Reports on Progress in Physics REVIEW Reports on progress in physics the complex dynamics of earthquake fault systems: new approaches to forecasting and nowcasting of earthquakes To cite this article: John B Rundle et al 2021 Rep. Prog. Phys. 84 076801 View the article online for updates and enhancements. This content was downloaded from IP address 169.237.222.219 on 28/05/2021 at 18:47 Reports on Progress in Physics Rep. Prog. Phys. 84 (2021) 076801 (37pp) https://doi.org/10.1088/1361-6633/abf893 Review Reports on progress in physics the complex dynamics of earthquake fault systems: new approaches to forecasting and nowcasting of earthquakes John B Rundle1,2,3,∗,SethStein4, Andrea Donnellan5, Donald L Turcotte2, William Klein6 and Cameron Saylor1 1 Department of Physics and Astronomy, University of California, Davis, CA 95616, United States of America 2 Department of Earth & Planetary Sciences, University of California, Davis, CA 95616, United States of America 3 Santa Fe Institute, 1399 Hyde Park Rd, Santa Fe, NM 87501, United States of America 4 Department of Earth and Planetary Sciences and Institute for Policy Research, Northwestern University, Evanston, IL 60208, United States of America 5 Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, United States of America 6 Department of Physics, Boston University, Boston, MA 02215, United States of America E-mail: [email protected] Received 20 August 2020, revised 31 March 2021 Accepted for publication 15 April 2021 Published 25 May 2021 Abstract Charles Richter’s observation that ‘only fools and charlatans predict earthquakes,’ reflects the fact that despite more than 100 years of effort, seismologists remain unable to do so with reliable and accurate results. Meaningful prediction involves specifying the location, time, and size of an earthquake before it occurs to greater precision than expected purely by chance from the known statistics of earthquakes in an area. In this context, ‘forecasting’ implies a prediction with a specification of a probability of the time, location, and magnitude. Two general approaches have been used. In one, the rate of motion accumulating across faults and the amount of slip in past earthquakes is used to infer where and when future earthquakes will occur and the shaking that would be expected. Because the intervals between earthquakes are highly variable, these long-term forecasts are accurate to no better than a hundred years. They are thus valuable for earthquake hazard mitigation, given the long lives of structures, but have clear limitations. The second approach is to identify potentially observable changes in the Earth that precede earthquakes. Various precursors have been suggested, and may have been real in certain cases, but none have yet proved to be a general feature preceding all earthquakes or to stand out convincingly from the normal variability of the Earth’s behavior. However, new types of data, models, and computational power may provide avenues for progress using machine learning that were not previously available. At present, it is unclear whether deterministic earthquake prediction is possible. The frustrations of this search have led to the observation that (echoing Yogi Berra) ‘it is difficult to predict earthquakes, especially before ∗Author to whom any correspondence should be addressed. Corresponding editor: Professor Michael Bevis. 1361-6633/21/076801+37$33.00 1 © 2021 IOP Publishing Ltd Printed in the UK Rep. Prog. Phys. 84 (2021) 076801 Review they happen.’ However, because success would be of enormous societal benefit, the search for methods of earthquake prediction and forecasting will likely continue. In this review, we note that the focus is on anticipating the earthquake rupture before it occurs, rather than characterizing it rapidly just after it occurs. The latter is the domain of earthquake early warning, which we do not treat in detail here, although we include a short discussion in the machine learning section at the end. Keywords: earthquakes, forecasting, nowcasting, machine learning (Some figures may appear in colour only in the online journal) 1. Introduction competition, among many other topics, below. But first we describe the history of earthquake prediction studies and the In order to provide an impartial test of earthquake predic- search for reliable precursors. tion the United States Geological Survey (USGS) initiated the Parkfield (California) Earthquake Prediction Experiment 2. History of earthquake precursor studies in 1985 (Bakun and Lindh 1985). Earthquakes on this section of the San Andreas fault had occurred in 1847, 1881, 1901, In the 1960s and 1970s, well-funded government earthquake 1922, 1934, and 1966. It was expected that the next earth- prediction programs began in the US, China, Japan, and the ± quake in the sequence would occur in 1988 5 years. An USSR. These programs relied on two approaches. One, based extensive array of instrumentation was deployed. The expected on laboratory experiments showing changes in the physical earthquake finally occurred on September 28, 