Abstract
Improved recovery of the spatial variability in the effective elastic thickness, Te, of the lithosphere is important to interpret Earth's structure and geodynamics. Here we use synthetic topography and gravity data to systematically compare the recovery of Te using the multitaper windowing scheme of Pérez‐Gussinyé et al. (2004, 2007) with the wavelet approach of Kirby and Swain (2008). We find that spuriously high Te estimates previously present in synthetic tests with the multitaper are artifacts arising from Fourier transform edge effects introduced by splitting the data into land and ocean areas prior to load deconvolution. Hence, we adopt the mixed land‐ocean loading scheme of Kirby and Swain (2008). Using this approach we find that underestimation of Te when the transitional wavelength approaches the window size is largely related to the limited size of the analysis window and not to random correlations of initial loading surfaces as occurs in the wavelet method. To attenuate this bias we apply a correction factor to our resulting multitapered Te and find that this technique more accurately recovers steep Te gradients and small‐scale Te anomalies than the wavelet method. Finally, we recalculate Te of South America using the multitaper and the mixed land‐ocean approach. The new Te distribution is very similar to that of Pérez‐Gussinyé et al. (2007, 2008), suggesting that Fourier transform edge effects are less severe in the real Earth than in synthetic data. This difference arises because coastlines are much less rough in the real Earth than in our synthetic data. On the basis of the synthetic results, we interpret that most of the small‐scale features present in the multitaper but not in wavelet Te estimates of South America presented by Tassara et al. (2007) are real. These include high Te over the Andean flat subduction zones and low Te along the dyke swarms of the Paraná flood basalts, the Tacutu graben, and Amazonian basin.

This publication has 24 references indexed in Scilit: