346
R. Arnone et al.
the bathymetry (3) model grid spacing (4) vertical layer spacing (5) boundary conditions etc. The physical model can run a set of ensemble (typically ∼40) to determine
the uncertainty in the currents forecast field (Rixen et al., 2009). Using the current
uncertainty, we can estimate the optical forecast for each of the ensemble currents
and define the mean and spread of the optical forecast. We did not do this in these
examples and assume the model currents were valid and correct. This is one source
of optical forecast uncertainty that was not taken into account.
Satellite retrievals of the bio-optical properties (backscattering coefficient or
chlorophyll) have an uncertainty based on the uncertainty of the algorithms. We
implemented a set of uniform optical relationships that are used in the QAA algorithms which are considered standard. However, these relationships can change with
location and bio-optical processes. The QAA algorithm has been shown to have
some degree of uncertainty, (Lee et al., 2010) however these relationships should
work well in the Gulf of Mexico. The uncertainty from these algorithms can influence the initialization and forecast. We assume the uncertainty of the algorithms
is similar from 1 day’s satellite image to another. Lastly, satellite retrievals require
processing for atmospheric correction in addition to the in-water algorithms. The
uncertainty in the aerosol models used for atmospheric correction is spatially and
temporally changing, especially in coastal areas where aerosol optical depths are
variable. We did not include this source of uncertainty of the atmospheric correction
in the MODIS backscattering retrievals that were used in the forecast.
19.6 Conclusion
The retrievals of bio-optical properties from ocean color satellite have made significant advances in defining a “nowcast” of coastal conditions. New capability is
required for coastal operations, coastal managers and researchers to forecast these
properties on time scales of hours to weeks. The coastal environment changes on
scales of hours, mostly as a result of the physical forcing associated with tidal,
discharge and currents. We coupled daily surface ocean color properties of the
backscattering coefficient at 551 nm and chlorophyll concentration to a physical circulation model and advected the field to determine an hourly forecast of the satellite
derived properties.
The forecast of bio-optical properties is reinitialized daily as new satellite observations enter the forecast. The methods to construct a gap filled initialization field
and integrate it into a coastal bio-optical forecast system required approximately
1–2 weeks spin-up time, and is based on availability of cloud free observations. We
illustrate a 1 month daily forecast of the surface particle backscattering properties
for October 2009.
A daily 24 h forecast of the backscattering coefficient was evaluated based on
comparison with the “next day’s” derived product. We conducted the validation and
uncertainty of daily bio-optical forecast for a 1 month period to estimate spatial statistical relationships of the forecast uncertainty. The October 2009 statistics (mean
R. Arnone et al.
the bathymetry (3) model grid spacing (4) vertical layer spacing (5) boundary conditions etc. The physical model can run a set of ensemble (typically ∼40) to determine
the uncertainty in the currents forecast field (Rixen et al., 2009). Using the current
uncertainty, we can estimate the optical forecast for each of the ensemble currents
and define the mean and spread of the optical forecast. We did not do this in these
examples and assume the model currents were valid and correct. This is one source
of optical forecast uncertainty that was not taken into account.
Satellite retrievals of the bio-optical properties (backscattering coefficient or
chlorophyll) have an uncertainty based on the uncertainty of the algorithms. We
implemented a set of uniform optical relationships that are used in the QAA algorithms which are considered standard. However, these relationships can change with
location and bio-optical processes. The QAA algorithm has been shown to have
some degree of uncertainty, (Lee et al., 2010) however these relationships should
work well in the Gulf of Mexico. The uncertainty from these algorithms can influence the initialization and forecast. We assume the uncertainty of the algorithms
is similar from 1 day’s satellite image to another. Lastly, satellite retrievals require
processing for atmospheric correction in addition to the in-water algorithms. The
uncertainty in the aerosol models used for atmospheric correction is spatially and
temporally changing, especially in coastal areas where aerosol optical depths are
variable. We did not include this source of uncertainty of the atmospheric correction
in the MODIS backscattering retrievals that were used in the forecast.
19.6 Conclusion
The retrievals of bio-optical properties from ocean color satellite have made significant advances in defining a “nowcast” of coastal conditions. New capability is
required for coastal operations, coastal managers and researchers to forecast these
properties on time scales of hours to weeks. The coastal environment changes on
scales of hours, mostly as a result of the physical forcing associated with tidal,
discharge and currents. We coupled daily surface ocean color properties of the
backscattering coefficient at 551 nm and chlorophyll concentration to a physical circulation model and advected the field to determine an hourly forecast of the satellite
derived properties.
The forecast of bio-optical properties is reinitialized daily as new satellite observations enter the forecast. The methods to construct a gap filled initialization field
and integrate it into a coastal bio-optical forecast system required approximately
1–2 weeks spin-up time, and is based on availability of cloud free observations. We
illustrate a 1 month daily forecast of the surface particle backscattering properties
for October 2009.
A daily 24 h forecast of the backscattering coefficient was evaluated based on
comparison with the “next day’s” derived product. We conducted the validation and
uncertainty of daily bio-optical forecast for a 1 month period to estimate spatial statistical relationships of the forecast uncertainty. The October 2009 statistics (mean