Traffic-related air pollution (TRAP), including tailpipe emissions and non-exhaust emissions, is a major health concern in urban areas. As tailpipe emissions decrease due to regulatory efforts, non-exhaust emissions have become a more significant source of pollution. Machine learning (ML) has been widely used to characterize and quantify traffic-related air pollutants, enhancing our understanding of their compositions and spatiotemporal distributions. We hence identified and reviewed over 50 peer-reviewed publications from 2020 to 2024 on applying ML for TRAP. Many recent studies are related to the emerging interest in non-exhaust emissions, which have not been summarized before. Thus, this review provides a structured synthesis of ML applications across key TRAP modeling, specifically pertaining to the underrepresented domain of non-exhaust emissions including: (1) TRAP spatial modeling, (2) identification of contributing factors to TRAP, (3) characterization of non-exhaust emissions, and (4) source tracking and apportionment. Across the literature, several persistent challenges are identified, such as data sparsity, inconsistent feature representation, and limited causal interpretability. The review also highlights methodological trends in TRAP modeling, integration of heterogeneous data, and model interpretability approaches that influence the reliability of ML-based TRAP models. With future advances in ML methods, these approaches will play an increasingly crucial role in improving TRAP exposure assessment and research.

1,4-dioxane (DX) is a compound used in the manufacturing of many synthetic compounds in dyes, cosmetics, and food additives. DX is a suspect carcinogen, and its high solubility and persistence in groundwater presents a challenge for traditional remediation efforts. Bioremediation offers a potential solution, as some strains of bacteria have been shown to be able to degrade DX. Among these is Pseudonocardia dioxanivorans CB1190, a bacteria known for its ability to use DX as a food source. Calculating the kinetics of DX degradation by CB1190 is key to bioremediation efforts, but the microbe’s gram-positive structure and flaky morphology pose challenges to accurate biomass collection and protein quantification. Prior research examining the kinetics of other DX-degrading bacteria have used various methods, with no definitive recommendation for CB1190. This research examined several methods of quantifying CB1190 biomass through protein concentration. Cell biomass from CB1190 grown on DX was extracted through sonification and dry weight measurement before total protein concentration was subsequently quantified using the Lowry protein assay and Bradford assay. Protein concentrations were then plotted against CB1190 16S DNA copies/mL. The Lowry protein assay showed the most accurate standard curve relationship between copies/mL and protein mass, indicating its efficacy for use in calculating the updated Vmax and Km values for DX degradation by CB1190, as well as future kinetics experiments involving CB1190.

Water quality in Santa Monica Bay (SMB) poses significant health risks for the public, particularly following rainfall events in Los Angeles. Fecal indicator bacteria, among other pollutants, contaminate runoff that is flushed out into the coastal ocean affecting communities that utilize SMB. Current Los Angeles public health guidelines recommend a uniform 72-hour “no-contact” period after rainfall. However, previous observational data and modeling work found that parts of SMB exceed the three day period. In this work, we analyze pollutant residence time using a 100 m resolution Regional Ocean Modeling System (ROMS) simulation of SMB with realistic river and storm-drain discharges, tides, wind, and surface wave effects. Using salinity as a proxy for pollutant concentration, we observe runoff plumes at fifteen outlets in the Bay to assess salinity level recovery time. Adding on to work done previously, we include surface wave forcing to better represent coastal dynamics. Results show strong variability of residence time, demonstrating that the current 72-hour period cannot be applied uniformly. Continuous runoffs such as Malibu Creek and Ballona Creek, display residence time around 6-7 days; while more exposed regions present recovery time between 2-4 days. Including wave effects reaffirms that exposure risk could last up to one week following rainfall. We emphasize the importance of increasing community sampled measurements and high resolution modeling to improve public health advisories in Los Angeles.