## No, David Attenborough, you can’t tell how heavy a dinosaur was from the circumference of its femur

### January 12, 2016

I was a bit disappointed to hear David Attenborough on BBC Radio 4 this morning, while trailing a forthcoming documentary, telling the interviewing that you can determine the mass of an extinct animal by measuring the circumference of its femur.

We all know what he was alluding to, of course: the idea first published by Anderson et al. (1985) that if you measure the life masses of lots of animals, then measuring their long-bone circumferences when they’ve died, you can plot the two measurements against each other, find a best-fit line, and extrapolate it to estimate the masses of dinosaurs based on their limb-bone measurements.

This approach has been extensively refined since 1985, most recently by Benson et al. (2014). but the principle is the same.

But the thing is, as Anderson et al. and other authors have made clear, the error-bars on this method are substantial. It’s not super-clear in the image above (Fig 1. from the Anderson et al. paper) because log-10 scales are used, but the 95% confidence interval is about 42 pixels tall, compared with 220 pixels for an order of magnitude (i.e. an increment of 1.0 on the log-10 scale). That means the interval is 42/220 = 0.2 of an order of magnitude. That’s a factor 10 ^ 0.2 = 1.58. In other words you could have two animals with equally robust femora, one of them nearly 60% heavier than the other, and they would both fall within the 95% confidence interval.

I’m surprised that someone as experienced and knowledgeable as Attenborough would perpetuate the idea that you can measure mass with any precision in this way (even more so when using only a femur, rather than the femur+humerus combo of Anderson et al.)

More: when the presenter told him that not all scientists buy the idea that the new titanosaur is the biggest known, he said that came as a surprise. Again, it’s disappointing that the documentary researchers didn’t make Attenborough aware of, for example, Paul Barrett’s cautionary comments or Matt Wedel’s carefully argued dissent. Ten minutes of simple research would have found this post — for example, it’s Google’s fourth hit for “how big is the new argentinian titanosaur”. I can only hope that the actual documentary, which screens on Sunday 24 January, doesn’t present the new titanosaur’s mass as a known and agreed number.

(To be clear, I am not blaming Attenborough for any of this. He is a presenter, not a palaeontologist, and should have been properly prepped by the researchers for the programme he’s fronting. He is also what can only be described as 89, so should be forgiven if he’s not quite as quick on his feet when confronted with an interviewer as he used to be.)

# Update 1 (the next day)

Thanks to Victoria Arbour for pointing out an important reference that I missed: it was Campione and Evans (2012) who expanding Anderson et al.’s dataset and came up with the revised equation which Benson et al. used.

# Update 2 (same day as #1)

It seems most commenters are inclined to go with Attenborough on this. That’s a surprise to me — I wonder whether he’s getting a free pass because of who he is. All I can say is that as I listened to the segment it struck me as really misleading. You can listen to it for yourself here if you’re in the UK; otherwise you’ll have to make do with this transcript:

“It’s surprising how much information you can get from just one bone. I mean for example that thigh bone, eight feet or so long, if you measure the circumference of that, you will be able to say how much weight that could have carried, because you know what the strength of bone is. So the estimate of weight is really pretty accurate and the thought is that this is something around over seventy tonnes in weight.”

(Note also that the Anderson et al./Campione and Evans method has absolutely nothing to do with the strength of bone.)

How long it was depends on whether you think it held its neck out horizontally or vertically. If it held it out horizontally, well then it would be about half as big again as the Diplodocus, which is the dinosaur that’s in the hall of the Natural History Museum. It would be absolutely huge.

Interviewer: And how tall, if we do all the dimensions?

Ah well that is again the question of how it holds its neck, and it could have certainly reached up about to the size of a four or five storey building.

Needless to say, the matter of neck posture is very relevant to our interests. I don’t want to read too much into a couple of throwaway comments, but the implication does seem to be that this is an issue that the documentary might spend some time on. We’ll see what happens.

# References

## How light could a giant azhdarchid be?

### October 19, 2015

I imagine that by now, everyone who reads this blog is familiar with Mark Witton’s painting of a giant azhdarchid pterosaur alongside a big giraffe. Here it is, for those who haven’t seen it:

(This is the fifth and most recent version that Mark has created, taken from 9 things you may not know about giant azhdarchid pterosaurs.)

It’s one of those images that really kicks you in the brain the first time you see it. The idea that an animal the size of a giraffe could fly under its own power seems ludicrous — yet that’s what the evidence tells us.

But wait — what do we mean by “an animal the size of a giraffe”? Yes, the pterosaur in this image is the same height as the giraffe, but how does its weight compare?

Mark says “The giraffe is a big bull Masai individual, standing a healthy 5.6 m tall, close to the maximum known Masai giraffe height.” He doesn’t give a mass, but Wikipedia, citing Owen-Smith (1988), says “Fully grown giraffes stand 5–6 m (16–20 ft) tall, with males taller than females. The average weight is 1,192 kg (2,628 lb) for an adult male and 828 kg (1,825 lb) for an adult female with maximum weights of 1,930 kg (4,250 lb) and 1,180 kg (2,600 lb) having been recorded for males and females, respectively.” So it seems reasonable to use a mass intermediate between those of an average and maximum-sized male, (1192+1930)/2 = 1561 kg.

So much for the giraffe. What does the azhdarchid weigh? The literature is studded with figures that vary wildly, from the 544 kg that Henderson (2010) found for Quetzalcoatlus, right down to the widely cited 70 kg that Chatterjee and Templin (2004) found for the same individual — and even the astonishing 50 kg that seems to be favoured by Unwin (2005:192). In the middle is the 259 kg of Witton (2008).

