[cig-commits] r13465 - seismo/3D/ADJOINT_TOMO/flexwin/latex

[alessia at geodynamics.org]

Author: alessia
Date: 2008-12-05 07:33:09 -0800 (Fri, 05 Dec 2008)
New Revision: 13465

Added:
   seismo/3D/ADJOINT_TOMO/flexwin/latex/response.pdf
   seismo/3D/ADJOINT_TOMO/flexwin/latex/response.tex
Modified:
   seismo/3D/ADJOINT_TOMO/flexwin/latex/discussion.tex
Log:
Writing the reviewer response letter

Modified: seismo/3D/ADJOINT_TOMO/flexwin/latex/discussion.tex
===================================================================
--- seismo/3D/ADJOINT_TOMO/flexwin/latex/discussion.tex	2008-12-05 01:29:30 UTC (rev 13464)
+++ seismo/3D/ADJOINT_TOMO/flexwin/latex/discussion.tex	2008-12-05 15:33:09 UTC (rev 13465)
@@ -12,7 +12,7 @@
 
 \subsection{Relevance to adjoint tomography}
 
-The full power of FLEXWIN can only be unleashed for problems --- such as 3D-3D tomography --- which do not require the separation (natural or otherwise) of seismic phases.  The specificity of adjoint tomography, among the 3D-3D tomographic methods, is to calculate the sensitivity kernels by interaction between the wavefield used to generate the synthetic seismograms and an adjoint wavefield whose source term is derived from measurements of misfit between the synthetic and observed seismograms \cite{TrompEtal2005, LiuTromp2006}.  The manner in which the adjoint sources are constructed is specific to each type of measurement (e.g. waveform difference, cross-correlation time-lag, multi-taper phase and amplitude anomaly), but once formulated can be applied indifferently to any part of the seismogram.  Adjoint methods have been used to calculate kernels of various body- and surface-wave phases with respect to isotropic elastic parameters and interface depths \citep{LiuTromp2006}, and with respect to anisotropic elastic parameters \citep{SieminskiEtal2007a,SieminskiEtal2007b}.  Adjoint methods allow us to calculate kernels for each and every wiggle on a given seismic record, thereby giving access to virtually all the information contained within.
+The full power of FLEXWIN can only be unleashed for problems --- such as 3D-3D tomography --- that do not require the separation (natural or otherwise) of seismic phases.  The specificity of adjoint tomography, among the 3D-3D tomographic methods, is to calculate the sensitivity kernels by interaction between the wavefield used to generate the synthetic seismograms and an adjoint wavefield whose source term is derived from measurements of misfit between the synthetic and observed seismograms \cite{TrompEtal2005, LiuTromp2006}.  The manner in which the adjoint sources are constructed is specific to each type of measurement (e.g. waveform difference, cross-correlation time-lag, multi-taper phase and amplitude anomaly), but once formulated can be applied indifferently to any part of the seismogram.  Adjoint methods have been used to calculate kernels of various body- and surface-wave phases with respect to isotropic elastic parameters and interface depths \citep{LiuTromp2006}, and with respect to anisotropic elastic parameters \citep{SieminskiEtal2007a,SieminskiEtal2007b}.  Adjoint methods allow us to calculate kernels for each and every wiggle on a given seismic record, thereby giving access to virtually all the information contained within.
 
 
 It is becoming clear, as more finite-frequency tomography models are published, that better kernels on their own are not the answer to the problem of improving the resolution of tomographic studies.  \cite{TrampertSpetzler2006} and \cite{BoschiEtal2007} investigate the factors limiting the quality of finite-frequency tomography images, and conclude that incomplete and inhomogeneous data coverage limit in practice the improvement in resolution that accurate finite-frequency kernels can provide.  The current frustration with the data-induced limitations to the improvements in wave-propagation theory is well summarized by \cite{Romanowicz2008}.  The ability of adjoint methods to deal with all parts of the seismogram indifferently means we can incorporate more information from each seismogram into a tomographic problem, thereby improving data coverage.

