Research project SR/12/140 (Research action SR)
Several satellites measure Sea Surface Temperature (SST), each of these with different technical specificities and error sources. Together with in situ data, they form a highly complementary data set. The creation of merged SST products, integrating the strengths of each of its components and minimising their weaknesses, is however not an easy task, but it is certainly a desirable goal that has generated a large amount of research over the last years.
Objective
The objectives of this project are:
(i) To develop a technology that allows to merge different data sets at very different sampling intervals (in space and time) and create an integrated product at the highest sampling frequency and with the highest quality possible.
(ii) To provide improved, merged analyses of variables such as SST and chlorophyll.
(iii) Obtain a better understanding of inter-sensor differences, and of the diurnal cycle of the studied variables.
(iv) To better understand the relation between variables (and take advantage of this improved knowledge to ameliorate the analyses).
(v) Using the above-mentioned developments, explore the capability of DINEOF to produce SST forecasts based on multi-variate EOFs and model forecasts.
(vi) Finally, to improve DINEOF to meet user needs and required precision.
Method
DINEOF is a technique to infer missing data is satellite data sets. In this project we will further develop DINEOF so that it can merge different data sets. First, an initial DINEOF reconstruction of a data set with a high spatial resolution will be made, and the EOF basis obtained will be used as the covariance matrix needed to subsequently include into the analysis other data sources (satellite and in situ). Error estimations for each data set will be used to weight their influence in the final product. Special attention will be given to the diurnal cycle and the effect of diurnal warming in the quality of the measurements, and multivariate DINEOF analyses will be performed to investigate the influence of variables like wind and turbidity in these warming events. Finally, by combining satellite SST fields and model forecasts using a multivariate DINEOF, we will investigate the capability of DINEOF to produce SST forecasts, which will be compared to the forectas provided by numerical analyses.
Result
Merged high-resolution (in space and time) SST data sets and error statistics will be obtained for the study zones, for variable time frames within 2008 and 2010. The improvements made to the base technique used throughout the project (DINEOF) will be made available freely and openly the the scientific community (source code and documentation). Also, statistical parameters characterising the diurnal cycle and the difference between skin and bulk temperature will be obtained through this project. Finally, the technology to forecast SST based on statistical information and model data will be obtained as well. The portability of the developed techniques to other variables and domains will be reported. All results will be published in international peer-reviewed journals.
