Brief communication : Appropriate messaging is critical for effective 1 earthquake early warning systems 2

Meng Zhang, Xue Qiao, Barnabas C. Seyler, Baofeng, Di, Yuan Wang, Ya Tang 3 1 Department of the Environment, College of Architecture and Environment, Sichuan University, Chengdu 610065, China 4 2 Institute of New Energy and Low-carbon Technology & Healthy Food Evaluation Research Center, Sichuan University, 5 Chengdu, China 6 3 State Key Laboratory of Hydraulics and Mountain River Engineering, Sichuan University, Chengdu 610065, China. 7 4 Institute for Disaster Management and Reconstruction, Sichuan University, Chengdu 610200, China. 8 Correspondence to: Ya Tang (tangya@scu.edu.cn) 9

Hoshiba and Ozaki, 2014; Kamigaichi et al., 2009). For example, 1) it is hard to provide timely warnings in areas closest to 47 epicentres (e.g., the blind zones); 2) when more than two earthquakes occur in close temporal or spatial proximity, the 48 estimation parameters become hard to process and the error substantially increases; 3) The unsaturated magnitude and 49 seismic intensity of large earthquakes (M>8) may be underestimated, such as the Tohoku-Oki Earthquake (Hoshiba and 50 Ozaki, 2014); and 4) The EEWSs may not work properly due to power failures, wiring disconnects, and high background 51 noise caused by large earthquakes and their aftershocks. 52 Recently, more and more scholars have devoted attention to increasing the effectiveness of EEWS by social means (e.g., 53 Santos-Reyes, 2019; Sutton et al., 2020), which can alleviate the limitations that are difficult to solve with technical 54 innovations. For example, the Japan's EEWS has significantly contributed to reducing social vulnerability to earthquakes 55 through nationwide participation. Most of the alerted respondents can understand and act to protect themselves due to their 56 education and training, although the magnitude of the 2011 Tohoku-Oki Earthquake was under-estimated due to technical 57 limitations, resulting in poor-quality alerts (Fujinawa and Noda, 2013;Hoshiba and Ozaki, 2014). In addition, the United 58 States' EEWS (ShakeAlert) enables recipients to immediately participate in the alert process and define the system capability 59 to enhance public participation, which is currently being tested in California, Oregon, and Washington states (Allen and 60 Melgar, 2019). Comparatively, Mexico's EEWS detected and issued warnings for the 2017 Puebla Earthquake; however, the 61 public took a negative attitude towards its performance since they received little information for either the EEWS or the 62 warnings themselves and had not been previously educated how to act during an emergency response (Santos-Reyes, 2019). 63 These events demonstrate that importance of EEWSs, but also show the critical importance of public awareness education 64 and training, to activate the benefits of EEWSs. 65

China's EEWS Development 66
The China's EEWS development is particularly challenging because several regions are prone to earthquakes, including 67 major metropolitan areas. Therefore, following the 2008 Wenchuan Earthquake, China's central government encouraged the 68 establishment of a national EEWS, initially focusing efforts on four seismic regions for pilot testing (Fig. 1a). With support 69 from the "National System for Fast Seismic Intensity Reporting and Earthquake Early Warning Project" led by the China 70 Earthquake Administration (CEA), a high-quality national seismological network was installed with 15,000 seismic 71 monitoring stations. The instruments aimed at quickly reporting earthquake intensities and earthquake early warnings in key 72 areas on the minute and second scales, respectively. EEWSs in the pilot regions (e.g., Fujian and Sichuan provinces,

Sichuan case 84
Sichuan is a major earthquake-prone region with 73 earthquakes having magnitudes above Ms 5.0 occurring since the 2008 85 Wenchuan Earthquake (Fig. 1b), based on China Earthquake Networks Center (CENC, http://www.ceic.ac.cn/history). Municipality, triggering an alert in some cities across the province, including Yibin (52 km from epicenter), Leshan (168 90 km), and Chengdu (245 km) (Fig. 1c). The alerts were issued approximately 10 s, 43 s, and 61 s prior to major shaking in the 91 https://doi.org/10.5194/nhess-2021-42 Preprint. Discussion started: 19 February 2021 c Author(s) 2021. CC BY 4.0 License. above cities, respectively. It was the first time that an alert system was triggered in Sichuan, which generated great public 92 interest and confusion. 93 In Chengdu, the provincial capital city, the alert was delivered in several ways, including broadcast sirens, as well as text 94 messages on televisions and cell phones that had special applications installed. Of these, the broadcast siren notified the most 95 people with speakers located in more than 110 residential areas. The alert began with a countdown, followed by loud alarm 96 sirens. However, few people understood what the siren pertained to or what was about to happen with only a countdown and 97 then siren. Only when the shaking began, did most people realize the alarm was intended to warn of an impending 98 earthquake. Most people reported that when the countdown over broadcast speakers began followed by the siren, they were 99 confused and unsure what to do. They did not know what was happening or what would happen, because the countdown and 100 siren were unaccompanied by clear audio messages with explanatory information. Many people interpreted the alarm as a 101 firemen's duty task, an air raid alert test, an explosion, theft alarm from a car or electric bicycle, or a special sales event. 102 Clearly, due to the diversity of reactions, the alert caused more confusion, fear, and disturbance than what was intended by 103 EEWS. Some people were less concerned with the earthquake than by the confusion over the loud countdown and siren, as it 104 was nearly midnight. 105 We examined the public perception of Sichuan's EEWS using an internet-based survey, conducted June 21-23 in Chengdu 106 with 770 participants. The demographics of the survey participants can be found in Table 1. The participants can be divided 107 into two groups: 1) those who heard the broadcast siren alert (Group A, n=261) and 2) those who did not (Group B, n=509) 108 (Fig. 2). We found that large majorities of both groups (Group A, 72%; Group B, 61%) did not understand the purpose of the 109 alert. There were only 55 (21%) from Group A who understood the alert and knew what actions to take. Of these, their 110 knowledge came from previous training (26), hearing a brief note at the beginning of the alert (11), advice from people 111 nearby when the alert was ongoing (7)

