Seismology isn't magic, it's geometry
When an earthquake happens, seismometers around the world pick up the energy. The real work is figuring out where that energy came from, how big it was, and what kind of fault moved. I spent years watching this process unfold during the 2011 Tohoku sequence and later working with smaller crustal swarms in California, so I can tell you where the standard methods break down and what actually works when the textbook answer is wrong. There are really two measurement families here, and they measure different things. The moment magnitude scale, or Mw, is what you see in the news. It's calculated from the seismic moment, which depends on the rigidity of the rock, the area of the fault that slipped, and the average amount of slip. The Mercalli intensity scale is completely different - it measures what the ground shaking actually did to buildings and people at a specific location. One earthquake can have one magnitude but a dozen different Mercalli values depending on where you stand. The standard triangulation method uses P-wave and S-wave arrival times at multiple stations. P-waves arrive first, then S-waves. The time gap between them tells you the distance from the station to the epicenter. Three stations give you three circles, and where they intersect is your epicenter. In practice, modern agencies use hundreds or thousands of stations and run inversion algorithms to refine the location, depth, and magnitude simultaneously. The USGS and EMSC do this automatically within minutes of the first wave arriving anywhere.
Where things get complicated
I worked through a sequence in the Salton Sea area where the standard hypocenter calculation kept drifting by two kilometers between iterations. The problem turned out to be a velocity model issue - the software assumed a standard crustal model, but the sedimentary basin beneath that region slows seismic waves significantly. When I ran the same data through a local tomographic velocity model, the locations snapped into place and matched the aftershock pattern we could see from instrumental microseismicity. This is the kind of thing nobody tells you in an intro seismology class. Magnitude saturation is another real problem. Moment magnitude saturates at around Mw 8.3 to 8.5 for certain source parameters, meaning a much larger earthquake can produce nearly the same magnitude reading as a slightly smaller one. During Tohoku, the early automated reports came out around Mw 8.0, then jumped to 8.4, and finally settled at 9.0-9.1 once the full waveform data was processed. If you're relying on early magnitude estimates for any kind of decision-making, you're working with incomplete information by definition.
Practical considerations
Local magnitude, or ML, is what most people think of when they hear "Richter scale," but Richter only works reliably for small to moderate earthquakes in California where his original calibration data came from. It breaks down completely for large events and in regions with different crustal properties. The body wave magnitude (mb) has similar limitations, saturating around Mw 6.5 to 7.0. Moment magnitude is the right tool for everything above roughly Mw 5 in most tectonic settings. Depth matters more than most people realize. A Mw 6.0 at ten kilometers depth will feel dramatically different at the surface than a Mw 6.0 at forty kilometers, even though the magnitude number is identical. Shallow earthquakes concentrate more energy into the near surface and generate higher frequency shaking that damages buildings more effectively. Deep earthquakes spread their energy over a larger volume of rock and much of it dissipates before reaching the surface. I've also seen people confuse fault slip measurements with magnitude entirely too often. A fault that moved two meters during an event doesn't automatically tell you the magnitude without knowing the rupture area. There are empirical relationships between moment and rupture dimensions, but they carry significant scatter. The 2015 Gorkha earthquake in Nepal had an average slip of roughly two meters over a rupture area of about 200 by 80 kilometers, producing an Mw of about 7.8. A smaller fault slipping the same two meters would be a completely different event.
Get the Full Details
For anyone actually working with seismic data rather than just reading about it, the IRIS Earthquake Browser and the USGS Earthquake Hazards Program website are the primary entry points. The data is freely available through their FDSN web services if you need raw waveform data for analysis. Most academic institutions also provide training resources through IRIS or through university geophysics departments, though honestly the steepest part of the learning curve is just getting comfortable with the file formats and coordinate systems, not the actual physics. The biggest blind spot in current measurement practice remains the early warning window. Automated systems can detect an earthquake and estimate its size within seconds, but those estimates are based on the first few seconds of waveforms. For large earthquakes, the early magnitude estimate is almost always too small. We're working on better real-time scaling relationships, but until then, any magnitude you see within the first thirty seconds of a large event should be treated as a lower bound rather than a final value.
