In this paper, we assess, in the framework of Global Navigation Satellite System (GNSS) meteorology, the accuracy of GNSS propagation delays corresponding to the Saastamoinen zenith hydrostatic delay (ZHD) model and the Vienna Mapping function VMF1/VMF3 (hydrostatic and wet), with reference to radiosonde ray-tracing delays over a three-year period on 28 globally distributed sites. The results show that the Saastamoinen ZHD estimates have a mean root mean square (RMS) error of 1.7 mm with respect to the radiosonde. We also detected some seasonal signatures in these Saastamoinen ZHD estimates. This indicates that the Saastamoinen model, based on the hydrostatic assumption and the ground pressure, is insufficient to capture the full variability of the ZHD estimates over time with the accuracy needed for GNSS meteorology. Furthermore, we found that VMF3 slant hydrostatic delay (SHD) estimates outperform the corresponding VMF1 SHD estimates (equivalent SHD RMS error of 4.8 mm for VMF3 versus 7.1 mm for VMF1 at 5° elevation angle), with respect to the radiosonde SHD estimates. Unexpectedly, the situation is opposite for the VMF3 slant wet delay (SWD) estimates compared to VMF1 SWD estimates (equivalent SWD RMS error of 11.4 mm for VMF3 versus 7.0 mm for VMF1 at 5° elevation angle). Our general conclusion is that the joint approach using ZHD models and mapping functions must be revisited, at least in the framework of GNSS meteorology.

High precision Global Positioning System (GPS) receivers, with the advantages of all-weather work and low cost, are now widely used to routinely monitor precipitable water (PW) vapor. They are so successful that the progressive phasing out of the costly and sparse in situ radio soundings (RS) is now a certainty. Nevertheless, the sub-daily to annual monitoring of high levels of the PW by GPS receivers in the tropics and the equatorial area still needs to be asserted in terms of metrology accuracy. This is the subject of this paper, which focuses on a tropical site located in mid-ocean (Tahiti). The metrology assessment was divided into two steps. Firstly, a GPS internal assessment, with an in-house processing based on the Bernese GNSS Software Version 5.2 and a comparison with the Center for Orbit Determination in Europe (CODE) products. Secondly, an external assessment, with a comparison with RS PW estimates. In contrast with previous works that only used PW estimates from the Integrated Global Radiosonde Archive (IGRA) website, we estimated the RS PW from the balloon raw data. This is especially important in tropical areas, where IGRA estimates only consider balloon measurements taken below approximately 5500 m. We show that, in our case, this threshold is one of the main sources of bias between GPS and RS estimates, and that the formula used to translate the GPS zenith wet delays (ZWD) to PW estimates also needs to be revisited for high level water vapor contents in the atmosphere.

China's lunar exploration program has developed and progressed for more than 13 years. Over this time, Precise Orbit Determination (POD) has guaranteed successful execution of the Chang’e missions, and is the basis for further scientific investigations using radio science data; for example, recovering the lunar gravity field to explore the inner structure and detect craters to study the evolutionary history of the moon. This study briefly reviews the Chang’e series mission orbit and radio tracking measurements. We reprocessed the tracking data and comprehensively summarized the evolution of the POD accuracy and the tracking system precision from the Chang’e 1 to Chang’e 4 missions using in‐house software. Our results show a progression of the tracking station, as the first Chang’e 1 mission only reached several millimeters per second level accuracy in the Doppler measurements and several meters level accuracy for the two‐way range measurements. Later on the accuracy of Chang’e 5T1 Doppler measurements reached about 0.35 mm/s and the accuracy of the Chang’e 3 two‐way range measurements below 0.7 m. Moreover, the newest Chang’e 4 relay satellite achieved POD residuals Root Mean Square for the two‐way range less than a meter and orbital accuracy at 200 m level, in contrast to meters level residual and the half‐kilometer level accuracy achieved during the Chang’e 1 mission. We can clearly see that the POD performance of the Chang’e spacecraft continuously improved. This research study is a reference for future Chinese Lunar exploration missions, as well as Chinese Mars and asteroid explorations.

