A deep learning model for reconstructing centenary water storage changes in the Yangtze River Basin.

Since 2002, the Gravity Recovery and Climate Experiment (GRACE) and its Follow-On mission (GRACE-FO) have facilitated highly accurate observations of changes in total water storage anomalies (TWSA). However, limited observations of TWSA derived from GRACE in the Yangtze River Basin (YRB) have hindered our understanding of its long-term variability. In this paper, we present a deep learning model called RecNet to reconstruct the climate-driven TWSA in the YRB from 1923 to 2022. The RecNet model is trained on precipitation, temperature, and GRACE observations with a weighted mean square error (WMSE) loss function. The performance of the RecNet model is validated and compared against GRACE data, water budget estimates, hydrological models, drought indices, and existing reconstruction datasets. The results indicate that the RecNet model can successfully reconstruct historical water storage changes, surpassing the performance of previous studies. In addition, the reconstructed datasets are utilized to assess the frequency of extreme hydrological conditions and their teleconnections with major climate patterns, including the El Niño-Southern Oscillation, Indian Ocean Dipole, Pacific Decadal Oscillation, and North Atlantic Oscillation. Independent component analysis is employed to investigate individual climate patterns' unique or combined influence on TWSA. We show that the YRB exhibits a notable vulnerability to extreme events, characterized by a recurrent occurrence of diverse extreme dry/wet conditions throughout the past century. Wavelet coherence analysis reveals significant coherence between the climate patterns and TWSA across the entire basin. The reconstructed datasets provide valuable information for studying long-term climate variability and projecting future droughts and floods in the YRB, which can inform effective water resource management and climate change adaptation strategies.

Revisiting sea-level budget by considering all potential impact factors for global mean sea-level change estimation.

Accurate estimates of global sea-level change from the observations of Altimetry, Argo and Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-on (GRACE-FO) are of great value for investigating the global sea-level budget. In this study, we analyzed the global sea-level change over the period from January 2005 to December 2019 by considering all potential impact factors, i.e. three factors for Altimetry observations (two Altimetry products, ocean bottom deformation (OBD) and glacial isostatic adjustment (GIA)), three factors for Argo observations (four Argo products, salinity product error and deep-ocean steric sea-level change), and seven factors for GRACE/GRACE-FO observations including three official RL06 solutions, five spatial filtering methods, three GIA models, two C20 (degree 2 order 0) products, Geocenter motion, GAD field and global mass conservation. The seven impact factors of GRACE/GRACE-FO observations lead to ninety combinations for the post-procession of global mean barystatic sea-level change estimation, whose rates range from 2.00 to 2.45 mm/year. The total uncertainty of global barystatic sea-level change rate is ± 0.27 mm/year at the 95% confidence level, estimated as the standard deviation of the differences between the different datasets constituting the ensembles. The statistical results show that the preferred GIA model developed by Caron et al. in 2018 can improve the closure of the global sea-level budget by 0.20-0.30 mm/year, which is comparable with that of neglecting the halosteric component. About 30.8% of total combinations (GRACE/GRACE-FO plus Argo) can close the global sea-level budget within 1-sigma (0.23 mm/year) of Altimetry observations, 88.9% within 2-sigma. Once the adopted factors including GRACE/GRACE-FO solutions from Center for Space Research (CSR), Caron18 GIA model, SWENSON filtering and Argo product from China Second Institute of Oceanography, the linear trend of global sterodynamic sea-level change derived from GRACE/GRACE-FO plus Argo observations is 3.85 ± 0.14 mm/year, nearly closed to 3.90 ± 0.23 mm/year of Altimetry observations.

Reduced misclosure of global sea-level budget with updated Tongji-Grace2018 solution.

