve this by using, a three-pronged, stakeholder analysis of Computer Science academics, ICT industry professionals, and citizen groups was undertaken to explore what they consider to be crucial computing ethics concerns. The overlap between these stakeholder groups are explored, as well as whether their concerns are reflected in the existing literature.Design/methodology/approachData collection was performed using focus groups, and the data was analysed using a thematic analysis. The data was also analysed to determine if there were overlaps between the literature and the stakeholders’ concerns and attitudes towards computing ethics.FindingsThe results of the focus group analysis show a mixture of overlapping concerns between the different groups, as well as some concerns that are unique to each of the specific groups. All groups stressed the importance of data as a key topic in computing ethics. This includes concerns around the accuracy, completeness and representativeness of data sets used to develop computing applications. Academics were concerned with the best ways to teach computing ethics to university students. Industry professionals believed that a lack of diversity in software teams resulted in important questions not being asked during design and development. Citizens discussed at length the negative and unexpected impacts of social media applications. These are all topics that have gained broad coverage in the literature.Social implicationsIn recent years, the impact of ICT on society and the environment at large has grown tremendously. From this fast-paced growth, a myriad of ethical concerns have arisen. The analysis aims to shed light on what a diverse group of stakeholders consider the most important social impacts of technology and whether these concerns are reflected in the literature on computing ethics. The outcomes of this analysis will form the basis for new teaching content that will be developed in future to help illuminate and address these concerns.Originality/valueThe multi-stakeholder analysis provides individual and differing perspectives on the issues related to the rapidly evolving discipline of computing ethics.

Overview and update of the SPARC Data Initiative: comparison of stratospheric composition measurements from satellite limb sounders

Abstract. The Stratosphere-troposphere Processes and their Role in Climate (SPARC) Data Initiative (SPARC, 2017) performed the first comprehensive assessment of currently available stratospheric composition measurements obtained from an international suite of space-based limb sounders. The initiative's main objectives were (1) to assess the state of data availability, (2) to compile time series of vertically resolved, zonal monthly mean trace gas and aerosol fields, and (3) to perform a detailed intercomparison of these time series, summarizing useful information and highlighting differences among datasets. The datasets extend over the region from the upper troposphere to the lower mesosphere (300–0.1 hPa) and are provided on a common latitude–pressure grid. They cover 26 different atmospheric constituents including the stratospheric trace gases of primary interest, ozone (O3) and water vapor (H2O), major long-lived trace gases (SF6, N2O, HF, CCl3F, CCl2F2, NOy), trace gases with intermediate lifetimes (HCl, CH4, CO, HNO3), and shorter-lived trace gases important to stratospheric chemistry including nitrogen-containing species (NO, NO2, NOx, N2O5, HNO4), halogens (BrO, ClO, ClONO2, HOCl), and other minor species (OH, HO2, CH2O, CH3CN), and aerosol. This overview of the SPARC Data Initiative introduces the updated versions of the SPARC Data Initiative time series for the extended time period 1979–2018 and provides information on the satellite instruments included in the assessment: LIMS, SAGE I/II/III, HALOE, UARS-MLS, POAM II/III, OSIRIS, SMR, MIPAS, GOMOS, SCIAMACHY, ACE-FTS, ACE-MAESTRO, Aura-MLS, HIRDLS, SMILES, and OMPS-LP. It describes the Data Initiative's top-down climatological validation approach to compare stratospheric composition measurements based on zonal monthly mean fields, which provides upper bounds to relative inter-instrument biases and an assessment of how well the instruments are able to capture geophysical features of the stratosphere. An update to previously published evaluations of O3 and H2O monthly mean time series is provided. In addition, example trace gas evaluations of methane (CH4), carbon monoxide (CO), a set of nitrogen species (NO, NO2, and HNO3), the reactive nitrogen family (NOy), and hydroperoxyl (HO2) are presented. The results highlight the quality, strengths and weaknesses, and representativeness of the different datasets. As a summary, the current state of our knowledge of stratospheric composition and variability is provided based on the overall consistency between the datasets. As such, the SPARC Data Initiative datasets and evaluations can serve as an atlas or reference of stratospheric composition and variability during the “golden age” of atmospheric limb sounding. The updated SPARC Data Initiative zonal monthly mean time series for each instrument are publicly available and accessible via the Zenodo data archive (Hegglin et al., 2020).

