The importance of good science journalism

This is an ideal case, of course, but no less important for that. In such an ideal scenario, time currently spent writing grant proposals would be spent on outreach. It's probably easier to turn scientists into journalists than the other way around, but at the same time, journalists bring a much wider perspective.


Scientists and journalists share a passion for questioning assumptions and biases. We are trained to uncover hidden narratives in the pursuit of deeper understandings. And we share an enthusiasm for revealing new knowledge that can be shared with the world.

These values cannot be taken for granted, especially in our current political environment. It is no secret that both scientists and journalists are facing a concerted wave of allegations around “bias” and “fake news.” This type of regressive criticism is not new. Throughout history, those who have sought to suppress the truth have endeavored to muffle the voices of scientists and journalists. Without these voices societies decay. Therefore, when confronted with the current assaults, we cannot allow ourselves to recoil into our protective harbors and wait for the storm to pass. We cannot wait for others to step into the void. We have to shrug off whatever reluctances we may have and find ways to share the stories of science with a world in desperate need of hearing them.

We recognize that there are differences between the ways journalists and scientists perform their professions, and those differences could serve as barriers for cooperation. Scientists often are wary of the way their work might be presented by journalists. They have seen nuanced research oversimplified or hyped for more dramatic (and sometimes misleading) headlines. At the same time, journalists can be frustrated by scientists who respond to straightforward questions with jargon and are unable or unwilling to explain the essence of their discoveries without caveats and qualifiers.

We believe, however, this mistrust is superficial and can be overcome for the benefit of all of society. Scientists and journalists share core aspirations. Both disciplines are about observing the world, questioning the unknown and collecting facts. Both scientists and journalists know their work is built on the work of others and they must find a way to share their discoveries. Scientists may tell their stories in papers they publish to share with their colleagues in the field. Journalists may tell their stories in print, radio or film, often trying to reach as wide an audience as possible. But the mission is the same. Knowledge can only have an impact if others hear about it.


As far as journalism goes, it seems to me that poor science reporting is (mostly) simply due to ignorance. Maybe there ought to be more outreach courses aimed at journalists. Of course you need experts to tell you about the technical details and the results themselves. But beyond that, many of the techniques of critical analysis aren't that hard.

From the political perspective something much more sinister looks to be going on. It feels less of a case of simple ignorance and more of wilful bullshitting : not caring about the evidence rather than (but not excluding the case of) not understanding it. It's not difficult to understand, say, that something being possible doesn't mean it isn't fantastically unlikely, or that because something did happen once it doesn't mean that it happened much less often than other incidents. False degrees of confidence in or against a result aren't because politicians don't understand this, it's because they don't care (not always out of malevolence or even stupidity, but sometimes). When you strip away the objective evidence, all you have is subjective, emotion-driven ideology. And where it may be difficult to argue with a fact, it's easy to argue with an emotion.

That's not to say that there aren't media outlets that are hugely partisan and essentially nothing but the mouthpieces of their favourite political tribe : there are. These institutions attempt to discredit science but only as part of a larger campaign of discrediting anyone and anything (from any field) that even hints at disagreement with their moral values. Anyone who says otherwise, regardless of their status, is branded as a member of the controlling elite, and conversely, anyone agreeing with them is One of The People. They don't actually care a damn about The People, of course; mostly this kind of rhetoric is used by people who are far more out of touch than the "elite" they like to deride. It is merely a rhetorical tool to sow division, nothing more. The underlying theme is one of avoiding and ignoring the evidence, because even imperfect evidence is, if analysed sensibly, a damn sight harder to argue with than a whimsical feeling.

So yes, improving science journalism is important. But this is only one expression of the root problem, not the problem itself.


https://blogs.scientificamerican.com/observations/what-journalists-and-scientists-have-in-common/

The Big One-Zero

Submitted my tenth paper as first author :

Faint and fading tails : the fate of stripped HI gas in Virgo cluster galaxies

Although many galaxies in the Virgo cluster are known to have lost significant amounts of HI gas, only about a dozen features are known where the HI extends significantly outside its parent galaxy. Previous numerical simulations have predicted that HI removed by ram pressure stripping should have column densities far in excess of the sensitivity limits of observational surveys. We construct a simple model to try and quantify how many streams we might expect to detect. This accounts for the expected random orientation of the streams in position and velocity space as well as the expected stream length and mass of stripped HI. Using archival data from the Arecibo Galaxy Environment Survey, we search for any streams which might previously have been missed in earlier analyses. We report the confident discovery of nine streams as well as sixteen other less sure detections. We show that these well-match our analytic predictions for which galaxies should be actively losing gas, however the mass of the streams is typically far below the amount of missing HI in their parent galaxies, implying that a phase change and/or dispersal renders the gas undetectable. By estimating the orbital timescales we estimate that dissolution rates of 1-10 M⊙ yr−1 are able to explain both the presence of a few long, massive streams and the greater number of shorter, less massive features.

