University of California Museum of Paleontology UCMP in the field See the world (and its fossils) with UCMP's field notes.
About UCMP People Blog Online Exhibits Public programs Education Collections Research
Education : Graduate programs : Louderback fund

The revolution in paleoclimatology
Daniel Karner's 2003 Louderback luncheon presentation

When Doris Sloan invited me to give a talk, I asked her what I should talk about. She said, "Talk about 15 minutes." So I will. Today I will talk about a revolution in the field of paleoclimatology, a discipline that is dominated by paleontologists and geologists alike. Hopefully, this means that you will all find something useful in what I have to say.

It pleases me to report that, while I was originally asked to give this talk at last year's Louderback luncheon, the revolution in paleoclimatology has continued through the last year, adding credence to the notion that this revolution is real, and is not just some failed coup d'etat.

The revolution that I am talking about has to do with the way in which we assign ages to past geologic events. While most fields of Earth science rely on the use of radioisotopic dating methods to assign ages, paleoclimatology has largely strayed from this type of dating method because paleoclimatologists thought that they had discovered something better — a dating method based on astronomy, which has virtually no calculation error. This has led to great consternation among geochronologists, as increasing the precision and accuracy of radioisotopic dating is something that geochronologists have been laboring to do for decades … they still aren't satisfied yet. Paul Renne can testify at great length to that last statement; few have exerted as much effort as he in improving the 40Ar/39Ar dating method. What if all those efforts to improve radioisotopic dating had been for naught, and the field of geochronology is abandoned for something called astrochronology? Many, or dare I say most, paleoclimatologists believe that this should happen. But despite the difficulty in improving their dating methods, the geochronologists are finally having their day. The advances that they have made in radioisotopic dating techniques have led to the discovery that something is fundamentally wrong with what the paleoclimatologists are doing.

So, let me describe briefly what has transpired in the field of paleoclimatology. A central problem in paleoclimatology is finding the cause of the Ice Ages. Indeed, it has been known for over two hundred years that great ice sheets, in some places over 3 km thick, once covered vast expanses on the Northern Hemisphere, and it's been over one hundred years since it was recognized that these great ice sheets grew and melted away many times. The cyclical nature of the advances and retreats of the great ice sheets were viewed even then as a natural extension of what was known about climate change today — that most of climate is driven by changes in Earth's orbit around the Sun. Indeed, the most logical mechanism to cause ice ages is to decrease the amount of sunlight that the Earth receives, particularly at high latitudes where the glaciers formed. Despite the elegance of this idea, it has fallen into disfavor at different times during the last century, only to be revitalized later in a somewhat different form.

In recent decades, the lion's share of ice age research has been done using deep-sea sediments, or more specifically, the oxygen isotopes found in calcium-carbonate tests of foraminifera, to evaluate global ice volume back through time. The logic is as follows: evaporation from the oceans preferentially removes water containing light oxygen and hydrogen isotopes, and transports this isotopically light water over land to form the ice sheets. As ice accumulates on the continents, the ocean water becomes relatively enriched in the heavier isotopes of oxygen and hydrogen, and the foraminifera secrete their calcium-carbonate tests in equilibrium with the ocean water. So as the ice sheets grow, the isotopes in the foraminifera tests become heavier. By measuring the oxygen isotopes of a particular species of foraminifera test in deeper and deeper layers of the ocean sediment, we can construct a history of ice ages versus depth in the sediment column. The big trick, and this is central to revolution, is how we transform that depth scale into something more meaningful — an age scale.

The standard way in which paleoclimatologists assign ages to their sediment cores is by correlating their records of climate change to predictions originally made in the early 20th Century by a Serbian mathematician named Muletin Milankovitch. The Milankovitch theory, as it is now widely called, claims that ice age cycles are being controlled by variations in the Earth's orbit, which lead to small changes in the amount and distribution of sunlight on the Earth. The Milankovitch theory predicted that warm Northern Hemisphere summers, which occur about every 23,000 years, would melt the great Northern Hemisphere ice sheets, and would keep new ones from forming. Subsequently it was discovered that the fundamental period of ice age cycles is not 23,000 years, but instead is about 100,000 years, and so it was necessary for John Imbrie at Brown University to develop a complex model that could salvage the Milankovitch theory in light of this new information. Indeed, the Milankovitch theory has necessarily undergone many revisions since its inception in order to be made consistent with new data. Today, a fairer definition that describes this theory might be something like, "Earth's orbital variations are somehow preserved in the geologic record." By all accounts this is a far cry from a predictive tool that originally was designed to explain the Ice Ages. It is my opinion that the present use of the term "Milankovitch theory" has strayed so far from its original definition that its meaning has become confusing to those inside and outside of the paleoclimatology community. I think that its present incarnation would be better described by a more generic term, like "orbital theory," which is purposely vague to allow for further revision of how orbital variations affect ice ages or other aspects of climate.

There were tremendous fringe benefits for those who believed in the Milankovitch theory. Because there is no appreciable friction in space, the timing of Earth's orbital variations can and have been calculated with a high degree of accuracy and precision back through the past several million years, and as such the paleoclimatologists could use this precise time scale of orbital variations to date their paleoclimate records. Indeed, if done correctly, and if these climate events truly were caused by Earth's orbital variations, and with a known phasing between forcing and response, then this approach would offer paleoclimatologists a dating method that has virtually no time scale error. Until relatively recently, radioisotopic dates for ice age cycle events were not good enough to test the validity of the way in which scientists were correlating the orbital cycles with ice ages.

