Iraqi Geological

Abstract


Introduction
A localized natural or artificial seismic source will often emit transient waveforms known as seismic signals.They may be used to find the source, examine source processes, and investigate the composition of the propagation medium.In contrast, "seismic noise" refers to unwanted ground motion components that don't fit into the mental model of the signal being studied.The data that is accessible, the purpose of this study, and the analytic approach all influence how to define and handle seismic noise.As a result, information that is deemed noise in one application may be a helpful signal in another.For instance, surface-wave tomography may be employed with long-period seismic noise whereas micro zonation studies in cities can use short-period seismic noise (Yanovskaya, 2012).
In seismic recordings, disturbing noise can roughly be defined as signals produced by seismometers' susceptibility to environmental factors (such as temperature, air pressure, magnetic field, etc.); Natural vibrations in the environment brought on by things (like the wind and ocean microseisms); artificial vibrations (from industry, traffic, etc.).Secondary signals are produced for a variety of causes, including scattering that occurs during wave propagation in an inhomogeneous medium (Bormann and Wielandt, 2013).
For that purpose and to provide more effective noise reduction in seismic data, several techniques are applied and developed, e.g.The Fourier method (Syberg, 1972), wavelet analysis (Mohammed and Al-Rahim, 2022), bidimensional empirical mode decomposition technique (Al-Rahim, 2016), Re-Fitting multiple filter technique (MFT) (Al-Bakr et al., 2022) and many others.
Seismicity in Iraq has also been studied by many researchers, including them (Al-Ridha and Jasem, 2013;Al-Ridha and Mohammed, 2015;Al-Ridha et al., 2012;Ramthan et al., 2021).The aim of this study is to use the Empirical Mode Decomposition (EMD) which is a new way for adaptively decomposing non-stationary and non-linear signals to reduce the passive noise in the recoeded seismic data.The final performance was evaluated and the noise reduction results were presented using one of IMOS dataset that is seismic event recorded at Kirkuk seismological station.
The waveforms data, gathered from these stations belong to IMOS.The network covers a wide area and is aimed at monitoring the active fault system and is interested in continuous background seismic activity, essentially including micro and moderate earthquakes.Most broadband stations usually record 100, 40, or 10 samples per second (SPS).The Quanterra Q330 digitizer acquisition system and STS-2 Streckeisen seismometers make up each station's digital recording system.All seismic stations are located in urban areas, and buildings have lately been built close to them.However, local noise sources vary from station to station, including roadways, population density, daily and seasonal fluctuations.Human activity, noisy vehicles, and equipment like diesel generators and water pumps are the main sources of noise close to stations (such as the BHD station) (Mohammed and Al-Rahim, 2022).

The Empirical Mode Decomposition (EMD)
Most analytical techniques are designed to analyze nonstationary but linear data or non-linear but stationary data.The EMD method works with nonstationary and nonlinear data acuratly.The use of such decomposition techniques frequently results in components that have no physical significance as well as harmonic, inherent to the Fourier transform.The Hilbert-Huang transform is an alternate method that (Huang et al., 1998) developed for data analysis.
The Hilbert transform is a useful technique for computing the instantaneous frequency since it makes data analytical.In comparison to the conventional Fourier and wavelet decompositions, the EMD is a type of basis decomposition that is frequently used for spatial or temporal data.Firstly, it works better when there are non-stationary, non-linear systems present.Second, no prior parametric formulation is necessary for the specification of the basic functions.Indeed, the basis functions are recursively resolved as the fluctuations are automatically and adaptively retrieved from the signal (Campi, 2022).
Breaking down signals into a group of suitable oscillatory functions known as intrinsic modes functions (IMFs) is the essential tenet of EMD.
The IMF represents a plain oscillatory mode, akin to the basic harmonic function, in contrast to wavelets or Fourier analysis.The IMF is more general since it can have time-varying amplitude and frequency instead of constant amplitude and frequency, as in a simple harmonic component.The execution of a class of locally modified basis functions, which will be applicable even in the case of non-stationary signals, is what makes this representation important (Damerval et al., 2005).
In other words, any function that verifies the following characteristics conforms to the traditional definition of an (IMF): • There can be no more than one difference between the number of zero-crossings and extrema.
• The envelope defined by the local maxima and minima has the same mean value of zero (Huang et al., 1998).
Realizing that the EMD decomposition results in a deterministic decomposition of the signal rather than the stochastic decomposition generally taken into account in a time-series structure is also crucial.This affects the possibility of forecasting and the creation of statistically significant confidence ranges for such a decomposition Fig. 2.
The ordered oscillatory index of the produced (IMFs) is one of its properties.To create (IMFs), the EMD method iteratively reduces oscillation indices.The base extraction procedure is an algorithm named sifting (Campi, 2022).