And now to pray to the Journal Gods for a fair and fast referee....

A Volumetric Law For Star Formation

It's well-known that there's a correlation between galaxy gas density and star formation rate. The problem is that no-one's clear on exactly what sort of correlation it is, or what sort of gas it is. In general, galaxies which have significant amounts of atomic hydrogen (which is relatively warm) tend to be forming stars, so there's definitely some sort of connection there. But recent studies have found the correlation is much better when considering only the colder, molecular hydrogen, which gives a much nicer linear relation. And that makes physical sense too, since to form a star you need higher gas density, which is easier if the gas is cold as it can't use thermal pressure to support itself against collapse. Perhaps most convincing were the discovery of holes in the atomic gas component of some spiral galaxies, which seem to be the result of star formation consuming all the gas. Also, the atomic gas has a strict upper density limit, beyond which all gas seems to become molecular.

The shape of the correlation between the gas and star formation rate is somewhat unclear as well. Mostly it's a nice power law, but there's some evidence for a density threshold below which the star formation activity drops sharply. This, say the authors of this work, is quite controversial (more so than I realised, and I'm supposed to know about this stuff), as is the choice of which gas component to use.

There are lots of uncertainties, but perhaps the main one is the gas density (of either the cold or the warm component). We can estimate the gas density per unit area (surface density) easily enough, but the true volume density is much harder because we can't directly measure the thickness of the gas disc. Here the authors attempt to overcome this. They assume the gas is in hydrostatic equilibrium, meaning that its outward pressure (due to thermal and other motions) is balanced by its tendency to collapse under gravity. This isn't straightforward : it requires detailed knowledge of the mass and distribution of stars and dark matter as well as that of the gas, and also it needs the velocity dispersion and overall rotation curve of the gas. This is currently only possible for quite nearby galaxies since you need very detailed, well-resolved data to do this properly. Even then there are still uncertainties and assumptions that have to be made.

This paper is under review, but it seems to me to be a careful, detailed work more in need of correcting typos than methodological revisions. After describing their methods with considerable precision, they find that both the atomic and molecular components show very clear, power-law correlations with star formation. There's no change of slope with density either. They say this could be because the thickness of the gas disc varies significantly depending on where you are in the galaxy : it's much fatter in the low-density outskirts than the centre. The surface density measurement would give a misleadingly high estimate in the outer regions.

The fact that both warm gas correlates with star formation is also very interesting. Previously the tendency had been to assume that this connection would be somewhat secondary : the picture has been shifting to the atomic gas having to transition to molecular gas before forming stars. So the correlation is expected to be rather rough, but in fact it's very clear - certainly no worse than that of the molecular gas. They give two interpretations :
- The warm gas is a good tracer of the cold gas. The connection between atomic gas and star formation could then still be indirect. It would also mean that there could be undetected cold gas (which is very difficult to detect directly - normally other components have to be used) in the outskirts of galaxies, where star formation activity has previously been puzzling.
- The warm gas can form stars directly. This is theoretically possible : there are conditions under which the warm gas could cool so quickly that there's no time for molecular gas to form and it goes directly into stars.

There's a lot of work to be done, but this isn't the only evidence for atomic gas being directly involved in star formation. The idea that actually this complex process is governed by a rather simple condition - true density - but that this condition is hard to measure is very appealing.
https://arxiv.org/abs/1810.03616

"Freedom from" versus "freedom to" in the world of science journals

A central charge, from some publishers and some academics is that Plan S is an infringement of academic freedom to choose how and where your work is published and it therefore unethical.

Blink.

Whut ?

Kudos to the author of this piece for the detailed legalistic analysis of why this is wrong, but surely from an ethical standpoint this is Bloody Obvious ?

As I understand it, the Plan calls for all publically-funded research to be made Open Access, i.e. freely available to the public. Since the public funded the research, they ought to get to read it if they want to. As far as I know it doesn't affect private research, and nor should it. If you privately commission a study, I don't think you're invariably, necessarily obligated to release its findings or data (though that's not to say there might not be some cases, e.g. legal proceedings, in which release could be demanded) to the public. But you wouldn't expect the researchers to turn around and say, "I've published the findings in this journal you can't read, because I'm exercising my academic freedom". You'd feel a common-sense entitlement to see what you'd paid for, unless for some strange reason you'd previously agreed an exception with the researchers. Could happen, but unlikely. Generally speaking you'd say the researcher was being unethical by denying your right to inspect the findings.

