paleoclimate

I have from time to time written short pieces here about climate fluctuations and their effect on human evolution. This topic was a major theme in the recent Nova miniseries on human origins.

Some scientists think that human intelligence is an adaptation to rapid climate changes, and that brain size increased during the Pleistocene as a direct result of climate fluctuations. I write, "scientists," because this really is a motley interdisciplinary crew -- archaeologists, geologists, psychologists and some paleoanthropologists in the mix.

Some others think that humans mainly suffered through climate by means of a few major events, global catastrophes like the Toba eruption. By decimating human populations, these big catastrophes caused all kinds of accidental changes to our biology. Why just today, you can read a story riffing on a Curtis Marean lecture, making it into this: "How shellfish saved the human race." Just one of many speculations about how climate brought humans to near-extinction in the last couple hundred thousand years.

These two kinds of thinking about paleoclimate and human evolution are not mutually exclusive, and many people kind of lump them together. Both kinds of arguments depend on models linking fitness, plasticity, and paleoclimate. They also depend on the amplitude of ancient climate change episodes and the relation between this amplitude and the geographic variation in climate known to have been tolerated by ancient humans. Genetic modeling and paleoclimate modeling, working together.

That's the way that testable science ought to work. But so far, proponents of these ideas have proposed only verbal arguments without any modeling or math at all. Without the numbers, I don't see any reason to believe that climate specifically affected the evolution of human behavior. Is it bunk? I'd like to know at least if any of it is falsifiable.

I'm going to review some literature in both areas during the next few weeks. I'm reasonably familiar with the literature on selection and plasticity, and it gives me a lot of skepticism about the prospects of explaining human intelligence as a function of climate variability. But I'm less familiar with the paleoclimate literature. Maybe hypotheses about climate forcing in human evolution are untestable -- but to arrive at this conclusion both the genetic and paleoclimate aspects need a realistic assessment.

Submillennial-scale variability

How can we start to understand the role of climate variability on Pleistocene human evolution? I'm going to begin with the topic of submillennial-scale climate fluctuations. The global climate fluctuates from year to year, from decade to decade, across centuries, and on much longer scales -- up to the 100,000-year duration of Pleistocene glacial episodes.

The longest climate cycles are relatively simple -- the paleoanthropological record is just good enough that we might be able to find associations between skeletal evolution, archaeological presence and the major glacials. These very large-scale temporal effects are not my immediate interest, but they will be worth considering very closely. Did the earliest habitation of Europe depend on climatic goodwill? To what extent were humans driven south out of Northern Europe by the last glaciation? What effects did periodic isolation from the Asian mainland have on Pleistocene Javan populations? Those are questions that implicate the largest-scale climate cycles, and I think they may be to a great extent answerable.

They are not, however, the kind of effects argued to have induced a consistent selection pressure on intelligence. Nor do they suggest a global bottleneck due to climate instability.

A sustained selection pressure would require climate instability on a shorter timescale -- measured in maybe tens of human generations or less, not thousands. This scale is submillennial at most, and possibly decadal.

The reconstruction of paleoclimate on these shorter timescales has become feasible during the last ten to fifteen years. For the last few millennia, tree ring records provide evidence about climate variation -- sometimes at annual intervals, sometimes time-averaged over five or ten year spans. Lake sediments sometimes preserve a record of climate variability, in the thickness or composition of layers laid down over time. Sea floor sediments also can inform about ancient climate, principally because windblown (or waterborne) sediments reflect continental aridity. Ice cores from glaciers and ice sheets also provide a record for reconstruction atmospheric temperatures and precipitation over some regions -- and in the exceptional cases of the Greenland and Antarctic ice sheets, climate reconstruction may be possible back to as far as 800,000 years ago.

But these records are not direct indicators of ancient temperature or precipitation; they are proxies. Each of them is localized in a particular place with its own unique regional deviations from the global mean. During the Pleistocene, the places were people lived were pretty far away from most of the available proxy records. There are exceptions, like the Nile delta sapropels and African lake sediments, which I'll discuss in some detail. But for the most part, our question is whether proxy records can test hypotheses about regional climate variability in the regions occupied by humans.

