Sythetic Aperture Radar application on landslides

Introduction

This project was my master thesis. It aimed to map the temporal evolution of the 2019 landslides in West Pokot, Kenya, using Sentinel-1 SAR imagery.

There has been a recurrence of landslide events in West Pokot, Kenya, from 2007 to 2020.

However, the landslide predictions of the events have led to various false alarms leading to mistrust of the local communities in the warnings.

The final goal of the project was to improve the monitoring capacity for early warning and risk reduction.

Amplitude Change Detection

In this study, I used three Ground Range detected SAR amplitude information to detect and map the 2019 landslide event.

The data processing and analysis was carried out in SNAP software and mapped in QGIS software.

A flowchart of the amplitude change detection done on SNAP

The map showed a significant amplitude change, with red and blue close to each other, in the regions around the landslide locations. The red patterns are on the left of the image's blue patterns.

The amplitude increase is noticed at the lower side on the slopes of the same aspect that detected amplitude decrease.

This pattern could be attributed to the movement due to the mudslides slope or the DEM inaccuracies applied during the terrain correction.

Both amplitude increase and reduction are associated with vegetation land cover removal due to landslides.

Amplitude ratio of SAR sensed on 17/11/2019 and 29/11/2019 for landslide detection

Based on an accuracy assessment, the amplitude change detection proved effective in mapping the 2019 landslide event with an accuracy of 63% in reference to the 500m buffer of NASA landslide locations.

Permanent Scatterers Interferomentry

The phase information of SAR was used to measure the surface deformation using the Interferometry Synthetic Aperture (InSAR).

I applied the Permanent Scatterers Interferometry (PSI), an InSAR technique, to monitor the landslides.

In particular, 30 Single Look Complex (SLC) SAR imagery were pre-processed using SNAP software and PSI was carried out using STaMPS tool.

A flowchart of the steps carried out in the Permanent Scatterer Interferometry Technique

The PS points have been reported to have limited spatial density in vegetated areas. However, my results show that the PS points were well distributed on the are of interest (AOI)when visually inspected.

The total PS in the AOI is 44139 PS points, an average of 197 PS per km sq. The PS density is reliable for analysing the deformation’s temporal and spatial pattern.

The velocity of the surface deformation of the whole study area is between 24mm and -24mm per year.

The results shows different velocities at different parts of the AOI. The velocities have a clear pattern in that we see a close range of values in specific regions.

The distribution and velocity of the PS points in the area of interest

The PSI results were further analysed by looking into the landslides driving factors: precipitation, soil and lithological units and land cover.

Soil Correlation

The analysis was performed by overlaying the PS velocity results on subsoil on QGIS.

The spatial extent of the subsoil guided in determining the soil's behaviour during rainfall.

The landslide affected area has dystric, humic and chromic cambisols. There were no distinct PS velocities observed on the different cambisol types.

The subsoil in LFA as Dystric, humic and chromic cambisols

The characteristics of the type of lithological units on the landslide area was determined, and its influence on landslide occurrence was based on the PS velocities.

The analysis was performed by overlaying the PS velocity results on the lithological units on QGIS.

The lithological units in the landslide region are acid metamorphic and gneiss rocks.

There was no distinct PS velocities pattern on the two lithological units. The acid and gneiss are metamorphic rocks, which explain no difference in the PS velocities.

The acid metamorphic rocks and gneiss as the lithological units in the LFA

The PS velocities on different cambisols and the metamorphic lithological unit had no distinct spatial pattern. However, most of the landslides were observed to have occurred on the boundary of different soils and lithological units.

Precipitation

The precipitation was identified as a trigger of the 2019 landslide due to high-intensity daily rainfall of 47mm (above the observed ‘normal’) one day before the landslide event.

The daily rainfall of the 12 days from 11/05/2019 to 11/17/2019 and 11/17/2019 to 11/29/2019 was also observed to vary.

The daily rainfall intensities between 11/05/2019 and 11/17/2019 were below 20mm.

From 11/17/2019 to 11/29/2019, daily rainfall intensity above 20mm was observed 11/22/2019. The rainfall intensity measured on 11/22/2019 was 47mm, with an increase of 38mm from 11/21/2019.

12-days of accumulated rain between 11/05/2019 to 11/17/2019 was 43mm, while between 11/17/2019 and 11/29/2019, there was a sudden increase of 60mm.

The increase in PSI displacement's magnitude and spatial extent between the SAR imagery before (11/17/2019) and after the (11/29/2019) landslide event was attributed to the 60mm increase in the accumulated rainfall over the 12 days.

The bar graphs of daily precipitation and plot of 12-day accumulated rainfall from 11/05/2019 to 11/29/2018

Land cover Correlation

The influence of different land cover on landslides in West Pokot was determined by selecting the PS points on each landcover and looking at the PS velocities on QGIS software.

The PS velocities varied over the different land covers in the AOI. Based on the PS points selected, 35.85%, 26.82%, 20.57%, 12.06% and 4.70% were on trees, crops, grass, shrubs, and built areas, respectively.

60% of the PS points on each landcover showed that the landslides rate was classified as ‘extremely slow’ according to literature.

You can access the full document through: https://drive.google.com/file/d/12A3BewPt-bKGpOV48p9x16rAKBHZqEzt/view?usp=share_link

Page updated

Google Sites

Report abuse