Very high-resolution remote sensing change detection has always been an important research issue due to the registration error, robustness of the method, and monitoring accuracy, etc. This paper proposes a robust and more accurate approach of change detection (CD), and it is applied on a smaller experimental area, and then extended to a wider range. A feature space, including object features, Visual Geometry Group (VGG) depth features, and texture features, is constructed. The difference image is obtained by considering the contextual information in a radius scalable circular. This is to overcome the registration error caused by the rotation and shift of the instantaneous field of view and also to improve the reliability and robustness of the CD. To enhance the robustness of the U-Net model, the training dataset is constructed manually via various operations, such as blurring the image, increasing noise, and rotating the image. After this, the trained model is used to predict the experimental areas, which achieved 92.3% accuracy. The proposed method is compared with Support Vector Machine (SVM) and Siamese Network, and the check error rate dropped to 7.86%, while the Kappa increased to 0.8254. The results revealed that our method outperforms SVM and Siamese Network.
Hydrological models play an essential role in data assimilation (DA) systems. However, it is a challenging task to acquire the distributed hydrological model parameters that affect the accuracy of the simulations at a grid scale. Remote sensing data provide an ideal observation for DA to estimate parameters and state variables. In this study, a special assimilation scheme was proposed to jointly estimate parameters and soil moisture (SM) by assimilating brightness temperature (TB) from the Soil Moisture and Ocean Salinity (SMOS) mission. Variable infiltration capacity (VIC) hydrological model and L-band microwave emission of the biosphere model (L-MEB) are coupled as model and observation operators, respectively. The scheme combines two stages of estimators, one for the static model parameters and the other for the dynamic state variables. The estimators approximate the posterior probability distribution of an unknown target through sequential Monte Carlo (SMC) sampling. Markov chain Monte Carlo (MCMC) and immune evolution strategy are embedded in both stages to solve particle impoverishment problems. To evaluate the effectiveness of the scheme, the estimated SM sets are compared with in-situ observations and SMOS products in Maqu on the Tibetan Plateau. Specifically, the root mean square error decreased from 0.126 to 0.087 m3m−3 for surface SM, with a slight impact on the root zone. The temporal correlation between DA results and in-situ measurements increased to 0.808 and 0.755 for surface SM (+0.057) and root zone SM (+0.040), respectively. The results demonstrate that assimilating TB has tremendous potential as an approach to improve the estimation of distributed model parameters and SMs of surface and root zone at a grid scale, and the immune evolution strategy is effective for increasing the accuracy of approximation in sampling.