And so if the public fund research and the researcher has the option and ability to make it publicly visible, then the default expectation should be that it will be public. For them to actively choose an alternative, unless their are compelling reasons to do so (currently Open Access is frickin' expensive), is clearly unethical. Removing a freedom to do an unethical thing isn't itself a fundamentally unethical act - it's normally known as, for instance, "justice". Unless I'm missing something obvious, I find this "infringes academic freedom" argument to be so stupid I can't believe this is really what's bothering them, even if they think it is (with the possible exception of a few Randian devotees).

People are weird.

http://occamstypewriter.org/scurry/2018/10/01/academic-freedom-and-responsibility-why-plan-s-is-not-unethical/

The Galaxy End Sequence

Disclaimer : the lead author was my third year academic tutor (easily one of Cardiff's best lecturers), but I was not involved in this research at all.

Galaxies are pretty complicated things that come in all shapes and sizes. But at least one thing has tended to be quite simple : their colours. By and large, they divide themselves into two distinct sequences when you plot their colour against total stellar mass - a narrow red sequence, and a fuzzy blue cloud. In between lies the "green valley", or transition region, where there are relatively few galaxies. Most galaxies in the blue cloud tend to be spirals and irregulars, whereas most on the red sequence tend to be "red and dead" ellipticals. The blue galaxies tend to be gas rich and actively forming stars, whereas the red ones basically don't.

These are not hard-and-fast rules by any means : there are plenty of exceptions, and pretty much all of parameter space is populated to some degree. But these guidelines are pretty good and do seem to be true most of the time. And various other studies have claimed different evidence of a "bimodality" in the galaxy population, with active galactic nuclei being far more common above a certain mass threshold.

A popular view of galaxy evolution is that when galaxies run out of gas, they stop forming stars. Since young stellar populations are dominated by short-lived, bright blue stars, galaxies with ongoing star formation tend to be bluer. And the collisional nature of the gas helps it to form complex structures like spiral arms. So when the galaxy runs out of gas, the bright blue stars quickly die off and only the less massive red stars survive. The transient spirals that were sustained by the gas quickly dissipate, and the galaxy becomes smooth, red, and dead. It ought to move from the blue cloud through the green valley and end up on the red sequence.

One particularly nice paper from 2009 found that galaxies in the green valley are dominated by galaxies which have less gas than expected. Since there aren't very many of these galaxies, this suggests that galaxy evolution is a rapid process that's dominated by a sudden event. This is supported by analyses of environmental processes like ram pressure stripping, which can quickly strip a galaxy of its entire gas content in certain circumstances.

This paper disputes this interpretation, calling (in a later paper in the sequence I've not yet read) for nothing less than a new paradigm for galaxy evolution. A bold statement, but potentially justified.

Here the authors have used the Herschel Reference Survey, which claims to be volume-limited (that is, it's found every bright galaxy within its survey volume). They've also got a whole slew of different estimators for the star formation activity, so their data is high quality stuff. They then plot how the star formation rate (normalised to the total stellar mass of each galaxy) varies with stellar mass. This is much more physics-based than the traditional plots, which use colour as a crude proxy for star formation activity. And they don't see any evidence of bimodality, they see a continuous sequence. Even the galaxy structures apparently vary very smoothly from the galaxies with highest to lowest star formation rates.

One issue that's often raised for galaxy morphology evolution is that the stellar density profiles of spirals and ellipticals is so different that it doesn't seem that removing the gas and quenching star formation would be enough to cause this change. The spirals structures could disperse, sure, but why would the end of star formation lead to the presence of a very dense central bulge ? Here the authors note that the galaxies may not be as different as they appear, with recent studies finding evidence of residual rotating discs in elliptical galaxies. That makes the smooth evolution of disc to elliptical at least somewhat more plausible, as do the discoveries of gas in elliptical and lenticular galaxies.

This is definitely very interesting, but I'd be a bit cautious (I may change my tune when I eventually read the other papers in this sequence). First, while the correlation is obvious the authors claim the slope of this trend is curved, but their best fit is patently lousy. I'll show their plot without their best-fit curve in the comments below - I could probably agree that there's a slight curve, but definitely not the one the authors fit. That doesn't really change the conclusions, but it does raise a flag about the statistical analysis.

Second, there does seem to be some hint of bimodality, even if it's only weak. It would be nice to see a 2D density histogram of their main plot to see if galaxy density really does vary or if this is just an illusion.

