Perpustakaan Digital ITB

Satellite altimetry has been a store for huge spatial data. The ability to cover oceans immediately enhances geodetic applications such as deriving gravity anomaly using height slopes constructed from sea surface heights. Least Square Collocation is one of methods to estimate gravity anomaly. This research attempts to evaluate several local covariance functions as kernel for Least Square Collocation. Data used in this experiment are sea surface height from Cryosat-2A, EGM2008 gravity anomaly model with no wavelength truncation, and EGM2008 gravity anomaly with truncation at degree 360. Hirvonen and Markov models are two covariance functions investigated in this experiment. To qualify their performance, they must satisfy several test: R-squared value, positive definiteness check, singularity and matrix condition number check, root mean square value, and spectral characteristics. The experiments predict gravity anomaly using matrices that combine several covariance functions: auto-covariace of height slope, auto-covariance of gravity anomaly, and cross-covariance between height slope and gravity anomaly. The result shows that at Banda Sea, height slopes are not enough to delete timevarying forces affecting sea surface height. Hence, some noises remain and distract covariance function prediction. Several algorithms and strategies handle the data processing. Even then, the running time is still long in which more improvements in the algorithm design are necessary.