Ozone pollution has become an increasingly severe issue in North China. To address the challenges of insufficient prediction accuracy under high-concentration conditions and imbalanced sample distribution, this study proposes a hybrid model named EX-PIM-XGB(Extreme sample enhancement - Permutation Importance Method Extreme Gradient Boosting), which integrates feature selection with extreme-sample augmentation for ozone forecasting. Using air quality and meteorological data from 79 monitoring stations during 2020-2024, the dataset was divided into four seasonal subsets-spring, summer, autumn, and winter-for model development. Results indicate that NO2 consistently exhibited high and stable importance across all seasons, with a maximum score of 2.78, followed by T2M and D2M, while UVB, CDIR, PM10,PM2.5 exhibited stronger seasonal variation. The R2 values of the spring, summer, autumn, and winter models were 0.90, 0.88, 0.93, and 0.86, and the RMSEs were 17.53, 19.93, 17.50, and 10.67. The enhanced model demonstrated significantly improved performance in high-concentration ranges; for instance, in summer, R2 increased from 0.79 to 0.90, RMSE decreased by 16%, and prediction accuracy improved by over 22%, outperforming both the conventional XGBoost model and the PIM-based variable selection model. These findings suggest that combining feature stability screening with extreme-sample augmentation effectively enhances the model's responsiveness to extreme pollution events, offering methodological insights and data-driven support for ozone pollution warning and control.

The advancement of explainable artificial intelligence (XAI) has emerged as a pivotal tool for unraveling the intrinsic mechanisms of deep learning-based models for predicting air quality. The Layer-wise Relevance Propagation (LRP) method enables the quantification of temporally resolved feature contributions in deep learning (DL) models predicting ozone (O₃) concentrations. In this study, a Random Forest-corrected (RF-corrected) model was employed to generate 9-km gridded O₃ data for the Yangtze River Delta (YRD) region from 2020 to 2023, filling gaps in unmonitored areas. These gridded data, combined with WRF-modeled meteorological parameters and ground observation data, served as inputs for an attention-based sequence-to-sequence (seq2seq) model to predict O₃ concentrations at 24-h (24 h), 48-h (48 h), and 72-h (72 h). For the 24 h predictions, the model demonstrates robust performance, with R of 0.85 (test set) and 0.89 (validation set), alongside RMSE of 19.81 μg/m3 and 19.22 μg/m3, respectively. LRP analysis revealed annual average contributions of 24.6 % from gridded features, 38.2 % from meteorological features, and 17.0 % from pollutant features to O₃ predictions. During 13:00-15:00, contributions increased to 30.0 %, 35.0 %, and 15.3 %, respectively, aligning with midday photochemical O₃ production and regional transport under favorable meteorological conditions. Further analysis integrating Potential Source Contribution Function (PSCF) values and industrial zone locations near monitoring sites demonstrated that gridded feature contributions effectively explain the influence of O₃ sources from different wind directions. These findings highlight that deep learning-derived feature insights, combined with spatial and chemical context, provide novel perspectives for identifying emission sources and transport mechanisms driving O₃ variability.

Extensive spatiotemporal analyses of long-trend surface ozone in the Yangtze River Delta (YRD) region and its meteorology-related and emission-related have not been systematically analyzed. In this study, by using 8-year-long (2015-2022) surface ozone observation data, we attempted to reveal the variation of multiple timescale components using the Kolmogorov-Zurbenko filter, and the effects of meteorology and emissions were quantitatively isolated using multiple linear regression with meteorological variables. The results showed that the short-term, seasonal, and long-term components accounted for daily maximum 8-hr average O3 (O3-8 hr) concentration, 46.4%, 45.9%, and 1.0%, respectively. The meteorological impacts account for an average of 71.8% of O3-8 hr, and the YRD's eastern and northern sections are meteorology-sensitive areas. Based on statistical analysis technology with empirical orthogonal function, the contribution of meteorology, local emission, and transport in the long-term component of O3-8 hr were 0.21%, 0.12%, and 0.6%, respectively. The spatiotemporal analysis indicated that a distinct decreasing spatial pattern could be observed from coastal cities towards the northwest, influenced by the monsoon and synoptic conditions. The central urban agglomeration north and south of the YRD was particularly susceptible to local pollution. Among the cities studied, Shanghai, Anqing, and Xuancheng, located at similar latitudes, were significantly impacted by atmospheric transmission-the contribution of Shanghai, the maximum accounting for 3.6%.

