Sea surface temperature (SST) is a key parameter in ocean hydrology. Currently, existing SST prediction methods fail to fully utilize the potential spatial correlation between variables. To address this challenge, we propose a spatiotenporal UNet (ST-UNet) model based on the UNet model. In particular, in the encoding phase of ST-UNet, we use parallel convolution with different kernel sizes to efficiently extract spatial features, and use ConvLSTM to capture temporal features based on the utilization of spatial features. Atrous Spatial Pyramid Pooling (ASPP) module is placed at the bottleneck of the network to further incorporate the multi-scale features, allowing the spatial features to be fully utilized. The final prediction is then generated in the decoding stage using parallel convolution with different kernel sizes similar to the encoding stage. We conducted a series of experiments on the Bohai Sea and Yellow Sea SST data set, as well as the South China Sea SST data set, using SST data from the past 35 days to predict SST data for 1, 3, and 7 days in the future. The model was trained using data spanning from 2010 to 2021, with data from 2022 being utilized to assess the model’s predictive performance. The experimental results show that the model proposed in this research paper achieves excellent results at different prediction scales in both sea areas, and the model consistently outperforms other methods. Specifically, in the Bohai Sea and Yellow Sea sea areas, when the prediction scales are 1, 3, and 7 days, the MAE of ST-UNet outperforms the best results of the other three compared models by 17%, 12%, and 2%, and the MSE by 16%, 18%, and 9%, respectively. In the South China Sea, when the prediction ranges are 1, 3, and 7 days, the MAE of ST-UNet is 27%, 18%, and 3% higher than the best of the other three compared models, and the MSE is 46%, 39%, and 16% higher, respectively. Our results highlight the effectiveness of the ST-UNet model in capturing spatial correlations and accurately predicting SST. The proposed model is expected to improve marine hydrographic studies.

An effective soil moisture retrieval method for FY-3E (Fengyun-3E) GNOS-R (GNSS occultation sounder II-reflectometry) is developed in this paper. Here, the LAGRS model, which is totally oriented for GNOS-R, is employed to estimate vegetation and surface roughness effects on surface reflectivity. Since the LAGRS (land surface GNSS reflection simulator) model is a space-borne GNSS-R (GNSS reflectometry) simulator based on the microwave radiative transfer equation model, the method presented in this paper takes more consideration on the physical scattering properties for retrieval. Ancillary information from SMAP (soil moisture active passive) such as the vegetation water content and the roughness coefficient are investigated for the final algorithm’s development. At first, the SR (surface reflectivity) data calculated from GNOS-R is calculated and then calibrated, and then the vegetation roughness factor is achieved and used to eliminate the effects on both factors. After receiving the Fresnel reflectivity, the corresponding soil moisture estimated from this method is retrieved. The results demonstrate good consistency between soil moisture derived from GNOS-R data and SMAP soil moisture, with a correlation coefficient of 0.9599 and a root mean square error of 0.0483 cm³/cm³. This method succeeds in providing soil moisture on a global scale and is based on the previously developed physical LAGRS model. In this way, the great potential of GNOS-R for soil moisture estimation is presented.

The fast radiative transfer model has a wide range of applications in remote sensing, such as satellite L2 product retrieval and satellite L1 data quality monitoring and assessment. The traditional radiative transfer model, particularly reflective solar bands, has the disadvantage of being time-consuming. To improve the calculation efficiency and realize fast observation simulation over the global ocean under clear-sky conditions, machine learning technology is applied. A fast radiative transfer simulation method based on the extreme gradient boosting (XGBoost) algorithm is proposed. The polynomial regression and XGBoost regression models in the machine learning field were used in the fast simulation process. By comparing and analysing the experimental results of the two regression models, it was revealed that the prediction results of the XGBoost algorithm were better than those of the polynomial regression model. The shorter the wavelength, the better the prediction performance, with a small error and a larger determination coefficient. The blue bands demonstrated the best results, with an R²(R-Squire) value of approximately 0.99. The deviation analysis between the predicted and simulated values demonstrated that there was no obvious functional dependence between the prediction error and the influencing factors.