Inter-comparison of phytoplankton functional type phenology metrics derived from ocean color algorithms and Earth System Models

Abstract

Ocean color remote sensing of chlorophyll concentration has revolutionized our understanding of the biology of the oceans. However, a comprehensive understanding of the structure and function of oceanic ecosystems requires the characterization of the spatio-temporal variability of various phytoplankton functional types (PFTs), which have differing biogeochemical roles. Thus, recent bio-optical algorithm developments have focused on retrieval of various PFTs. It is important to validate and inter-compare the existing PFT algorithms; however direct comparison of retrieved variables is non-trivial because in those algorithms PFTs are defined differently. Thus, it is more plausible and potentially more informative to focus on emergent properties of PFTs, such as phenology. Furthermore, ocean color satellite PFT data sets can play a pivotal role in informing and/or validating the biogeochemical routines of Earth System Models. Here, the phenological characteristics of 10 PFT satellite algorithms and 7 latest-generation climate models from the Coupled Model Inter-comparison Project (CMIP5) are inter-compared as part of the International Satellite PFT Algorithm Inter-comparison Project. The comparison is based on monthly satellite data (mostly SeaWiFS) for the 2003–2007 period. The phenological analysis is based on the fraction of microplankton or a similar variable for the satellite algorithms and on the carbon biomass due to diatoms for the climate models. The seasonal cycle is estimated on a per-pixel basis as a sum of sinusoidal harmonics, derived from the Discrete Fourier Transform of the variable time series. Peak analysis is then applied to the estimated seasonal signal and the following phenological parameters are quantified for each satellite algorithm and climate model: seasonal amplitude, percent seasonal variance, month of maximum, and bloom duration. Secondary/double blooms occur in many areas and are also quantified. The algorithms and the models are quantitatively compared based on these emergent phenological parameters. Results indicate that while algorithms agree to a first order on a global scale, large differences among them exist; differences are analyzed in detail for two Longhurst regions in the North Atlantic: North Atlantic Drift Region (NADR) and North Atlantic Subtropical Gyre West (NASW). Seasonal cycles explain the most variance in zonal bands in the seasonally-stratified subtropics at about 30° latitude in the satellite PFT data. The CMIP5 models do not reproduce this pattern, exhibiting higher seasonality in mid and high-latitudes and generally much more spatially homogeneous patterns in phenological indices compared to satellite data. Satellite data indicate a complex structure of double blooms in the Equatorial region and mid-latitudes, and single blooms on the poleward edges of the subtropical gyres. In contrast, the CMIP5 models show single annual blooms over most of the ocean except for the Equatorial band and Arabian Sea.