Monday, May 20, 2019

What Were Adzebills?

New Zealand is renowned for its unique avifauna, hosting many distinctive bird clades found nowhere else in the world. And prior to the settlement of New Zealand by humans 700-800 years ago, this ensemble of unusual birds would have been even more diverse than it is now. The most famous of these recently lost birds are the moa (a group of large flightless birds) and the Haast's eagle (the largest known raptorial bird) that preyed upon them.

Less well known but no less remarkable were the two species of adzebills. Named after their robust, downcurved bills, these were another group of flightless birds. Although smaller than the largest moa species, adzebills probably weighed around 15-20 kg, making them large by the standards of most birds. The chemical composition of their bones suggests that they ate small animals, which they may have dug out of the ground or decaying wood by using their strong beaks and feet.

Skeleton of a South Island adzebill from the Auckland War Memorial Museum, under CC BY 4.0.

Unlike moa, adzebills clearly weren't paleognaths (a group of birds that also includes ostriches and kiwi), but their relationships to living birds have otherwise been difficult to figure out. Their overall anatomy doesn't obviously resemble any other group of birds; however, it has long been noted that they share similarities with members of the Gruiformes, a group that includes cranes, rails, and their close relatives. An alternative idea that became particularly popular starting in the 1980s is that adzebills were instead more closely related to the kagu, an unusual flightless bird from New Caledonia.

Initially, these ideas probably wouldn't have seemed dramatically discordant, given that the kagu and its closest living relative, the sunbittern of South America, were once thought to be gruiforms. However, with the advent of molecular phylogenetics, current evidence now suggests that they are more closely related to the marine tropicbirds. (In fact, many other ground-dwelling birds, such as bustards and seriemas, were traditionally classified as gruiforms, but are no longer considered members of that group.) Thus, a close relationship with the kagu would put adzebills in a quite different part of the bird family tree compared to potential affinities among gruiforms. Furthermore, a few researchers have even entertained the possibility that adzebills were closely related to yet a different group, the galloanserans (which includes chickens and ducks).

Adzebills went extinct recently enough that some genetic material can be extracted from their remains. Since the 1990s, small snippets of adzebill DNA have been available for study, and analyses that have included these have generally supported the gruiform hypothesis, placing adzebills as close kin of cranes and rails. However, the limited amount of data has prevented a confident assessment of where adzebills belong within Gruiformes.

In a recent study, Alexander Boast and colleagues sequenced nearly complete mitochondrial genomes from the two recent adzebill species. When they analyzed these sequences alongside those of a diverse sampling of other gruiforms, they consistently found a surprising result: the closest living relatives of adzebills are the flufftails, a group of small, rail-like birds from Africa.

Phylogeny of gruiform birds plotted against divergence times estimated by Boast et al. (2019), from their study.

Boast et al. made some novel findings about the relationships among living gruiforms as well. They found that the gray-throated rail (Canirallus oculeus), long thought to be closely related to the flufftail genus Mentocrex, is in fact a true rail. Flufftails in general were formerly considered to be a type of rail, and, especially given the absence of genetic samples for some obscure rail-like birds, it is evident that teasing the two groups apart remains an ongoing process.

But at least the adzebill problem is solved, right? Well... a second study on the phylogenetic position of adzebills has come out this year, and it came to a decidedly different conclusion. For this paper, Grace Musser and Joel Cracraft assessed the likely affinities of adzebills by assembling a new dataset of anatomical features in neornithean birds.

I was quite excited about this study when I first learned of it from Musser's presentation at SVP 2017. Our current understanding of neornithean phylogeny has been greatly refined by analysis of molecular data; however, morphological phylogenetic datasets for neornitheans (i.e.: the only way we can evaluate the phylogenetic relationships of most fossil neornitheans) have remained underdeveloped by comparison. Musser and Cracraft's dataset contains information on 368 skeletal characteristics, making it larger than nearly all other available morphological datasets for neornitheans.

One morphological phylogenetic dataset on neornitheans that is larger than Musser and Cracraft's was published in Livezey and Zusi (2006). In fact, with over 2,900 characters, it is one of the largest morphological phylogenetic datasets of any kind. Although impressive in scope, however, Livezey and Zusi's study has been criticized for containing numerous errors as well as failing to recover many clades that are otherwise well supported by both molecular and morphological studies. I thus appreciated the fact that Musser and Cracraft built much of their dataset by reassessing the characters used in Livezey and Zusi (2006), building upon that previous work while (hopefully) not repeating its mistakes.

So how do the results of Musser and Cracraft's morphological analysis compare to those of recent molecular analyses? They certainly get closer to molecular results than any previous morphological analysis on neornitheans has gotten. Like Jarvis et al. (2014) and Prum et al. (2015) (the two largest recent molecular studies on bird phylogeny), nightjars and their kin were found to be an early-diverging branch among neoavians. Musser and Cracraft's dataset also placed the sunbittern, kagu, and seriemas outside of Gruiformes. Among gruiforms, rails were mostly recovered forming a clade excluding flufftails (though the Nkulengu rail, Himantornis haematopus, was unexpectedly found closer to flufftails), with Canirallus recovered as a true rail like in Boast et al.'s study.

