科研成果

In this work, we develop a distributed least-square approximation (DLSA) method that is able to solve a large family of regression problems (e.g., linear regression, logistic regression, and Cox’s model) on a distributed system. By approximating the local objective function using a local quadratic form, we are able to obtain a combined estimator by taking a weighted average of local estimators. The resulting estimator is proved to be statistically as efficient as the global estimator. Moreover, it requires only one round of communication. We further conduct a shrinkage estimation based on the DLSA estimation using an adaptive Lasso approach. The solution can be easily obtained by using the LARS algorithm on the master node. It is theoretically shown that the resulting estimator possesses the oracle property and is selection consistent by using a newly designed distributed Bayesian information criterion. The finite sample performance and computational efficiency are further illustrated by an extensive numerical study and an airline dataset. The airline dataset is 52 GB in size. The entire methodology has been implemented in Python for a de-facto standard Spark system. The proposed DLSA algorithm on the Spark system takes 26 min to obtain a logistic regression estimator, which is more efficient and memory friendly than conventional methods. Supplementary materials for this article are available online.

Links between the optical properties and chemical compositions of brown carbon (BrC) are poorly understood because of the complexity of BrC chromophores. We conducted field studies simultaneously at both vehicle-influenced site and biomass burning-affected site in China in polluted winter. The chemical compositions and light absorption values of functionalized aromatic compounds, including phenyl aldehyde, phenyl acid, and nitroaromatic compounds, were measured. P-phthalic acid, nitrophenols and nitrocatechols were dominant BrC species, accounting for over 50% of the concentration of identified chromophores. Nitrophenols and nitrocatechols contributed more than 50% of the identified BrC absorbance between 300 and 400 nm. Oxidation of biomass burning-related products (e.g., pyrocatechol and methylcatechols) and anthropogenic volatile organic compounds (e.g., benzene and toluene) generated similar BrC chromophores, implying that these functionalized aromatic compounds play an important role in both environments. Compared with the biomass burning-affected site (22%), functionalized aromatic compounds at vehicle-influenced site accounted for a higher percentage of BrC absorption (25%). This research improves our understanding of the links between optical properties and composition of BrC, and the difference between BrC chromophores from BB-influenced area and vehicle-affected area under polluted atmospheric conditions. © 2021 Elsevier B.V.

Background Air pollutants, particularly fine particulate matters (PM2.5) have been associated with mental disorder such as depression. Clean air policy (CAP, i.e., a series of emission-control actions) has been shown to reduce the public health burden of air pollutions. There were few studies on the health effects of CAP on mental health, particularly, in low-income and middle-income countries (LMICs). We investigated the association between a stringent CAP and depressive symptoms among general adults in China. Methods We used three waves (2011, 2013 and 2015) of the China Health and Retirement Longitudinal Study (CHARLS), a prospective nationwide cohort of the middle-aged and older population in China. We assessed exposure to PM2.5 through a satellite-retrieved dataset. We implemented a difference-in-differences (DID) approach, under the quasi-experimental framework of the temporal contrast between 2011 (before the CAP) and 2015 (after the CAP), to evaluate the effect of CAP on depressive symptoms. The association was further explored using a mixed-effects model of the three waves. To increase the interpretability, the estimated impact of PM2.5 was compared to that of aging, an established risk factor for depression. Findings Our analysis included 15,954 participants. In the DID model, we found a 10-µg/m3 reduction of PM2.5 concentration was associated with a 4.14% (95% CI: 0.41–8.00%) decrement in the depressive score. The estimate was similar to that from the mixed-effects model (3.63% [95% CI, 2.00–5.27%]). We also found improved air quality during 2011–2015 offset the negative impact from 5-years’ aging. Interpretation The findings suggest that implementing CAP may improve mental wellbeing of adults in China and other LMICs.

A set-pair Lovász extension is established to construct equivalent continuous optimization problems for graph k-cut problems.

