Landslide Studies in Garhwal Himalayas

Abstract

Natural degradation processes such as landslides in hilly terrains are one of the most important landscape-building forces. Landslides emit signals that can be recorded by very sensitive sensors due to their movement. Earthquake-related ground vibrations are recorded using seismological devices. During the landslide's movement, a similar type of ground motion is created. As a result, seismological equipment can be utilised to record signals generated during various landslide events (remotely and in a non-invasive way). With the help of seismic sensors already deployed in the Himalayan region, the past data can be processed using automatic seismological processing chains to detect such landslide events and construction of landslide catalogues using seismology is now possible. By studying the signals recorded by seismological array and studying the ground motion generated by landslides, detection of landslides on seismogram has to be done. The data set for signal analysis and studies which are considered in this thesis are picked from the past events occurred in the region of Garhwal Himalayas. There are basically two events whose data has been collected from the seismological observatory of IIT Roorkee. First one is the earthquake event which took place at a distance of approximately 25 km from Nizmola station of seismological array network on 08 June 2020. Second one is the landslide event which took place in Chamoli region on 07 February 2021. The reason for taking smaller data set is the current prevailing pandemic conditions. Due to corona more data could not be compiled. Two events are picked up and their signals are recorded. After recording those signals, analysis has been done using time frequency method and the differentiation between earthquakes and lanslides has been done. For the analysis of signals, various methodologies have been developed by the researchers which are explained in the further chapter of this thesis work. Those methodologies are stated in brief here:  Feature Extraction- In order to assess the differences and similarities between the different classes, nine signal characteristics were chosen. The nine characteristics were chosen because they match to expert criteria for identifying seismic signals, as well as the fact that they may be employed in automatic classification systems. iv  High Frequency Data Analysis Method- From seismic data recorded continuously on a sparse local seismometer array, this method combines seismic imaging and travel time inversion to determine the locations and times of earthquakes and landslides.  Event Detection Method- This method proves to be very helpful in differentiating between various seismic events and applying STA-LTA ratio algorithm to signal data. With the help of this method, classification of rockfall and earthquake events can be done by visualising their signal shape, waveform and PSD plot.  Time Frequency Analysis Method- The representation of a signal in time and/or frequency domain is insufficient for signal analysis. The Fourier transform, which converts a signal from the time domain to the frequency domain, is a relationship between time and frequency. However, it does not reveal how these frequency contents evolve over time. These limitations greatly affect the proper understanding of the signal. To address this issue, joint time frequency representations were developed, which provide a more accurate representation of both time and frequency and may be applied to non-stationary seismic signals. This is why, in this thesis, the Time Frequency Analysis approach is used to analyse seismic data. This study helps on detecting the significance of landslides in the hilly terrains through experiences from the past incidences of landslides along with their impact. By getting the exact timing of event and monitoring data from sensors installed in the particular area and after studying their features and analysing them on various parameters, methodology for landslide signal analysis can be developed.