Natural wetland ecosystems provide not only important habitats for many wildlife species, but also food for migratory and resident animals. In Shanghai, the Chongming Dongtan International Wetland, located at the mouth of the Yangtze River, plays an important role in maintaining both ecosystem health and ecological security of the island. Meanwhile it provides an especially important stopover and overwintering site for migratory birds, being located in the middle of the East Asian-Australasian Flyway. However, with the increase in development intensity and human activities, this wetland suffers from increasing environmental pressure. On the other hand, biological succession in the mudflat wetland makes Chongming Dongtan a rapidly developing and rare ecosystem in the world. Therefore, studying the wetland spatio-temporal change is an important precondition for analyzing the relationship between wetland evolution processes and human activities. This paper presents a novel method for analyzing land-use/cover changes (LUCC) on Chongming Dongtan wetland using multispectral satellite images. Our method mainly takes advantages of a machine learning algorithm, named the Kernel Extreme Learning Machine (K-ELM), which is applied to distinguish between different objects and extract their information from images. In the K-ELM, the kernel trick makes it more stable and accurate. The comparison between K-ELM and three other conventional classification methods indicates that the proposed K-ELM has the highest overall accuracy, especially for distinguishing between Spartina alternflora, Scirpus mariqueter, and Phragmites australis. Meanwhile, its efficiency is remarkable as well. Then a total of eight Landsat TM series images acquired from 1986 to 2013 were used for the LUCC analysis with K-ELM. According to the classification result, the change detection and spatio-temporal quantitative analysis were performed. The specific analysis of different objects are significant for learning about the historical changes to Chongming Dongtan and obtaining the evaluation rules. Generally, the rapid speed of Chongming Dongtan’s urbanization brought about great influence with respect to natural resources and the environment. Integrating the results into the ecological analysis and ecological regional planning of Dongtan could provide a reliable scientific basis for rational planning, development, and the ecological balance and regional sustainability of the wetland area.

Point cloud fitting plays an important role in various applications of laser scanning technology. Currently, the most dominant algorithms are the linear least squares (LS) method and the Random Sample Consensus (RANSAC) algorithm, in which the random noises of the point cloud coordinates and the influence of the corresponding stochastic model cannot be fully considered. In this study, the mathematical model of point cloud fitting is uniformly described using the nonlinear Gauss-Helmert (NGH) model, in which the nonlinear relationship between the measurements and parameters as well as the effects of all random errors can be considered. In addition, the covariance matrix of point cloud coordinates obtained from the precision of the original measurements is employed to construct the corresponding stochastic model. By further introducing an equivalent weight scheme of robust estimation, a novel point cloud fitting method is proposed based on the robust NGH (RNGH) model. Unlike the previous studies, the RNGH-based algorithm can effectively resist the negative effects of outliers with a full consideration of all random errors in various linear and nonlinear fitting problems. To verify the performance of the proposed RNGH, comparative experiments of point cloud fitting are employed in both simulated and real-world scenarios, respectively. The results demonstrate that the proposed point cloud fitting algorithm has significant advantages over LS and RANSAC in terms of the accuracy and robustness.

The spatiotemporal changes of lake water resources objectively reflect not only the process of the water resources balance, but also the ecological environment changes in the lake area. In recent decades, climate changes and human activities have caused great impacts on the spatial distribution of the earth’s water resources and the spatiotemporal process of the surface water cycle, which has caused a series of ecological crises and environmental problems, such as the drying-up of inland lakes, the disappearance of the oasis, water shortage or flooding and water pollution. Therefore, monitoring and fully understanding the dynamic changes of lakes is of great scientific significance for grasping regional water balance, water resources management, and sustainable development of the ecological environment. In this study, we focus on using multi-source satellite data on the estimation of water volume and multi-timescale variations analysis for large scale lakes. This study combines the problems in the practical application of “African Water Action”, taking the largest lake in Africa, Lake Victoria, as the study area, and utilizes long-term serial multi-source satellite data of the past 15 years (2003–2017), including Moderate-resolution Imaging Spectroradiometer (MODIS), Jason-1/-2/-3 and Gravity Recovery and Climate Experiment (GRACE) to perform the comprehensive analysis on the water volume change estimation. Firstly, the satellite altimetry data of Jason-1/-2/-3 and MODIS imagery was used to calculate series of water level, and to extract series of water surface area, respectively. On this basis, a more accurate regression model between the area and water level variation ($\Delta H$) was constructed. Then, the model between water volume variation ($\Delta V$) and $\Delta H$, derived from area-$\Delta H$ model, was applied to calculate the relative water volume of Lake Victoria. Meanwhile, terrestrial water storage (TWS) changes between 2003 and 2016, derived from GRACE data, were also used for a comparative verification of the $\Delta V$ results. The results show the long-term series change trends of $\Delta V$ and the TWS are the same. Finally, the multi-timescale analysis of water volume changes was carried out on different time scales, such as the inter-annual, inter-monthly, and variation period.

Variance component estimation (VCE) is widely applied to adjust random models in the fusion processing of multiple classes of observations. In our previous study, the least squares variance component estimation (LS-VCE) for the classical Gauss-Markov (GM) model was extended to a universal adjustment model: nonlinear Gauss-Helmert (GH) model. However, due to its limited ability to resist outliers, the accuracy of the estimated variance components and parameters will be negatively affected in the presence of outliers. In this paper, the variance inflation principle of robust estimation is further introduced based on our previous study and a robust LS-VCE method for the nonlinear GH model is proposed. To avoid the emergence of negative variance components, the non-negative estimation of the variance components is achieved as well. Unlike the existing studies, the new method can simultaneously mitigate the negative influence of outliers while reasonably adjusting the relative weighting ratios among different classes of observations in the nonlinear GH model. Finally, case studies of point cloud fitting based on original observations and geodetic symmetric transformation are carried out to validate the performance of the proposed method. The results show that when the observations are polluted by outliers, the accuracy of the parameters obtained by the new method has considerable improvement compared with that from the generalized total least squares and the LS-VCE method for the nonlinear GH model. Since the linear/nonlinear GM and errors-in-variables (EIV) models can be treated as the special cases of the nonlinear GH model, the proposed method possesses a wide range of applicability.