Sea surface salinity retrieval based on Levenberg Marquardt algorithm using satellite data

• Main Author: Abd. Rahim, Noorlida
• Format: Thesis
• Language: English
• Published: 2014
• Subjects: G70.39-70.6 Remote sensing
• Online Access: http://eprints.utm.my/id/eprint/51409/25/NoorlidaAbdRahimMFGHT2014.pdf; http://eprints.utm.my/id/eprint/51409/; http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:89334

Soil Moisture Ocean Salinity satellite exploits the frequency of 1.4 gigahertz which represents the best conditions for salinity retrieval. The new challenge is to interpret the observed brightness temperature into the salinity. The main objective of this study is to measure the sea surface salinity in the South China Sea using the Levenberg Marquardt algorithm. The methodology of this study involves the mapping of this algorithm to solve the non-linear least squares in order to obtain the salinity. The salinity was estimated based on the comparison between brightness temperature measured and brightness temperature simulated value of the successive iteration. The difference between both brightness temperature values is compared to the desired threshold at each iteration, this recursive process either updates the brightness temperature simulated or finally terminated if the brightness temperature difference is lower or higher than that threshold respectively. The salinity values estimated from the designed of Levenberg Marquardt algorithm tools were assembled, thus maps of sea surface salinity were produced. Some accuracy analyses were carried out to identify the appropriateness of a Levenberg Marquardt algorithm for the salinity retrieval. The results of the regression analysis and Pearson Correlation Coefficient indicate that sea surface salinity measured performs high correlation with the sea truth data, which is 0.9042 and ±0. 9509 psu, respectively. The analysis of variance by testing the hypothesis indicates that there is no substantial difference between the mean of sea surface salinity from the satellite and sea truth data. The root mean square error of measured sea surface salinity is smaller compared to the sea truth data values. In conclusion, the appropriateness of Levenberg Marquardt algorithm in inverting the salinity in the non-linear technique proved as a solution for ill-posed inversion that estimates the sea surface salinity from the Soil Moisture Ocean Salinity brightness temperature.