A global framework for subsurface soil moisture estimation: Coupling fractal Richards equation with Bayesian optimization

作者：Ziyue Zhu, John Eylander, Venkataraman Lakshmi · 发表于：Remote Sensing of Environment · 年份：2026 · DOI：10.1016/j.rse.2026.115318 · 被引用次数：5 · 研究领域：Soil Moisture and Remote Sensing、Climate change and permafrost、Plant Water Relations and Carbon Dynamics

Global subsurface soil moisture (SM) monitoring remains a critical gap in Earth observation, as current satellite missions such as Soil Moisture Active Passive (SMAP) only reliably measure near-surface SM (∼0–5 cm). To address this limitation, we present a global, satellite-based framework that extends surface SM to 20 cm and 50 cm at 400 m daily temporal repeat (2017–2020). The method integrates the Exponential Filter (ExpF) with a fractal-diffusion representation of vertical transfer (ExpF–FRE) and replaces the empirical ExpF parameter—the characteristic transfer time T, which controls surface–depth coupling—with a physically derived, pixel- and depth-specific timescale T opt . We compute T opt from an effective vertical diffusivity informed by globally available soil hydraulics, the satellite surface SM boundary (SMAP-derived, downscaled to 400 m), and dynamic land-surface temperature (MODIS) and leaf area index (VIIRS). A Bayesian optimization over a small set of global coefficients anchors the physics to observations once, after this procedure, the inference uses only satellite inputs. Comprehensive validation against extensive in situ networks and two existing subsurface SM products (ERA5-Land and GLDAS-Noah) demonstrates robust model performance, with global-scale mean correlation coefficients (R) of approximately 0.804 and 0.623, unbiased root-mean-square errors (ubRMSE) of 0.033 and 0.041 m 3 /m 3 , alongside minor bias −0.005 and − 0.011 m 3 /m 3 at 20 cm and 50 cm,...