The Earth‐ionosphere waveguide (EIW) determines the propagation of Very Low Frequency (VLF; ∼3−30 kHz) waves. Characterizing the waveguide is a longstanding challenge due to its large spatial scale and the complex variability of the lower ionosphere. Here we apply a novel linear basis function regression technique to characterize attenuation in the EIW using broadband measurements of lightning‐generated radio waves. The process begins with defining a basis function set, which ideally encompasses a feature set that can predict the variability seen in VLF attenuation properties. With this basis set defined, a system of linear equations is then constructed using sensor pair observations to eliminate the dependency on source amplitude in each observation. Using this formalism, an empirical attenuation model for broadband signals from lightning is constructed and the dependence on attenuation properties with the boundary conditions is explored. The empirically derived results show attenuation rates over ice that are 12 dB/Mm higher compared to paths over saltwater. Well‐known east/west attenuation rate asymmetry stemming from anisotropic reflection coefficients of the ionosphere is also demonstrated and investigated. For example, under a daytime ionosphere, westward‐propagating waves suffer up to 2.8 dB/Mm greater attenuation compared to eastward‐propagating waves. The trained model is used for propagation corrections in a global lightning locating system (LLS), but this technique can be expanded to further study VLF attenuation rates by employing different sets of basis functions.