Session: B.2 - Solid Earth Sciences
Title: GRACE Science Objectives in Polar Regions and the Annoying Problem of GIA Uncertainty
Presenter: Ivins, Erik
Co-Authors: L. Caron
Abstract: Since the launch of GRACE in spring of 2002 there have been dozens of peer reviewed articles discussing scientific results that are dependent upon having a model of the secular rate of change that follows from employing one or more models of glacial isostatic adjustment (GIA) that are touted as having either uncertainties so small as to be ignored, or tested by sampling the range of different model predictions. Rarely do we see the opportunity to be able to provide justifiable uncertainties. Using the new Bayesian statistical approach of Caron et al. [2017a, b], we see how such statistics may elevate GRACE science in its goal of assessing secular changes in polar regions. For example, the Bayesian statistics apply directly to the question of the mass balance of ice sheets. We also discuss the more nuanced problem of over interpreting the statistics in places of known anomalous mantle, recovered from both seismic tomographic imaging and forward model adjustments using GNSS station displacements. Additionally, we offer insights on the role to be played by systematically filling in gaps in GIA models wherein both the lateral mantle viscosity variations, and the Little Ice Age loads are both of significance. Here, targeted GRACE-GIA analysis with GNSS station uplift data must be implemented [e.g., Ivins et al., 2011; Richter et al., 2016].
Back to top
Title: Using New Data to Construct the Bayesian Statistics for GIA with Applications to GRACE
Presenter: Caron, Lambert
Co-Authors: E. Ivins, S. Adhikari
Abstract: Traditionally glacial isostatic models have incorporated data sets by examining a host of parameters for Solid Earth 1-D mantle viscosity and lithospheric thickness on the one hand, and timing and magnitude of ice gain and collapse on the other. Generally, a ‘best-fitting’ global model has been sought by tedious experimentation with parameter adjustment. More recently, sophisticated mathematical approaches are emerging which develop adjoint properties of forward solutions (Al-Attar and Tromp, 2014; Martinec et al., 2014). However, there has been limited attempts to tap into the wealth of new data sets in combination with advanced methods capable of capitalizing on the potential rewards that may come from those data. Here we free both ice history parameters simultaneously with mantle structure in order to use more than 450 GNSS bedrock uplift time series for vertical land motion (VLM) and ~11,500 relative sea-level (RSL) data to compute probability distribution functions from order 100,000 simulations. The result is a GIA model with Bayesian statistics, and all the information that such statistics convey about model prediction uncertainty. In this talk we describe how the model functions and suggest ways in which the GRACE communities may take advantage of such model prediction and uncertainty.
Back to top
Poster Title: Spatial correlations in ground deformation for terrestrial reference frame estimation
Presenter: Chin, T. Mike
Co-Authors: C. Abbondanza, R.S. Gross, M.B. Heflin, J.W. Parker, B.S. Soja, X. Wu
Abstract: A terrestrial reference frame (TRF) is estimated from space-geodetic station-position data sets that are several decades long. The primary model upon which the data are combined has been the linear motion due mostly to plate tectonics and post-glacial rebound. In recent TRF realizations such as ITRF2014 and JTRF2014, models of local motions are introduced to improve the estimation accuracy. These additional models include post-seismic displacements as well as annual, semi-annual, and random ground deformations due to atmospheric, oceanic, and ground water storage loading. Although such motions are often regionally correlated, the models of these ground deformations are formulated to be strictly local in space at present. We are investigating application of the GRACE data to improve the TRF stochastic models (used in JTRF2014) by determining spatial statistics of the deformation of the Earth's surface caused by mass loading. A potential target of improvement is the non-uniform distribution of the geodetic observation sites, which can introduce bias in estimated TRFs. In this presentation, several GRACE data sets (including MasCon and vertical deformation solution) are used to determine spatial correlation in ground deformation at geodetic station sites. Analogous empirical correlations have also been obtained from the recent GFZ loading model outputs. After removing linear trends as well as annual and semi-annual periodic terms from each data set, the spatial correlation patterns from these data sets generally agree with each other in regional (sub-continental) range, but some major disagreement can be observed over longer distances or inter-continentally. We are applying a regional filter on the correlation maps to extract only the continental scale spatial patterns common in all data sets, and this filtered version of spatial patterns will be the basis of the new TRF stochastic model.
Back to top