Skip to content

Because of the ubiquitous computer chip, it has become much easier to construct models of natural phenomena than to study them in situ. This is a growing problem for science because it leads to an over dependence on modeling and diminishes motivation for actually getting out into the big messy world. A pair of articles in Nature Geoscience, focused on the ocean’s nitrogen cycle, serve to underscore the problems that can arise when multiple models disagree with each other and with nature. More proof that computer models do not provide scientific evidence, just tales from the chip.

The ocean’s nitrogen budget has eluded accurate measurement. The nutrients nitrate (NO3) and phosphate (PO4) are key factors limiting the growth of marine biota, and thus the build-up of organic matter in the ocean. Cyanobacteria are key to this process because they capture nitrogen in a process known as nitrogen fixation. These photosynthetic bacteria, which can be found in both ocean and freshwater environments, convert elemental nitrogen (N2) into a biologically available form that other bacteria can consume. Those bacteria return fixed nitrogen to its elemental form in a process known as denitrification.

It is assumed that, with the exception of nitrogen fixation and denitrification, all biological processes consume nitrogen and phosphorus at a fixed ratio of 16:1. This is called the Redfield ratio. Unfortunately, the actual balance between nitrogen fixation and denitrification in the ocean is unknown. Determining the balance by direct measurement is a difficult task—measurements must be taken over vast distances and long periods of time. So naturally, scientists have turned to computer modeling.


Nitrification in the global nitrogen cycle.

A 2007 model analysis of surface nitrate-to-phosphate by ratios Curtis Deutsch et al. suggests that the Pacific contributes the largest share of global nitrogen fixation, driven by the upwelling of nitrogen-deficient waters (see “Spatial coupling of nitrogen inputs and losses in the ocean” in Nature). But analysis of excess nitrate in subsurface waters suggest that nitrogen fixation outweighs denitrification in the Atlantic Ocean and that denitrification dominates in the Pacific and Indian Oceans (see “Global patterns of marine nitrogen fixation and denitrification” by N. Gruber and J. L. Sarmiento in Global Biogeochemical Cycles). This is in line with other available rate measurements but disagrees with the Deutsch et al. model results.

To further muddy the waters, a new model study has reported that the relative abundance of fast- and slow-growing phytoplankton could determine the amount of elemental nitrogen fixed in the Pacific Ocean. Matthew M. Mills and Kevin R. Arrigo, both from the Department of Environmental Earth System Science at Stanford University, suggest that a small shift in the nitrate-to-phosphate uptake ratio of phytoplankton has a large effect on calculated nitrogen fixation rates. Their motivation to develop a new model is described in “Magnitude of oceanic nitrogen fixation influenced by the nutrient uptake ratio of phytoplankton,” appearing in the may 16, 2010, eddition of Nature Geoscience:

Despite the prevalence of non-Redfield nutrient utilization, few ecosystem models, and virtually no geochemical or ocean general circulation models that include biological processes, consider non-Redfield nutrient consumption, assuming instead that phytoplankton N/P stoichiometry is Redfield (diazotrophs notwithstanding). Considering the predominance of marine environments that support non-Redfield N/P utilization by non-diazotrophic phytoplankton, a better description of how phytoplankton stoichiometry affects oceanic nutrient inventories is critical to understanding the marine N cycle.

In response to this challenge, Mills and Arrigo developed a dynamic five-box ecosystem model with three phytoplankton groups: diatoms, picocyanobacteria and diazotrophs. The model includes “flow between compartments representing the deep ocean, the nearshore subsurface oxygen minimum zone (OMZ) where fixed N is lost through microbial processes, the coastal upwelling zone dominated by diatoms with low N/P requirements, transitional waters between the coastal upwelling and the oligotrophic ocean, and the oligotrophic gyre dominated by picocyanobacteria (for example, Prochlorococcus) with high N/P requirements.” A functional diagram of the model is shown below.

Using different sets of assumptions about phytoplankton N/P ratios, multiple simulations were run. Changes in N and P inventories during yearly cycles were simulated in each box. N2 fixation and the fraction of the N requirement satisfied by that fixation was calculated from diazotroph growth. The amount of N2 fixation required in each box to restore PO4concentrations to the observed World Ocean Atlas (WOA) 2005 levels was also calculated by the model. In the end, results from the various simulations were compared.

