[Paper Review] pcigale: porting Code Investigating Galaxy Emission to Python

Y. Roehlly, D. Burgarella+4 more|arXiv (Cornell University)|Sep 24, 2013

Astronomy and Astrophysical Research1 references3 citations

TL;DR

pcigale is a Python reimplementation of the CIGALE SED fitting code, designed to improve modularity, extensibility, and usability in astrophysical research. It decouples SED creation from analysis, enabling flexible model combinations, Bayesian fitting, and theoretical SED generation, with performance trade-offs for enhanced developer accessibility and integration with the astrophysics Python ecosystem.

ABSTRACT

We present pcigale, the port to Python of CIGALE (Code Investigating Galaxy Emission) a Fortran spectral energy distribution (SED) fitting code developed at the Laboratoire d'Astrophysique de Marseille. After recalling the specifics of the SED fitting method, we show the gains in modularity and versatility offered by Python, as well as the drawbacks compared to the compiled code.

Motivation & Objective

• To overcome limitations in the original Fortran-based CIGALE code, including poor modularity and difficulty in adding new models or analysis methods.
• To enable easy integration of new stellar population models, attenuation laws, IR templates, and AGN components through a modular object-oriented design.
• To decouple SED creation from statistical analysis, allowing independent development of fitting algorithms and SED construction pipelines.
• To support alternative uses beyond fitting, such as generating synthetic galaxy catalogues for evolutionary modelling.
• To leverage Python’s ecosystem for better accessibility, maintainability, and community-driven development in astrophysics.

Proposed method

• The core of pcigale is a pcigale.sed object that tracks SED parameters and component contributions through a chain of processing modules.
• SED creation is modular: each component (e.g., star formation history, stellar emission, attenuation, IR re-emission) is added via dedicated, pluggable modules inheriting from an abstract base class.
• Analysis is handled by separate modules, including a Bayesian-like fitting engine and a SED generation module for creating artificial galaxy catalogues.
• A command-line script automates configuration and execution, guiding users through initialization, configuration, and execution steps.
• The code uses Python’s object model and integrates with astropy for cosmology, file I/O, and data handling, enhancing usability and extensibility.
• Performance is optimized via Numpy array operations, though initial execution remains slower than the compiled Fortran version.

Experimental results

Research questions

• RQ1 How can a spectral energy distribution (SED) fitting code be redesigned to support modular, extensible, and pluggable components for stellar populations, dust attenuation, and IR emission?
• RQ2 To what extent does porting a high-performance Fortran SED code to Python compromise computational speed while improving maintainability and extensibility?
• RQ3 Can a decoupled architecture separate SED creation from statistical analysis, enabling independent development of fitting methods and model components?
• RQ4 How effectively can Python’s ecosystem and object-oriented design facilitate the integration of new astrophysical models and templates in SED fitting?
• RQ5 What are the practical benefits of using a high-level language like Python for SED fitting in large-scale galaxy surveys?

Key findings

• The pcigale framework successfully decouples SED creation from analysis, enabling independent development and reuse of SED modules and analysis methods.
• The modular design allows seamless addition of new models, such as updated stellar population templates, attenuation laws, and IR/AGN templates, without modifying core logic.
• A command-line interface simplifies configuration and execution, reducing barriers for non-expert users and streamlining batch processing.
• The use of Python and astropy enables better integration with modern astrophysical data workflows and enhances code readability and maintainability.
• While the initial performance is slower than the Fortran version, performance bottlenecks can be mitigated through Numpy-optimized operations and selective code rewrites.
• The framework supports both fitting observed photometry and generating theoretical SEDs, enabling synthetic catalog generation for evolutionary modelling.