Monitoring the Evolution of the Kaiwhata Landslide in New Zealand using Object-based Image Analysis and Sentinel-2 Time Series

Kiarash POOLADSAZ

Master's thesis

Abstract:

Landslides are geological events that occur frequently in mountainous and hilly areas of New Zealand, causing significant damage and changes to the landscape. These events can be disastrous, and monitoring their evolution and subsequent impacts is essential to mitigate hazards that could arise in later reactivation phases or similar cases. The abundance of time-series remote sensing data has facilitated the mapping and monitoring of landslides. Applying object-based image analysis (OBIA) allows for a more detailed understanding of the complex natural phenomena and help to integrate several type of features that are essential in the analysis and as the result, it facilitates to semi-automatically map the evolution of natural phenomena such as landslides. In this particular case, the focus is on the Kaiwhata landslide, which is located in the Wairarapa region of New Zealand. The data from Sentinel-2 and PlanetScope have been used from 2017 to 2021 to semi-automatically map the Kaiwhata landslide and monitor its evolution and impacts on the upstream area. To achieve this, the author has defined an OBIA workflow, consisting of rulesets that applied to the time-series data to extract the landslides and landslide-dammed lakes. The workflow has been defined using different multiresolution segmentation (MRS) values, classification parameters, and class refinements. The desired classes have been visualized not only in static two-dimensional (2D) maps but also in interactive three-dimensional (3D) models to influence the augmented visual impression and semantic augmentation of the time-series landslide and landslide-dammed lake results. The results of the OBIA mapping reveal the extent and the gradual increase in the area of the landslides, with two major changes occurring in June 2019 and November 2020. These changes were followed by the formation of temporary landslide-dammed upstream lakes along the Kaiwhata River, due to the intense rainfall. Overall, OBIA allows for a more detailed understanding of the landslides and their evolution and help to identify patterns and features that are not easily detectable by pixel-based approaches. The use of OBIA and time-series remote sensing data can provide valuable insights into the evolution of landslides and their impacts on the landscape.

Abstract:

Landslides are geological events that occur frequently in mountainous and hilly areas of New Zealand, causing significant damage and changes to the landscape. These events can be disastrous, and monitoring their evolution and subsequent impacts is essential to mitigate hazards that could arise in later reactivation phases or similar cases. The abundance of time-series remote sensing data has facilitated the mapping and monitoring of landslides. Applying object-based image analysis (OBIA) allows for a more detailed understanding of the complex natural phenomena and help to integrate several type of features that are essential in the analysis and as the result, it facilitates to semi-automatically map the evolution of natural phenomena such as landslides. In this particular case, the focus is on the Kaiwhata landslide, which is located in the Wairarapa region of New Zealand. The data from Sentinel-2 and PlanetScope have been used from 2017 to 2021 to semi-automatically map the Kaiwhata landslide and monitor its evolution and impacts on the upstream area. To achieve this, the author has defined an OBIA workflow, consisting of rulesets that applied to the time-series data to extract the landslides and landslide-dammed lakes. The workflow has been defined using different multiresolution segmentation (MRS) values, classification parameters, and class refinements. The desired classes have been visualized not only in static two-dimensional (2D) maps but also in interactive three-dimensional (3D) models to influence the augmented visual impression and semantic augmentation of the time-series landslide and landslide-dammed lake results. The results of the OBIA mapping reveal the extent and the gradual increase in the area of the landslides, with two major changes occurring in June 2019 and November 2020. These changes were followed by the formation of temporary landslide-dammed upstream lakes along the Kaiwhata River, due to the intense rainfall. Overall, OBIA allows for a more detailed understanding of the landslides and their evolution and help to identify patterns and features that are not easily detectable by pixel-based approaches. The use of OBIA and time-series remote sensing data can provide valuable insights into the evolution of landslides and their impacts on the landscape.

Keywords

Landslide, New Zealand, Earth Observation (EO), Object-based Image-Analysis (OBIA), Time Series Analysis, 3D Visualization

Language used: English

Date on which the thesis was submitted / produced: 24. 5. 2023

Thesis defence

• Supervisor: RNDr. Jan Brus, Ph.D.

Palacký University Olomouc

Faculty of Science

Master programme / field:
Geoinformatics and Cartography / Geoinformatics and Cartography