How do scientists determine earthquake probabilities?
Summary
Scientists determine where faults are and how much they need to move in order to release the strain that accumulates because of plate tectonic motions. Different faults release the energy in different ways, so scientists study past earthquakes on each fault to predict the size of possible earthquakes that could occur in the future. They then use information about how long it's been since the last earthquake to estimate the probability that an earthquake will occur in the next few years.

Read the description below for more detail.

You might ask, "What are the chances that a big earthquake will happen in my neighborhood?" Scientists answer this question using a detailed knowledge of plate tectonics and the earthquake cycle.

Mapping faults
To begin with, they need to know where the faults are in your neighborhood because earthquakes happen when faults move. Geologists map fault locations using a combination of photos taken from satellites and airplanes along with exploring faults on the ground.

Fault Activity
Not all faults produce large earthquakes. As plates move over time, plate boundaries also change. For example, there are faults in the Appalachian mountains that record the location of an ancient plate boundary where two continents collided together millions of years ago. The faults are still there, but they no longer move. Determining which faults are active requires careful investigation.

Scientists can tell if the faults in your neighborhood are active by using seismometers to monitor for smaller earthquakes and by looking for evidence of rupture from recent earthquakes in the layers of rock near the surface.

Strain Buildup: Plate Tectonics
We not only want to know if the fault in your neighborhood is active, but we want to know how much this fault might slip in the future. Plate motions due to plate tectonics cause strain to slowly accumulate along faults near plate boundaries. Earthquakes are nature's way of releasing that strain. Scientists measure the amount of plate motion in order to calculate the amount of energy that needs to be released by earthquakes. Today, they can directly measure the rate of plate motion with very high-tech surveying equipment. But there are other ways to determine long-term plate motion by measuring the distance between features that have been offset by faults. These include offset fences, offset river channels, and even volcanoes and other geologic units cut by the fault. By measuring the amount of offset and the age of the offset feature, we can calculate the rate of plate motions.

Strain Release: Earthquake History
To predict when the next earthquake will occur in your neighborhood, scientists need to know about the earthquakes that relieve strain. We need to know about how often earthquakes occur in your area and when the last big earthquake occurred. If the last earthquake was 200 years ago and we find that earthquakes happen repeatedly every 150-250 years in your neighborhood, then it might be time for another earthquake (a high probability). However, if a large earthquake already happened just 20 years ago, the chances of another earthquake tomorrow are a lot lower (low probability).

So how do we find out about earthquakes that happened hundreds of years before we had seismometers? This is where geologists get creative by asking for help from historians. Early California explorers recorded feeling large earthquakes in their diaries. The priests at California's Missions kept very good records, including every time a large earthquake caused damage to their Mission buildings. By collecting information from as many historical sources as possible, geologists can get a detailed record of where, when, and how big earthquakes were in the past.

In other parts of the world like China, Japan, and the Middle East, written records go back thousands of years. But in California, the oldest written records are from the late 1700's -- that's not long enough for finding out the average time between earthquakes that might be 300-500 years. To find out about even older earthquakes, geologists use information recorded in the layers of rocks. By digging down beneath the surface, we can actually see evidence of ancient earthquakes. That evidence is usually small cracks in the ground where a previous earthquake ruptured the surface. Over time, sediments bury what used to be the surface with a new layer. Scientists can then use radiometric dating (such as Carbon-14 dating) to date each of these layers and then determine when earthquakes happened in the past.

Fault Segments
Some faults are really long, but earthquakes don't always rupture the whole fault at once. For example, the 1868 earthquake on the Hayward fault only ruptured the southern half of the fault. These shorter ruptures release less energy than if the entire fault were to slip at once, so shorter ruptures result in smaller magnitude earthquakes. Scientists are still studying the things that cause fault ruptures to stop, so we don't entirely understand why faults divide into segments. We do know one thing: Looking at earthquakes around the world, we see that ruptures often stop at slight bends where faults are not perfectly straight. Mapping faults is therefore crucial to figuring out if a fault is straight enough to allow really huge earthquakes, or where there are segment boundaries that might make ruptures shorter. The map here colors each segment in the San Francisco Bay Area a different color.

Scenarios
An earthquake "scenario" is one possible earthquake that could occur. Take a look at the Hayward fault, which has two segments: northern and southern. They can either rupture together ("Hayward northern + southern" scenario) or only one segment might rupture during a particular earthquake (for example, the "southern Hayward" scenario). The scenario that ruptures both segments results in a larger magnitude earthquake (because more energy is released by two segments than one alone). Because these scenarios release a different amount of energy, there is a different probability that each scenario will occur. That's why there is a probability for each scenario. Note that a certain fault segment might be included in several different scenarios.

Putting it all together
In the end, scientists take all the information that we know and perform calculations using a computer model to estimate probabilities for each earthquake scenario. The probability is a lot like the chance of rain that a weather report gives, but there is a slight difference. Weather forecasters can look at the history of rainfall for the past 100 years in some places, but they only need to predict one week into the future. Earthquake scientists have only had seismometers for about the same time (100 years), but we are trying to predict earthquakes decades into the future. It's harder for them to recognize patterns leading up to earthquakes because we have recorded so many fewer earthquakes than weather forecasters have recorded rainstorms. The computer model used by earthquake scientists helps them put a range on the probability. In the end, we might tell you that there is a 5-10% chance of a large earthquake happening in your neighborhood during the next 30 years.

Will an earthquake happen today in California? Look at the 1 day earthquake forecast.
For a detailed introduction to earthquake probabilities at a scientific level, visit the (USGS).