A Dynamic Approach to Retrieving Snow Depth Based on Integration of Remote Sensing and Observed Data

Both the observed data and remote sensing data have respective different advantages and disadvantages. Based on integration of observed and remote sensing data, a temporal spatial dynamic approach to retrieve snow depth is explored by skillfully combining observation station data in China and brightness temperature (T_b) from the Special Sensor Microwave Imager (SSM/I). The aim is to utilize the dynamic scheme of the statistical relation to overcome the complexity of the physical relation between T_b and snow depth, accordingly, to improve the retrieval precision in marginal regions of snow cover and the regions where there are few observation stations. The dynamic scheme is implemented by the following steps: For the first time, according to the linear relationship between observed snow depth and T_b difference at each station, the retrieval coefficients of all stations at this time can be achieved, which guarantees the coefficients' spatial difference. Second, after reasonable influencing radius decided, by using of Cressman interpolation algorithm, the retrieval coefficients at all grid points at this time can be obtained, which guarantees the coefficients' spatial continuity. Third, unreasonable stations and grids are eliminated through quality control. Last, for the next time, the previous steps are repeated, and so on, which guarantees temporal dynamics. Its biggest characteristic is that the retrieval coefficients are not fixed, but variable with time and space, which overcomes the errors from regional and temporal (seasonal) differences of the physical features. By comparing it with another retrieval approach, the primary analysis indicates that the error of the snow data through the dynamic approach to retrieving snow depth based on integrated observed and remote sensing data is generally smaller, and the accuracy percentage is higher. Compared to observed data, it has a continuous snow depth distribution that is more reasonable than that of observed field, and in the regions where there are few stations, more appropriate snow depth data could still be obtained. Moreover, compared with the results from direct remote sensing retrieval approach and visible snow cover, the distribution of snow cover obtained by the approach is closer to real field, while the results from static remote sensing retrieval approach and visible snow cover usually underestimate snow cover extent in North China and Central China, and the retrieval result in the western China is also improved using the dynamic approach.