Wednesday, September 30, 2015

The Big LGM Bottleneck

The 1000 Genomes paper has been published and has few surprises, given that the data that went into it has mostly been widely used for many years.

The one really striking point made by the paper, however, is that modern humans experienced the most intense bottleneck in the history of the species during the Last Glacial Maximum (LGM) and the several thousand years thereafter (i.e. 20,000 to 15,000 years before present).  The effective population size of each of the non-African populations in the sample at that time depth was under 1,500.  The bottleneck was less intense in Africa where the effective population size was somewhat in excess of 4,500 at the bottom (but still more intense than the bottleneck at any point our species history going all of the way back to Homo Erectus). Basically, our entire species rebooted at that point, while we were still all hunter-gatherers.

Effective population size is not strictly comparable to census population.  There were probably more than 7,000 people alive outside of Africa even at the most dire moment of the LGM, and probably more than 10,000 people alive in Africa at that point.  But, it is quite surprising that the bottleneck during the LGM was more severe, for example, than the bottleneck at the point of Out of Africa migration, or the Toba eruption that may have facilitated the passage of modern humans from South Asia to Southeast Asia and beyond by the Southern route.

The bottleneck around the time of the settling of the New World in the population ancestral to the first wave of Native American migration, of course, is well known.  The number of people belonging to the founding population of the Americas at its lowest point may have numbered in the hundreds, or even less.

Likewise, it is widely known that Northern Europe and Northern Asia was entirely depopulated and covered with a glacier at that point in time with relict populations surviving only in three Southern European refugia (the Franco-Cantabrian refugia, one in Italy, and one in the mountains of far Southeast Europe).

But, the intensity of the bottleneck in Asia was a surprise to me.  South Asia, Southeast Asia and East Asia were not rendered uninhabitable during the Last Glacial Maximum.  They were not buried under a glacier.  The weather in India wouldn't have been all that much worse than the weather in much of Africa.  Yet, South Asians, Southeast Asian, and East Asians suffered a bottleneck just as severe.


Nirjhar007 said...

Andrew, this is just amazing, i never expected it also.

Daniel de França MTd2 said...

It was not a during the Last Glacial Maximum, there was activity going on in many places and it was quite stable temperature wise.

I could be either in the Heinrich event 1 or During the Younger Dryas. But to be bigger than what happened in Toba, I can only think of a meteor impact.

Daniel de França MTd2 said...

What I mean is that a huge tsunami drowned people in many. It is also worth noting that the Linguistic variability of Australia is too low, given the time of its seatlement. Compare with the variety of Papua, just to the north.

andrew said...

"I could be either in the Heinrich event 1 or During the Younger Dryas."

Fair enough. H1 at 16,800 years before present is actually a better fit to the proposed mutation rate date than the LGM which is on the old side of the range at 20,000 years before present. Younger Dryas at roughly 13,000 years before present is on the young side, but is so dramatic that it is worth considering and the Younger Dryas cold snap seems like a more plausible source of mayhem than the H1 heat wave (although H1 was almost equal in the magnitude of the change to Younger Dryas). Given that the LGM and H1 and the Younger Dryas come back to back to back, it is entirely fair to think that the one-two-three punch, rather than any one event individually, may have been what made the bottleneck so extreme.

H1 selected for genetics and cultural traits that were good for high temperatures, presumably selecting against genetic or cultural adaptations better suited for low temperatures that developed during the LGM. Then, these newly hot adapted populations were then immediately faced with the opposite selection pressure after having culled those genes and cultural adaptations best suited for the Younger Dryas situation. Combined, the LGM-H1-Younger Dryas sequence selected for only the most flexible populations in the most climate buffered regions, rather than populations adapted for either extreme.

If the 20,000-15,000 years before present is really just as plus or minus one standard deviation estimate, it could fit either of these events within two standard deviations and honestly we can't claim to be too certain about the margin of error in these mutation rate dates anyway, although whole genome mutation rate dating is probably more consistent than Y-DNA mutation rate dating which probably has more selection pressure and demonstrated considerable variability in mutation rates between loci on that gene.

