*3.3. Technical Route*

In order to make the structure of the article clearer, the technical route is shown in Figure 6. In this study, we made full use of the available remote sensing images to examine the Brumadinho tailings dam failure disaster. Firstly, we collected different remote sensing data from different data sources. Secondly, we used the true color remote sensing images from NASA to investigate the hazard chain along the gully where the dam failure occurred. Thirdly, considering that the river width is narrow and the medium resolution remote sensing image cannot meet the needs, we used the multi-temporal high-resolution remote sensing images from Google Earth to interpret the transport process of waste along the Paraopeba River. Fourthly, we used the original Landsat 8 images to carry out the analysis of waste diffusion in the reservoirs. Through the above procedure, the whole disaster process was clearly recovered using RS techniques. Last but not least, we used the global tailings dam failures database to examine tailings-dam-failure trends based on spatiotemporal pattern mining, and found that this area where the Brumadinho tailings dam failure occurred belonged to the Consecutive Hot Spot area with a relatively high risk.



**Figure 6.** Technical route.

#### **4. Results**

*4.1. Hazard Chain Caused by This Event*

Hazard chains are the hot topic in the broader realm of natural hazards [57] (e.g., earthquake-induced chains [58], glacial-outburst-induced chains [59], and volcano-eruptioninduced chains [60]). In this disaster, the dam released a mudflow of tailings after the dam failure (Figure 7b-A). The high-speed mudflow struck the mine's administrative area (Figure 7b-C) [61], destroyed the railway bridge (Figure 3b,c and Figure 7b-D), and continued to move downstream. At about 3:50 pm on 25 January 2019, the mud reached and came into the Paraopeba River (Figure 7b-E). On 27 January 2019, around 5:30 am, sirens

were sounded for the stability of the mine's adjacent Dam VI [16] (Figure 7b-B), where increased water levels were observed.

This hazard chain contained three stages including dam failure, mudflow, and hyperconcentrated flow with tailings waste (Figure 7b). The occurrence of this hazard chain is the result of a combination of many factors. Since the tailings contained a certain amount of water, it created conditions for the mudflow after the tailings dam failure. The tailings waste entered the river following the original branch channel, which in turn enlarged the impact of this disaster with extremely high turbidity and metal concentrations, lower dissolved oxygen, and change of microbial communities which would impact the growth and reproduction of aquatic creatures [62,63].

**Figure 7.** Comparison of pre- and post-disaster Landsat 8 images [64]. (**a**) Pre-disaster remote sensing image (14 January 2019) [27]; (**b**) post-disaster remote sensing image (30 January 2019) [28]. A = location of the destroyed tailings dam and the tailings pond "Barragem I" on 25 January 2019. B = location of tailings pond "Barragem VI" which appeared as an early warning of stability on 27 January 2019. C = location of the destroyed cantina and office buildings. D = location of the destroyed railway bridge (Figure 3b,c). E = location of the entry point of the mudflow into the Paraopeba River.

#### *4.2. Transport Process of Waste in the Rio Paraopeba River*

Through the examination of multi-temporal Google Earth images, the transport process of waste in the Paraopeba River can be observed (Figure 8). Comparing Figure 8A-1 with 8A-2, a large amount of waste entered the river several days after the failure, which might block the river for a certain period. Figure 8A-3 shows that due to the increase of precipitation in the later period (Figure 13), the water level of the river increased and eroded a new channel.

Figure 8B-1 shows the location of the waste as of 2 February 2019. It can be observed that the color of the right river section is vermeil, compared with the left section (red circle). The length of AB reach is 131.48 km with a height difference of about 60 m. It took less than a week for the waste to transport from A to B. After the waste entered the Paraopeba River, the transport speed of waste in the water was affected by many factors, such as concentration and stream gradients [65,66]. In addition, according to remote sensing images of Google Earth, it appears that the barrier of some small river dams in the Paraopeba River might also slow down the movement of the waste. Figure 8B-1 is a true-color image of tailings in rivers, and the change of water color in the circle position can be observed. In order to make the watercolor contrast more obvious, considering that the green and red bands are sensitive to the sediment [67], the ratio between these bands were calculated. As can be seen in Figure 8B-2, there is an obvious change in water color at the red circle.

**Figure 8.** Transport process of waste in the Paraopeba River. A—1, 2, 3 = images of the entry point of the mudflow into the Paraopeba River, the yellow lines are used to mark the river boundary; B—1, 2 = images on 2 February 2019 around point B in the Paraopeba River, the watercolor change can be seen; C—1, 2, 3 = images of the entry point into the Retiro Baixo.

Comparing Figure 8C-1 with 8C-2, the influence of waste on the water body was obvious after it entered the reservoir of hydropower station—Retiro Baixo. A few months later, the watercolor recovered due to the deposition of sediments. Based on the analysis of Figure 8 and multi-temporal Google Earth images, a table of sediment transport time node was generated (Table 4) in order to make the interpretation of sediment transport clear.

**Table 4.** Sediment transport time nodes.


Note: The brown color indicates that the location exhibits sediments on that specific date, while the blue color implies that there was no sediment transported to this location or sediments had settled down.

### *4.3. Diffusion of Waste in the Reservoirs of Two Hydroelectric Plants*