It occurred to me that I could visualise these mass estimates by shrinking the giraffe in Mark’s image down to the various proposed masses, and seeing how credible it looks to imagine these reduced-sized giraffes weighting the same as the azhdarchid. The maths is simple. For each proposed azhdarchid mass, we figure out what it is as a proportion of the giraffe’s 1561 kg; then the cube root of that mass proportion gives us the linear proportion.

• 544 kg = 0.389 giraffe masses = 0.704 giraffe lengths
• 259 kg = 0.166 giraffe masses = 0.549 giraffe lengths
• 70 kg =0.0448 giraffe masses = 0.355 giraffe lengths

Let’s see how that looks.

On the left, we have Mark’s artwork, with the giraffe massing 1561 kg. On the right, we have three smaller (isometrically scaled) giraffes of masses corresponding to giant azhdarchid mass estimates in the literature. If Don Henderson (2010) is right, then the pterosaur weighs the same as the 544 kg giraffe, which to me looks pretty feasible if it’s very pneumatic. If Witton (2008) is right, then it weighs the same as the 259 kg giraffe, which I find hard to swallow. And if Chatterjee and Templin (2004) are right, then the giant pterosaur weighs the same as the teeny tiny 70 kg giraffe, which I find frankly ludicrous. (For that matter, 70 kg is in the same size-class as Georgia, the human scale-bar: the idea that she and the pterosaur weigh the same is just silly.)

What is the value of such eyeball comparisons? I’m not sure, beyond a basic reality check. Running this exercise has certainly made me sceptical about even the 250 kg mass range which now seems to be fairly widely accepted among pterosaur workers. Remember, if that mass is correct then the pterosaur and the 259 kg giraffe in the picture above weight the same. Can you buy that?

Or can we find extant analogues? Are there birds and mammals with the same mass that are in the same size relation as these images show?

# References

• Chatterjee, Sankar, and R. J. Templin. 2004. Posture, locomotion, and paleoecology of pterosaurs. Geological Society of America, Special Paper 376. 68 pages.
• Henderson, Donald M. 2010. Pterosaur body mass estimates from three-dimensional mathematical slicing. Journal of Vertebrate Paleontology 30(3):768-785.
• Witton, Mark P. 2008. A new approach to determining pterosaur body mass and its implications for pterosaur flight. Zitteliana 28:143-159.

## Sauropods’ neutral neck postures were really weird

### November 5, 2014

Last night, I submitted a paper for publication — for the first time since April 2013. I’d almost forgotten what it felt like. But, because we’re living in the Shiny Digital Future, you don’t have to wait till it’s been through review and formal publication to read it. I submitted to PeerJ, and at the same time, made it available as a preprint (Taylor 2014).

It’s called “Quantifying the effect of intervertebral cartilage on neutral posture in the necks of sauropod dinosaurs”, and frankly the results are weird. Here’s a taste:

Taylor (2014:figure 3). Effect of adding cartilage to the neutral pose of the neck of Apatosaurus louisae CM 3018. Images of vertebra from Gilmore (1936:plate XXIV). At the bottom, the vertebrae are composed in a horizontal posture. Superimposed, the same vertebrae are shown inclined by the additional extension angles indicated in Table 1. If the slightly sub-horizontal osteological neutral pose of Stevens and Parrish (1999) is correct, then the cartilaginous neutral pose would be correspondingly slightly lower than depicted here, but still much closer to the elevated posture than to horizontal. (Note that the posture shown here would not have been the habitual posture in life: see discussion.)

A year back, as I was composing a blog-post about our neck-cartilage paper in PLOS ONE (Taylor and Wedel 2013c), I found myself writing down the rather trivial formula for the additional angle of extension at an intervertebral joint once the cartilage is taken into account. In that post, I finished with the promise “I guess that will have to go in a followup now”. Amazingly it’s taken me a year to get that one-pager written and submitted. (Although in the usual way of things, the manuscript ended up being 13 pages long.)

To summarise the main point of the paper: when you insert cartilage of thickness t between two vertebrae whose zygapophyses articulate at height h above the centra, the more anterior vertebra is forced upwards by t/h radians. Our best guess for how much cartilage is between the adjacent vertebrae in an Apatosaurus neck is about 10% of centrum length: the image above shows the effect of inserting that much cartilage at each joint.

And yes, it’s weird. But it’s where the data leads me, so I think it would be dishonest not to publish it.

I’ll be interested to see what the reviewers make of this. You are all of course welcome to leave comments on the preprint itself; but because this is going through conventional peer-review straight away (unlike our Barosaurus preprint), there’s no need to offer the kind of detailed and comprehensive comment that several people did with the previous one. Of course feel free if you wish, but I’m not depending on it.

# References

Gilmore Charles W. 1936. Osteology of Apatosaurus, with special reference to specimens in the Carnegie Museum. Memoirs of the Carnegie Museum 11:175–300 and plates XXI–XXXIV.

Stevens, Kent A., and J. Michael Parrish. 1999. Neck posture and feeding habits of two Jurassic sauropod dinosaurs. Science 284(5415):798–800. doi:10.1126/science.284.5415.798

Taylor, Michael P. 2014. Quantifying the effect of intervertebral cartilage on neutral posture in the necks of sauropod dinosaurs. PeerJ PrePrints 2:e588v1 doi:10.7287/peerj.preprints.588v1

Taylor, Michael P., and Mathew J. Wedel. 2013c. The effect of intervertebral cartilage on neutral posture and range of motion in the necks of sauropod dinosaurs. PLOS ONE 8(10):e78214. 17 pages. doi:10.1371/journal.pone.0078214

## Gender balance at SVPCA

### September 17, 2014

I’ve always thought of SVPCA as a pretty well gender-balanced conference: if not 50-50 men and women, then no more than 60-40 slanted towards men. So imagine my surprise when I ran the actual numbers.

