observation bias
Recently Published Documents

Abstract The observed distributions of the source properties from gravitational-wave (GW) detections are biased due to the selection effects and detection criteria in the detections, analogous to the Malmquist bias. In this work, this observation bias is investigated through its fundamental statistical and physical origins. An efficient semi-analytical formulation for its estimation is derived, which is as accurate as the standard method of numerical simulations, with only a millionth of the computational cost. Then, the estimated bias is used for unmodeled inferences on the binary black hole population. These inferences show additional structures, specifically two peaks in the joint mass distribution around binary masses ∼10 M ⊙ and ∼30 M ⊙. Example ready-to-use scripts and some produced data sets for this method are shared in an online repository.

Fisheries researchers have focused on the value of information (VOI) in fisheries management and trade-offs since scientists and managers realized that information from different resources has different contribution in the management process. We picked seven indicators, which are log-normal annual catch observation error (Cobs), annual catch observation bias (Cbias), log-normal annual index observation error (Iobs), maximum length observation bias (Linfbias), observed natural mortality rate bias (Mbias), observed von Bertalanffy growth parameter K bias (Kbias), and catch-at-age sample size (CAA_nsamp), and built operating models (OMs) to simulate fisheries dynamics, and then applied management strategy evaluation (MSE). Relative yield is chosen as the result to evaluate the contribution of the seven indicators. Within the parameter range, there was not much information value reflected from fisheries-dependent parameters including Cobs, Cbias, and Iobs. On the other hand, for fisheries-independent parameters such as Kbias, Mbias, and Linfbias, similar tendency of the information value was showed in the results, in which the relative yield goes down from the upper bound to the lower bound of the interval. CAA_nsamp had no impact on the yield after over 134 individuals. The VOI analysis contributes to the trade-offs in the decision-making process. Information with more value is more worthy to collect in case of waste of time and money so that we could make the best use of scientific effort. But we still need to improve the simulation process such as enhancing the diversity and predictability in an OM. More parameters are on the way to be tested in order to collect optimum information for management and decision-making.

Social attention deficits represent a central impairment of patients suffering from autism spectrum disorder (ASD), but the nature of such deficits remains controversial. We compared visual attention regarding social (faces) vs. non-social stimuli (objects), in an ecological diagnostic context, in 46 children and adolescents divided in two groups: ASD (N = 23) and typical neurodevelopment (TD) (N = 23), matched for chronological age and intellectual performance. Eye-tracking measures of visual scanning, while exploring and describing scenes from three different tasks from the Autism Diagnostic Observation Schedule (ADOS), were analyzed: “Description of a Picture,” “Cartoons,” and “Telling a Story from a Book.” Our analyses revealed a three-way interaction between Group, Task, and Social vs. Object Stimuli. We found a striking main effect of group and a task dependence of attentional allocation: while the TD attended first and longer to faces, ASD participants became similar to TD when they were asked to look at pictures while telling a story. Our results suggest that social attention allocation is task dependent, raising the question whether spontaneous attention deficits can be rescued by guiding goal-directed actions.

Climate change is set to increase landslide frequency around the globe, thus increasing the potential exposure of people and material assets to these disturbances. Landslide hazard is commonly modelled from terrain and precipitation parameters, assuming that shorter, more intense rain events require less precipitation volume to trigger a slide. Given the extent of non-catastrophic slides, an operable vulnerability mapping requires high spatial resolution. We combined heterogeneous regional slide inventories with long-term meteorological records and small-scale spatial information for hazard modelling. Slope, its (protective) interaction with forest cover, and altitude were the most influential terrain parameters. A widely used exponential threshold to estimate critical precipitation was found to incorrectly predict meteorological hazard to a substantial degree and, qualitatively, delineate the upper boundary of natural conditions rather than a critical threshold. Scaling rainfall parameters from absolute values into local probabilities (per km²) however revealed a consistent pattern across datasets, with the transition from normal to critical rain volumes and durations being gradual rather than abrupt thresholds. Scaled values could be reverted into site-specific nomograms for easy appraisal of critical rain conditions by local stakeholders. An overlay of terrain-related hazard with infrastructure yielded local vulnerability maps, which were verified with actual slide occurrence. Multiple potential for observation bias in ground-based slide reporting underlined the value of complementary earth observation data for slide mapping and early warning.

<p>Our understanding of subsurface processes suffers from a profound observation bias: ground-motion sensors are rare, sparse, clustered on continents and not available where they are most needed. A new seismic recording technology called distributed acoustic sensing (DAS), can transform existing telecommunication fiber-optic cables into arrays of thousands of sensors, enabling meter-scale recording over tens of kilometers of linear fiber length. DAS works in high-pressure and high-temperature environments, enabling long-term recordings of seismic signals inside reservoirs, fault zones, near active volcanoes, in deep seas or in highly urbanized areas.</p><p>In this talk, we will introduce this laser-based technology and present three recent cases of study. The first experiment is in the city of Stanford, California, where DAS measurements are used to provide geotechnical information at a scale normally unattainable (i.e., for each building) with traditional geophone instrumentation. In the second study, we will show how downhole DAS passive recordings from the San Andreas Fault Observatory at Depth can be used for seismic velocity estimation. In the third research, we use DAS (in collaboration with Fujitec) to understand the ocean physics and infer seismic properties of the seafloor under a 100 km telecommunication cable.</p>

