Efficient stochastic estimation of the model resolution matrix diagonal and generalized cross–validation for large geophysical inverse problems

Abstract

[1] In recent years, larger geophysical data sets and novel model parameterizations have dramatically increased both the data and model space dimensions of many inverse problems. Because of their relatively low computational expense, trade-off curve corner estimation for choosing regularized models and "checkerboard" tests for evaluating model resolution are commonly applied, despite their limitations. We present and demonstrate a low-cost method for accurately estimating the diagonal elements of the model resolution matrix and for implementing generalized cross-validation (GCV) for optimal regularization parameter selection. The ability to estimate the diagonal of the resolution matrix and GCV function thus facilitates the introduction of additional tools for diagonal resolution analysis and regularization evaluation, even for very large inverse problems, with storage and computational costs comparable to those required for obtaining model solutions. We demonstrate the method using a Tikhonov regularized teleseismic body wave velocity inversion example with approximately 260,000 model parameters, where we validate randomly selected R m diagonal elements against explicitly calculated values and compare GCV-estimated regularized model results to those obtained through traditional methods.Citation: MacCarthy, J. K., B. Borchers, and R. C. Aster (2011), Efficient stochastic estimation of the model resolution matrix diagonal and generalized cross-validation for large geophysical inverse problems,

Mantle-driven dynamic uplift of the Rocky Mountains and Colorado Plateau and its surface response: Toward a unified hypothesis

“…Figure 1B shows the P-wave velocity anomaly of the Rocky Mountain region as derived from teleseismic P-wave residual measurements using the ~70-km-spacing EarthScope Transportable Array (TA) seismic array and all available PASSCAL experiment data (Schmandt and Humphreys, 2010). The tomographic image in Colorado is refi ned based on data from the ~25-km-spacing combined CREST/TA array (Aster et al, 2009;MacCarthy, 2010;MacCarthy et al, 2011). In this image, very sharp velocity gradients are associated with the Lees Ferry knickpoint on the Colorado River (LF of Figs.…”

Deconvolution of gene expression from cell populations across the C. eleganslineage

“…This transformation can be represented as a matrix ( A † A , where A † is the pseudoinverse of the sort matrix, A ), known in geophysical modeling as the model resolution matrix [23]. This resolution matrix depends on both the sort markers used, and the underlying expression pattern for a given gene, resulting in a distinct resolution matrix for each deconvolved gene.…”

Joining statistics and geophysics for assessment and uncertainty quantification of three‐dimensional seismic Earth models

“…A breakthrough was reached in 2005 with the first implementation by Tromp et al [56] of the adjoint method proposed by Tarantola [50], which makes accurate calculation of Frechet kernels in 3D Earth models possible. A new approach based on randomization theory has been recently proposed to estimate resolution matrix of a given 3D Earth model [29,55]. The method has been developed further to provide other resolution measures such correlation length and interparameter tradeoffs, and this with a limited computational cost [12,14].…”