To reconcile the last few messages, the question is whether Kepler can perform its planet-seeking duties given the recent failure. There's no doubt it can do _something_, although which somethings it can do, and their merit, is an open question.

In my view, it's very likely that this spells the end of Kepler doing what it was doing. It has gathered extensive data useful for determining intrinsic planetary frequency for orbital periods up to a certain limit, and a range of planet sizes that at least includes terrestrial-sized planets in short-period orbits. What we would hope for with more data would be to gain more observations of the same stars to see if we could detect a third transit for those which have only shown us two, and a fourth for those that have only shown us three, etc. For smaller planets in particular, Kepler was barely, if at all, delivering the needed fidelity when it was at its best. There is a significant challenge remaining in analyzing the existing data for orbital periods of more than about 270 days because the instrument was far noisier than hoped.

I hope that the data from quarters 13-15 ends up being at least as good as the data from Q1-Q12. As it stands right now, whether or not we have even a single good "earth like" candidate is still up in the air, and the hunt continues using the data on the ground. But it sounds like if Kepler will have more useful observations in its future, they will like involve repurposing the instrument for some quite different purpose.

While I'm disappointed with this (apparent) abrupt end to Kepler's unique capabilities, the Kepler team can be proud of what was accomplished in the time they had. The mission was well worth it.



I was under the impression that most of the problem was the unanticipated variability of main sequence stars. Perhaps that's wrong?

Regardless, looking to the future, and the follow up. As JRehling stated, an earth size planet transiting a sun like star turned out to be a marginal detection event for Kepler. Nevertheless, those events were probably picked up in the data most of the time, although they were closer to the noise level than it was hoped for at the beginning of the mission. They said that three transits, or four, would not be enough to confirm such a planet with high reliability, but still you would get candidates worth looking at. Since we would know which stars had the candidates (far less than the 150,000 being monitored) and a fairly good guess at the timing for possible future transits if the planet is real ...

... well, surely something can be done? If not with Kepler in its current state, then with some other present or future instrument?

Kepler did prove that the concept works. We can detect planets via the transit method. Now that the concept is proven, follow-up spacecraft can be designed to reduce the noise level even further and begin the systematic process of "mapping" local planetary systems.

While the mission tag-line, "finding new Earths," is what grabs the popular interest (and helps in obtaining funding), I'm a firm believer that it's just as important to find and categorize all of the planetary systems around us as it is to find "new Earths." After all, it's not like we have the propulsion technology to *reach* any exoplanets any time soon, so finding the planets that are there, colonizable or not, is providing a lot of basic data that we'll need if we're ever to truly understand how planetary systems are formed. (I mean, who would have guessed that binary star systems have just as many planets as single star systems? That was a rather non-intuitive thing that I know the astrophysicists weren't expecting.)

-the other Doug

Holder,

An unexpected source of variability exists in portions of the detector surface by which false dimming events occur purely due to instrument noise, often enough that three such events may occur as to resemble a planet where no real object exists. Because this occurs primarily in very limited zones of the detector surface, it is impossible to get the series of transits that a real planet would generate unless:

1) All observations of false transits occur when the spacecraft is in the same one out of four seasonal rotation orientations. This means it happens at a very high rate for periods of about one year, and quite rarely for other periods. Unfortunately, that is exactly the period of greatest interest!

2) Because some of the noisiest areas happen to be aligned 180° opposite one another, the problem happens in isolated cases for periods of about 6 months.

3) There is apparently some background level of higher-than-expected noise such that instrument-derived false positives can actually occur anywhere on the instrument surface, but this is probably quite rare if four or more transits have been observed, so with three years of data, periods up to ~270 days should be free of these. Once again, we see the problem occur at precisely the period range of greatest interest.

Note: The above are my observations from studying the data for the last few months in preparation for a possible publication, and I think these are quite well known to anyone else studying the data, but (3) has not been published by anyone so far as I know.

The distribution of planets Super Earth sized or greater has definitely been characterized out to ~270 days. The distribution of larger planets is known even further out, via Kepler and still other methods. The distribution of Earth sized planets has been characterized out to ~90 days. So working around those noise problems may deliver a scientific bonanza yet to come from the data on the ground.

