SeisComP Moment Tensors - geofon.gfz-potsdam.de fileEvolution 2007-2009 first automatic version based on Dreger codes / scripts simple data integration instead of deconvolution solutions
Post on 08-Sep-2019
14 Views
Preview:
Transcript
Goal
● (C)MT computation in realtime● More data in → more reliable solution out
● Fast and stable● Cover full magnitude range● Local and global scale● Interactive analysis tools● Full integration into SeisComP
Evolution
2007-2009● first automatic version based on Dreger codes / scripts
● simple data integration instead of deconvolution
● solutions were not convincing
2009-2010● development of interactive analysis tools
● new MT inversion code based on SeisComP framework
● 1-3 components, full waveforms
● algorithm not stable enough
● gained experience with long period data and inversion process
Evolution
2010● integration of Kiwi tools into SeisComP (http://kinherd.org) by
Sebastian Heimann (not yet complete)
● complete rewrite of inversion code
● Gemini PREM Greens function database up to 90 deg distance and 0-700 km depth
● 1-3 components, separation of body- and surface waves
● automatic tool for hypocenter and centroid solutions
● analysis tools need a rewrite due to new algorithm
Hypocenter algorithm
● Time domain● Usage of full responses● Separation of body- and surface waves● Clean data set (SNR, amplitude outliers, ...)● Adaptive depth search● Filtertests● Optimization of solution
Centroid algorithm
● Time domain● Starts from hypocenter solution● Adaptive 4D gridsearch (location, depth & time)
that follows local maxima● Grid size derived from rupture area computed
from relationships by Wells & Coppersmith (1994)
Where are we now?
● Automatic (C)MT tool in testing mode● Processing of results within SeisComP● Dissemination of results to a web page● Promising results in comparison to GlobalCMT
and USGS● No interactive analysis tools again
Future plans
● Analysis tools● W-Phase● Automated tests with global data and reference
solutions from GlobalCMT● Greens function database for other earth
models (e.g. IASP91)● Improve speed (CPU: multithreading, ALG: start
with body waves only)● Automatic detection of periods instead of
testing different predefined periods
Acknowledgment
● John Clinton, ETHZ● Initial support and filter settings
● Sebastian Heimann, Hamburg University● Deconvolution, filtering, decimation, inversion,
science, everything!
● John Ristau, GNS Science● Joachim Saul, German Research Centre for
Geosciences● Greens function database, scientific support and
comments
top related