A comparison between phytoplankton community structures derived from a global 3D ecosystem model and satellite observation
Introduction
Due to the geographically large spatial scale of the oceans and the inter-disciplinary nature of marine ecosystems, there is an inherent necessity for an integrated analysis using in situ and satellite observations and modeling in order to understand, represent and forecast processes involved in the global marine ecosystems. In such a circumstance, one of the strategies to achieve a scientific objective may be to scale up in situ observation to satellite observation in order to increase a spatial coverage and temporal resolution of the observation, and then apply them to modeling (e.g. re-analysis), which is expected to give a better understanding of a temporally continuous three-dimensional structure of the marine ecosystems and its variability in the future environment.
A large amount of global in situ data of phytoplankton, i.e. primary producers in the marine ecosystems, has been obtained through pigment biomass measurements. Taking advantage of the availability of the global data, O'Reilly et al. (1998) scaled up the in situ observations of the pigment biomass to satellite observations, which enormously improved the spatial coverage and temporal frequency of data sampling (approximately 8 days of observations covers the majority of the globe). Furthermore, the global satellite observations of phytoplankton have been extended to obtain fractional contributions of different phytoplankton functional types (PFTs) to the pigment biomass for a better understanding of their biogeochemical and ecological functions in the oceans (e.g. Bracher et al., 2009, Brewin et al., 2010, Devred et al., 2006, Kostadinov et al., 2010, Hirata et al., 2011, Uitz et al., 2006).
In parallel, marine biogeochemical–ecosystem models, which explicitly represent physiology of marine organisms and ecosystem processes as well as emergent properties such as a composition of the phytoplankton community, have also been advanced: e.g. NEMURO (Aita et al., 2007, Hashioka et al., 2009), PlankTOM-5 (Le Quéré et al., 2005, Le Quéré and Pesant, 2008), BEC (Doney et al., 2001, Moore et al., 2002), PISCES (Aumont and Bopp, 2006), MEM-OU (Shigemitsu et al., submitted for publication). These biogeochemical–ecosystem models are coupled with the ocean general circulation models (OGCM), and are able to compute a realistic distribution of the PFTs. Thus, it is now possible to compare, not only Chlorophyll-a pigment biomass of the total phytoplankton community but also fractional pigment biomass of each PFT derived from the modeling and the satellite observation, towards applications to a global analysis of detailed marine ecosystems.
Comparisons between modeling and observations have been done in terms of primary productivity, chlorophyll-a, chemical species etc. (e.g. Doney et al., 2009, Friedrichs et al., 2009). In these analyses, standard univariate metrics such as correlation coefficient, the root mean square error, bias etc. have been used to assess a model skill. While these are useful metrics for grid-by-grid comparison for a time series, spatial scales at which the PFTs derived from the model and the observations agree at a certain time step are not addressed. Vichi et al. (2011) used a multi-dimensional scaling approach to compare biogeochemical provinces and spatial distribution, which can be applied to the PFTs too. However, the spatial scales agreed between model and observation was not quantified. As a result, the quantification of “spatial pattern matching” tends to remain unresolved in marine ecosystem model–observation comparisons, and an assessment of a spatial similarity between model and observation still relies on visual and subjective comparison. Here, we attempt to bridge the univariate and pattern comparisons using the PFTs derived from modeling and satellite observation to quantify the spatial similarity. We first derive conventional univariate statistical metrics. In addition, the spatial Principal Component Analysis (PCA) is performed to find major variability in spatial pattern of the PFTs and then a two-dimensional wavelet method is used to quantify the spatial scale(s) at which the modeled PFTs agree to the satellite PFTs, to see that our analysis presented here provide a novel and significant value on conventional comparisons between the model and satellite observation.
Section snippets
Model data
A new Marine Ecosystem Model (called MEM-OU) based on a marine biogeochemical model has been developed (Shigemitsu et al., submitted for publication) with the aim of a better representation of biogeochemical dependencies and phytoplankton physiology. The MEM-OU includes silica and dissolved and particulate iron cycles (Fig. 1), notably incorporating the optimum nutrient uptake kinetics (Smith and Yamanaka, 2007). Phytoplankton is represented by two phytoplankton types in the MEM-OU; large
Methodology
Spatial distributions of pigment biomass of PL and PS derived from the model and the satellite, as well as of the sum of PL and PS (= Tchla), are compared by bias assessment, correlation statistics, the principal component analysis (PCA) and the two-dimensional Harr wavelet analysis (2D-WA). Due to a limitation in the satellite observation at high latitudes in winter, only 47°S to 47°N is considered to match up grids between model and satellite data for the comparison.
Since our primary interest
Climatological mean
Fig. 3 shows a climatological pigment biomass of Tchla, PL and PS over 1998–2007 derived from a satellite observation and model hindcast. The satellite observation exhibits relatively high pigment biomass at higher latitudes for Tchla, PL and PS (> 0.5 mg m− 3) but relatively low in the subtropical gyres (< 0.2 mg m− 3), which is consistent with a decadal in situ observation (Aiken et al., 2009). In the satellite Tchla and PS, the Equatorial Pacific also show an elevated pigment biomass (> 0.2 mg m− 3) due
Discussion
Using the standardized data derived from the model and the satellite, the wavelet analysis presented here quantitatively identified spatial scales where the model has skill to represent spatial distributions of Tchla, PL and PS. The information lost during the standardization (e.g. such as the bias), has to be assessed by other means (see Fig. 4). However a spatial distribution of the bias in relative unit is still considered in the wavelet analysis. Thus, the wavelet analysis does not provide
Summary and conclusions
The marine ecosystem model, MEM-OU, was applied to a three dimensional scale by coupling an OGCM, and climatological fields of the PFTs obtained from the model were compared with those from the satellite observation. Spatial distribution of phytoplankton communities generally agree between model and satellite except for the equatorial ocean where relatively larger bias was found. Nonetheless, the model clearly captured a dominant distribution of PFTs at the global scale across the oceans. The
Acknowledgment
This work was financially supported by the JSPS Institutional Program for Young Researcher Overseas Visits, Grant-in-Aid for Young Scientist (Start-up) and JAXA Global Observation Mission - Climate project. A development of the ecosystem model was funded by the Japan Science Technology Agency (JST) under Core Research for Evolution Science and Technology program (CREST).