1. Delegates. I went through the delegate list at the back of the abstracts book, counting the men and women. Those I knew, or whose name made it obvious, I noted down; the half-dozen that I couldn’t easily categorise, I have successfully stalked on the Internet. So I now know that there were 39 women and 79 men — so that women made up 33% of the delegates, almost exactly one third.

Official conference photo, SVPCA 2014, York, UK.

2. Presentations. There were a total of 50 presentations in the three days of SVPCA: 18 on days 1 and 3, and 14 on day 2, which had a poster session in place of the final session of four talks. I counted the presenters (which were usually, but not always, the lead authors). Here’s how the number of talks by women broke down:

Day one: 2 of 18
Day two: 8 of 14
Day three: 3 of 18

In total, this gives us 13 of 50 talks by women, or 26%.

3. Presenter:delegate ratios. Since 37 of the 79 attending men gave talks, that’s 47% of them; but only 13 of the 39 attending women gave talks, which is 33%. On other words, a man attending SVPCA was 40% more likely to give a talk than a woman.

I’m not sure what to make of all this. I was shocked when I found that only one ninth of the first day’s talks were by women. It’s a statistical oddity that women actually dominated day two, but day three was nearly as unbalanced as day one.

Since SVPCA accepts pretty much every submitted talk, the conference itself can’t be blamed for the imbalance. (At least, not unless you think the organisers should turn down talks by men just because they’re men, leaving blank spots in the program.) It seems that the imbalance more likely reflects that of the field in general. Maybe more disturbing is that the proportion of women giving talks was rather less than the proportion attending (26% vs. 33%) which suggests that perhaps women feel more confident about attending than about presenting.

It would be interesting to know how these numbers compare with SVP’s.

## How long was the torso of Dreadnoughtus?

### September 15, 2014

In a comment on the last post, on the mass of Dreadnoughtus, Asier Larramendi wrote:

The body mass should be considerably lower because the reconstructed column don’t match with published vertebrae centra lengths. 3D reconstruction also leaves too much space between vertebrae. The reconstruction body trunk is probably 15-20% longer than it really was. Check the supplementary material: http://www.nature.com/srep/2014/140904/srep06196/extref/srep06196-s1.pdf

So I did. The table of measurements in the supplementary material is admirably complete. For all of the available dorsal vertebrae except D9, which I suppose must have been too poorly preserved to measure the difference, Lacovara et al. list both the total centrum length and the centrum length minus the anterior condyle. Centrum length minus the condyle is what in my disseration I referred to as “functional length”, since it’s the length that the vertebra actually contributes to the articulated series, assuming that the condyle of one vertebra sticks out about as far as the cotyle is recessed on the next vertebra. Here are total lengths/functional lengths/differences for the seven preserved dorsals, in mm:

• D4 – 400/305/95
• D5 – 470/320/150
• D6 – 200/180/20
• D7 – 300/260/40
• D8 – 350/270/80
• D9 – 410/ – / –
• D10 – 330/225/105

The average difference between functional length and total length is 82 mm. If we apply that to D9 to estimate it’s functional length, we get 330mm. The summed functional lengths of the seven preserved vertebrae are then 1890 mm. What about the missing D1-D3? Since the charge is that Lacovara et al. (2014) restored Dreadnoughtus with a too-long torso, we should be as generous as possible in estimating the lengths of the missing dorsals. In Malawisaurus the centrum lengths of D1-D3 are all less than or equal to that of D4, which is the longest vertebra in the series (Gomani 2005: table 3), so it seems simplest here to assign D1-D3 functional lengths of 320 mm. That brings the total functional length of the dorsal vertebral column to 2850 mm, or 2.85 m.

At this point on my first pass, I was thinking that Lacovara et al. (2014) were in trouble. In the skeletal reconstruction that I used for the GDI work in the last post, I measured the length of the dorsal vertebral column as 149 pixels. Divided by 36 px/m gives a summed dorsal length of 4.1 m. That’s more than 40% longer than the summed functional lengths of the vertebrae calculated above (4.1/2.85 = 1.44). Had Lacovara et al. really blown it that badly?

Before we can rule on that, we have to estimate how much cartilage separated the dorsal vertebrae. This is a subject of more than passing interest here at SV-POW! Towers–the only applicable data I know of are the measurements of intervertebral spacing in two juvenile apatosaurs that Mike and I reported in our cartilage paper last year (Taylor and Wedel 2013: table 3, and see this post). We found that the invertebral cartilage thickness equaled 15-24% of the length of the centra.* For the estimated 2.85-meter dorsal column of Dreadnoughtus, that means 43-68 cm of cartilage (4.3-6.8 cm of cartilage per joint), for an in vivo dorsal column length of 3.28-3.53 meters. That’s still about 15-20% shorter than the 4.1 meters I measured from the skeletal recon–and, I must note, exactly what Asier stated in his comment. All my noodling has accomplished is to verify that his presumably off-the-cuff estimate was spot on. But is that a big deal?

Visually, a 20% shorter torso makes a small but noticeable difference. Check out the original reconstruction (top) with the 20%-shorter-torso version (bottom):

FWIW, the bottom version looks a lot more plausible to my eye–I hadn’t realized quite how weiner-dog-y the original recon is until I saw it next to the shortened version.