Characterizing Extreme Geomagnetic Storms Using Extreme Value Analysis: A Discussion on the Representativeness of Short Data Sets

AbstractOne of the main goals when studying Space Weather is to characterize extreme events occurrences and related characteristics. To do so, dedicated statistical methods from the so‐called extreme value analysis (EVA) field have been developed. In this study we used Ca index, derived from aa, in order to characterize geoeffectiveness from the radiation belts point of view with a 150‐year‐long data set. The analysis performed in this study thus focuses on this newsworthy index to provide clues on the reliability of EVA methods. The first main result we present here is that the 1‐in‐10‐, 1‐in‐50‐, and 1‐in‐100‐year events, respectively, match Ca values of 100.39, 131.39, and 142.84 nT. Consequently, the only 1‐in‐100 event observed during the Space Era would be the “Halloween Storm” in 2003 that reached a Ca value of 147.6 nT. The second main result highlighted in this work is that performing the same analysis with shorter subsets (20 years long) can give significantly different results for two reasons. The first reason is that some short time periods do not display the same distribution of events as the full period. The second reason is that the choice of the correct threshold (when using a Peaks Over Threshold approach) is made difficult with a short data set and leads to inaccurate results. This is a strong result as for accurate estimation of the induced effects of extreme events in radiation belts, we may only rely on short flux data sets from one or another mission (mostly shorter than 20 years).

A Continuing Investigation of Diurnal and Location Trends in an Ice Crystal Icing Engine Event Database

<div class="section abstract"><div class="htmlview paragraph">Due to ongoing efforts by the aviation industry, much has been learned over the last several years regarding jet engine power loss and compressor damage events caused by the ingestion of high concentrations of ice crystal particles into the core flow path. Boeing has created and maintained a database of such ice crystal icing (ICI) events to aid in analysis and further study of this phenomenon. This article provides a general update on statistics derived from the Boeing event database, and provides more details on specific event clusters of interest. A series of three flight campaigns have, over the past five years, collected in-situ data in deep convective clouds that will be used for the assessment of the new FAA CFR Part 33 ice crystal environmental envelope Appendix D, and the equivalent EASA CS-25 Appendix P. The most recent Boeing engine event study in 2015 focused on oceanic cloud systems that caused events in Southeast Asia, a region expected to have similar cloud properties as the first flight campaign in Darwin, Australia. The current study will examine ICI engine events over South America and Africa, which are caused by large deep convective continental clouds of the type that were not the main focus of the three Appendix D/P flight campaigns. Continental event clouds are compared to the previously studied oceanic event clouds, in order to ascertain whether any systematic differences exist that should be considered when assessing the representativeness of the flight campaign datasets. The results may also be useful to those developing nowcasting algorithms identifying high ice concentration regions using satellite measurements.</div></div>

Comparing forwarder boom-control systems based on an automatically recorded follow-up dataset

Crane work is the most time-consuming work element in forwarding. Hence, assist systems like boom-tip control are of interest. The first commercially available boom-tip control for forwarders was introduced in 2013. In this study we analysed whether replacing conventional boom control (CBC) with John Deere’s version of boom-tip control (named Intelligent Boom Control, IBC), increases crane-work productivity. We used data automatically gathered from 10 final-felling stands, covering typical logging conditions for southern, central and northern Sweden. Two John Deere 1510E and two John Deere 1910G forwarders were operated by seven experienced operators during the follow-up study, covering 1238 loads in total. A split-plot design was applied to isolate effects of the boom-control system being used (CBC, IBC). We found that using IBC for loading work (crane work and driving included) saved 5.2% of productive machine time compared to using CBC (p ≤ 0.05). The corresponding saving when using IBC for unloading work was 7.9% (p ≤ 0.05). Depending on geophysical factors, this corresponds to approximately 4% savings in productive machine time for forwarding as a whole, including pure transport (with and without load). Moreover, the study introduced in cut-to-length context a novel field-study design to collect a large follow-up dataset in the course of ordinary forwarding operations. We found the study design to be a cost-efficient way to combine the representativeness of conventional follow-up datasets with the ability to establish causal relationships. Establishment of causal relationships has traditionally been possible only through observational time studies or standardized experiments.

