The partial pressure of ocean surface CO₂ (pCO₂) plays an important role in quantifying the carbon budget and assessing ocean acidification. For such a vast and complex ocean system as the global ocean, most current research practices tend to study the ocean into regions. In order to reveal the overall characteristics of the global ocean and avoid mutual influence between zones, a holistic research method was used to detect the correlation of twelve predictive factors, including chlorophyll concentration (Chlor_a), diffuse attenuation coefficient at 490 nm (Kd_490), density ocean mixed layer thickness (Mlotst), eastward velocity (East), northward velocity (North), salinity (Sal), temperature (Temp), dissolved iron (Fe), dissolved silicate (Si), nitrate (NO₃), potential of hydrogen (pH), phosphate (PO₄), at the global ocean scale. Based on measured data from the Global Surface pCO₂ (LDEO) database, combined with National Aeronautics and Space Administration (NASA) Ocean Color satellite data and Copernicus Ocean reanalysis data, an improved optimized random forest (ORF) method is proposed for the overall reconstruction of global ocean surface pCO₂, and compared with various machine learning methods. The results indicate that the ORF method is the most accurate in overall modeling at the global ocean scale (mean absolute error of 6.27μatm, root mean square error of 15.34μatm, R² of 0.92). Based on independent observations from the LDEO dataset and time series observation stations, the ORF model was further validated, and the global ocean surface pCO₂ distribution map of 0.25° × 0.25° for 2010 to 2019 was reconstructed, which is of significance for the global ocean carbon cycle and carbon assessment.

Sulfur dioxide (SO₂) is a serious air pollutant emitted from different sources in many developing regions worldwide, where the contribution of different potential influencing factors remains unclear. Using Shandong, a typical industrial province in China as an example, we studied the spatial distribution of SO₂ and used geographical detectors to explore its influencing factors. Based on the daily average concentration in Shandong Province from 2014 to 2019, we explored the influence of the diurnal temperature range, secondary production, precipitation, wind speed, soot emission, sunshine duration, and urbanization rate on the SO₂ concentration. The results showed that the diurnal temperature range had the largest impact on SO₂, with q values of 0.69, followed by secondary production (0.51), precipitation (0.46), and wind speed (0.42). There was no significant difference in the SO₂ distribution between pairs of sunshine durations, soot emissions, and urbanization rates. The meteorological factors of precipitation, wind speed, and diurnal temperature range were sensitive to seasonal changes. There were nonlinear enhancement relationships among those meteorological factors to the SO₂ pollution. There were obvious geographical differences in the human activity factors of soot emissions, secondary production, and urbanization rates. The amount of SO₂ emissions should be adjusted in different seasons considering the varied effect of meteorological factors.

With the rapid development of the global navigation satellite system (GNSS), high-rate GNSS has been widely used for high-precision GNSS coseismic displacement retrieval. In recent decades, relative positioning (RP) and precise point positioning (PPP) are mainly adopted to retrieve coseismic displacements. However, RP can only obtain relative coseismic displacements with respect to a reference station, which might be subject to quaking during a large seismic event. While PPP needs a long (re)convergence period of tens of minutes. There is no convergence time needed in the variometric approach for displacements analysis standalone engine (VADASE) but the derived displacements are accompanied by a drift. Temporal point positioning (TPP) method adopts temporal-differenced ionosphere-free phase measurements between a reference epoch and the current epoch, and there is almost no drift in the displacement derived from TPP method. Nevertheless, the precise orbit and clock products should be applied in the TPP method. The studies in recent years are almost based on the postprocessing precise orbits and clocks or simulated real-time products. Since 2013, international GNSS service (IGS) has been providing an open-access real-time service (RTS), which consists of orbit, clock and other corrections. In this contribution, we evaluated the performance of real-time coseismic displacement retrieval based on TPP method with IGS RTS correction products. At first, the real-time precise orbit and clock offsets are derived from the RTS correction products. Then, the temporal-differenced ionosphere-free (IF) combinations are formed and adopted as the TPP measurements. By applying real-time precise orbit and clock offsets, the coseismic displacement can be real-timely retrieved based on TPP measurements. To evaluate the accuracy, two experiments including a stationary experiment and an application to an earthquake event were carried out. The former gives an accuracy of 1.8 cm in the horizontal direction and 4.1 cm in the vertical direction during the whole period of 15-min. The latter gives an accuracy of 1.2 cm and 2.4 cm in the horizontal and vertical components, respectively.

