Tropospheric ozone in the surface air has become the primary atmospheric pollutant in Hangzhou, China, in recent years. Previous analysis is not enough to decode it for better regulation. Therefore, we use the traditional atmospheric model, Weather Research and Forecasting coupled with Community Multi-scale Air Quality (WRF-CMAQ), and machine learning models, Extreme Learning Machine (ELM), Multi-layer Perceptron (MLP), Random Forest (RF) and Recurrent Neural Network (RNN) to analyze and predict the ozone in the surface air in Hangzhou, China, using meteorology and air pollutants as input. We firstly quantitatively demonstrate that the dew-point deficit, instead of temperature and relative humidity, is the predominant meteorological factor in shaping tropospheric ozone. Urban heat island, daily direct solar radiation time, wind speed and wind direction play trivial role in impacting tropospheric ozone. NO₂ is the primary influential factors both for hourly ozone and daily O₃-8 h due to the titration effect. The most environmental-friendly way to mitigate the ozone pollution is to lower the volatile organic compounds (VOCs) with the highest ozone formation potentials. We deduce that the tropospheric ozone formation process tends to be not only non-linear but also non-smooth. Compared with the traditional atmospheric models, machine learning, whose characteristics are rapid convergence, short calculating time, adaptation of forecasting episodes, small program memory, higher accuracy and less cost, is able to predict tropospheric ozone more accurately.

The heavy metal concentration of soil samples often exhibits a skewed distribution, especially for soil samples from mining areas with an extremely high concentration of heavy metals. In this study, to model soil contamination in mining areas using reflectance spectroscopy, the skewed distribution was corrected and heavy metal concentration estimated. In total, 46 soil samples from a mining area, along with corresponding field soil spectra, were collected. Laboratory spectra of the soil samples and the field spectra were used to estimate copper (Cu) concentration in the mining area. A logarithmic transformation was used to correct the skewed distribution, and based on the sorption of Cu on spectrally active soil constituents, the spectral bands associated with iron oxides were extracted from the visible and near-infrared (VNIR) region and used in the estimation. A genetic algorithm was adopted for band selection, and partial least squares regression was used to calibrate the estimation model. After transforming the distribution of Cu concentration, the accuracies (R2) of the estimation of Cu concentration using laboratory and field spectra separately were 0.94 and 0.96. The results indicate that Cu concentration in the mining area can be estimated using reflectance spectroscopy following correction of skewed distribution.

Ambient submicron particles (PM₁) exert significant impacts on visibility degradation during severe pollution episodes of urban China. The U.S. IMPROVE algorithms are widely used for assessing the extinction effect of atmospheric aerosols, but only suitable for fine particulate matter. A proper algorithm for PM₁ extinction estimation is lacking and becomes urgent, especially after the online measurement of PM₁ species is routine by aerosol mass spectrometers. Here we conducted three-month in-situ measurements to explore mass scattering efficiencies (MSE) of PM₁ major species at a supersite of eastern China. Results indicated that MSEs of ammonium sulfate and nitrate increase quickly and then keep stable with the mass accumulation, while those of organic matter keep at ∼5.5 m²/g but with a large vibration in the whole mass range. The algorithm for reconstructing PM₁ dry scattering coefficient was derived from the integral of the variation patterns for the three PM₁ species. The algorithm was then validated and compared with other empirical algorithms through separate field measurements. Good correlations between the reconstructed and measured dry scattering coefficient were observed with R square higher than 0.9 and slope of 1.01–1.05, indicating that the reconstructed algorithm can predict the dry scattering coefficient well based on PM₁ chemical composition measurements in urban China. Our study is expected to provide observed insights on the variation of MSE in the wide mass range especially in the high region, as well as accurate formulas for ambient PM₁ dry scattering apportionment.

