The relationship among meteorological, agricultural, and in situ news-generated big data on droughts

Abstract

The purpose of this study is to evaluate the effectiveness of agricultural drought risk management using news media data (NMD) by elucidating the relationships among the standardized precipitation index (SPI), the agricultural reservoir storage deficit index (RDI), and the NMD obtained from news sources. For a severe drought that occurred in South Korea from 2014 to 2016, the SPI and RDI were calculated, and the NMD were collected. In drought-affected areas, the receiver operating characteristic (ROC) was used to assess the performance of NMD and to replicate the temporal drought trends using the SPI and RDI. The ROC analysis of NMD and drought indices showed a hit rate above 0.65, and the hit rate showed the highest value (0.75) in SPI-12. The central region of South Korea showed the highest number of news postings during the 12 months in which SPI-12 remained in the severe drought category. For the southern region of South Korea, large amounts of NMD were collected when the RDI had the lowest value. The amount of NMD was sensitive to spring drought from March, the delayed Jangma in late June, the dry Jangma in July, and the absence of a typhoon in September of 2015. The study results showed that NMD are closely related to both meteorological drought and agricultural drought conditions. As NMD represent the in situ drought experienced by the people, these data can provide a useful drought indicator along with government-published drought information.