| Snow is an important water resource and plays a critical role in the hydrologic cycle.Accurately determining the large-scale snow cover information on a global scale is of great importance for the study of meteorology,hydrology and global change.The traditional ground-based snow monitoring techniques are limited by spatial and temporal sensitivity,while the snow depth measurements estimated by spaceborne microwave sensors are imprecise and costly.With the unceasing development of GNSS,several investigations have demonstrated that the reflected signals can be used as Earth's surface environments detection.As a new emerging discipline,the GNSS reflectometry(GNSS-R)shows great potential in snow depth monitoring and snow water equivalent estimation by its superiority of low-cost and high-resolution.However,the effect of terrain variation has been ignored by most existing GNSS-R model of snow depth estimation.In addition,the lack of effective quality control will introduce large errors and uncertainties in the results.Moreover,current research mainly focuses on the estimations of snow depth rather than providing spatial information on snow density and SWE(Snow Water Equivalent).In an attempt to counter the above problems,the research on the inversion of snow depth and SWE by GNSS-R method was carried out on the basis of the former studies.Some key issues including,the analysis to characteristic of multi-GNSS reflected signals,correction of terrain topography,snow density and snow water equivalent inversion.The main contents and achievement of this paper are shown as follows:(1)In this paper,we discuss the geometric and physical models of GNSS-R as well as the properties of reflected signals.In addition,the method of retrieving snow depth from various observations is shown.(2)We present the principles and models of the traditional snow depth retrieval method based on SNR and phase combinations.Moreover,the performance evaluations of the traditional technique are also shown.Considering that the combination observables of L4 method degraded by inter-frequency ionospheric delays,the triplefrequency method performs significantly better than the L4 method and has similar performance as the SNR method.For L4 method,the STD is 4.5 cm,and for triplefrequency method and SNR method the STD is about 2.5 cm.(3)In this study,we investigate the effects of terrain variation on snow depth estimation and propose an improved method to estimate snow depth.A clustering algorithm and normalization approach are applied to compensate for the effects of terrain variation.In terms of snow depth retrieval accuracy,the proposed method performed better than the traditional method by experiment for various systems and snow depth retrieval models.Compared with the slightly improvement for GPS,the performance of the proposed method is much better than the traditional one for the systems with longer revisit periods,such as BDS and Galileo.The RMSE of the snow depth estimate reduces from 4.3 cm to 2.6 cm for Galilep,and 5.6 cm to 3.5 cm for BDS,respectively.In addition,the corresponding correlation coefficient increases from0.93 to 0.97 for Galileo,0.91 to 0.95 for BDS,respectively.(4)The relationship among snow depth,snow density and snow water equivalent was discussed in detail.We also evaluate the performance of the traditional technique and the proposed one.The experiment results show that the proposed method performed better than the traditional one,in the case of deviation in the snow depth estimation,and the correlation coefficient is about 0.63.Since snow depth have a far greater impact on SWE,the SWE estimation retrieval from different models have the similar results,and all correlation coefficients exceed 0.95. |
PDF Full Text Request