Interactive Application
Explore the full analysis dashboard below. Navigate between tabs to view 3D echograms, 2D cross-sections, layer details, echo-free zone analysis, and forward models. The app may take a moment to load on first visit.
Hosted on Hugging Face Spaces. If the embed doesn't load, open it directly.
Overview
The Autonomous Phase-Sensitive Radio Echo Sounder (ApRES) is an FMCW radar system that provides millimeter-precision measurements of internal ice deformation and basal melt rates. Deployed on the Mercer Ice Stream in Antarctica, the instrument transmits daily chirps at 200–400 MHz, yielding a depth profile of internal reflection layers that can be tracked over time to measure vertical ice velocities.
This project focuses on the echo-free zone, the region below the deepest clearly identifiable internal layer where coherent reflections vanish. By applying phase-sensitive analysis, GMM-based noise decomposition, and SVD denoising to the radar data, we explore whether meaningful phase signals persist in this zone and what they reveal about the ice dynamics near the bed.
The interactive dashboard visualizes the full analysis pipeline: from raw echograms through layer detection, phase tracking, velocity estimation, and forward modeling of the expected ice deformation profile.
Key Methods
Two complementary techniques form the core of the velocity analysis: one operating in the phase domain, the other using optimal path finding through the echogram.
Phase-Slope Velocity Estimation
Rather than tracking individual layers, the phase-slope method estimates vertical velocity directly from the local gradient of the phase field across depth and time. By fitting a linear slope to the unwrapped phase over sliding depth windows, it produces a continuous velocity-vs-depth profile without requiring discrete reflector identification.
Crucially, the phase-slope approach confirms the presence of coherent signal even in the echo-free zone, at depths where no clear internal layers are visible in the amplitude echogram. The phase field retains measurable, physically consistent velocity information well below the last identifiable layer.
Viterbi Layer Tracking
The Viterbi algorithm finds the globally optimal path through the echogram by treating layer tracking as a hidden Markov model. The state space is the set of depth bins at each time step, and transition costs penalize large depth jumps while observation costs reward high amplitude and phase coherence.
This produces robust, continuous layer trajectories that can bridge gaps in amplitude, making it well suited for tracking layers that weaken or fragment near the echo-free zone boundary. The tracked positions then feed into phase-difference velocity estimation.
Key Finding · Coherent Signal in the Echo-Free Zone
The phase-slope analysis reveals that the so-called "echo-free zone" is not truly echo-free. Coherent phase signals persist well below the deepest identifiable amplitude layer, producing velocity estimates that smoothly continue the profile measured in the layered ice above. This opens the question of what generates these signals and how they can be used to constrain ice dynamics closer to the bed.
Ongoing Research · Nature of the EFZ Signal
A central open question is the physical origin of the coherent signal in the echo-free zone. Three hypotheses are under investigation:
- Specular layers: intact, continuous dielectric contrasts that are simply too weak to appear above the noise floor in amplitude but remain phase-coherent.
- Rayleigh scatterers: volumetric scattering from distributed inhomogeneities (e.g. crystal fabric variations, dust inclusions) that produce a diffuse but statistically coherent return.
- Broken / disrupted layers: formerly continuous layers that have been folded, sheared, or fragmented by ice dynamics, producing intermittent but still detectable reflections.
To distinguish between these mechanisms, we are applying MCMC (Markov Chain Monte Carlo) methods for probabilistic model fitting and CLEAN deconvolution, an algorithm originally developed for radio astronomy, to decompose the radar return into discrete point-like reflectors and continuous background. The relative strength and spatial distribution of the CLEAN components will help determine which scattering regime dominates at each depth.