424 John Woods
file of solar irradiance (Fasham 1990). So the case study focused on ocean optics.
The three sources of optical errors are:
E d the optical code used in the WB model to compute irradiance profile from the
surface irradiance and the chlorophyll profile.
E h this has two components in the LE method: (a) the resolution ofthe grid, and
(b) the number of particles used to simulate the plankton populations, and therefore
the chlorophyll profile.
E S the optical boundary conditions were computed from data on monthly mean
cloud cover, using a well-known cloud-radiation code from Paltridge and Platt
(1976).
The question is: How large are these three errors, and which contributed most to
the error in the Spring bloom error in the Azores VPE? We follow the analysis of
Liu et al (2001).
The optical code in the WB model (E d )
Errors in the model equations can be assessed by using independent data sets that
isolate one function, in this case bio-optical feedback, which computes the profile
of irradiance from the profile of chlorophyll concentration. A suitable data set had
been collected during the JGOFS spring bloom experiment (Ducklow and Harris
1993). It comprised simultaneous profiles of chlorophyll concentration and solar
irradiance (upwelling and downwelling in the photosynthetically active range 400700 nm). Liu et al (1999) developed a new optical code based on the radiative
transfer equation, which fitted the JGOFS data much better than the simpler MoreI
(1991) code used in the WB model. Substituting this new code reduced the optical
contribution to E d from 100% to less than 5%. The SeaWIFS ETT was repeated
for a new version of the Azores VPE based on this new optical code. It tumed out
that the gap between observation and simulation ofthe spring bloom was increased
slightly by this reduction in E d (for reasons diagnosed in Liu et al 2001).
Optimising the infant mortality parameter (18.9) increased primary production so
that the available nitrogen was consumed faster, with the re suIt that the spring
bloom peaked earlier. This reduced the SeaWIFS ETT error by 10% (five days in
fifty).
The LE method (E h )
The Eulerian grid used in the WB model to define the profiles of chlorophyll
concentration and solar irradiance has a layer spacing of one metre. Higher resolution is used in the top metre. That is needed to reduce optically-driven errors in
computing the turbocline depth. As the result there is no significant contribution to
E h from grid resolution.
However, the accuracy of the computed chlorophyll profile depends on the number of particles used to simulate the plankton. The number of particles varies
according to the particle management rules (see 18.6.1). Tests showed that the
default rule, which ensured that there are never fewer than 20 particles per onemetre-thick layer, gave an error (E h ) which significantly affected the simulation of
file of solar irradiance (Fasham 1990). So the case study focused on ocean optics.
The three sources of optical errors are:
E d the optical code used in the WB model to compute irradiance profile from the
surface irradiance and the chlorophyll profile.
E h this has two components in the LE method: (a) the resolution ofthe grid, and
(b) the number of particles used to simulate the plankton populations, and therefore
the chlorophyll profile.
E S the optical boundary conditions were computed from data on monthly mean
cloud cover, using a well-known cloud-radiation code from Paltridge and Platt
(1976).
The question is: How large are these three errors, and which contributed most to
the error in the Spring bloom error in the Azores VPE? We follow the analysis of
Liu et al (2001).
The optical code in the WB model (E d )
Errors in the model equations can be assessed by using independent data sets that
isolate one function, in this case bio-optical feedback, which computes the profile
of irradiance from the profile of chlorophyll concentration. A suitable data set had
been collected during the JGOFS spring bloom experiment (Ducklow and Harris
1993). It comprised simultaneous profiles of chlorophyll concentration and solar
irradiance (upwelling and downwelling in the photosynthetically active range 400700 nm). Liu et al (1999) developed a new optical code based on the radiative
transfer equation, which fitted the JGOFS data much better than the simpler MoreI
(1991) code used in the WB model. Substituting this new code reduced the optical
contribution to E d from 100% to less than 5%. The SeaWIFS ETT was repeated
for a new version of the Azores VPE based on this new optical code. It tumed out
that the gap between observation and simulation ofthe spring bloom was increased
slightly by this reduction in E d (for reasons diagnosed in Liu et al 2001).
Optimising the infant mortality parameter (18.9) increased primary production so
that the available nitrogen was consumed faster, with the re suIt that the spring
bloom peaked earlier. This reduced the SeaWIFS ETT error by 10% (five days in
fifty).
The LE method (E h )
The Eulerian grid used in the WB model to define the profiles of chlorophyll
concentration and solar irradiance has a layer spacing of one metre. Higher resolution is used in the top metre. That is needed to reduce optically-driven errors in
computing the turbocline depth. As the result there is no significant contribution to
E h from grid resolution.
However, the accuracy of the computed chlorophyll profile depends on the number of particles used to simulate the plankton. The number of particles varies
according to the particle management rules (see 18.6.1). Tests showed that the
default rule, which ensured that there are never fewer than 20 particles per onemetre-thick layer, gave an error (E h ) which significantly affected the simulation of