Accuracy assessment of SAR data-based snow-covered area estimation method

Employment of satellite radar-based remote sensing data for snow monitoring during the snow melt season has been widely studied by several investigators. Several methods for the estimation of snow-covered area (SCA) fraction have been developed for different types of regions. One common deficiency with the SCA estimation methods has been the lack of statistical accuracy analyses for them. In order to incorporate SCA estimates for operational use, one vital requisite is a thorough statistical analysis of the SCA estimation accuracy. This shortcoming has been addressed for boreal forest region, as an extensive statistical accuracy analysis has been carried out for the Helsinki University of Technology (TKK)-developed SCA method. The TKK SCA method was developed for boreal forest regions, and it is studied here with 24 European Remote Sensing 2 synthetic aperture radar intensity images, on a boreal-forest-dominated test area located in northern Finland. The performance of the SCA method is investigated by using reference data acquired through hydrological modeling. The accuracy analysis is carried out for several statistical variables, and the statistical interpretation is done with respect to several affecting parameters. The accuracy analysis shows a high correlation coefficient between the SCA estimates and the reference data and root mean square error values of 0.213 for open areas and 0.179 for forested areas. In addition, the TKK method employs two reference images for the SCA estimation, and the usability of multiyear reference image utilization was analyzed and proven feasible in this study.