1. INTRODUCTION
Although science and technology continue to progress, the world is still severely affected by natural disasters. Numerous studies have explored the mechanism of natural disasters (Shuin, Hotta, Suzuki, & Ogawa, 2012; Oku & Nakakita, 2013; Zanandrea, Michel, Kobiyama, & Cardozo, 2019; Nam, Kim, Kang, & Kim, 2019); however, factors affecting disasters are complex. In addition to the environmental condition of the catchment area, anthropogenic activity is a prominent factor. Taiwan, an island surrounded by ocean, is no exception; moreover, Taiwan faces the threat of natural disasters because of its geographical location. The climate is affected by the monsoon throughout the year, mainly the northeast monsoon prevailing from mid-October to April of the following year and the southwest monsoon prevailing from mid-June to September. Specifically, July to September is the typhoon season (Kuo, Lee, & Lu, 2016), and during this period, three to four typhoons striking Taiwan is common. Rainfall during the passage of typhoons is characterized by its long duration and high intensity; such rainfall patterns often cause severe disasters (Yang et al., 2011; Milliman, Lee, Huang, & Kao, 2017).
To mitigate disasters during the passage of typhoons, experiential and physical models have been developed by several studies and can be used to analyze the processes of disaster occurrence (Chen, Jhong, Wu, & Chen, 2013; Yang et al., 2015). Specifically, experiential models can be used to obtain data in a short time; however, this type of model requires a large amount of observational data, and thus is time-consuming. Physical models are usually developed based on a fine theoretical foundation; however, the parameters are usually overly complex. Therefore, both model types have advantages and disadvantages. After the United States (US) government passed the Flood Control Act of 1936, numerous American hydrologists vigorously developed rainfall-runoff models. Finally, in 1956, the US Soil Conservation Service (now Natural Resources Conservation Service) developed the curve number (CN) method, which has been widely applied to simulate or predict runoff from rainfall (Huang, Gallichand, Wang, & Goulet, 2006; Ali, Khan, Aslam, & Khan, 2011; Deshmukh, Chaube, Hailu, Gudeta, & Kassa, 2013; Lal, Mishra, & Pandey, 2015; Ozdemir & Elbaşı, 2015).
Sediment yield (SY) information is essential for environmental and flood control; thus, catchment survey projects usually require it as a reference for catchment planning and design. In the 1960s, a model for assessing SY was developed in response to the need for environmental management (Wischmeier & Smith, 1965). The model effectively simplified the calculation of catchment areas by using natural empirical parameters, and it can be applied to estimate SY during rainstorms or throughout the year. In recent decades, numerous models have been developed to simulate SY, including the universal soil loss equation, revised universal soil loss equation, and modified universal soil loss equation (Chandramohan, Venkatesh, & Balchand, 2015; Furl, Sharif, & Jeong, 2015; Zerihun, Mohammedyasin, Sewnet, Adem, & Lakew, 2018). For the conservation of soil and water resources in catchment areas, runoff and SY are crucial information; thus, studies have used the CN method to estimate the SY in catchment areas (Mishra, Tyagi, Singh, & Singh, 2006; Gajbhiye, Mishra, & Pandey, 2014). In addition to providing runoff information, the CN method can be used to estimate SY.
Although hydrological and SY models have been well established, they require high-quality input data, which are difficult to acquire from long-term monitored catchment areas. Therefore, this study sought to estimate the SY of a catchment area by using big data and the CN method. In addition, the conservation of soil and water resources in a catchment area is essential; thus, this study proposed a hierarchical sediment risk management method based on the concept of subdivisions management and established a management mechanism to serve as a reference for future disaster management.