Publications 2015

Filtering remotely sensed chlorophyll concentrations in the Red Sea using a space-time covariance model and a Kalman filter
D. Dreano, B. Mallick, and I. Hoteit
Spatial Statistics, Volume 13, Pages 1–20, 2015
D. Dreano, B. Mallick, and I. Hoteit
Chlorophyll concentration; Covariance models; Geostatistics; Kalman filter; Space-time statistics
A statistical model is proposed to filter satellite-derived chlorophyll concentration from the Red Sea, and to predict future chlorophyll concentrations. The seasonal trend is first estimated after filling missing chlorophyll data using an Empirical Orthogonal Function (EOF)-based algorithm (Data Interpolation EOF). The anomalies are then modeled as a stationary Gaussian process. A method proposed by Gneiting (2002) is used to construct positive-definite space-time covariance models for this process. After choosing an appropriate statistical model and identifying its parameters, Kriging is applied in the space-time domain to make a one step ahead prediction of the anomalies. The latter serves as the prediction model of a reduced-order Kalman filter, which is applied to assimilate and predict future chlorophyll concentrations. The proposed method decreases the root mean square (RMS) prediction error by about 11% compared with the seasonal average.

DOI: 10.1016/j.spasta.2015.04.002