Estimating upper ocean phosphate concentrations using ARGO float temperature profiles

The ARGO free-drifting profiling float array, with >3125 floats deployed between 60°N and 60°S latitudes at about 3° resolution as of May 2008 and each float profiling through 2000 m every 10 days, provides a comprehensive four-dimensional view of temperature and salinity in the world ocean. The resulting dataset complements satellite-based sea surface temperature (SST) measurements and similarly will complement future satellite-based sea surface salinity measurements. Although plans exist to add biogeochemical sensors to future floats, cost and depth restrictions may limit comprehensive upgrades to a fraction of all floats deployed after 2008. Temperature–nutrient (TN) relationships provide a mechanism to estimate nutrient concentrations from temperature to supplement sparser nutrient concentration measurements potentially obtained using non-chemical approaches like ISUS-based nitrate. Both negative and positive aspects of applying a temperature–phosphate (TP) linear regression matrix with global coverage (70°N and 70°S) are examined. The TP linear regression matrix was derived by combining an existing 1° latitude and longitude table of phosphate depletion temperatures (PDT) or X-intercepts with representative TP linear regression slopes derived from the GEOSECS dataset. Temperatures from datasets with associated latitude and longitude coordinates and, in some cases, measured phosphate concentrations ([PO4]) were matched with calculated TP linear regression slopes and Y-intercepts in the global matrix with 1° resolution using MSExcel Lookup worksheet functions to calculate TP-estimated [PO4]. The mean deviation of TP-estimated [PO4] <3.0 μM from measured [PO4] is 0.18±0.18 μM at Hawaii (HOT) and 0.04±0.08 μM at Bermuda (BATS) time series stations and 0.28±0.27 μM over all considered World Ocean Circulation Experiment (WOCE) stations representing the different ocean basins. In general, TP-estimated [PO4] represents measured [PO4] more accurately in the southern hemisphere than in the northern hemisphere. For the World Ocean Atlas 2005 (WOA05), a TP-estimated [PO4] map based on annual statistical mean SST approximates an annual statistical mean measured [PO4] map in overall geographic pattern but less so in absolute concentration. ARGO 0–10 m temperature data and derived TP-estimated [PO4] maps from all 2006 compare more favorably with the WOA05 annual statistical mean SST and measured [PO4] maps. For winter 2006, ARGO 0–10 m temperature and derived TP-estimated [PO4] maps favorably compare with MODIS mean SST and derived TP-estimated [PO4] maps. ARGO 30–50 m and 75–100 m temperatures from winter 2006 and derived TP-estimated [PO4] demonstrate that the ARGO dataset provides a subsurface nutrient complement to nutrient inferences based on MODIS SST. The TP-estimated [PO4] approach is responsive to changing conditions since the independent variable, temperature, integrates ambient environmental variability as expressed in the vertical thermal structure at the time of the ARGO profile. TP-estimated [PO4] can enhance the interpretation of developing ARGO-based ecosystem applications by providing more specific estimates of nutrient availability than temperature alone.