In terms of body mass, the difference is major. You’ll recall that I estimated the torso volume of Dreadnoughtus at 32 cubic meters. Lopping off 20% means losing 6.4 cubic meters–about the same volume as a big bull elephant, or all four of Dreadnoughtus‘s limbs put together. Even assuming a low whole-body density of 0.7 g/cm^3, that’s 4.5 metric tons off the estimated mass. So a ~30-ton Dreadnoughtus is looking more plausible by the minute.

For more on how torso length can affect the visual appearance and estimated mass of an animal, see this post and Taylor (2009).

* I asked Mike to do a review pass on this post before I published, and regarding the intervertebral spacing derived from the juvenile apatosaurs, he wrote:

That 15-24% is for juveniles. For the cervicals of adult Sauroposeidon we got about 5%. Why the differences? Three reasons might be relevant: 1, taxonomic difference between Sauroposeidon and Apatosaurus; 2, serial difference between neck and torso; 3, ontogenetic difference between juvenile and adult. By applying the juvenile Apatosaurus dorsal measurement directly to the adult Dreadnoughtus dorsals, you’re implicitly assuming that the adult/juvenile axis is irrelevant (which seems unlikely to me), that the taxonomic axis is (I guess) unknowable, and that the cervical/dorsal distinction is the only one that matter.

That’s a solid point, and it deserves a post of its own, which I’m already working on. For now, it seems intuitively obvious to me that we got a low percentage on Sauroposeidon simply because the vertebrae are so long. If the length-to-diameter ratio was 2.5 instead of 5, we’d have gotten 10%, unless cartilage thickness scales with centrum length, which seems unlikely. For a dorsal with EI of 1.5, cartilage thickness would then be 20%, which is about what I figured above.

Now, admittedly that is arm-waving, not science (and really just a wordy restatement of his point #2). The obvious thing to do is take all of our data and see if intervertebral spacing is more closely correlated with centrum length or centrum diameter. Now that it’s occurred to me, it seems very silly not to have done that in the actual paper. And I will do that very thing in an upcoming post. For now I’ll just note three things:

1. As you can see from figure 15 in our cartilage paper, in the opisthocoelous anterior dorsals of CM 3390, the condyle of the posterior vertebra is firmly engaged in the cotyle of the anterior one, and if anything the two vertebrae look jammed together, not drifted apart. But the intervertebral spacing as a fraction of centrum length is still huge (20+4%) because the centra are so short.
2. Transferring these numbers to Dreadnoughtus only results in 4.3-6.8 cm of cartilage between adjacent vertebrae, which does not seem unreasonable for a 30- or 40-ton animal with dorsal centra averaging 35 cm in diameter. If you asked me off the cuff what I thought a reasonable intervertebral spacing was for such a large animal, I would have said 3 or 4 inches (7.5 to 10 cm), so the numbers I got through cross-scaling are actually lower than what I would have guessed.
3. Finally, if I’ve overestimated the intervertebral spacing, then the actual torso length of Dreadnoughtus was even shorter than that illustrated above, and the volumetric mass estimate would be smaller still. So in going with relatively thick cartilage, I’m being as generous as possible to the Lacovara et al. (2014) skeletal reconstruction (and indirectly to their super-high allometry-derived mass estimate), which I think is only fair.

### September 11, 2014

In the paper describing the new giant titanosaur Dreadnoughtus, Lacovara et al. (2014) use the limb bone allometry equation of Campione and Evans (2012) to derive a mass estimate for the holotype individual of 59.3 metric tons. This is presumably the “middle of the road” value spat out by the equation; the 95% confidence interval on either side probably goes from 40 to 80 metric tons or maybe even wider.

I decided to see if 59 metric tons was plausible for Dreadnoughtus by doing Graphic Double Integration (GDI) on the published skeletal reconstruction and body outline (Lacovara et al. 2014: fig. 2). The image above is the one I used, so if you like, you can check my numbers or try your hand at GDI and see what you get.

First up, I have to congratulate Lacovara et al. for the rare feat of having everything pretty much to scale, and a properly-sized scale bar. This is not always the case. Presumably having a 3D digital model of the reconstructed skeleton helped — and BTW, if you haven’t downloaded the 3D PDFs and played with them, you are missing out bigtime.

Here are my measurements of various bits in the picture and the scale factors they give:

Meter scale bar: 37 pixels – 1.0 meters – 37 px/m
Human figure: 66 pixels – 1.8 meters – 37 px/m
Scapula: 62 pixels – 1.7 meters – 36 px/m
Humerus: 58 pixels – 1.6 meters – 36 px/m
Femur: 70 pixels – 1.9 meters – 37 px/m
Cervical: 45 pixels – 1.1 meters – 41 px/m (not included in average*)
Neck: 407 pixels – 11.3 meters – 36 px/m
Post-cervical vertebral column: 512 pixels – 13.8 meters – 37 px/m
Total length: 922 pixels – 26.0 meters – 35 px/m
AVERAGE 36 px/m

* I didn’t include the cervical because when I measured it I sorta guessed about where the condyle was supposed to be. That was the odd measurement out, and I didn’t want to tar Lacovara et al. for what might well be my own observer error.