Nonetheless, the ice age timing predictions had been made. The "validation" of correlating ice ages with the Milankovitch theory came in 1984 with publication of the SPECMAP stack. SPECMAP was constructed from oxygen isotope data collected from five deep-sea sediment cores. By averaging together the isotope data from all these cores, and then correlating as many features as possible with the calculated sunlight curve for the last 800,000 years the inventors of the SPECMAP stack felt that they had created an accurate estimate, both in amplitude and in time, of global ice volume. Indeed, even today most paleoclimate studies use the SPECMAP stack as a reference against which they compare their work. In some instances, they even use SPECMAP as a basis for discounting their own results or the results of others.

But in hindsight, the SPECMAP stack had serious flaws, and so should only have been considered a working hypothesis. The isotopic data used to create the SPECMAP stack were from planktonic foraminifera. We now know that planktonic foraminifera record many regional features of climate, such as the migration of monsoons and changes in sea surface temperature, which obscure the global ice volume signature. More recently, radioisotopic dating has improved to the point of being able to test the validity of the SPECMAP time scale, and has been able to show that the SPECMAP chronology was wrong.

The revolution in paleoclimatology had one of its most important battles in 1992, and I do not use the word "battle" lightly here. In 1992, several members of the US Geological Survey, including Ken Ludwig who now is at the Berkeley Geochronology Center, reported that the end of the second to last ice age at around 135,000 years ago, occurred about 10,000 years earlier than was predicted by the Milankovitch sunlight theory.

Ken had been using uranium-series decay methods to date calcite that had precipitated in a cave found in the Mojave Desert; the cave is appropriately named Devils Hole. Surprisingly, as the calcite grew on the walls of Devils Hole, the isotopes of oxygen that went into this calcite also changed, and they did so in a pattern that was quite similar to the ice age record recovered from oxygen isotopes from deep-sea sediment. The breakthrough at Devils Hole was that this particular type of calcite could be dated accurately by uranium-series methods, whereas most calcite in the oceans could not. This study was a major blow to the standard Milankovitch theory, as it showed that warming in the Northern Hemisphere had occurred while northern hemisphere sunlight was at a minimum. Something was amiss!

What the Devils Hole work did was provide the rest of the scientific community the opportunity to question the validity of the Milankovitch theory of ice ages. The year 1992 was when I started graduate school here at Berkeley and my decision to work in the field of paleoclimatology was directly influenced by the Devils Hole study. I was invited, even encouraged, to become an iconoclast of paleoclimatology. Fortunately, the resident paleoclimatologist in our department had resigned for a position at MIT the week before I arrived, so nobody was around to dissuade me from following the encouragements of my advisor. I chose to work in Italy along the Tyrrhenian Sea coastline west of Rome, where the rises and falls of sea level that accompanied ice age cycles controlled the deposition and erosion of the coastal sediments. I was able to date those rises and falls by identifying volcanic ash layers in the coastal sediments, and dating them using 40Ar/39Ar dating methods at the Berkeley Geochronology Center, where Paul Renne was one of my graduate advisors. Our work has been able to provide radioisotope dates for three other glacial-interglacial transitions, and like Devils Hole, our data suggests that the end of ice age cycles occur systematically earlier than when the Northern Hemisphere is supposed to warm up by sunlight reaching high northern latitudes. In addition to our work, there are now many radioisotopic dates from coral terraces found at interglacial sea-level elevations, which prove the early warming at 135,000 years ago.

What these new dates mean is greatly underappreciated by most scientists inside and outside of paleoclimatology; perhaps this revolution has been so quiet because it was initiated by geochronologists and physicists, who are not mainstream paleoclimatologists. What this means is that at the very least there is an unaccounted for variable whose effect on ice ages is at least as large as the traditional Milankovitch mechanism, and this variable has dominated the ice age cycles for the last million years. Since the Milankovitch theory already includes all orbital variations that affect the amount and distribution of sunlight that reaches Earth, this unaccounted for variable cannot be attributed to changes in sunlight.

A clue to this missing variable may lie in studying its periodicity. When we examine the ice age cycle data for the last million years, we find that it is dominated by a 100,000 periodicity, and this periodicity is not consistent with any component of the Milankovitch theory or predictions based on it. However, the cycle remains phase-locked, meaning that if you were to superimpose a sine wave with a period of 100,000 years on top of the data, you would find that, while the climate signal wanders a little bit, it eventually snaps back to its 100,000 periodicity. This type of regularity can only be caused by some forcing signal that is external to the Earth — it is probably astronomical in origin.

At present there is no model that satisfactorily explains this 100,000 ice age cycle. While Richard Muller in the Physics Department here at Berkeley has shown that this periodicity matches that of Earth's orbital inclination, we have not been able to show a causal relationship between inclination and ice ages. This means that there is great opportunity within this field for scientists to formulate new models for the cause of the ice ages, but these new models must make testable predictions if they are to advance this field. Regardless of which model turns out to be right, I am certain that it will revolutionize our understanding of long-term changes in Earth's climate.

To summarize, the field of paleoclimatology has seen several hypotheses for the causes of the ice ages come and go, with the most recent one being disproved by scientists who were not experts within that field. I am sure that many paleoclimatologists have kept quiet about this matter, and have been hoping that someone would identify some flaw with how the geochronologists were analyzing their data. But given the fact that many different radioisotope dating labs have come to this same conclusion in the past several years, I don't think that's going to happen.