Spatiotemporal analysis of change in Landsat time series in the eastern Hindu Kush region

Saeed Akhtar Khan1, Kim André Vanselow2, Oliver Sass1, Cyrus Samimi1
1 Department of Geography, University of Bayreuth, Bayreuth, Germany
2 Institute of Geography, Friedrich-Alexander-University of Erlangen-Nürnberg, Erlangen, Germany

O 2.6 in Session 2: Tracing Changes in Space and Time

29.10.2020, 12:15-12:30, online via Zoom

Introduction

The land cover change in the eastern Hindu Kush region is driven by human activity and impacts of environmental change. Floods and landslides are recurring natural disasters due to which vegetation in the valleys have undergone numerous changes. So far little research has been carried out to map past disasters and their impact on the environment and on human land-use. The aim of this study is to explore the abrupt changes in vegetation and possible drivers such as floods and landslides.

Material and Methods

In this study we used Breaks For Additive Seasonal and Trend (BFAST) for time series analysis of Landsat data. The BFAST tool iteratively decomposes the time series into trend, seasonal and remainder components. Here, we analyse the trend component to discover abrupt changes caused by disasters.

All available Landsat surface reflectance derived data was accessed from USGS for the years 1988 to 2019. Three spectral indices, Normalised Difference Vegetation Index (NDVI), Modified Soil Adjusted Vegetation Index (MSAVI) and Normalised Difference Water Index (NDWI) were used in time series analysis.

Results

We applied BFAST to NDVI, MSAVI and NDWI indices and mapped changes for five selected locations. Produced results are magnitude, timing and number of breakpoints. Detected changes are then interpreted visually and through secondary data. The results show that the timing of the breakpoints detected correspond with known timing of the past disasters in these locations. The impact of disasters is mapped through changes in magnitude of pixels in the study area.

Conclusion

BFAST was successfully applied to five selected locations to detect changes in Landsat time series. In applying BFAST, we were able to detect past disasters (e.g. floods) in the study area. BFAST is well suited to detect disturbances and map their spatiotemporal patterns in large areas. The results of the study will be cross-checked with data from interviews with locals in the area.



Keywords: land cover change; disasters; vegetation; Landsat time series; Hindu Kush

Export as iCal: Export iCal