Here’s the chopped-up Dreadnoughtus I used for my estimate. Just for the heck of it, for the first time out I assigned all of the body regions circular cross-sections. We’ll come back to how realistic this is later. Here’s what I got for the volumes of the various bits:

Neck: 13.9
Body: 32.1
Tail: 4.0
Limbs: 6.8
TOTAL: 57.0 m^3

Okay, this is looking pretty good, right? Lacovara et al. (2014) got 59.3 metric tons using limb allometry, I got a volume of 57 cubic meters using GDI. If Dreadnoughtus was the same density as water — 1 metric ton per cubic meter — then my estimated mass would be 57 tons, which is crazy close given all of the uncertainties involved.

BUT there are a couple of big buts involved. The first is that a lot of sauropods had distinctly non-round body cross-sections (Diplodocus, Camarasaurus). So assuming circular cross-sections might inflate the body well beyond its likely volume. Second is that sauropods were probably much less dense than water (discussed here, here, and here, and see Wedel 2005 for the full scoop). What are the implications for Dreadnoughtus?

### Round and Round

It turns out that circular cross-sections are probably defensible for some parts of Dreadnoughtus. By playing around with the 3D PDF of the assembled skeleton I was able to get these orthogonal views:

I don’t remember what the pixel counts were for the max height and max width of the torso, but they were pretty close. I measured at several points, too: front of the pelvis, max extent of ribcage, mid-scap. This is probably not super-surprising as the fatness of titanosaurs has been widely noted before this. Here’s a cross-section through the torso of Opisthocoelicaudia at D4 (Borsuk-Bialynicka 1977: fig. 5) — compare to the more taconic forms of Diplodocus and Camarasaurus linked above.

Okay, a round torso on Dreadnoughtus I can buy. A round neck and tail, not so much. Look at the skeletal recon and you can see that even with a generous allowance for caudofemoralis muscles on the tail, and diapophyses on the cervical vertebrae, no way were those extremities circular in cross-section. Just off the cuff I think a width:height ratio of 2:3 is probably about right.

But there are some body regions that probably were round, or close enough as to have made no difference, like the head and limbs. So I actually toted up the volume three times: once with circular cross-sections throughout (probably too fat), once with a 2:3 width:height ratio in the neck, trunk, and tail (probably too thin, at least in the torso), and once with the 2:3 ratio only in the neck and tail (my Goldilocks version). Here are the numbers I got:

### Air Apparent?

Now, for density. Birds are usually much less dense than water — lotsa cited data in this hummingbird post, the punchline of which is that the average whole-body density of a bunch of birds is 0.73 g/cm^3. Why so light? In part because the lungs and air sacs are huge, and account for 15-20% of the whole-body volume, and in part because many of the bones are pneumatic (= air-filled). For a really visceral look at how much air there can be in the bones of birds, see this post, and this one and this one for sauropods.

In my 2005 paper (almost a decade old already — gosh!), I found that for Diplodocus, even a fairly conservative estimate suggested that air inside the bones accounted for about 10% of the volume of the whole animal in life. That may be higher than in a lot of birds, because sauropods were corn-on-the-cob, not shish-kebabs. And that’s just the air in the bones — we also have several lines of evidence suggesting that sauropods had air-sacs like those of birds (Wedel 2009). If the lungs and air sacs occupied 15% of the volume of the whole animal, and the air in the bones occupied another 10%, that would give a whole-body density pretty close to the 0.73 g/cm^3 found for birds. Sauropods might have been lighter still — I didn’t include visceral, intermuscular, or subcutaneous diverticula in my calculations, because I couldn’t think of any way to constrain their volumes.

What about Dreadnoughtus? As Lacovara et al. (2014) describe, the cervical, dorsal, and sacral vertebrae and sacral ribs are honeycombed with pneumatic camellae (small, thin-walled chambers). And the dorsal ribs have pneumatic foramina and were probably at least partly hollowed-out as well. The caudal vertebrae do not appear to have been pneumatic, at least internally (but diverticula going into the tail can be cryptic — see Wedel and Taylor 2013b). Diplodocus has a big, long, highly pneumatic tail, but Dreadnoughtus has a much longer neck, both proportionally and absolutely, and pneumatic dorsal ribs. So this one may be too close to call. But I also ran the numbers for T. rex way back when and found that air in its vertebrae accounted for 7% of its body volume (this abstract). Pessimistically, if we assume Dreadnoughtus had small lungs and air sacs (maybe 10% of whole-body volume) and not much air in the bones (7%), it’s whole-body density was probably still closer to 0.8 g/cm^3 than to 0.9. Optimistically, a lot of titanosaurs were radically pneumatic and they have may have had big air sac systems and extensive diverticula to match, so a bird-like 0.7-0.75 g/cm^3 is certainly not beyond the bounds of possibility.

This table shows a spectrum of masses, based on the three body volumes from GDI (columns) and some possible whole-body densities (rows). Note that the columns are not in the same order as in the first table — I lined them up from most t0 least voluminous here. The 57-ton estimate is the max, and that assumes that the neck and tail were both perfectly round, and that despite the lungs, air sacs, and air reservoirs inside the bones, the whole-body density of Dreadnoughtus was still 1.0 g/cm^3, neither of which are likely (or, I guess, that a real Dreadnoughtus was significantly fatter than the one shown, and that all of that extra bulk was muscle or some other heavy tissue). The 28t mass in the lower left corner is also unrealistic, because it assumes a tall, narrow torso. My pick is the 36t estimate at the bottom of the middle column, derived from what I think are the most defensible volume and density. Your thoughts may differ — the comment thread is open.

