Analysis of Twitter Data Using Evolutionary Clustering during the COVID-19 Pandemic

People started posting textual tweets on Twitter as soon as the novel coronavirus(COVID-19)emerged.Analyzing these tweets can assist institutions in better decision-making and prioritizing their tasks.Therefore,this study aimed to analyze 43 million tweets collected between March 22 and March 30,2020 and describe the trend of public attention given to the topics related to the COVID-19 epidemic using evolutionary clustering analysis.The results indicated that unigram terms were trended more frequently than bigram and trigram terms.A large number of tweets about the COVID-19 were disseminated and received widespread public attention during the epidemic.The high-frequency words such as“death”,“test”,“spread”,and“lockdown”suggest that people fear of being infected,and those who got infection are afraid of death.The results also showed that people agreed to stay at home due to the fear of the spread,and they were calling for social distancing since they become aware of the COVID-19.It can be suggested that social media posts may affect human psychology and behavior.These results may help governments and health organizations to better understand the psychology of the public,and thereby,better communicate with them to prevent and manage the panic.

Extreme Wind Variability and Wind Map Development in Western Java,Indonesia

Wind-related disasters are one of the most frequent disasters in Indonesia.It can cause severe damages of residential construction,especially in the world's most populated island of Java.Understanding the characteristics of extreme winds is crucial for mitigating the disasters and for defining structural design standards.This study investigated the spatiotemporal variations of extreme winds and pioneered a design wind map in Indonesia by focusing on western Java.Based on gust data observed in recent years from 24 stations,the extreme winds exhibit a clear annual cycle where northwestern and southeastern sides of western Java show out-of-phase relationship due to reversal monsoons.Meanwhile,extreme wind occurrences are mostly affected by small-scale weather systems,regardless of seasons and locations.To build the wind map,we used bias-corrected gust from ERA5 and applied the Gumbel method to predict extreme winds with different return periods.The wind map highlights some drawbacks of the current national design standards,which use single wind speed values regardless of location and return period.Beside a fundamental improvement for wind design,this study will benefit disaster risk mapping and other applications that require extreme wind speed distribution.