Summary of Chapter 5
In chapter five of The Signal and the Noise, the author Nate Silver explains the difficulties of predicting earthquakes. Due to the lack of comprehensive understanding of earth’s crust, seismologists are unable to make timely predictions on earthquakes. Many people have tried but failed, and their false predictions cause huge consequences. Giampaolo Giuliani’s false alarm prediction caused the public to panic and put the local government under high pressure. Fear of making another false alarm warning, the authority hesitated to issue an earthquake warning when the real earthquake hit, causing huge casualties and severe damage to the city. Other examples of failed predictions mentioned by Silver include Brain Brandy’s prediction on Lima Peru, Parkfield, and Mojave Desert, California. Although the current technologies and understandings are insufficient to predict earthquakes, seismologists can forecast earthquakes in a larger time frame using statistical analysis. In the chapter, Silver also discusses the current model and method used by seismologists to forecast and the downside of this approach.
One of the key challenges with earthquake predictions is the narrow definition of prediction. According to Silver, prediction and forecast have different meanings in seismology. Prediction needs to be accurate to a given time and place. Forecast, on the other hand, can simply be a probability over a long timeframe. Given the high accuracy requirement in seismology, earthquake prediction seems impossible with the current technology,
Another difficulty with earthquake prediction is the lack of observation of the earth’s crust. Similar to the weather forecast, Chaos theory applies to seismic activities as well, meaning one crustal activity might lead to another. Unlike the weather, the earth’s crust is underneath the earth, making it difficult to observe. Due to a lack of understanding, seismologists depend solely on statistical methods to forecast future earthquakes. Since the statistical method only provides the probability of the event, thus, with the current technology, seismologists are not able to predict earthquakes.
Furthermore, there are many downsides to the current statistical model. The earthquake model, Gutenberg-Richter law, uses historical earthquake data to identify earthquake patterns and then calculates the probability of earthquakes. However, some major earthquakes only happen once in a thousand years. With limited data, the pattern generated is not as accurate because the model might treat those rare events as outliers. Oftentimes, those outliers got ignored. The earthquake that struck Tohoku Japan is a classic example of ignoring the outlier as seismologists believed that it was rare. Semistoligists did pick up some patterns of the foreshock. However, the pattern of the foreshock was not as discernable, and thus it was overlooked. As a result, the earthquake caused severe damage. One lesson from the Tohoku earthquake is that every data point should not be overlooked. Sometimes, patterns are not as obvious as the public expects. Even some semiologists tried to find the perfect pattern by manipulating the data because it would give them a good reputation in the academic world. However, the perfect pattern found by overfitting or underfitting the data usually produces less accurate forecasts, which do more harm to the public.
Since earthquake prediction is still at the preliminary stage, one way to reduce the catastrophic damage caused by earthquakes is through government policies. Silver mentions one of the best measures can be to increase the readiness for earthquakes. Some of the sample policies to improve readiness include setting an earthquake-resisting building requirement in areas where earthquakes are likely to happen to reduce the damage to housing; issuing earthquake protocols and instructions to local government; setting a minimum requirement on the emergency supplies needed to be stocked by each city; educating the public on how to evacuate when an earthquake strikes. Another measure the government should take is to carefully evaluate the forecast before issuing an earthquake warning to avoid causing public panic.
The key takeaway from Chapter 5 is when using a statistical model to forecast the future try to avoid finding the perfect pattern by manipulating data. The real world is more complicated than the model and often filled with outliers. Moreover, sometimes patterns are less discernable and to be sure to examine all data points carefully.