The same question arises in the context of recent climate change. We have several proxy records for temperature and precipitation during the last 1000 years, and we want to accurately reconstruct global mean temperature means and variability. The question was broached in a 2009 editorial essay by Hughes and Ammann, as follows:

The last millennium paleoclimate reconstruction (PR) challenge Given the suite of uncertainties and limitations connected to proxy and short instrumental data, it is very difficult to derive a reconstruction method that is capable of explicitly dealing with all issues, and thus some compromises must, inevitably, be made. At the same time, many existing methods might be reasonably well suited to study particular aspects of past variability and change (Ammann and Wahl 2007). However, we do not have a solid quantitative understanding of the strengths and weaknesses of each method. When different methods are applied to the same time period and spatial domain, how different are their results and inherent uncertainties, and to what is each method particularly sensitive? Differences could result from limitations in the included proxy climate records’ ability to record climate across different time scales; the sampling distribution (network); or possibly the method itself or its underlying assumptions. To assess the influence of all these uncertainties, the last millennium paleoclimate reconstruction (PR) challenge is being developed to offer an experimental test bed for systematically gauging the ability of the various methods to recover the true underlying climate. The climate data are drawn from output of fully coupled climate system models that provide geophysically realistic spatiotemporal variations across various climate fields. Thus the target of reconstructions will be fully known. ... The key question, ultimately, is to find out how well the suite of reconstruction methods, if applied to a realistic set of pseudo-proxy data, are able to reproduce the underlying climate, not knowing the actual target climate (achieved in a double-blind approach) (Hughes and Ammann 2008:256-257).

In other words, even considering the last 1000 years with its abundance of proxy data, the climate models are not yet capable of replicating regional variability or time-series variance of temperature records with high fidelity. I'll add (from elsewhere in the editorial) that the general question of the magnitude and duration of external forcing of climate (say, fluctuations in solar output) are also not well integrated.

The last 1000 years (and more broadly, the Holocene) has sometimes been described as relatively stable compared to the higher-amplitude climate fluctuations of the Late Pleistocene. In that light, the prospects seem poor that we could accurately assess the regional effects of Pleistocene climate instability, using many fewer proxies than are available for the last 1000 years.

But for particular times and places, the situation may be better. Also, the general concept of global climate instability may itself allow us to work with genetic models, if we only knew the likely amplitude and periodicity of changes. So I'll be reading further into the climate variation of the last 1000 years and how it may have been different in the Pleistocene.

Meanwhile, the Hughes and Ammann editorial is interesting in several sections, as the authors discuss possible methods for increasing the fidelity of climate models as applied to paleoclimate data. I was following up on the literature citing one of these methods when I found the essay. On this, I'll just add that the statistical methods for dealing with time series of temperature proxies are tricky. Some methods may raise the prediction value of a proxy in different regions, while decreasing the appearance of yearly or decadal-scale variation in the global mean. So one must be careful about the choice of statistics when assessing paleoclimate variability.

References:

Hughes MK, Ammann CM. 2009. The future of the past -- an earth system framework for high resolution paleoclimatology: editorial essay. Climatic Change 94:247-259. doi:10.1007/s10584-009-9588-0

In 2007, R. B. Firestone and colleagues published evidence of an extraterrestrial impact, roughly coincident with the onset of the cold climate event known as the Younger Dryas. This event, around 12,900 years ago, is around about the time of some (but not all) megafaunal extinctions in North America, it is also around the time (but not precisely) of the Clovis culture. The paper argued that the impact event may have "contributed to end-Pleistocene megafaunal extinctions and adaptive shifts among PaleoAmericans in North America".

Last year, I reported on widespread dissatisfaction with this impact hypothesis. Some critics didn't think that there was any evidence of megafaunal trauma from the impact, some didn't think that the dates matched any "adaptive shifts", and in particular the end of the Clovis culture.

And then others didn't think that there had been an impact at all. These were in some ways the most worrisome, because they directly questioned the supposed evidence in support of an extraterrestrial event -- "microspherules" of magnetic material, clustered in sedimentary contexts at precisely 12,900 years ago in sites across much of North America.

Now, in the Proceedings of the National Academy of Sciences (where Firestone and colleagues originally published their observations), Todd Surovell and colleagues have published a remarkable paper that tests the Firestone impact hypothesis: "An independent evaluation of the Younger Dryas extraterrestrial impact hypothesis." Most critiques attempt to find an alternative explanation for a set of original observations. In this paper, Surovell and colleagues merely attempt to replicate the original observations at multiple sites, and fail -- as their abstract tersely states,

We were unable to reproduce any results of the Firestone et al. study and find no support for Younger Dryas extraterrestrial impact.

Just like that -- it's about as hard-hitting as you're going to see in a scientific research paper.