The proportion of more active volatile organic compounds (VOCs) species decreases with altitude, whereas radiation increases vertically due to aerosol scattering near the surface, exerting opposite effects on vertical O 3 generation. Using the observation‐based model (OBM) constrained by vertical profiles, we identified that the aerosol radiation effect (ARE) has a stronger impact on photochemical characteristics and O 3 ‐NOx‐VOC sensitivity than the VOCs reactivity effect (VRE). ARE is the dominant factor, promoting the formation of O 3 and ·OH (increases by 36.7% and 106.3%), enhancing high‐altitude oxidation, and shifting parts of the VOC‐limited regimes to the NOx‐limited regimes (VOCs/NOx ratio decreases by 21.1%). Additionally, rapid NOx depletion with altitude leads to stronger NOx limitation aloft, amplifying ARE's effect on O 3 ‐NOx‐VOC sensitivity (ratio decreases by 28.1%). These findings improve the understanding of vertical ozone generation conditions and suggest that more attention should be paid to vertical environments in surface ozone management.

Based on the observational data of volatile organic compounds （VOCs）, conventional air pollutants, and ERA5 meteorological reanalysis data at three sites, namely, Caochangmen （CCM）, Pukou （PK）, and Xianlin University Town （XL）, in Nanjing from 2015 to 2021, the ozone generation and depletion mechanisms in ozone-polluted days under stable weather conditions were investigated using the observation-based model （OBM-MCM）. The results showed that ① Significant year-by-year differences exist in the frequency of stable weather on ozone-polluted days for the three sites. The maximum number of stable days occurred in 2019, with 46 d （66.7%）, 50 d （64.9%）, and 54 d （69.2%） at the CCM, PK, and XL sites, respectively. ② Significant differences exist between the net O3 production rates for the CCM, PK, and XL sites during the polluted period, with the highest rate of 2.5×10-9 h-1 at the CCM site and the lowest rate of 1.4×10-9 h-1 at the XL site. Additionally, the O3 production and depletion rate at the XL site were lower compared to those at the other two sites. ③ The reactions of HO2·+NO and ·OH+NO2, respectively, contributed the most to O3 production and depletion. The HO2·+NO reaction contributed to O3 production by 69% （CCM）, 68% （PK）, and 71% （XL）, and the ·OH+NO2 reaction contributed to O3 depletion by 67% （CCM）, 63% （PK）, and 62% （XL）. ④ The modeling study observed that ozone pollution under stable weather conditions was mainly affected by local photochemistry processes； therefore, local emission reduction is very important for O3 pollution mitigation.

Based on in-situ vertical observations of volatile organic compounds (VOCs) in the lower troposphere (0–1.0 km) in Nanjing, China, during the summer and autumn, we analyzed the VOCs vertical profiles, diurnal variation, and their impact factors in meteorology and photochemistry. The results showed that almost all the concentrations of VOC species decreased with height, similar to the profiles of primary air pollutants, as expected. However, we found the ratios of inactive species (e.g., acetylene) and secondary VOCs (e.g., ketones and aldehydes) in total VOCs (TVOCs) increased with height. Combined with satellite-retrieved data, we found the average HCHO tropospheric column concentrations were 2.0 times higher in the summer than in the autumn. While the average of tropospheric NO2 column concentrations was 3.0 times lower in the summer than in the autumn, the seasonal differences in the ratio of oxygenated VOCs (OVOCs) to NO2 (e.g., HCHO/NO2) shown in TROPOMI satellite-retrieved data were consistent with in-situ observations (e.g., acetone/NO2). On average, during autumn daytime, the mixing layer (ML), stable boundary layer (SBL), and residual layer (RL) had OH loss rates (LOH) of 6.9, 6.3, and 5.5 s−1, respectively. The LOH of alkenes was the largest in the ML, while the LOH of aromatics was the largest in the SBL and RL. At autumn night, the NO3 loss rates (LNO3) in the SBL and RL were 2.0 × 10−2 and 1.6 × 10−2 s−1, respectively, and the LNO3 of aromatics was the largest in the SBL and RL. In the daytime of summer, the LOH of VOCs was ~40% lower than that in autumn in all layers, while there was no significant difference in LNO3 at night between the two seasons. This study provides data support and a theoretical basis for VOC composite pollution control in the Nanjing region.

