A comparison of global estimates of marine primary production from ocean color

https://doi.org/10.1016/j.dsr2.2006.01.028Get rights and content

Abstract

The third primary production algorithm round robin (PPARR3) compares output from 24 models that estimate depth-integrated primary production from satellite measurements of ocean color, as well as seven general circulation models (GCMs) coupled with ecosystem or biogeochemical models. Here we compare the global primary production fields corresponding to eight months of 1998 and 1999 as estimated from common input fields of photosynthetically-available radiation (PAR), sea-surface temperature (SST), mixed-layer depth, and chlorophyll concentration. We also quantify the sensitivity of the ocean-color-based models to perturbations in their input variables. The pair-wise correlation between ocean-color models was used to cluster them into groups or related output, which reflect the regions and environmental conditions under which they respond differently. The groups do not follow model complexity with regards to wavelength or depth dependence, though they are related to the manner in which temperature is used to parameterize photosynthesis. Global average PP varies by a factor of two between models. The models diverged the most for the Southern Ocean, SST under 10C, and chlorophyll concentration exceeding 1 mg Chl m-3. Based on the conditions under which the model results diverge most, we conclude that current ocean-color-based models are challenged by high-nutrient low-chlorophyll conditions, and extreme temperatures or chlorophyll concentrations. The GCM-based models predict comparable primary production to those based on ocean color: they estimate higher values in the Southern Ocean, at low SST, and in the equatorial band, while they estimate lower values in eutrophic regions (probably because the area of high chlorophyll concentrations is smaller in the GCMs). Further progress in primary production modeling requires improved understanding of the effect of temperature on photosynthesis and better parameterization of the maximum photosynthetic rate.

Introduction

Although photosynthesis is a key component of the global carbon cycle, its spatial and temporal variability is poorly constrained observationally. Furthermore it is unclear how this variability may respond to potential scenarios of climate change. Global net primary production, the carbon fixed through photosynthesis and available for higher trophic levels, occurs in both terrestrial (52%) and marine ecosystems (48%) (Field et al., 1998). The highly dynamic nature of marine photosynthesis is revealed by considering that the annual mean value of 45–50 Gt C is carried out by a phytoplankton biomass of 1Gt. Ship resources cannot resolve low-frequency spatial and temporal variability, much less make direct observations of mesoscale variability beyond isolated snapshots. The chronic undersampling of ship-based estimates of global primary production requires significant extrapolations, making it essentially impossible to quantify basin-scale variability from in situ measurements.

Fortunately, satellites provide a solution (McClain et al., 1998). Sensors that measure ocean color are presently used to estimate chlorophyll concentration in the upper ocean. Integrated biomass can be obtained from ocean color by assuming a vertical profile and a carbon to chlorophyll relationship. To go from biomass, a pool, to photosynthesis, a rate, a time dependent variable is needed. Solar radiation is an obvious choice, and simple mechanistic models compute productivity from biomass, photosynthetically available radiation (PAR), and a transfer or yield function which incorporates the physiological response of the measured chlorophyll to light, nutrients, temperature, and other environmental variables. As a variable amenable to remote sensing, sea-surface temperature (SST) is often used to parameterize the photosynthetic potential.

There exist a range of modeling approaches, e.g., Platt and Sathyendranath (1993), Longhurst et al. (1995), Howard and Yoder (1997), Antoine and Morel (1996), Behrenfeld and Falkowski (1997a), or Ondrusek et al. (2001). These models can be distinguished by the degree of explicit resolution in depth and irradiance as described by Behrenfeld and Falkowski (1997b). While vertically and spectrally explicit models incorporate information about algal physiology and its dependence on environmental factors, the paucity of measurements of physiological characteristics on the global scale hinders their full application. A common parameter in many simpler models is the maximum observed photosynthetic rate (normalized by biomass) within the water column (PoptB). Another parameter, PmaxB, is derived from short-term light-saturated incubations; consequently extant measurements are fewer. PmaxB is defined as the maximum rate of photosynthesis when light is not limiting, while PoptB represents the effective photoadaptive yield in the field for specific light conditions.