Of course, this paper only examined one out of a number of observations that Firestone and colleagues had adduced in support of the impact hypothesis. But in the introduction to their paper, Surovell and colleagues reference several other recent studies that re-examined other aspects of the evidence:

A series of critiques of the original Firestone et al. article (1) have been published recently (8-10). Pinter and Ishman (8) argue that the suite of markers used to indicate impact are inconsistent with "any single impactor or any known event." Furthermore, they provide alternative explanations for many of the observed marker peaks. For example, glassy and metallic microspherules are known components of atmospheric dust derived from the constant influx of micrometeorites. An independent evaluation of the charcoal evidence was recently published by Marlon et al. (9). Examining concentrations of charcoal from 35 pollen cores across North America, they found no evidence for large-scale, continent-wide wildfires specifically associated with the onset of the [Younger Dryas].

In the current case, the results are very simple: they went looking for a spike in the number of impact-generated particles coincident with the Younger Dryas. They looked at seven sites with long and continuous records of sedimentation across that interval. They found the supposed impact-generated particles, but not patterned with any kind of spike.

They suggest a different model for the presence and accumulation of the magnetic particles:

Alternatively, it may be that the presence, absence, and relative abundance of magnetic materials, especially the spherules, is due to characteristics of the parent material and depositional environment instead of some sort of continent-wide extraterrestrial process. The characteristics of the local depositional setting before, during, and after 12.9 ka have not been addressed by the proponents of the impact hypothesis. The zones producing the YDB ‘‘impact markers’’ are typically associated with soils (stable surfaces) or shifts in the depositional environment (e.g., alluvial to lacustrine conditions at Blackwater Draw, Lubbock Lake, Murray Springs, and Lake Hind; buried soils in the Carolina Bays and at Lommel, Belgium).

One might imagine atmospheric particles accumulating on stable paleosols over long stretches of time, generating a local spike in the number of such particles in the stratigraphic column. In any event, the data presented here don't bear out the hypothesis of any unusually large impact event.

I'm not a geologist, and I have no special insight into the analyses here, beyond reading the charts. But remember that the impact hypothesis made a tremendous media splash. Maybe more damaging to the scientific side of things, the hypothesis that the Younger Dryas cold period came from an extraterrestrial force, made it seem for a moment less necessary to investigate terrestrial sources of cooling at the terminal Pleistocene. The science will correct itself, but the public perception of the climate changes at the end of the Ice Ages will need quite a bit more nursing to get a more realistic perspective on the story.

References:

Firestone RB and lots of others. 2007. Evidence for an extraterrestrial impact 12,900 years ago that contributed to the megafaunal extinctions and the Younger Dryas cooling. Proc Nat Acad Sci USA, 104:16016-16021. doi:10.1073/pnas.0706977104

Kerr RA. 2008. Experts find no evidence for a mammoth-killer impact. Science 319:1331-1332. doi:10.1126/science.319.5868.1331

Surovell TA, Holliday VT, Gingerich JAM, Ketron C, Haynes CV, Jr, Hilman I, Wagner DP, Johnson E, Claeys P. 2009. An independent evaluation of the Younger Dryas extraterrestrial impact hypothesis. Proc Nat Acad Sci USA (early) doi:10.1073/pnas.0907857106

So a couple of weeks ago, the Journal of Biogeography published a paper arguing that humans and orangutans are sister taxa.

This week, the journal has published a paper on the biogeography of Sasquatch.

Yes, that's correct. The same journal that published the updated "red ape" paper has now published a Bigfoot paper.

OK, Journal of Biogeography, I'll be your straight man. I mean, I bow to no one in my ability to snark on human evolution, but this is like some sort of karmic singularity.

Besides, this Bigfoot paper is really good. The authors intend it as a "tongue-in-cheek" example, and it works on that basis, as an illustration of the "garbage in, garbage out" principle. Biogeographers have been using increasingly sophisticated computer algorithms to predict the "ecological niche" of species. The algorithms take information about sightings or recorded incidences of a species, find commonalities among those sightings against maps of other ecological data (rainfall, forest type, presence of other species, etc.), and spit out an ecogeographic distribution for the target species.

Of course, the algorithm will come to some result, regardless of how well each piece of data in the system is known. And the algorithms are complex enough that the creeping effects of errors may be hard to evaluate:

While the value of publicly available sample locality data is not questioned, the consequent introduction of errors in the accuracy of specimen identity and georeferencing could be problematic for developing ENMs [that is, "ecological niche models"] from public data sources (Graham et al., 2004; Soberón & Peterson, 2004). Although georeferencing inaccuracies can be identified in databases from qualitative or quantitative accuracy thresholds (e.g. http://manisnet.org/GeorefGuide.html), poor taxonomy and/or misidentification may be less detectable. This issue may be particularly problematic, for example, with cryptic species or subspecies that are morphologically similar but may have very distinct ecological requirements and geographic distributions, or for those data sources that contain indirect observations rather than references only to physical specimens.

