• Journal of the Earthquake Engineering Society of Korea
• /
• v.25 no.2
• /
• pp.71-81
• /
• 2021

This paper presents a real-time, false-pick filter based on deep learning to reduce false alarms of an onsite Earthquake Early Warning (EEW) system. Most onsite EEW systems use P-wave to predict S-wave. Therefore, it is essential to properly distinguish P-waves from noises or other seismic phases to avoid false alarms. To reduce false-picks causing false alarms, this study made the EEWNet Part 1 'False-Pick Filter' model based on Convolutional Neural Network (CNN). Specifically, it modified the Pick_FP (Lomax et al.) to generate input data such as the amplitude, velocity, and displacement of three components from 2 seconds ahead and 2 seconds after the P-wave arrival following one-second time steps. This model extracts log-mel power spectrum features from this input data, then classifies P-waves and others using these features. The dataset consisted of 3,189,583 samples: 81,394 samples from event data (727 events in the Korean Peninsula, 103 teleseismic events, and 1,734 events in Taiwan) and 3,108,189 samples from continuous data (recorded by seismic stations in South Korea for 27 months from 2018 to 2020). This model was trained with 1,826,357 samples through balancing, then tested on continuous data samples of the year 2019, filtering more than 99% of strong false-picks that could trigger false alarms. This model was developed as a module for USGS Earthworm and is written in C language to operate with minimal computing resources.

• Journal of the Earthquake Engineering Society of Korea
• /
• v.28 no.4
• /
• pp.183-191
• /
• 2024

Accurate seismic vulnerability assessment requires high quality and large amounts of ground motion data. Ground motion data generated from time series contains not only the seismic waves but also the background noise. Therefore, it is crucial to determine the high-pass cut-off frequency to reduce the background noise. Traditional methods for determining the high-pass filter frequency are based on human inspection, such as comparing the noise and the signal Fourier Amplitude Spectrum (FAS), f² trend line fitting, and inspection of the displacement curve after filtering. However, these methods are subject to human error and unsuitable for automating the process. This study used a deep learning approach to determine the high-pass filter frequency. We used the Mel-spectrogram for feature extraction and mixup technique to overcome the lack of data. We selected convolutional neural network (CNN) models such as ResNet, DenseNet, and EfficientNet for transfer learning. Additionally, we chose ViT and DeiT for transformer-based models. The results showed that ResNet had the highest performance with R² (the coefficient of determination) at 0.977 and the lowest mean absolute error (MAE) and RMSE (root mean square error) at 0.006 and 0.074, respectively. When applied to a seismic event and compared to the traditional methods, the determination of the high-pass filter frequency through the deep learning method showed a difference of 0.1 Hz, which demonstrates that it can be used as a replacement for traditional methods. We anticipate that this study will pave the way for automating ground motion processing, which could be applied to the system to handle large amounts of data efficiently.

• Journal of the Korean Geotechnical Society
• /
• v.35 no.11
• /
• pp.121-129
• /
• 2019

An Earthquake Early Warning (EEW) system is a technology that alerts people to an incoming earthquake by using P waves that are detected before the arrival of more severe seismic waves. P-wave analysis is therefore an important factor in the production of rapid seismic information as it can be used to quickly estimate the earthquake magnitude and epicenter through the amplitude and predominant period of the observed P-wave. However, when a large-magnitude teleseismic earthquake is observed in a local seismic network, the significantly attenuated P wave phases may be mischaracterized as belonging to a small-magnitude local earthquake in the initial analysis stage. Such a misanalysis may be sent to the public as a false alert, reducing the credibility of the EEW system and potentially causing economic losses for infrastructure and industrial facilities. Therefore, it is necessary to develop methods that reduce misanalysis. In this study, the possibility of seismic misclassifying teleseimic earthquakes as local events was reviewed using the Filter Bank method, which uses the attenuation characteristics of P waves to classify local and outside Korean peninsula (regional and teleseismic) events with filtered waveform depending on frequency and epicenter distance. The data used in our analysis were analyzed for maximum P_v values using 463 events with local magnitudes (2 < M_L ≦ 3), 44 (3 < M_L ≦ 4), 4 (4 < M_L ≦ 5), 3 (M_L > 5), and 89 outside Korean peninsula earthquakes recorded by the KMA seismic network. The results show that local and telesesimic earthquakes can be classified more accurately when combination of filtering bands of No. 3 (6-12 Hz) and No. 6 (0.75-1.5 Hz) is applied.

