La lecture à portée de main
Description
Sujets
Informations
Publié par | karlsruher_institut_fur_technologie |
Publié le | 01 janvier 2010 |
Nombre de lectures | 17 |
Langue | English |
Poids de l'ouvrage | 11 Mo |
Extrait
Broadband Seismic Noise:
Classification and Green‟s Function Estimati on
Zur Erlangung des akademischen Grades eines
DOKTORS DER NATURWISSENSCHAFTEN
von der Fakultät für Physik des Karlsruher Instituts für Technologie (KIT)
genehmigte
DISSERTATION
von
Diplom-Geophysiker Jörn Christoffer Groos
aus
Mannheim
Tag der mündlichen Prüfung: 3. Dezember 2010
Referent: PD Dr. Joachim Ritter
Korreferent: Prof. Dr. Friedemann Wenzel
‘Noise is that part of the data that we choose not to explain.’
Scales & Snieder (1998)
Abstract
Several efforts are undertaken in seismology to turn ambient seismic noise into signal. A
key motivation for this approach is to overcome obstacles hampering established methods
of active source and passive earthquake seismology. Techniques based on seismic noise
are independent from earthquake activity or active seismic sources. This is a significant
advantage in areas where the natural seismicity is low and/or where active seismic
sources (explosions, large vibrators) can‟t be used. Passive seismic mno eaise sure ments
can be conducted also in sensitive areas such as city centres and nature reserves due to
their low environmental impact. Seismic noise is a low-cost and easy-to-measure signal
which is available everywhere and at every time.
At the same time the social and economical importance of seismic hazard assessment
and mitigation in (mega)cities is rapidly increasing due to the exploding urbanisation,
especially close to major fault systems. Site effect analyses, wave propagation scenarios
and early warning concepts are high-priority issues for such urban regions. Therefore, the
number of passive seismic measurements in urban environments is permanently
increasing to provide the required information about the underground by utilising seismic
noise. Urban seismic noise is evolving to one of the most important signals of modern
seismology.
This thesis aims at a better determination and understanding of the spatial and temporal
variations of the amplitudes as well as the statistical properties of the (urban) seismic
noise wave field. A good knowledge of these spatial and temporal variations of the
seismic noise is crucial to identify noise sources on the one hand and to be able to
consider the actual noise conditions by the utilisation of seismic noise on the other hand.
A new statistical time series classification is presented which is capable to distinguish
between corrupt and non-corrupt time series as well as to classify non-corrupt time series
in six meaningful noise classes. The time series classification is used to conduct a
comprehensive analysis of the spatial and temporal variations of the seismic noise
between 8 mHz and 45 Hz in the metropolitan area of Bucharest, Romania. This analysis
improves the understanding of the statistical properties of the urban seismic noise due to
temporally and spatially varying noise sources. The combination of the time series
classification with an unsupervised neural network technique, the Self-Organizing Map
method, is demonstrated to be a promising approach to enhance the analysis of complex
urban seismic noise data sets.
The time series classification is furthermore used to realise a data selection approach for
the estimation of Green‟s functions from seismic noise c-rcorrosselation functions. The
implementation of this data selection approach involves a comprehensive evaluation of
the common seismic noise cross-correlation processing. Based on this evaluation a more
flexible processing scheme is realised and critical parameters of the processing such as
the time window length are identified. Furthermore, a wave form preserving time domain
normalisation and a second data selection approach are presented and evaluated in this
context to improve the calculation of seismic noise cross-correlation functions.
Concluding, an effective time series classification for seismic noise time series is
proposed in this thesis. It is demonstrated that the time series classification can be used
to obtain new insights into the temporal and spatial variations of (urban) seismic noise.
The time series classification provides furthermore valuable data selection capabilities for
all methods utilising seismic noise.
i
Contents
Abstract .............................................................................................................................. i
Contents ........................... iii
1 Introduction .............................................................................................................. 1
2 Seismic noise .......... 5
2.1 „Seismic noise‟, an ambiguous te ................................rm ................................ 5
2.2 The seismic noise wave field .......... 6
2.2.1 The sources of the seismic noise wave field .......... 6
2.2.2 The composition of the seismic noise wave field ................................... 7
2.3 Utilisation of seismic noise .............................................. 8
2.3.1 Spectral H/V ratio .................................................. 8
2.3.2 Array methods ....................... 9
2.3.3 Seismic interferometry ..........................................10
2.3.4 Other applications using seismic noise .................................................14
3 Data ........................................................................................17
3.1 URS Bucharest ..............................................................17
3.2 GSN data .......................................................................19
3.3 Inconsistencies in SEED instrument response metadata ...............................20
3.3.1 The SEED format .................................................20
3.3.2 The seismological measuring system ...................................................21
3.3.3 The transfer function: Description of the seismometer ..........................23
3.3.4 The digital low-pass decimation filter ....................25
3.3.5 The inconsistencies ..............................................................................27
3.3.6 Some remarks about evalresp ..............................................................30
3.3.7 Occurrence of the inconsistencies ........................30
3.3.8 Summary ................................32
4 Time series classification ................................33
4.1 Time series properties used for quantification and classification ....................33
4.2 Preparation of the time series prior to the classification .................................35
4.3 Observed deviations from the Gaussian distribution ......35
4.4 Noise classification scheme ...........................................37
4.4.1 Corrupt time series ...............................................37
4.4.2 Non-corrupt time series ........37
4.5 Classification of synthetic data .......................................39
4.6 Summary of chapter 4 ...................................................41
5 Analysis of urban seismic noise ..............43
5.1 Time-Frequency analysis ...............................................44
5.1.1 Urban seismic noise above 1 Hz („microtremo ................................r‟) ...45
5.1.2 Urban seismic noise 0.6-1 Hz (natural and man-made sources) ...........45
iii
5.1.3 Urban seismic noise below 0.6 Hz („microseis ................................m‟) . 47
5.2 Analysed frequency bands and time windows ............... 48
5.3 Analysis of selected working days ................................. 50
5.3.1 Microtremor (1-45 Hz) .......................................... 51
5.3.2 Transitional range (0.6-1 Hz) 52
5.3.3 Microseism (0.04-0.6 Hz) ..................................... 54
5.3.4 Frequency range 0.008-0.04 Hz ........................... 58
5.4 Analysis of the complete URS data set ......................................................... 60
5.4.1 General discussion of the seismic noise amplitudes ............................. 60
5.4.2 Comparison of all / selected working days (low wind velocities) ........... 60
5.4.3 Cn of working days / Sundays (low wind velocities) ............... 62
5.4.4 Comparison of working days with high / low wind velocities ................. 64
5.4.5 Cn of the vertical and horizontal components ........................ 65
5.4.6 Time-domain H/V ratio ......................................................................... 67
5.5 Analysis with Self-Organizing Maps (SOMs) ................. 70
5.5.1 The Self-Organizing Map ..... 70
5.5.2 Application to the URS data set............................................................ 71
5.5.3 Results of the SOM analysis ................................ 72
5.6 Summary of chapter 5 ................................................... 78
6 Improved calculation of seismic noise cross-correlation functions .......................... 81
6.1 Data processing ..........................................