Speaker
Description
HydroDrone is an ESA-funded UAV-based C-band SAR campaign designed to investigate short-term temporal decorrelation over vegetated scenes across timescales ranging from tens of seconds to minutes and several hours. Characterizing the temporal evolution of vegetation coherence at these timescales is important for understanding the dynamics of vegetated scatterers and their impact on interferometric observations. This is particularly relevant for SAR systems characterized by long coherent integration times, for which scene changes occurring during image formation may result in significant coherence loss. Future geosynchronous SAR concepts represent a notable example, as their coherent integration times may be comparable to the short-term temporal decorrelation timescales investigated here, making the characterization of these dynamics critical for reliable image formation.
Previous experimental campaigns have investigated radar backscatter variability and temporal decorrelation over vegetation using ground-based SAR, airborne SAR, and tower-mounted radar systems. However, these experiments have generally been characterized either by limited spatial coverage and imaging capability or by revisit intervals on the order of several minutes, leaving the intermediate temporal regime between tens of seconds and a few minutes poorly characterized through high-resolution SAR observations. HydroDrone was conceived to address this gap by exploiting the flexibility of a UAV platform together with a repeated circular external-looking acquisition geometry. This configuration enables a wide surrounding area to be repeatedly observed while maintaining a very short revisit interval. In the configuration considered here, approximately 1 km$^2$ is imaged with a circular pass duration of about 20 s. Consecutive passes are designed to reproduce the same viewing geometry, thereby minimizing spatial decorrelation and facilitating the isolation of temporal effects.
The circular external-looking geometry requires a dedicated SAR processing formulation. In contrast to conventional linear trajectories, the tangential wavenumber support depends on both the target position and the radius of the circular trajectory. Using the full antenna beamwidth, the reference tangential resolution can be expressed as:
$ρ_{t}=λ/[4 \ \sin(α_{ant}/2)] r_p/r_s $
where $r_p$ is the target radial distance from the trajectory center and $r_s$ is the trajectory radius. SAR images are focused using time-domain back-projection while adapting the processed wavenumber support to enforce a spatially uniform tangential resolution across the scene.
For interferometric applications, residual UAV trajectory inaccuracies represent an additional critical issue, since navigation uncertainties introduce phase errors that may severely degrade the interferometric products. HydroDrone therefore employs a multi-aperture InSAR trajectory refinement strategy based on short overlapping sub-apertures. For each sub-aperture, the residual interferometric phase is modeled in the local tangential-radial-vertical coordinate system as
$\phi_{err}=K_t \ dS_t+K_r \ dS_r+K_z \ dS_z $ .
The residual three-dimensional displacement of the phase center is estimated through an iterative coarse-to-fine search. The corrections retrieved over the individual sub-apertures are subsequently interpolated along the complete trajectory, and the SAR data are refocused using the refined platform trajectory.
The HydroDrone project encompasses eight campaign days distributed throughout the year to capture seasonal variations in vegetation and soil conditions, with two campaigns performed in each season approximately 45~days apart. Each campaign day consists of three UAV-SAR flights carried out in the morning, around mid-day, and in the afternoon. Each flight ideally comprises 30 consecutive circular passes over the same scene, providing dense interferometric sampling from approximately 20 s to about 10 min within a single flight and extending to several hours when acquisitions from different flights are combined. The SAR measurements are complemented by meteorological observations, soil-moisture measurements, ground-level photographs, and high-resolution UAV optical surveys to support the interpretation of the observed temporal decorrelation.
Preliminary results from the first processed campaign day demonstrate the effectiveness of the proposed trajectory refinement, with a clear improvement in interferometric coherence and near-zero residual phase over stable targets after calibration. Coherence matrices computed over different vegetated areas reveal markedly different temporal behaviors. Forest exhibits consistently low coherence over the investigated temporal scales, while the cornfield shows a rapid coherence decay, particularly under mature crop conditions. In contrast, medium-height vegetation and meadows generally exhibit higher temporal stability, although their behavior changes with seasonal vegetation conditions. These results highlight the sensitivity of short-term interferometric coherence to both vegetation type and seasonal state. Thermal-noise and spatial-decorrelation contributions are also assessed and found to be small compared with the measured coherence loss, indicating that the observed behavior is predominantly associated with temporal decorrelation. These results demonstrate the capability of the HydroDrone acquisition concept to densely characterize vegetation coherence evolution from tens of seconds to several hours.