A series of round-robin experiments have been carried out to evaluate and compare models which estimate primary productivity from ocean color (Campbell et al., 2002). In these experiments, in situ measurements of carbon uptake were used to test the ability of the participating models to predict depth-integrated primary production (PP) based on information accessible via remote sensing. The first round-robin experiment used data from only 25 stations. The second primary production algorithm round robin (PPARR2) used data from 89 stations with wide geographic coverage (Campbell et al., 2002). There were 10 participant teams and 12 models.

Eight models were within a factor of 2.4 (based on one standard deviation in log-difference errors) of the 14C measurements (Campbell et al., 2002). Biases were a significant source of error. If biases were eliminated, 10 of the 12 model estimates would be within a factor of two of the in situ data. The algorithms performed best in the Atlantic region, which has historically contributed the most data for parameterization. The equatorial Pacific and the Southern Oceans presented the worst results. The Southern Ocean data included both the lowest and highest values of primary production, so the poor performance may be related to this dynamic range. The high-nutrient low-chlorophyll (HNLC) conditions observed in both the equatorial Pacific and the Southern Oceans may contribute to the higher model-data misfit, as most models were not developed with data subject to micronutrient limitation. Likewise, globally-tuned parameterizations of temperature and of the vertical extent of surface biomass are likely to fail in both regions.

The third primary production algorithm round robin (PPARR3) compares output from 24 ocean-color-based models and model variants from the US, Europe, Japan, and Brazil (Table 1). The first part of PPARR3 is a comparison of monthly global primary production fields generated by the different algorithms while part 2 is a sensitivity analysis. These two parts do not use in situ data to quantify model performance. Therefore, it is not possible to define a ‘best’ model. Part 3 is a ground-truth comparison like PPARR1 and PPARR2. We compare modeled PP and a high quality database of 14C measurements from the tropical Pacific (Le Borgne et al., 2002). The poor performance of the PPARR2 models in the tropical Pacific and the plentiful high-quality data led us to emphasize this region within PPARR3. An upcoming manuscript (Friedrichs et al., in prep) will present the results of part 3 of PPARR3 and recommend the best performing model for the tropical database comparison. A future study will look at a broader range of in situ data.

Circulation and nutrient fields are necessary to fully quantify oceanic carbon fluxes and biological productivity. In an effort to bring the ocean-color-based productivity modelers together with ecosystem and biogeochemical modelers, we invited the latter group to participate so we can compare their modeled primary production fields for the same time period with those of the ocean-color models. Our sole criterion for participation were that the models simulate global primary production fields.

In this paper, we describe PPARR3 results from parts 1 and 2, i.e. a global intercomparison of models for eight months and a sensitivity analysis. Although a comparison with in situ data is needed to quantify the performance of the models, the intercomparison enables us to discern the conditions under which the models have divergent results. By comparing the model output, we can distinguish groups, which in turn can be understood on the basis of the sensitivity analysis. Here we address the observed spatial, seasonal, and interannual variability among the participating models.

Section snippets

Participating models

The participating models are of all types discussed by Behrenfeld and Falkowski (1997b): wavelength- and depth-integrated (WIDI, 14 models), wavelength-integrated and depth-resolved (WIDR, five models), and wavelength- and depth-resolved (WRDR, five models). The list of models is given in Table 1, classified by model type, with the name of the participant(s), and the PPARR3 parts to which they have contributed. Seven general circulation models coupled to biogeochemistry (GCM-based) have

Relationships among models

In PPARR2, production estimated from ocean-color algorithms was found to be highly correlated and the correlation was independent of model complexity (Campbell et al., 2002). In an attempt to group the models on the basis of related output, here we estimated pair-wise correlation and root-mean-square (rms=(modeli-modelj2)/n) corresponding to the monthly global PP fields in January and July 1998 (Fig. 3). The correlation coefficient and rms between any pair of models is generally inversely