In the present work, we performed chassis dynamometer experiments to investigate the emissions and secondary organic aerosol (SOA) formation potential of intermediate-volatility organic compounds (IVOCs) from an on-road Chinese gasoline vehicle. High IVOC emission factors (EFs) and distinct volatility distribution were recognized. The IVOC EFs for the China V vehicle ranged from 12.1 to 226.3 mg per kilogram fuel, with a median value of 83.7 mg per kilogram fuel, which was higher than that from US vehicles. Besides, a large discrepancy in volatility distribution and chemical composition of IVOCs from Chinese gasoline vehicle exhaust was discovered, with larger contributions of B14-B16 compounds (retention time bins corresponding to C14-C16 n-alkanes) and a higher percentage of n-alkanes. Further we investigated the possible reasons that influence the IVOC EFs and volatility distribution and found that fuel type, starting mode, operating cycles and acceleration rates did have an impact on the IVOC EF. When using E10 (ethanol volume ratio of 10 %, v=v) as fuel, the IVOC EF of the tested vehicle was lower than that using commercial China standard V fuel. The average IVOC-to-THC (total hydrocarbon) ratios for gasoline-fueled and E10-fueled gasoline vehicles were 0:07 ± 0:01 and 0:11 ± 0:02, respectively. Cold-start operation had higher IVOC EFs than hot-start operation. The China Light-Duty Vehicle Test Cycle (CLTC) produced 70 % higher IVOCs than those from the Worldwide Harmonized Light Vehicles Test Cycle (WLTC). We found that the tested vehicle emitted more IVOCs at lower acceleration rates, which leads to high EFs under CLTC. The only factor that may influence the volatility distribution and compound composition is the engine aftertreatment system, which has compound and volatility selectivity in exhaust purification. These distinct characteristics in EFs and volatility may result in higher SOA formation potential in China. Using published yield data and a surrogate equivalent method, we estimated SOA formation under different OA (organic aerosol) loading and NOx conditions. Results showed that under low-and high-NOx conditions at different OA loadings, IVOCs contributed more than 80 % of the predicted SOA. Furthermore, we built up a parameterization method to simply estimate the vehicular SOA based on our bottomup measurement of VOCs (volatile organic compounds) and IVOCs, which would provide another dimension of information when considering the vehicular contribution to the ambient OA. Our results indicate that vehicular IVOCs contribute significantly to SOA, implying the importance of reducing IVOCs when making air pollution controlling policies in urban areas of China. © 2021 Author(s).

To investigate the composition, variation, and sources of nitrated phenols (NPs) in the winter of Beijing, gas-phase NPs were measured by a chemical ionization long time-of-flight mass spectrometer (CI-LToF-MS). A box model was applied to simulate the secondary formation process of NPs. In addition, the primary sources of NPs were resolved by a non-negative matrix factorization (NMF) model. Our results showed that secondary formation contributed 38 %, 9 %, 5 %, 17 %, and almost 100% of the nitrophenol (NP), methyl-nitrophenol (MNP), dinitrophenol (DNP), methyl-dinitrophenol (MDNP or DNOC), and dimethyl-nitrophenol (DMNP) concentrations. The phenol-OH reaction was the predominant loss pathway (46.7 %) during the heavy pollution episode, which produced the phenoxy radical (C6H5O). The phenoxy radical consequently reacted with NO2 and produced nitrophenol. By estimating the primarily emitted phenol from the ratio of phenol=CO from freshly emitted vehicle exhaust, this study proposed that oxidation of primary phenol contributes much more nitrophenol (37 %) than that from benzene oxidation (< 1 %) in the winter of Beijing. The latter pathway was widely used in models and might lead to great uncertainties. The source apportionment results by NMF indicated the importance of combustion sources (> 50 %) to the gas-phase NPs. The industry source contributed 30% and 9% to DNP and MDNP, respectively, which was non-negligible. The concentration weighted trajectory (CWT) analysis demonstrated that regional transport from provinces that surround the Yellow and Bohai seas contributed more primary NPs to Beijing. Both primary sources and secondary formation at either local or regional scale should be considered when making control policies of NPs. © 2021 Kai Song et al.