Toxigenic Bloom Predictive Models - Monterey Bay

Source: Lane et al., 2009

Annual model

LOGIT(p) = 9.763 – 1.700[ln(silicic acid)] + 1.132
[ln(chl a)] – 0.800(temp) + 0.006(upwelling)

Spring model

LOGIT(p) = 5.835 + 1.398[ln(chl a)] – 1.135[ln(silicic
acid)] – 0.549(temp)

Fall–Winter model

LOGIT(p) = 10.832 – 5.026[ln(Pajaro River)] – 3.893
[ln(silicic acid)] + 1.972[ln(chl a)] + 0.652(nitrate)

Logistic regression models are of the form:

LOGIT(p)= ln[p/(1-p)]= β₀+ β₁z₁+ β₂z₂+…+β_kz_k

where p is the probability of the condition being modeled (here, the probability of a toxigenic Pseudo-nitzschia bloom), β₀ is a constant, and β₁,β_2,...β_k are the regression coefficients of z₁,z₂,...z_k, respectively. The independent variables are defined below.

The toxigenic bloom predictive models presented here were developed from a combined 8.5 years of phytoplankton sampling in Monterey Bay. Out of the 1071 surface water samples (depth ≤5m) analyzed, 576 contained toxigenic Pseudo-nitzschia cells. As not all data contained the same suite of environmental variables, it was necessary to evaluate the minimum combination of variables needed for the development of a successful model. Cases included in the models here required samples from Monterey Bay surface waters with toxigenic Pseudo-nitzschia species (P. australis and/or P. multiseries) concurrent with measurements of water temperature, chlorophyll a, and macronutrients (silicic acid and nitrate, specifically).