Signal decomposition of hypertemporal sentinel-1 time series to optimize information retrieval for land applications
The overall goal of this project is to develop and test an innovative method to optimize the extraction of spatio-temporal information from Sentinel-1 time series. The proposed approach is based on the temporal decomposition of SAR time series. The extracted spatio-temporal information will be investigated with respect to its suitability as a data basis for specific applications. In parallel, a novel speckle filter can be developed using temporal decomposition. This will also be pursued within this project. The speckle filter will have the property to filter exclusively over time. Thus, the geometric resolution of the input data is completely preserved. An extension of the filter for spatio-temporal filtering is also planned.
The temporal signal will be decomposed into individual components of different frequencies. Components of high frequency reflect random backscatter variations (speckle, impact effects etc.), components of medium or low frequency are dominated by bio-physical processes such as soil moisture changes or plant growth. Which components are related to which biophysical processes, and which temporal sampling density must be given in order to be able to represent the influence of these control variables, is to be analyzed in the context of this project.
According to these results the research of the usability of the temporal components with regard to the application development will be carried out. The main applications are the mapping of land cover and land use including REDD+ questions, the detection of land cover changes, analyses of the dynamics of wetlands and the derivation of soil moisture indicators.
Project runtime: 2019 - 2022
Spokeperson: apl. Prof. Dr. habil. Christian Thiel