19 Forecasting the Coastal Optical Properties Using Satellite Ocean Color
339
Depending on the area and cloud coverage, the bio-optical coverage takes about
1–2 weeks before a coherent bio-optical field initialization and forecast field can
be established. Additionally, as increased satellite bio-optical data enters the initialization process, the bio-optical forecast improves, and it is then used in the next
day’s initialization field. This forecast “spin-up” time has been used in a variety of
“cloudy” coastal regions with partial ocean coverage. A similar procedure to generate the initialization field is used in weather forecasting. The procedure has the
advantage that for each day the “best” and most recent data enters the forecast.
However, the entire bio-optical properties along a coast line can change from 1 day
to the next, for example, if a “cloud free” scene enters the initialization field when
the previous days initialization was based only on a forecast. Because the initialization is performed daily, reinitialization to observation conditions is rapid and the
forecast improves. An example of the initialization field of backscattering is shown
in Fig. 19.2 (panel b).
19.3 Forecasting Using Eulerian Advection
The forecasting of bio-optical properties is performed by applying a simple advection approach to satellite derived bio-optical products using the NCOM forecast
circulation model in order to forecast the surface optical properties. As was
described earlier, the theoretical basis for this, assumes bio-optical properties are
controlled solely by the physical circulation within a 24 h cycle. The bio-optical
processes such as phytoplankton growth and decay and CDOM production and oxidation are not considered and are remissive. Previous efforts used a Lagrangian
advection approach; however, this process was shown to produce significant errors
along coastal boundaries in addition to the high computing requirements. For these
reasons we switched to an eulerian approach (Arnone et al., 2006).
The NCOM model and the satellite derived bio-optical initialization field pixel
grid are established based on the grid resolution of the image (in this case 1 km).
The vertical fluxes into and out of the grid cells of the bio-optical properties are estimated from the horizontal fields. We apply a “thin-layer approximation” to extend
surface 2D advection to 3 dimensions. The advection is performed on a surface
layer, for convenience, at 1 m thickness, which goes up and down with free sea
surface such that the vertical velocity is 0 at the surface. The vertical velocity at
the base of the layer can be determined from divergence/convergence of horizontal
currents following the volume conservation. The vertical flux is estimated assuming
a uniform concentration of field at vertical. Forward time stepping with first-order
upwind advection is applied to the vertical advection. For the horizontal advection
of the satellite field, a third-order upwind advection scheme with flow adjustment is
applied to reduce diffusion and to prevent “a numerical overshoot”.
So, e.g., as the surface layer bio-optical properties move offshore and diverge
from coastal boundaries, the vertical flux replenishes the bio-optical concentration
from a vertical upwelling (flux) of the subsurface bio-optical property. Similarly,
the vertical flux of bio-optical concentration can account for downwelling flux into
339
Depending on the area and cloud coverage, the bio-optical coverage takes about
1–2 weeks before a coherent bio-optical field initialization and forecast field can
be established. Additionally, as increased satellite bio-optical data enters the initialization process, the bio-optical forecast improves, and it is then used in the next
day’s initialization field. This forecast “spin-up” time has been used in a variety of
“cloudy” coastal regions with partial ocean coverage. A similar procedure to generate the initialization field is used in weather forecasting. The procedure has the
advantage that for each day the “best” and most recent data enters the forecast.
However, the entire bio-optical properties along a coast line can change from 1 day
to the next, for example, if a “cloud free” scene enters the initialization field when
the previous days initialization was based only on a forecast. Because the initialization is performed daily, reinitialization to observation conditions is rapid and the
forecast improves. An example of the initialization field of backscattering is shown
in Fig. 19.2 (panel b).
19.3 Forecasting Using Eulerian Advection
The forecasting of bio-optical properties is performed by applying a simple advection approach to satellite derived bio-optical products using the NCOM forecast
circulation model in order to forecast the surface optical properties. As was
described earlier, the theoretical basis for this, assumes bio-optical properties are
controlled solely by the physical circulation within a 24 h cycle. The bio-optical
processes such as phytoplankton growth and decay and CDOM production and oxidation are not considered and are remissive. Previous efforts used a Lagrangian
advection approach; however, this process was shown to produce significant errors
along coastal boundaries in addition to the high computing requirements. For these
reasons we switched to an eulerian approach (Arnone et al., 2006).
The NCOM model and the satellite derived bio-optical initialization field pixel
grid are established based on the grid resolution of the image (in this case 1 km).
The vertical fluxes into and out of the grid cells of the bio-optical properties are estimated from the horizontal fields. We apply a “thin-layer approximation” to extend
surface 2D advection to 3 dimensions. The advection is performed on a surface
layer, for convenience, at 1 m thickness, which goes up and down with free sea
surface such that the vertical velocity is 0 at the surface. The vertical velocity at
the base of the layer can be determined from divergence/convergence of horizontal
currents following the volume conservation. The vertical flux is estimated assuming
a uniform concentration of field at vertical. Forward time stepping with first-order
upwind advection is applied to the vertical advection. For the horizontal advection
of the satellite field, a third-order upwind advection scheme with flow adjustment is
applied to reduce diffusion and to prevent “a numerical overshoot”.
So, e.g., as the surface layer bio-optical properties move offshore and diverge
from coastal boundaries, the vertical flux replenishes the bio-optical concentration
from a vertical upwelling (flux) of the subsurface bio-optical property. Similarly,
the vertical flux of bio-optical concentration can account for downwelling flux into