The application of functional linear models on remote sensing data in oceanography

Abstract

With the development of technology, functional data analysis gain importance in analysing large data sets where the observations are sampled over a dense time interval, space or on a spectrum that consists of different frequency channels. Remote Sensing (RS) data obtained from satellites are an example of spectral data. Particulary in oceanography, RS data are used to predict ocean characteristic parameters such as Sea Surface Temperature (SST), Chlorophyll-a content (Chl-a) and Total Suspended Solids (TSS). Different functional linear regression models for scalar responses are applied on the Remote Sensing (RS) data obtained from full spatial resolution (FRS) MEdium Resolution Imaging Spectrometer (MERIS) on board the Envisat multispectral platform. The purpose is to predict the amount of TSS in the coastal zone adjacent to the Guadalquivir estuary. The models are compared by using Mean Error of Prediction (MEP) computed from Leave One Out Cross Validation (LOOCV).