The Shuttle Radar Topographic Mission (SRTM) digital elevation model (DEM) is a widely utilized product for geological, climatic, oceanic, and ecological applications. However, the accuracy of the SRTM DEM is constrained by topography and vegetation. Using machine learning models to correct SRTM DEM with high-accuracy reference elevation observations has been proven to be useful. However, most of the reference observation-aided approaches rely on either parametric or non-parametric regression (e.g., a single machine learning model), which may lead to overfitting or underfitting and limit improvements in the accuracy of SRTM DEM products. In this study, we presented an algorithm for correcting SRTM DEM using a stacking ensemble machine learning algorithm. The proposed algorithm is capable of learning how to optimally combine the predictions from multiple well-performing machine learning models, resulting in superior performance compared to any individual model within the ensemble. The proposed approach was tested under varying relief and vegetation conditions in Hunan Province, China. The results indicate that the accuracy of the SRTM DEM productions improved by approximately 46% using the presented algorithm with respect to the original SRTM DEM. In comparison to two conventional algorithms, namely linear regression and artificial neural network models, the presented algorithm demonstrated a reduction in root-mean-square errors of SRTM DEM by 28% and 12%, respectively. The approach provides a more robust tool for correcting SRTM DEM or other similar DEM products over a wide area.

This paper presents a terrain-related method to simultaneously correct the global error trends and local linear/nonlinear terrain-related errors of the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM), which have not been focused in most of the previous methods. To meet this goal, an adaptive strategy for modelling the SRTM DEM errors is first proposed, especially over mountainous areas, based on the Bayesian information criterion. Then, the M-estimator, instead of the ordinary least squares’ solver, is utilized to estimate the model parameters of the constructed model to improve the estimation robustness. The proposed method was tested over the Zhangjiajie areas of China, where the ground surface terrain varies from plains to steep mountains. The results show that the errors of the SRTM DEM over the region of interest decrease from 10.1 m to approximately 8.1 m after correction with the proposed method, indicating an improvement of approximately 20%. In addition, compared with two existing and common methods that can correct the global error trends and local linear terrain-related errors of the SRTM DEM, the accuracy of the corrected SRTM DEM improves by approximately 29% and 27.9% over mountainous areas with slopes larger than 20° when the proposed method is used. Moreover, the proposed method will be beneficial to the correction of other airborne or spaceborne DEM products, especially over mountainous areas.

Potential of precipitable water vapor (PWV) maps retrieved by remote sensing satellites can address the geometry defect of GNSS observations in tropospheric tomography. The second-generation geostationary meteorological satellite Fengyun-4A (FY-4A) of China can provide PWV products with high spatial (4 km) and temporal (15 min) resolutions. This paper presents the first study on water vapor tomography by integrating GNSS measurements and FY-4A products using node-based parameterized method. Layer PWV (LPW) products of FY-4A instead of the total PWV are adopted that can increase the rank of the tomographic equation significantly. The integrated tomography model is validated with observational data collected over the three-month period of June to August 2020 from 124 GNSS stations in Hunan Province, China. Assessments using radiosonde and European Centre for Medium-Range Weather Forecasts ReAnalysis 5 (ERA5) data demonstrate the better performance of the integrated model against the traditional model using GNSS data alone. In the assessment with radiosonde profiles, the integrated model improves the tomographic solutions upon the traditional model by 40.58% and 36.33% for 30 and 15 min resolutions, respectively. Root mean square errors (RMSEs) of density differences between ERA5 and the integrated model vary from 1.24 g/m ³ to 2.82 g/m ³ throughout the study area. RMSEs vertically decrease from ~5 g/m ³ at the bottom to ~0.5 g/m ³ at the top layer of about 10 km. This work demonstrates the benefit of high-quality FY-4A products to GNSS tomography because they can effectively mitigate the ill-posed problem of inverse process.

Hunan continuously operating reference station network is one of the most important infrastructure of regional geospatial datum in Hunan province, China. It provides the official 24 hour RTK service to the public. How to revealing the user behavior pattern by spatio-temporal analysis on location based big data, is significant for the service promotion. With procedures such as cleaning, sampling, et al., the usage count, fixing rate and network delay data from August, 2017 to July, 2018, are firstly analyzed on multiple spatial and temporal scales. The results show that, the user behavior is strongly correlated to the surveying field work habits. Overall, the usage count is much more in the central and eastern regions, developed and plain or hill area. While it is less in the western, underdeveloped, mountain and lake area. The suburbs is the most popular area. The usage count is also correlated to the local economic profile. Meanwhile, the Huaihua and Shaoyang city need to be paid more attentions in promotions. The hot spots revolution in 24 hours can be divided into 6 stages as sleeping, recovery, first and second busy stages, adjustment and dormancy, when the hot spot successively increased and decreased around the Changsha- Zhuzhou-Xiangtan urban agglomeration and other 11 urban centers in Hunan province..

In most previous studies of tropospheric tomography, water vapor is assumed to have a homogeneous distribution within each voxel. The parameterization of voxels can mitigate the negative effects of the improper assumption to the tomographic solution. An improved parameterized algorithm is proposed for determining the water vapor distribution by Global Navigation Satellite System (GNSS) tomography. Within a voxel, a generic point is determined via horizontal inverse distance weighted (IDW) interpolation and vertical exponential interpolation from the wet refractivities at the eight surrounding voxel nodes. The parameters involved in exponential and IDW interpolation are dynamically estimated for each tomography by using the refractivity field of the last process. By considering the quasi-exponential behavior of the wet refractivity profile, an optimal algorithm is proposed to discretize the vertical layers of the tomographic model. The improved parameterization algorithm is validated with the observational data collected over a 1-month period from 124 Global Positioning System (GPS) stations of Hunan Province, China. Assessments by GPS, radiosonde, and European Centre for Medium-Range Weather Forecasts (ECMWF) ReAnalysis 5 (ERA5) data, demonstrate that the improved model outperforms the traditional nonparametric model and the parameterized model using trilinear interpolation. In the assessment by GPS data, the improved model performs better than the traditional model and the trilinear parameterized model by 54% and 10%, respectively. Such improvements are 31% and 10% in the validation by radiosonde profiles. In comparison with the ERA5 reanalysis, the improved model yields a minimum overall root mean square (RMS) error of 8.94 mm/km, while those of the traditional and trilinear parametrized models are 10.79 and 9.73 mm/km, respectively. The RMS errors vertically decrease from ~20 mm/km at the bottom to ~5 mm/km at the top layer.

With the rapid development of the ground infrastructure as CORS and mobile communication stations, the high precision LBS expand from the limited fields to the public. The comprehensive LBS platforms which meet the demands such as multi-precision level, massive user and desirable scalability, become critical for the public applications. In this paper, a new high precision LBS architecture based on GNSS CORS network is presented. The platform is deployed in the cloud computational environment. It is divided into two parts: data processing and allocation system individually. The virtual grids differential corrections are involved to optimize the real-time CORS stations data processing. Besides, the API/SDK module is introduced as a new service mode, which integrates the meter, decimeter, centimeter level LBS applications, respectively realized by the coordinates, pseudo range and carrier phase differential. With the daily needs of HNCORS network, the platform “SoBDS” is established. The experiments for decimeter level example shows that, the platform could support one million users, and it is compatible and scalable. The real-time positioning accuracy is ±0.425m and ±0.436m in east and north. The total accuracy is ±0.612m and ±0.773m in horizontal and vertical direction.