Signal Denoising using Empirical Mode Decomposition
This section will explain the practical steps for applying the EMD method using Matlab.
• The Waveform is first converted from SAC to ASCII format.
• The EMD algorithm is applied to the seismic event file in ASCII format.
• Sifting starts when this algorithm is put into practice.The Hilbert-Huang transformation's filtering procedure is connected to the IMF components' signal reconstruction characteristic.The EMD primarily functions as a binary filter bank, which implies that low-order IMF components catch high oscillations while higher-order IMF components collect low-frequency waves (Flandrin et al., 2004a).• To this end, reconstructing the signal is carried out by those components that contain the signal energy, and leaving the noise captured in IMF of a higher order, provides a noise-free equivalent to the raw data.• After the screening process is completed and the IMFs are obtained, the de-noised seismic signal is determined, which is either the result of an assembly of some IMF components or one of the IMF will represent the clean signal and can be denoted by D. • If the seismic signal is very noisy and the pure signal cannot be determined, then step 5 can be repeated using the signal D and repeat the sifting process in steps 2 and 3 until the noise-free seismic signal represented by IMF is obtained.

Results
In the current study, the seismic record by IKRK station is used.The date of the seismic event is 10 February 2021; the origin time is 04:58:40; the coordinates are 36.388as latitude and 41.286 as longitude, magnitude is 4.2 and the depth is 10 km.
The unfiltered vertical component (Z) of the above-mentioned seismic event (which was recorded at the Kirkuk station) is shown in Fig. 3.The seismogram of this componentshows that the noise interferes with the main earthquake signal in the high-frequency segment.This seismic signal was imported into MATLAB R2021a and taken as input data for decomposition using different methods of analysis such as EMD.
Comprehensive ambient noise processing techniques are preferable to individual processing techniques when it comes to identifying noise-induced interference with seismic signals and how it affects the results of increasing the signal-to-noise ratio.These techniques make it possible to identify, isolate, and remove unnecessary information while extracting useful signal information.The kind of data, the characteristics of the station locations, and the kind of processing technology employed all have a significant role in its accuracy and efficacy.
The suggested seismic signal from Kirkuk station was subjected to the EMD algorithm; in this setting, (Flandrin et al., 2004b)characterized EMD to work as a binary filter bank alone.Thus, EMD may be used to extract valuable information that is inherent in the signal and has gotten more attention in the noise reduction area of the signal.This can be understood as having several filters for interfering frequency content.
The total decomposition result consisted of a total of 19782 samples and IMFs values were obtained at 30 levels.Fig. 4.
IMF1 is not useful for seismic signal extraction because it is noisy, thus IMF4 can be chosen as the best low-noise level from from the other (IMFs) levels.Fig. 5.
This experimental procedure is optional and theoretical and is based on the researcher's bias of choosing a clean signal.It is also possible to sift the selected clean signal again and apply the EMD technique to it again when needed to obtain better and suitable results for working on it.

Discussion
The kind of data, the characteristics of the station locations, and the kind of processing technology employed all have a significant role in its accuracy and efficacy.In this study, the EMD method was applied and discussed as follows: 1. EMD is a fresh method for adaptively decomposing non-stationary and non-linear signals.It breaks down a signal depending on its time scale properties.Additionally, it offers improved frequency resolution in the time-frequency domain and may obtain the frequency distribution at any moment.Because the EMD technique outperforms wavelet representation, a new dimension in the signal analysis is made possible.As shown in Fig. 4.
2. Despite its successes in practice, the EMD methodology is simply an algorithmic approach without a direct, systematic optimization of the process or a solid theoretical foundation.The abundance of EMD literature bemoans the absence of a solid mathematical foundation for the concept of EMD.Numerous problems including theoretical analysis and comprehension, performance improvement, optimization, and other related aspects need to be resolved.
3. Using EMD, it is possible to eliminate specific components such as noise and reconstruct the signal.You can also extract relevant components for further analysis.
This decomposition is based on the direct extraction of energy related to multiple intrinsic time scales, which are the most crucial system factors, which are expressed in IMFs form, from which instantaneous frequencies can be determined.

Conclusions
Through the results obtained, this technique proves to be beneficial for the reconstruction of nonlinear and non-stationary signals.The original signal's entire and nearly orthogonal basis is removed using EMD filters.These so-called IMFs guarantee completeness and are hence enough to characterize the signal.The fundamental property of EMD is that different frequencies in time may be kept since all of the functions into which a signal is divided are in the time domain and have the same length as the original signal.So, the EMD is suited for analyzing earthquake motion records.

Fig. 1 .
Fig.1.Geographical distribution of stations in the study area.

Fig. 3 .
Fig.3.The unfiltered vertical component (Z) of the seismic event recorded by IKRK station.The date of the seismic event is 10 February 2021; the origin time is 04:58:40; the coordinates are 36.388as latitude and 41.286 as longitude, magnitude 4.2 and the depth is 10 km.show how the noise interferes with the main earthquake signal in the high-frequency segment.

Fig. 4 .
Fig.4.The EMD algorithm applied to the proposed seismic signal recorded in Kirkuk station.