COROT, WASP, OGLE, SWEEPS, even the amateur built HATNet all proved that the concept works long before Kepler was launched.

Kepler has discovered many Exoplanets....but it certainly didn't 'prove' the concept - we already knew it worked - Kepler would never have been funded if we hadn't have already known that.

Say the team does resort to using a telescope as a scanning platform, which seems likely because besides the reaction wheels the spacecraft seems to be in fine condition. I imagine that many more short period planets could be found because the telescope is not looking at one part of the sky anymore.

But for Cyngus, wouldn't it be helpful to the Kepler team to analyze their data as soon as possible? They could have already found an earth already but would not have decoded it yet. Once you know something is there it becomes a lot easier to find again. That is a theme that I found reoccurs a lot in astronomy.

These planets are not going anywhere. I'm sure the science team will want to get the data processed quickly - but unlike finding and asteroid or comet for the first time ....we know where these planets are. Unlike an asteroid or comet discovery - these planets do not have to be 'found' again once disocvered. We know where they are.

Right, so if we know they are there then it means it we wont need to look as hard to find them again. If Kepler becomes less accurate as a result of the wheel failure, then by using the data we already have (as well as the data Kepler took before the wheel failed), it can still continue its mission.

Kepler is looking for three transits to confirm a candidate. Say it has already identified one and another hidden in the database. After decoding that transit, there are two transits, so you look for a third. But if you had one transit then it less likely you are to focus on one star. So what I'm suggesting is that all the high quality data should be decoded and the Kepler team should start focusing on identifying current KOIs instead of finding new ones.

The new telescope is going to have limits, but I feel it could still be a statistics mission, just at a reduced rate.

I'm not sure you understand the difficulties here. It's not that the loss of pointing precision keeps the spacecraft from monitoring more stars simultaneously... it keeps it from monitoring any stars with the precision needed to detect transiting planets. Stars having their light spread variably and unpredictably across multiple pixels adds a lot of extra noise. This noise blinds the spacecraft to transits. Even if Kepler could continue to take light curve data, the extra noise would rule out detecting extrasolar planets (at least small ones, not sure if it can detect hot Jupiters).

Honestly, I think the planets at the appropriate distance from K / M dwarfs are at least as interesting as the ones around Sunlike stars, since the K and especially M stars are far more common. The red dwarfs will presumably dominate the planet population of the Galaxy.

Based off the latest interviews these are the scenarios they're trying next to recover full operation:
-re-activate the 3rd reaction wheel which failed last year and had metal-on-metal friction. They're hoping the lubricant spread over the rest period and restores full function
-See if they can use the thrusters to compensate for the missing reaction wheels.

Unsaid in the interview, but I suppose they could see if they can do anything with a noisy reaction wheel (or if they could salvage any data while scanning the field as mentioned previously).

There's added complexity, Morpheus, because for small planets, a "transit" observation is often noise only marginally distinguishable from the background noise. Three years of data for a star might have literally hundreds of possible transit events, and any pair of them, into the tens of thousands, might be two transits of the same planet, or one real transit and one false one, or two false ones.

For candidates with higher signal-to-noise, it's more obvious that a real transit has been observed, and these cases are also interesting. But I'd note that the distribution of jovians has already been constrained to periods longer than Kepler has been observing. So its the smaller planets that are of greater interest.

No doubt, it would be great to get additional data of top-notch quality even after an arbitrary delay, because it could still be coupled with older data to find long-period planets.

To highlight the strict requirements of nominal Kepler operations: They involve holding the craft steady to a small fraction of a single pixel for about 90 days at a time. Motion by half a pixel would cause about half of the observed stars to cross the boundary between one instrument pixel and another, which would enormously complicate the analysis in the best case, and seriously complicate it in the case of small planets with a lot of noise in the best case. Moreover, the motion of the target on a scale less than a pixel is being used now to help identify astrophysical false positives, because binary stars often show a detectable motion during the course of observations.

