Estimation of primary production in the southern argentine continental shelf and shelf-break regions using field and remote sensing data

Abstract

The Argentine continental shelf and shelf-break regions comprise a large and rich biological area of the ocean. However, field estimations of primary production are scarce, making remote sensing of ocean color a valuable tool to provide synoptic maps of primary production in this ecologically relevant region. Field studies performed during spring 2005, and summer and winter 2006 showed a high spatial and seasonal variability in the daily integrated water column primary production, chlorophyll-a and biomass-normalized photosynthetic parameters. Using field measurements, five different and relatively simple (non-spectral and vertically homogeneous biomass) models were tested: three chlorophyll-, one carbon- and one absorption-based model. The chlorophyll-based &lsquoBIOM&rsquo model developed by Platt and Sathyendranath (Science, 241:1613&ndash1620, 1988) provided the closest estimates to the field values, and was selected as the local algorithm. Its performance was assessed using simultaneous satellite-derived products and field photosynthetic parameters as input. Close values compared to the field estimates were obtained using BIOM (Absolute Percent Difference error, APD ~ 10%), even though satellite-derived products used as input to the model (i.e. chlorophyll-a concentration, diffuse attenuation coefficient in the photosynthetically active radiation range &mdash PAR-, and PAR irradiance) showed relative high errors (APD ~ 40%, 20% and 50%, respectively). Provided that an efficient way to assign the physiological parameters in a pixel-by-pixel basis is found, this model seems to be the best to produce primary production maps from remote sensing of ocean color in the southern Argentine shelf and shelf-break regions. Highlights •Five models to estimate primary production in the Argentine Sea were tested. •A non-spectral and uniform biomass profile model (BIOM) provided the best estimates. •Good performance was found when applied to MODIS data and field P-I parameters. •Errors in MODIS Chla, Kd(PAR), and PAR products were 40%, 20% and 50%, respectively.