Lunar craters are essential for spacecraft landing navigation and lunar exploration missions. Deep learning holds great promise in the crater detection task, but still faces some challenges. One major issue is that using incomplete crater catalogs to create the crater dataset, which negatively impacts the training of deep learning models. Additionally, deep learning models are prone to missing detections in overlapping crater scenes. To address these challenges, we propose a novel Lunar Crater Detection Neural Network (LCD-Net), with two key innovations. The first innovation is the fusion of a Reserved Negative Sampling (RNS) module, which alleviates the problem of incomplete crater annotation in the dataset and detects a large number of potential new craters. The second innovation is the addition of an Adaptive Non-Maximum suppression (A-NMS) module, which optimizes the detection of overlapping craters and reduces the miss rate. We use lunar digital elevation model (DEM) to construct the crater dataset for training and evaluating the detection model. The experimental results demonstrate that LCD-Net achieves recall rates to 96% and 95% on the validation and the test sets, respectively. These results are significantly better than the state-of-the-art methods. Furthermore, LCD-Net also ables to detect tens of thousands of credible new lunar craters that can be used as an expansion to existing crater catalogs, making them more suitable for deep-learning dataset annotation. Overall, our work provides an efficient and reliable solution for lunar crater detection.

Almost no in situ seismic data are available for Venus; therefore, we can constrain the deep structure of this planet only from the geodetic data obtained by spacecraft. Of particular interest is Venus’s core‐mantle boundary, which holds clues about the origin and evolution of this celestial body. In this study, we build a series of different mantle/core models of Venus based on several mantle composition models and compare their associated Love numbersk₂with the observed values. Due to the large uncertainty in the observed values ofk₂, the state of Venus’s core cannot be reliably constrained. However, the expected precision ofk₂obtained by the EnVision mission will sufficiently reduce the acceptable model space and contribute to estimating the mantle viscosity structure. Based on current geodetic data, we find that the bottom of Venus’s mantle may not feature a phase transition from perovskite to post‐perovskite if the FeO content of the mantle is less than 8.1 wt%; this region may be different from Earth’s D” layer. Furthermore, we find that the combination of observedk₂andQcan be used to distinguish whether the lowermost part of Venus’s mantle is a high‐temperature basalt layer or a thin thermal boundary layer if observations ofQcan be obtained.

Snowfall data are vital in calculating the surface mass balance of the Antarctic Ice Sheet (AIS), where in-situ and satellite measurements are sparse at synoptic timescales. CloudSat data are used to construct Antarctic snowfall data at synoptic timescales to compensate for the sparseness of synoptic snowfall data on the AIS and to better understand its surface mass balance. Synoptic CloudSat snowfall data are evaluated by comparison with daily snow accumulation measurements from ten automatic weather stations (AWSs) and the fifth generation of the European Centre for Medium-Range Weather Forecasts climate reanalysis (ERA5) snowfall. Synoptic snowfall data were constructed based on the CloudSat measurements within a radius of 1.41°. The results show that reconstructed CloudSat snowfall at daily and two-day resolutions cover about 28% and 29% of the area of the AIS, respectively. Daily CloudSat snowfall and AWS snow accumulation have similar trends at all stations. While influenced by stronger winds, >73.3% of extreme snow accumulation events correspond to snowfall at eight stations. Even if the CloudSat snowfall data have not been assimilated into the ERA5 dataset, the synoptic CloudSat snowfall data are almost identical to the daily ERA5 snowfall with only small biases (average root mean square error and mean absolute error < 3.9 mm/day). Agreement among the three datasets suggests that the CloudSat data can provide reliable synoptic snowfall data in most areas of the AIS. The ERA5 dataset captures a large number of extreme snowfall events at all AWSs, with capture rates varying from 56% to 88%. There are still high uncertainties in ERA5. Nevertheless, the result suggests that ERA5 can be used to represent actual snowfall events on the AIS at synoptic timescale.