Use of real-time observations in an operational ocean data assimilation system: the Mediterranean case

Abstract

Real-time observations are essential for operational forecasting that in turn can be used to predict changes of the state of the ocean and its associated biochemical fi elds. In addition, real-time observations are useful to detect changes in the past with the shortest delay, to standardize practices in data collection and to exchange data between remote regions of the ocean and seas. Th e drawback is that real-time observations could be less accurate than their delayed mode counterparts due to the time constraints for data dissemination. In situ real-time data are usually decimated to be transmitted in real time (loss of accuracy and resolution), whereas satellite data are corrected with approximate algorithms and less ancillary data. Delayed mode quality control analysis increases the value of the observational data set, fl agging outliers and producing climatological estimates of the state of the system. Th us real-time data, together with a modelling system and the climatological estimates, give the appropriate information for scientifi c studies and applications. Th e principles of operational science started to develop in the 1940s and 1950s, based on the combined use of real-time data and modelling systems that can extend the information from observations in space and time. Operational science is based on a sound knowledge of the dynamics and processes for the space/timescales of interest and operational meteorology and oceanography have started to implement these principles to weather and ocean forecasting activities. In the past 20 years, operational meteorology has become a reality with a network of in situ and satellite observations that has made the weather forecast capable of extending the theoretical limit of predictability of the atmosphere (only one-two days theoretically, now forecasts are useful for more than fi ve days on average). Today meteorological observations are mainly used in their assimilated form even if observations are still collected for specifi c process-oriented studies. Recently the meteorological re-analysis projects (Gibson et al., 1997; Kalnay et al., 1996) have released a wealth of data to be understood and analysed. Th ese data sets are coherent and approximately continuous (daily), fi lling the observational gaps in space and time with a dynamical interpolation scheme. Th e model and the real-time observations are fused in one best estimate of the state of the system by data-assimilation techniques that have been developed to a great degree of sophistication in recent years (Lorenc, 2002). Th e re-analysis data are now forming the basic reference data set to understand climate variability in the atmosphere and upper oceans. Ch20.indd 73Ch20.indd 733 3/7/07 9:58:01 AM Habwatch 734 Dynamical interpolation/extrapolation of observational data for operational forecasting in the ocean began to be investigated at the beginning of the 1980s and the fi rst successful forecasts were carried out in the open ocean (Robinson and Leslie, 1985). Th ese exercises required real-time data that were initially collected with rapid ship surveys realizing adaptive sampling schemes and collecting a combination of traditional recoverable and expendable instruments (CTD, XBTs). At the same time but in a totally independent way, shelf scale and coastal real-time data from moored and drifting sensors such as meteorological buoys and sea-level stations started to be used for shelf scale storm surge operational forecasting (Prandle, 2002). Operational oceanography is now building on this experience and considers real-time measurements from opportunity platforms and satellites in a manner very similar to operational meteorology. Th is chapter aims to show the use of real-time observations in a state-of-the-art ocean-predicting system realized in the Mediterranean. We discuss the pre-processing schemes required to properly assimilate the observations into an operational nowcasting/ forecasting system, elucidate the role and impact of diff erent observations in the assimilation system and show the use of real-time data to evaluate quality of the modelling system. We start with the description of the Mediterranean Forecasting System (MFS) real-time observing system and pre-processing quality control in Section 20.2, we then describe the modelling and assimilation system in relation to the impact of diff erent real-time observations in Section 20.3. In Section 20.4 we evaluate the consistency, quality and accuracy of the forecasting system using model-data intercomparison and Section 20.5 offers conclusions