The author revisits Sverdrup’s Critical Depth Hypothesis using satellite products and concludes that it should be replaced by a Dilution-Recoupling Hypothesis:

[First] winter through spring changes in phytoplankton biomass are highly correlated with changes in PAR, and, [second], the largest changes in Cphyt coincide with spring MLD shoaling. Similar relationships are found for all 12 bins and may at first appear to confirm the Critical Depth Hypothesis, but in fact they do not. The key problem is that correlation between phytoplankton standing stock (Cphyt), PAR, and MLD shoaling does not imply that a similar relationship exists for net growth rates (r). This is a crucial point because Sverdrup’s Critical Depth Hypothesis is a statement that “shoaling of the mixed layer above the critical depth horizon initiates a spring bloom because it demarcates the first time when r becomes positive.”

In each case, Cphyt is already increasing when MLDs are at their maximum. [...] Indeed, in 104 out of the total 108 annual cycles available for the 12 bins, the onset of net positive growth in Cphyt is found to coincide with the cessation of mixed layer deepening. In no case is positive net growth delayed until significant shoaling of the mixed layer occurs.

The satellite record clearly indicates that North Atlantic bloom initiation is not a springtime event, but rather occurs in midwinter, at the latest. This finding strongly refutes the Critical Depth Hypothesis, but raises the new question: “Why does cessation of mixed layer deepening initiate a net increase in phytoplankton concentration?” To answer to this question, it is important to recognize (as alluded to previously) that net phytoplankton specific growth rates are not accurately reflected by changes in carbon concentration when a population is being diluted. [...] As explained by Evans and Parslow (1985), dilution effects must be considered in the North Atlantic when mixed layer deepening entrains phytoplankton-free water from below. If this dilution effect is sufficiently large, it is possible for phytoplankton concentration to decrease (e.g., in late autumn and early winter) despite net in situ population growth. To achieve a more complete understanding of North Atlantic seasonal phytoplankton cycles, it is therefore necessary to first account for dilution effects and then evaluate variability in r directly (as opposed to simply viewing log-transformed plots of Cphyt).To conduct such an analysis using satellite data, Eq. 3 can be used to estimate r from changes in Cphyt during periods of mixed layer shoaling and when mixed layer deepening entrains significant phytoplankton biomass from below. During periods when mixing entrains relatively phytoplankton-free water, r must instead be estimated from changes in mixed layer integrated phytoplankton biomass, which accounts for dilution of the population over a larger volume.

[D]uring each seasonal cycle r is negative over most of the summer and early autumn, then increases in parallel with increasing MLD and decreasing PAR to become positive by late autumn (horizontal dashed line indicates r=0), and thereafter remains positive until early summer. [...] Clearly, the positive net growth phase is not initiated by increasing PAR or mixed layer shoaling. The NA-5 record also illustrates the high temporal variability in r and shows that the peak in r is not consistently found in spring.

[T]he positive growth phase (December to June) begins roughly when the mixed layer penetrates below the euphotic layer [see Fig. 4B reproduced below]. Another characteristic of the mean annual cycles in r is a slight midwinter depression (e.g., January through March in Fig. 4B), which becomes more prominent at higher latitudes and gives the positive growth phase a “dual peaked” appearance (Fig. 4B). The mean annual cycle also shows the small and brief positive growth period associated with the September–October “fall bloom” (Fig. 4B).

This comparison illustrates the general inverse relationship between mu and r and strongly implies that the positive growth phase leading to the spring peak in phytoplankton biomass is less a consequence of springtime increases in mu as it is a reflection of seasonal changes in loss rates (l). In other words, the only way for r to increase while mu is decreasing is for the fraction of mu escaping losses to increases as mu decreases. [...T]he essential requirement to achieve comparable values of r from winter through spring (Fig. 4B) is that phytoplankton loss rates (l) must covary closely with mu, and herein lies the crucial flaw in the Critical Depth Hypothesis: Sverdrup assumed l to be constant over time.

I stopped on p. 985, the explanation of the Dilution-Recoupling Hypothesis.