This last table is just a quick-and-dirty comparison of how the volume of the body breaks down among its constituent parts in Plateosaurus (from this post), Giraffatitan (from Taylor 2009), and Dreadnoughtus (based on my “tall neck and tail” GDI). Dreadnoughtus seems to have a more voluminous neck and a less voluminous trunk, proportionally, than Giraffatitan, but I think a lot of that is down to the very fat fleshy envelope drawn around the cervicals of Dreadnoughtus. We are fortunate to count some fearsomely talented paleoartists among our readers — I’ll look forward to seeing what you all come up with in your independent skeletal recons.

So, what’s the take-home? Based on the data available, I don’t think the holotype individual of Dreadnoughtus massed anything like 59 metric tons. I think 35-40 metric tons is much more defensible. But I’m happy to have my errors pointed out and new data and arguments brought to the fore. Your thoughts are most welcome.

## How heavy was Giraffatitan brancai? I mean, really?

### June 9, 2014

We’ve touched on this several times in various posts and comment threads, but it’s worth taking a moment to think in detail about the various published mass estimates for the single specimen MB.R.2181 (formerly known as HMN SII), the paralectotype of Giraffatitan brancai, which is the basis of the awesome mounted skeleton in Berlin.

Here is the table of published estimates from my 2010 sauropod-history paper, augmented with the two more recent estimates extrapolated from limb-bone measurements:

Author and date Method Volume (l) Density (kg/l) Mass (kg)
Janensch (1938) Not specified `40 t’
Colbert (1962) Displacement of sand 86,953 0.9 78,258
Russell et al. (1980) Limb-bone allometry 13,618
Anderson et al. (1985) Limb-bone allometry 29,000
Paul (1988) Displacement of water 36,585 0.861 31,500
Alexander (1989) Weighing in air and water 46,600 1.0 46,600
Gunga et al. (1995) Computer model 74,420 1.0 74,420
Christiansen (1997) Weighing in air and water 41,556 0.9 37,400
Henderson (2004) Computer model 32,398 0.796 25,789
Henderson (2006) Computer model 25,922
Gunga et al. (2008) Computer model 47,600 0.8 38,000
Taylor (2009) Graphic double integration 29,171 0.8 23,337
Campione and Evans (2012) Limb-bone allometry 35,780
Benson et al. (2014) Limb-bone allometry 34,000

(The estimate of Russell et al. (1980) is sometimes reported as 14900 kg. However, they report their estimate only as “14.9 t”; and since they also cite “the generally accepted figure of 85 tons”, which can only be a reference to Colbert (1962)”, we must assume that Russell et al. were using US tons throughout.)

The first thing to notice is that there is no very clear trend through time, either upwards or downwards. Here’s a plot of mass (y-axis) against year of estimate (x-axis):

I’ve not even tried to put a regression line through this: the outliers are so extreme they’d render it pretty much useless.

In fact, the lowest and highest estimates differ by a factor of 5.75, which is plainly absurd.

But we can go some way to fixing this by discarding the outliers. We can dump Colbert (1962) and Alexander (1989) as they used overweight toys as their references. We more or less have to dump Russell et al. (1980) simply because it’s impossible to take seriously. (Yes, this is the argument from personal incredulity, and I don’t feel good about it; but as Pual (1988) put it, “so little flesh simply cannot be stretched over the animal’s great frame”.) And we can ignore Gunga et al. (1995) because it used circular conic sections — a bug fixed by Gunga et al. (2008) by using elliptical sections.

With these four unpalatable outliers discarded, our highest and lowest estimates are those of Gunga et al. (2008) at 38,000 kg and Taylor (2009)at 23,337. The former should be taken seriously as it was done using photogrammetrical measurements of the actual skeletal mount. And so should the latter because Hurlburt (1999) showed that GDI is generally the least inaccurate of our mass-estimation techniques. That still gives us a factor of 1.63. That’s the difference between a lightweight 66 kg man and and overweight 108 kg.

Here’s another way of thinking about that 1.63 factor. Assuming two people are the same height, one of them weighing 1.62 times as much as the other means he has to be 1.28 times as wide and deep as the first (1.28^2 = 1.63). Here is a man next to his 1.28-times-as-wide equivalent:

I would call that a very noticeable difference. You wouldn’t expect someone estimating the mass of one of these men to come up with that of the other.

So what’s going on here? I truly don’t know. We are, let’s not forget, dealing with a complete skeletal mount here, one of the very best sauropod specimens in the world, which has been extensively studied for a century. Yet even within the last six years, we’re getting masses that vary by as much as the two dudes above.

## How much is our intuition about sauropod mass worth?

### June 5, 2014

As promised, some thoughts on the various new brachiosaur mass estimates in recent papers and blog-posts.

Back in 2008, when I did the GDI of Giraffatitan and Brachiosaurus for my 2009 paper on those genera, I came out with estimates of 28688 and 23337 kg respectively. At the time I said to Matt that I was suspicious of those numbers because they seemed too low. He rightly told me to shut up and put my actual results in the paper.

More recently, Benson et al. (2014) used limb-bone measurements to estimate the masses of the same individuals as 56000 and 34000 kg. When Ian Corfe mentioned this in a comment, my immediate reaction was to be sceptical: “I’m amazed that the two more recent papers have got such high estimates for brachiosaurs, which have the most gracile humeri of all sauropods“.

So evidently I have a pretty strong intuition that Brachiosaurus massed somewhere in the region of 35000 kg and Giraffatitan around 30000 kg. But why? Where does that intuition come from?

