Exoplanet Demographics Combining Multiple Detection Method
PI: Eric Ford (Astronomy & Astrophysics)
Targeting a postdoc (or assistant research professor) who already has partial support lined up from grants or other internal funds.
Level of effort: 3-6 months of effort from a postdoc or assistant research professor
Summary Sentence: This project aims to develop simulation-based inference (SBI) tools for characterizing the intrinsic distribution of exoplanets while combining observational constraints from multiple exoplanet detection techniques.
Description: The PI pioneered the application of Approximate Bayesian Computing (ABC) to improve the characterization of the intrinsic distribution of exoplanets as a function of size and orbital separation using data from NASA’s Kepler mission (Hsu et al. 2018, 2019). While Kepler observations still provide the largest, homogeneous sample of exoplanets, other exoplanet detection methods are complementary, offering greater sensitivity to planets that orbit further from their host star and the ability to characterize additional properties (e.g., planet mass; Crass et al. 2021). Unfortunately, each detection technique has its own set of selection effects and detection biases, making rigorous analysis of the intrinsic population very challenging.
This project aims to build SBI tools that can combine observational constraints from multiple exoplanet surveys including qualitatively different exoplanet detection methods. We can provide packages and example codes to generate simulated planetary systems (He et al. 2019, 2020), to simulate observations by the Kepler mission (Hsu et al. 2019), to simulate observations by radial velocity surveys (Hara & Ford 2023), and to perform ABC (Hsu et al. 2019, He et al. 2019, 2020). We can also point to tutorials for performing SBI in other contexts. An early task for the Junior Researcher will be to assess whether it is best to evolve the tools for performing ABC or to develop more flexible tools using more general forms of Simulation-based Inference (SBI; e.g., normalizing flow as a neural posterior) given the size of the available data and dimensionality of the parameter space describing the intrinsic distribution of exoplanets. Then the ICDS junior researcher would integrate the above tools with an ABC or SBI method to perform inference on a hierarchical Bayesian model for the intrinsic exoplanet population. They will implement end-to-end tests using simulated data, and document and release their code so that other research projects and teams can apply it as additional data becomes available and/or extend it to incorporate additional exoplanet detection methods.
Expertise/skills of interest:
- Statistics: Uncertainty Quantification, Hierarchical Bayesian Modeling, Approximate Bayesian Computing and/or Simulation-based Inference
- Astronomy: Exoplanets, Radial velocity method, Transit method, Transit timing method.
- Programming experience using Julia and Python or substantial experience with one and willingness to learn the other.
Expectations:
- Postdoc or assistant research professor with experience and/or training in Astronomy & Astrophysics, Physics, or a related field.
- Write code organized into small functions with appropriate documentation and unit tests that will be released as open-source software.
- Produce Jupyter notebooks, scripts, and project environment files that support conclusions and make results readily reproducible.
- Weekly meeting and project updates with faculty advisor(s). Participate in group meetings (~1 hour roughly twice a month).
Goal: Characterize the intrinsic distribution of exoplanets while combining observational constraints from multiple exoplanet detection techniques. Develop a toolbox that can support future funding proposals to NASA ROSES and/or the NSF AAG calls.
Specific Objectives:
1. Develop a module providing a forward model for generating simulated planetary systems, building on previous models for simulating data from transiting exoplanet surveys (He et al. 2019, 2020).
2. Develop a module providing a forward model for generating simulated observations from transit (Hsu et al. 2019) and radial velocity surveys (Hara & Ford 2023), building on previous codes.
3. Develop a module to perform Approximate Bayesian Computing or Simulation-based Inference to characterize properties of the intrinsic distribution of exoplanets.
4. Verify accuracy and quantify expected uncertainty due to finite sample sizes using simulated datasets and simplified detection efficiency model.
5. Optimize and parallelize above code as necessary to allow for practical application to realistic datasets.
6. Begin to explore incorporating more realistic (and much more computationally expensive) planet detection efficiency models into above framework. [if time permits]
Engagement: Ford is an ICDS co-hire and an Associate Director for the Center for Astrostatistics. Ford regularly participates in activities for ICDS co-hires and has served on numerous committees and in leadership roles related to ICDS.