Objectives Global navigation satellite system (GNSS) interferometric reflectometry(IR) technique has been proved to be able to monitor sea level. Improving accuracy is the key to GNSS-IR sea level retrieval based on signal-to-noise ratio (SNR) data.

Methods We propose a new quality control method considering the sharpness of spectrum peak on the basis of the common SNR spectrum quality control methods, and those methods jointly control the quality of the initial retrievals. Then, a second-order dynamic sea surface correction model considering inter-frequency bias is established by combining the new quality control method used for weighting with inter-frequency bias correction, achieving multi-frequency and multi-system data fusion and error correction of initial retrievals.

Results The GNSS data collected from SC02 in USA and HKQT in Hong Kong, China was processed in the experiment. The accuracy of initial retrievals is generally improved by more than 1 cm after using the new quality control method. The second-order dynamic sea surface correction model considering inter-frequency bias is applied to initial retrievals and improves accuracy by more than 3 cm. When the observation environment and data quality are favorable, the accuracy of GNSS-IR sea level retrieval reach centimeter level, but the retrievals are poor when the wind speed is more than 20 m/s. Under the circumstance of larger time window, the proposed model has better dynamic sea surface correction effect than the first-order model, and the retrieval accuracy is significantly improved than the model without taking into account the inter-frequency bias.

Conclusions The new quality control method can effectively control the occurrence of gross errors. The second-order dynamic sea surface correction model considering inter-frequency bias has better correction effect than the conventional first-order model and the model without considering the inter-frequency bias.