Observations with a bit less steadiness than the nominal operations can still be useful (indeed, the steadiness varied considerably throughout the primary mission) but at some point, it would invalidate most or all possible detections of earth-sized planets. The larger planets would tolerate less steadiness, but there's also less interest there. And as an overarching problem: As we consider longer periods, the probability of a geometric alignment decreases, so that even with the most ideal continued operations, getting observations of planets beyond about 3-5 AU starts to become unlikely, and for small planets, very dicey.

So all told, I'm quite pessimistic that a significant addition of science can happen unless near-nominal pointing can be resumed.

Maybe if the metal on metal momenum wheel can be made to work, we turn it on when we expect a special transit will happen. Then turn it off and wait to do the same when the next similar opportunity comes.

If we already expect a transit - why not observe it from a telescope on Earth?

Kepler was designed for survey - not targeted observation. It will probably be best used hereafter as an NEO surveying telescope....a use that would make the most of it's large sensor and comparatively large FOV, whilst accommodating it's degraded pointing accuracy.

"If we already expect a transit - why not observe it from a telescope on Earth?"
Actually, if I understand correctly, CHEOPS will partly do a transit survey in such a way, based on radial velocity measurements. http://cheops.unibe.ch/index.php/science/c...arget-selection You would need to go to space for the detection of transits of rocky planets.

That article states "With the imminent arrival of new, dedicated planet hunting spectrographs (HARPS-N, ESPRESSO, APF), the size of the sample of identified stars hosting small mass planet will grow even more rapidly."

HARPS-N, ESPRESSO and APF are ground based.

Indeed, but they cannot observe the transit.

Ooh but they can!
Exoplanets and the Rossiter-McLaughlin Effect
http://arxiv.org/abs/astro-ph/0612744

And they have. Some two dozen or so planets have had their transits spectroscopically detected (yielding the projected angle between the stellar rotation and orbital axis as a bonus!).

And furthermore, the R-M effect for a low-mass, long-period planet can actually have a higher amplitude than the planet's Keplerian RV signal. If you can time the spectroscopic observations to observe the transit, and if you detect the R-M effect, you
1) Confirm the planet candidate as a bona-fide planet.
2) Constrain the orbit axis inclination relative to the stellar rotation axis, and thus the dynamical history of the system.

Edit: Attached RV + R-M Effect model for an Earth analogue orbiting a Sun-analogue. Notice how high the R-M effect amplitude is relative to the Keplerian RV curve.

But the R-M effect cannot be used to determine the radius of the planet, which is what CHEOPS will do, or study the planet's atmosphere, which follow-up obervations of transiting planets around bright stars will do. Also, in these cases they already knew the planet was transiting and often make dedicated observations to observe the R-M effect. It is very hard to determine from just radial velocity surveys and will be an ineffective way to search for transiting planets, especially small ones. Anyway, all of this won't work well for Kepler, since it is made to study faint stars, for which there are no radial velocity measurements.
If it will study NEOs, perhaps it can also help New Horizons find a new target, or is it too late for that or not suitable?

Kepler's results call for follow-up observations of the same field both as a way of finding new planets and verifying those which are candidates but unconfirmed.

Stellar parameters are very noisy for Kepler targets in general: Errors of about 20% or more are common, which means that any given "Earth" which is found might actually be sized significantly larger than supposed.

In many cases, there is no way to distinguish, using Kepler data alone, between an earth-sized planet orbiting the target star and a Jupiter-sized planet orbiting a background star which is 1% as bright as the target star (or a Neptune orbiting a background star which is 6% as bright, etc.) Ground-based spectroscopy can help resolve these.

We will also undoubtedly have many prospects which are individually of low probability, but cumulatively will produce some real planets. In these cases, possible transits of very low signal-to-noise ratio will provide a specific expectation of when future transits might occur, so observatory time can be slated for the precise time when another transit should occur, if the object is real.

In fact, every system that has shown any sign of real planets will be worth following up, because of the likelihood of favorable orbital inclination for planets which were not found by Kepler, either in longer periods, or of smaller radius.

So from 190,000 original targets, Kepler will give us 3,000 worth following up, for various reasons, and this is much more tractable.

The down side: The typical distance of Kepler systems is about 2,500 light years, so many kinds of observations will be difficult or even impossible. Ultimately, much of the science determining the characteristics of exoplanets will focus on the worlds about 10-100 times closer than that.

