GNSS-IR Snow Depth Retrieval Based on the PSO-NFP Method With Multi-GNSS Constellations

Abstract

The Global Navigation Satellite System interferometric reflectometry (GNSS-IR) method with high spatial and temporal resolution is used to derive snow depth as a complement to existing snow products because of its ease of implementation. GNSS-IR snow depth retrieval accuracy is affected by the land cover and terrain irregularities on the reflecting surface. The number of full waveforms of the signal-to-noise ratio (SNR) is a reliable indicator of reflector height (RH). More importantly, this feature can be extracted in real time. Considering the presence of noise in the received signal, a fitting process is essential. In this article, we propose to use the particle swarm optimization (PSO) algorithm to fit the SNR oscillatory term and extract the number of fit peaks (NFP) to describe the number of full waveforms. Based on signal optimization, retrieval results are enhanced by exploiting the relationship between the good NFP (G-NFP) derived from the historical data and snow depth, and the operation is devoid of a priori constraints. The validation experiment used GNSS data from the P351 station of the EarthScope Plate Boundary Observatory (PBO) network and in situ snow depth measurements from the Snowdrift Telemetry (SNOTEL) network for 2020–2022. Snow depth retrieval results from 2020 to 2021 were used as historical data to derive the G-NFP distribution statistically. Statistically, each NFP corresponds to roughly 25 cm of snow depth change. The G-NFP distribution was then used in the snow depth retrieval process for 2022. The experimental results show that the root-mean-square errors (RMSEs) for global positioning system (GPS)-S1C, GLONASS-S1C, beidou navigation satellite system (BDS)-S2I, and Galileo-S1C based on the PSO-NFP method are 10, 13, 11, and 12 cm, respectively. Compared to the conventional method (CM), the accuracies have improved by approximately 38%, 43%, 35%, and 33%. Moreover, during the snow-free state, the retrieval accuracies based on the PSO-NFP method are improved by approximately 60% compared to the CM. The results show that the proposed method is very suitable for GNSS stations with large snow depth and terrain fluctuations and improves the retrieval results in the snow-free state. Moreover, NFP does not require prior data and can be extracted in real time, indicating its strong generality and potential to serve as a fundamental metric for other snow depth retrieval methods.