Marci Robinson. American Scientist. Volume 99, Issue 3. May/Jun 2011.
Over the course of geologic history, global temperatures have increased and decreased in response to a multitude of forces. Changes in the positions of continents, which can modify circulation in ocean basins and affect the global distribution of heat, have been a major cause of global climate change. Variations in Earth’s orbital configurations, which guide the glacial-interglacial cycles, have been important. So have changes in the atmospheric concentration of carbon dioxide, a greenhouse gas that traps heat in our atmosphere. Earth’s present warming trend is happening too quickly to be related to the first two factors. So most scientists attribute the change to the rapid increase in greenhouse gases—primarily from fossilfuel combustion—accumulating in our atmosphere.
What will this mean for our future? Answering that question has fallen in large part to the Intergovernmental Panel on Climate Change (TPCC), the international body that synthesizes volumes of climate research into a single report every five to six years. In 2007, the IPCC Fourth Assessment Report predicted a mean annual global temperature increase of between 1.1 degrees and 6.4 degrees Celsius by the end of the 21st century. These conclusions were arrived at using general circulation model, or GCM, experiments. GCMs are numerical models of the planet that simulate physical processes in the ocean, atmosphere, cryosphere and at the surface.
Such a wide range of predicted temperature change, of course, produces a wide range of climate-effect forecasts. A climate that is on average 1 degree warmer than today is likely to be very different than a climate that is 6 degrees warmer. That could represent the difference between 18 centimeters and 59 centimeters of sea-level rise, for example, in the next 90 years. How much faith should.be placed in different model results? How do we quantify their uncertainty?
One way to seek a more accurate view of the future is to look closely at the past, to a time when Earth’s temperatures matched what may occur by the end of this century. Temperature records, dating only from 1850, don’t carry us back far enough. Nor can ice cores, which capturé atmospheric carbon dioxide levels dating back only 800,000 years. But studies based on chemical ratios preserved in tiny marine fossils, along with other proxies, are filling in some blanks. The results are becoming a useful “ground truth” for climatologists attempting to model our future climate. At the same time, the models are helping paleoclimate researchers such as myself improve our understanding of the past.
The most powerful proxy tool available to help scientists identify past warm periods is the marine oxygen isotope record, which is based on the ratio of two oxygen isotopes—oxygen-16 and oxygen-18—preserved in the shells of microfossils. During evaporation, the lighter isotope—16O—is preferentially removed from the water, leaving the remaining sea water enriched with 18O. The evaporated water contributes to the snow that builds ice caps. During cold periods, ice caps grow, trapping the lighter isotope, and sea level falls, which leaves seawater enriched in the heavier isotope.
Marine organisms preserve the oceans’ isotopie signatures when they precipitate calcium-carbonate shells. Shells formed during glacial periods have a higher proportion of heavy oxygen than shells formed during interglacial periods. Because changes in the isotopie signature are synchronous across all oceans, records from different parts of the world can be correlated. The efforts of many scientists analyzing millions of shells spanning millions of years of geologic time have yielded an oxygen-isotope record that roughly estimates global changes in temperature and sea level.
According to this record, the middle portion of the Pliocene Epoch—about three million years ago—is the most recent period when global temperatures were sustained at levels we may see at the end of this century. Three million years may. sound like the very distant past but, geologically speaking, it is not. At that time, Earth’s continents were in basically the same positions they are now, so ocean circulation patterns (the main vehicle for the transportation of heat) were much the same. Many of the plants and animals that populate our world had already evolved, which makes direct comparisons of living species to their fossil counterparts easy. This symmetry makes the Pliocene a good model of what our future could look like.
Atmospheric carbon-dioxide concentrations during the Pliocene were only slightly higher than they are today. Still, global temperatures during the mid-Pliocene were on average 2 to 3 degrees warmer than today, and sea level was on average 25 meters higher. This apparent paradox—a warmer planet with comparable carbon dioxide levels—concerns some researchers who wonder whether even small increases in carbon dioxide can significantly alter our climate.
