Ecosystem experiments   Email     Print   PDF

K. Popova, M. Steele, F. Dupont, D. Holland, T. Reddy, C. Hill, E. Hunke

Recognizing that marine ecosystem modeling is complex and that the ecosystems come in many forms, even in the Arctic Ocean environment, the AOMIP has decided to formulate a set on coordinated experiments to incorporate a relatively simple ecosystem modeling in their regional models of the Arctic Ocean. These experiments are important to our understanding of the changing Arctic marine environment. The arctic ecosystems are often highly complex and are affected by both cyclic and stochastic influences. Computer models, combined with suitable data-collection programs, can help in deepening our understanding of these systems and how they will react to various influences (from climatologic to human). The first-order proposed set up experiments are outlined below.

Participants:

• Katya Popova (NOCS, UK)
• Mike Steele (UW, USA)
• Frederick Dupont (BIO, Canada)
• David Holland (NYU, USA)
• Tasha Reddy (Calgary University, Canada & NYU, USA)
• Chris Hill (MIT)
• Elizabeth Hunke (LANL)

The proposed plan for intercomparison of ecosystem models is split into two phases. The first phase involves participants with fully coupled physical and biological models. The second phase is based on the results of the first one and aims at the whole AOMIP community.

The working group included representatives from six pan-Arctic or Global modeling projects with fully coupled ecosystems of various complexity (see separate table). All the participants acknowledged potential problems with intercomparison of different ecosystem models embedded into different physical models of different resolution forcing by different atmospheric fields.

In discussion it was decided that three physical factors were likely to play a disproportionate role in Arctic productivity, and that the collation and intercomparison of these with primary production should provide a focus for the research:

• Justification: winter mixing provides the main mechanism of the nutrient supply to the photic zone; combined with deep nutrient distribution (either from Levitus climatology or from model output) provides a good estimate of amount of nutrients available for the phytoplankton primary production
• Potential problems:  estimating maximum on the base of different average period of the standard output (e.g. monthly means vs 5 days means); difference in definitions of the UML depth between the models
• To be discussed: do we need to use unified definition of the UML depth?

• Justification: In some areas (mostly on the shelf break) upwelling provides additional significant source of nutrients
• Potential problems:  Not usually included into the standard model output; will require recalculation from horizontal velocity fields
• To be discussed: what is the optimal depth to analyse vertical velocity at (100m?)

• Justification: primary production is limited by light availability especially in the areas with permanent or seasonal ice cover. Integration over the period when UML is shallow will take care of the Nordic Seas where deep convection prevents Primary Production no matter how much of the short-wave radiation is available [Strictly speaking instead of 80m or any other fixed depth we should use a depth of the photic zone, however it is variable and a function of phytoplankton which seems to be an unnecessary complication]
• Potential problems:  ecosystem models are probably substantially different in their calculation of the photosinthetically active radiation in a grid cell with a fractional ice cover.
• To be discussed: is 80m (see above) a good estimate of the photic zone?

1. The working hypothesis of the Phase 1 is that 60-80% of the variability of the primary production can be explained by the variability in the three physical factors mentioned above. Thus we can have a constructive way forward for comparison of the various ecosystem models by comparing the relevant physics first. Then we can proceed by explaining the rest by difference in our ecosystem models or additional physical factors (e.g. horizontal nutrient transport).
2. Provided that (1) is correct all models can train a regression model using three 2D fields described above and Levitus nutrient climatology to estimate 2D annual mean primary production.
3. On the basis of model comparison with observations, the best model will be selected, and its regression will be used in phase two.

             Deadline1: participants are providing 3 physical fields identified above (UML, w, short-wave rad) as well as mean annual primary production. Additional fields of interest (to be discussed): grazing, f-ratio, Chlorophyll (or biomass), nutrients.

            Deadline 2: analysis of the fields and attempt at creating regression models

            Deadline 3: validation(*) of the models; selection of the best one (if at all possible)  to use its regression model in Phase 2.

(*)Model validation [was not discussed by the working group, please add your suggestions]

• UML depth climatology (monthly means)
• Satellite-derived Chl-a (monthly means)
• Satellite derived primary production (?) and synthesis by Carmack et al. (2006)
• Nutrient climatology

Phase 2 (Leading author: M.Steele, UW)

• which geographical areas are the most sensitive to the errors in the physical models
• how sensitive ecosystem model to the errors in the physical fields
• what level of ecosystem model complexity in Arctic is appropriate in the climate modelling

Phase 3 (was not discussed during the working group, leading author was not identified)