Procolophonidae, a clade of stem reptiles, are hypothesized to be some of the first highly specialized herbivores to evolve following the end-Permian mass extinction event. That hypothesis is largely based on qualitative observations of tooth shape, which are highly subjective and not generalizable. Quantitative studies of reptilian tooth shape have employed relatively sophisticated methods to capture tooth complexity, but these approaches often require expensive equipment and software and are time intensive. In this study, we built a predictive model based on extant lizards to quantitatively predict the diets of procolophonids using simple measures of tooth morphology. We use linear discriminant analysis (LDA) to predict dietary ecology from tooth dimensions and phylogenetic MANOVA to test for significant differences in tooth dimensions for different diet categories. We report two key findings: (1) procolophonids are largely predicted as herbivorous but occupy a different area of the LDA space from extant lizards, and (2) simple metrics return similar results as complex methods, but with less confidence. We hypothesize that Triassic flora posed different mechanical and processing challenges from modern plants, which contributed to the unique tooth morphologies of procolophonids and likely other Triassic taxa.

A stratigraphic section made of Quaternary alluvial–lacustrine sediments belonging to the Baza Formation (South Spain) has been logged and studied for paleomagnetism, rock magnetic cyclostratigraphy, and electron spin resonance (ESR) quartz dating. Our results indicate that the section, which is found in the vicinity of a number of paleontological and archaeological localities, falls within the mid-Early Pleistocene (Calabrian), within the Matuyama Chron, and runs to the Jaramillo Subchron, encompassing the Cobb Mountain Subchron. The magnetostratigraphic results combined with rock magnetic cyclostratigraphy and ESR provide solid timelines, which allow gross accumulation rates to be estimated, and revealing an upsection decrease of sedimentation in accordance with the lithological and paleodepositional changes. Our study furnishes new chronologies to better understand the timing of the latest stages of endorheic sedimentation that precedes the capture of the Baza Basin by the Gualdalquivir River in the Middle Pleistocene.

In January 1852, as searches continued for Sir John Franklin’s missing Arctic expedition, large quantities of preserved (that is, canned) meats supplied to the Royal Navy were found putrid and caused concern that the expedition had received similar meats with fatal consequences. Whilst a Parliamentary enquiry concluded correctly that the expedition had received good-quality meats, it neglected the fact that some 5% of all canned meats were condemned on ships due to damage and corrosion. As the Franklin expedition would be no exception, the study applies recent evidence of the expedition’s victualling schedule to estimate the number of cans condemned by the time at Beechey Island when a decision would be made whether enough remained to sustain the mission. It also estimates the vitamin B1 (thiamine) content of the meat because high temperatures during canning would have degraded or destroyed that vitamin, and vitamin C, both being essential to health. Any reduction in general rations would add to the decline in the quality of the diet. The expedition’s unique circumstances of long entrapment without recourse to hunting to supplement such deficiencies, or to escape, would prove fatal regardless of the good quality of the canned provisions.

We present a critical survey on the consistency of uncertainty quantification used in deep learning and highlight partial uncertainty coverage and many inconsistencies. We then provide a comprehensive and statistically consistent framework for uncertainty quantification in deep learning that accounts for all major sources of uncertainty: input data, training and testing data, neural network weights, and machine-learning model imperfections, targeting regression problems. We systematically quantify each source by applying Bayes’ theorem and conditional probability densities and introduce a fast, practical implementation method. We demonstrate its effectiveness on a simple regression problem and a real-world application: predicting cloud autoconversion rates using a neural network trained on aircraft measurements from the Azores and guided by a two-moment bin model of the stochastic collection equation. In this application, uncertainty from the training and testing data dominates, followed by input data, neural network model, and weight variability. Finally, we highlight the practical advantages of this methodology, showing that explicitly modeling training data uncertainty improves robustness to new inputs that fall outside the training data, and enhances model reliability in real-world scenarios.

This study presents the first sponge biodiversity inventory of Los Picos reef in Veracruz, Mexico. Although the Veracruz Reef System is known for its high sponge diversity, several recently discovered submerged reefs – including Los Picos – had remained biologically uncharacterised until this investigation. Our comprehensive inventory documents 37 species, identified at the species level, all belonging to Demospongiae; 15 of which are new records for the Mexican coast, and 13 for the Gulf of Mexico (GoM). Six species are described as new: Psammocinia alcoladoi sp. nov., distinguished by a dermal surface armoured with sand and spongin filaments, lightly fasciculated primary fibres, and non fasciculated secondary with long conules and slender fiber diameters; Hyatella hyattus sp. nov., distinguished by its soft and lobular habitus, and slender fiber diameters; Zyzzya marinagreenae sp. nov., is an open fistula with acanthostrongyles irregularly spined and both, acanthostrongyles and isochelae, smaller in size; Desmapsamma paulumharenae sp. nov., has an encrusting shape with larger spicules than D. anchorata; Phorbas veracruzanus sp. nov., consists of conule-shaped processes on an embedded layer and morphometric differences on spicules; and Timea citlallitzina sp. nov., stands out by the lumpy tips of the oxyaster type. The whole, highlighted by two genera, Psammocinia and Zyzzya, both reported for the first time in the GoM. Samples were obtained by SCUBA surveys at 10–16 m depth, between August and October 2017.

Plastic pollution is an emerging threat to our marine and terrestrial environments. International policy development to mitigate this pollution is currently underway; however, the process is complex. In parallel, local, national and regional initiatives are being developed. The research community is advocating for strong regulations based on robust evidence of the harm and severity of plastic pollution on oceans, land and for humans. However, the research community comprises different disciplines, research traditions and methodological approaches. This heterogeneity is in its core a great advantage, but only when we manage to build on each other where “synthesis” occurs. Currently, there remains a divide between the physical and social sciences, with research predominantly skewed toward the physical sciences, focusing on concentrations, fate and impact of plastic pollution.

In this article, I question the lack of philosophical discussions about the foundation of our research on plastic pollution, our knowledge claims, our views of “clean,” “pristine” and nature itself, as well as of the need for such discussions. I also propose a possible way forward to bridge the gap between disciplines and research traditions, fostering real interdisciplinarity.