Abstract: Oasis agriculture in arid regions persistently faces dual challenges of water scarcity and secondary soil salinization, making accurate acquisition of spatiotemporal soil moisture distribution characteristics critical for sustaining agricultural development. Geophysical detection methods such as ground penetrating radar (GPR) have become viable approaches for mesoscale soil moisture monitoring. However, conventional GPR methods suffer performance degradation in high-conductivity saline soils due to electromagnetic wave attenuation and scattering. Additionally, existing GPR early-time signal studies are predominantly limited to homogeneous or low-loss media, lacking systematic validation in oasis farmland environments characterized by high salinity and spatial heterogeneity, which severely constrains the widespread application of this method in saline soil moisture inversion. This study constructs a multi-scale saline soil model based on elliptical random media, integrating forward simulations with field experiments in Xinjiang oasis cotton fields to quantitatively compare the performance of antenna frequencies (250 MHz and 1000 MHz) and early-time signal indices including carrier frequency amplitude (CFA) and average envelope amplitude (AEA) in saline soil moisture inversion. The results demonstrate that: (1) Both early-time signal methods maintain stable inversion accuracy (R²>0.60) in saline soils with conductivities up to 0.36 S/m, confirming the feasibility of GPR early-time signal methods for quantitative soil moisture inversion in highly saline environments; (2) The accuracy of saline soil moisture inversion based on early-time signals is highly sensitive to antenna frequency and medium heterogeneity, with the 250 MHz low-frequency antenna exhibiting superior depth adaptability and anti-interference capability for monitoring the 0-30 cm plough layer, while the 1000 MHz antenna shows heightened sensitivity to scattering effects from small-scale surface heterogeneities; (3) The CFA index demonstrates higher sensitivity to dielectric constant variations under low soil moisture conditions (<15%), whereas the AEA index exhibits enhanced stability in high moisture environments (>20%). This study proposes an optimized strategy for GPR early-time signal saline soil moisture inversion indices tailored to different degrees of soil heterogeneity and moisture conditions, and validates through field experiments the practical applicability of 250 MHz antennas combined with CFA and AEA indices in oasis farmland environments, providing innovative technical support for precision water-salt management in arid oasis agriculture.