More concerning is that the galaxy sample includes the Virgo cluster, which has a much higher galaxy density than the general field so here we expect environmental processes to be different. When they do the same plot without the cluster members, they claim to see the same trends, but I'm not so sure (see their figure 3). I would say it's at least arguable, though I'd fully accept that it's not certain, that two populations are evident : there appears to be a group of actively star-forming galaxies and a smattering of others. It would have been nice to also plot only cluster members as well - I suppose we'd see the reverse, a population dominated by dead galaxies with a smattering of live ones. Unfortunately they don't do this. I for one would be very surprised indeed if galaxy evolution was a largely continuous process inside clusters - I would almost say that doesn't make any sense.

So I don't know. It's intriguing, but I'm not so sure their sample is as good or uniform as they claim. I'm not at all sure about this business of specific star formation rate (s.f.r per unit mass). While s.f.r. will vary in galaxies simply because of their mass anyway (bigger galaxies have more gas available for star formation), galaxy size will affect their environmental susceptibility as well. So I'm a bit worried that this might be misleading in some way, though I can't quite put my finger on it... what we're missing is how individual galaxies evolve, of course, we only have this statistical picture of what they're doing now. Also, Virgo cluster galaxies do seem to either have lots of HI (warm gas) or none at all, with not much middle ground. It would be interesting to see if this is also true for the colder molecular gas, which is thought to be less vulnerable to stripping.

Interesting stuff, but I reserve the right to remain unconvinced for no good reason...
http://adsabs.harvard.edu/abs/2017MNRAS.465.3125E

M33 in VR : final version

Before I left for the science castle [an ALMA conference], I left the M33 data cube in VR rendering with higher resolution and wider camera separation for greater depth effect. I also saved it as an mp4, which means I can add the necessary metadata for this to work on YouTube. So you don't need a headset for this one, you can view it in regular browsers and use the mouse to look around. You can even turn on red-green 3D if you really want. Of course it's a lot better if you use an actual headset, and having it on YouTube should make viewing it a lot easier that way too.

This is still just a proof-of-concept test but I'm quite happy with the result. It could be fun to make this into a more fully developed, explanatory tour. What would be really nice would be to use the full AGES cube (this is only about 3% of the total), though that will require a different technique because otherwise I'll exceed the 1,024 image texture limit in Blender. I'll see how well Cycles handles image sequences as volumetrics. In the meantime I have a more detailed data cube to try out.
https://www.youtube.com/watch?v=ZI6dACWVni8

Science castle FTW !

Every year, the European ALMA ARC nodes hold an all-hands meeting to discuss mostly very uninteresting logistical issues. The ALMA Regional Centres are responsible for providing support to ALMA users, from writing telescope proposals to helping with the data reduction - that one being especially important. They're also involved with software development and whatnot. Meetings tend to have very little science content so they're not really worth lengthy descriptions, but this year the venue was especially nice : Chateaux Liblice, just north of Prague. This is one of two castles owned by the Czech Academy of Sciences and I'm fully in favour of science institutions being castle-based.

The actual seminar room was entirely normal and not worth describing. But the coffee and dining areas, those were proper marble-columned chateaux awesomeness.

I didn't get a room in the castle itself, but in a small town a few miles away. I didn't care though because it was super-opulent. It had TWO balconies ! TWO ! Whoever heard of such a thing ?

Such opulence, gentle reader, is not standard. Usually we get mid-level hotels, and sometimes extremely bad ones. This one was probably the most luxuirant conference venue I've ever had.

M33 VR looking shiny and nice

Some major improvements to the M33 VR render. Low resolution video but that really doesn't matter because the data is low resolution anyway. Much better colour scheme so you see a lot more detail in this one, and the data range now shows enough noise to give a better sense of depth. Plus the colour scheme is just much prettier.

Unfortunately I forgot to render this in the .mp4 format required for YouTube so you'll probably still have to download this one, and it's only suitable for headsets. I'll try and get the YouTube version working next week.

I think this would be a very nice way to give a tour through a data cube. A full AGES data cube (e.g. https://www.youtube.com/watch?v=1YWGZhXe_gA) would be a lot of fun, but that would require breaking the image texture limit that Blender < 2.78 can handle. So either I reinstall Linux on my work machine, or try and get the images sequences to process as Cycles volumetrics instead of textured planes.

Proof of concept : M33 HI data cube in VR.

Proof of concept : M33 HI data cube in VR. Has a lot of little flaws but the basic concept works : you fly through the data, it's visibly 3D, and you get full 360 coverage. Needs a headset to view this one. Once I iron out the problems (data has been smoothed too much, the colour scheme gets rid of too much noise which would provide useful reference points for depth information, and there's probably too much saturation) I'll upload to YouTube with metadata, so you can pan around in a regular web browser.

