Research papersBridging the gap between GRACE and GRACE follow-on monthly gravity field solutions using improved multichannel singular spectrum analysis
Introduction
The monthly gravity field models, derived from the Gravity Recovery and Climate Experiment (GRACE) satellite mission (Tapley et al., 2004) spanning March 2002 to October 2017, were commonly applied in many areas such as hydrology, climatology, marine science, and geophysics for its high resolution and accuracy. For example, GRACE has provided new insights into the Total Water Storage Change (TWSC) (Syed et al., 2008, Landerer and Swenson, 2012, Guo et al., 2018), hydrological mass variations (Abd-Elbaky and Jin, 2019, Chen, 2019, Chandan and Nagesh, 2018), drought or flood detection (Yirdaw et al., 2008, Thomas et al., 2014, Forootan et al., 2019), and global ocean mass change (Chen et al., 2018a, Tamisiea, 2011, Cazenave, 2018), the Antarctic ice sheet mass changes (Chen et al., 2009, Gao et al., 2015, Tapley et al., 2019), the Glacial Isostatic Adjustment (GIA) (Steffen et al., 2008), etc. However, the GRACE mission operated from March 2002 to October 2017, and then the GRACE-FO mission was launched in May 2018. Therefore, there exists about one year of missing data between the two GRACE missions (Li et al., 2020), which leads to the inability to supply continuous geophysical information. How to bridge the gap effectively is valuable to acquire continual GRACE geophysical signals.
Many efforts were carried out to explore the potential of bridging the gap between the two GRACE missions with some alternative observations or approaches. In general, the gap-filling methods can be classified into three aspects. Firstly, satellite laser ranging (SLR) or GPS (Global Positioning System) observations to low Earth-orbiting satellites, such as Swarm, can provide the temporal gravity solutions with a lower spatial resolution (Bezděk et al., 2016, Jäggi et al., 2016, Lück et al., 2018, Teixeira et al., 2019), which provides an opportunity to bridge the gap (Forootan et al., 2020). The potential to bridge the gap of the two GRACE missions with Swarm was investigated by Lück et al., 2018, Forootan et al., 2020 and successfully implemented by Meyer et al. (2019). Secondly, one can reconstruct the GRACE TWSC by determining the relationships between TWSC and corresponding climatic and hydrological variables specifically including rainfall, temperature, etc, which is normally called data-driven approaches (Humphry & Gudmundsson, 2019, Hasan et al., 2019, Hasan & Tarhule, 2020). For example, the Artificial Neural Network (ANN) was adopted to learn the relationship between TWSC and related variables (Ahmed et al., 2019). Forootan et al. (2014) first applied the Independent Component Analysis (ICA) (Forootan and Kusche, 2012) to separate the GRACE signals into their original sources, then produced the reconstruction and derived the relations based on the autoregressive exogenous (ARX) (Ljung, 1987). Li et al. (2020) applied data-driven methods for reconstructing and predicting GRACE-Like gridded TWSC using climate inputs. Besides, other approaches, including the Multiple Linear Regression (MLR) approach (Myers, 1986), Convolutional Neural Network (CNN) (Sun et al., 2019) and Principal Component Analysis (PCA) (Wold, 1987), were also used to extrapolate the GRACE time series of gridded TWSC outside the GRACE period by constructing relationship models with indicators such as precipitation, land surface temperature etc (Li et al., 2020).
Besides the above approaches, many interpolation methods, such as linear interpolation (Zotov and Shum, 2010, Zotov, 2012), cubic spline interpolation (Guo et al., 2018) and least-squares fitting (Rangelova et al., 2010), were used to interpolate missing data with neighboring data. However, the performance of filling gap with these interpolation approaches is basically dependent on the lengths of time series and gaps, availability of neighboring data, and so on (Semiromi and Koch, 2019). In contrast to the above methods, a data-adaptive method SSA\MSSA, can better decompose the time series into a trend, periodic components and noise, and distinguish spatiotemporal patterns (Zotov and Shum, 2010, Zotov, 2012). Three classified SSA-based approaches have been proposed for filling the data gap. The first approach is an iterative interpolation that first needs to fill the missing data with arbitrary numbers (Golyandina, 2010, Golyandina and Zhigljavsky, 2013, Prevost et al., 2019, Wang et al., 2020). The second one is an SSA-based forecasting approach, used for bridging the TWSA gap of two GRACE missions (Li et al., 2019). The third approach named Improved SSA (ISSA) was first proposed in Shen et al. (2015) for stationary time series and performed the algorithm only with available observations. It should be noted that all three SSA-based approaches assumed that the reconstructed or forecasted component can be approximately represented by a time series with finite rank (Golyandina and Zhigljavsky, 2013).