From the modeling results, Mills and Arrigo concluded that the prevailing paradigm that the marine N inventory is controlled by a simple feedback between global rates of nitrification and denitrification is incorrect. “Our results indicate that non-Redfield nutrient utilization by non-diazotrophic phytoplankton represents a mechanism that can decouple the processes of N2 fixation and denitrification, and consequently, the steady-state oceanic N inventory,” they state at the end of their report. “Furthermore, the reduction in xsPO4 by diatoms and the corresponding decrease in downstream N2fixation may help explain the apparent imbalance between rates of N2 fixation and fixed-N loss (which is substantially larger) in the contemporary ocean.”


Distribution of excess phosphate in the surface of the Pacific Ocean.

What we have here are two different models that yield two different explanations for the mismatch between predicted and observed phosphate levels in the surface waters of the Pacific Ocean: Deutsch et al.’s proposed upwelling of nitrogen-deficient waters coupled with consumption of excess phosphate by nitrogen fixers, which require no bioavailable nitrogen; and Mills and Arrigo’s non-Redfield uptake of nitrogen and phosphorus, which varies the nitrogen and phosphorus requirements during the growth cycle of non-nitrogen fixers. Which is correct, if either? That very question was addressed by Wolfgang Koeve and Paul Kähler in a view point letter to Nature Geoscience, “Ocean science: Balancing ocean nitrogen.” Here is their take on the situation:

[W]hich model scenario matches reality is uncertain. A comparison of model results with nitrogen fixation measurements could provide the answer. Unfortunately, most measurements of nitrogen fixation have been confined to the Atlantic Ocean, where Trichodesmium—once believed to be the globally dominant species of nitrogen-fixing phytoplankton—was known or expected to occur, leaving the Pacific undersampled. Although remote sensing measurements provide information on the global distribution of Trichodesmium, the data cannot be easily reconciled with the two model scenarios or the distribution of dust inputs from the atmosphere, thought to supply iron for which nitrogen fixers are known to have high requirements.

As I said: Two models, two answers, neither in full agreement with observation. Now recall Mills and Arrigo’s statement that virtually no geochemical or ocean general circulation models that even include biological processes are correct. The general circulation models (GCM) used by the IPCC and others to predict climate change have at their hearts two different models: an atmospheric circulation model and an ocean circulation model. The nitrogen cycle has significant impact on marine organisms and through them the absorption or emission of CO2 by the ocean. Here is blunt evidence that all of the ocean models in use are incorrect when it comes to biological processes. Furthermore, there is no known correct model to replace them with.

In a soon to be published PNAS article, two mathematicians have tried to address the errors inherent in climate modeling. After an analysis using empirical information theory, Andrew J. Majda and Boris Gershgorin concluded that “there are intrinsic model errors in the AOS models for the climate system.” They attribute this intrinsic error to fundamentally incomplete knowledge of the climate system and limited observational data. They put it this way:

The central difficulty in climate change science is that the dynamical equations for the actual climate are unknown. All that is available from the true climate in nature are some coarse-grained observations of functions such as mean or variance of temperature, tracer greenhouse gases such as carbon dioxide, or the large scale horizontal winds. Thus, climate change science must cope with predicting the coarse-grained dynamic changes of an extremely complex system only partially observed from a suite of imperfect models for the climate.

And yet, we are asked to gamble humanity’s future on the output from IPCC climate models. There is a further warming at the end of Koeve and Kähler’s commentary on the two models presented above: “Their work reminds us that models are tools to test hypotheses and show their consequences, but need to be checked by independent field data. They cannot be viewed as evidence.”

Real scientists know that model output is not proof of how the real world works. Such output are just numbers spun by computer code, and the interpretation of those numbers by climate change alarmists merely tales from the chip. Climate scientists know they have no proof of future global warming, only model projections—and that means global warming is not really science at all.

Be safe, enjoy the interglacial and stay skeptical.

The Resilient Earth, 12 August 2010