An Iterative ICA-Based Reconstruction Method to Produce Consistent Time-Variable Total Water Storage Fields Using GRACE and Swarm Satellite Data

Forootan, Ehsan; Schumacher, Maike; Mehrnegar, Nooshin; Bezděk, Aleš; Talpe, M.; Farzaneh, Saeed; Zhang, Chaoyang; Zeng, Yu; Shum, C. K.

doi:10.3390/rs12101639

Cited by 38 publications

(26 citation statements)

References 112 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On contrary, the matrix

can be determined based on the condition that the column vectors of

are as independent as possible. There are several methods to compute the matrix

, including the FastICA [ 24 , 25 , 26 , 27 ], joint approximate diagonalization of eigenmatrices algorithm [ 29 , 37 , 38 ], kernel independent component analysis [ 39 , 40 ], etc. In this work, the FastICA algorithm, which shows obvious advantages in computation, robustness and convergence rate [ 24 ], is used to estimate the rotation matrix

.…”

Section: Methodsmentioning

confidence: 99%

“…Data missing, however, inevitably occurs in GNSS coordinate time series, indicating that the interpolation should be implemented to fill the data gaps beforehand. Several commonly used interpolation methods in ICA, such as cubic spline interpolation [ 28 ], regularized EM interpolation [ 22 ], or iterative interpolation [ 29 ], may introduce false information and cause deviations in the extracted signals especially for the case of a large amount of missing data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

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

Feng¹,

Shen²,

Wang³

2021

Sensors

View full text Add to dashboard Cite

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.

show abstract

“…On contrary, the matrix

can be determined based on the condition that the column vectors of

are as independent as possible. There are several methods to compute the matrix

.…”

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

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

Feng¹,

Shen²,

Wang³

2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Forootan et al. (2020) recently proposed a reconstruction method using only data from European Space Agency's Swarm Earth explorer mission. Launched in 2013, Swarm consists of three identical satellites, with two out of three orbiting the Earth side‐by‐side in nearly polar orbits at ∼400 km and one flying at a higher altitude of ∼500 km.…”

Section: Introductionmentioning

confidence: 99%

“…Of the three methods considered, they found that both autoregressive exogenous (ARX) and MLP network methods simulate the target variables better than multilinear regression (MLR), but the methods are not robust and have a tendency of overfitting. Forootan et al (2020) recently proposed a reconstruction method using only data from European Space Agency's Swarm Earth explorer mission. Launched in 2013, Swarm consists of three identical satellites, with two out of three orbiting the Earth side-by-side in nearly polar orbits at ∼400 km and one flying at a higher altitude of ∼500 km.…”

mentioning

confidence: 99%

Reconstruction of GRACE Total Water Storage Through Automated Machine Learning

Sun

Scanlon

Save

et al. 2021

Water Resources Research

View full text Add to dashboard Cite

The Gravity Recovery and Climate Experiment (GRACE) satellite mission and its follow‐on, GRACE‐FO, have provided unprecedented opportunities to quantify the impact of climate extremes and human activities on total water storage at large scales. The ∼1‐year data gap between the two GRACE missions needs to be filled to maintain data continuity and maximize mission benefits. In this study, we applied an automated machine learning (AutoML) workflow to perform gridwise GRACE‐like data reconstruction. AutoML represents a new paradigm for optimal algorithm selection, model structure selection, and hyperparameter tuning, addressing some of the most challenging issues in machine learning applications. We demonstrated the workflow over the conterminous U.S. (CONUS) using six types of machine learning models and multiple groups of meteorological and climatic variables as predictors. Results indicate that the AutoML‐assisted gap filling achieved satisfactory performance over the CONUS. On the testing data, the mean gridwise Nash‐Sutcliffe efficiency is around 0.85, the mean correlation coefficient is around 0.95, and the mean normalized root‐mean‐square‐error is about 0.09. Trained models maintain good performance when extrapolating to the mission gap and to GRACE‐FO periods (after June 2017). Results further suggest that no single algorithm provides the best predictive performance over the entire CONUS, stressing the importance of using an end‐to‐end workflow to train, optimize, and combine multiple machine learning models to deliver robust performance, especially when building large‐scale hydrological prediction systems and when predictor importance exhibiting strong spatial variability.

show abstract

“…field. The almost one-year data gap from July 2017 is completely covered by the Swarm, so the Swarm orbital measurements can help bridge the gravity data gap (Forootan et al, 2020;Lück et al, 2018;Richter et al, 2021;Teixeira da Encarnação et al, 2020). However, Swarm's data quality is only good at low degrees, for example, below degree 12 (Teixeira da Encarnação et al, 2020).…”

mentioning

confidence: 99%

Filling the Data Gaps Within GRACE Missions Using Singular Spectrum Analysis

Sneeuw

2021

JGR Solid Earth

View full text Add to dashboard Cite

The GRACE and GRACE-FO missions are devoted to continuously monitor the global gravity field with unprecedented spatiotemporal resolution (Tapley et al., 2004). Since all forms of net mass transports come with changes in gravity, the measurement of GRACE missions from the space is valuable for many previously unobservable regions and signals, as long as the magnitude of mass change is large enough (Chen et al., 2007;Yi, Song, et al., 2017). The revolutionary GRACE missions are so successful that the product has been widely used in multiple fields, greatly promoting our understanding of terrestrial water storage, polar and alpine ice melting, sea-level change, seismic activities and other fields (Rodell et al., 2018;Tapley et al., 2019;Wouters et al., 2014).The GRACE product of the earth's gravity field began in April 2002. After several months of testing, the monthly data availability of this product was perfect until 2010, already exceeding the initially designed 5-year lifetime. Nonetheless, the GRACE satellites have sustained for seven more years, with regular dormant periods to prolong the lifespan of the degraded power system (Flechtner et al., 2014). The availability of GRACE data ended in June 2017, but the GRACE-FO satellites were launched in early 2018 and provided their first product in June 2018, resulting in an 11-month gap between the two missions. The time series extension due to GRACE-FO improves the reliability of most GRACE applications and enhances the ability to identify unrecognized processes. However, the disturbing data gaps, especially the almost one-year gap between missions, have impeded the continuous analysis of the 18-year-long observations of GRACE and GRACE-FO to date. Moreover, as we will show below, uneven sampling due to gaps can bias trend estimates. In total, we identified 35 data gaps (Figure 1), most of which are isolated or paired and thus, not a big problem given the nominal temporal resolution of the data. On the contrary, to fill the nearly one-year gap is challenging.The Swarm mission, composed of a constellation of three identical satellites launched in 2013, was designed to study the earth's magnetic field, while its orbital perturbations can also be used to retrieve the gravity

show abstract

An Iterative ICA-Based Reconstruction Method to Produce Consistent Time-Variable Total Water Storage Fields Using GRACE and Swarm Satellite Data

Cited by 38 publications

References 112 publications

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

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

Reconstruction of GRACE Total Water Storage Through Automated Machine Learning

Filling the Data Gaps Within GRACE Missions Using Singular Spectrum Analysis

Contact Info

Product

Resources

About