Session: A.1 - Analysis Techniques
Title: An improved approach for estimating trends in mass variations derived from GRACE
Presenter: Didova, Olga
Co-Authors: B. Gunter; R. Riva; R. Klees; L. Roese-Koerner
Abstract: GRACE plays a central role in estimating mass trends. Reliable estimation of these trends and corresponding uncertainties is crucial towards understanding the underlying processes driving these variations, particularly in the context of climate change. The importance of realistic uncertainties is further amplified when combining GRACE with other data. Therefore, the signal and noise must be modeled appropriately. Typically, trends and co-estimated seasonal signals are modeled deterministically with the remaining signal appearing as noise. In this study, a new approach is presented, in which all signal constituents are modeled stochastically allowing them for physically natural variations in time. For this purpose, state space models are defined and solved through the use of a Kalman filter. Special attention is paid towards carefully estimating the noise parameters, which is an essential step in the pursued approach. The results suggest that the developed technique allows for a more reliable trend estimation compared to the traditionally used least-squares adjustment technique.
Back to top
Decorrelation and filtering of GRACE time variable gravity field solutions using full covariance information
Presenter: Horvath, Alexander
Co-Authors: M. Murböck; R. Pail
Abstract: Aiming for an as accurate as possible estimation of mass trends in Antarctica or other regions, based on global GRACE gravity field solutions, calls for best possible post processing strategies. Decorrelation filters employing static covariance information have already been developed in the past (e.g. DDK filter series by Jürgen Kusche), but covariance information for a decade long recent time series was not publicly available since the publication of the ITSG temporal gravity field model in October 2014.
With this work we aim to use this time series with its evolving correlation structures due to changing mission configuration (e.g. orbital height) and instrument characteristics over time.
Proper reduction of correlated errors is a crucial step towards trend estimation. For this purpose we analyzed the existing series of DDK filters based on static or simplified assumptions on the correlation structure of spherical harmonic coefficients and target signals. To analyze the potential gain using month to month full covariance information we have tested the impact of certain simplifications (e.g. the ones applied for the DDK filters) with respect to the full covariance information in a closed loop simulator. Variables investigated within this experiment are for the error covariance matrices full, order-block and diagonal structures and for the signal variance matrices “real”, mean and “l-rule” (degree dependent) signal content.
Based on the outcome of the simulated results we computed new decorrelation filters using full error covariance information and investigated the impact on basin scale mass change estimates in the Antarctic region. These results are compared to the ones obtained from DDK, Swenson & Wahr type and other filters as well as independently derived results from e.g. radar altimetry.
Back to top
Title: Discussion on the compared virtues of un-stabilized and stabilized GRACE solutions
Presenter: Bourgogne, Stéphane
Co-Authors: J. Lemoine; S. Bourgogne; S. Bruinsma; R. Biancale
Abstract: The GRACE PI groups (JPL, CSR and GFZ) as well as many international groups (UBERN, TUGRAZ, TONGJI,...) produce unconstrained GRACE solutions which need to be filtered using dedicated filters (e.g. Swenson & Wahr or DDK-type filters) in order to be used. On the other hand some groups have adopted a different approach where a regularization process is applied during the solution computation, either in the form of a stabilization constraint (GRGS RL02) or using Truncated Singular Value Decomposition (GRGS RL03-v1) or by changing the set of recovered coefficients (Ramillien & Seoane or GSFC's mascons).
In all cases the whole process results in a smoothing of the recovered signal leading to a damping of the solutions at the shortest wavelengths, compared to the true signal. However this damping is not the same in all cases and depends strongly on the method used. We propose here to discuss the ability of the different strategies to recover the smallest wavelengths observable with GRACE. This will be done by analyzing the residual spectral content of the solutions as well as by confronting the different solutions with in situ data (altimetry and ocean bottom pressure data in particular).
Back to top
Title: Improved estimates of regional mass change: Mascons, forward models, and iteration
Presenter: Loomis, Bryant
Co-Authors: S. B. Luthcke; T. Sabaka
Abstract: The GRACE time-variable gravity mascon solution from the NASA Goddard Space Flight Center employs forward models, solution iteration, and optimized regularization for the direct measurement of 1-arc-degree global mass flux parameters each month. Unlike spherical harmonics, mascon estimation enables the application of a regularization matrix in the least-squares parameter adjustment that is informed by knowledge of the geophysical signals, resulting in solutions with a significantly improved signal-to-noise ratio. A primary benefit of mascons to the scientific end user is that regional mass change is easily computed without the need to select and design one of the many post-processing filters available in the literature required for analysis of the traditional spherical harmonic solutions. As the scientific application of mascon solutions increases in popularity it is critical that mascon solutions are appropriately validated to ensure that the regularization procedure has reduced the noise without significantly altering the recovered signal. A common GRACE validation approach is to perform comparisons to independent datasets from other geodetic missions and in situ measurements with appropriate corrections applied. Though useful, such comparisons have limitations and challenges due to the uniqueness of the GRACE data information content and its spatial and temporal resolution.
In this study we seek to validate our current mascon solution using the spherical harmonics obtained from the same Level-1B data processing. We present the results of a simple validation procedure that compares the time series of regional mass change obtained from our latest regularized mascon solution to those obtained by summing the un-regularized “Delta” spherical harmonics determined on the final iteration of our estimation procedure and the full-resolution background model applied in the data reduction. Several post-processing filtering methods are applied to both the “Delta” and non-iterated spherical harmonics. As a result of this analysis we make the following conclusions: 1.) Mascon solution iteration is an effective way to reduce the uncertainty of regional mass change estimates; 2.) The influence of the selected filtering technique is significantly reduced when applied to the “Delta” spherical harmonics as compared to the non-iterated solution; and 3.) We demonstrate the successful validation of our mascon solution, most notably through the analysis of inland and marginal seas, which are ideal test cases due to their relatively small geographic size and potential for signal leakage.
Back to top
Title: Progress towards high resolution mass grid solutions from GRACE for global applications
Presenter: Save, Himanshu
Co-Authors: S. Bettadpur, B. D. Tapley
Abstract: The range-rate data from Gravity Recovery and Climate Experiment (GRACE) is inverted into global equal-area mass grid solution at the Center for Space Research (CSR). The mass anomalies for the equal area regions, that are roughly 120 km wide, are estimated while stabilizing the ill-posed estimation problem using Tikhonov Regularization. These solutions are intended to be used for applications in Hydrology, Oceanography, Cryosphere etc without any need for post-processing. This paper evaluates these solutions with emphasis on spatial and temporal characteristics of the signal content. These solutions will be validated against multiple models and in-situ data sets.
Back to top