Volume 28 Issue 5 - January 30, 2015 PDF
Accurate Forecasting of the Satellite-Derived Seasonal Caspian Sea Level Anomaly Using Polynomial Interpolation and Holt-Winters Exponential Smoothing
Moslem Imani, Rey-Jer You, Chung-Yen Kuo*
Department of Geomatics, National Cheng-Kung University
Font Enlarge
The Caspian Sea located at the boundary of Asia and Europe and surrounded by five countries, Azer¬baijan, Iran, Kazakhstan, Russia, and Turkmenistan, plays a key role in their interaction and is the largest enclosed inland water body in the world with a water surface of 370000 km2 and an average depth of 187 m. The water level fluctuation in Caspian Sea changing rapidly at around 1 meter in seasonal and interannual time scales resulting from the interactions of geographical and meteoro¬logical phenomena including river discharge, evaporation, precipitation, and water temperature, is relatively large as compared to Kaohsiung tide gauge records of 20-30 cm for example. Since the accurate prediction of sea level changes is extremely difficult in Caspian Sea due to large sea level fluctuations and is considerably important for coastal zone management and ship navigation activities, it is quite necessary to find an optimal forecasting technique for Caspian Sea forecasting based on available information and data. In the study, Holt-Winters exponential smoothing (HWES) for not only forecasting and smooth¬ing the analyzed function but also presenting data conve¬niently and eliminating random errors, are used to analyze and forecast the variations in Caspian Sea using sea level anomalies (SLA). SLA derived from 15-year (1993 - 2008) Topex/Poseidon and Jason-1 satellite altimeter has been pre-processed by a least squares polynomial interpolation to fill spatial and temporal data gaps. Fig. 1 shows that a 3-year forecasting SLAs (2005 - 2008) derived by HWES agree well with the altimeter observed SLAs with a correlation coefficient of 0.86 and a root mean square error of 7 cm.

Fig 1. Comparison of HWES forecasting and altimeter observed time sea level time series ( Imani et. Al., 2013)
< Previous
Next >
Copyright National Cheng Kung University