Search for: All records

Estimating the uncertainty of a model’s prediction on a test point is a crucial part of ensuring reliability and calibration under distribution shifts. A minimum description length approach to this problem uses the predictive normalized maximum likelihood (pNML) distribution, which considers every possible label for a data point, and decreases confidence in a prediction if other labels are also consistent with the model and training data. In this work we propose IF-COMP, a scalable and efficient approximation of the pNML distribution that linearizes the model with a temperature- scaled Boltzmann influence function. IF-COMP can be used to produce well-calibrated predictions on test points as well as measure complexity in both labelled and unlabelled settings. We experimentally validate IF-COMP on uncertainty calibration, mislabel detection, and OOD detection tasks, where it consistently matches or beats strong baseline methods. 

Abstract Furans are a major class of volatile organic compounds emitted from biomass burning. Their high reactivity with atmospheric oxidants leads to the formation of secondary organic aerosol (SOA), including secondary brown carbon (BrC) that can affect global climate via interactions with solar radiation. Here, we investigate the optical properties and chemical composition of SOA generated via photooxidation of furfural, 2‐methylfuran, and 3‐methylfuran under dry (RH < 5%) and humid (RH ∼ 50%) conditions in the presence of nitrogen oxides (NO_x) and ammonium sulfate seed aerosol. Dry furfural oxidation has the greatest BrC formation, including reduced nitrogen‐containing organic compounds (NOCs) in SOA, which are dominated by amines and amides formed from reactions between carbonyls and ammonia/ammonium. Based on the products detected, we propose novel formation pathways of NOCs in furfural photooxidation, which can contribute to BrC via accretion reactions during the photochemical aging of biomass burning plumes. 

Abstract. The formation of secondary organic aerosol (SOA) from the oxidation of β-pinene via nitrate radicals is investigated in the Georgia Tech Environmental Chamber (GTEC) facility. Aerosol yields are determined for experiments performed under both dry (relative humidity (RH) < 2 %) and humid (RH = 50 % and RH = 70 %) conditions. To probe the effects of peroxy radical (RO₂) fate on aerosol formation, "RO₂ + NO₃ dominant" and "RO₂ + HO₂ dominant" experiments are performed. Gas-phase organic nitrate species (with molecular weights of 215, 229, 231, and 245 amu, which likely correspond to molecular formulas of C₁₀H₁₇NO₄, C₁₀H₁₅NO₅, C₁₀H₁₇NO₅, and C₁₀H₁₅NO₆, respectively) are detected by chemical ionization mass spectrometry (CIMS) and their formation mechanisms are proposed. The NO⁺ (at m/z 30) and NO₂⁺ (at m/z 46) ions contribute about 11 % to the combined organics and nitrate signals in the typical aerosol mass spectrum, with the NO⁺ : NO₂⁺ ratio ranging from 4.8 to 10.2 in all experiments conducted. The SOA yields in the "RO₂ + NO₃ dominant" and "RO₂ + HO₂ dominant" experiments are comparable. For a wide range of organic mass loadings (5.1–216.1 μg m^&minus;3), the aerosol mass yield is calculated to be 27.0–104.1 %. Although humidity does not appear to affect SOA yields, there is evidence of particle-phase hydrolysis of organic nitrates, which are estimated to compose 45–74 % of the organic aerosol. The extent of organic nitrate hydrolysis is significantly lower than that observed in previous studies on photooxidation of volatile organic compounds in the presence of NO_x. It is estimated that about 90 and 10 % of the organic nitrates formed from the β-pinene+NO₃ reaction are primary organic nitrates and tertiary organic nitrates, respectively. While the primary organic nitrates do not appear to hydrolyze, the tertiary organic nitrates undergo hydrolysis with a lifetime of 3–4.5 h. Results from this laboratory chamber study provide the fundamental data to evaluate the contributions of monoterpene + NO₃ reaction to ambient organic aerosol measured in the southeastern United States, including the Southern Oxidant and Aerosol Study (SOAS) and the Southeastern Center for Air Pollution and Epidemiology (SCAPE) study.