|
Assimilation
of chlorophyll and nutrients into a 3-D marine carbon cycle model
(NASA-Carbon cycle)
Box
model assimilation
To investigate the performance of the
ecosystem model parameterization in the GCM, a simplified box model has
been developed with the same biogeochemical parameterizations as in the
3-D model. Significant reduction of the cost functions were generated
with a reasonable fit to the observed chlorophyll concentration in the
BATS region [Schartau et al.,
2001]. The mixing paramater [Fasham
et al., 1990] and the phytoplankton growth are appeared to be
dominant in controlling the amplitude of the seasonal phytoplankon
bloom. When SeaWiFS observation was assimilated into the box model, the
biological parameters with the phytoplankton growth rate parameter are
sensitive.
(a)
(b)
Figure 1. Chlorophyll concentration resulted from
assimilating BATS (a) and SeaWiFS (b) chlorophyll-a data into a
box model.
|
Experiment
|
Parameter Set
|
Cost
Function Reduction
|
|
BATS
|
SeaWiFS
|
|
MIX
|
GOWTH
|
BIO
|
Monthly
|
Biweekly
|
Monthly
|
|
E1
|
X
|
|
|
44%
|
40%
|
54%
|
|
E2
|
|
X
|
|
54%
|
59%
|
67%
|
|
E3
|
|
|
X
|
7%
|
18%
|
59%
|
|
E4
|
X
|
X
|
|
56%
|
57%
|
54%
|
|
E5
|
X
|
|
X
|
42%
|
37%
|
53%
|
|
E6
|
|
X
|
X
|
35%
|
13%
|
63%
|
|
E7
|
X
|
X
|
X
|
50%
|
48%
|
56%
|
|
E8
|
X
|
X
|
X
|
67%
|
66%
|
92%
|
Table 3. List of experiment performance when
different parameters are optimized in the assimilation. Assimilation E8
uses optimized parameters from E1 and E2.
Since it is computationally
inefficient to explore the cost function minimum using the 3-D
model, we used the box model to generate 10,000 iterations of
identical twin experiments using randomly perturbed control parameters.
The frequency distribution of the minimized cost function shows that
most of the assimilation did not reached the global minimum (Figure 2).
Figure 2. Minimum cost function distribution from
10,000 iterations of box model assimilation.
References
Fasham, M. J. R., H. W. Ducklow, and S. M. McKelvie (1990), A
nitrogen-based model of plankton dynamics in
the ocean mixed layer, J. Mar. Res.,
48, 591-639.
Schartau, M., A. Oschlies, and J. Willebrand (2001). Parameter
estimates of a zero-dimensional ecosystem model applying the adjoint
method, Deep Sea Res. Part. II, 48,
1760-1800.
|
|