• Geophysics and Geophysical Exploration
• /
• v.20 no.1
• /
• pp.25-32
• /
• 2017

We present a 3D SH-wave velocity model of the crust and uppermost mantle and seismic radial anisotropy beneath East Asia. The SH-wave velocity structure model was built using Love-wave group-velocity dispersion data from earthquake data recorded at broadband seismic networks of Korea, Japan, and China. Love-wave group-velocity dispersion curves were obtained by using the multiple filtering technique in the period range of 3 to 150 s for 3,369 event-station pairs. The inverted model using these data sets provides a crust and upper mantle SH-wave velocity structure down to 100 km depth. At 10 ~ 40 km depths SH-wave velocity beneath the East Sea is higher than beneath the Japanese island region. We estimated the Moho beneath the East Sea to be between 10 ~ 20 km depth, while Moho beneath the Korean Peninsula at around 35 km based on the depth where high-velocity anomalies are detected. We estimated the lithosphere-asthenosphere boundary beneath the East Sea to be at around 50 km based on the depth where strong low-velocity anomalies are observed. Widespread low-velocity anomalies are found between 50 ~ 100 km depth in the study region. Positive radial anisotropy ($V_{SV}$ > $V _{SH}$) is observed down to 35 km depth, while negative radial anisotropy ($V_{SV}$ > $V _{SH}$) is observed for deeper depth.

• Journal of Environmental Impact Assessment
• /
• v.19 no.5
• /
• pp.475-481
• /
• 2010

Lots of pollutants typically originating from urban transportation are accumulating on the paved surfaces during dry periods and are washed-off directly to the river during a storm. Also, paved surfaces are contributing to increase in peak flows and volume of stormwater flows. These are the main reasons why the water quality of rivers and lakes remain polluted and still below standards. Currently, several management practices are being applied in developed countries but the design standards are still lacking. This research was conducted to develop a treatment technology that can be useful to address the problems concerning runoff quality and quantity. A lab scale infiltration device consisting of a pretreatment tank and media zone was designed and tested for various flow regimes characterizing the low, average and high intensity rainfall. Based on the experiments, the high intensity flow resulted to increase in outflow event mean concentration (EMC) of pollutants, about twice as much as the average outflow EMC. However, 78 to 88% of the total suspended solids were captured and retained in the pretreatment tank because of sedimentation. The removal of heavy metals such as zinc and lead was greatly affected by the vertical placement of woodchip layer prior to the media zone. It was observed that the high carbon content (almost 50%) in the woodchip provided opportunity for enhancing its uptake of metal by adsorption. The findings implied that the reduction of pollutants can be greatly achieved by means of proper pretreatment to allow for settling of particles with a combination of using high carbon source media like woodchip and a geotextile mat to reduce the flow before filtering into the media zone and finally discharging to the drainage system.

• Economic and Environmental Geology
• /
• v.52 no.4
• /
• pp.313-322
• /
• 2019

SAR (Synthetic Aperture Radar) remote sensing data is a very useful tool for near-real-time identification of landslide affected areas that can occur over a large area due to heavy rains or typhoons. This study aims to develop an effective algorithm for automatically delineating landslide areas from the polarimetric SAR data acquired after the landslide event. To detect landslides from SAR observations, reduction of the speckle effects in the estimation of polarimetric SAR parameters and the orthorectification of geometric distortions on sloping terrain are essential processing steps. Based on the experimental analysis, it was found that the IDAN filter can provide a better estimation of the polarimetric parameters. In addition, it was appropriate to apply orthorectification process after estimating polarimetric parameters in the slant range domain. Furthermore, it was found that the polarimetric entropy is the most appropriate parameters among various polarimetric parameters. Based on those analyses, we proposed an automatic landslide detection algorithm using the histogram thresholding of the polarimetric parameters with the aid of terrain slope information. The landslide detection algorithm was applied to the ALOS-2 PALSAR-2 data which observed landslide areas in Japan triggered by Typhoon in September 2011. Experimental results showed that the landslide areas were successfully identified by using the proposed algorithm with a detection rate of about 82% and a false alarm rate of about 3%.