Which is why I am eagerly awaiting TESS, which will look at much brighter stars with far more easily studied planets.

But that's a topic for another thread.

If you already have the transit ephemeris, you already have a preliminary radius for the planet. The R-M effect amplitude is actually (partially) a function of the planet's radius.


Oh but you can! Since the R-M effect amplitude is radius-dependent at the wavelength of observation, you can use the R-M effect to constrain atmospheric properties (though it is harder to do than with photometry).
http://arxiv.org/abs/arXiv:0903.2217

But we need not limit ourselves to spectroscopy. Ground-based photometry can be done. For some of the smaller worlds, it will be challenging to do, but still possible.

All you said was that transits can't be spectroscopically detected. I was merely demonstrating that they can, not arguing in favour of a spectroscopic transit survey.

It's wise to keep in mind that Kepler's prime mission goal was to determine the overall statistical frequency of planet distribution, and that goal's been well met. Pretty much every other finding has been icing on that cake.

Hungry: You're right. I was too hasty in my reply again and should have expanded a bit to make my point. Very interesting about the atmospheric detection using R-M. But, like you say, it will all be hard. I still think these instruments would not be suitable to detect the transit or determine the radius of the kind of planets that CHEOPS will target, without any prior knowledge of the transit, which was (supposed to be) my point to Doug.

In principle, it is of course also possible to image these planets But I would still normally say it is not possible (now). I guess that's just the way I talk... :/

Anyway, sorry to stray you off topic. I thought I was just mentioning an interesting side point about CHEOPS.

We have prior knowledge of the transits thanks to Kepler's data to date. That was my point. Nothing to do with detection - simply a comment on the suggestion that Kepler be fired up just to observe the odd transit already known to be occurring. I'm really not sure how or why the discussion went down the CHEOPS route. FWIW - CHEOPS is still only in a study phase- it hasn't been selected for flight.

nprev, Kepler has certainly filled out our knowledge of intrinsic planetary frequency, or ƒ, but there are a lot of crucially interesting blanks:

1) ƒ as a function of stellar class and metallicity (and planet radius, which is known to be inter-related).
2) ƒ for earth-sized planets at any orbital period beyond ~100 days. This was a primary goal and pending further analysis, is unmet.

Much more to say about (2):

We know the distribution of Super Earths, within some bounds, into the habitable zone, and further analysis will certainly improve that information. Then it's a reasonable hypothesis that the distribution of Earths as a function of increasing orbital period might follow the same trend as Super Earths, but until we have data, that's only a hypothesis.

My belief is that further analysis will almost definitely give us information on ƒ[Super Earth] to periods of 300-500 days, which will help set expectations of ƒ[Earth] for those periods, but not provide any hard evidence.

If, as is likely, analysis of past observations can cover periods of 300-500 days, then even the non-detection of any Earths will provide an upper bound on ƒ[Earth] for those periods, which would still be useful (in a pessimistic sense).

[Fressin et al, 2013] showed that there is a significant rate of false positives wherein a larger body transits a background star (or dimmer binary companion), so their method is to estimate the rate of true positives in each size category, from largest to smallest. This provides estimated rates of false positives, but doesn't identify which candidates are false positives, so if the number of detected Earths in the 300-500 day range is, as is extremely likely, a small integer (possibly zero), it will not be easy to tell if one or all of those are false positives. But we'll probably end up either with zero detections (and this an upper bound on ƒ[Earth]) or a very small number of systems for detailed follow-up observations, which would have a good chance of distinguishing between actual Earths and false positives.

I'm sorry. I was just whimsically wondering that if we had, say a month of off and on third reaction wheel use (I'm no expert on this), what would be the best use of fine pointing. Since the data seems to just be getting to confirming earth type planets I was looking for a way to maximise their confirmation.

Before each quarter of Kepler data is collected, there's a phase of getting the craft pointed correctly for that quarter. The time between quarterly science operations can be as short as about 20 hours, which includes data downlink, but it's not as straightforward as, say, getting one image of Dione from Cassini, where the pointing is tolerant by many pixels.

Given any highly interesting systems, earth-based observation is merited in any case. The Kepler field is going to garner a lot of observation time in the years to come, not because it's intrinsically unique, but because of Kepler having given us indications of which stars might have transiting planets.

