In this work, we present a time-constrained approach for the estimation of soil dielectric permittivity ($\varepsilon$) in agricultural fields in central Argentina. This method combines a two-layer scattering model (based on the Fresnel approximation, which accounts for vegetation attenuation and surface roughness), in the context of a Bayesian inference scheme. To constrain the inference process, we introduce a semi-empirical prior distribution for $\varepsilon$ on day $t+1$, based on its value on day $t$, derived from in situ $\varepsilon$ data. This distribution automatically accounts for the asymmetrical dynamics of soil permittivity. Validation against measured soil $\varepsilon$ in the study areas demonstrates the potential of the proposed method.