An evaluation of novel remotely sensed data to improve and verify ocean-atmosphere forecasting
The aim of this study is to evaluate the use of novel remote observations and spatial data analysis to improve the skill of an ocean forecasting system for the central Mediterranean Sea. A high-resolution (0.042 by 0.042ๆ ocean forecasting system was setup consisting of an atmosphere model (NCEP Eta model) that was coupled to an ocean model (Princeton Ocean Model). This coupling consisted of the provision of surface atmospheric fluxes predicted at 3-hourly intervals to drive forward the ocean model. This research study dealt with a variety of aspects to improve this forecasting system using an inter-disciplinary approach. The main aspect of this thesis is an evaluation of novel, remotely- sensed data acquired by an orbiting passive microwave sensor as a tool to assess and improve ocean forecasting. Thus, SST derived by the Tropical Microwave Imager onboard the TRMM satellite was evaluated for its potential to define one of the lower boundary conditions of the Eta model. The impact was positive, and resulted in an average improvement of the skill of the model to predict lower surface marine winds by approximately 10%. TMI-data proved extremely useful to derive instantaneous turbulent heat fluxes and other surface geophysical fields that were needed to diagnose and fine-tune the skill of the Eta model to forecast these fields. The TMI SST product also proved to be a valuable data source for data assimilation by the ocean model. An optimised data assimilation scheme was derived resulting in a bias of just -0.05 С after a 15-day model integration run. This thesis shows how spatial data analysis can provide more detailed information about the high-resolution forecasts and their quality in addition to standard verification tools. Routines that explore the spatial data of the forecasts, observations and their relationship were developed and applied. Geostatistical analysis was used to model the spatial structure of the residual fields of the predictions and observations, and to translate the degree of spatial correlation in numerical and graphical terms.