Nonetheless, there are some notable differences. For example, seriemas were not found as close relatives to the other telluravians included in Musser and Cracraft's dataset (in this case vultures and courols), and grebes and loons were recovered as close relatives, a historically popular idea that is now considered outdated. (The convergent diving adaptations shared between grebes and loons have long been recognized as confounding factors in morphological phylogenies of birds.) Analyzing the morphological dataset in combination with a relatively small molecular dataset put some of these relationships (e.g.: the affinities of seriemas) more in line with molecular trees, though even this wasn't enough to pull grebes and loons apart in this study.

Simplified results of the phylogenetic analyses run by Musser and Cracraft (2019) compared to those of recent molecular analyses of modern birds.

Evidently, there is still much work to be done in the field of avian morphological phylogenetics. There are several bird groups that would have been interesting to see included in this study; flamingos, pigeons, and bustards come to mind. Regardless, I consider the assembly of this dataset to be a step in the right direction and the most valuable contribution of Musser and Cracraft's paper, irrespective of what it has to say about adzebills.

Speaking of which, what does it say about adzebills? Musser and Cracraft's analyses are consistent with molecular studies in putting adzebills among gruiforms. However, their results found the closest living relatives of adzebills to be the trumpeters, a group of South American, fruit-eating gruiforms that are more closely related to cranes than to rails. The analyses identified up to 15 similarities between adzebills and trumpeters, particularly in the hip and hindlimb bones. Their results also conflicted with molecular analyses when it came to another extinct New Zealand gruiform: Hawkins's rail (Diaphorapteryx hawkinsi) was found outside of a clade including both rails and flufftails, instead of being a true rail as previous studies suggested. (Note that even though Musser and Cracraft did run a combined molecular-morphological dataset, they did not include any molecular data from adzebills or Diaphorapteryx, as they had used nuclear instead of mitochondrial genes.)

Musser and Cracraft's study was evidently submitted for publication late enough to take into account the findings of Boast et al. (2019), because they explicitly tested the possibility of Boast et al.'s results by running additional analyses in which a close relationship between adzebills and flufftails was enforced. When this was done, the resulting tree was 18 steps longer (i.e.: it implied 18 more evolutionary changes) compared to the phylogenies in which adzebills were closely related to trumpeters. Phylogenetic trees that require so many more steps to explain are often thought to be less likely in phylogenetic studies. For this reason, Musser and Cracraft considered the potential adzebill-flufftail relationship to be inconsistent with their morphological dataset.

However, given that adzebills were aberrant forms that had been isolated from their closest relatives for more than 16 million years, it wouldn't seem farfetched to me if they really had undergone so many evolutionary changes since their last common ancestor with whatever their closest living relatives turn out to be. It's not as though there aren't any anatomical similarities that might indicate an adzebill-flufftail relationship: Musser and Cracraft found 16 shared features between the two groups. Although none of these features are unique to adzebills and flufftails, the same is true of the similarities between adzebills and trumpeters.

What would be particularly helpful for resolving this conundrum would be the discovery of early members of the adzebill lineage that shed light on their ancestral anatomy. Unfortunately, the oldest known fossil adzebills were already fairly similar to recent species, and thus aren't very informative in that regard.

Phylogenetic tree of gruiforms showing the conflicting positions for adzebills found by Boast et al. (2019) and Musser and Cracraft (2019).

Whether adzebills are more closely related to flufftails or trumpeters, either option raises some interesting biogeographic implications, as they imply that the closest living relatives of adzebills are either restricted to Africa (flufftails) or South America (trumpeters). Interestingly, there are parallels for both of these cases among other New Zealand birds: it is now thought that moa were most closely related to the South American tinamous, whereas kiwi are most closely related to the Malagasy elephant birds. What happened in the distant past that led to these unusual distributions in several different groups of birds?

My best guess is that the ancestral groups that gave rise to each of these closely related pairs were once more widespread across the Southern Hemisphere but have since died out, leaving recent descendants only in geographically restricted ranges. Antarctica would be a prime suspect for the place of origin of these groups, with its complete glaciation during the Neogene being a potential cause of extinction for their ancestral populations. Boast et al. (2019) and Musser and Cracraft (2019) both entertain similar possibilities, but we don't yet have the fossils to confirm this scenario. Oh, for a productive deposit of early Paleogene bird fossils from the Southern Hemisphere!

The two recent studies on the affinities of adzebills may not exactly agree with one another, but it's exciting that we've now received two new phylogenetic datasets devoted to resolving this problem in quick succession. If nothing else, we can probably be pretty confident in identifying adzebills as gruiforms now, and we can expect that their closest living relatives are likely one of the other Southern Hemisphere gruiform groups. Furthermore, these new datasets have paved the way for future studies to build upon them (or even combine them together). I certainly foresee Musser and Cracraft's dataset playing a big role in my future research...