Junyang Gou


Last Name

Gou

First Name

Junyang

ORCID

0000-0002-7599-0577

Organisational unit

09707 - Soja, Benedikt / Soja, Benedikt

Filters

2021 - 202644
Reset filters

Search Results

Publications 1 - 10 of 44
  • Downscaling GRACE-derived ocean bottom pressure anomalies using self-supervised data fusion
    Item type: Journal Article
    Gou, Junyang; Börger, Lara; Schindelegger, Michael; et al. (2025)
    Journal of Geodesy
    The gravimetry measurements from the Gravity Recovery and Climate Experiment (GRACE) and its follow-on (GRACEFO) mission provide an essential way to monitor changes in ocean bottom pressure (pb), which is a critical variable in understanding ocean circulation. However, the coarse spatial resolution of the GRACE(-FO) fields blurs important spatial details, such as pb gradients. In this study, we employ a self-supervised deep learning algorithm to downscale global monthly pb anomalies derived from GRACE(-FO) observations to an equal-angle 0.25◦ grid in the absence of high-resolution ground truth. The optimization process is realized by constraining the outputs to follow the large-scale mass conservation contained in the gravity field estimates while learning the spatial details from two ocean reanalysis products. The downscaled product agrees with GRACE(-FO) solutions over large ocean basins at the millimeter level in terms of equivalent water height and shows signs of outperforming them when evaluating short spatial scale variability. In particular, the downscaled pb product has more realistic signal content near the coast and exhibits better agreement with tide gauge measurements at around 80% of 465 globally distributed stations. Our method presents a novel way of combining the advantages of satellite measurements and ocean models at the product level, with potential downstream applications for studies of the large-scale ocean circulation, coastal sea level variability, and changes in global geodetic parameters.
  • Improving the Accuracy of GNSS Orbit Predictions using Machine Learning Approaches
    Item type: Other Conference Item
    Gou, Junyang; Rösch, Christine; Shehaj, Endrit; et al. (2022)
    EGUsphere
    Precise orbit determination is vital for the increasingly vast number of space objects around the Earth. Moreover, accurate orbit prediction of GNSS satellites is essential for many real-time geodetic applications, including real-time navigation. The typical way to obtain accurate orbit predictions is using physics-based orbit propagators. However, the prediction errors accumulate with time because of insufficient modeling of the changing perturbing forces. Motivated by the rapid expansion of computing power and the considerable data volume of satellite orbits available in recent years, we can apply machine learning (ML) and deep learning (DL) algorithms to assess if they can be used to further reduce orbit errors. In this study, we focus on the orbit prediction of GNSS constellations. We investigate the potential of using different ML and DL algorithms for improving the accuracy of the ultra-rapid products from IGS. As ground truth we consider the IGS final products, and the differences between the ultra-rapid and final products are computed and serve as targets for the ML/DL methods. In this context, we combine the advantages of physics-based and data-driven ML/DL methods. Since the major errors of GNSS orbits are expected to be caused by the deficiency of solar radiation pressure models, we consider different related parameters as additional features to implicitly model the solar impact, such as the C0,0 terms of global ionosphere maps. In order to accurately model the effect of solar radiation pressure on the radial, along-track and cross-track components of the satellite orbit system, the geometric relation between the Sun, the satellite and the Earth are also considered. Furthermore, the performances of different ML/DL algorithms are compared and discussed. Due to the temporal characteristics of the problem, certain sequential modeling algorithms, such as Long Short-Term Memory and Gated Recurrent Unit, show superiority. Our approach shows promising results with average improvements of over 40% in 3D RMS within the 24-hours prediction interval of the ultra-rapid products.
  • Machine learning-based prediction of geodetic variables at the Geodetic Prediction Center of ETH Zurich
    Item type: Other Conference Item
    Soja, Benedikt; Crocetti, Laura; Gou, Junyang; et al. (2024)
  • Machine learning-based GRACE accelerometer transplant
    Item type: Other Conference Item
    Behzadpour, Saniya; Gou, Junyang; Kiani Shahvandi, Mostafa; et al. (2023)
  • Assessment of ocean bottom pressure variations in CMIP6 HighResMIP simulations
    Item type: Journal Article
    Liu , Le; Schindelegger , Michael; Börger , Lara; et al. (2025)
    Ocean Science
    Ocean bottom pressure (pb) variations from high-resolution climate model simulations under the CMIP6 (Coupled Model Intercomparison Project Phase 6) HighResMIP protocol are potentially useful for oceanographic and space-geodetic research, but the overall signal content and accuracy of these pb estimates have hitherto not been assessed. Here, we compute monthly pb fields from five CMIP6 HighResMIP models at 1/4° grid spacing over both historical and future time spans and compare these data, in terms of temporal variance, against observation-based pb estimates from a 1/4° downscaled GRACE (Gravity Recovery and Climate Experiment) product and 23 bottom pressure recorders, mostly in the Pacific. The model results are qualitatively and quantitatively similar to the GRACE-based pb variances, featuring – aside from eddy imprints – elevated amplitudes on continental shelves and in major abyssal plains of the Southern Ocean. Modeled pb variance in these regions is ∼ 10 %–80 % higher and thus overestimated compared to GRACE, whereas underestimation relative to GRACE and the bottom pressure recorders prevails in more quiescent deep-ocean regions. We also form variance ratios of detrended pb signals over 2030–2049 under a high-emission scenario relative to 1980–1999 for three selected models and find statistically significant increases in future pb variance by ∼ 30 %–50 % across deep Arctic basins and the southern South Atlantic. The strengthening appears to be linked to projected changes in high-latitude surface winds and, in the case of the South Atlantic, intensified eddy kinetic energy. The study thus points to possibly new pathways for relating observed pb variability from (future) satellite gravimetry missions to anthropogenic climate change.