There were 6 severe haze events over a large area of the Yangtze River Delta (YRD) region in January 2013. In this study, based on the hourly concentrations of trace gases and PM2.5 at 10 observation stations (8 city stations, 1 regional background station and 1 island station) during Jan. 1–31, 2013 as well as the concentrations of water-soluble ions at 5 stations (4 city stations and 1 regional background station) during Jan. 18–24, 2013 in the YRD region, the regional characteristics of the air pollutants during heavy haze episodes were investigated in combination with the atmospheric circulation patterns. The concentrations of PM2.5 on haze days were 1.6–2.4-fold higher than on clear days. The concentration of PM2.5, SO2, NO2 and CO increased significantly, with average values of 128.6, 48.5, 78.1 µg m−3 and 1.5 mg m−3 on haze days, and were 64.6, 36, 52.5 µg m−3 and 1.1 mg m−3 on clear days. The PM2.5 concentration of ten observation sites had positive correlations with CO and NO2, and had weakly negative correlations with O3. The sources of PM2.5, SO2, NO2 and CO were strong in inland cities and weak in coastal cities, and the sources of O3 were mainly from Wuxi, Suzhou and southeast of An’hui. The mass and water-soluble ion concentrations were both centralized in PM2.1 during the haze events; additionally, the NH4+, SO42− and NO3− ions were dominant, constituting 86–90.9

In this study, a Kolmogorov-Zurbenko （KZ） filter was proposed to decompose the original ozone （O3） sequence to improve the accuracy of ozone long-term series prediction and select relevant meteorological features. Furthermore, the enhanced maximal minimal redundancy （mRMR） feature selection technique was combined with the support vector regression （SVR） approach to select the most illuminating meteorological features. Subsequently, from May to August 2023, during high ozone concentration periods, a long short-term memory network （LSTM） was utilized to assess and predict high ozone concentration periods at the monitoring stations of Jingan （urban area）, Pudong-Chuansha （suburban area）, and Dianshan Lake （suburban area） in Shanghai. The results showed that pressure, temperature, humidity, boundary layer height, and wind direction were the best combinations of O3 baseline and short-term components, as chosen by feature screening. The R2 values for Jingan Station, Pudong-Chuansha Station, and Dianshan Lake Station were 0.86, 0.83, and 0.85, respectively. The RMSE values were 18.26, 18.74, and 20.02 μg·m-3, respectively. These findings suggest that decomposing the original O3 sequence improved the prediction accuracy of ozone concentrations. Additionally, as indicated by the R2 and RMSE values found for every monitoring station, feature screening preserved the model's predictive performance.

Aerosol physicochemical properties during two dust storms (DS1: March 30-31, and DS2: May 7-8) are measured in 2021 in Nanjing, aiming to investigate the impacts of dust storms on aerosol chemical compositions and optical hygroscopicity in the Yangtze River Delta (YRD). During DS1, the dust air masses are transported in the lower atmosphere and pass through the inlands and sea areas before reaching the YRD region. During DS2, the dust air masses are transported in the upper atmosphere and pass through the inlands only. Both of them are accompanied by an increase in black carbon (BC) mass fraction and a decrease in nitrate mass fraction in fine particles (PM2.5, particles with diameters less than 2.5 mu m) near the surface. However, the impacts on the mass fractions of organics or sulfate are adverse between DS1 and DS2. During dust-influence periods the enlarged mass ratios of sulfate to nitrate (greater than 4) promote the occurrence of deliquescent behavior of ambient aerosols although dust aerosols can significantly suppress aerosol optical hygroscopicity. To improve model simulations of aerosol optical hygroscopicity, it is necessary to use a segment parameterization to describe aerosol light scattering enhancement factor during dust-influence periods. The closure study of optical hygroscopicity parameters with two different methods reveals the impact of aerosol scattering angstrom ngstrom exponent on the estimation of optical hygroscopicity parameter. The results highlight that the impact of dust storms on aerosol chemical compositions and optical hygroscopicity are different for DS events with different transport pathways. Dust storm (DS) is one of the severest natural disasters. Studying the change of aerosol properties during DS is important to understand the role of dust in the earth-atmosphere system. This study chooses two DS events (DS1 and DS2) in March and May 2021 to investigate the impacts of DS on aerosol chemical compositions and optical hygroscopicity in the YRD region. It is found that DS1 transporting in the lower atmosphere makes a larger increase of coarse-mode particles near the surface than DS2 transporting in the upper atmosphere. The weak impact of DS2 on the surface aerosols leads to a small change in aerosol chemical compositions of fine particles near the surface. The enlarged mass ratio of sulfate to nitrate (greater than 4) during dust periods can promote ambient aerosol deliquescence, which is important to quantify aerosol optical properties. To provide a reference for improving the model simulation of aerosol optical hygroscopicity, different parameterization schemes are derived to fit the change of aerosol light scattering enhancement factor with RH during dust and no-dust periods. The closure study of aerosol optical hygroscopicity parameters suggests that aerosol scattering angstrom ngstrom exponent plays an important role in the overestimation of optical hygroscopicity parameters. The dust storms influencing the YRD region have different transport pathways and impacting atmospheric layersThe enlarged mass ratio of sulfate to nitrate in fine particles makes the deliquescence phenomenon appear more in dust-influence periodsAerosol scattering angstrom ngstrom exponent is an important factor to estimate optical hygroscopicity parameters