Added: seismo/3D/ADJOINT_TOMO/flexwin/latex/response.pdf
===================================================================
(Binary files differ)


Property changes on: seismo/3D/ADJOINT_TOMO/flexwin/latex/response.pdf
___________________________________________________________________
Name: svn:mime-type
   + application/octet-stream

Added: seismo/3D/ADJOINT_TOMO/flexwin/latex/response.tex
===================================================================
--- seismo/3D/ADJOINT_TOMO/flexwin/latex/response.tex	                        (rev 0)
+++ seismo/3D/ADJOINT_TOMO/flexwin/latex/response.tex	2008-12-05 15:33:09 UTC (rev 13465)
@@ -0,0 +1,177 @@
+\documentclass{letter}
+\usepackage{a4wide}
+\usepackage{setspace}
+\onehalfspacing
+\renewcommand{\deg}{$^\circ$}
+%\begin{document}
+\address{Alessia {\sc Maggi},\\ \'Ecole et Observatoire des Sciences de la Terre\\ 5 rue Ren\'e {\sc Descartes}\\ 67084 {\sc Strasbourg} cedex\\ E-mail : alessia at sismo.u-strasbg.fr}
+\signature{A. {\sc Maggi}}
+\begin{document}
+\begin{letter}{Jeannot Trampert\\Editorial Office\\ Geophysical Journal International}
+
+\opening{Dear Jeannot,}
+
+on behalf of my coauthors, I am sending you the following revised manuscript
+for your consideration,
+
+``An automated time-window selection algorithm for seismic tomography'', by
+Alessia Maggi, Carl Tape, Min Chen, Daniel Chao, and Jeroen Tromp.
+
+Jeroen Ritsema and an anonymous reviewer made many helpful comments on our
+original manuscript, which we have encorporated in this revised version.  The
+revisions have been moderate in extent, and have led to a modified
+introduction, the insertion of a new discussion section on the relevance of our
+algorithm for tomography, and the addition of an appendix on the considerations
+that need to be taken into account in order to tune the algorithm correctly.
+
+In the six months since our first submission, we have made several minor
+adjustments to the FLEXWIN source code.  The revised manuscript
+reflects the most up-to-date version of the code.  
+
+In the following, we reply in detail to the major comments made by the
+reviewers.  Quoted comments are in italics.
+
+Reviewer 1 (Anonymous):
+
+
+\begin{enumerate}
+%
+\item {\em Besides describing the authors' new measurement approach as embodied in
+the FLEXWIN algorithm, this paper also seems to be meant as a tutorial/manual
+for potential users of the software.  My main comment is that, in this
+manual-like function, the manuscript is frustatingly terse or incomplete in
+places.}
+\\
+Our aim is indeed to publish a pedagogical paper that explains enough of the
+workings of the algorithm to convince the readers to use it, without
+turning the paper into a user manual per se (a complete user manual will be
+distributed with the source code).  The algorithm is not a black box, and needs
+to be configuered and tuned for each tomographic problem.  We have discussed
+the general considerations the user should bear in mind when tuning FLEXWIN in
+a new appendix (Appendix A: Tuning considerations).  This appendix complements the method
+section of the main text (Section 2: The selection algorithm), expanding on the
+concepts, and giving advice on starting values for the tuning parameters.  The
+user functions that made up the bulk of the appendix in the originally
+submitted version of the manuscript have been incorporated into this new
+appendix, expanded, and explained in greater detail.
+%
+\item{\em How would the user have to proceed to re-tune the 19 free parameters
+of the algorithm to work with some arbitrary other data set?}
+\\
+The tuning procedure is discussed in great detail in the new Appendix A: Tuning
+considerations.  In particular, the order in which the tuning parameters are
+discussed in this appendix does not follow the order in which they are used by the
+algorithm, but rather the practical sequence we suggest the user adopt in the
+tuning stage.
+%
+\item{\label{3D}\em What are the advantages of this relatively complex approach compared
+to simpler schemes that select time windows based on phase arrivals that are
+known to be useful or robust?}
+\\
+FLEXWIN was designed to solve the problem of automatically picking windows for
+tomographic problems in which phase separation and identification are not
+necessary.  It provides a window-selection solution that is midway between
+full-waveform selection -- which carries the risk of including high noise
+portions of the waveform that would contaminate the tomography -- and the
+selection of known phases or phase-groups based on a-priori arrival times --
+which carries the risk of missing the information contained in the
+non-traditional phases produced by fully 3D structures.
+For problems involving only traditional phases whose separation on the
+seismogram occurs naturally (under certain frequency and epicentral distance
+conditions), the advantage of using FLEXWIN over manual or specifically designed automated
+windowing would be the encapsulation of the selection criteria in well-defined
+user files, leading to greater clarity and portability between studies using
