Optimized multi-satellite merger of primary production estimates in the California Current using inherent optical properties

Building a multi-decadal time series of large-scale estimates of net primary production (NPP) requires merging data from multiple ocean color satellites. The primary product of ocean color sensors is spectral remote sensing reflectance (Rrs). We found significant differences (13-18% median absolute percent error) between Rrs estimates at 443 nm of different satellite sensors. These differences in Rrs are transferred to inherent optical properties and further on to estimates of NPP. We estimated NPP for the California Current region from three ocean color sensors (SeaWiFS, MODIS-Aqua and MERIS) using a regionally optimized absorption based primary production model (Aph-PP) of Lee et al. (2011). Optimization of the Aph-PP model was required for each individual satellite sensor in order to make NPP estimates from different sensors compatible with each other. While the concept of Aph-PP has advantages over traditional chlorophyll-based NPP models, in practical application even the optimized Aph-PP model explained less than 60% of the total variance in NPP which is similar to other NPP algorithms. Uncertainties in satellite Rrs estimates as well as uncertainties in parameters representing phytoplankton depth distribution and physiology are likely to be limiting our current capability to accurately estimate NPP from space. Introducing a generic vertical profile for phytoplankton improved slightly the skill of the Aph-PP model.