Trapped Lee-Wave Resonance Determines Antarctic Megadune Wavelength

Shannon T. Wong 

Snow megadunes cover 5×105 km2 of the East Antarctic plateau, biasing surface mass balance estimates and overprinting ice-core signals—yet their 2–5 km wavelength has lacked a quantitative selection mechanism for two decades. We first falsify the standard formula λ∗ = 2πU0/N through a calibration-free spatial test: eight REMA 2 m tiles across two independent locations at 81.5–82.0°S show no organised 2–5 km signal while the formula predicts 4–5 km megadunes at every tile (p < 10−8). We then derive the correct mechanism: finite inversion depth Hinv and saltation saturation length Lsat produce a dispersion peak shifting > 60% with Hinv, vanishing below HinvN1/U0 < π/2. Leave-oneout ERA5 cross-validation gives RMSE= 0.17 km with perfect rank ordering across all sectors (Spearman ρ = 1.00, p = 0.042), versus ρ = 0.80 for the single-scale formula, which mis-ranks sectors driven by Hinv variation—a quantity immune to ERA5 calibration multipliers. A fifth independent site extends validation 1500 km beyond the nearest calibration anchor.