历史地震搜索

Search and filter the historical earthquake database by country, magnitude, year, and impact.

Analysis

探索历史地震记录

历史地震记录可追溯到4,000多年前，最早有记载的地震事件大约在公元前2150年的中国。该数据库来源于NOAA国家环境信息中心（NCEI）重要地震数据库，收录了超过5,700次造成人员伤亡、重大损失或震级达7.5级以上的地震。地震学的仪器观测时代始于19世纪末，因此更早的记录依赖于文字记载、地质证据和考古发现。

研究历史地震揭示了重要规律。全球地震记录表明，相同的断层系统会反复产生大地震，通常复发周期为几十年到几百年。例如，卡斯卡迪亚俯冲带最近一次在1700年发生M9.0级地震破裂，大约每200-500年产生一次此类事件。历史分析还表明，地震死亡人数更多地取决于建筑物的脆弱性和人口密度，而非单纯的震级——2010年海地地震（M7.0）造成超过20万人死亡，而更强的2010年智利地震（M8.8）造成525人死亡，这主要归功于智利更优的建筑规范。

历史数据的关键洞见

有记录以来最致命的地震是1556年中国陕西大地震（估计M8.0-8.3），造成约83万人死亡——大多数遇难者居住在黄土崖壁上开凿的窑洞中，这些窑洞在地震中坍塌。
海底地震引发的海啸在历史上造成了沿海大地震的大部分伤亡，如2004年印度洋地震（M9.1，约23万人死亡）。
地震目录随时间推移越来越完整：1900年以前的记录主要是非常大或非常致命的事件，而仪器观测时代逐步记录了越来越小的地震。
地震空白区——长期未发生破裂的断层段——是值得高度关注的区域，因为累积的应变终将释放。

常见用途

出于学术、新闻或政策目的研究特定国家或地区的地震历史。
识别地震复发规律以了解长期地震风险。
查找引发海啸的历史地震用于沿海灾害评估。
比较各次地震的伤亡数据，研究建筑规范和防灾准备如何影响结果。

How to Use

1
Set Your Search Filters
Filter the USGS and ISC historical catalog by region (country, bounding box, or fault system), magnitude range (Mw ≥ 4.0 for complete global coverage post-1976), date range, and depth. The USGS catalog extends to 1900 for events Mw ≥ 6.0.
2
Sort and Explore Results
Sort results by magnitude, date, depth, or fatalities. Click any event to access the full USGS event page with focal mechanism, ShakeMap, PAGER loss estimates, and Did You Feel It reports if available.
3
Export for Analysis
Export filtered results as CSV or GeoJSON for use in mapping tools or statistical analysis. Note that catalog completeness varies by region and time period—pre-1960 catalogs are substantially incomplete below M6.5 in many regions.

About

Historical earthquake catalogs are the foundational datasets of earthquake science, encoding the accumulated seismic history of a region across centuries of observations. The USGS Earthquake Hazards Program catalog, the ISC Bulletin, and regional catalogs such as the Japan Meteorological Agency (JMA) catalog are continuously updated as new events occur and historical events are re-analyzed with modern methods. Catalog completeness—the minimum magnitude above which all earthquakes in a region are reliably recorded—is the key limitation for time-series analysis; completeness typically increases (improves) over time with denser networks and is spatially variable, being highest in Japan, California, and parts of Europe.

The shift from analog to digital seismograph networks beginning in the 1970s–80s dramatically improved both completeness and parameter accuracy. Modern moment tensor catalogs (Harvard CMT, GCMT, USGS W-phase CMT) provide standardized source mechanism solutions for all M ≥ 5.5 events since 1976, enabling global analyses of focal mechanism populations, stress field orientations, and tectonic regime classification. The IRIS SYNGINE and IRIS DMC provide open access to seismic waveform archives, enabling retrospective analysis of historical events with contemporary methods.