Here's a big update from Kepler: The KOIs from Q1-Q12.

http://spaceref.com/exoplanets/kepler-deli...et-hunters.html

Some key takeaways from the article and my commentary:

1) As I've mentioned before, there is a major issue of instrument noise adding false positives to the data.
2) KOIs are not to be confused with candidates. Many of these KOIs are sure to be instrument-related false positives and many others are sure to be astrophysical false positives (e.g., eclipsing binary stars).
3) There will be a release later this summer of Q1-Q16 results. This will be, probably, the last set of observations from Kepler. It will resolve a lot of the false positives in the Q1-Q12 results.

No updates on the reaction wheels yet (no news is good news?), but found this interesting post-reaction-wheel proposal to point it at the galactic bulge for a microlensing survey:
http://arxiv.org/pdf/1306.2308v1.pdf

"In contrast to Kepler's current primary hunting ground of close-in planets, its microlensing planets would be in the cool outer parts of solar systems, generally beyond the snow line. The same survey would yield a spectacular catalog of brown-dwarf binaries, probe the stellar mass function in a unique way, and still have plenty of time available for asteroseismology targets."

Some updates from the Kepler Asteroseismic Workshop in Sydney: the Ames folks told that expect news on the timescales of months. It will take considerable time to figure out if the reaction wheels can be brought back online or not. Wheel #4 is jammed so the only way is to apply all torque there is and just crunch whatever debris is blocking it. Wheel #2 is still movable so any action will be much more delicate and well considered (meaning slow).

According to CNN, tests on the two failed reacton wheels have begun. On thursday they tried to spin the wheel that failed in may. It spun counterclockwise but not clockwise.

Further details of additional reaction wheel tests can be found on the following NASA page LINK

Wish them luck....

Mission update:
http://spaceref.com/exoplanets/kepler-miss...nting-test.html

Slow and steady....

It's now been accepted that Kepler will not resume operations as it was designed for. The 16 quarters of data on the ground complete Kepler's data set for that mission, which consisted of its primary mission and a relatively brief extension before the reaction wheel failure that caused the loss of fine pointing capability.

It remains open what use Kepler will serve, but it clearly maintains significant capabilities that don't require the exceptional fine pointing that defined its intended use. Looking for exoplanets via microlensing events seems to be one possibility, but it could potentially do any of hundreds of things that involve visible light frequencies and a very wide field of view.

The four-year duration of observations allows four transits to have been recorded for planets with orbital periods of a year, and this may be crucial to its mission to find planets of earthlike size and temperature. Unexpectedly high levels of noise (partly due to the instrument, partly due to stars themselves -- the Sun was the only star we knew so well before Kepler) compromise the validity of many cases where only three possible transits were detected. A fourth transit is amazingly powerful in ascertaining the validity of an astrophysical event, because three false transits are far more likely to occur by chance.

I plan to submit a publication soon on the results from the first twelve quarters and I know a torrent of publications will follow as the additional four quarters' data is analyzed.

Is the Kepler team still taking suggestions on what to do? I am personally am a fan of using the telescope for microlensing. There could be so many applications from that alone.

On a happier note, there are still going to be quite a few discoveries in store. Most interesting right now are these ultra, ultra ,short period candidates.

http://www.space.com/22429-alien-planets-k...-lava-iron.html

With planets this close to the star it might be possible to use radial velocity to find their masses, despite how small they are.

They have put out an AO for interested members of the scientific community.

People are now starting to post their ideas for what to do with Kepler (white papers) on astro-ph: (just browse through the last few days of http://arxiv.org/list/astro-ph.EP/recent ; or do a search). Interesting to see the differences.

Edit- here's a quick list (might be missing some)

http://arxiv.org/abs/1309.1177
http://arxiv.org/abs/1309.1176
http://arxiv.org/abs/1309.1096
http://arxiv.org/abs/1309.1078
http://arxiv.org/abs/1309.0918
http://arxiv.org/abs/1309.0654

Here's a link to my latest analysis of Kepler data and the Eta Earth question: How common are earthlike planets around sunlike stars?

http://www.spacedaily.com/reports/Kepler_F...Earths_999.html

This was written concurrent with an effort to publish that work in a journal.