Soviet scientist Mikhail Budyko pioneered climate studies of the mid-Pliocene with climate change in mind, but he was fueled by optimism rather than alarm. In the 1970s, he concluded that a world warmed by greenhouse gases could benefit regions of the Soviet Union, and he was the first to suggest that using reconstructions of a previous warm period on Earth could be a useful predictor of 21st-century conditions. After encountering Budyko’s preliminary data-based climate reconstruction during a U.S. and Soviet scientific exchange, Dick Poore of the U.S. Geological Survey (USGS) and David Rind of NASA concluded that the USGS could develop a more quantitative data set for the Pliocene. Since its founding in 1989, the USGS Pliocene Research, Interpretation and Synoptic Mapping (PRISM) project has pursued an unparalleled estimation of what a warmer world looked like.
A fundamental tenet in geology is uniformitarianism, meaning that, the same natural laws and physical processes that operate today were active in the past. By making a few assumptions concerning the stability of ocean chemistry and individual species’ ecological tolerances, fossil studies can reconstruct environments of the past at specific locations. When many of these individual reconstructed environments of the same age but from different locations are strung together they give us a picture of global conditions in the past.
This research starts with collecting sediment samples from all over the world. In the deep ocean, most samples come from deep-sea cores drilled by the internationally funded Integrated Ocean Drilling Program (IODP) and its precursors. A European consortium, China, Korea, Australia, New Zealand and India help fund the drilling program, with the U.S. National Science Foundation and Japan’s Ministry of Education, Culture, Sports, Science and Technology at the helm. Any researcher can request IODP samples.
IODP expeditions lower coring equipment from research ships to the sea floor and drill down hundreds to thousands of meters to remove ocean sediment containing millions of years of fossil deposits. Several dating techniques allow us to determine how deep in the core to look for Pliocene fossüs. (In lakes, the coring process is similar but on a smaller scale. On land, the work can involve deploying hammers and shovels on the side of a road that cuts through a Pliocene deposit.)
Marine cores are usually full of planktic foraminif era, or forams, singlecelled organisms that float at or near the ocean surface. A uniquely shaped and ornamented calcium-carbonate shell characterizes each of the nearly 40 species, making it easy to identify them. All are about the size of a grain of sand. Because each species lives in a well-defined range of environmental conditions, the types of species found tells us something about the temperature (or salinity or productivity) at the place and time the forams were alive.
If a fossil assemblage is 60 percent one species and 40 percent another, and if that combination is found in contemporary times where the water temperature is 14 degrees, then you assign a temperature of 14 degrees to that sample. More likely, however, the fossil sample will have as many as 40 species, and that combination will exist in today’s ocean over a range of temperatures. The best match to the assemblage may represent a combination of temperatures. A factor analysis is also useful: Modern assemblage data and the associated physical environments undergo a multivariate regression that is applied to the fossil data. That yields an estimate of the temperature in which the assemblage formed.
Most reconstructed ocean temperatures in this research are derived from forams, but other microfossil groups (such as diatoms, radiolaria, ostracods and pollen) are used in much the same way. Another way to estimate paleotemperature is through foram and ostracod shell chemistry. Although made primarily of calcium, carbon and oxygen, they contain a small amount of magnesium. Some magnesium ions replace calcium ions in the calcium carbonate lattice structure. The rate at which this happens is closely tied to the water temperature at the time the shell is secreted. Applied to fossils at the ocean floor as well as those at the surface, this correlation helps us reconstruct deep-ocean thermal grathents. Also useful are alkenones, long-chained organic compounds synthesized by a small group of algae that dwell near the surface of the ocean. The number of double carbon bonds in the chains, or the degree of alkenone unsaturation, varies linearly with the water temperature during growth and provides ocean-temperature estimates that are independent of marine microfossils.
