110
Chapter 7: Provinces: The Secondary Compartments
readjusted using an iterative procedure until, ideally, a single SST peak appears in each
province; chlorophyll concentration is taken as an indicator of water mass boundaries in
the same way where SST cannot be used. Applied to the North Atlantic provinces, and
for spring and autumn of 2003, this enabled their boundaries to be refined to match
the distribution of properties in the ocean in those periods, and suggested that another
entity (“Slope Water”) might usefully be integrated into the partition. Finally, the use
of an algorithm that isolates diatom-dominated regional phytoplankton biomass from
other populations shows that the dynamic boundaries derived from the temperature field
in the NW Atlantic were a good match for boundaries between diatom-dominated and
other phytoplankton populations.
Ways of Testing Static Province Boundaries
in the Open Ocean
It would be useful to have some confirmation that the boundaries selected intuitively
by analysis of climatological data, and the number of provinces so identified, were not
entirely subjective. This can be done in several different ways: (i) to test statistically
that conditions differ in adjacent provinces, (ii) to compare data on the distribution of
individual biota with boundaries between provinces, and (iii) to use analytical techniques
to partition a relevant global data set.
A Statistical Test
Confirmation of the reality of boundaries was obtained by distributing the individual
data from a suitable global archive into compartments representing provinces, and then
investigating statistically the similarities and differences between the individual data sets
that then represented the attributes of individual provinces.
This was done with the 21,872 sets of parameters descriptive of the chlorophyll profile
obtained from the global archive described previously. These values were aggregated
seasonally, centered on the 15th day of January, April, July, and October to represent
the boreal (austral) winter (summer), spring (fall), summer (winter), and fall (spring) in
each province. Of the resulting 204 possible cases (four seasons in each of 51 provinces),
for 145 there were >25 profiles, for 38 there were <25 profiles, and for 19 cases there
were no profiles at all, though some of these represented polar provinces during winter
darkness. This seemed a much better result than might be expected.
Because the depth and relative strength of the deep chlorophyll maximum layer is an
integral of the nutrient, light, and density environments of phytoplankton and because it
is the most consistent feature of chlorophyll profiles, it seemed to be a good criterion for
testing the proposed boundaries. The depth of the chlorophyll maximum is described by
the parameter Z m . Where this is a positive integer, it represents the subsurface depth of
the layer; where Z m is a negative integer, the chlorophyll maximum lies at the surface.
It is not surprising, therefore, that of the parameters describing the chlorophyll profiles,
Z m consistently showed (as I shall discuss) the greatest differences between regions and
seasons when partitioned by the proposed set of boundaries.
Some of the resulting seasonal frequency distributions of Z m were bimodal, suggesting
nonuniform conditions for that season with a province. The reason for this was resolvable
by identifying the data creating the minor peak. In most cases, this showed that the
anomalous data were a discrete set of observations, concentrated in one part of the
province, often over a restricted period. In such cases, it was usually possible to obtain
Chapter 7: Provinces: The Secondary Compartments
readjusted using an iterative procedure until, ideally, a single SST peak appears in each
province; chlorophyll concentration is taken as an indicator of water mass boundaries in
the same way where SST cannot be used. Applied to the North Atlantic provinces, and
for spring and autumn of 2003, this enabled their boundaries to be refined to match
the distribution of properties in the ocean in those periods, and suggested that another
entity (“Slope Water”) might usefully be integrated into the partition. Finally, the use
of an algorithm that isolates diatom-dominated regional phytoplankton biomass from
other populations shows that the dynamic boundaries derived from the temperature field
in the NW Atlantic were a good match for boundaries between diatom-dominated and
other phytoplankton populations.
Ways of Testing Static Province Boundaries
in the Open Ocean
It would be useful to have some confirmation that the boundaries selected intuitively
by analysis of climatological data, and the number of provinces so identified, were not
entirely subjective. This can be done in several different ways: (i) to test statistically
that conditions differ in adjacent provinces, (ii) to compare data on the distribution of
individual biota with boundaries between provinces, and (iii) to use analytical techniques
to partition a relevant global data set.
A Statistical Test
Confirmation of the reality of boundaries was obtained by distributing the individual
data from a suitable global archive into compartments representing provinces, and then
investigating statistically the similarities and differences between the individual data sets
that then represented the attributes of individual provinces.
This was done with the 21,872 sets of parameters descriptive of the chlorophyll profile
obtained from the global archive described previously. These values were aggregated
seasonally, centered on the 15th day of January, April, July, and October to represent
the boreal (austral) winter (summer), spring (fall), summer (winter), and fall (spring) in
each province. Of the resulting 204 possible cases (four seasons in each of 51 provinces),
for 145 there were >25 profiles, for 38 there were <25 profiles, and for 19 cases there
were no profiles at all, though some of these represented polar provinces during winter
darkness. This seemed a much better result than might be expected.
Because the depth and relative strength of the deep chlorophyll maximum layer is an
integral of the nutrient, light, and density environments of phytoplankton and because it
is the most consistent feature of chlorophyll profiles, it seemed to be a good criterion for
testing the proposed boundaries. The depth of the chlorophyll maximum is described by
the parameter Z m . Where this is a positive integer, it represents the subsurface depth of
the layer; where Z m is a negative integer, the chlorophyll maximum lies at the surface.
It is not surprising, therefore, that of the parameters describing the chlorophyll profiles,
Z m consistently showed (as I shall discuss) the greatest differences between regions and
seasons when partitioned by the proposed set of boundaries.
Some of the resulting seasonal frequency distributions of Z m were bimodal, suggesting
nonuniform conditions for that season with a province. The reason for this was resolvable
by identifying the data creating the minor peak. In most cases, this showed that the
anomalous data were a discrete set of observations, concentrated in one part of the
province, often over a restricted period. In such cases, it was usually possible to obtain