한반도 연안 위성 및 수치모델 기반 해수면온도의 정확도에 관한 연구

Abstract

This study investigates the accuracy of daily satellite-composit (OSTIA, AVHRR, G1SST, FNMONC-S) and model-reanalysis (HYCOM, JCOPE2, FNMOC-M) sea surface temperature (SST) data in the coastal regions around the Korean Peninsula (KP) using data observed from buoys and ocean research station over 2011-2013. Analysis results reveals that among the satellite-composit data OSTIA has the lowest bias (0.33oC) and RMSE (0.94oC), respectively and FNMOC-S has the highest corrections (R=0.99), while G1SST, JCOPE2, and AVHRR have relatively larger RMSE (1.40oC, 1.69oC, 1.59oC, respectively). In particular, G1SST, which has the highest spatial resolution (~1km), shows lower accuracy than OSTIA and FNMOC-S with lower resolution. This suggests that the spatial resolution of data is not a critical factor determining the accuracy of SST data in this region. Comparing the results according to regions, the errors were large in the western coastal regions of the KP where depth is shallow and tidal action is strong (for example, Chilbaldo and Dukjukdo buoy), while the errors were low in the East sea and open oceans where depth is deep (for example, Donghae, Ullengdo, Marado). Among the model-reanalysis data, FNMOC-M has the highest accuracy (RMSE=1.06oC, R=0.987) followed by HYCOM (RMSE=1.38oC, R=0.978). JCOPE2 shows the lowest accuracy even though JCOPE2 used the largest data in the western North Pacific for data assimilation (RMSE=1.68oC, R=0.969). It is found that the main sources of large errors in the SST data around the KP are from the rapid SST change during the event of tidal mixing, upwelling, and typhoon-induced cooling. Tidal mixing breaking stratifications in summer is known to produce large surface cooling in the western coast of the KP where tidal current is very strong and depth is shallow. Particularly, major errors of Chilbaldo located in the strong tidal regions occur during the strong tidal periods (high and spring tides). During the passage of two Typhoons, Bolaven and Tembin in 2012, a rapid SST drop of about 8oC has been observed at Mardo buoy. This is well simulated from the most data except JCOPE2. From spatial distribution of G1SST, however, we found abnormal and unrealistic high SST in the regions where the typhoon-induced large cooling occurs, which is thought to be a result of replacing the cooled regions by using climatological SST data in their SST algorism. In Pohang buoy, a significant SST drop (maximum 9oC decrease) was observed in July 2013 due to upwelling. This event is poorly simulated from most data although HYCOM has a similar tendency (but not for magnitude). This means that most SST data sets have problem to simulate upwelling in the East Sea, which requires cautiones to use SST data in this region. In general, daily mean SST data are widely used because of the nature of SST with slow temporal and spatial variability and the lack of high frequency ocean data. However, when (or where) the diurnal variation of SST is dominant, using the daily mean data may have significant limitations, particularly in simulating and explaining weather phenomena varying with short time scale such as torrential rains. The present study investigated the characteristics of diurnal SST variations in the seas around the Korean peninsula using the ocean buoys from the Korea Meteorological Administration (KMA). The diurnal variations were the largest in summer and the smallest in winter. Spatially they are large in the Yellow and South Seas and small in the East Sea. Among all the buoys, Chilbal-do and Geomun-do buoys reveal the largest variation, at which the magnitude reached up to 8℃ in summer, while Donghae, Mara-do, and Ulleung-do buoys show a rather small diurnal variation within 5℃. The magnitudes of diurnal SST variations are mainly related to the variations of solar radiation with high and low peaks in 2-4 PM and 7-9 AM, respectively. In the Yellow Sea and the South Sea, tidal mixing in summer contributed to additional diurnal variations. These overall results suggest that an improved weather prediction in Korea, particularly during summer, requires the consideration of diurnal SST variation.