The global sea-level budget is studied using the Gravity Recovery and Climate Experiment (GRACE) solutions, Satellite Altimetry and Argo observations based on the updated budget equation. When the global ocean mass change is estimated with the updated Tongji-Grace2018 solution, the misclosure of the global sea-level budget can be reduced by 0.11-0.22 mm/year compared to four other recent solutions (i.e. CSR RL06, GFZ RL06, JPL RL06 and ITSG-Grace2018) over the period January 2005 to December 2016. When the same missing months as the GRACE solution are deleted from altimetry and Argo data, the misclosure will be reduced by 0.06 mm/year. Once retained the GRACE C20 term, the linear trends of Tongji-Grace2018 and ITSG-Grace2018 solutions are 2.60 ± 0.16 and 2.54 ± 0.16 mm/year, closer to 2.60 ± 0.14 mm/year from Altimetry-Argo than the three RL06 official solutions. Therefore, the Tongji-Grace2018 solution can reduce the misclosure between altimetry, Argo and GRACE data, regardless of whether the C20 term is replaced or not, since the low-degree spherical harmonic coefficients of the Tongji-Grace2018 solution can capture more ocean signals, which are confirmed by the statistical results of the time series of global mean ocean mass change derived from five GRACE solutions with the spherical harmonic coefficients truncated to different degrees and orders.

Separation of geophysical signals in the LAGEOS geocentre motion based on singular spectrum analysis.

We recompute the 26-yr weekly Geocentre Motion (GCM) time-series from 1994 to 2020 through the network shift approach using Satellite Laser Ranging (SLR) observations to LAGEOS1/2. Then the Singular Spectrum Analysis (SSA) is applied for the first time to separate and investigate the geophysical signals from the GCM time-series. The Principal Components (PCs) of the embedded covariance matrix of SSA from the GCM time-series are determined based on the w-correlation criterion and two PCs with large w-correlation are regarded as one periodic signal pair. The results indicate that the annual signal in all three coordinate components and semi-annual signal in both X and Z components are detected. The annual signal from this study agrees well in both amplitude and phase with those derived by the Astronomical Institute of the University of Bern and the Center for Space Research, especially for the Y and Z components. Besides, the other periodic signals with the periods of (1043.6, 85, 28), (570, 280, 222.7) and (14.1, 15.3) days are also quantitatively explored for the first time from the GCM time-series by using SSA, interpreting the corresponding geophysical and astrodynamic sources of aliasing effects of K1/O1, T2 and Mm tides, draconitic effects, and overlapping effects of the ground-track repeatability of LAGEOS1/2.

Independent Component Extraction from the Incomplete Coordinate Time Series of Regional GNSS Networks.

Independent component analysis (ICA) is one of the most effective approaches in extracting independent signals from a global navigation satellite system (GNSS) regional station network. However, ICA requires the involved time series to be complete, thereby the missing data of incomplete time series should be interpolated beforehand. In this contribution, a modified ICA is proposed, by which the missing data are first recovered based on the reversible property between the original time series and decomposed principal components, then the complete time series are further processed with FastICA. To evaluate the performance of the modified ICA for extracting independent components, 24 regional GNSS network stations located in North China from 2011 to 2019 were selected. After the trend, annual and semiannual terms were removed from the GNSS time series, the first two independent components captured 17.42, 18.44 and 17.38% of the total energy for the North, East and Up coordinate components, more than those derived by the iterative ICA that accounted for 16.21%, 17.72% and 16.93%, respectively. Therefore, modified ICA can extract more independent signals than iterative ICA. Subsequently, selecting the 7 stations with less missing data from the network, we repeatedly process the time series after randomly deleting parts of the data and compute the root mean square error (RMSE) from the differences of reconstructed signals before and after deleting data. All RMSEs of modified ICA are smaller than those of iterative ICA, indicating that modified ICA can extract more exact signals than iterative ICA.

Specific Direction-Based Outlier Detection Approach for GNSS Vector Networks.

In this paper we propose an outlier detection approach for GNSS vector networks based on the specific direction (i.e., SD approach), along which the test statistic constructed reaches the maximum. We derive the unit vector of this specific direction in detail, and prove that the unit vector is the same as that determined by the outlier estimates in three-dimensional (3D) approach, while the distribution of the maximum test statistic in this direction is the square root of Chi-squared distribution. Therefore, eliminating an outlier along this specific direction can get the same result as that of eliminating all three components of outlier vector in 3D approach. The mathematical equivalence of SD approach and 3D approach is further demonstrated by a real GNSS network. Moreover, preliminary application of the SD approach to detect the abnormal antenna height measurement is carried out in terms of numerical simulations of multiple baseline solutions, and it shows that the SD approach can effectively detect baselines that are directly infected by corresponding receiver antenna height errors.