Epidemiologists use prediction models to downscale (i.e., interpolate) air pollution exposure where monitoring data is insufficient. This study compares machine learning prediction models for ground-level ozone during wildfires, evaluating the predictive accuracy of ten algorithms on the daily 8-hour maximum average ozone during a 2008 wildfire event in northern California. Models were evaluated using a leave-one-location-out cross-validation (LOLO CV) procedure to account for the spatial and temporal dependence of the data and produce more realistic estimates of prediction error. LOLO CV avoids both the well-known overly optimistic bias of k-fold cross-validation on dependent data and the conservative bias of evaluating prediction error over a coarser spatial resolution via leave-k-locations-out CV. Gradient boosting was the most accurate of the ten machine learning algorithms with the lowest LOLO CV estimated root mean square error (0.228) and the highest LOLO CV Rˆ2 (0.677). Random forest was the second best performing algorithm with an LOLO CV Rˆ2 of 0.661. The LOLO CV estimates of predictive accuracy were less optimistic than 10-fold CV estimates for all ten models. The difference in estimated accuracy between the 10-fold CV and LOLO CV was greater for more flexible models like gradient boosting and random forest. The order of estimated model accuracy depended on the choice of evaluation metric, indicating that 10-fold CV and LOLO CV may select different models or sets of covariates as optimal, which calls into question the reliability of 10-fold CV for model (or variable) selection. These prediction models are designed for interpolating ozone exposure, and are not suited to inferring the effect of wildfires on ozone or extrapolating to predict ozone in other spatial or temporal domains. This is demonstrated by the inability of the best performing models to accurately predict ozone during 2007 southern California wildfires.

Indoor and outdoor fine particulate matter (PM2.5) are both leading risk factors for death and disease, but making indoor measurements is often infeasible for large study populations.We developed models to predict indoor PM2.5 concentrations for pregnant women who were part of a randomized controlled trial of portable air cleaners in Ulaanbaatar, Mongolia. We used multiple linear regression (MLR) and random forest regression (RFR) to model indoor PM2.5 concentrations with 447 independent 7-day PM2.5 measurements and 87 potential predictor variables obtained from outdoor monitoring data, questionnaires, home assessments, and geographic data sets. We also developed blended models that combined the MLR and RFR approaches. All models were evaluated in a 10-fold cross-validation.The predictors in the MLR model were season, outdoor PM2.5 concentration, the number of air cleaners deployed, and the density of gers (traditional felt-lined yurts) surrounding the apartments. MLR and RFR had similar performance in cross-validation (R2 = 50.2%, R2 = 48.9% respectively). The blended MLR model that included RFR predictions had the best performance (cross validation R2 = 81.5%). Intervention status alone explained only 6.0% of the variation in indoor PM2.5 concentrations.We predicted a moderate amount of variation in indoor PM2.5 concentrations using easily obtained predictor variables and the models explained substantially more variation than intervention status alone. While RFR shows promise for modelling indoor concentrations, our results highlight the importance of out-of-sample validation when evaluating model performance. We also demonstrate the improved performance of blended MLR/RFR models in predicting indoor air pollution.

Satellite aerosol products have been widely used to retrieve ground PM₂.₅ concentration because of their wide coverage and continuous spatial distribution. While more and more studies have focused on the retrieval algorithms, the foundation for the retrieval—relationship between PM₂.₅ and satellite aerosol optical depth (AOD) has not been fully investigated. In this study, the relationships between PM₂.₅ and AOD were investigated in 368 cities in mainland China from February 2013 to December 2017, at different temporal and regional scales. Pearson correlation coefficients and the PM₂.₅/AOD ratio were used as indicators. Firstly, we established the relationship between PM₂.₅ and AOD in terms of the spatio-temporal variations, and discuss the impact of some potential factors for a better understanding of the spatio-temporal variations. Spatially, we found that the correlation is higher in the Beijing-Tianjin-Hebei and Chengyu regions and weaker in coastal areas. The PM₂.₅/AOD ratio shows an obvious north–south difference, with the ratio in North China higher than South China. Temporally, the correlation coefficient tends to be higher in May and September, with the PM₂.₅/AOD ratio higher in winter and lower in summer. As for interannual variations, we detected a decreasing tendency for the PM₂.₅-AOD correlation and PM₂.₅/AOD ratio for recent years. Then, to determine the impact of the weakening of the PM₂.₅-AOD relationship on PM₂.₅ remote sensing retrieval performance, a geographically weighted regression (GWR) retrieval experiment was conducted. The results showed that the performance of retrievals is also decreasing while PM₂.₅-AOD relationship getting weaker. Our study investigated the PM₂.₅-AOD relationship over a large extent at the city scale, and investigated the temporal variations in terms of interannual variations. The results will be useful for the satellite retrieval of PM₂.₅ concentration and will help us to further understand the PM₂.₅ pollution situation in mainland China.

