The South China Sea (SCS) is one of the largest marginal seas in the world. It includes a river-dominated, highly productive ocean margin on the northern shelf and an oligotrophic ocean-dominated basin along with other sub-regions with various features. It was challenge to estimate the air-sea CO2 flux in this area. We developed a retrieval algorithm for sea surface p CO ₂ by a combination of our previously established semi-mechanistic approach (MeSAA) and machine learning (ML) method, named MeSAA-ML-SCS, built upon a large dataset of sea surface partial pressure of CO ₂ ( p CO ₂ ) collected from in situ measurements during 44 cruises/legs to the SCS in the last two decades. We set several semi-analytical parameters, includes: p CO _{2_therm} represented the combined effect of thermodynamics and the atmospheric CO ₂ forcing on seawater p CO ₂ ; upwelling index (UISST) and mixing layer depth (MLD) to characterize the mixing processes; chlorophyll-a concentration (Chl-a) with remote sensing reflectance at 443 and 555 nm (Rrs(443) and Rrs(555)), which were proxies of biological effects and other characteristics for distinguishing shelf, basin, and sub-regions. We set the difference between seawater p CO ₂ and atmospheric p CO ₂ (Δ p CO ₂ ^Sea-Air ) as the output, and the seawater p CO ₂ was finally obtained by summing atmospheric p CO ₂ and Δ p CO ₂ ^Sea-Air . We compared several ML models, and the XGBoost model was confirmed as the best. Independent cruise-based datasets that are not involved in the model training were used to validate the satellite products, with low root mean square error (RMSE = 11.69 μatm) and mean absolute percentage deviation (APD = 1.59%). The increasing trend of time-series satellite-derived p CO ₂ (2.44 ± 0.24 μatm/yr) were validated by the in situ data at the Southeastern Asia Time-series Study (SEATS) station, showing good consistency. Results indicate that the SCS as a whole is a source of atmospheric CO ₂ , releasing an average of 12.34 ± 3.11 Tg C/yr from a total area of 2.87 × 10 ⁶ km ² , while the northern shelf act as a sink (2.02 ± 0.64 Tg C/yr). With the forcing of increasing atmospheric CO ₂ , the area-integrated CO ₂ efflux over the entire SCS is decreasing with a rate of 0.41 Tg C/yr during 2003–2019. This shared long time series, high-accuracy dataset (1 km) can be helpful to further improve our understanding of the air-sea CO ₂ exchange dynamics in the SCS.

Seawater pH is a direct proxy of ocean acidification, and monitoring the global pH distribution and long-term series changes is critical to understanding the changes and responses of the marine ecology and environment under climate change. Owing to the lack of sufficient global-scale pH data and the complex relationship between seawater pH and related environmental variables, generating time-series products of satellite-derived global sea surface pH poses a great challenge. In this study, we solved the problem of the lack of sufficient data for pH algorithm development by using the massive underway sea surface carbon dioxide partial pressure (pCO₂) dataset to structure a large data volume of near in situ pH based on carbonate calculation between underway pCO₂ and calculated total alkalinity from sea surface salinity and relevant parameters. The remote sensing inversion model of pH was then constructed through this massive pH training dataset and machine learning methods. After several tests of machine learning methods and groups of input parameters, we chose the random forest model with longitude, latitude, sea surface temperature (SST), chlorophyll a (Chla), and Mixed layer depth (MLD) as model inputs with the best performance of correlation coefficient (R² = 0.96) and root mean squared error (RMSE = 0.008) in the training set and R² = 0.83 (RMSE = 0.017) in the testing set. The sensitivity analysis of the error variation induced by the uncertainty of SST and Chla (SST ≤ ±0.5 °C and Chla ≤ ±20%; RMSE_SST ≤ 0.011 and RMSE_Chla ≤ 0.009) indicated that our sea surface pH model had good robustness. Monthly average global sea surface pH products from 2004 to 2019 with a spatial resolution of 0.25° × 0.25° were produced based on the satellite-derived SST and Chla products and modeled MLD dataset. The pH model and products were validated using another independent station-measured pH dataset from the Global Ocean Data Analysis Project (GLODAP), showing good performance. With the time-series pH products, refined interannual variability and seasonal variability were presented, and trends of pH decline were found globally. Our study provides a new method of directly using remote sensing to invert pH instead of indirect calculation based on the construction of massive underway calculated pH data, which would be made useful by comparing it with satellite-derived pCO₂ products to understand the carbonate system change and the ocean ecological environments responding to the global change.