Representativeness of variation benchmark datasets

BackgroundBenchmark datasets are essential for both method development and performance assessment. These datasets have numerous requirements, representativeness being one. In the case of variant tolerance/pathogenicity prediction, representativeness means that the dataset covers the space of variations and their effects.ResultsWe performed the first analysis of the representativeness of variation benchmark datasets. We used statistical approaches to investigate how proteins in the benchmark datasets were representative for the entire human protein universe. We investigated the distributions of variants in chromosomes, protein structures, CATH domains and classes, Pfam protein families, Enzyme Commission (EC) classifications and Gene Ontology annotations in 24 datasets that have been used for training and testing variant tolerance prediction methods. All the datasets were available in VariBench or VariSNP databases. We tested also whether the pathogenic variant datasets contained neutral variants defined as those that have high minor allele frequency in the ExAC database. The distributions of variants over the chromosomes and proteins varied greatly between the datasets.ConclusionsNone of the datasets was found to be well representative. Many of the tested datasets had quite good coverage of the different protein characteristics. Dataset size correlates to representativeness but only weakly to the performance of methods trained on them. The results imply that dataset representativeness is an important factor and should be taken into account in predictor development and testing.

A novel method to determine the minimum number of sequences required for reliable microbial community analysis

Although high-throughput sequencing is an efficient approach to study the structure of microbial communities in detail, it is still impossible to enumerate all individuals using this method. Therefore, it is a common strategy to generate sampling datasets that are representative of the assemblages. However, the representativeness of these sampling datasets has never been assessed. In this study, we developed a method to determine the minimum number sequences that are required to be analyzed to obtain a reliable description of microbial community structure. First, a set of datasets from microbial communities were constructed by in silico sampling at different depths. Second, the correlation equation between dissimilarity of the sampling datasets and sampling depths was fitted, and thereby the minimum number of 16S rRNA gene sequences was predicted. Finally, we verified the method using empirical data of microbiota from a laboratory mesocosm. Our method showed that at least 5,528,079 sequences were required to reliably characterize microbial communities inhabiting the mesocosms. However, if only dominant OTUs (>1%) were considered, thousands of sequences were enough. This promising method provides a criterion to ensure sequencing sufficiency when analyzing the structure of natural microbial communities.

Global biodiversity monitoring: From data sources to Essential Biodiversity Variables

Essential Biodiversity Variables (EBVs) consolidate information from varied biodiversity observation sources. Here we demonstrate the links between data sources, EBVs and indicators and discuss how different sources of biodiversity observations can be harnessed to inform EBVs. We classify sources of primary observations into four types: extensive and intensive monitoring schemes, ecological field studies and satellite remote sensing. We characterize their geographic, taxonomic and temporal coverage. Ecological field studies and intensive monitoring schemes inform a wide range of EBVs, but the former tend to deliver short-term data, while the geographic coverage of the latter is limited. In contrast, extensive monitoring schemes mostly inform the population abundance EBV, but deliver long-term data across an extensive network of sites. Satellite remote sensing is particularly suited to providing information on ecosystem function and structure EBVs. Biases behind data sources may affect the representativeness of global biodiversity datasets. To improve them, researchers must assess data sources and then develop strategies to compensate for identified gaps. We draw on the population abundance dataset informing the Living Planet Index (LPI) to illustrate the effects of data sources on EBV representativeness. We find that long-term monitoring schemes informing the LPI are still scarce outside of Europe and North America and that ecological field studies play a key role in covering that gap. Achieving representative EBV datasets will depend both on the ability to integrate available data, through data harmonization and modeling efforts, and on the establishment of new monitoring programs to address critical data gaps.