GPS has become a very effective tool to remotely sense precipitable water vapor (PWV) information, which is important for weather forecasting and nowcasting. The number of geodetic GNSS stations set up in China has substantially increased over the last few decades. However, GPS PWV derivation requires surface pressure to calculate the precise zenith hydrostatic delay and weighted mean temperature to map the zenith wet delay to precipitable water vapor. GPS stations without collocated meteorological sensors can retrieve water vapor using standard atmosphere parameters, which lead to a decrease in accuracy. In this paper, a method of interpolating NWP reanalysis data to site locations for generating corresponding meteorological elements is explored over China. The NCEP FNL dataset provided by the NCEP (National Centers for Environmental Prediction) and over 600 observed stations from different sources was selected to assess the quality of the results. A one-year experiment was performed in our study. The types of stations selected include meteorological sites, GPS stations, radio sounding stations, and a sun photometer station. Compared with real surface measurements, the accuracy of the interpolated surface pressure and air temperature both meet the requirements of GPS PWV derivation in most areas; however, the interpolated surface air temperature exhibits lower precision than the interpolated surface pressure. At more than 96% of selected stations, PWV differences caused by the differences between the interpolation results and real measurements were less than 1.0 mm. Our study also indicates that relief amplitude exerts great influence on the accuracy of the interpolation approach. Unsatisfactory interpolation results always occurred in areas of strong relief. GPS PWV data generated from interpolated meteorological parameters are consistent with other PWV products (radio soundings, the NWP reanalysis dataset, and sun photometer PWV data). The differences between them were approximately 1~3 mm at most at our selected stations, and GPS data processing is the main factor influencing the agreement of the GPS PWV results with those of other methods.

Having a high-quality historical air pollutant dataset is critical for environmental and epidemiological research. In this study, a novel deep learning model based on convolutional neural network architecture was developed to estimate ground-level ozone concentrations across eastern China. A high-resolution maximum daily average 8-h (MDA8) surface ground ozone concentration dataset was generated with the support of the total ozone column from the satellite Tropospheric Monitoring Instrument, meteorological data from the China Meteorological Administration Land Data Assimilation System, and simulations of the WRF-Chem model. The modeled results were compared with in situ measurements in five cities that were not involved in model training, and the mean R² of predicted ozone with observed values was 0.9, indicating the good robustness of our model. In addition, we compared the model results with some widely used machine learning techniques (e.g., random forest) and recently published ozone datasets, showing that the accuracy of our model is higher and that the spatial distributions of predicted ozone are more coherent. This study provides an efficient and exact method to estimate ground-level ozone and offers a new perspective for modeling spatiotemporal air pollutants.

Water-vapor-weighted mean temperature, T_m, is the key variable for estimating the mapping factor between GPS zenith wet delay (ZWD) and precipitable water vapor (PWV). For the near-real-time GPS–PWV retrieval, estimating T_m from surface air temperature T_s is a widely used method because of its high temporal resolution and fair degree of accuracy. Based on the estimations of T_m and T_s at each reanalysis grid node of the ERA-Interim data, we analyzed the relationship between T_m and T_s without data smoothing. The analyses demonstrate that the T_s–T_m relationship has significant spatial and temporal variations. Static and time-varying global gridded T_s–T_m models were established and evaluated by comparisons with the radiosonde data at 723 radiosonde stations in the Integrated Global Radiosonde Archive (IGRA). Results show that our global gridded T_s–T_m equations have prominent advantages over the other globally applied models. At over 17 % of the stations, errors larger than 5 K exist in the Bevis equation (Bevis et al., 1992) and in the latitude-related linear model (Y. B. Yao et al., 2014), while these large errors are removed in our time-varying T_s–T_m models. Multiple statistical tests at the 5 % significance level show that the time-varying global gridded model is superior to the other models at 60.03 % of the radiosonde sites. The second-best model is the 1^∘ × 1^∘ GPT2w model, which is superior at only 12.86 % of the sites. More accurate T_m can reduce the contribution of the uncertainty associated with T_m to the total uncertainty in GPS–PWV, and the reduction augments with the growth of GPS–PWV. Our theoretical analyses with high PWV and small uncertainty in surface pressure indicate that the uncertainty associated with T_m can contribute more than 50 % of the total GPS–PWV uncertainty when using the Bevis equation, and it can decline to less than 25 % when using our time-varying T_s–T_m model. However, the uncertainty associated with surface pressure dominates the error budget of PWV (more than 75 %) when the surface pressure has an error larger than 5 hPa. GPS–PWV retrievals using different T_m estimates were compared at 74 International GNSS Service (IGS) stations. At 74.32 % of the IGS sites, the relative differences of GPS–PWV are within 1 % by applying the static or the time-varying global gridded T_s–T_m equations, while the Bevis model, the latitude-related model and the GPT2w model perform the same at 37.84 %, 41.89 % and 29.73 % of the sites. Compared with the radiosonde PWV, the error reduction in the GPS–PWV retrieval can be around 1–2 mm when using a more accurate T_m parameterization, which accounts for around 30 % of the total GPS–PWV error.

Directly establishing the relationship between satellite data and PM_2.5 concentration through deep learning methods for PM_2.5 concentration estimation is an important means for estimating regional PM_2.5 concentration. However, due to the lack of consideration of uncertainty in deep learning methods, methods based on deep learning have certain overfitting problems in the process of PM_2.5 estimation. In response to this problem, this paper designs a deep Bayesian PM_2.5 estimation model that takes into account multiple scales. The model uses a Bayesian neural network to describe key parameters a priori, provide regularization effects to the neural network, perform posterior inference through parameters, and take into account the characteristics of data uncertainty, which is used to alleviate the problem of model overfitting and to improve the generalization ability of the model. In addition, different-scale Moderate-Resolution Imaging Spectroradiometer (MODIS) satellite data and ERA5 reanalysis data were used as input to the model to strengthen the model’s perception of different-scale features of the atmosphere, as well as to further enhance the model’s PM_2.5 estimation accuracy and generalization ability. Experiments with Anhui Province as the research area showed that the R² of this method on the independent test set was 0.78, which was higher than that of the DNN, random forest, and BNN models that do not consider the impact of the surrounding environment; moreover, the RMSE was 19.45 μg·m⁻³, which was also lower than the three compared models. In the experiment of different seasons in 2019, compared with the other three models, the estimation accuracy was significantly reduced; however, the R² of the model in this paper could still reach 0.66 or more. Thus, the model in this paper has a higher accuracy and better generalization ability.

Ozone (O₃), whose concentrations have been increasing in eastern China recently, plays a key role in human health, biodiversity, and climate change. Accurate information about the spatiotemporal distribution of O₃ is crucial for human exposure studies. We developed a deep learning model based on a long short-term memory (LSTM) network to estimate the daily maximum 8 h average (MDA8) O₃ across eastern China in 2020. The proposed model combines LSTM with an attentional mechanism and residual connection structure. The model employed total O₃ column product from the Tropospheric Monitoring Instrument, meteorological data, and other covariates as inputs. Then, the estimates from our model were compared with real observations of the China air quality monitoring network. The results indicated that our model performed better than other traditional models, such as the random forest model and deep neural network. The sample-based cross-validation R² and RMSE of our model were 0.94 and 10.64 μg m⁻³, respectively. Based on the O₃ distribution over eastern China derived from the model, we found that people in this region suffered from excessive O₃ exposure. Approximately 81% of the population in eastern China was exposed to MDA8 O₃ > 100 μg m⁻³ for more than 150 days in 2020.