As you probably know, I have been working as of late on the puzzle of the early human mtDNA, in the context of Africa, using particularly Behar 2008 as reference.
I published some stuff previously (L0, L1, L2 & L5, L3'4'6 and considerations on Arabian L(xM,N)) and went into a lengthy discussion with Terry about where exactly did the various L lineages coalesce and, of course, human prehistoric boating abilities (an impossible for him before the arrival to almost mythical archipelago of Wallacea, aka Eastern Indonesia).
So I took my time to estimate the composite centroids of each lineage, at every level, from bottom up. Some centroids are plainly valid as likely origins for the major clades, while a few others are more questionable (see below). After a subjective but common-sensical revision of these I came up with the following maps that reconstruct early human expansion:
There seem to be basically two moments of demographic expansion, the first one happened at about 9 (coding region) mutations downstream of the most recent common ancestor (mitochondrial Eve), where the nodes of L0, L1 and L2"6 are in the phylogenetic tree, maybe corresponding with some favorable conditions, which are hard to unravel.
This map shows the earliest expansion at the very phylogenetic origin of the species (purple arrow) and the subsequent three populations generated, defined by the mtDNA lineages L0, L1 and L2"6, all three showing some sign of expansion around the 9th CR mutation (i.e. approximately at the same time, if the molecular clock makes any sense whatsoever).
Of course, L1 and L2"6 shared phylogeny for 5 CR mutations (L1"6), what seems to mean that they had reached Central Africa by that time.
The location that may cause more controversy is that of L0 around Lake Tanganyika, because this lineage is most commonly associated with Khoisan people. However only the L0d1'2 subslineage is totally exclusive of this macro-ethnicity and other L0 lineages are scattered towards the Nile and even as far as Kuwait. The overall centroid, as well as those of L0d and L0a'b'f'k, showed up around there, so I decided that it stands that way, even if admittedly origins around Ethiopia-Sudan or Southern Africa can also be argued for (as did Doron Behar).
The second major expansion is located at around the 20th CR mutation downstream from "Eve" and in my opinion may be correlated with the Abbassia Pluvial, some 120-90,000 years ago, which probably improved living conditions allowing for significative demographic growth.
It is also the prelude of the migration out of Africa into Asia.
At this moment we see L0 finally splitting up clearly between a northern and a southern group, the latter leading to the Khoisan peoples.
We also see signal of expansion of L1c (but not yet L1b), which correlates best with Pygmies (and also some other peoples of the jungle belt essentially).
And we see the split of L2'3'4'6 (the branching of L5 is earlier, just marked it for the record), first between Western (L2) and Eastern (L3'4'6) populations, centered at Central Africa and what seems to be Eritrea respectively. Then L3'4 split up, with L3 expanding rapidly in Ethiopia, Sudan and towards Lake Chad and L4 scattering along the Rift Valley down to Tanzania, where it is still the most important lineage among the Hadza and Sandawe. L3, of course, also expanded eastward towards Asia where its sublineages M and N would find enormous opportunities.
The minor lineage L6 would not expand till a later moment, so at this time it was yet some "private" L3'4'6*. I suspect that it also benefited from the Out of Africa migration, along with L4b and some L0 subclades, because it's highest basal diversity is in Yemen, suggesting it coalesced there.
A word of caution must be said about the area of origin of L2. The actual composite centroid happens to be at the Niger but this location is the product of basal sublineage L2e, which is only represented by one individual in Behar's data, who happens to live at Guinea Bissau. But the major subclade, only diverged from L2 by a single CR mutation (unlike L2e, which has a very long stem), is L2a"d and this one has a clear center at Chad and, secondarily, the CAR.
Similarly, I ended up with a composite centroid at Yemen for L3 after considering M and N. But I can't ignore the absence of basal L3 sublineages in South Arabia, so I decided that the result without M and N is much more likely to reflect the reality.
The actual centroids:
Prior to the above maps, and after hard work, I came up with this preliminary map. The composite centroids here are the "raw" ones, without further consideration:
But, as said above, I have serious caveats about L2, L3 and L6 specially. I think that the areas depicted in the other maps reflect better the likely reality, with L2 coalescing at Chad, L3 at Sudan and L6 most probably at Yemen after the out-of-Africa episode.
As said before, I also have some caveats about the coalescence area of L0 but I can't come up with a clear alternative (it'd be either Ethiopia or Southern Africa but both weight about the same), so I left it that way.
Of course, the ultimate place of origin of Humankind (Homo sapiens), or at least of the most recent common ancestor (mtDNA Eve), remains a mystery. But guess that somewhere in the vast geography of Eastern Africa is a quite reasonable conjecture (Southern and Central Africa are also possibilities).
Erratum: there is an error with minor lineage L5 (thanks to Terry for noticing) but luckily doesn't affect much the overall picture. I made a meaningful error when calculating the centroid of L5a, what, in the maps here pushes the centroid of L5 significantly but not dramatically to the south, to South Sudan at the border with SW Ethiopia to be precise. In turn this affects the centroid of L2"6, L1"6 and "Eve", pushing them also proportionately southwards (not too much but something anyhow). The rest of the nodes are not affected, as they depend only of the downstream geography.
I think it's no big deal but I might correct the maps later on... if my perfectionist side prevails over my lazy one.