"if you see a horizontal streak moving"

No. This does simply not work out from an SNR perspective: the more pixels a fixed total signal is spread across, the lower the achievable SNR of any derived measurement. Even if signal trivially adds and noise only adds in quadrature, you are spreading the same signal out over many more pixels, hence much greater total noise for a given fixed signal.

Wow. Mercury really is more interesting than I expected.



Bad pun?

I guess JunoCam being less of a scientific priority versus imaging systems on Dawn probably helps; the investigators don't have to worry about being pipped at the post by amateur researchers. It makes sense to use a public outreach instrument for...um... public outreach. Good news for us, anyway.

On the processing side, given that it is a pushbroom system, the processing might be a little bit more involved for those who want to work from the really low-level raw data. I might do some research into processing methods, as it seems like an interesting challenge.

Does anyone have any idea where to find the description of the compression algorithm? Apparently it is in a document called:
MDIS Compression Description, Pat Murphy
(Which doesn't appear to be available anywhere on the internet.)

Except that the vast majority of the data will probably be GCMS results, which are a little less photogenic, albeit very scientifically valuable.

Thank-you john_s, I didn't realise that the smear was quite so bad! Do you just use a naive subtraction or deconvolution type approach (or something a bit more advanced)? Looks like some sort of denoising approach might be necessary to help get rid of the smear removal artefacts. Might be an interesting project to play around with in my spare time

The images don't really look all that high quality -- did they not do proper bias/flat field calibration or is there something I'm missing? Certainly I would expect a fair bit of shot noise, but...

The air pressure tends to be substantially lower at the poles, even at the same altitude. For example, the summit of Denali (at about 6.2 km) has an air pressure equivalent to that at 6.9km altitude in the Himalayas (say Everest).

It might have been useful if Kepler had been designed to detect the beginning of a transit and increase the sample rate. By buffering data at a higher sample rate for say, half an hour, you could detect the transit with a high statistical certainty but not lose the beginning of the transit.

On the other hand, once the period is known it's pretty easy to train a large aperture telescope on the star and get photometric readings with a much better SNR and temporal resolution. I don't think Kepler gets a great SNR for dim targets, especially at high temporal resolutions, so there may not have been any significant benefit from doing this.

Got to wonder what 'no star apparent at target location' really means.

Some very interesting papers there, especially considering how little Kepler data it was based on.

The problem with having only one spectra is that you can get the mean colour of Uranus perfect, but that doesn't necessarily imply that you get the colour of any specific features correct.

I was thinking of something more along these lines (although this is probably too simple a model to work very well):

* Take a low res spectral cube of Uranus in the visible (say 15x15x15). assume that each spatial pixel is a sample from a multivariate Gaussian distribution. Determine the mean and covariance matrix from the data. Assume that the Voyager pictures of Uranus actually look vaguely like Uranus when you took the spectral cube (not likely I know).

* Each one of the Voyager images corresponds to sampling the inner product of the combined filter/imager spectral response with the actual spectra at that pixel. Thus a sequence of images can be considered a matrix transform for each pixel. The matrix transformation can then be written up as an optimisation problem trying to maximise the posterior probability of the spectra for a pixel versus the measured data using the known spectral relationships as a prior and a positivity constraint. This would require some sort of noise modelling (I have no idea what the noise characteristics of a Vidicon tube are). Unfortunately the positivity constraint means that it would probably require some sort of iterative solution.

* Given the estimated spectra for each pixel, conversion to RGB (or whatever format you want) is a pretty standard linear transformation.

Any thoughts? The premise is that missing colours like red (and indistinguishable colours) can be inferred from the presence of of other colours given prior knowledge of what Uranus looks like.

Since the Voyager camera had very little true red response, it's a bit difficult to do anything. Essentially it comes down to inverting a highly singular matrix to recover the spectral curve for each pixel; a pretty standard inversion problem, but no simple regularisation techniques will get you what you want because it's so underdetermined.

I suppose you could probably use low res multispectral images to create a statistical model of how red correlates with the various other colours Voyager actually recorded and try to recreate the missing colours. I'm a bit too busy at the moment to give it a go, but if suitable reference data is available then it ought to be possible at least. Might be worth a paper if someone actually does it!

They don't have a lot of choice as there aren't a lot of good target craters with an adequate concentration of water and the ones with water don't necessarily have the best geometry. I'm not saying they haven't made the best decision possible, just that it doesn't look like an easy trade-off.

Thanks, I wasn't aware of the new release.

I wouldn't say that there are any advantages to the viewing geometry over Caebus A though -- even with the valley the Earth mask height is still very poor and they didn't specify the actual sun mask height. It might be better than previous, but without a specific figure it could still be a lot worse than Caebus A and a similar backdrop could have as easily been present at the previous crater (Not trying to be argumentative or anything , it's just hard to distill a lot of information from a normal press release).

No, that's not true unfortunately. This document (http://www.nasa.gov/pdf/386497main_target-selection_web2.pdf) shows that the new target crater Caebus has *far* worse viewing conditions than Caebus A.

Caebus A Sun Mask Height: 630m, Caebus A Earth Mask Height: 330m
Caebus Sun Mask Height 1420m, Caebus Earth Mask Height 3070m

This is quite bad news -- they're sacrificing a lot of water sensitivity due to the sunlight/Earth masking. There must have been something seriously wrong with Caebus A to warrant this, as the mask heights at Caebus A are absolutely perfect for an experiment like this. I'm guessing that new LRO data must have either indicated a very uneven distribution or lack of water in the old target.

One thing I've been wondering is how much Kepler can tell about the shapes of stars. Since so many stars spin really fast (often ~90% of breakup velocity) there are a lot of oblate stars out there. Now the Von Zeipel effect means that the equatorial regions of oblate stars are darker than the polar regions (due to a lower temperature) -- this on its own doesn't help us though because a spinning oblate star doesn't result in any apparent change in brightness from the point of view of Kepler. However, I have been wondering if there is a star analogue of Haumea -- which is a Kuiper belt object that spins so fast that hydrostatic equilibrium results in a more cigar shaped/rugby ball object (prolate isn't correct because that implies a different rotational axis) rather than a oblate spheroid). I presume that this occurs at even higher spin rates, just before breakup velocity. If stars like this exist, then different regions of the star's equator would be at different temperatures -- this means that if the spin axis is suitably aligned it should be possible to determine the stars spin rate and shape based on the light curve. This could be backed up by further investigations using optical interferometry.

Anyway... Just a random thought! If this does occur, it must be in the lightcurve data somewhere.

There's an interesting Van Karman lecture on using optical interferometry to measure star shapes
http://www.jpl.nasa.gov/events/lectures/mar06.cfm
I have no idea why the main Von Karman page doesn't link to the pre-2008 lectures any more -- I even sent a note to the webmaster, sigh!