Titolo della tesi: Time series analysis of synthetic aperture radar data for surface soil moisture retrieval at high resolution
Earth observation (EO) of surface soil moisture (SSM) is of great importance for improving the understanding of the hydrological cycle and supporting land applications such as numerical weather prediction (NWP), flood forecasting and drought monitoring and prediction.
The thesis investigates the use of dense time series of synthetic aperture radar (SAR) data for the retrieval of SSM at high spatial and temporal resolution. The study encompasses analytic development and validation aspects, supported by experimental activity. Besides, an emphasis is placed on the sensitivity of C- and L- band SAR data to SSM as a function of the incidence angle and the vegetation cover.
Indeed, the recent availability of Sentinel-1 (S-1) data, having 30% of its swath over land imaged at incidence angles higher than 40°, has urged to bring fresh evidence about the capabilities of C-band SAR systems to retrieve SSM of agricultural surfaces observed over a fairly broad range of incidence angles. This need is even more compelling looking at the future polar SAR systems, such as the next generation of the S-1 system. They will be characterised by a further increase of the swath width, which implies a shallower incidence at far range. Moreover, new concepts of geostationary SAR platforms, such as the one investigated in the ASI SINERGY project (Synthetic aperture Instrument for Novel Earth Remote-sensed MeteoroloGy and HydrologY), feature incidence angles increasing with the latitude of the observed area and, therefore, may reach very high values. Under these circumstances, there is a growing interest to understand the physical mechanisms limiting the SSM retrieval over agricultural surfaces at high incidence angles.
The issue has been further elaborated on with a view to the future ROSE-L mission. Indeed, an extension to the case of L-band SAR systems has been carried out in the context of the ESA SARSense project (Technical Assistance for Airborne Measurements during the SAR Sentinel Experiment). The potential of L- versus C-band SAR data for retrieving SSM underneath crops has been investigated. To this purpose, the data set collected during the SARSense campaign over the Selhausen site (Germany) in 2019 has been exploited.
A second important focus of the thesis has been on developing a mathematical scheme for SSM retrieval that combines the information content of temporal changes of the SAR interferometric phase and intensity. Again, the underlying motivation is the new availability of dense time series of SAR data that makes timely the integration of incoherent and coherent change detection approaches for surface parameter retrieval.
A third pillar of the study concerns the experimental assessment of SAR SSM, which is approaching a mature phase. In the near future, various products will likely be available at a large scale on a continuous temporal basis. That requires the development of a validation protocol tailored to high resolution SSM products. Worldwide, there exist a number of hydrologic networks, continuously recording in situ SSM observations, which are used to test the accuracy of retrieved SSM data. However, an important open issue is the mismatch between the point scale in situ observations and the spatial resolution of the satellite SSM. Such a mismatch generates the spatial representativeness error, which has been deeply investigated in the context of the ESA EXPLOIT-S-1 project (Exploitation of Sentinel-1 for Surface Soil Moisture retrieval at high resolution) and represents a third important contribution to this study.
Structure of the thesis
The thesis consists of 6 chapters. The first chapter is an introduction that briefly describes the important role of SSM for science and applications. It also contains a concise state of the art on the approaches developed for SSM retrieval from SAR data. Finally, the main questions driving the study are listed.
Chapter 2 investigates, through an experimental and numerical study, the S-1 sensitivity to SSM over agricultural fields observed at low (i.e., 33°) and high (i.e., 43°) incidence angles and quantifies the impact of the incidence angle on the SSM retrieval accuracy. The subsequent chapter deals with the analysis of SAR data acquired by spaceborne and airborne systems during the SARSense campaign, in the context of the ESA SARSense project.
In Chapter 4, a combined incoherent and coherent change detection method to retrieve SSM from S-1 data is developed, implemented, and assessed on synthetic and experimental data.
Chapter 5 develops a method for modelling the spatial representativeness error across scales and evaluates its impact at ~1 km resolution.
The final chapter provides a summary of the thesis and presents directions for further research.