I can only assume that my strongly held ideas are based only on what I’d heard before. Back when I did my 2008 estimate, I probably had in mind things like Paul’s (1998) estimate of 35000 kg for Brachiosaurus, and Christiansen’s (1997:67) estimate of 37400 for Giraffatitan. Whereas by the time the Benson et al. paper came out I’d managed to persuade myself that my own much lower estimates were right. In other words, I think my sauropod-mass intuition is based mostly on sheer mental inertia, and so should be ignored.

# References

## How big was Brachiosaurus altithorax? I mean, really?

### June 2, 2014

Way back in November 2011, I got this inquiry from Keiron Pim:
I’m currently writing a popular guide to dinosaurs, to be published by Random House next autumn [Ed.: available now at amazon.com and at amazon.co.uk]. I’ve been writing about [Brachiosaurus and Giraffatitan], and have read your 2009 study vindicating the proposal to separate them into two genera.
[…]
I know you consider Brachiosaurus likely to have been bigger (and note that the specimen was not fully grown), with a longer trunk and tail – but most of the sources I can find give both animals the same body length, generally around 26m. Presumably this doesn’t reflect your work, and your calculations are different.

I replied at the time, and said that I’d post that response here on SV-POW!. But one thing and another prevented me from getting around to it, and I forgot all about it until recently. Since we’re currently in a sequence of Brachiosaurus-themed posts [part 1, part 2, part 3, part 4, part 5, part 6], this seems like a good time to fix that. So here is my response, fresh from November 2011, lightly edited.

Well, Giraffatitan has only been recognised as a separate animal at all in the last couple of years, and nearly everything that has been written about “Brachiosaurus“, at least in the technical literature, is actually about Giraffatitan. So existing sources that give the same length for both are probably not making a meaningful distinction between the two animals.

First, on Giraffatitan: Janensch (1950b:102) did a great job of measuring his composite mounted skeleton. His figure for the total length of Giraffatitan along the neural canal is 22.46 m, and is certainly the best estimate in the literature for an actual brachiosaur specimen (and quite possibly the best for any sauropod).

I don’t know where the figure of 26 m comes from, but as Janensch (1961:213) notes, the isolated fibula XV2 of Giraffatitan in the Berlin collection is 134 cm long, compared with 119 cm for that of the mounted skeleton. This is 1.126 times as long, which if scaled isometrically would yield a total length of 25.29 m.  So that is defensible, but 26 m is not, really.

I would advise sticking with Janensch’s published figure of 22.46 m, as it’s based on good material, and also because it forms the basis of my comparative estimate for a Brachiosaurus of similar limb length.

Now in my 2009 paper I estimated with reasonable rigour that the torso of Brachiosaurus was probably about 23% longer than that of Giraffatitan, yielding 4.82 m rather than 3.92 that Janensch gave for Giraffatitan. On much less solid evidence, I tentatively estimated that the tail of Brachiosaurus might have been 20-25% longer than that of Giraffatitan. Given the paucity of evidence I would play safer by going with the lower end of that estimate, which would give a tail length of 9.14 m compared with Janensch’s 7.62 for Giraffatitan. Riggs (1904) tells us that the sacrum of Brachiosaurus is 0.95 m long, which is slightly less than 1.07 m for Giraffatitan. Finally, since we know nothing of the head and neck of Brachiosaurus, the null hypothesis has to be that they were similar in proportion to those of Giraffatitan.

Putting it all together, Brachiosaurus may have been longer in the torso by 0.9 m, and in the tail by 1.52 m, but shorter in the sacrum by 0.12 m — for a total additional length of 2.3 m. That would make Brachiosaurus 24.76 m long, which is 10% longer than Giraffatitan.

Note that all the Brachiosaurus figures are given with much greater precision than the sparse data we have really allows.  I think you could round Janensch’s 22.46 m for Giraffatitan to 22.5 and be pretty confident in that number, but you shouldn’t really say anything more precise than “maybe about 25 m” for Brachiosaurus.

Finally, you correctly note that the Brachiosaurus specimen was not fully grown — we can tell because its coracoid was not fused to the scapula. But the same is true of the mounted Giraffatitan, so these two very similarly sized animals were both subadult. How much bigger did they get?  We know from the fibula that Giraffatitan got at least 12-13% bigger than the well-known specimen, and I’d be pretty happy guessing the same about Brachiosaurus.  And I wouldn’t rule out much bigger specimens, either.

# References

## How slender were the humeri of Giraffatitan?

### May 23, 2014

Continuing with what seems to have turned out to be Brachiosaur Humerus Week here on SV-POW! (part 1, part 2, part 3), let’s consider the oft-stated idea that brachiosaurs have the most slender humeri of any sauropod. For example, Taylor (2009:796) wrote that:

Discarding a single outlier, the ratio of proximodistal length to minimum transverse width (Gracility Index or GI) in humeri of B. brancai [i.e. Giraffatitan] varies between 7.86 for the right humerus HMN F2 and 9.19 for the left humerus HMN J12, with the type specimen’s right humerus scoring 8.69, slightly more gracile than the middle of the range […] For the B. altithorax type specimen, the GI is 8.50, based on the length of 204 cm and the minimum transverse width of 24 cm reported by Riggs (1904:241). However, the B. altithorax humerus looks rather less gracile to the naked eye than that of B. brancai, and careful measurement from Riggs’s plate LXXIV yields a GI of 7.12, indicating that the true value of the minimum transverse width is closer to 28.5 cm. As noted by Riggs (1903:300-301), the surface of the distal end of this humerus has flaked away in the process of weathering. Careful comparison of the humeral proportions with those of other sauropods (Taylor and Wedel, in prep.) indicates that the missing portion of this bone would have extended approximately a further 12 cm, extending the total length to 216 cm and so increasing the GI to 7.53 – still less gracile than any B. brancai humerus except the outlier, but more gracile than any other sauropod species except Lusotitan atalaiensis (8.91), and much more gracile than the humerus of any non-brachiosaurid sauropod (e.g., Diplodocus Marsh, 1878 sp., 6.76; Malawisaurus dixeyi Jacobs, Winkler, Downs and Gomani, 1993, 6.20; Mamenchisaurus constructus Young, 1958, 5.54; Camarasaurus supremus Cope, 1877, 5.12; Opisthocoelicaudia skarzynskii Borsuk-Bialynicka, 1977, 5.00 – see Taylor and Wedel, in prep.)