Abstract Background Modelling species richness across an elevation gradient has long attracted attention, and at same time places some significant obstacles to research. Many interpretations of patterns and corresponding mechanisms for species distributions are made without consideration of multiple confounding factors. What are factors that affect species richness with elevation? The answer may contribute to better understanding of the elevational distribution patterns and mechanisms.In this study, we performed the research on species richness of nematode-trapping fungi (NTF) across an elevation gradient in Yunnan, China.Results The results showed that sampling patterns, sampling altitude range, and human disturbance in sampling site could affect the resulting patterns of species richness significantly.Conclusion The results suggested that future studies on the elevational gradients of species richness should address these factors, and try to adopt the high-sampling patterns to reduce the observation bias.

Abstract. Data assimilation algorithms rely on a basic assumption of an unbiased observation error. However, the presence of inconsistent measurements with nontrivial biases or inseparable baselines is unavoidable in practice. Assimilation analysis might diverge from reality since the data assimilation itself cannot distinguish whether the differences between model simulations and observations are due to the biased observations or model deficiencies. Unfortunately, modeling of observation biases or baselines which show strong spatiotemporal variability is a challenging task. In this study, we report how data-driven machine learning can be used to perform observation bias correction for data assimilation through a real application, which is the dust emission inversion using PM10 observations. PM10 observations are considered unbiased; however, a bias correction is necessary if they are used as a proxy for dust during dust storms since they actually represent a sum of dust particles and non-dust aerosols. Two observation bias correction methods have been designed in order to use PM10 measurements as proxy for the dust storm loads under severe dust conditions. The first one is the conventional chemistry transport model (CTM) that simulates life cycles of non-dust aerosols. The other one is the machine-learning model that describes the relations between the regular PM10 and other air quality measurements. The latter is trained by learning using 2 years of historical samples. The machine-learning-based non-dust model is shown to be in better agreement with observations compared to the CTM. The dust emission inversion tests have been performed, through assimilating either the raw measurements or the bias-corrected dust observations using either the CTM or machine-learning model. The emission field, surface dust concentration, and forecast skill are evaluated. The worst case is when we directly assimilate the original observations. The forecasts driven by the a posteriori emission in this case even result in larger errors than the reference prediction. This shows the necessities of bias correction in data assimilation. The best results are obtained when using the machine-learning model for bias correction, with the existing measurements used more precisely and the resulting forecasts close to reality.

Abstract. Data assimilation algorithms rely on a basic assumption of an unbiased observation error. However, the presence of inconsistent measurements with nontrivial biases or inseparable baselines is unavoidable in practice. Assimilation analysis might diverge from reality, since the data assimilation itself cannot distinguish whether the differences between model simulations and observations are due to the biased observations or model deficiencies. Unfortunately, modeling of observation biases or baselines which show strong spatiotemporal variability is a challenging task. In this study, we report how data-driven machine learning can be used to perform observation bias correction for data assimilation through a real application, which is the dust emission inversion using PM10 observations. PM10 observations are considered as unbiased, however, a bias correction is necessary if they are used as a proxy for dust during dust storms since they actually represent a sum of dust particles and non-dust aerosols. Two observation bias correction methods have been designed in order to use PM10 measurements as proxy for the dust storm loads under severe dust conditions. The first one is the conventional chemical transport (CTM) model that simulates life cycles of non-dust aerosols. The other one is the machine learning model that describes the relations between the regular PM10 and other air quality measurement. The latter is trained by learning two-year's historical samples. The machine learning based non-dust model is shown to be in better agreements with observations compared to the CTM. The dust emission inversion tests have been performed, either through assimilating the raw measurements, or the bias-corrected dust observations using either the CTM or machine learning model. The emission field, surface dust concentration and forecast skill are evaluated. The worst case is when we directly assimilate the original observations. The forecasts driven by the posterior emission in this case even results in larger errors than the reference prediction. This shows the necessities of bias correction in data assimilation. The best results are obtained when using a machine learning model for bias correction, with the existing measurements used more precisely and the resulting forecasts close to reality.

Abstract. Marine planktonic diatoms, as today's ocean main carbon and silicon exporters, are central to developing an understanding of the interplay between the evolution of marine life and climate change. The diatom fossil record extends as far as the Early Cretaceous, and the late Paleogene to Recent interval is relatively complete and well documented. Their early Paleogene record, when diatoms first expanded substantially in the marine plankton, is hampered by decreased preservation (notably an episode of intense chertification in the early Eocene) as well as by observation bias. In this article, we attempt to correct for the latter by collecting diatom data in various Paleocene samples from legacy Deep Sea Drilling Project and Ocean Drilling Program deep-sea sediment sections. The results show a different picture from what previous analyses concluded, in that the Paleocene deep-sea diatoms seem in fact to have been as diverse and abundant as in the later Eocene, while exhibiting very substantial survivorship of Cretaceous species up until the Eocene.