assimilation efficiencies (e.g., models that used an ‘Eppley-type’ formulation
clustered together). The most comprehensive study demonstrating the second point
was made by Saba et al. (2011). Output from 21 satellite-based ocean color models
were analyzed against[1000 in situ
14 C measurements from diverse ocean regions
spanning the Black Sea, Southern Ocean, high latitude North Atlantic, and subtropical Pacific and Atlantic. Some of the results are reproduced in Fig. 8.2 which
show the ensemble average root mean square error (RMSE) in each region. Lower
RMSE results from better predictive ability of NPP models (no distinction between
models is made here).
The last finding listed above from the PPARR exercises underscores that a
model’s ability to accurately predict NPP is highly dependent upon the uncertainties inherent in the input data (Fig. 8.2) (Saba et al. 2011). While this may seem
obvious, it is often difficult to propagate errors through complex analytical formulations resulting in model NPP estimates being reported with no accompanying
error estimates. Saba et al. (2010) found that the majority of models were unable to
accurately reproduce the observed trends in NPP at open ocean sites in the North
Pacific (Station ALOHA near Hawaii) and the Atlantic (BATS near Bermuda).
However, holding all other properties the same, but employing in situ estimates of
Chl in the NPP models allowed many of the models to match the sign, and to a
lesser extent, the magnitude of the observed trends in NPP. Siegel et al. (2001) also
demonstrated the influence of using time-varying or site-averaged photosynthetic
parameters in NPP models at BATS and found the predictive ability dropped significantly (r
2 = 0.80 to r
2 = 0.27) when mean values were used.
8.5 Major Findings
One of the fundamental realizations provided by remote sensing estimates of
marine NPP is that the ocean contribution to biospheric annual NPP is roughly
equivalent to that from terrestrial sources. Field et al. (1998) first synthesized
Fig. 8.2 Average root mean
square difference (RMSD) of
21 satellite NPP models
evaluated against field 14C
measurements for several
different locations. Error bars
represent 2x standard error.
Blue portions of each bar
represent maximum potential
reduction in RMSD if
uncertainties in model input
and field measurements are
accounted for. For details
regarding specific models and
field dataset used, see Saba
et al. (2011)
212
T. K. Westberry and M. J. Behrenfeld
clustered together). The most comprehensive study demonstrating the second point
was made by Saba et al. (2011). Output from 21 satellite-based ocean color models
were analyzed against[1000 in situ
14 C measurements from diverse ocean regions
spanning the Black Sea, Southern Ocean, high latitude North Atlantic, and subtropical Pacific and Atlantic. Some of the results are reproduced in Fig. 8.2 which
show the ensemble average root mean square error (RMSE) in each region. Lower
RMSE results from better predictive ability of NPP models (no distinction between
models is made here).
The last finding listed above from the PPARR exercises underscores that a
model’s ability to accurately predict NPP is highly dependent upon the uncertainties inherent in the input data (Fig. 8.2) (Saba et al. 2011). While this may seem
obvious, it is often difficult to propagate errors through complex analytical formulations resulting in model NPP estimates being reported with no accompanying
error estimates. Saba et al. (2010) found that the majority of models were unable to
accurately reproduce the observed trends in NPP at open ocean sites in the North
Pacific (Station ALOHA near Hawaii) and the Atlantic (BATS near Bermuda).
However, holding all other properties the same, but employing in situ estimates of
Chl in the NPP models allowed many of the models to match the sign, and to a
lesser extent, the magnitude of the observed trends in NPP. Siegel et al. (2001) also
demonstrated the influence of using time-varying or site-averaged photosynthetic
parameters in NPP models at BATS and found the predictive ability dropped significantly (r
2 = 0.80 to r
2 = 0.27) when mean values were used.
8.5 Major Findings
One of the fundamental realizations provided by remote sensing estimates of
marine NPP is that the ocean contribution to biospheric annual NPP is roughly
equivalent to that from terrestrial sources. Field et al. (1998) first synthesized
Fig. 8.2 Average root mean
square difference (RMSD) of
21 satellite NPP models
evaluated against field 14C
measurements for several
different locations. Error bars
represent 2x standard error.
Blue portions of each bar
represent maximum potential
reduction in RMSD if
uncertainties in model input
and field measurements are
accounted for. For details
regarding specific models and
field dataset used, see Saba
et al. (2011)
212
T. K. Westberry and M. J. Behrenfeld