  • Applications of high-resolution total water storage anomalies from self-supervised data assimilation
    Item type: Conference Poster
    Gou, Junyang; Soja, Benedikt (2023)
    Total water storage anomalies (TWSAs) describe the variations of the terrestrial water cycle, which are essential for better understanding our climate system. Recently, we have developed a self-supervised data assimilation algorithm to combine the advantages of the TWSAs obtained from the Gravity Recovery And Climate Experiment (GRACE) satellite mission and the WaterGAP Global Hydrology Model (WGHM) model, resulting in a high-quality TWSA product with a spatial resolution of 0.5 degrees. In this presentation, we discuss the potential applications and contributions of the downscaled TWSAs, from understanding long-term changes in water storage to monitoring short-term climate extremes. Benefiting from the strong performance in retaining the long-term trends, the downscaled TWSAs allow us to study the water changes on a local scale. For example, the significant negative trends in the High Plains aquifer are partially smoothed out in the GRACE TWSAs since the positive trends in the neighbouring cells caused by progress from dry to wet periods average them out. These signals can be better separated from the downscaled TWSAs. In the aspect of short-term variations, we can derive the well-established GRACE-based indices, such as the flooding potential index and drought severity index, using downscaled TWSAs. The high-resolution indices visually agree with the GRACE-derived ones on larger scales while opening the window to monitor extreme environmental events with higher spatial resolution. As a result, hazard monitoring and prevention could be better targeted and more efficiently conducted by different organisations and governments.
  • Improving the spatial resolution of global mass changes observed by GRACE(-FO) using deep learning - from terrestrial water to the ocean
    Item type: Other Conference Item
    Gou, Junyang; Börger, Lara; Schindelegger, Michael; et al. (2024)
  • Downscaling total water storage anomalies by fusing GRACE measurements and hydrological models using deep learning algorithms
    Item type: Other Conference Item
    Gou, Junyang; Soja, Benedikt (2022)
    The Gravity Recovery And Climate Experiment (GRACE) satellite mission provides a unique opportunity to monitor the monthly global gravity variations, which can be converted to total water storage anomalies (TWSAs). Monitoring TWSAs is an essential way to study the hydrological cycle, which is crucial in many fields, such as climate change, ecological systems, and water resource management. However, the coarse resolution of GRACE TWSAs impedes its regional application. In this study, we develop a deep learning model to fuse GRACE measurements and WaterGAP Hydrological Model (WGHM) simulations. Besides, additional hydrological data from the Global Land Data Assimilation System (GLDAS) are considered as features. The proposed deep learning model combines the principles of convolutional neural networks, residual learning, and an encoder-decoder structure. A novel loss function is designed to maximize the similarities to the WGHM TWSAs while minimizing the average deviations from the GRACE TWSAs within an area. The result is a global model that can produce TWSAs over all the land areas except Greenland with a resolution of 0.5 degrees. Our validations prove that the downscaled TWSAs have a better agreement with the GRACE TWSAs on larger scales while keeping a high spatial correlation to the WGHM TWSAs. Compared to the GRACE TWSAs, the spatial correlation with the WGHM TWSAs is improved from 0.43 to 0.71. On the other hand, the basin- and continent-wise average TWSAs of our downscaled solutions agree better with the GRACE TWSAs with RMS errors lower than 20 mm in most continents, equal to improvements of more than 50% compared to the WGHM TWSAs. Moreover, our method can estimate the trends and seasonal signals in the TWSA time series more accurately without sensing any temporal information. The correlation with the GRACE TWSA trends has been improved from 0.34 (WGHM TWSAs) to 0.80 (our model).
  • Enhancing the spatial resolution of GRACE ocean bottom pressure using deep learning algorithms
    Item type: Other Conference Item
    Gou, Junyang; Börger, Lara; Schindelegger, Michael; et al. (2023)
    The gravimetry measurements from the Gravity Recovery and Climate Experiment (GRACE) satellite mission provide an essential way to monitor changes in ocean bottom pressure (OBP), which is a key variable in understanding ocean circulation. However, the coarse spatial resolution of GRACE OBP hinders resolving mass transports with refined details, particularly on the continental slope. By contrast, classical ocean forward models or reanalyses provide small-scale OBP information, but typically suffer from other problems (e.g., uncertainties in forcing fields, bathymetry, or structural errors in the dynamical formulation). In this study, we downscale the GRACE measured OBP to the eddy-permitting resolution of 0.25º using a self-supervised deep learning model by considering inputs from external high-resolution ocean models. The proposed deep learning model combines the principles of convolutional neural networks, residual learning, and an encoder-decoder structure. By a specific design of the loss function, the model learns to retain the short spatial scale signals contained in the ocean model and calibrate their magnitudes based on GRACE measurements over an area larger than the effective resolution of GRACE. We will compare the downscaled OBP signals to in-situ observations obtained from globally distributed bottom pressure recorders. The possibility of using the downscaled OBP changes for monitoring the meridional overturning circulation via boundary pressures on the continental slope in the North Atlantic will also be discussed.
  • AI-based EOP Prediction by ETH Zurich: the Second EOP Prediction Comparison Campaign and Beyond
    Item type: Conference Poster
    Soja, Benedikt; Kiani Shahvandi, Mostafa; Gou, Junyang; et al. (2023)
Publications 1 - 10 of 44