Implications of model similarity and divergence

Global PP estimates from the twenty-four ocean-color-based models range over a factor of two (Fig. 5A), from values less than 40 Gt C y-1 (models #14, 6, and 12) to those exceeding 60 Gt C y-1 (#15, 17, 13, and 21). HYR variants include the lowest and highest global estimates. VGPM variants tend to be average or low, HYR variants and WIDR can be high or low, and WRDR tend to be average or high. VGPM variants that use a variant of the sixth-order polynomial expression tend to be low because they

Acknowledgments

We thank Tony Lee of JPL and Stephen Yeager and Scott Doney (now at WHOI) of NCAR for generously providing modeled fields of mixed-layer depth. We are grateful to the SeaWiFS Science Project for the SeaWiFS data set and to the NOAA/NASA AVHRR Pathfinder project for SST, made available by the JPL-PODAAC. We thank two anonymous reviewers for helpful comments that improved an earlier version. Funding was provided by the NASA Ocean Biogeochemistry Program within the Carbon Cycle Science Program.

References (54)

  • D. Antoine et al.

    Oceanic primary production 2. Adaptation of a spectral light-photosynthesis model in view of application to satellite chlorophyll observations

    Global Biogeochemical Cycles

    (1996)
  • D. Antoine et al.

    Oceanic primary production 2. Estimation at global scale from satellite (Coastal Zone Color Scanner) chlorophyll

    Global Biogeochemical Cycles

    (1996)
  • K.R. Arrigo et al.

    Primary production in Southern Ocean waters

    Journal of Geophysical Research

    (1998)
  • Asanuma, I., Nieke, J., Matsumoto, K., Kawano, T., 2003. Optical properties control primary productivity model on the...
  • Aumont, O., Maier-Reimer, E., Blain, S., Pondaven, P., 2003. An ecosystem model of the global ocean including Fe, Si, P...
  • M. Behrenfeld et al.

    Photosynthetic rates derived from satellite-based chlorophyll concentration

    Limnology and Oceanography

    (1997)
  • M. Behrenfeld et al.

    A consumer's guide to phytoplankton primary productivity models

    Limnology and Oceanography

    (1997)
  • W.H. Berger

    Global maps of primary productivity

  • Bidigare, R.R., Prézelin, B.B., Smith, R.C, 1992. Bio-optical models and the problems of scaling. In: Primary...
  • Bopp, L., Kohfeld, K., Le Quéré, C, Aumont, O., 2003. Dust impact on marine biota and atmospheric CO2 during glacial...
  • Buitenhuis, E.T., Le Quéré, C., Aumont, O., Beaugrand, G., Bunker, A., Hirst, A., Ikeda, T., O’Brien, T., Piontkovski,...
  • Campbell, J., Antoine, D., Armstrong, R., Balch, W., Barber, R., Beherenfeld, M., Bidigare, R., Bishop, J., Carr,...
  • M.-E. Carr

    Estimation of potential productivity in Eastern Boundary Currents using remote sensing

    Deep-Sea Research II

    (2002)
  • J.E. Cloern et al.

    An empirical model of the phytoplankton chlorophyllicarbon ratio—the conversion factor between productivity and growth rate

    Limnology and Oceanography

    (1995)
  • H. Dierssen et al.

    Bio-optical properties and remote sensing ocean color algorithms for Antarctic Peninsula waters

    Journal of Geophysical Research

    (2000)
  • H. Dierssen et al.

    Optimizing models for remotely estimating primary production in Antarctic coastal waters

    Antarctic Science

    (2000)
  • R.W. Eppley

    Temperature and phytoplankton growth in the sea

    Fisheries Bulletin

    (1972)
  • Cited by (527)

    • Editorial Reflections on 70 years of Deep-Sea Research

      2023, Deep-Sea Research Part I: Oceanographic Research Papers
    View all citing articles on Scopus
    View full text