"it was quite stable temperature wise."

The temperatures were not, in fact, stable in the 20,000-15,000 years BP time frame. In fact, around 15,000 years BP there was a difference in average temperature over a 100 year period equal to about 5°C, which is bigger than at any other point in post-Out of Africa history, and there was no sustained period from 20,000 to 15,000 years BP when the difference in average temperature over a 100 year period was less than about 2.5°C, more than double the levels seen in the Holocene except for the last 500 years or so (during which global temperature has steadily risen due to air pollution but hasn't fluctuated in such a dramatic fashion both up and down).

"to be bigger than what happened in Toba, I can only think of a meteor impact."

This would certainly be a viable theory to explain the Younger Dryas. I can imagine others, but a detailed examination of the causes of the LGM, H1 or the Younger Dryas, let alone all three, is beyond the scope of this particular post.

"a huge tsunami drowned people"

Color me skeptical on that point. The likelihood that a sufficient percentage of people live that close to the coast, and the fact that there is almost always considerable region variation due to ocean floor topography makes that unlikely to be a major cause at a point in time when humans had a range from Tasmania to Korea to Japan to Iberia. Also, even a one time 50% cull in global population wouldn't have that much impact. The Black Plague in Europe, for example, didn't do too much to change the effective population size of Europe. Similarly, the European diseases that caused genocidal deaths at first contact with the Native Americans wasn't enough to reduce effective population size for Native Americans all that much. On an order of magnitude level, you need a sustained and severe event that lasts for many (at least 3-4 and probably more) generations and 95%-100% wipes out most populations, to get the kind bottleneck that appears in the collective history book of our genes.

andrew said...

"the Linguistic variability of Australia is too low, given the time of its seatlement. Compare with the variety of Papua, just to the north."

Not convinced on this score. Australia is a huge continent, but the portion of it that is well suited for sustainable and continuous human habitation is much smaller, and there are relatively few insurmountable geographic barriers in the region where Australia displayed low linguistic variability. Continuously interacting populations without strong geographic barriers in a reasonably compact core area are going to from a single linguistic community.

In contrast, Papua New Guinea is incredibly lush by comparison, has several ecological niches that vary with altitude, and has mountains and valleys that create natural barriers isolating particular linguistic areas. Also, some of the apparent diversity in Papuan languages is due to a lack of resources in the form of professionally trained linguists and a lack of data on languages in one of the few parts of the world where there are many uncontacted populations, that will likely end up being linked into language families and macro-language families over time. Right now, the focus for linguists who are in triage mode there is on documenting languages before they go extinct. It has taken a couple of centuries for linguists to put together some of the major families of indigenous American languages (particularly within the Amerind grouping) and it may be several decades before the lumpers who face the more onerous burden of proof in linguistics overcome the splitters when it comes to the classification of the Papuan languages.

Daniel de França MTd2 said...

Araound 15bp there were variations due Old Dryas and Bolling Allerod. I am not sure if these were classified as the last glacial maximum. I think more in the period after H2 and before H1 , at least, these are what I have in mind. I thought of H1 due the cultural discontinuity between Solutrean and Magdalenian.

I am thinking more of a deep penetration of tusnamis of higher than 1 km high, enough to send waves across Australia, up to Himalayas on India and deep into Africa. That and the severe winter by sending dust into air. I think there was likely a sequence of impacts.

The idea is that, like Mars' moons, Earth had moons and some rings, which were slowing losing energy and falling, like Phobos:

Notice that it will break apart when it gets closer to earth and notice how small that period is in geological time. So, probably, humans had several bottlenecks.

Besides asteroid, I don't know what would cause such a bottleneck.

Daniel de França MTd2 said...

The position of Dixon is pretty much unique . As for Papuan languages, they are indeed classified in macro groups, though the classification parameters are the relationship between pronouns. It seems like they get chucks with as much variation as chunks within nostratic. As far as I know, there is an on going effort to study them. I wouldn't say it would take centuries.

andrew said...