The improved MSSA (Wang et al., 2020), which is the extension of ISSA (Shen et al., 2015), has been successfully used to process the incomplete GRACE monthly solutions. The improved MSSA can fill the missing data especially for consecutive data gaps more exact than the other interpolation and iteration MSSA approaches (Wang et al., 2020). This will provide a new method to fill the large gap between the two GRACE missions accurately. The missing months between GRACE and GRACE-FO account for 14.35% of the total data from April 2002 to March 2020. To test whether improved MSSA can effectively fill the large gap, we will deeply investigate it in this study. As we know that the GRACE gravity field solutions can be validated by comparing with some independent geophysical data, for example, the in-situ data such as ocean bottom pressure and Global Navigation Satellite Systems (GNSS)-derived vertical loads (Chambers and Willis, 2010, Chen et al., 2018b), and the hydrological models such as GLDAS Noah and WGHM data (Li et al., 2020), etc. However, according to Forootan et al. (2020), the in-situ data is more suited to evaluate the low degree spectral variations. Therefore, we use the GLDAS Noah models to demonstrate the performance of improved MSSA and Swarm solutions. The rest of this paper is organized as follows: the datasets used in this study are briefly described in Section 2. We briefly introduced the improved MSSA in Section 3. The results of the filled gap using improved MSSA are presented and compared with the Swarm data in Section 4, the simulation experiment is performed in Section 5 for further evaluating the improved MSSA and finally, the conclusions are shown in Section 6.
Section snippets
GRACE and GRACE-FO data
Many institutions provided different time-variable monthly gravity field solutions due to adopting different processing strategies, known as GRACE level 2 (L2) products. In this paper, we mainly focus on bridging the gap between two GRACE missions using improved MSSA and select the CSR RL06 monthly gravity field models being truncated to d/o 60 from April 2002 to March 2020, which can be directly downloaded by the website of the International Centre for Global Earth Models (ICGEM) (//icgem.gfz-potsdam.de/home
Methodology
Improved MSSA can be directly used to process the incomplete GRACE monthly gravity field solutions without either data interpolation or iteration (Wang et al., 2020). It has great potential for bridging the large consecutive gap of the two GRACE missions relative to other interpolation or iteration MSSA approaches.
Bridging the gap between GRACE and GRACE-FO missions using improved MSSA
To explore the potential of improved MSSA for filling the gap between two GRACE missions, the CSR RL06 monthly gravity field models are analyzed from April 2002 to March 2020 with 33 months missing data. The GRACE/GRACE-FO and Swarm monthly SH coefficients are truncated to d/o 60 and 40, respectively, and are subtracted with the corresponding mean-field (Bettadpur, 2012). The background of the Swarm solution is unified to GRACE and GRACE-FO. All the C20 coefficients of GRACE, GRACE-FO and Swarm
Simulation experiment analysis
The improved MSSA has been compared with interpolation MSSA and iteration MSSA in filling the missing data via randomly deleting some months of data by Wang et al. (2020), and the results showed that improved MSSA outperforms interpolation MSSA and iteration MSSA in filling the missing data. To demonstrate the performance of improved MSSA in filling the gaps as large as that between GRACE and GRACE-FO, we generate the gap by deleting 11 consecutive months (2006.08~2007.06) from CSR RL06 monthly
Conclusions
In this study, the gap between two GRACE missions is bridged by improved MSSA to acquire continual global mass change signals, and results are compared with Swarm monthly solutions. The time series of 18-year (2002.04~2020.03) CSR RL06 monthly gravity field solutions (d/o 60) and the combined monthly Swarm solutions (d/o 40) are analyzed. The infilled gravity field model in May 2018 agrees with Swarm solution in the distribution of spatial signals, indicating the reliability of the infilled
CRediT authorship contribution statement
Fengwei Wang: Methodology, Investigation, Validation, Writing - original draft. Yunzhong Shen: Conceptualization, Methodology, Supervision, Writing - review & editing. Qiujie Chen: Formal analysis, Writing - review & editing. Wei Wang: Data curation, Validation.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work is funded by the National Natural Science Foundation of China (Grant No. 41731069 and 41974002), Dr. Qiujie Chen is supported by the Alexander von Humboldt Foundation in Germany. The Center for Space Research at the University of Texas is acknowledged for supplying the released RL06 GRACE and GRACE-FO monthly gravity field model. The Astronomical Institute, Czech Academy of Sciences (ASU) is also acknowledged for providing the combined monthly Swarm L2 gravity model. The GLDAS Noah
References (64)
- et al.
