Day 4

Detailed paper information

Back to list

Paper title Using Sentinel-1 and MODIS data for mapping and monitoring approaches in the aftermath of the tailings dam failure in Brumadinho, Brazil
Authors
  1. Torben Dedring Ruhr Universität Bochum Speaker
  2. Valerie Graw Ruhr University Bochum
  3. Andreas Rienow Ruhr University Bochum
Form of presentation Poster
Topics
  • D1. Managing Risks
    • D1.01 Satellite EO for Geohazard Risks
Abstract text Tailings are the main waste stream in the mining sector and are commonly stored behind earth embankments termed as Tailings Storage Facilities (TSFs). The failure of a tailings dam can cause ecological damages, economic loss and even casualties. The Tailings Dam Failure (TDF) in Brumadinho (Brazil) is one of the most recent and largest TDFs and caused at least 270 casualties, economic loss, and ecological damages. Earth observation can contribute to disaster risk reduction after TSFs throughout different phases of the disaster management cycle by providing timely and continuous information about the situation on-site.
We exploited and compared different processing techniques for Sentinel-1 data to extract information for rapid mapping activities. Regarding incoherent change detection algorithms, we calculated the log ratio of intensity and the intensity correlation normalised difference, while a normalised coherence difference and a multi-temporal approach were tested as an instance for coherent change detection algorithms. All algorithms were tested regarding their informative value using the Receiver Operating Characteristic curve. The analysis showed that incoherent methods delivered a better basis for rapid mapping activities in this case with an Area Under the Curve of up to 0.849 under a logistic regression classifier. The dense vegetation cover in this region caused low coherence values also in non-affected areas, which made the coherence-based methods less meaningful.
For long term monitoring of the vegetation cover after the TDF, the Standard Vegetation Index (SVI) was calculated in the Google Earth Engine based on 16-day Enhanced Vegetation Index data captured by the MODIS sensor. Even though the SVI is commonly used for drought monitoring, we tested its capabilities for recovery monitoring in Brumadinho. The TDF caused a severe drop in the SVI values, which remained at a low level. The analysis shows that the vegetation cover has not reached the pre-TDF conditions yet.
The presentation focuses on the results coming from the Sentinel-1-based mapping approach as well as the possibilities and limitations of vegetion recovery monitoring with MODIS data, but also briefly discusses the potential of a GIS-based modelling approach to emphasise the ubiquity of geospatial data throughout the disaster management cycle regarding TDFs.