**3. Materials and Methods**

#### *3.1. Materials*

#### 3.1.1. High-Resolution Remote Sensing Images

Freely available remote sensing (RS) imagery can be used to investigate natural hazards such as landslides [9,22], debris flows [8,9] and mountain fires [23]. Google Earth is an important open source of high-resolution remote sensing imagery, and more importantly, it can provide multi-temporal remote sensing datasets [24–26]. In this study, seven highresolution remote sensing images from Google Earth were used to track the movement of sediment in the channel after the tailings entered the Paraopeba River: Three images at the confluence of the debris flow gully and the Paraopeba River (collected at different times), one image at the stage of sediment transport in the river, and three others covering the area where the river enters the Retiro Baixo reservoir (different times).

#### 3.1.2. Medium-Resolution Remote Sensing Images

Landsat satellite images of National Aeronautics and Space Administration (NASA) are important medium-resolution image datasets that can be used to investigate natural hazards. Landsat 8, as an American Earth observation satellite, is the eighth satellite launched on 11 February 2013 in the Landsat program. It has two sensors including the Operational Land Imager (OLI) and the Thermal Infrared Sensor (TIRS). The OLI consists of 8 bands with a spatial resolution of 30 m, and a 15-m panchromatic band. The TIRS can provide 100-m thermal infrared images.

In this study, Landsat 8 images were used to investigate the hazard chain. They were true color synthesized images of the pre- and post-disaster and released by the researchers of NASA without copyright protection [27,28]. Landsat 8 images were also used to examine the sediment concentration of pre- and post-disaster in the reservoirs. Three Landsat 8 images of the pre- and post-disaster (Table 1) were downloaded from the USGS Global Visualization Viewer (GloVis) website [29], and used to examine the diffusion of waste in the reservoirs of two hydroelectric plants.


**Table 1.** Landsat 8 remote sensing images used for the examination of sediment concentration [29].

#### 3.1.3. Global Tailings Dam Failures Database

There are about 3500 active tailings ponds in the world, among which 2000 experience about two to five known "major" failures, and 35 "minor" failures annually [30]. During the period from 2007 to 2017, there were at least 10 very serious mine tailings dam failures involving multiple loss of life, with approximately 20 lives per incident, a release of at least 1 million cubic meters of waste each time, and a travel of 20 km or longer every waste movement [31]. Based on the world mine tailings failure data with more than 300 records during the period from 1915 to 2019 [14], a global spatial geographic database of tailings dam failure was made using ArcGIS. Using the "Natural Break" classification method, the tailings dam failure records could be divided into five categories (Figure 4).

**Figure 4.** Tailings dam failures (1915–2019) with their occurrence times.