We study the evolution of large systems of sticky patches of pollution that float on the sea surface and are moved around by realistic wind and simulated surface currents. The study area is the Gulf of Finland in the eastern Baltic Sea that hosts extremely heavy cargo and tanker traffic. It is assumed that patches merge at collision. Collisions are enhanced by wind impact that depends on the patch size. We implement a space partitioning algorithm for fast simulations of large systems of >8000 patches and perform a series of simulations with different initial size and location of the patches. The resulting empirical distribution of the number of patches of different size is governed universally by a stretched-exponential power law f(x) = A exp (−xb). This indicates that the evolution of large systems of pollution patches on the sea surface exhibits classic self-organization and scale invariance properties. This may be the key effect explaining how the patchiness of waste in marine environment is formed under the impact of different drivers.

Biodegradation is important for the fate of oil spilled in marine environments, yet parameterization of biodegradation varies across oil spill models, which usually apply constant first-order decay rates to multiple pseudo-components describing an oil. To understand the influence of model parameterization on the fate of subsurface oil droplets, we reviewed existing algorithms and rates and conducted a model sensitivity study. Droplets were simulated from a blowout at 2000 m depth and were either treated with sub-surface dispersant injection (2% dispersant to oil ratio) or untreated. The most important factor affecting oil fate was the size of the droplets, with biodegradation contributing substantially to the fate of droplets ≤0.5 mm. Oil types, which were similar, had limited influence on simulated oil fate. Model results suggest that knowledge of droplet sizes and improved estimation of pseudo-component biodegradation rates and lag times would enhance prediction of the fate and transport of subsurface oil.

The paper reviews the existing applications of sensing technologies for measuring construction off-road vehicle emissions (COVE) such as earthmoving equipment. The current literature presented different measurement methods and reported the results of utilisation of new technologies for measuring COVE. However, previous papers used different technology applications covering only a part of the monitoring process with its own limitations. Since technologies are advancing and offering novel solutions, there is an urgent need to identify the gaps, re-evaluate the current methods, and develop a critical agenda for automating the entire process of collecting emissions data from construction sites, and monitoring the emission contributors across cities. This paper systematically identifies relevant papers through a search of three key databases—Web of Science, Engineering Valley and Scopus—covering the publications in the last decade from 2008 to 2017. An innovative robust research method was designed to select and analyse the relevant papers. The identified papers were stored in a data set, and a thematic algorithm employed to find the clusters of papers which might be potentially relevant. The selected papers were used for further micro-thematic analysis to find key relevant papers on COVE, and the gap in the literature. A sample of relevant papers was found relevant to COVE and critically reviewed by coding and content analysis. This paper critically reviews the selected papers and also shows that there is a considerable gap in the applications of new technologies for measuring in-use COVE in real time based on real activities toward automated methods. This review enables practitioners and scholars to gain a concrete understanding of the gap in measuring COVE and to provide a significant agenda for future technology applications.

New optical tools are proposed for the real-time diagnosis of water quality without traditional sampling and laboratory physico-chemical analysis. In particular, the optical decision-making system is developed as an operational tool for in-situ assessment of water quality in natural water areas. Specifically, three versions of this system are analyzed using 8-channel and 35-channel photometric devices and a 128-channel spectroelipsometric device. The spectral images obtained from these devices are the basis for the diagnosis of water quality by using new algorithms to detect these spectral images. In more detail, four algorithms are developed to identify optical spectral images of water objects. The effectiveness of these optical tools is demonstrated in experimental water quality control sites in water systems operating under different climatic conditions. Finally, the results for the determination of the various pollutants at the surface of these water bodies are presented.