Based on the sounding data of VOCs in the lower troposphere (0-1000 m) in the northern suburb of Nanjing in the autumn of 2020, the vertical profile distribution, diurnal variation, and photochemical reactivity of VOCs in this area were analyzed. The results showed that the volume fraction of VOCs decreased with the increase in height (72.1×10-9±28.1×10-9-56.4×10-9±24.8×10-9). Alkanes at all heights accounted for the largest proportion (68%-75%), followed by aromatics (10%-12%), halohydrocarbons (10%-11%), alkenes (3%-7%), and acetylene (2%). The diurnal variation of the boundary layer had a great influence on the VOCs profile. The lower boundary layer in the morning and evening caused the volume fraction of VOCs to accumulate near the ground and lower in the upper layer. The vertical distribution of VOCs was more uniform in the afternoon. In the morning, the volume fraction proportion of alkenes (alkanes) with strong (weak) photochemical reactivity decreased (increased) with the increase in height, indicating that the photochemical aging of VOCs in the upper layer was significant. In the afternoon, the vertical distribution of VOCs volume fraction and OFP in the lower troposphere were more uniform. Affected by the surrounding air masses with different sources, the volume fraction and component proportion of VOCs at each height were significantly different. The alkanes in rural air masses were vertically evenly distributed, and the proportion increased gradually with the height. The vertical negative gradient of VOCs volume fraction in the urban air mass was the largest, the volume fraction of VOCs near the ground was high, and it was rich in aromatics. The proportion of aromatics increased with the increase in VOCs volume fraction between 200-400 m height of industrial air mass. The near-surface VOCs volume fraction of the highway traffic air mass was high, and alkanes accounted for the largest proportion.

With the development and improvement of optimization algorithms, machine learning models have become more important in air quality prediction. These models can be used to address problems associated with atmospheric environmental pollutants like ozone and facilitate the development of appropriate control policies. Firstly, the data sets of meteorological and pollution in Nanjing and four nearby cities from 2018 to 2021 were divided into four sections to build mixed models. The extreme gradient boosting (XGBoost) algorithm was improved by the permutation importance method (PIM) and Pearson correlation coefficient to acquire the most important features of the seasonal model and analysis of the effect on ozone prediction. The results reflect that the NO2, PM2.5, and PM10 parameters have an important influence, accounting for an average of 23% of prediction, and the average influence of the meteorological parameters of the solar radiation exceeds 11%. Secondly, the particle swarm optimization (PSO), grey wolf optimizer (GWO), and ant lion optimizer (ALO) algorithms were used to optimize the SVR model parameters to predict hourly ozone concentration in January, April, July, and September 2022. The R2 values are 0.95, 0.92, 0.85, and 0.84, respectively, and the RMSE values are 5.2, 12.4, 18.0 and 16.6 μg/m3, respectively. Finally, the impact of hourly ozone concentration on population health was studied by a risk assessment of the population density-weighted pollution exposure. The result shows that the influence on humans gradually decreased yearly from 2018 to 2021, the average variation from 2.6% to −1.1%, and health risks during the evening rush hours are still prevalent and require further attention.

