Downward continuation of airborne gravity data based on sparse spherical radial basis function

Downward continuation (DWC) of airborne gravity data is essential for high-resolution gravity field recovery but is an ill-posed problem sensitive to noise.We propose a novel approach using spherical radial basis functions (SRBFs) with L1-norm regularization (Lasso) to create sparse DWC models, contrasting with traditional Tikhonov (L2) regularization.Simulation experiments using EGM2008 show that the Lasso method achieves accuracy comparable to the Tikhonov method.Crucially, it simultaneously produces parsimonious models: at a 2′ resolution, sparsity rates of 56-70 % were achieved across different flight heights, drastically reducing the number of active parameters.The Lasso-based SRBF method effectively mitigates model complexity without sacrificing precision.This data-driven strategy aligns with the Occam's razor principle, favoring simpler models with greater generalization potential, and presents a viable alternative for airborne gravity DWC.