Session Presentations

Youn-Seo KooㆍDae-Ryun ChoiㆍHui-Young YunㆍGeon-Woo YoonㆍJoung-Beom LeeㆍHee Yong KwonㆍSuk Hyun YuㆍJi-Seok Koo

Ambient exposure to PM2.5 is harmful effect on public health and the forecasting PM2.5 is essential to protect the public health in advance. Current PM2.5 forecasting systems in operation are mainly based on the physicochemical models such as CMAQ and WRF. The forecasting accuracy of them however, have substantial limits due to uncertainty of input data of anthropogenic emissions and meteorological fields as well as the physicochemical models themselves. In order to overcome the drawback of the prognostic model predictions and to take advantage of the recent advance in machine learning algorithm, the PM2.5 forecasting system using physicochemcal models of CMAQ and WRF as well as ensemble machine learning method is being developed and tested for national PM2.5 forecasting in Korea. The various levels of big data based on observation and model outputs and machine learning methods to reflect temporal variations and spatial characteristic PM2.5 between the source and receptor regions.

Jiani Yang, Yifan Wen, Yuan Wang, Shaojun Zhang, Joseph P. Pinto, Elyse A. Pennington, Zhou Wang, Ye Wu, Stanley P. Sander, Jonathan H. Jiang, Jiming Hao, Yuk L. Yung, and John H. Seinfeld

The large fluctuations in traffic during the COVID-19 pandemic provide an unparalleled opportunity to assess vehicle emission control efficacy. Here we develop a random-forest regression model, based on the large volume of real-time observational data during COVID-19, to predict surface-level NO₂, O₃, and fine particle concentration in the Los Angeles megacity. Our model exhibits high fidelity in reproducing pollutant concentrations in the Los Angeles Basin and identifies major factors controlling each species. During the strictest lockdown period, traffic reduction led to decreases in NO₂ and particulate matter with aerodynamic diameters <2.5 μm by –30.1% and –17.5%, respectively, but a 5.7% increase in O₃. Heavy-duty truck emissions contribute primarily to these variations. Future traffic-emission controls are estimated to impose similar effects as observed during the COVID-19 lockdown, but with a smaller magnitude. Vehicular electrification will achieve further alleviation of NO₂ levels.

Yannic Lops, Arman Pouyaei, Masoud Ghahremanloo, Yunsoo Choi, Arman Pouyaei, Jia Jung

Nitrogen Oxides (NO_X) are highly reactive and are one of the major criteria pollutants with significant negative impact on human health and the environment. The spatial coverage and quality of satellite remote sensing Nitrogen Dioxide (NO₂) measurements are essential as an indicator for NO_X concentrations. Unfortunately, satellite-derived measurements are negatively impacted by surface reflectivity and cloud cover, which lead to biases. Thus, these measurements would need to be discarded and significantly increase the amount of missing data within a satellite-derived image. This work employs the Partial Convolutional Neural Network to impute missing remote sensing data of the TROPOspheric Monitoring Instrument (TROPOMI) NO₂ measurements efficiently and accurately for the entire Contiguous United States. The neural network model was trained by employing simulated NO₂ from the Community Multiscale Air Quality Modeling System and missing masks from the TROPOMI dataset.

Jiachen Liu, Eric Chen, Shannon L. Capps

Secondary organic aerosols (SOAs) are generated by complex transformations of organic molecule emissions and are often a significant portion of atmospheric particulate matter (PM) concentrations. High PM concentrations can lead to regional haze and pose health risks to the public. Understanding the sensitivities of SOA formation with respect to precursor concentrations is therefore essential for evaluating the highest priority of SOA precursor emissions to reduce. Existing methods of calculating the sensitivity coefficients include the finite difference method (FDM), the direct decoupled method (DDM), and the adjoint method. The FDM suffers from truncation and cancellation errors, while DDM and the adjoint method are relatively difficult to implement and maintain in chemical transport models. Here, we propose an alternative approach to calculating first- and second-order sensitivities. This method is an operator overloading approach which calculates the exact first- and second-order sensitivity coefficients of select precursors through one single run of the model (Fike & Alonso, 2011). We applied this method in the Community Multiscale Air Quality (CMAQ) model v.5.3.2 to formulate the CMAQ-hyd model. In this work, we apply a box model version of the CMAQ-hyd to compute the exact sensitivities of SOA concentrations with respect to their precursor concentrations. The sensitivity coefficient results could aid in evaluating the current Regional Haze Rule and potentially new NAAQS rules for the total atmospheric PM_2.5 concentration in the future.

Zhen Liu, Pingkuan Di, Jeremy Avise

We present results from our ongoing effort of investigating the potential of CMAQ sensitivity analysis in supporting air quality policy making and planning in the state of California. The model code for CMAQ-DDM-3D (v5.2) was adapted with the SAPRC07tic-AERO6i chemical mechanism, and then used for reduced order modeling (ROM), source attribution, and process analysis. A reduced order model (ROM) was constructed to establish first-order source-receptor relationships between emissions of NOx and VOCs from 58 counties in California and gridded surface concentrations for the full suite of CMAQ output species at a 12 km resolution. The ROM is shown to allow efficient projections of changes in air quality due to emission perturbations at 58 counties as resulting from a future regulation, while its accuracy is illustrated by comparing against results from full CMAQ modeling. Source contributions from 58 counties to PM and toxic VOC concentrations are estimated with the ROM, considering both transport and chemistry with a first-order approximation. Finally, we discuss the potential of using sensitivity analysis to help understand chemical and transport processes (e.g., nitrate formation and its dependency on emissions of precursors).