Based on the air quality data and conventional meteorological data of the Nanjing Region from January 2015 to December 2016, to analyze the characteristics of O3 concentration changes in the Nanjing Region, a light gradient boosting machine (LightGBM) model was established to predict O3 concentration. The model was compared with three machine learning methods that are commonly used in air quality prediction, including support vector machine, recurrent neural network, and random forest methods, to verify its effectiveness and feasibility. Finally, the performance of the prediction model was analyzed under different meteorological conditions. The results showed that the variation in O3 concentration in Nanjing had significant seasonal differences and was affected by a combination of its pre-concentration, meteorological factors, and other air pollutant concentrations. The LightGBM model predicted the ground-level O3 concentration in the Nanjing area more precisely to a large extent (R2=0.92), and the model outperformed other models in prediction accuracy and computational efficiency. In particular, the model showed a significantly higher prediction accuracy and stability than that of other models under a high-temperature condition that was more likely prone to ozone pollution. The LightGBM model was characterized by its high prediction accuracy, good stability, satisfactory generalization ability, and short operation time, which broaden its application prospect in O3 concentration prediction.

天气雷达实验中三维回波的显示与交互能够提升学生对雷达探测方式和天气学分析的认知.根据雷达探测原理及其数据结构特点,提出极坐标系下三角面片顶点位置的计算方法,简化了坐标转换和空间插值;运用相邻体素状态的快速判别算法和三角剖分构型的判定算法,进一步提升计算性能.测试结果表明,本算法的计算速度是传统移动立方体算法的4倍,内存占用仅为传统算法的56％,能够满足师生教学过程中雷达数据渲染与交互的实时性要求.该算法也可应用于数值天气模式、气象卫星等资料的三维渲染,具有一定的推广应用价值.

Visibility data are fundamental meteorological observation data widely used in many fields. When using visibility data, it is often necessary to calculate the average visibility, which used to be the arithmetic average of the visibility data directly. In this study, we first analyze the relationship between the visibility, the extinction coefficient, and the atmospheric compositions. Then we propose to use the harmonic average of visibility data as the average visibility, which can better reflect changes in atmospheric extinction coefficients and aerosol concentrations. It is recommended to use the harmonic average visibility in the studies of climate change, atmospheric radiation, air pollution, environmental health, etc.

Abstract Aerosol optical effects can trigger complex changes in solar shortwave radiation (SW) in the atmosphere, resulting in significant impacts on the photochemistry and vertical structure of ozone. This paper provides observational evidence of aerosol absorbing and scattering effects on modifying the SW and ozone profiles in the low troposphere. Using field vertical measurements and observation‐based model simulations, we demonstrated that absorbing aerosols decreased SW, resulting in substantial inhibition of ozone production throughout the boundary layer (BL). A similar inhibition effect occurred within the lower BL under sufficient scattering aerosols. However, the scattering augmentation effect played an additional role in enhancing the photolysis rate and promoting ozone generation in the upper BL. Hence, the observational evidence as well as our model simulations disentangled the radiative effects of different types of aerosols on the vertical structures of ozone.

AMA GC5000BTX was used to monitor the mixing ratio of benzene, toluene, ethylbenzene, m,p-xylene, o-xylene, and styrene (BTESX) in the atmosphere of the northern suburb of Nanjing from January 2014 to December 2016. The temporal variation characteristics of BTESX and the influence of meteorological elements on it were analyzed, and the characteristic ratio method (T/B) was used to qualitatively analyze the source of BTESX. Finally, the human exposure analysis and evaluation method of EPA was used to evaluate the health risk of BTESX. The results showed that during the observation period, the average mixing ratio of BTESX was (7.28±6.63)×10-9, and the mixing ratio of benzene was the highest at (2.45±3.91)×10-9. The mixing ratio of other species from large to small was toluene>ethylbenzene>m,p-xylene>o-xylene>styrene, which were (2.41±2.61)×10-9, (1.37±1.28)×10-9, (0.51±0.48)×10-9, (0.3±0.36)×10-9, and (0.22±0.42)×10-9, respectively. Due to the existence of stable aromatic sources, the monthly and seasonal variation in BTESX mixing ratio was not as obvious as that of other species (NOx, CO, SO2, PM2.5, etc.). The weekend effect of BTESX and other pollutants was not significant. The mixing ratio of BTESX was largely affected by the short distance transportation of chemical enterprises and traffic trunk roads in the northeast, resulting in a large mixing ratio of BTESX in the northeast. The mixing ratio of BTESX was jointly affected by relative humidity and temperature, and its high value area was mainly located in the range of 30%-70% relative humidity. In this range of relative humidity, the high value range of BTESX volume fraction increased with the elevation of temperature. The HI (hazard index) of BTESX in different seasons was within the safety range recognized by EPA, whereas the R (carcinogenic risk of benzene) value was higher than the safety threshold specified by EPA. At the same time, the HI and R values were higher in summer, to which great attention should be paid.