Global warming and sea-level rise (SLR) induced by rising atmospheric CO₂ concentrations can cause coral bleaching, death, and submergence of the world’s coral reefs. Adopting the GIS and RS methods, we modeled how these two stressors combine to influence the future growth of the atolls and table reefs of three archipelagoes in the South China Sea (SCS), based on geomorphic and ecological zones. A large-scale survey of the coral communities in Xisha Islands in 2014, Dongsha Islands in 2014–2016 and Nansha Islands in 2007 provided zone-specific process datasets on the range of reef accretion rates. Sea surface temperature and extreme (minimum and maximum) SLR data above 1985–2005 levels by 2100 in the SCS were derived from the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5) models forced with the Representative Concentration Pathways (RCPs). Our model projected that: (1) the Xisha Islands and Dongsha Islands may have a better growth status, because the reef flat biotic sparse zone may be recolonized with hard coral and become a biotic dense zone; (2) the southern Nansha Islands reefs have a risk of stopping growing due to their earlier annual bleaching years. The increasing of water depths of these reefs is stronger in the RCP with more emissions. Our approach offers insights into the best-case and worst-case impacts of two global environmental pressures on potential future reef growth under a changing climate.

Air pollution is a serious problem in China, and the government has taken a series of steps to solve it. However, it is still u2nclear how the situation has improved after years of atmospheric pollution control. Shandong Province, which has the second largest population and the highest coal consumption in China, was chosen to analyze the spatiotemporal air quality trends. We obtained daily air quality index (AQI) values from 91 stations in the province from 1 January 2014, to 31 December 2019, based on an adaptive data analysis method, empirical mode decomposition (EMD). The distribution of AQI in Shandong Province was heterogeneous at both spatial and temporal scales. All the stations could be divided into four clusters whose AQI trends decreased from 75 to 53, 95 to 68, 128 to 82, and 148 to 82, respectively. The overall trend of pollution became more serious from east to west in the province. The AQI is the largest in winter, followed by spring and autumn, and the smallest index occurs in summer. There are four types of annual trends of the AQI of each city. The overall downward trend indicates that the air quality of each city was improving annually.

Artificial neural networks are widely applied for prediction, function simulation, and data classification. Among these applications, the wavelet neural network is widely used in image classification problems due to its advantages of high approximation capabilities, fault-tolerant capabilities, learning capacity, its ability to effectively overcome local minimization issues, and so on. The error function of a network is critical to determine the convergence, stability, and classification accuracy of a neural network. The selection of the error function directly determines the network’s performance. Different error functions will correspond with different minimum error values in training samples. With the decrease of network errors, the accuracy of the image classification is increased. However, if the image classification accuracy is difficult to improve upon, or is even decreased with the decreasing of the errors, then this indicates that the network has an “over-learning” phenomenon, which is closely related to the selection of the function errors. With regards to remote sensing data, it has not yet been reported whether there have been studies conducted regarding the “over-learning” phenomenon, as well as the relationship between the “over-learning” phenomenon and error functions. This study takes SAR, hyper-spectral, high-resolution, and multi-spectral images as data sources, in order to comprehensively and systematically analyze the possibility of an “over-learning” phenomenon in the remote sensing images from the aspects of image characteristics and neural network. Then, this study discusses the impact of three typical entropy error functions (NB, CE, and SH) on the “over-learning” phenomenon of a network. The experimental results show that the “over-learning” phenomenon may be caused only when there is a strong separability between the ground features, a low image complexity, a small image size, and a large number of hidden nodes. The SH entropy error function in that case will show a good “over-learning” resistance ability. However, for remote sensing image classification, the “over-learning” phenomenon will not be easily caused in most cases, due to the complexity of the image itself, and the diversity of the ground features. In that case, the NB and CE entropy error network mainly show a good stability. Therefore, a blind selection of a SH entropy error function with a high “over-learning” resistance ability from the wavelet neural network classification of the remote sensing image will only decrease the classification accuracy of the remote sensing image. It is therefore recommended to use an NB or CE entropy error function with a stable learning effect.