Towards inferring elastic structural variations from Earth’s response to surface mass loading

Abstract:

We explore the sensitivity of surface mass loading displacement response to perturbations in elastic structure, with the goal to refine profiles of elastic moduli and density through the crust and upper mantle. Examples of surface mass loads include tidal and non-tidal ocean loads, atmospheric loads and hydrological loads. Using software developed in-house (LoadDef), we derive sensitivity kernels for Love numbers and load Green’s functions (LGFs) using calculus of variations and finite difference methods. Perturbations to the two elastic moduli and density exhibit unique LGF sensitivity patterns, retaining the possibility that the material parameters may be independently constrained given a spatially distributed set of sufficiently accurate loading response observations. To further elucidate the ability to invert for structure in a particular region, a thorough investigation into model resolution must also be performed. We garner a more palpable sense for the effects of structural variations on the response to surface mass loading by calculating and comparing sets of predicted ocean tidal loading (OTL) displacement responses across a global network of land-based locations, generated from convolutions of an ocean tide model with LGFs derived from a variety of reference Earth models. We find that discrepancies between predictions for the M2 harmonic differ by less than 0.2 mm at over 95% of the locations considered, a value generally exceeded, albeit not substantially, by current observational and forward modeling errors. Although predicted discrepancies can reach 2 mm or more at some coastal locations, errors in the ocean tide models and convolution algorithms are also largest near the coasts. As a case study, we examine the residuals between Global Positioning System (GPS) observations and modeled predictions of OTL response across the South American continent. A comparison of ocean models suggests that a common mode (mean displacement) accounts for a dominant portion of the prediction residuals, perhaps due to larger modeling uncertainties at the polar regions. Removing the network mean reduces the sensitivity to a particular ocean model significantly, except at a few coastal stations. Preliminary examinations of the observational residuals show only limited spatial and inter-harmonic coherency.