Abstract. Visibility data are fundamental meteorological data widely used in many fields such as climate change, atmospheric radiation, atmospheric pollution, and environmental health. Calculating the average visibility is typically the first step when using visibility data. However, this study proves that the algorithm previously used to calculate average visibility is incorrect, leading to a non-negligible error in average visibility data. Moreover, the use of this incorrect algorithm not only artificially reduces the reliability of visibility data, but also affects the credibility and even the correctness of the conclusions reached in previous studies using visibility data. Therefore, we present the correct algorithm for average visibility, which should be applied to both future and previous research to significantly increase the reliability and application scope of visibility data.

In this study, 56 volatile organic compounds species (VOCs) and other pollutants (NO, NO2, SO2, O3, CO and PM2.5) were measured in the northern suburbs of Nanjing from September 2014 to August 2015. The total volatile organic compound (TVOC) concentrations were higher in the autumn (40.6 ± 23.8 ppbv) and winter (41.1 ± 21.7 ppbv) and alkanes were the most abundant species among the VOCs (18.4 ± 10.0 ppbv). According to the positive matrix factorization (PMF) model, the VOCs were found to be from seven sources in the northern suburbs of Nanjing, including liquefied petroleum gas (LPG) sources, gasoline vehicle emissions, iron and steel industry sources, industrial refining coke sources, solvent sources and petrochemical industry sources. One of the sources was influenced by seasonal variations: it was a diesel vehicle emission source in the spring, while it was a coal combustion source in the winter. According to the conditional probability function (CPF) method, it was found that the main contribution areas of each source were located in the easterly direction (mainly residential areas, industrial areas, major traffic routes, etc.). There were also seasonal differences in concentration, ozone formation potential (OFP), OH radical loss rate (LOH) and secondary organic aerosols potential (SOAP) for each source due to the high volatility of the summer and autumn temperatures, while combustion increases in the winter. Finally, the time series of O3 and OFP was compared to that PM2.5 and SOAP and then they were combined with the wind rose figure. It was found that O3 corresponded poorly to the OFP, while PM2.5 corresponded well to the SOAP. The reason for this was that the O3 generation was influenced by several factors (NOx concentration, solar radiation and non-local transport), among which the influence of non-local transport could not be ignored.

A persistent fog-haze process associated with high pollution occurred in the northern suburbs of Nanjing from November to December 2013. Based on the comprehensive chemical and microphysical observations during the intense observation period, the composition characteristics, and variation rules of volatile organic compounds (VOCs) in the atmosphere under four weather conditions (slight haze, haze, fog, and dense fog) were compared and analyzed, the influencing factors for VOCs during extremely dense fog were discussed in more detail. The average concentrations of VOCs displayed as alkanes > aromatics > alkenes > alkynes, and their concentrations were ranked as dense fog > fog > haze > slight haze, the main factor contributing to the difference in concentrations of VOCs under different weather conditions is the boundary layer characteristics and photochemical reaction rate. Microphysical parameters such as liquid water content (LWC) were negatively correlated with VOCs concentration in dense fog (LWC>0.008 g m −3 ). Also, the concentration of VOCs showed an oscillating decrease in extremely dense fog (LWC>0.12 g m −3 ), and the total VOCs removal rate was close to 30%, which may be attributed to an indirect/direct removal effect, in which the enhanced collision and deposition of fog droplets promote the redistribution of VOCs gas-aqueous/particle partitioning, and remove them from the atmosphere by fog water.