I wrote the SpaceDaily piece immediately before the Exopag conference which has some excellent papers on the Eta Earth question (and other topics I haven't gotten around to):

http://exep.jpl.nasa.gov/exopag/exopag8/agenda/

I think my article and the Eric Ford paper at Exopag converge on one message: Even though Kepler's data collection of this kind is complete, the most exciting discoveries (especially regarding Eta Earth) are still to come from analysis of the data and follow up observations. We may already have the first earthlike planet on a short list of possible planets, but confirmation will take some time.

oh... wow!
Extrasolar planets: An infernal Earth



An Earth-sized planet with an Earth-like density
A rocky composition for an Earth-sized exoplanet

<VERY HOT EARTH-SIZED, EARTH-MASS PLANET>

An interesting story. A breakdown of what this really reveals:

Nobody is surprised that worlds like this exist. The accomplishment is mainly in measuring a specific example. Kepler produces loose measures of planets' radii with no information about mass, and therefore, no information about density.

There may be terrestrial planets where the optical radius is very different from the lithosphere's radius. Titan is the closest thing in our solar system, where the haze extends hundreds of km above the surface. There is suspicion that some exoplanets may have an even greater discrepancy.

However, on a world so incredibly hot as this one, it doesn't seem like there was much chance of an atmosphere as we know it, so a high haze layer would have to be something really bizarre, like gaseous iron.

Beyond that, it's interesting to see what the range of density is for terrestrial planets, regardless of temperature. Our solar system has nearly a three-way tie at the top of the density distribution, and it's unknown what the distribution might be in general. Only rare exoplanets permit a good measurement of both radius and mass with current methods, so the data is still trickling in.

[quote name='Paolo' date='Oct 30 2013, 09:02 PM' post='204213']
oh... wow!
Extrasolar planets: An infernal Earth

GREAT ! They found where Crematoria is... Now we can save Riddick !
Click to view attachment Click to view attachment Click to view attachment Click to view attachment

Haha. Those are some nice pictures. They remind me of science fiction stories about Mercury back in the fifties. Though this planet makes Mercury seem like Earth in comparision.

The second Kepler science conference begins today. The last one was in 2011, so I am looking forward to some exciting announcements!

http://nexsci.caltech.edu/conferences/KeplerII/index.shtml

Pretty cool:

More on Earth sized planets found by Kepler.
Berkeley UC: http://newscenter.berkeley.edu/2013/11/04/...itable-planets/

Among the great cases of hazardous interface between science and the media, we have the terms "earthlike", "Earths", "Earth-size" and so on.

The Petigura, et al, study which stirred great interest with its release yesterday is using the terms in a much looser fashion than some previous work.

Their definition of Earth-size is 1-2 RE (1 RE = the same radius as Earth).

Their definition of "lukewarm" or "habitable" (loaded term there, obviously) is receiving 1/4 to 4x the light that Earth receives from the Sun, which for a star exactly as bright as the Sun would be 0.5 AU out to 2.0 AU.

It's fine to define these however one wishes and proceed, but there is a noteworthy lack of standard. The work of Fressin, et al, used 0.8-1.25 RE as their definition of Earth-size. Neither of these two ranges is a proper subset of one another, and given a power law distribution of sizes, the size range of the newer study is simply larger and should boost the numbers considerably.

Meanwhile, the temperature range (or rather, equilibrium temperature range, since we have no solid evidence of what climate dynamics may exist in these worlds) is curiously large, including worlds getting twice the radiative heating of Venus.

Finally, the work itself cites a value of 11±4% of sunlike (we could also examine the definition of sunlike) stars possessing such planets whereas the popular press circulates double that number.

Given ranges of size like those of Fressin, et al, and equilibrium temperatures ranging from Venus to Mars, that 11% would be considerably reduced, to under 5%.

There's no right or wrong here – any sort of science that can examine such worlds (none of which has actually been discovered yet: the study depends on a reasonable extrapolation of candidate planets) is still far off – but the definitions have a fuzziness to them about which anyone with the opportunity to do so should inform interested individuals.

Well said JRehling.

