surface wave inversion

SWinvert: a workflow for performing rigorous 1-D surface wave inversions | Geophysical Journal International | Oxford Academic (oup.com)

inversion algorithm

• Local search algorithms, also known as linearized inversion methods, rely on an initial starting model, which is assumed to be close to the true solution. This assumption may or may not be valid and is unverifiable for real data.
• Global search algorithms rely upon upper and lower limits of each model parameter, which as a whole is often referred to as the parameterization space. Unlike local search methods, global search methods do not require the user to provide a single initial starting model.

これは以前に別のところで聞いたことがあります。表面波から逆計算する際には最小二乗法、アレーから逆算する場合はGAを使用していたのも、よくよく考えると意味はありませんでした。

• When developing trial inversion parameterizations all high-quality site-specific information (geology, boring logs, etc.) should be used to constrain the layering and develop reasonable parameter limits.
• When parameterizing Vs, if limited site specific information is available or site specific information does not extend to a sufficient depth, the layering by number (LN) and layering ratio (LR) parameterizations are recommended.
• At a minimum the lowest misfit model from multiple parameterizations should be reported to quantify the inter-parameterization uncertainty. Furthermore, the inter- and intra-parameterization uncertainty should be reported qualitatively, such as with a plot of 100 lowest misfit Vs profiles for each parameterization, like that in Figure 10c and 11c, and quantitatively, such as with a plot of σln,V s , like that in Figure 10d and 11d, to communicate to the end user the relative (un)certainty of the inversion results.
  

特別ではないですが、経験的な内容だからか見かけない内容です。
特に最後。分散曲線を再現する１次元速度モデルはたくさんあるということがあまり浸透していません。