Validating the EUMETSAT HydroSAF Snow Recognition Product over Mountainuous Areas of Turkey

An algorithm has been running in order to produce real-time snow cover maps from MSG-SEVIRI sensor imagery, covering whole Europe, for more than two years under the framework of EUMETSAT Hydrology-SAF (HydroSAF) Project. Hydrological processes and climate in the mountainous areas are highly affected by the seasonal snow cover. Due to lack of enough field observations because of the inaccessibility of high mountains, it is convenient to monitor the amount of snow with remote sensing satellite data besides setting up and managing ground weather stations. Developed algorithm is based on a multi-spectral thresholding method which uses visible, shortwave-infrared and near-infrared channels of MSG-SEVIRI. For a single day, 32 successive satellite images which have 15 minutes time interval between each of them are interpreted in order to produce a daily snow cover map. The algorithm uses Nowcasting-SAF (SAFNWC) cloud products in classifying the clouds.

In this study 2007-2008 and 2008-2009 snow melting seasons are considered for the validation and evaluation purposes of the HydroSAF snow recognition product. The validation is performed for the mountainous region in the eastern part of Turkey on a daily basis by using the ground observations from 30 climate stations operated by Turkish State Meteorological Service (TSMS). The snow depth was recorded to the nearest 1 cm and reported in integer form. Besides the validation of snow product with ground data, the utility of the snow product in deriving the snow depletion curves (SDC) is evaluated. Other satellite snow products namely, MODIS 8-day snow cover data (MOD10C2) are also used in deriving the snow depletion curves.

Results show high agreement between ground snow measurements and HydroSAF snow recognition product. The overall accuracies for 2008 and 2009 are calculated as 90.96 % and 80.59 % respectively. The commission error for 2008 is 8.12 % whereas for 2009 it is calculated as 17.03 %. The high cloud coverage percentage observed in 2009 caused a higher false alarm rate in the snow classification. Moreover, SDCs derived from HydroSAF and MOD10C2 snow recognition products display similar trends, especially in the melting period of snow. Refinement and enhancement of the HydroSAF snow recognition product with additional validation studies and inclusion of probable HydroSAF weekly snow recognition product are set as future goals during the extension period of HydroSAF Project.