Search for: All records

Abstract The West Antarctic Ice Sheet (WAIS) is thinning at an accelerating rate, driven by melting at its margins by warm Circumpolar Deep Water (CDW). However, this understanding is largely based on observations from recent decades, leaving the long-term influence of ocean temperature on WAIS stability uncertain. Here we reconstruct bottom water temperatures and water mass properties over the past 18 kyr using benthic foraminiferal Mg/Ca and δ¹³C records from sediment cores in the Amundsen Sea. Our data indicate that warm CDW occupied the continental shelf between ~ 18.0 and 10.1 kyr BP, coincident with major WAIS retreat from the shelf break to near its present-day grounding-line position along the Marie Byrd Land coast. Bottom waters cooled after ~ 10.1 kyr BP and remained relatively stable thereafter, with no evidence for substantial grounding-line migration. Continued atmospheric warming across West Antarctica until a mid-Holocene thermal maximum (~6–3 kyr BP) without further retreat indicates that ocean heat was the primary driver of WAIS variability since the Last Glacial Maximum. 

Vision-language models (VLMs) like CLIP have been cherished for their ability to perform zero-shot visual recognition on open-vocabulary concepts. This is achieved by selecting the object category whose textual representation bears the highest similarity with the query image. While successful in some domains, this method struggles with identifying fine-grained entities as well as generalizing to unseen concepts that are not captured by the training distribution. Recent works attempt to mitigate these challenges by integrating category descriptions at test time, albeit yielding modest improvements. We attribute these limited gains to a misalignment between image regions and textual descriptions, which stems from CLIP's global alignment objective. In this paper, we propose GRAIN, a new pretraining strategy aimed at aligning representations at both fine and coarse levels simultaneously. Our approach learns to jointly ground textual descriptions in image regions along with aligning overarching captions with global image representations. To drive this pre-training, we leverage frozen Multimodal Large Language Models (MLLMs) to derive large-scale synthetic annotations. We demonstrate the enhanced zero-shot performance of our model compared to current state-of-the art methods across 11 diverse image classification datasets. Additionally, we introduce Products-2023, a newly curated, manually labeled dataset featuring novel concepts, and showcase our model's ability to recognize these concepts by benchmarking on it. Significant improvements achieved by our model on other downstream tasks like retrieval further highlight the superior quality of representations learned by our approach. 

Abstract Speciation processes in plants can be difficult to evaluate, but are essential to understanding evolutionary processes that lead to diversification. Determining the juncture at which a genetically and/or morphologically divergent population can be reliably considered a separate species is often challenging. This is particularly so with respect to recent divergences amongst closely related taxa wherein factors such as incomplete lineage sorting may yield confounding results. Taxa in theCymopterus terebinthinus(Apiaceae) species complex have long puzzled botanists. Named entities in this group display similar, yet apparently distinct morphologies that have been classified as varieties under various generic names highlighting long‐standing nomenclatural instability. Previous phylogenetic studies have challenged the monophyly of this complex. This study aims to clarify taxonomic boundaries and infer evolutionary relationships among the fourC. terebinthinusvarieties andC. petraeusby applying phylogenetic inference and incorporating ecological, morphological, and geographical evidence. We sampled from populations of all varieties ofC. terebinthinusandC. petraeusfor target capture with the Angiosperms353 bait kit. We performed phylogenetic analyses with maximum likelihood (RAxML and IQ‐TREE) and coalescent‐based phylogenetic analysis (ASTRAL). We also conducted principal component analysis of soil samples and climatic variables. We find thatC. terebinthinusand its varietal infrataxa comprise a monophyletic clade that includesC. petraeus. Clade groupings correspond to previous taxonomic assignments and morphology. Clades are often closely associated with geographical variables and at times correlated with ecological variables. Exceptions to this are here attributed to various evolutionary factors that often confound other phylogenetic analyses such as incomplete lineage sorting, introgression, and paralogous loci. Our findings suggests that geographical factors might play a major role in genetic and morphological differentiation in this complex. Despite finding well‐supported clades that correspond to defined morphological characters; further sampling amongC. petraeuspopulations is required to make taxonomic decisions. 

Abstract The genusTauschiahas long been a source of taxonomic consternation for researchers. The group of species currently included in this genus are distributed primarily across the western United States and Mexico, but a few species occur in Central America and northern South America. Its circumscription is highly problematic, and its species have been moved countless times between more than a dozen genera. The advent of molecular phylogenetics has allowed some testing of generic boundaries inTauschiaand related taxa, but the sampling of previous studies was limited to a few species representing too small of a range to sort out the confusion. Here, we expand the sample size to include plants from throughout the range of the genus and use this to examine relationships among species ofTauschia, as well as to the larger clades to which it belongs within tribe Selineae. We also detail the complex taxonomic history ofTauschiaand related genera, provide a complete synonymy of the genus as it is currently defined, and confirm the polyphyly ofTauschiavia phylogenetic analysis of nuclear and cpDNA sequences. 

Sedimentary records can illuminate relationships between the climate, topography, and glaciation of West Antarctica by revealing its Cenozoic topographic and paleoenvironmental history. Eocene fluvial drainage patterns have previously been inferred using geochemical provenance data from an ~44– to 34–million year deltaic sandstone recovered from the Amundsen Sea Embayment. One interpretation holds that a low-relief, low-lying West Antarctic landscape supported a >1500-kilometer transcontinental river system. Alternatively, higher-relief topography in central West Antarctica formed a drainage divide between the Ross and Amundsen seas. Here, zircon U-Pb data from Amundsen Sea Embayment sediments are examined alongside known regional bedrock provenance signatures. These analyses suggest that all observed provenance indicators in the Eocene sandstone derive from West Antarctic rocks. This implies that a local river system flowed off a West Antarctic drainage divide, helping constrain the mid-Late Eocene evolution of West Antarctic topography with implications for the history of rifting and the characteristics of sediments infilling interior basins.