Why develop high-level goals? • Operational decisions flow from goals, in principle. • Develop metrics of success against which we can assess our work. • Increase probability of achieving high-impact results.
Does 40-ish targets make sense? • The number 40 is set by a certain fixed idea about observing: • Every star is observed every possible night for 20-ish minutes. • Every star gets some 1000 measurements. • Are there better or different observing choices? • How about 80 stars observed every second night? • Observe the same star many times per night, at least some nights? • Vary exposure times adaptively or randomly? • Adapt the target list over the decade? • Answers should and can flow from our high-level goals.
Possible kinds of goals: • (G1) Find as many planets as we possibly can in a decade of observing. • (in some ranges of stellar and planetary properties, maybe?) • (G1d) Or some future-discounted version of that. • (G2) Determine the rate at which Sun-like stars host Earth-like planets. • (G3) Determine planet occurrence as a function of star properties. • (G4) Learn everything we can about the planetary systems of 40-ish particular stars.
Combining goals • In principle we don’t need to choose only one goal. • But if we combine goals, we must do so in a convex manner. • It’s sometimes useful to use an explicit cost-benefit framework. • Goals have values expressed in Euro. • Operational activities have costs expressed in Euro.
(G4) Learn everything about 40-ish stars • This is the simplest possible high-level goal for T.H.E. • It motivates thinking very hard about which 40 stars we should target! • (note to self: Observing proposals!) • Even under this goal, the observing plan of one identical observation per star per night will not deliver the most informative data set. • Mitigation of p-modes and star spots. • Avoiding unintended periodic observing patterns and aliases.
(G1) Find as many planets as we can. • Probably we would restrict this goal: • Find as many planets as we can (in the decade) in X mass range and Y period range around stars of type Z. • Motivates choosing targets wisely. • Motivates adaptively dropping and adding targets as we learn things. • Motivates a stellar (and not just instrument) commissioning period!
(G1d) A future-discounted version of (G1). • In reality we do have a discount rate! • It is useful to have good science results early, for morale and funding. • Student and postdoc career considerations drive discount rates up. • Motivates adaptive scheduling that is responsive to early results. • Motivates (probably) even more aggressive adaptation in general. • Works well with cost–benefit approaches. • Breaks some operational degeneracies.
(G2) Determine the planet occurrence rate. • Again, we would restrict this goal: • What is the rate at which stars of type Z host planets with properties X and Y? • Earth-like planets around Solar twins? • Motivates choosing a 10-year plan up-front and sticking to it! • Adaptation is possible, but rules must be agreed-upon in advance. • This is a very hard way to live. I’m willing to try this. • Radical idea: Do more than one survey. • What? See next slide.
Radical idea: Do more than one survey! • Instead of a 10-year survey, do a 1-year survey, a 2.5-year survey, and a 6.5-year survey. • The goals and targets of each survey could be informed by the results of the previous survey. • Many stars could still easily get 10 years of observing. • This would permit much more flexible adaptation. • This would involve extremely hard turn-around efforts at each survey- to-survey boundary. • We might want to negotiate about our time-sharing.
Adaptation and statistics. • In general the statistical goal of determination of the occurrence rate is at odds with the discovery goal of finding planets. • Occurrence rate determinations require that we can figure out what the survey would have found in different universes. • Counterfactuals. • We cannot do things in operations by hand: • No choosing targets by hand. • No adjustment of targets or operations during the entire survey. • (Hence maybe split to multiple surveys?)
(G3) Rate as a function of star properties. • Again, we would restrict this: • Earth-like planet occurrence rate, say, as a function of star properties. • Star properties might include mass, age, and composition (in some ranges). • This is similar to (G2) but requires diversity in the stellar targets. • Motivates observing more stars, less frequently.
Yes, I do live in the real world. • In reality, projects I have worked on decide what they can do, and then reverse engineer high-level goals that are appropriate. • That might sound like cheating, but it’s not; it’s healthy. • We are permitted to use a consequence of a high-level goal as an argument in favor of that goal. Or against that goal.
What do I want? • I want T.H.E. to succeed. • I want clearly defined and achievable metrics of success. • I want as much insurance as possible against “getting unlucky”. • I want early science results that are exciting and valuable, especially for early-career team members. • I want long-term legacy value for our data and our work. • Operations must be (close to perfectly) algorithmic.
Recap: • (G1) Find as many planets as we possibly can in a decade of observing. • (G1d) Or some future-discounted version of that. • (G2) Determine the rate at which Sun-like stars host Earth-like planets. • (G3) Determine planet occurrence as a function of star properties. • (G4) Learn everything we can about the planetary systems of 40-ish particular stars.
dwhgg
ssii
tagtag
ynakano
langstat
lcolladotor
konakalab
productmarketing
afnizarnur
lara
akmur
marcduiker
aarron
iamctodd
sferik
addyosmani
jrom
dotmariusz
michaelherold