2004. No precur- properties of rocks prior to fracture, involved searching for pre- sory observations outside the expected background levels were cursors or observable behavior that precedes earthquakes. A observed (Bakun et al 2005). The earthquake had not been second was based on the idea of the seismic cycle, in which predicted. This result has been interpreted to imply the infeasi- strain accumulates over time following a large earthquake. bility of deterministic short-term prediction of earthquakes on Hence areas on major faults that had not had recent earth- a consistent basis. quakes could be considered ‘seismic gaps’ likely to have large Successful near-term predictions of future earthquakes, earthquakes. which have happened on occasion, are very limited. A notable The idea that earthquake prediction was about to become exception was the M = 7.3 Haicheng earthquake in north- reality was promoted heavily in the media. US Geological east China that occurred on February 4, 1975. This predic- Survey director William Pecora announced in 1969 ‘we are tion resulted in the evacuation of the city which saved many predicting another massive earthquake certainly within the lives. It was reported that the prediction was based on fore- next 30 years and most likely in the next decade or so’ shocks, groundwater anomalies and animal behavior. It should on the San Andreas fault. California senator Alan Cranston, be noted, however, that no prediction was made prior to the prediction’s leading political supporter, told reporters that ‘we occurrence of the M = 7.8 Tangshan earthquake in China on have the technology to develop a reliable prediction system July 28, 1976. Reports suggest the death toll in this case was already at hand.’ Although the President’s science advisor as high as 600 000. questioned the need for an expensive program given the low It seems surprising that it is not possible to make reliable death rate from earthquakes, lobbying prevailed and funding short-term predictions of the occurrence of a major earthquake poured into the US program and similar programs in other (Kanamori 2003, Keilis-Borok 2002, Mogi 1985, Scholz 2019, countries. Turcotte 1991). Based on analog laboratory experiments, pre- To date this search has proved generally unsuccessful. As cursory micro cracking expressed as small earthquakes should a result, it is unclear whether earthquake prediction is even occur, and precursory strain would also be expected. Fore- possible. In one hypothesis, all earthquakes start off as tiny shocks occur prior to about 25% of major earthquakes, but it is earthquakes, which happen frequently. However, only a few difficult to distinguish foreshocks from background seismicity cascade via a failure process into large earthquakes. This since they are all ‘just earthquakes’. hypothesis draws on ideas from nonlinear dynamics or chaos An important recent development in this area was the theory, in which small perturbations can grow to have unpre- regional earthquake likelihood models (RELM) test of earth- dictable large consequences. These ideas were posed in terms quake forecasts in California. Forecasts had to be sub- of the possibility that the flap of a butterfly’s wings in Brazil mitted prior to the start of the evaluation period, so this might set off a tornado in Texas, or in general that minus- was a true prospective evaluation. Six participants submit- cule disturbances do not affect the overall frequency of storms ted forecasts for 7700 cells. Two of the forecasts showed but can modify when they occur (Lorenz 1995). In this view, promise, these being the pattern informatics (PI) and epidemic because there is nothing special about those tiny earthquakes type aftershock sequence (ETAS) forecasts. We discuss this that happen to grow into large ones, the interval between large 2 Rep. Prog. Phys. 84 (2021) 076801 Review earthquakes is highly variable, and no observable precursors However, this phenomenon has not been substantiated as a should occur before them. If so, earthquake prediction is either general phenomenon. Similar difficulties beset reports of a impossible or nearly so. decrease in the electrical resistivity of the ground before Support for this view comes from the failure to observe a some earthquakes, consistent with large-scale microcracking. compelling pattern of precursory behavior before earthquakes Changes in the amount and composition of groundwater have (Geller 1997). Various possible precursors have been sug- also been observed. For example, a geyser in Calistoga, Cal- gested—and some may have been real in certain cases—but ifornia, changed its period between eruptions before the Mw none have yet proved to be a general feature preceding all 6.9 1989 Loma Prieta and Mw 5.7 1975 Oroville, California, earthquakes, or to stand out convincingly from the normal earthquakes. range of the Earth’s variable behavior. In many previous cases, Efforts have also been made to identify ground deforma- it was not realized that a successful prediction scheme must tion immediately preceding earthquakes.