Hydrological mass variations in the Nile River Basin from GRACE and hydrological models
Geod. Geodyn.
(2019) Satellite Gravimetry and mass transport in the earth system
Geod. Geodyn.
(2019)- et al.
Understanding the global hydrological droughts of 2003–2016 and their relationships with teleconnections
Sci. Total Environ.
(2019) - et al.
Swarm kinematic orbits and gravity fields from 18 months of GPS data
Adv. Space Res.
(2016) - et al.
On the capabilities of the multi-channel singular spectrum method for extracting the main periodic and non-periodic variability from weekly GRACE data
J. Geodyn.
(2012) - et al.
Glacial isostatic adjustment in Fennoscandia from GRACE data and comparison with geodynamical models
J. Geodyn.
(2008) - et al.
Singular-spectrum analysis: A toolkit for short, noisy chaotic signals
Physica D
(1992) Principal component analysis
Chemomet. Intell. Lab. Syst.
(1987)- et al.
GRACE satellite observations of terrestrial moisture changes for drought characterization in the Canadian Prairie
J. Hydrol.
(2008) Forecasting GRACE Data over the African Watersheds Using Artificial Neural Networks
Remote Sensing
(2019)
Time-variable gravity fields derived from GPS tracking of Swarm
Geophys. J. Int.
A global evaluation of ocean bottom pressure from GRACE, OMCT, and steric-corrected altimetry
J. Atmos. Ocean. Technol.
Analyzing large-scale hydrologic processes using GRACE and hydrometeorological datasets
Water Resour. Manag.
Optimized smoothing of gravity recovery and climate experiment (Grace) time-variable gravity observations
J. Geophys. Res-Sol. EA.
Quantification of ocean mass change using gravity recovery and climate experiment, satellite altimeter, and Argo floats observations
J. Geophys. Res-Sol. EA.
Accelerated Antarctic ice loss from satellite gravity measurements
Nat. Geosci.
Reducing leakage error in GRACE-observed long-term ice mass change: a case study in West Antarctica
J. Geodesy
Validation of the EGSIEM GRACE gravity fields using GNSS coordinate timeseries and in-situ ocean bottom pressure records
Remote Sens.
An optimized short-arc approach: Methodology and application to develop refined time series of Tongji-Grace2018 GRACE monthly solutions
J. Geophys. Res-Sol. EA.
An improved GRACE monthly gravity field solution by modeling the non-conservative acceleration and attitude observation errors
J. Geodesy
Deceleration in the Earth's oblateness
J. Geophys. Res-Sol. EA.
Satellites measure recent rates of groundwater depletion in California’s Central Valley
Geophys. Res. Lett.
Separation of global time-variable gravity signals into maximally independent components
J. Geodesy
Multivariate prediction of total water storage changes over west Africa from multi-satellite data
Surv. Geophys.
An iterative ICA-based reconstruction method to produce consistent time-variable total water storage fields using Grace and swarm satellite data
Remote Sens.
Ice sheet mass balance in Antarctica measured by GRACE and its uncertainty
Chinese J. Geophy.
On the choice of parameters in singular spectrum analysis and related subspace-based methods
Stat. interface
Singular Spectrum Analysis for Time Series
Cited by (35)
Analysis of gap filling techniques for GRACE/GRACE-FO terrestrial water storage anomalies in Canada
2024, Journal of HydrologySpatial and temporal downscaling schemes to reconstruct high-resolution GRACE data: A case study in the Tarim River Basin, Northwest China
2024, Science of the Total EnvironmentTotal water storage anomalies reconstruction using noise-augmented u-shaped network: A case study in the Yangtze River Basin
2024, Computers and GeosciencesReconstructing GRACE-derived terrestrial water storage anomalies with in-situ groundwater level measurements and meteorological forcing data
2023, Journal of Hydrology: Regional Studies