Evidence regarding vegetation, land ice, sea level, deep-ocean temperature, and topography are also in play in the PRISM mid-Pliocene reconstruction. On land, the distribution of vegetation and land ice is reconstructed through fossil pollen. The extent of land ice on Greenland and Antarctica is loosely defined by the vegetation distribution, but the volume of land ice is closely tied to Pliocene sea level as well. More physical evidence comes from the geological remains of mid-Pliocene shorelines. For instance, the Orangeburg Scarp in the United States, which runs from Florida to Virginia, marks the edge of the Pliocene-era Atlantic Ocean. Today Interstate 95 runs east of the scarp.
Testing the Models
PRISM so far has reconstructed the midPliocene climate with data from 86 marine and 202 land sites. The Pliocene climate coming into view is very different from modern climate in some areas and very similar in others. Temperatures near the equator generally were much like they are today, but temperatures at the poles were much warmer. Moving away from the equator, the temperature difference becomes more extreme.
In the North Atlantic and Arctic oceans, for instance, temperatures at the surface were substantially warmer. These warmer conditions were reflected in the vegetation of Iceland and Greenland, which hosted boreal forest rather than polar tundra vegetation, indicating a Pliocene mean annual temperature at least 10 degrees warmer than what is seen today. In the Southern Hemisphere, Antarctica was vegetated with shrubs along the West Antarctic peninsula and along the coast of Wilkes Land where today there is only ice.
Temperatures in the tropics, for the most part, mimic today’s temperatures. In the western tropical Pacific Ocean, near Papua New Guinea, surface water temperatures were similar to present readings—about 29 degrees. In the modern Pacific Ocean, the surface water in the west is a few degrees warmer than in the east, except during an El Niño event when the temperature grathent flattens, an atypical pattern of warmer than normal sea-surface temperatures that alters global weather patterns. But the normal state during the warmest parts of the Pliocene was more like a modern El Niño event. The eastern equatorial Pacific Ocean, near Panama, Colombia and Ecuador, was as warm during the Pliocene as it was in the western equatorial Pacific Ocean, about 1.5 degrees to 2.5 degrees warmer than it is today.
At western continental margins along the coasts of Peru and California, today we find cold, nutrient-rich upwelling zones (and the world’s most productive fisheries). Pliocene upwelling was nutrient-rich, judging by paleoproductivity proxies, but as much as 7 degrees warmer. PRISM is incorporating 27 new data sites into its reconstruction, many of them along continental margins, to better understand the nature of the warm upwelling zones of the Pliocene. The Indian Ocean is woefully underrepresented in the current PRISM reconstruction but will come into much sharper view when six new sites are added, as planned. A new focus of research is the historical context of the modern Indian Ocean temperature dipole, another episodic pattern with a strong surface temperature grathent that affects local weather patterns and regional climate.
Increasingly, climate modelers are testing their accuracy by applying their computational tools to making predictions about the Pliocene climate. And geologists and the modelers are benefitting. For example, in 2004 a significant discrepancy was observed between PRISM’s estimates of Pliocene temperatures in the eastern equatorial Pacific and temperatures calculated by the United Kingdom’s Hadley Centre Coupled Model (HadCM3). The accuracy of the Hadley model is vital since it has been used extensively for climate prediction and sensitivity studies. In its Pliocene simulation of the eastern equatorial Pacific, HadCM3 had concluded that waters west of Panama, Colombia and Ecuador were warmer than they are today. The PRISM reconstruction had not captured that result.
This prompted the author and Harry Dowsett, a micropaleontologist who leads the PRISM group, to search for clues. In IODP’s records, we found three cores containing Pliocene sediments in the equatorial region (0 degrees to 6 degrees north or south latitude), one very near the west coast of South America at 84 degrees west longitude, one at 95 degrees west longitude, and one at 110 degrees west longitude. We analyzed the planktic foram assemblages from all three, along with alkenones from the easternmost site that is today within an upwelling region.
