整合臺灣高密度觀測網對地震預警系統效能的評估

Abstract

地震預警系統的原理主要是透過資訊傳遞的速度遠快於具破壞性的地震次達波，進而對震度可能較大的地區做出預警。臺灣目前以中央氣象局即時地震觀測網測站利用區域型方法提供臺灣各縣市地震預警資訊；另外臺灣大學則是以微機電地震儀組成的P-Alert地震觀測網提供區域型和現地型的預警資訊，能在同一地區並存運作兩套如此高密度的地震觀測網在世界上著實罕見。密集的地震觀測網能夠盡早檢測到事件的發生，進而縮小未能在次達波前收到預警訊息的區域；而地震預警系統的定位結果又會受到測站分布對事件的包覆程度影響，藉由高密度的地震觀測網有望提供快速且可靠的定位結果，本研究希望利用Earthworm平台整合這兩個高密度地震觀測網，並對2013-2021年133個ML大於5.0的事件進行離線測試，以確定在整合如此高密度的地震預警系統時，其時效性是否仍能提升。研究顯示，與現行系統相比合併的地震觀測網，平均定位誤差自10.06公里增加至10.24公里(增加1.79%)，震源誤差標準差自9.72公里增加至10.62公里(增加9.26%)，規模估計誤差自0.30增加至0.36(增加20.00%)，產生報告的所需時間自14.04秒減少至11.55秒(減少17.73%)，報告時間標準差自4.04秒降至3.89秒(減少3.71%)，雖能提供更快速的地震預警訊息，但會減少震源參數和規模估計的準確度。對於規模估計準確度的問題，在經過測站修正後，能夠降低些許的規模估計誤差，但效果並不顯著，說明在合併高密度地震觀測網時規模估計的準確度仍是挑戰。研究也發現以測站空餘角可作為提早判斷報告是否趨於穩定的指標，並藉此建立一新的初步警報決定流程，能夠讓系統提供更快速且誤差與原方法相似的報告。

The concept of the earthquake early warning system is the information transmitted speed outpaces the S-wave of an earthquake, providing vital warning messages to the near-quake area to prevent potential disasters. Taiwan currently employs earthquake early warning messages for counties and cities through the Central Weather Bureau (CWB) real-time earthquake seismic network using a regional approach. Additionally, the P-Alert seismic network, equipped with Micro Electro Mechanical Systems (MEMS) sensors developed by the National Taiwan University (NTU), provides early warning messages using both regional and on-site methods. Remarkably, Taiwan operates such a high-density seismic network in the same area, enabling early detection of seismic events and reducing the area that might miss early warning messages before the S-wave arrival. Moreover, the location accuracy of the EEW system will be affected by the coverage gap which is related to the distribution of stations. In other words, a high-density seismic network is expected to provide fast and reliable source parameters. This study integrated these two high-density seismic networks via the Earthworm platform. An offline test was implemented by 133 events with a local magnitude (ML) larger than 5.0 from 2013 to September 2022. The primary aim is to assess whether the integration of such a high-density seismic network can significantly improve the timeliness of earthquake early warning messages. The results indicate that the Integrated seismic network results increase in the average location error from 10.06 km to 10.24 km (1.79%) and the magnitude estimation error from 0.30 to 0.36 (20%), respectively. However, the reporting time demonstrates a notable decrease from 14.04 to 11.55 seconds (17.73%). Thus, the Integrated seismic network offers the advantage of faster earthquake early warning messages, although it may lead to some compromise in magnitude estimation accuracy. To address the issue of magnitude estimation accuracy, this study applies the station correction methods. Although this approach shows potential for reducing magnitude errors, it yields non-significant improvements within the Integrated seismic network, suggesting that magnitude estimation remains a noteworthy challenge in this context. Moreover, this study introduces the station coverage gap (GAP) as an indicator of the reliability of EEW reports and revises a novel process to determine preliminary reports, letting the EEW system provide faster warnings while maintaining errors similar to the process before revision. In conclusion, the integration of high-density seismic networks in Taiwan presents promising results in accelerating earthquake early warning messages. Nevertheless, the trade-off between speed and accuracy, particularly in magnitude estimation, warrants further research to enhance the overall performance of the EEW system. Continued investigation into station correction methods is essential to achieve more precise magnitude estimations. Additionally, the application of the GAP parameter as an indicator of EEW report reliability opens new avenues for enhancing the efficiency and dependability of early warning systems.