Open data policies have transformed earthquake science's ability to mine historical records. The USGS ComCat (Comprehensive Earthquake Catalog) provides API access to over 3 million events with searchable parameters; the ISC catalog contains 10+ million events since 1900. Machine learning approaches applied to these large catalogs are identifying previously undetected seismicity patterns, precisely relocating historical events using modern velocity models, and extracting fault geometry from waveform similarity clustering. The emerging field of earthquake forensics combines these catalog analyses with paleoseismic field observations, InSAR surface deformation records, and geodetic strain rate models to produce increasingly comprehensive seismic source models for hazard assessment.

FAQ

How far back does the historical earthquake record go?

The instrumental seismic record begins in 1900, following the development of the Wiechert and Milne seismographs. However, historical documentary records of earthquakes in densely populated regions extend much further: Chinese records document earthquakes to approximately 700 BCE; Middle Eastern records to around 2000 BCE; and European records to classical antiquity. Japan's catalog contains documented events back to the 7th century CE. These historical records allow paleoseismologists to estimate recurrence intervals for major fault systems, though magnitude estimates from pre-instrumental records carry large uncertainties (±0.5 magnitude units). The International Seismological Centre (ISC) bulletin, the most comprehensive global catalog, provides systematic coverage from 1900 with completeness above M7.0 and from approximately 1960 with completeness above M5.0 globally.

Which country has the most earthquakes?

Japan experiences the world's highest earthquake frequency among populated nations, recording roughly 1,500 earthquakes per year detectable by standard seismographs (and far more by Japan's ultra-dense Hi-net network). Indonesia, China, Iran, Turkey, and the United States (particularly Alaska and California) also rank among the most seismically active countries. By total seismic energy release, the Pacific subduction zones dominate: Chile, Alaska, Japan, and Indonesia collectively account for the majority of global seismic moment release. By impact on population, Turkey, Iran, China, and Italy have historically experienced the most damaging earthquakes relative to their land area and population exposure, due to a combination of high hazard, dense population in hazardous areas, and historically vulnerable building stock.

What was the most damaging earthquake in history?

The 1556 Shaanxi earthquake in China is estimated to have killed approximately 830,000 people—the highest death toll in the historical earthquake record. The catastrophic losses resulted from the earthquake striking a densely populated region where much of the population lived in yaodong, cave dwellings carved into soft loess cliffs that collapsed during the event. The 1976 Tangshan earthquake (China) caused an officially reported 242,419 deaths (other estimates suggest up to 650,000). In the instrumental era, the 2010 Haiti earthquake (M7.0) caused approximately 160,000–316,000 deaths, driven by densely built poor-quality construction on soft sediments near the epicenter. Economic losses are dominated by developed-world events: the 1995 Kobe and 2011 Tohoku earthquakes each caused losses exceeding US$100–200 billion.

How do scientists compile historical earthquake catalogs?

Historical earthquake catalog compilation is a multi-source integration process. Instrumental records from seismograph networks are processed by national and international agencies (USGS, GFZ, ISC, EMSC) and assembled into standardized bulletins with hypocentral parameters and magnitudes. For the pre-instrumental period (pre-1900), catalog compilers systematically search documentary sources: monastery chronicles, government administrative records, newspaper archives, scientific expedition reports, and indigenous oral traditions. Magnitude and location are estimated from isoseismal maps drawn from intensity reports, using regression relations that convert felt-area distributions to Mw equivalents. Paleoseismic data—evidence of fault rupture preserved in trenched sediment exposures—extends the record into prehistory but provides only magnitude bounds and recurrence intervals, not precise dates.

What can historical earthquake data tell us about future seismicity?

Historical earthquake data underlies all probabilistic seismic hazard assessment through several pathways. The observed frequency-magnitude distribution on a fault system constrains the Gutenberg-Richter a and b parameters, quantifying the relative frequency of small versus large events. Recurrence intervals for characteristic earthquakes on specific faults—derived from combining instrumental, historical, and paleoseismic observations—determine the long-term rate of large event production. Fault interaction studies use historical sequences to quantify Coulomb stress transfer between faults: a large earthquake alters stress on neighboring faults, sometimes bringing them closer to failure ('stress loading') and sometimes moving them away ('stress shadowing'). However, these inputs produce probabilistic forecasts with return periods measured in decades to centuries—they do not enable short-term deterministic prediction of specific future events.