Building Interpretable Machine Learning Models to Identify Chemometric Trends in Seabirds of the North Pacific Ocean.

Marine environmental monitoring efforts often rely on the bioaccumulation of persistent anthropogenic contaminants in organisms to create a spatiotemporal record of the ecosystem. Intercorrelation results from the origin, uptake, and transport of these contaminants throughout the ecosystem and may be affected by organism-specific processes such as biotransformation. Here, we explore trends that machine learning tools reveal about a large, recently released environmental chemistry data set of common anthropogenic pollutants measured in the eggs of five seabird species from the North Pacific Ocean. We modeled these data with a variety of machine learning approaches and found models that could accurately determine a range of taxonomic and spatiotemporal trends. We illustrate a general workflow and set of analysis tools that can be used to identify interpretable models which perform nearly as well as state-of-the-art "black boxes." For example, we found shallow decision trees that could resolve genus with greater than 96% accuracy using as few as two analytes and a k-nearest neighbor classifier that could resolve species differences with more than 94% accuracy using only five analytes. The benefits of interpretability outweighed the marginally improved accuracy of more complex models. This demonstrates how machine learning may be used to discover rational, quantitative trends in these systems.

Levels and profiles of perfluorinated alkyl acids in liver tissues of birds with different habitat types and trophic levels from an urbanized coastal region of South Korea.

Contamination status and characteristics of perfluorinated alkyl acids (PFAAs) including perfluorinated carboxylic acids (PFCAs) and perfluorinated sulfonic acids (PFSAs) was examined using liver tissue of birds - black-tailed gulls (Larus crassirostris), domestic pigeons (Columba livia var. domestica), pacific loons (Gavia pacifica), herons (Ardea cinerea), and egrets (Egretta garzetta and Ardea alba) - with different trophic levels, habitat types and migratory behaviors from an industrialized coastal region of South Korea. A wide range of PFAAs (1.09 ng/g to 1060 ng/g; median = 52.6 ng/g) were detected in bird livers from the Korean coasts with high detection frequency. Accumulation features of PFAAs in birds indicated that primarily trophic position and secondly habitat type influence the levels and composition of PFAAs, e.g., relatively high PFAA levels and high composition of odd-numbered long carbon chain PFCAs (perfluoroundecanoic acid (PFUnDA) and perfluorotridecanoic acid (PFTriDA)) and PFOS in higher trophic and marine birds. The prevalence of long carbon chain (≥14) PFCAs likely implies a wide use of fluorotelomer-based substances in Korea. Interspecies comparison in the accumulation profile of persistent organic pollutants (including polychlorinated biphenyls (PCBs), organochlorine pesticides, polybrominated diphenylethers (PBDEs), and PFAAs) reveals relatively high load of PFAAs in inland (pigeons) and estuarine (egrets/herons) species compared to marine bird species, indicating wide use of PFAAs in the terrestrial environment.

Relationship between the Pacific Decadal Oscillation (PDO) and persistent organic pollutants in sympatric Alaskan seabird (Uria aalge and U. lomvia) eggs between 1999 and 2010.

Although climate change occurs alongside other anthropogenic ecosystem impacts, little is known about how sea-surface temperature variability influences the ecotoxicology of persistent organic pollutants (POPs). We analyzed POP contaminant levels, and stable isotopes δ15N and δ13C as measures of trophic position, in eggs collected from the Gulf of Alaska and Bering Sea between 1999 and 2010 from two similar avian species with different trophic positions: common murres (Uria aalge) and thick-billed murres (Uria lomvia). The ebb and flow of the Pacific Decadal Oscillation (PDO), a long-lived El Niño-like pattern of climate variability in the Pacific Ocean, predicted both trophic position and polychlorinated biphenyl (PCB) levels in thick-billed murres, but not in common murres. There was a similar pattern of association of the PDO with organochlorine pesticide levels in thick-billed murres, but not in common murres. The magnitude of association in thick-billed murres of PDO with the level of a specific PCB congener was a function of the number of chlorine groups on the PCB congener. Although this statistical analysis does not account for all factors contributing to climate variation, this contrast between the species suggests that facultative changes in foraging behavior, reflected in trophic position, can determine how POPs flow through and thereby alter ecosystems under climate change.