The analysis revealed that the eastern equatorial Pacific was in fact warmer during the Pliocene than we’d estimated. The westernmost site showed a 2.5 degrees temperature increase over modern averages. Proxies at the middle site estimated a 2.8 degree temperature increase over modern averages—an incredible agreement between our foram analysis and other data. And at the easternmost site, temperature estimates for the Pliocene were 1.5 and 1.9 degrees warmer than modern measurements. The model projection was correct.
In another example, the PRISM reconstruction observed extreme warmth near Iceland and Svalbard, but the 2004 HadCM3 simulation did not. To address this discrepancy, we started looking into ocean circulation patterns and how the features on the ocean floor in this region have changed since the Pliocene.
North Atlantic deep water forms as cold, salty and more dense surface water sinks just north of Iceland. The Greenland-Scotland Ridge, a bathymétrie feature related to the Icelandic hot spot, runs from Greenland to Scotland, through Iceland. This ridge traps the newly formed deepwater as it tries to flow southward along the ocean floor. Geophysical research pioneered by Peter Vogt in the 1970s shows that the Greenland-Scotland Ridge was lower during the Pliocene. Geothermal activity under Iceland has caused the ridge to rise by about 300 meters in the past 3 million years. Additional research by Jim Wright and Ken Miller at Rutgers University showed that changes in the ridge affect deepwater circulation. A pilot study published this year by the author and Dowsett, climate modelers Paul Valdes of the University of Bristol, Alan Haywood of the University of Leeds and Dan Hill of the British Geological Survey, and Steve Jones, a geophysicist at the University of Birmingham, incorporated this change. The adjustment to the model’s boundary conditions brought the data and model simulation into agreement.
Comparing data to the model output is essential to improve confidence in climate-model simulations for the future, but these comparisons are in their infancy Subtle aspects of the comparisons require further attention. We are still addressing some basic differences between the models and the reconstruction, such as how to define “modern” and how to apply seasonal temperature cycles. The data-point-based nature of the reconstruction is also an issue. Both marine and terrestrial cores are scattered unevenly across the globe, clustering in the North Atlantic Ocean and in Western Europe, respectively. Temperature estimates must be extrapolated between points. The biggest data-free space is the middle part of the South Pacific.
In addition, some assumptions are inherent in all proxy data as well as in climate modeling. We assume, for instance, that the ecological tolerances of individual species of foraminifera and the species that produce alkenones haven’t changed significantly over time. We assume they lived in the same sorts of environments in the Pliocene as they do now. Likewise, the modelers assume that parameters of climate phenomena, such as cloud formation, can be defined mathematically. Without these assumptions, we wouldn’t be able to pursue the science.
Uncertainty in climate projections comes from two sources: the accuracy of the prescribed boundary conditions and the ability of the model to simulate the highly complex global climate system. In this case, the boundary conditions are the individual data sets of the PRISM reconstruction. These are used to set up the model experiments. In the PRISM sea-surface-temperature data set, we rninimize uncertainty by utilizing multiple proxies: microfossil assemblage data, magnesium-calcium ratios and alkenones. This approach reduces the overall error compared with results based on a single proxy. It also expands temperature estimation into regions where other proxies are not suitable. For example, planktic foraminifera techniques are not well suited for estimating temperature at high latitudes where foram specimens are often scarce and where assemblages become dominated by a single, often extinct, species. Similarly, alkenone-based techniques are ineffective in the warm tropics because they do not record temperatures above roughly 28 degrees.
Uncertainties in model simulations can be addressed through model ensembles. In multi-model ensembles, also known as model intercomparisons, different climate models run identical experiments with the same set of boundary conditions and prescribed emissions scenarios, as in the EPCC projections at the beginning of this article.