"I am thinking more of a deep penetration of tsunamis of higher than 1 km high, enough to send waves across Australia, up to Himalayas on India and deep into Africa."

This is a basically impossible scenario outside of Hollywood movie studio special effects. The only impact Earth has ever had of that would have a sufficient scale to do that was the impact that created the moon something like 4 billion years ago (which, of course, was even more intense than that).

The impact that killed the off the dinosaurs and cleared the decks for mammals 60 mya is the only known tsunami that had a wave height higher than about 700 meters (it was 5 km), and even huge megatsunamis don't go all that far inland. The k-t boundary tsunami might have made it 125 km inland.

A tsunami that was 1 km high might make it 25km inland before petering out. Basically, once a tsunami makes landfall it has no more water to supply it so its wave height falls as a function of 1/d where d is the distance from shore (it actually falls off faster than that due to rising land height as you move inland and because some of the water retreats back to sea or doesn't make the entire trip inland or is absorbed by indentations in the terrain).

There is no doubt that an extra-terrestrial impact on Earth ca. 20,000-15,000 years ago (if there was one or a series of such events) was much, much smaller than the k-t boundary impact. Scientists may not have documented every significant extraterrestrial impact in the last 2 million years or so, but there is no way that they could have missed anything that big and any tsunami that had that much of an impact on humans would have had a bigger impact on other flora and fauna than anything that happened around then other than hunting from humans. It wouldn't have been specific to megafauna as the fossil record indicates.

The 1958 Lituya Bay megatsunami in Alaska is the largest recorded tsunami and was about 524 meters of wave height due to funneling down a long, narrow and shallow bay. Without something to channel it, the waves don't get as high.

andrew said...

An illustration of a 300 foot tsunami that hit the entire East Coast complete with animation is found here.

Also, no impact anywhere on Earth, no matter how big, could create that kind of tsunami on all of the shorelines on Earth. A strike in the Indian Ocean, for example, would spare the East Asian and Siberian coasts, Japan, most of the coast of Australia and the coast of Vietnam and either the west coast of India, and the coast of Africa that is shielded by Madagascar which would have been uninhabited at the time as well as the West coast and Southern coasts of Africa, the Mediterranean coast, the Atlantic coast of Europe. And, the more centrally such a tsunami was located (so as to have an impact on more places) the smaller it would be when it arrived as the radius of the tsunami horizon increased. The worst tsunamis usually involve a shift at the continental shelf close and have a fairly focused area of land impact.

andrew said...

The vast majority of deadly tsunamis (99%) strike within 250km of the source of the event. And, very few tsunamis have an area of innundation more than 10 miles - even 6 km inland is extraordinary and requires special geography in the affected area. You can scale this up for a bigger event, but only so much. An omni-directional tsunami event caused by a meteor is going to get weaker as a function of 1/r with distance from the impact, and impact strength is going to impact the area of innundation. Delta communities might be wiped out, but not a huge share of the entire population of humanity in Asia.

Daniel de França MTd2 said...

I found this, in your link, perhaps the last of the asteroids that hit earth: Scientists believe that an asteroid struck the Indian Ocean about 4,800 years ago. The tsunami that resulted is theorized to have been approximately 600 feet (180 m) high.

I don't think there were tsunamis everywhere, but there were tropical places even during the harshest times of the ice age. I cannot come up with a way of killing these people other than literally drowning them. People outside tropics can be killed by a cold event, following this.

andrew said...

You start with a bunch of people who learn how to survive with plants and animals that thrive in a sunny tropical environment. Then, you have dust in the air obscuring sunlight and lower temperatures that turns their world into a cloudy, subtropical environment over the scope of a few decades or a century.

The ecosystem gets out of whack and their traditional plants and animals are no longer present in sufficient numbers to support their traditional means of subsistence and starvation and illness vulnerability associated with malnutrition follow.

A few isolated groups of humans with innovative hunting and gathering geniuses in their communities discover different plants and animals that they can cobble together now that their historic diet isn't sufficient to feed their community, and so they don't die.

