Combining Machine Learning and Satellite Observations to Predict Spatial and Temporal Variation of near Surface OH in North American Cities

• DOI: 10.1021/acs.est.1c05636
• Publication Date: 2022-03-18T18:17:17Z
• Authors: Qindan Zhu; Joshua L. Laughner; Ronald C. Cohen
• Keywords: Isoprene; Atmospheric chemistry; Chemical Transport Model

The hydroxyl radical (OH) is the primary cleansing agent in atmosphere. abundance of OH cities initiates removal local pollutants; therefore, it serves as key species describing urban chemical environment. We propose a machine learning (ML) approach an efficient alternative to simulation using computationally expensive transport model. ML model trained on parameters simulated from WRF-Chem model, and suggests that six predictive are capable explaining 76% variability. tropospheric NO2 column, HCHO J(O1D), H2O, temperature, pressure. then use observations column OMI input enable measurement-based prediction daily near surface at 1:30 pm time across 49 North American over course 10 years between 2005 2014. result validated by comparing predictions measurements isoprene, which has source uncorrelated with removed rapidly almost exclusively daytime. demonstrate predicted is, expected, anticorrelated isoprene. also show this consistent our understanding chemistry given solely data-driven nature.