Register now After registration you will be able to apply for this opportunity online.
Implement better vegetation model for the GNSS-IR soil moisture retrieval
This master thesis aims to improve the retrieval of soil moisture using the GNSS Interferometric Reflectometry (GNSS-IR) method by the development of a new, machine-learning based model for the correction the vegetation influence on the retrieval.
GNSS interferometric reflectometry (GNSS-IR) is an innovative remote sensing technique enabling users to infer information about soil moisture, snow depth, or vegetation water content. The method uses reflected signals (~20 cm wavelength, L-band) that are transmitted by GNSS satellites. Signal to Noise Ratio (SNR) observations collected by GNSS receivers are sensitive to the interference between the direct signal and the reflected signal (often referred to as “multipath”). The interference pattern changes with the elevation angle of the satellite, the signal wavelength, and the height of the GNSS antenna above the reflecting surface. As discussed in several scientific studies, vegetation surrounding the GNSS site and its reflection zones can have a significant impact on ground reflection properties and thus derived soil moisture products. This leads to systematic errors and biases, which must be accounted for, to guarantee high-quality results. Although efforts to model and remove vegetation effects have been made in the past, problems still exist for varying vegetation and soil types.
GNSS interferometric reflectometry (GNSS-IR) is an innovative remote sensing technique enabling users to infer information about soil moisture, snow depth, or vegetation water content. The method uses reflected signals (~20 cm wavelength, L-band) that are transmitted by GNSS satellites. Signal to Noise Ratio (SNR) observations collected by GNSS receivers are sensitive to the interference between the direct signal and the reflected signal (often referred to as “multipath”). The interference pattern changes with the elevation angle of the satellite, the signal wavelength, and the height of the GNSS antenna above the reflecting surface. As discussed in several scientific studies, vegetation surrounding the GNSS site and its reflection zones can have a significant impact on ground reflection properties and thus derived soil moisture products. This leads to systematic errors and biases, which must be accounted for, to guarantee high-quality results. Although efforts to model and remove vegetation effects have been made in the past, problems still exist for varying vegetation and soil types.
The primary goal of this master thesis project is to improve the vegetation model for GNSS-IR. The thesis should also explore the possibilities and limitations of using of machine-learning models for this task. The specific objectives include:
1. Literature review on already existing vegetation modelling efforts and pre-selection of a suitable machine-learning approach
2. Preparation and pre-processing of GNSS SNR data from global and/or regional GNSS networks (e.g. in Switzerland)
3. Implementing and training a specific machine learning model using these datasets
4. Performance assessment for the new approach against soil moisture time series estimated using the current GNSS-IR modelling approach and from in-situ sensors.
The primary goal of this master thesis project is to improve the vegetation model for GNSS-IR. The thesis should also explore the possibilities and limitations of using of machine-learning models for this task. The specific objectives include: 1. Literature review on already existing vegetation modelling efforts and pre-selection of a suitable machine-learning approach 2. Preparation and pre-processing of GNSS SNR data from global and/or regional GNSS networks (e.g. in Switzerland) 3. Implementing and training a specific machine learning model using these datasets 4. Performance assessment for the new approach against soil moisture time series estimated using the current GNSS-IR modelling approach and from in-situ sensors.
Dr. Matthias Aichinger-Rosenberger (maichinger@ethz.ch)
Laura Crocetti (lcrocetti@ethz.ch)
Prof. Dr. Benedikt Soja (soja@ethz.ch)
Dr. Matthias Aichinger-Rosenberger (maichinger@ethz.ch) Laura Crocetti (lcrocetti@ethz.ch) Prof. Dr. Benedikt Soja (soja@ethz.ch)