Trust and trustworthy artificial intelligence: A research agenda for AI in the environmental sciences.

Demands to manage the risks of artificial intelligence (AI) are growing. These demands and the government standards arising from them both call for trustworthy AI. In response, we adopt a convergent approach to review, evaluate, and synthesize research on the trust and trustworthiness of AI in the environmental sciences and propose a research agenda. Evidential and conceptual histories of research on trust and trustworthiness reveal persisting ambiguities and measurement shortcomings related to inconsistent attention to the contextual and social dependencies and dynamics of trust. Potentially underappreciated in the development of trustworthy AI for environmental sciences is the importance of engaging AI users and other stakeholders, which human-AI teaming perspectives on AI development similarly underscore. Co-development strategies may also help reconcile efforts to develop performance-based trustworthiness standards with dynamic and contextual notions of trust. We illustrate the importance of these themes with applied examples and show how insights from research on trust and the communication of risk and uncertainty can help advance the understanding of trust and trustworthiness of AI in the environmental sciences.

Genus-wide analysis of Fusarium polyketide synthases reveals broad chemical potential.

The genus Fusarium includes pathogens of global concern to animal and plant health. Natural products (NPs) synthesized by Fusarium can contribute to pathogenesis or competitiveness of the fungus in the environment and to animal diseases, including cancer and neural tube defects. Polyketide synthases (PKSs) are a family of large, multi-domain enzymes that are required for synthesis of most fungal NPs. To gain insight into the NP potential of Fusarium, we retrieved 2974 PKS gene sequences from the genomes of 206 Fusarium species. Phylogenetic analysis resolved these PKSs, along with 118 previously described PKSs from other fungi, into 123 clades. Based on results from previous studies, we propose that PKSs in the same clade generally synthesize the same polyketide, which is structurally distinct from polyketides synthesized by PKSs in other clades. We predict that the 123 clades potentially produce 113 structurally distinct families of polyketide-derived NPs because some NPs (e.g., zearalenone) require two PKSs for their synthesis. Collectively, the clades include PKSs required for synthesis of six NPs whose production has not previously been reported in Fusarium, including two NPs with significant pharmaceutical interest: chaetoviridin and a statin. Our results highlight the NP diversity of Fusarium and the potential of the genus to produce metabolites with medical and other applications.

Quantification of Total Particulate Matter and Benzene-Soluble Fraction Inhalation Exposures in Roofing Workers Performing Tear-off Activities.

Asphalt shingle removal (tear-off) from roofs is a major job task for an estimated 174,000 roofers in the United States. However, a literature search showed that there are no published studies that characterize worker inhalation exposures to asphalt particulates during shingle tear-off. To begin to fill this gap, the present study of inhalation exposures of roofers performing asphalt shingle tear-off was undertaken. The airborne agents of interest were total particulate matter (TP) and organic particulates measured as the benzene-soluble fraction (BSF) of total particulate. The study's objectives were to measure the personal breathing zone (PBZ) exposures of roofing tear-off workers to BSF and TP; and to assess whether these PBZ exposures are different from ambient levels. Task-based PBZ samples (typical duration 1-5 hours) were collected during asphalt shingle tear-off from roofs near Houston, Texas and Denver, Colorado. Samples were analyzed for TP and BSF using National Institute of Occupational Safety and Health (NIOSH) Method 5042. As controls, area samples (typical duration 3-6 hours) were collected on the ground near the perimeter of the tear-off project Because of the presence of significant sources of inorganic particulates in the work environment, emphasis was placed on the BSF data. No BSF exposure higher than 0.25 mg/m3 was observed, and 69% of the PBZ samples were below the limit of detection (LOD). Due to unforeseen confounding, however, statistical comparisons of on-the-roof PBZ samples with on-the-ground area samples posed some special challenges. This confounding grew out of the interaction of three factors: statistical censoring from the left; the strong inverse correlation between LOD concentration and sampling duration; and variation in sampling durations between on-the-ground area samples and on-the-roof PBZ samples. A general linear model analysis of variance (GLM-ANOVA) was applied to help address the confounding. The results of this analysis indicate that personal sample BSF results were not statistically significantly different from the background/area samples.

Species-specific identification of penicillium linked to patulin contamination.

Certain species of Penicillium have been reported to produce the mycotoxin patulin, and research was undertaken to identify these with the use of oligonucleotide primer pairs. Species examined were found in food, plants, and soil and were reported to produce patulin. Penicillium expansum is the most commonly detected species linked to the presence of patulin in apple juice. At least 10 different enzymes are involved in the patulin biosynthetic pathway, including the isoepoxydon dehydrogenase (idh) gene. Based on nucleotide sequences previously determined for the idh gene in Penicillium species, PCR primers were designed for the species-specific detection of patulin-producing species. The 5' primers were based on differences in the second intron of the idh gene. To ensure that the primer pairs produced a PCR product restricted to the species for which it was designed, and not to unrelated species, all of the primer pairs were tested against all of the Penicillium species. With one exception, it was possible to detect a reaction only with the organism of interest. The primer pair for Penicillium griseofulvum also amplified DNA from Penicillium dipodomyicola, a closely related species; however, it was possible to distinguish between these two species by doing a second amplification, with a different primer pair specific only for P. dipodomyicola. Consequently, with different primer sets, it was possible to identify individual patulin-producing species of Penicillium.

Enhancing understanding and improving prediction of severe weather through spatiotemporal relational learning.

Severe weather, including tornadoes, thunderstorms, wind, and hail annually cause significant loss of life and property. We are developing spatiotemporal machine learning techniques that will enable meteorologists to improve the prediction of these events by improving their understanding of the fundamental causes of the phenomena and by building skillful empirical predictive models. In this paper, we present significant enhancements of our Spatiotemporal Relational Probability Trees that enable autonomous discovery of spatiotemporal relationships as well as learning with arbitrary shapes. We focus our evaluation on two real-world case studies using our technique: predicting tornadoes in Oklahoma and predicting aircraft turbulence in the United States. We also discuss how to evaluate success for a machine learning algorithm in the severe weather domain, which will enable new methods such as ours to transfer from research to operations, provide a set of lessons learned for embedded machine learning applications, and discuss how to field our technique.