Rapid Assessment of the Coastal Ocean Environment
207
the statistical properties of meso-scale features may be correct, but their occurrence
must not be regarded as a realistic synoptic prediction.
In an alternative and more realistic approach, real-time ocean assessment is initialIy based on observations, which are presumed to be essentialIy coincident in
time, and with spatial resolution high enough to resolve the processes of interest.
Ocean models initialized with the observations are then used to produce parameter
fields that are consistent with the observations. This process, often called "nowcast", balances the fields in a way that they obey the laws of physics and therefore
are more than just interpolations between observations.
Starting from the nowcast, the ocean model is integrated over a certain number
of time steps, driven by its own internal dynamics and by predictions of the external forcing fields. Inevitably, the period for which reliable forecasts can be produced by this technique is limited by three factors: a) the quality and density of
observations, b) the appropriateness of physical approximations in the forecast
model and c) the quality and resolution of forecasts for the forcing fields. Before
model results would degrade too much, model fields must be adjusted by new
observations. The melding of new data into a model run is a non-trivial scientific
issue, it is the third component of the ocean forecast system. An oceanographic
observation and prediction system (OOPS) for REA thus consists of Ist an observational network, 2nd a numerical prediction model and 3rd a data assimilation
scheme.
11.4.2 Data Acquisition and Management
In an REA scenario, very detailed results may be required for a small ocean area.
Since the area of interest cannot be isolated from the surrounding ocean, the influences from outside must be taken into account by appropriate boundary conditions.
A request for continuous in situ measurements along the boundary would be unreasonable. Boundary conditions can instead be obtained from a larger scale model
running on a coarser grid and either assimilated from time to time into the small
scale model as if it were measured data, or the different scale models can be synchronized leading to frequent adjustments at the boundary of the small model
domain. Synchronized models usually have the same number and vertical position
of layers. Their horizontal grid spacing differs by a factor of 3 in order to obtain
matching points on the staggered computation grid. Model nesting (recall section
11.2) may be carried out also with two-way data flow, thus improving the fields of
the larger scale model with results from the more ac curate small scale. The observational network required for nested models, includes an appropriate number of
measurement stations in the outer domain in order to avoid discrepancies between
water mass properties that might otherwise occur especialIy at depth. If climatological data is also used it must be checked and adjusted as necessary for compatibility with the new observations in both domains. Simple adoption of spun-up
climatological data could result in unrealistic density driven currents.
Data acquired for assimilation into a model need not satisfy the conditions for
sampling frequency in space and time that must be observed for initial conditions
207
the statistical properties of meso-scale features may be correct, but their occurrence
must not be regarded as a realistic synoptic prediction.
In an alternative and more realistic approach, real-time ocean assessment is initialIy based on observations, which are presumed to be essentialIy coincident in
time, and with spatial resolution high enough to resolve the processes of interest.
Ocean models initialized with the observations are then used to produce parameter
fields that are consistent with the observations. This process, often called "nowcast", balances the fields in a way that they obey the laws of physics and therefore
are more than just interpolations between observations.
Starting from the nowcast, the ocean model is integrated over a certain number
of time steps, driven by its own internal dynamics and by predictions of the external forcing fields. Inevitably, the period for which reliable forecasts can be produced by this technique is limited by three factors: a) the quality and density of
observations, b) the appropriateness of physical approximations in the forecast
model and c) the quality and resolution of forecasts for the forcing fields. Before
model results would degrade too much, model fields must be adjusted by new
observations. The melding of new data into a model run is a non-trivial scientific
issue, it is the third component of the ocean forecast system. An oceanographic
observation and prediction system (OOPS) for REA thus consists of Ist an observational network, 2nd a numerical prediction model and 3rd a data assimilation
scheme.
11.4.2 Data Acquisition and Management
In an REA scenario, very detailed results may be required for a small ocean area.
Since the area of interest cannot be isolated from the surrounding ocean, the influences from outside must be taken into account by appropriate boundary conditions.
A request for continuous in situ measurements along the boundary would be unreasonable. Boundary conditions can instead be obtained from a larger scale model
running on a coarser grid and either assimilated from time to time into the small
scale model as if it were measured data, or the different scale models can be synchronized leading to frequent adjustments at the boundary of the small model
domain. Synchronized models usually have the same number and vertical position
of layers. Their horizontal grid spacing differs by a factor of 3 in order to obtain
matching points on the staggered computation grid. Model nesting (recall section
11.2) may be carried out also with two-way data flow, thus improving the fields of
the larger scale model with results from the more ac curate small scale. The observational network required for nested models, includes an appropriate number of
measurement stations in the outer domain in order to avoid discrepancies between
water mass properties that might otherwise occur especialIy at depth. If climatological data is also used it must be checked and adjusted as necessary for compatibility with the new observations in both domains. Simple adoption of spun-up
climatological data could result in unrealistic density driven currents.
Data acquired for assimilation into a model need not satisfy the conditions for
sampling frequency in space and time that must be observed for initial conditions