Earthquakes In New Zealand History

Working With Earthquakes In New Zealand History Data

I've spent more time than I care to admit digging through seismic records for this region. New Zealand sits on one of the most active fault systems in the world, and the historical record is... complicated. The early data is messy because the people recording it were working with wood-pile seismographs and guesswork. Still, if you need to make sense of it, here's how I approach it. The first thing most people get wrong is assuming the magnitudes are directly comparable across centuries. They're not. A 6.5 in 1855 doesn't mean the same thing as a 6.5 in 2016. The instruments changed. The calibration changed. The population density changed, which affects how we estimate historical event intensity from damage reports. GNS Science maintains the primary database, but it's not always straightforward to navigate. I usually pull from their QuakeMap portal first, then cross-reference with the International Seismological Centre for events before New Zealand had consistent instrumental coverage. The ISC fills gaps from the late 1800s onward, but you have to verify coordinates — some early events have location uncertainty measured in tens of kilometers.

I once tried to map seismicity along the Marlborough Fault System for a client who wanted a probability model going back to 1800. The problem was that several events listed in the historical catalog simply didn't exist in modern catalogs. What looked like separate earthquakes turned out to be different interpretations of the same event by different colonial reporters. I ended up spending three weeks reconciling newspaper archives from Papers Past against the official catalog. My workaround was to build a deduplication pass that compared intensity reports against known fault traces, which eliminated roughly 18% of the catalog entries as likely duplicates or mislocated events.

How To Actually Use This Data

If you're building a hazard model, start with the instrumental period — roughly post-1940 for most of NZ — and work backward. The pre-instrumental record relies heavily on felt reports and macroseismic intensity data, which introduces massive uncertainty. A 1931 Napier earthquake of magnitude 7.8 is well-documented. But events in the 1700s? We're working with Maori oral history and very sparse European records, if any. One thing beginners consistently miss is that New Zealand has two fundamentally different seismic regimes. The North Island is dominated by subduction-related and intraplate deformation — think Hikurangi Subduction Zone events and the Taupo Volcanic Zone. The South Island is primarily strike-slip driven by the Alpine Fault and the Marlborough system. Mixing these together in a single model without accounting for the structural difference gives you garbage results. Keep them separate until you understand each regime individually. Another counter-intuitive point: the longest-recurrence faults aren't always the most dangerous for near-term hazard. The Alpine Fault has a recurrence interval around 200 to 300 years for major events, and the last big one was 1717. That makes it overdue by some measures. But the Wairarapa and Kermadec segments have shown in the geological record that they can go longer between great events — sometimes 500 to 1000 years. Yet they also have higher maximum magnitudes when they do break. So "overdue" is a misleading frame. Both are capable of M8+ events right now.

Practical Steps For Working With The Records

Download the EQC and GNS Science catalog in CSV or SHP format. Run a location filter to remove events deeper than 30km if you're focused on crustal seismicity — deep events in the subducting Pacific plate behave differently and skew your statistics. Apply a magnitude-of-completeness check using the Wenner-Müller method or similar, because the catalog isn't complete below a certain magnitude for earlier time periods. I usually end up excluding anything before 1900 below M5.5 because the catalog simply isn't reliable at lower magnitudes during that period. For the pre-1900 data, use the KNDS (Keil and Jones) compiled catalog with caution. It's the most comprehensive attempt at compiling historical events, but it includes events from oceanic areas that may not have affected NZ at all. Double-check everything against contemporary newspaper accounts where possible. Papers Past is genuinely useful here — it's free and underappreciated.

Where This Approach Breaks Down

The biggest limitation is that no amount of data cleaning fixes the fundamental lack of records before European contact. We simply don't know how many significant earthquakes occurred in the centuries before 1840. Paleoseismology helps — trenching across fault traces gives us event timing from sediment disturbance — but it only covers specific locations and has its own error bars, often plus or minus 50 to 100 years on individual events. Another issue is that magnitude scaling relationships for New Zealand's unique tectonic settings aren't perfectly calibrated. The Atkinson and Wald (2007) attenuation relationships are the standard, but they were derived primarily from data in the western US. Using them for the Hikurangi subduction interface introduces systematic error. If you're doing this professionally, budget extra time for sensitivity analyses around ground motion prediction. And honestly, if you're looking for a clean dataset to feed into a machine learning model, you're going to be disappointed. The catalog is inconsistent, magnitudes are recalcitrant, and the metadata quality varies wildly depending on the decade and region. There's no workaround for that except manual curation, and it takes a long time.