邓攀, 王泽民, 安家春, 张辛, 于秋则, 孙伟. 利用小波分解的GNSS-R雪厚反演改进算法[J]. 武汉大学学报 ( 信息科学版), 2021, 46(6): 863-870. DOI: 10.13203/j.whugis20190181
引用本文: 邓攀, 王泽民, 安家春, 张辛, 于秋则, 孙伟. 利用小波分解的GNSS-R雪厚反演改进算法[J]. 武汉大学学报 ( 信息科学版), 2021, 46(6): 863-870. DOI: 10.13203/j.whugis20190181
DENG Pan, WANG Zemin, AN Jiachun, ZHANG Xin, YU Qiuze, SUN Wei. An Improved Algorithm Based on Wavelet Decomposition to Retrieve Snow Depth Using GNSS-R Signals[J]. Geomatics and Information Science of Wuhan University, 2021, 46(6): 863-870. DOI: 10.13203/j.whugis20190181
Citation: DENG Pan, WANG Zemin, AN Jiachun, ZHANG Xin, YU Qiuze, SUN Wei. An Improved Algorithm Based on Wavelet Decomposition to Retrieve Snow Depth Using GNSS-R Signals[J]. Geomatics and Information Science of Wuhan University, 2021, 46(6): 863-870. DOI: 10.13203/j.whugis20190181

利用小波分解的GNSS-R雪厚反演改进算法

An Improved Algorithm Based on Wavelet Decomposition to Retrieve Snow Depth Using GNSS-R Signals

  • 摘要: 基于全球导航卫星系统反射测量(global navigation satellite system reflectometry, GNSS-R)数据的雪厚反演具有低成本、低功耗、全天时采集数据的特点, 但利用GNSS信噪比观测值进行雪厚反演时, 观测值受噪声信号功率影响较大, 反演精度较低。基于此, 提出一种基于小波分解的雪厚反演改进算法, 利用小波分解良好的去噪效果, 在不改变原始信号中的频率组成的情况下, 较好地将噪声功率与信号功率分离。通过北极黄河站2017年年积日第32—100天采集的信噪比数据对此算法进行验证, 由于黄河站雪厚变化复杂, 同时对比分析了不同积雪状态下该算法的适用性。结果表明, 所提反演算法与现有的雪厚反演算法相比, 单天时间尺度上的反演结果与实测值的最大偏差由13.71 cm下降到9.43 cm, 反演结果与实测值的中误差由7.08 cm下降到5.98 cm, 反演结果本身的标准差由8.19 cm下降到7.07 cm, 数据利用率由82.60%提升到89.31%。在雪面消融、积累、稳定时, 反演结果与实测值的中误差分别由9.02 cm、10.30 cm、7.59 cm下降到5.82 cm、5.64 cm、7.17 cm, 平均绝对误差分别由6.77 cm、7.52 cm、7.00 cm下降到5.39 cm、4.72 cm、6.73 cm。可见, 在复杂的积雪变化下, 所提改进算法反演结果的精度和可靠性有明显的改善。

     

    Abstract:
      Objectives  Snow depth retrieval using global navigation satellite system (GNSS) reflected signals has the characteristics of low cost, low power consumption and all-day data collection. Because the signal-to-noise ratio (SNR) observations of GNSS are greatly affected by the noise power, we aim to retrieve snow depth in low noise and analyze snow depth variations in different snow surface characteristics.
      Methods  An improved algorithm is proposed to retrieve snow depth by introducing wavelet decomposition before spectral analysis, which can effectively separate the noise power from the signal power without changing the frequency composition of the original signal. This algorithm is verified by the SNR data collected at the Arctic Yellow River Station in the Winter and Spring of 2017. Furthermore, considering the complex variation of snow depth at the Yellow River Station, the applicability of this improved algorithm under different snow surface conditions was analyzed.
      Results  Compared with the original algorithm, the maximum bias of the improved algorithm between the estimation and the measurement decreases from 13.71 cm to 9.43 cm, the root mean squared error (RMSE) decreases from 7.08 cm to 5.98 cm, the standard deviation decreases from 8.19 cm to 7.07 cm, and the data utilization rate increases from 82.60% to 89.31%. In the snow surface conditions of ablation, accumulation and stabilization, the RMSE of improved algorithm decreases from 9.02 cm, 10.30 cm, 7.59 cm to 5.82 cm, 5.64 cm, 7.17 cm, and mean absolute error decreases from 6.77 cm, 7.52 cm, 7.00 cm to 5.39 cm, 4.72 cm, 6.73 cm, respectively.
      Conclusions  It can be seen that the proposed algorithm can provide more accurate and reliable snow depth even if in the conditions of ablation and accumulation.

     

/

返回文章
返回