The vertical distribution of absorbing aerosols has nonnegligible impact on the atmospheric correction of satellite ocean color remote sensing, especially for the water-leaving radiance retrieval at blue and ultraviolet bands. In this study, we investigated the impact of the vertical distribution of absorbing aerosols on the satellite-measured radiance at the top of the atmosphere (TOA) and the retrieved water-leaving radiance. First, the global occurrence frequency of absorbing aerosols was mapped, and it was found that the annual averaged occurrence frequencies of absorbing aerosols were >30% over the coasts of the Sahara and Arabian Desert, China, South-Central Africa, and the Indian Peninsula. Second, a new aerosol classification algorithm was developed to establish absorbing aerosol optical models based on AERosol RObotic NETwork (AERONET) site observations. Finally, the effects of the vertical distribution of absorbing aerosols on the upward radiance at the TOA at 412 nm and the retrieved water-leaving radiance were evaluated. The results showed that the influence of the vertical distribution on the TOA radiance could be up to 8% for dust and 10% for fine-dominated absorbing aerosols (FDAs), which was comparable with the influence of aerosol optical depth. Not considering absorbing aerosols during atmospheric correction might produce $\sim 4$ %–10% errors in the water-leaving radiance retrieval at 412 nm over turbid waters under the assumption of a Gaussian distribution. Simplified exponential and two-layer atmospheric vertical distribution models can lead to errors of water-leaving radiance retrieval up to $\sim 12$ %–15% and $\sim 30$ %–40%, respectively. Overall, the imperfect vertical distribution assumption and aerosol model selection might induce uncertainty in water-leaving radiance retrieval from 10% to 80% under absorbing aerosol conditions.

With the implementation of the 2018–2020 Clean Air Action Plan (CAAP) the and impact from COVID-19 lockdowns in 2020, air pollution emissions in central and eastern China have decreased markedly. Here, by combining satellite remote sensing, re-analysis, and ground-based observational data, we established a machine learning (ML) model to analyze annual and seasonal changes in primary air pollutants in 2020 compared to 2018 and 2019 over central and eastern China. The root mean squared errors (RMSE) for the PM_2.5, PM₁₀, O₃, and CO validation dataset were 9.027 μg/m³, 20.312 μg/m³, 10.436 μg/m³, and 0.097 mg/m³, respectively. The geographical random forest (RF) model demonstrated good performance for four main air pollutants. Notably, PM_2.5, PM₁₀, and CO decreased by 44.1%, 43.2%, and 35.9% in February 2020, which was likely influenced by the COVID-19 lockdown and primarily lasted until May 2020. Furthermore, PM_2.5, PM₁₀, O₃, and CO decreased by 16.4%, 24.2%, 2.7%, and 19.8% in 2020 relative to the average values in 2018 and 2019. Moreover, the reduction in O₃ emissions was not universal, with a significant increase (~20–40%) observed in uncontaminated areas.

Knowledge of the vertical distribution of absorbing aerosols is crucial for radiative forcing assessment, and its quasi real-time prediction is one of the keys for the atmospheric correction of satellite remote sensing. In this study, we investigated the seasonal and interannual changes of the vertical distribution of global absorbing aerosols based on satellite measurement from the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) and proposed a neural network (NN) model to predict the vertical distribution of global absorbing aerosols. Gaussian fitting was proposed to derive the maximum fitted particle number concentration (MFNC), altitude corresponding to MFNC (MFA), and standard deviation (MFASD) for vertical distribution of dust and smoke aerosols. Results showed that higher MFA values of dust and smoke aerosols mainly occurred over deserts and tropical savannas, respectively. For dust aerosol, the MFA is mainly observed at 0.5 to 6 km above deserts, and low MFNC values occur in boreal spring and winter while high values in summer and autumn. The MFA of smoke is systematically lower than that of dust, ranging from 0.5 to 3.5 km over tropical rainforest and grassland. Moreover, we found that the MFA of global dust and smoke had decreased by 2.7 m yr⁻¹ (statistical significance p = 0.02) and 1.7 m yr⁻¹ (p = 0.02) over 2007–2016, respectively. The MFNC of global dust has increased by 0.63 cm⁻³ yr⁻¹ (p = 0.05), whereas that of smoke has decreased by 0.12 cm⁻³ yr⁻¹ (p = 0.05). In addition, the determination coefficient (R²) of the established prediction models for vertical distributions of absorbing aerosols were larger than 0.76 with root mean square error (RMSE) less than 1.42 cm⁻³, which should be helpful for the radiative forcing evaluation and atmospheric correction of satellite remote sensing.