References (32)
- et al.
Phytoplankton pigments and functional types in the Atlantic Ocean: a decadal assessment, 1995–2005
Deep Sea Res. II
(2009) - et al.
Interdecadal variation of the lower trophic ecosystem in the northern Pacific between 1948 and 2002, in a 3-D implementation of the NEMURO model
Ecol. Model.
(2007) - et al.
Marine ecosystem models for earth systems applications: the MarQUEST experience
J. Mar. Syst.
(2010) - et al.
A three-component model of phytoplankton size class for the Atlantic Ocean
Ecol. Model.
(2010) - et al.
Skill metrics for confronting global upper ocean ecosystem‐biogeochemistry models against field and remote sensing data
J. Mar. Syst.
(2009) - et al.
Assessing the uncertainties of model estimates of primary productivity in the Tropical Pacific Ocean
J. Mar. Syst.
(2009) Fast spin up of ocean biogeochemical models using matrix-free Newton–Krylov
Ocean Model.
(2008)- et al.
An objective methodology for the classification of ecological pattern into biomes and provinces for the pelagic ocean
Remote. Sens. Environ.
(2008) - et al.
Globalizing results from ocean in situ iron fertilization studies
Global Biogeochem. Cycles
(2006) - et al.
Quantitative observation of cyanobacteria and diatoms from space using PhytoDOAS on SCIMACHY data
Biogeosciences
(2009)
A new intensity-scale approach for the verification of spatial precipitation forecasts
Meteorol. Appl.
A two-component model of phytoplankton absorption in the open ocean: theory and applications
J. Geophys. Res.
Marine biogeochemical modeling
Oceanography
NCEMP/NCAR 40-year reanalysis project
Bull. Amer. Meteor. Soc.
Global variability of phytoplankton functional types from space: assessment via the particle size distribution
Biogeosciences
K-1 coupled model (MIROC) description
Cited by (14)
Modelling size-fractionated primary production in the Atlantic Ocean from remote sensing
2017, Progress in OceanographyCitation Excerpt :High rates of primary production from large cells are observed in spring periods in each hemisphere and in upwelling regions such as the Benguela (Hirata et al., 2009). Output from size-fractionated primary production models, such as that illustrated in Fig. 16, has applications for multi-phytoplankton biogeochemical model evaluation (Ward et al., 2012; Hirata et al., 2013; de Mora et al., 2016), and may even be useful in a data assimilation scheme, to improve simulations of biogeochemical rates (Xiao and Friedrichs, 2014). Our approach to modelling size-fractionated primary production is based on an established and proven primary production model (Platt et al., 1980).
Influence of light in the mixed-layer on the parameters of a three-component model of phytoplankton size class
2015, Remote Sensing of EnvironmentCitation Excerpt :Observations on phytoplankton size, well resolved in time and space, are required to address key questions pertaining to our understanding of marine biogeochemisty and marine ecology, and how they are likely to be influenced by climate variability (Brewin et al., 2012; Finkel et al., 2010; Kostadinov, Siegel, & Maritorena, 2010; Siegel et al., 2014). These observations may be used directly or indirectly to address such questions, through the validation of, or assimilation into, size-based marine biogeochemistry models (Hirata et al., 2013; Holt et al., 2014; Ward et al., 2012; Xiao & Friedrichs, 2014). A common method to infer phytoplankton size structure in situ is through measurements of phytoplankton biomass (typically indexed through the chlorophyll concentration, the principal photosynthetic pigment in phytoplankton), partitioned into key size classes (Sheldon, 1972; Sieburth, Smetacek, & Lenz, 1978; Uitz, Claustre, Morel, & Hooker, 2006).
Comparison of two methods to derive the size-structure of natural populations of phytoplankton
2013, Deep-Sea Research Part I: Oceanographic Research PapersCitation Excerpt :Size-fractionated chlorophyll approaches complement ocean colour remote sensing, considering spectral variations in ocean colour are driven by changes in chlorophyll in open-ocean waters (Morel and Prieur, 1977), and methods have been developed to estimate size-fractionated chlorophyll and size-fractionated primary production using satellite observations (Claustre et al., 2005; Silió-Calzada et al., 2008; Uitz et al., 2006, 2008, 2009, 2010; Hirata et al., 2009; Brewin et al., 2010; Hirata et al., 2011). Measurements of size-fractionated chlorophyll have also been used to estimate export production (Guidi et al., 2009) and used for the validation of multi-phytoplankton biogeochemical models (Ward et al., 2012; Hirata et al., 2013). Two frequently used approaches to estimate the biomass of different size-classes of phytoplankton in situ are size-fractionated filtration (SFF) and High Performance Liquid Chromatography (HPLC) analysis of pigments.
Factors Regulating the Relationship Between Total and Size-Fractionated Chlorophyll-a in Coastal Waters of the Red Sea
2019, Frontiers in Microbiology