As the December 2011 deadline approaches for the first draft of the Working Group I Contribution to the IPCC Fifth Assessment Report, the paleoclimate community is gearing up. Model intercomparisons are being completed and analyzed. As part of the Pliocene Model Intercomparison Project (PlioMIP) launched by Dowsett of PRISM, Mark Chandler of Columbia University and the Goddard Institute of Space Studies, and Haywood of the University of Leeds, 17 international climate-modeling groups are currently using the PRISM reconstruction to set boundary conditions in identical experiments. The range in model results will identify model-dependent variability that can then be eliminated from the uncertainty measure. The remaining range will act as error bars to quantify the uncertainty associated with past and future climate simulations.
Another way to bracket model simulation uncertainty is through perturbed physics ensembles; Perturbed physics ensembles use only one model at a time and run a series of experiments in which only one variable is changed. The idea is to compare a range of model results, all using different values of that one variable, to the PRISM reconstruction. The best match gives us the most appropriate value for that variable. Thus, that bit of uncertainty is removed.
James Pope, a doctoral student at the University of Leeds, has taken on the task of using a perturbed physics ensemble to estimate uncertainties in Pliocene climate model projections using the HadCM3 and PRISM boundary conditions. This is the first time a perturbed physics ensemble has been applied to a warmer world with high levels of carbon dioxide. In the first experiments, now underway, the variable is climate sensitivity. Climate sensitivity is defined as the global mean annual temperature response to a doubling of atmospheric carbon dioxide. Uncertainty in the value of climate sensitivity is a big part of what creates uncertainty in climate model simulations.
A pilot study by Pope identified lowsensitivity (climate sensitivity = 2.1 degrees) and high-sensitivity (climate sensitivity = 7.1 degrees) model simulations of Pliocene climate as brackets. The high-sensitivity simulation was a close match to the PRISM sea-surface temperatures, but the control, an unperturbed simulation, better captured Pliocene vegetation distribution. These results show promise that a best fit to the data will likely be produced as part of a full ensemble of model simulations.
Making the Best of It
The Pliocene climate is the closest natural analog to the climate we expect near the end of this century and it is an appropriate target for ground-truthing model simulations. But it’s important to remember that the Pliocene is unlike the 21st century in key ways. For one, Earth then had not plunged into the glacial-interglacial cycles that have characterized the past few million years. So, rather than trending toward warmer conditions as we are today, the Pliocene was instead the last period before prolonged cooling that led to the ice ages.
More importantly, thé Pliocene warm period was characterized by a stable climate state very different from today’s harried state of climate disequilibrium. Atmospheric carbon dioxide concentrations, for instance, were relatively stable during the Pliocene, and the climate had adjusted to them. The release of carbon dioxide was natural and gradual, from processes such as volcanic emissions and the decay of plant and animal matter. This is very different from today, as carbon dioxide is released much more rapidly through fossil-fuel use and deforestation. The current climate system is still adjusting and will take a very long time to equilibrate.
In addition, we identified a natural, geologic contributor to Pliocene warmth that acts on a very slow time scale. The changes in the height of the Greenland-Scotland Ridge, resulting from the movement of magma below Iceland, likely affected the transport of warm surface water into the Arctic Ocean. The warmth seen at these high latitudes was a result of changes that took place on hundred-thousand to million-year timescales. Our climate will not be responding to this particular climate forcing in the near future.
Though the Pliocene is not the perfect analog for the near future, it helps us focus on what may occur. A rise in sea level, for example, and a poleward expansion of tropical temperatures and plants are aspects of the Pliocene that we’re likely to see as our climate warms. Earth’s high latitudes will likely warm more than the low latitudes, and our ice caps will melt. By how much, we don’t yet know. Minor changes in oceanic circulation may have major effects on regional phenomena such as seasonal changes in thermocline depth, where the rate of temperature decrease is the largest, and upwelling. Yes, uniformitarianism is often defined to mean that the present is the key to the past. But we might be wise to remember another accurate description: The past is a key to understanding the future.