Back in 1960's when Stephen Dole published 'Habitable Planets for Man' http://www.rand.org/pubs/commercial_books/CB179-1.html
I repeatedly grabbed this book out of the local library.

Dole's definition of habitable was a world humans could walk on and breath free. Stirred my imagination no end.

At that time this sort of study was considerd fringe science.

Now here we are almost 50 years later streaming science presentations from a conference with 400 attendees
going over varients of this topic.

Star trek without the starships exploring a universe vast and unforgiving. And we are participants at the beginning of it all.

Good time to be alive!

Craig

That's an interesting paper... Here are the relevant sentences:
"Previously, we showed that 11 ± 4% of stars harbor a planet having an RP=1−2 R⊕ (Earth radius) and FP=1−4 F⊕ (Earth flux)."
"If one was to adopt FP=0.25−4 F⊕ as the HZ and extrapolate from the FP=1−4 F⊕ domain, then the occurrence of Earth-size planets in the HZ is 22 ± 8% for Sun-like stars."

They basically took the cumulative number of planets per period and extrapolated it into the 200-400-day range where there is no data yet. That can be then converted into the <1 Earth-flux regime. Bit of a ballooning here, isn't it?

They also hid other values into a table:

The latter two are extreme cases for the inner and outer edges with weird atmospheres and stuff. So the "classical", narrow-HZ definition gives only 8.6%. Main point? We know loads more about exoplanets than about habitability...

That's a nice breakdown.

I read the paper last night: There's a short paper and a massive collection of additional materials. Some highlights I had not previously noticed:

Petigura and Marcy have invented their own lightcurve analysis pipeline to bypass the Kepler team's pipeline which does not adequately address instrumental noise.

The ten most favorable candidates they found amounted to 8 which were in the Kepler team's TCE release from last December and 2 which were not in the TCE release. That gives a slight idea how the Venn diagrams of discoveries from the two approaches might overlap.

None of those ten meet the stricter definitions of earth-sized in the habitable zone. One is pretty close, being just a little hot. A few have earth like equilibrium temperature, but significantly larger radius. As always, the uncertainties of size and equilibrium temperature of Kepler candidates is very large, so any particular candidate may be much larger, much smaller, or even not really exist.

All of the ten are associated with K-class stars. Generally speaking, the transiting method is more likely to show you more planets, for a given temperature, around smaller and cooler stars, because the orbital periods are shorter and completeness is improved more by the shorter orbital periods and semi major axes than it is hurt by the smaller star radius allowing fewer cases of geometrical alignment. So all of their analyses are about K stars and they extrapolate to G stars.

Broadly speaking, K and G stars should have similar prospects for earthlike environments, and probably K stars with terrestrial planets will greatly outnumber G stars with terrestrial planets. That trend continues for M stars (red dwarfs) but they will likely have planets that are tidally locked in their habitable zones, for what that's worth. (Always? Sometimes? This isn't certain.)

Extrapolation is always a bit dodgy, and the current result depends upon extrapolation in three different parameters: stellar class, planet radius, and planet temperature. It remains the case that every candidate deemed an "earth like" planet is estimated as being either hotter or larger. Certainly the extrapolations made by this work are smaller than the extrapolations performed by earlier estimates of Eta Earth.

I've seen it paraphrased on Twitter (so I pass this on with a grain of salt) that Borucki has concluded that discovering any planets of earthlike size and temperature will depend on future missions. The most promising angle, it seems to me, for truly confirming any Kepler candidate with earthlike statistics would be if it had an orbit synchronous with an interior planet that was easier to confirm with doppler-based methods, and that would make it much more credible that the earthlike planet truly exists, since random noise would not synchronize with a real planet.

The analysis goes on. I think at this point we can say that only rather devious characteristics in the planetary frequency function would make the extrapolations in Petigura and Marcy invalid, and barring such deviousness, Eta Earth is going to vary from about 4% to 25%, depending on the definitions one uses.

Any comments on the estimates of how close the nearest Earthlike planet might be? Since the number of sunlike stars should increase with the cube of the distance, that might reduce the amount of uncertainty for that parameter. (Of course, it also increases the size of the search that would be needed to identify such a planet.)