This story, as I have told it to this point, is a straightforward account of a recent advance in our understanding of how the oceans work. But there is a very rich subtext accompanying this tale---with lessons for scientists, the agencies that fund them, and the public officials who rely on them for setting public policy.
The part of the story that I have left out has its origins in
paleooceanography. By measuring the composition of materials
trapped in ice cores from the Antarctic, including the CO
content of ancient air, paleooceanographers can deduce the
chemistry of the oceans and atmosphere of past eras. These studies
have revealed that during the last glacial maximum, 18,000 years
B.P. (before present), CO
levels in the atmosphere were only
200 ppm. This contrasts with today's (pre-industrial) atmospheric
concentrations of 280 ppm, and previous interglacial atmospheres
(Raynaud et al. 1993) which had similar concentrations. Although
most people agree that the deep ocean CO
reservoir must have
played a role in glacial/interglacial transitions in atmospheric
CO
, the mechanism is unclear (Sarmiento and Bender, 1994).
One of the leading theories suggests that changes in primary productivity in the Southern ocean played a key role (Knox and McElroy, 1984; Sarmiento and Toggweiler, 1984; Siegenthaler and Wenk, 1984), but explanations for what triggers the changes in productivity have been lacking (Keir, 1988; Toggweiler and Sarmiento, 1985).
These observations were not lost on Martin as he was
revitalizing the iron hypothesis. Although the Southern ocean of
today receives one of the lowest dust inputs in the world, this was
not always the case (Martin, 1990a, 1992). During the last glacial
maximum, for example, arid land regions were 5 times larger than
today, and 50 times today's values of wind-borne iron was deposited
on waters around the Antarctic continent (Petit et al., 1981; De
Angelis et al., 1987). Martin hypothesized that increased supplies
of iron to the Southern ocean (where nitrates and phosphates are
present in excess) during the last glacial maximum stimulated the
biological pump and contributed to the drawdown of atmospheric
CO
to 200 ppm (Martin, 1990a). Indeed, when one looks at
the relationship between iron concentrations in Antarctic ice cores
and the corresponding atmospheric CO
concentrations, the
inverse relationship is striking (Fig. 8). Martin also reminded us
that this change in CO
concentration could have played
a role in the changing temperatures during the interglacial/glacial
transition (Shafter, 1989).
Although the suggested causality chain---iron availability
changeproductivity change 
CO
change
climate change---was, and still is, viewed as
oversimplified at best (Boyle, 1988; Sarmiento and Bender, 1994),
Martin's ideas were evolving in an atmosphere of public concern and
scientific debate about the threat of greenhouse warming from
anthropogenically-elevated CO
concentrations. As such, the
idea had a certain appeal. Noting that the difference between
today's concentrations (355 ppm) and pre-industrial concentrations
(280 ppm) is of the same order as the difference between the latter
and those during the glacial maximum (200 ppm), the obvious next
step was to suggest that if one could fertilize the Southern ocean
with iron, one could cool the Earth. Few would have had the
chutzpah to suggest this publicly, but Martin---who enjoyed nothing
more than a good-humored debate---did. At a 1988 seminar at the
Woods Hole Oceanographic Institution he suggested that fertilizing
the Southern ocean with 300,000 tons of iron could remove 2 billion
tons of CO
from the atmosphere. ``Give me a tankerload of
iron,'' he said, ``and I'll give you an ice age.'' (Martin, 1990b).
This more or less facetious remark triggered a series of events
(see Chisholm and Morel, 1991 for a detailed account) that
ultimately led to the iron fertilization experiment described
above. The so-called ``geritol solution'' to global warming was
rapidly picked up by the scientific grapevine and the press, and
was the focus of a National Academy workshop in late 1989. Soon
thereafter Martin casually slipped the idea into the scientific
literature (Martin, 1990a; Martin et al., 1990).
At this point a significant fraction of the scientists in the field was becoming somewhat unnerved by how much attention the ``geritol solution'' was getting in the face of very little evidence that it would work (and a consensus opinion that even if it did work, it was not a wise way to go). In response, the American Society of Limnology and Oceanography (``ASLO'') organized a Special Symposium in Feb. 1991 charged with gathering together experts on the marine food web to discuss in detail what was known about the regulation of productivity in high-nutrient-low-chlorophyll areas such as the Southern ocean. The proceedings of the symposium, published as a special issue of Limnology and Oceanography (Chisholm and Morel, 1991), reveal the true complexity of these ecosystems, and the difficulties involved in arriving at a consensus about what regulates them. The group concluded that the evidence for iron playing a role in regulating productivity in the HNLC areas of the oceans was compelling enough to warrant a small-scale fertilization experiment. They also drafted and adopted a resolution urging ``...all governments to regard the role of iron in marine productivity as an area for further research and not to consider iron fertilization as a policy option that significantly changes the need to reduce emissions of carbon dioxide.''
Thus the stage was set for what has come to be called the ``IRONEX'' experiment, the results of which I have summarized above. Martin had to secure the funding for the expedition, and there is no doubt that having the endorsement of the broader oceanographic community smoothed the way for this. Also, since the idea of an ocean fertilization experiment was no longer enmeshed in the more loaded issue of climate engineering, it could develop independently of the potential consequences of its success.
Looking back, however, we cannot help but wonder if this experiment would have come about had John Martin not thrown the climate change question into the public eye. His casting the hypothesis in a context relevant to public concerns catalyzed the scientific community---not the public---to take careful stock of our understanding of the regulation of open ocean ecosystems. This drew the attention of the federal funding agencies (who were involved in the ASLO Symposium), giving them an opportunity to witness the broadest possible ``peer review'' before endorsing the experiment. The experiment was a success beyond everyone's expectations (except, perhaps, John Martin's), and the issue of engineering climate has receded from view, for now. I am sure the debate will be rekindled sometime in the future, but by then our understanding of how the marine food web is regulated will be more advanced---through insights gained from this, and other, ocean fertilization experiments.
Acknowledgments. This paper is dedicated to the late John Martin, whom I thank for his friendship, inspiration, and wonderful good humor throughout all the struggles. I thank Sara Tanner and Steve Fitzwater for sampling help. The bottle experiments reported from the IRONEX cruise were designed and orchestrated by Kenneth Coale, and the pigment data were supplied by Robert Bidigare. Thanks also to Zackary Johnson and Kent Bares for help with the data analysis and graphics, and Edmund Carlevale for editorial advice. This effort was supported in part by grants from NSF, ONR, and the EPA Global Change Research Program (IN-0826-NAEX).