Improving the Accuracy of GNSS-IR-Based Soil Moisture Retrieval by Mitigating Phase Biases From Nonenvironmental Factors

The phase derived from the signal-to-noise ratio (SNR) is a critical parameter for soil moisture (SM) monitoring with the Global Navigation Satellite System Interferometric Reflectometry (GNSS-IR) technique. While most existing studies have focused on the impact of environmental factors on phase estimation, little attention has been paid to non-environmental phase biases. For the first time, this study provides a detailed analysis of two non-environmental phase bias sources: one arising from the nonlinear least squares (NLS) estimation method and the other from satellite revisit trajectory variations. To mitigate the two phase biases, we propose a reparameterized phase estimation (RPE) method and an adjacent-day SNR correlation-based phase bias correction (ASC-PBC) method. Their performance is evaluated using nearly one year of in situ observations from four GNSS stations with varying topography and vegetation cover. Experimental results demonstrate that the RPE method provides more stable and accurate phase estimates than the conventional NLS method, thereby enhancing SM retrieval accuracy. Specifically, the NLS method only yields 12 satellite trajectories with phase–SM correlation coefficients above 0.5, while the RPE method produces 34, a nearly twofold increase. Accordingly, the root mean square error (RMSE) of SM retrieved by each trajectory decreases from 0.068 cm3/cm3 to 0.058 cm3/cm3, with the mean absolute error (MAE) declining from 0.056 cm3/cm3 to 0.046 cm3/cm3. Meanwhile, the ASC-PBC method reduces the RMSE of SM retrieved from the trajectories affected by satellite revisit trajectory variations by 4.3%–10.9% and the MAE by 4.0%–13.2%. Overall, after mitigating the non-environmental phase biases, the average RMSE and MAE of SM retrieved from all trajectories at four stations decrease by 25.6% and 23.5%, respectively. This confirms that the performance of GNSS-IR-based SM retrieval can be significantly improved by mitigating non-environmental phase biases, even in a complex observation environment.