Implicit in this (though not spelled out, I admit) is that the humeri of brachiosaurs are slender proportional to their femora. So let’s take a look at the humerus and femur of Giraffatitan, as illustrated in Janensch’s beautiful 1961 monograph of the limbs and girdles of Tendaguru sauropods:

The first thing you’ll notice is that the humerus is way longer than the femur. That’s because Janensch’s Beilage A illustrates the right humerus of SII (now properly known as MB R.2181) while his Beilage J illustrates the right femur of the rather smaller referred individual St 291. He did this because the right femur of SII was never recovered and the left femur was broken, missing a section in the middle that had to be reconstructed in plaster.

(What’s a Beilage? It’s a German word that seems to literally mean something like “supplement”, but in Janensch’s paper it means a plate (full-page illustration) that occurs in the main body of the text, as opposed to the more traditional plates that come at the end, and which are numbered from XV to XXIII.)

How long would the intact SII femur have been? Janensch (1950b:99) wrote “Since the shaft of the right femur is missing for the most part, it was restored to a length of 196 cm, calculated from other finds” (translation by Gerhard Maier). Janensch confused the left and right femora here, but assuming his length estimate is good, we can upscale his illustration of St 291 so that it’s to SII scale, and matches the humerus. Here’s how that looks:

Much more reasonable! The humerus is still a little longer, as we’d expect, but not disturbingly so.

Measuring from this image, the midshaft widths of the femur and humerus are 315 and 207 pixels respectively, corresponding to absolute transverse widths of 353 and 232 mm — so the femur is broader by a factor of 1.52. That’s why I expressed surprise on learning that Benson et al (2014) gave Giraffatitan a CF:CH ratio (circumference of femur to circumference of humerus) of only 1.12.

Anyone who would like to see every published view of the humeri and femora of these beasts is referred to Taylor (2009:fig. 5). In fact, here it is — go crazy.

Taylor (2009: figure 5). Right limb bones of Brachiosaurus altithorax and Brachiosaurus brancai, equally scaled. AC, humerus of B. altithorax holotype FMNH P 25107; DF, femur of same; GK, humerus of B. brancai paralectotype HMN SII; LP, femur of B. brancai referred specimen HMN St 291, scaled to size of restored femur of HMN SII as estimated by Janensch (1950b:99). A, D, G, L, proximal; B, E, H, M, anterior; C, K, P, posterior; J, O, medial; F, I, N, distal. A, B, D, E modified from Riggs (1904:pl. LXXIV); C modified from Riggs (1904:fig. 1); F modified from Riggs (1903:fig. 7); GK modified from Janensch (1961:Beilage A); LP modified from Janensch (1961:Beilage J). Scale bar equals 50 cm.

Notice that the femur of Giraffatitan, while transversely pretty broad, is freakishly narrow anteroposteriorly. The same is true of the femur of Brachiosaurus, although it’s never been shown in a published paper — I observed it in the mounted casts in Chicago.

Weird.

# Calculations

So let’s take a wild stab at recalculating the mass of Giraffatitan using the Benson et al. formula. First, measuring the midshaft transverse:anteroposterior widths of the long bones gives eccentricity ratios of 2.39 for the femur and 1.54 for the humerus (I am not including the anterior prejection of the deltopectoral crest in the anteroposterior width of the humerus) . Dividing the absolute transverse widths above by these ratios gives us anteroposterior widths of 148 for the femur and 150 mm for the humerus. So they are almost exactly the same in this dimension.

If we simplify by treating these bones as elliptical in cross section, we can  approximate their midshaft circumference. It turns out that the formula for the circumference is incredibly complicated and involves summing an infinite series:

But since we’re hand-waving so much anyway, we can use the approximation C = 2π sqrt((a²+b²)/2). where a and b are the major and minor radii (not diameters). For the femur, these measurements are 176 and 74 mm, so C = 848 mm; and for the humerus, 116 and 75 mm yields 614 mm. (This compares with FC=730 and HC=654 in the data-set of Benson et al., so we have found the femur to be bigger and the humerus smaller than they did.)

So the CF:CH ratio is 1.38 — rather a lot more than the 1.12 reported by Benson et al.  (Of course, if they measured the actual bones rather than messing about with illustrations, then their numbers are better than mine!)

And so to the mass formula, which Campione and Evans (2012) gave as their equation 2:

log BM = 2.754 log (CH+CF) − 1.097

Which I understand to use base-10 logs, circumferences measured in millimeters, and yield a mass in grams, though Campione and Evans are shockingly cavalier about this. CH+CF is 1462; log(1462) = 3.165. That gives us a log BM of 7.619, so BM = 41,616,453 g = 41,616 kg.

# Comparison with Benson et al. (2014)

Midshaft measurements and estimates for SII long bones (all measurements in mm)
SV-POW! Benson et al.
Femur Humerus Femur Humerus
Transverse diameter 353 232 240