A few other isolated groups of humans live in places where conditions are optimal and so the same plant and animal species are around but there isn't as much food available to hunt and gather, but they have very low population density to start with, and manage to reduce their population density further without falling below the critical mass to sustain the species (perhaps by killing off other tribes who are competing with them for resources or perhaps with gradually shrinking numbers of children who survive to adulthood); so their effective population size shrinks, but they don't go extinct.

But, the human communities that fail to innovate as starvation and malnutrition based illness thins their numbers (the vast majority) and who falling in population density so quickly that their community size falls below the critical mass necessary to hunt, to gather, to defend themselves from other predatory species, and to guard pregnant women, and to raise children into functional adults at a rate high enough to replace themselves go extinct.

Daniel de França MTd2 said...

That's a situation similar to famines, which I've never seen to be so deadly. Maybe we could compare with other bottlenecks. For example, it seems there was bottlenecks which happened everywhere around 3000bc, which more or less coincides with the crater I pointed in the link you provided. It seems that around this time there was also a large inundation that affected Mesopotamia.

andrew said...

Twelve famines have killed 15%-50% of the population in the affected areas in the last 350 years alone. One in China (back when its population was much smaller than it is today) killed 60 million people. And, the last 350 years have really been quiet decedent climate-wise by historical standards. And, people were much more technologically advanced for pretty much all of the last 350 years than they were 13,000-20,000 years ago, during which there were multiple successive climate events in close succession that were profoundly worse than anything humanity has experienced in all of history.

Famines (often with associated plagues) are far more deadly than almost any other natural disaster, and are much more long lived and broad in geographic extent than almost any other kind of natural disaster. The geographic scope of a typical drought is hundreds of times greater than the geographic scope of a flood.

Daniel de França MTd2 said...

I am not sure if there would be a considerable difference between a famine and a unique case of bad climate events. Even if there were many of them, the growth until saturation of the carrying capacity doesn't take too long, maybe a few centuries, and I think they were quite interspersed to allow the growth.
The carrying capacity of hunters and gathers is around 0.1 - 1 people/km^2, so the world population was around a few million. You would need an event that would kill 99.9% of the people.

andrew said...

The effective population size of modern humans outside Africa prior to the LGM was probably on the order of 4500 or so. Reducing effective population size by 65% or so doesn't require you to kill 99.9% of the people, even though it does require you to kill more than 65% of them unless the population is perfectly structured. I suspect that a well placed 90%-95% culling of population could be quite sufficient to get a 65% reduction in effective population size.

Daniel de França MTd2 said...

Why would the population be so low? That's not what we see in hunter and gather societies nowadays. And there is no reason to be the case back then. There was nothing stopping the growth of population much lower than the carrying capacity. In Australia, the pre colonization estimates for the population were bigger than 200.000 people.

andrew said...

Effective population size and census population are two very different things. First, effective population excludes individuals who aren't of reproductive age (children and post-menopausal women).

Second, effective population size is the harmonic mean of the population over a long period of time which in practice is equivalent to somewhat more than the smallest population that an unstructured population has had at its lowest point. The effective population of the several billion people on Earth today is in the 10s of thousands.

In an unstructured population, the effective population isn't reduced much until the census population reaches something close to twice the all time low of the effective population.

But, if the population is structured (i.e. if particular lineages are sorted into separate populations), then the effective population size of the subpopulations will be some fraction of the effective population size of the pre-schism source population, and the effective population of the whole is reduced when a subpopulation's census population gets close to twice the effective population size of the subpopulation.

Daniel de França MTd2 said...

An harmonic sum will yield a result with a bias to the lowest values a few drops won't make much of a difference. The number of generations is too big, and as I mentioned, it doesn't take many of them to achieve a saturation, since the growth is exponential.

You'd also have to consider the time depth involved to define "today".

andrew said...

A new megatsunami from 73,000 years ago has just been discovered about 30 miles from its source which was about 800 feet high and made it about 1/2 mile inland.