EEWS Limitations and Implications from Sichuan 124
The Changning Earthquake's example highlights some challenges with Sichuan's EEWS. We are not arguing against issuing 125 earthquake alerts, however, this event and the resulting confusion raises four important issues that should be addressed 126 moving forward: 127 First, a big gap exists between the intention of EEWS and its reality in Sichuan. An effective EEWS should be sufficiently 128 tested and publicized widely (Kamigaichi et al., 2009), so that when an alert is issued people understand its meaning and 129 have enough time to take appropriate actions. When installed in a residential area, inhabitants should be notified about the 130 system, and most importantly, informed about what actions they should take after receiving an alert, but before shaking 131 begins. In the case of the EEWS's alert in Chengdu following the Changning Earthquake, inadequate efforts had been made 132 to adequately inform the public prior to the earthquake, so few people were able to understand or respond appropriately to stable, useful, and suitable ways. As our case study shows, some claimed that the earthquake itself did not scare them as 145 much as the blaring siren did. It seemed unnecessary to use sirens on loudspeakers that day. While the advantage of using 146 sirens is that it rapidly reaches people simultaneously, the use of such "shocking" alarms is needed only with high risks and 147 likelihood of considerable damage. For those that may not lead to causalities or considerable social or economic losses, use 148 of more "gentle" alert channels are recommended. Alerts delivered over the radio, TV, SMS messages, emails, and 149 smartphone applications have shown greater effectiveness in documented cases (Hoshiba and Ozaki, 2014). 150 Third, at what level the alert should be triggered is a key issue. It is essential to avoid the fabled "boy crying wolf" or over-151 alerting, which can lead to public frustration and apathy, so alert messages should not be issued unless the shaking is 152 expected to cause considerable damage. The Changning Earthquake did not cause strong motion or significant damage in 153 Chengdu, but 15% and 24% of the participants from Groups A and B 1 were terrified by the alarm sound, respectively. There 154 were no specific criteria for when to issue EEW alarms at that time. The provincial standard was only issued in April 2019, 155 so it had not yet been formally implemented. According to this standard (draft version) 2 , a warning should only be issued (to 156 the general public) when the seismic intensity is expected to be VI on the Chinese scale. However, despite the higher level in 157 Yibin, the seismic intensity in Chengdu was lower than VI (Fig. 1c), so the alerts should not have been issued in Chengdu. In 158 addition, there continues to be insufficient guidance about how to handle false alarms, updates, and canceled warnings. 159 Fourth, earthquake alerts should be released by an authoritative government agency. The public should be informed that only 160 alerts from the authorized body are reliable. But it was unclear who was the authority that released the alert on June 17, 2019. 161 There can be many third-party warning service providers, who forward EEW messages by multiple transmission routes. Yet, 162 according to Sichuan's draft standard, the publishing body should only be the Provincial Earthquake Warning Release 163 Center. In addition, the Sichuan case shows that one region may have multiple EEWSs (Wang and Lin, 2020), which will 164 raise greater challenges regarding best practices for issuing EEW and popularizing how to interpret them. Therefore, greater 165 supervision and management systems are urgently needed in Sichuan's EEW practice. 166 The most important component of a successful EEWS is a group of users who want alerts and can define the necessary 167 capabilities of the system, and next is the physical infrastructure and sensor system (Allen and Melgar, 2019). The 168 Changning Earthquake warning event showed that the transmission and utilization of EEW lagged behind the technological 169 development and physical construction. The public in affected areas were not well-informed by EEWS alerts, nor were they 170 adequately trained on how to respond. Therefore, we highlight the successful public education and preparedness training 171 from Japan's seismic culture, because the relatively poor understanding of an EEWS by the public can result in confusion. 172 Yet, beyond what actions are necessary to take in response to warnings (Ji et al., 2019; Sutton et al., 2020), the public also 173 needs education regarding the technical limitations and accuracy of EEWSs (Kamigaichi et al., 2009). We also suggest that 174 Chinese scholars should focus more efforts on the public response to and perception of EEWSs to get more insights for 175 issuing alerts, managing emergencies, and making policy. 176 In addition, due to differences in geological setting, socio-economic development status, and population density, losses 177 caused by earthquakes of the same magnitude can vary greatly. Therefore, it is also very important to decide where an 178 EEWS should be set up. Since earthquakes are disasters faced by many countries, collaboration in development and 179 application of EEWSs among countries or regions should be encouraged, so that appropriate efforts are made to reduce loss 180 of life and property when earthquakes occur, despite their inability to reduce losses in epicenter areas. 181

Conclusion 182
The Changning Earthquake warning event demonstrated that EEWSs are not simply technological engineering infrastructure, 183 but they are also social systems for disaster mitigation. There will be no substantive benefit without proper knowledge and 184 appropriate emergency responses by the public, even if the warning is issued accurately and timely, as evidenced by the facts 185 of Mexico and Chengdu, China. Although authoritative government agencies have emphasized that information release 186