T. Akutsu, M. Ando, M. Aoumi, A. Araya, Y. Aso, L. Baiotti, R. Bajpai, K. Cannon, A. H. -Y. Chen, D. Chen, H. Chen, A. Chiba, C. Chou, M. Eisenmann, K. Endo, T. Fujimori, S. Garg, D. Haba, S. Haino, R. Harada, H. Hayakawa, K. Hayama, S. Fujii, Y. Himemoto, N. Hirata, C. Hirose, H. -F. Hsieh, H. -Y. Hsieh, C. Hsiung, S. -H. Hsu, K. Ide, R. Iden, S. Ikeda, H. Imafuku, R. Ishikawa, Y. Itoh, M. Iwaya, H-B. Jin, K. Jung, T. Kajita, I. Kaku, M. Kamiizumi, N. Kanda, H. Kato, T. Kato, R. Kawamoto, S. Kim, K. Kobayashi, K. Kohri, K. Kokeyama, K. Komori, A. K. H. Kong, T. Koyama, J. Kume, S. Kuroyanagi, S. Kuwahara, K. Kwak, S. Kwon, H. W. Lee, R. Lee, S. Lee, K. L. Li, L. C. -C. Lin, E. T. Lin, Y. -C. Lin, G. C. Liu, K. Maeda, M. Meyer-Conde, Y. Michimura, K. Mitsuhashi, O. Miyakawa, S. Miyoki, S. Morisaki, Y. Moriwaki, M. Murakoshi, K. Nakagaki, K. Nakamura, H. Nakano, T. Narikawa, L. Naticchioni, L. Nguyen Quynh, Y. Nishino, A. Nishizawa, K. Obayashi, M. Ohashi, M. Onishi, K. Oohara, S. Oshino, R. Ozaki, M. A. Page, K. -C. Pan, B. -J. Park, J. Park, F. E. Pena Arellano, N. Ruhama, S. Saha, K. Sakai, Y. Sakai, R. Sato, S. Sato, Y. Sato, Y. Sato, T. Sawada, Y. Sekiguchi, N. Sembo, L. Shao, Z. -H. Shi, R. Shimomura, H. Shinkai, S. Singh, K. Somiya, I. Song, H. Sotani, Y. Sudo, K. Suzuki, M. Suzuki, H. Tagoshi, K. Takada, H. Takahashi, R. Takahashi, A. Takamori, S. Takano, H. Takeda, K. Takeshita, M. Tamaki, K. Tanaka, S. J. Tanaka, A. Taruya, T. Tomaru, T. Tomura, S. Tsuchida, N. Uchikata, T. Uchiyama, T. Uehara, K. Ueno, T. Ushiba, H. Wang, T. Washimi, C. Wu, H. Wu, K. Yamamoto, T. Yamamoto, T. S. Yamamoto, R. Yamazaki, Y. Yang, S. -W. Yeh, J. Yokoyama, T. Yokozawa, H. Yuzurihara, Z. -C. Zhao, Z. -H. Zhu, Y. -M Kim

Transient noise ("glitches") in gravitational wave detectors can mimic or obscure true signals, significantly reducing detection sensitivity. Identifying and excluding glitch-contaminated data segments is therefore crucial for enhancing the performance of gravitational-wave searches. We perform a noise analysis of the KAGRA data obtained during the O3GK observation. Our analysis is performed with hierarchical veto (Hveto) which identifies noises based on the statistical time correlation between the main channel and the auxiliary channels. A total of 2,531 noises were vetoed by 28 auxiliary channels with the configuration (i.e., signal-to-noise threshold set to 8) that we chose for Hveto. We identify vetoed events as glitches on the spectrogram via visual examination after plotting them with Q-transformation. By referring to the Gravity Spy project, we categorize 2,354 glitches into six types: blip, helix, scratchy, and scattered light, which correspond to those listed in Gravity Spy, and dot and line, which are not found in the Gravity Spy classification and are thus named based on their spectrogram morphology in KAGRA data. The remaining 177 glitches are determined not to belong to any of these six types. We show how the KAGRA glitch types are related to each subsystem of KAGRA. To investigate the possible correlation between the main channel and the round winner - an auxiliary channel statistically associated with the main channel for vetoing purposes - we visually examine the similarity or difference in the glitch pattern on the spectrogram. We compare the qualitative correlation found through visual examination with coherence, which is known to provide quantitative measurement for the correlation between the main channel and each auxiliary channel. Our comprehensive noise analysis will help improve the data quality of KAGRA by applying it to future KAGRA observation data.