What better way to illustrate the problem, than by applying the analytical methods straight to a "species" whose existence is, shall we say, "questionable."

The authors additionally raise one particular element of confusion that may enter into ecological niche modeling -- sightings of similar species may confound each other. They consider the example of black bears -- another large North American mammal that occurs in many of the areas where Bigfoot sightings are reported. They run the same analysis on black bears as they did on Sasquatch, finding a large overlap in distributions. But interestingly, they have fewer observations for bears than they do for Bigfoot -- leading them to an underestimate of the actual range of black bears. They reflect on the possible interpretations of the overlap:

Thus, the two 'species' do not demonstrate significant niche differentiation with respect to the selected bioclimatic variables. Although it is possible that Sasquatch and U. americanus share such remarkably similar bioclimatic requirements, we nonetheless suspect that many Bigfoot sightings are, in fact, of black bears.

From my perspective, this paper is important for two reasons, neither of them really having to do with large North American primates. First is a social and legal angle. Increasingly, the habitat distributions of endangered or threatened species are evaluated on the basis of similar computer models of ecological niche. In particular, the changes in species distributions under scenarios of climate change are modeled in this way. This computer modeling has the appearance of objectivity, and it certainly allows reams of data to be statistically simplified into human-readable maps. That makes the results of such analyses really valuable to cases where political and legal units need to make decisions about how to comply with threatened species regulations.

But if the data going into the model aren't correct, then the predictions of the models won't reflect reality. The question is, how much will they be wrong?

In this case, the limited black bear dataset leads to a substantial underestimation of the black bear habitat niche. And possible confusion between Sasquatch and black bear sightings raises the possibility that any rare species will be significantly overrepresented in ecological niche modeling when such confusions are possible. Neither of these outcomes tells us how bad such errors are likely to be, but they point to real weaknesses in the maps generated by these computer algorithms.

I expect that smart lawyers will be finding ways to use this Bigfoot paper a lot.

Second, I think the paper is important at this moment in paleoanthropology. Late last year, I wrote about a paper that evaluated ecological niche models for Neandertals and people who made the Aurignacian ("'Competitive exclusion' and the extinction of Neandertals: should we believe it?"). The paper, by William Banks and colleagues, had used the observed distribution of archaeological sites between certain radiocarbon date intervals to estimate an ecological niche model for the two hominid groups.

I think that paper was very good work, but it obviously is subject to uncertainties in the initial observations -- just as the Sasquatch data are. The archaeological observations add even more uncertainties of dating and certainty of association between biology and archaeology. And the Sasquatch data set is much, much larger in terms of sighting numbers; meaning that the archaeological cases ought to have more error, when it comes to evaluating niche flexibility.

For the purposes of the Banks et al. (2008) paper, I think their conclusions are pretty secure. They tested the hypothesis that the reduction of Neandertal occurrences over time could be explained by climate change; they were able to reject that hypothesis by showing the ecological niche breadth of earlier Neandertals included paleoenvironments that were very widespread across Europe throughout the period when they were declining. It's a nice demonstration.

But the Sasquatch example shows that we have to evaluate the ability of ecological niche modeling to test each hypothesis, on the basis of the data that are likely to be available. Can the model show that the spread of Aurignacian people caused the Neandertals to decline? That depends on our confidence about the dating and biological associations at early Aurignacian sites. The computer algorithm gives a structured way to reduce information that is already manifest in the data.

References:

Lozier JD, Aniello P, Hickerson MJ. 2009. Predicting the distribution of Sasquatch in western North America: anything goes with ecological niche modelling. J Biogeogr (early online) doi:10.1111/j.1365-2699.2009.02152.x

Banks WE, d’Errico F, Peterson AT, Kageyama M, Sima A, et al. (2008) Neanderthal Extinction by Competitive Exclusion. PLoS ONE 3(12): e3972. doi:10.1371/journal.pone.0003972

The Tswaing Crater is around 40 km from Pretoria, South Africa. It was created by an asteroid impact some 200,000 years ago, which released roughly the energy of the Tunguska explosion of 1908. The crater's floor has a salt pan, where people have gone to gather salt since MSA times. The floor has been cored, with analyses of sediment salinity and pollen, giving a record of climate over the last 200,000 years. For example, a 2007 paper by Kristen and colleagues:

Sediments from Lake Tswaing (25°24'30'' S, 28°04'59'' E) document hydrological changes in southern Africa over the last 200 Ka. Using high-resolution XRF- scanning, basic geochemistry (TIC, TOC, TN), organic petrology and rock-eval pyrolysis, we identify intervals of decreased carbonate precipitation, increased detrital input, decreased salinity and decreased autochthonous (algal and bacterial) organic matter content that represent periods of less stable water column stratification and increased rainfall. Between 200 and 80 Ka BP, these intervals appear to be contemporaneous with local summer insolation maxima, indicating a strong influence of precessional variability (~23 Ka) on African subtropical climate. This influence weakens during the last glacial period (~80 to 10 Ka BP), when humid intervals at 73 to 68 Ka, 54 to 50 Ka, 37 to 35 Ka and 15 to 10 Ka BP are largely out of phase with insolation changes, and presumably reflect southward displacement of the ITCZ (Inter Tropical Convergence Zone) and/or changes in ocean circulation.

I'm pointing to this study because it is one that documents wetter periods during the span between the Howieson's Poort (roughly 60,000 years ago) and the Last Glacial Maximum (around 18,000 years ago). There are some who have claimed that this was a long span of aridity in southern Africa -- but more recent evidence makes it clear that the climate was not unimodal but fluctuated as in earlier and later time frames. Also, the climate was simply not arid, compared to "megadrought" periods documented in East Africa before 70,000 years ago.

Thanks to a reader, I've been reading an excellent 2008 paper by Peter Mitchell, which documents archaeological sites and paleoclimate data leading to the conclusion that habitation in southern Africa was not significantly interrupted during late MSA times. I'll refer to it more extensively later, but in the meantime I'm noting some recent work that Mitchell may not have had available when he was writing his article.

In a similar vein, I can point to an article by Louis Scott and colleagues (2008), which examined pollen records from Tswaing Crater as well as the Wonderkrater spring, and speleothem isotope evidence from Lobatse Cave, Botswana. Correlating the records from different sites in the same general area -- in this case, all from the savanna biome of southern Africa -- is very important. Distant climate records, such as the Greenland or Vostok ice cores, give some indication of global climate fluctuations, but it is not usually obvious how these fluctuations will affect specific regions of the world. One lake sediment core from southern Africa helps to show the local climatic fluctuations, but some may be highly localized, while others may reflect regional water and temperature variations. Hence, the correlations among many sites in a single region allow us to talk about climate fluctuations on a scale relevant to human populations.

Again, the region-wide picture in southern Africa between 60,000 and 20,000 years ago does not yield a picture of static, arid or cool conditions. The time period covers almost two full precessional cycles of insolation in southern Africa, and thus covers a wide range of local climate variation. To sum up, I'll cite Mitchell (2008:54), who relied on some earlier work from the Tswaing Crater record:

Peak annual precipitation may have reached 650–720 mm, with the late MIS 3 peak at the low end of this range. The minimum precipitation experienced was about 535 mm; today’s figure for comparison is 630 mm. Thus, although there were certainly periods when rainfall was reduced compared to the present , such reductions still exceeded by some margin the levels experienced by much of Limpopo Province or the highveld today, and for about a third of the time rainfall was actually higher than at present. Moreover, a generally cooler climate should have reduced evapotranspiration, and thus enhanced effective precipitation, more than these raw estimates suggest.

These sources are relevant for the savanna of the northern and eastern parts of South Africa. The region is ecologically diverse, and Mitchell considers different parts of the region in turn.

References:

Kristen I, Fuhrmann A, Thorpe J, Röhl U, Wilkes H, Oberhänsli H. 2007. Hydrological changes in southern Africa over the last 200 Ka as recorded in lake sediments from the Tswaing impact crater. S Afr J Geol 110:311-326. doi:10.2113/gssajg.110.2-3.311

Mitchell P. 2008. Developing the archaeology of marine isotope stage 3. S Afr Archaeol Soc Goodwin Ser 10:52-65.

Scott L, Holmgren K, Partridge TC. 2008. Reconciliation of vegetation and climatic interpretations of pollen profiles and other regional records from the last 60 thousand years in the Savanna Biome of Southern Africa. Palaeogeography, Palaeoclimatology, Palaeoecology 257:198-206. doi:10.1016/j.palaeo.2007.10.018