MARINE CHEMISTRY 56, 77-96

Temporal variability in dissolved phosphorus concentrations in the subtropical North Pacific Ocean

David M. Karl, Georgia Tien

SOEST, Department of Oceanography, University of Hawaii, Honolulu, HI 96822, USA

Dissolved phosphorus pool measurements were made on approximately monthly intervals during the period October 1988-November 1994 at an oligotrophic, deep-ocean station near Hawaii (Station ALOHA; 22'45'N, 158'W). Data include depth profiles (0-200 m) of soluble reactive phosphorus (SRP), measured by two independent techniques (a high-sensitivity, magnesium-induced coprecipitation (MAGIC) method, and standard autoanalysis procedure), and total dissolved phosphorus (TDP) based on UV photo-oxidation treatment followed by automated SRP analysis. Considerable variability in SRP was observed both with depth and with time over the 6-yr period. Upper water column (0-100 m) SRP concentrations varied in time from > 150 to < 10 nM, and generally conformed to one of two distinct concentration vs. depth patterns: "Type I" profiles were characterized by uniform SRP concentrations with depth (concentration gradients <= 0.05 µmol n^-4 SRP) and "Type II" profiles were characterized by distinct near-surface (0-30 m) SRP concentration maxima. The Type I profiles were further divided into low (Type I-L) or high (Type I-H) categories based on whether the average SRP concentration was < 60 or > 60 nM, respectively. Throughout the entire 6-yr observation period, the upper water column (0-100 m) inventory of SRP displayed a systematic decrease from ~10 to ~5 mmol m^-2 P, a result that is consistent with the hypothesized role of N₂ fixation and P control of new and export production at Station ALOHA. Three independent, but not mutually exclusive, models are presented to explain these time-varying SRP concentrations: (1) convective mixing; (2) atmospheric deposition; and (3) upward P flux. The upward P flux model including both passive (upward particle flux) and active (biological transport) processes, is the most consistent model for the available data set.