Abstract's details
Using deep learning with CryoSat radar altimetry to adjust elevations and map surface penetration
CoAuthors
Event: 2022 Ocean Surface Topography Science Team Meeting
Session: Science IV: Altimetry for Cryosphere and Hydrology
Presentation type: Type Poster
Contribution: PDF file
Abstract:
Over the past 30 years, altimetry has revolutionised our ability to monitor surface conditions, quantify changes of the world’s ice masses and its impact on sea level, water availability, and glacial risks. With two high resolution altimeters currently active – the interferometric radar altimeter CryoSat-2 and the laser altimeter IceSat-2 – the present period offers a unique opportunity to co-exploit the observations made by the two sensors and improve the monitoring of ice height and trends.
Recent advances in swath altimetry processing, using the interferometric synthetic aperture radar (SARIn) mode of CryoSat-2, have enabled improved spatial resolution of surface elevation. Meanwhile, IceSat-2 provides enhanced resolution compared to the previous generation thanks to its six laser beams. However, Radar and laser altimeters have different intrinsic properties and behaviours. Joining and interpreting their measurements requires careful consideration of factors such as differences in electromagnetic interaction with the surface, impact of weather, and footprint size.
Here we use a Deep Neural Network to combine elevation measurements acquired by ESA’s CryoSat-2, SARIn waveform parameters, NASA’s Operation Ice Bridge, IceSat-2, and surface conditions over the Greenland Ice Sheet. We explore the difference between radar and laser altimetry and its relationship with surface condition, the impact of penetration of radar waves into snow and firn, and the respective measurement uncertainties.
While neural networks have been increasingly utilised in a wide variety of academic and commercial applications, their use for correcting elevation bias within the cryosphere is novel. The modelled elevation correction will be used to generate time-dependent Digital Elevation Models. Finally, we explore the potential to map ice, snow and firn surface conditions based on the relative differences between laser and radar instruments.
Recent advances in swath altimetry processing, using the interferometric synthetic aperture radar (SARIn) mode of CryoSat-2, have enabled improved spatial resolution of surface elevation. Meanwhile, IceSat-2 provides enhanced resolution compared to the previous generation thanks to its six laser beams. However, Radar and laser altimeters have different intrinsic properties and behaviours. Joining and interpreting their measurements requires careful consideration of factors such as differences in electromagnetic interaction with the surface, impact of weather, and footprint size.
Here we use a Deep Neural Network to combine elevation measurements acquired by ESA’s CryoSat-2, SARIn waveform parameters, NASA’s Operation Ice Bridge, IceSat-2, and surface conditions over the Greenland Ice Sheet. We explore the difference between radar and laser altimetry and its relationship with surface condition, the impact of penetration of radar waves into snow and firn, and the respective measurement uncertainties.
While neural networks have been increasingly utilised in a wide variety of academic and commercial applications, their use for correcting elevation bias within the cryosphere is novel. The modelled elevation correction will be used to generate time-dependent Digital Elevation Models. Finally, we explore the potential to map ice, snow and firn surface conditions based on the relative differences between laser and radar instruments.