+different inversion methods.  These points are made and argued in detail in the
+revised manuscript (Section 4: Using FLEXWIN for tomography).
+%
+\item{\em The extent to which the choice and functioning of [the tuning
+parameters] is motivated and explained varies.
+For example, equations 16-19 and why this weighting gives the "best"
+time windows is not clearly explained.
+}
+\\
+The revised manuscript provides an extensive discussion of the tuning
+parameters in Appendix A. 
+The equations the reviewer is referring to are equations 20-22 in the revised
+manuscript.  For an explanation on why we choose this weighting scheme, we
+refer the reviewer to paragraph 2 of page 12 of the manuscript, reproduced
+here: For our purposes, an optimal set of time windows contains only windows
+that have passed all previous tests, that do not overlap with other windows in
+the set, and that cover as much of the seismogram as possible.  When choosing
+between candidate windows, we favour those within which the observed ans
+synthetic seismograms are most similar.  Furthermore, should we have the choice
+between two short windows and a longer, equally well-fitting one covering the
+same time-span, we may wish to favour the longer time window as this poses a
+stronger constraint on the tomographic inversion.
+%
+\item{\em It would be nice to have an explicit discussion of what prompted them
+to choose different values for the different parameters in Table 3, depending
+on data set and frequency content.  That way a user could get the feel for how
+these parameters interdepend in practice.  Some of this information is
+dispersed in small pieces across the text and appendix A.  But many values for
+settings are not motivated at all,  starting with the water level.}
+\\
+The reviewer suggests that the user would get the feel for how to set the
+tuning parameters if we presented examples of incorrect tuning, and the steps
+taken to improve it.  Although this trial, error, and correction approach is
+indeed a good way to get a feel for the algorithm in practice, we believe that
+the reader of our manuscript would be better served by an approach in which we
+give descriptions of how each parameter influences the window selection
+(Section 2) and a practical set of starting parameter values and tuning steps
+(revised Appendix A, in which we have also motivated the suggested starting
+parameter value for the water value).
+%
+\item{\em From the discussed examples it emerges that, although the algorithm
+could be made to work on arbitrary data sets, the seismologist still needs to
+contribute quite detailed expertise on each specific data set as to what are
+useful pieces of the time series.  So since we already need to understand our
+data set and its arrivals in detail, why not say "I want to fit P, PP, S, SS,
+surface waves, and if the station is in a basin then I also attempt to fit
+reverberations at low frequencies"?}
+\\
+Yes, the seismologist does need to contribute detailed expertise in order to
+tune FLEXWIN.  If, given the understanding the seismologist has of the data and
+the tomographic probem at hand, he wants to fit only certain named phases, then
+he should write the time dependence of the tuning parameters to reflect this
+wish (see Appendix A for details).  As mentioned in the response to
+point~\ref{3D} above, FLEXWIN was desiged to cope with the complex set of
+arrivals generated by fully 3D strucutres, for which attempting to fit named phases
+would not be productive.
+%
+\item{\em Incomplete list of references: although itself a data processing
+paper, the present work does not put itself in the context of other recent work
+on measurements and data processing for waveform-based tomography.  Earlier
+work may not be geared specifically at adjoint inversions, but the challenge of
+selecting only usable parts of the seismograms has been similar.  More
+specifically, there are several recent papers that use cross-correlations on
+certain subsets of the time and frequency ranges contained in braodband
+seismograms.  When the authors do reference such papers, it is with regard to
+modelling rather than signal processing.}
+\\
+The revised manuscript contains a brief overview of previous studies that
+discuss automated data selection for body and surface wave studies
+(Introduction, page 2, paragraph 3).  These studies are also discussed in
+Section 4 (Using FLEXWIN for tompgraphy), and place the FLEXWIN
+algorithm in the context of recent work.
+%
+\item{\em Also, it is stated without giving references that the STA:LTA ratio
+is a standard tool (in earthquake detection).  Since it plays a central role in
+this paper, some references would be appropriate.}
+\\
+We give references to studies of automated earthquake detection algorithms, and
+phase characterization algorithms that use the STA:LTA tool in the revised
+manuscript (page 6, paragraph 2).  We also discuss in some detail the behaviour
+of the particular implementation of the STA:LTA that we use in our algorithm
+(page 6, last paragraph).
+\end{enumerate}
+
+
+
+
+
+
+\closing{Sincerely yours,}
+
+
+\end{letter}
+\end{document}