Affordable Access

Developing a data assimilation system for the BSH operational model of the North and Baltic Seas: Results and Perspectives.

Publication Date
  • Computer Science
  • Earth Science
  • Geography


Within the DeMarine-Environment project--- as a part of the European Global Monitoring for Environment and Security (GMES) initiative,--- a data assimilation (DA) system has been developed for the operational circulation model of the German Maritime and Hydrographic Agency(BSH). In order to improve forecast of hydrographic characteristics in the North and Baltic Seas, Singular Evolutive Interpolated Kalman (SEIK) filter algorithm has been locally implemented for assimilating the NOAA AVHRR-derived sea surface temperature (SST). The degree of the forecast quality improvement depends on the assumption about model and data error statistics. ~27% of forecast error reduction has been achieved for SST forecast over the period of October 2007 - September 2008 under best-tuned assumptions on standard SST data error of 0.8oC and exponential correlation function within radius of 100km. Pilot real-time data assimilative pre-operational runs manifest much higher quality of the SST forecast in March 2011 in comparison with the regular BSH forecast without DA. On average, over that period, the root mean square (RMS) error has been decreased from 0.8oC to 0.5oC. The experiments conducted with different timing and frequency of data assimilation and variable forecasting periods show that the data assimilation system enables one to correct the systematic model uncertainties and, due to memory on the corrections, better predict over periods of up to 5 days. Our results also apparently illustrate the bias in AVHRR daytime product, but, however, reveal low informative influence of the data on the forecasting system when daytime SSTs are assimilated additionally to 'midnight' observations. To this end, it is worth noting that despite the high dependency of forecast quality improvement on the assumption about model uncertainties and data errors, statistics of which are not always, if ever, known a priori, the combination of the information from two different sources- the model and the data,- provided by data assimilation, might itself improve our understanding of both these sources and help to optimize the system.

There are no comments yet on this publication. Be the first to share your thoughts.


Seen <100 times