This one is created using the bare minimum display code of FRELLED (http://www.rhysy.net/frelled-1.html) converted to use Blender's Cycles engine, which can handle the equirectangular camera format needed for 360 spherical stereo video. This has to be rendered rather than using realtime capture (though the Cycles camera supports equirectangular display in the realtime preview, it doesn't seem to allow for capturing preview animations like the OpenGL view does). It also requires having all three projections visible at once, so this is rather slow.

Previously I was hell-bent on getting the ALFALFA data catalogue rendered in VR, but the limitation was that Blender versions 2.78 and below don't allow more than 1,000 image textures. And my work machine, which can comfortably handle intensive processing jobs for days on end without batting an eye, won't let me install 2.79 (which doesn't have a texture limit) unless I do so massive upgrading of my Linux installation. Fortunately, while many HI data cubes have the equivalent of more than 1,000 images, most of them don't need it - in fact, removing most of them actually results in a more detailed, less saturated appearance of the final renders.

More on the data cube on display here.

It's always in the last place you look

Traditionally* the missing satellite problem refers to the lack of satellite galaxies detected in our Local Group compared to theoretical predictions. It's only in this nearest region that we can reach the sensitivity needed to be able to spot the faintest smudges of the smallest galaxies. Which means it's always been possible that our Local Group is in some way peculiar, with other galaxies having multitudinous swarms of tiny satellites that we just can't see. It would certainly be great to have better data of more distant galaxies so we can see how typical our little patch of the cosmos really is.

* Since 1999. That's traditional now.

This paper presents observations of the NGC 3175 group of galaxies, which is reasonably similar to our own group but about 14 Mpc (45 million light years) away. Using deep, but not extraordinarily so, 2.5 hour observations on a 2.2 m telescope, they claim to have detected ~550 candidate dwarf galaxies : far more than the 100-odd satellites in the Local Group. So, super interesting ?

I'd have to say no, not really. I'm not at all convinced by the methodology here. They don't state the surface brightness sensitivity of their observations and their detection method seems dicey. All (or almost all) of those recent "ultra diffuse galaxies" have been detected by basically the same method : doing deep observations of both the target field and a background control field. That way the search method can be tested. What it should do - and what the UDG papers demonstrate successfully - is find stuff only in the target field, so you can be pretty confident that most of the detections are really associated with the known group or cluster of galaxies. That means you can say they're probably at a similar distance, which is much easier to measure for the big bright galaxies (and damned hard for the faintest ones).

That's not what they do here : they just have the target field, and no control. They decide if the galaxies are likely to be in the group or not based on their structural parameters. The UDG papers do this too, but they don't rely on them this heavily. While I don't doubt that this can indeed remove lots of contaminating background objects, e.g. angularly tiny galaxies with spiral arms are almost certainly distant background objects, it's not at all clear it's a good way to remove all the background objects, much less how much foreground contamination there is. The discovery of large numbers of UDGs, which could easily be mistaken for dwarves except they're known to actually be large and far away, makes this kind of method questionable at best. It's also rather worrying that they dismiss UDGs as unimportant :
"...these UDGs are just a subset of dwarf elliptical galaxies found mostly in rich clusters of galaxies."
Oh well, one man's revolution in the field is another's blasé event of tedious mediocrity.

(The third author has mentioned the unimportance of UDGs before and is quite clearly bitter about his own similar discoveries being overlooked - to some extent quite understandably so, but it's the large numbers of these objects that's really got people interested.)

The other crucial thing lacking here is a discussion on the environment of the group. On the small scale, it would be extremely interesting to see the positions of the galaxies - with >500 objects we could perhaps see those damn planes I hate so much or if they were at least denser nearer to the centre of the group. And we need a discussion of the background environment - maybe there are background clusters or groups we might expect to host UDGs and so cause contamination. They could have at least have calculated the distance beyond which their candidates would be unfeasibly large. But they don't do any of that. The radii given in their (truncated - though it's possible the full version will be available when it's actually published in MNRAS) main table (which is not referenced in the text) are very much smaller than that of UDGs, and their Sersic indicies (a measure of the shape of their light profile) are similar, so it's entirely reasonable to suggest that many of them could be much more distant objects.

Assuming this result is correct, then the numbers they found are a bit less than the classical simulations predict, but a bit more than the more modern versions which have far more physics. But there are so many uncertainties at the moment that this could simply be meaningless, so there's not much point speculating about it.
http://adsabs.harvard.edu/abs/2018arXiv180809020K