Reconstructing Star Formation Histories of High-Redshift Galaxies: A Comparison of Resolved Parametric and Non-Parametric Models

Abstract

We investigate the optimal approach for recovering the star formation histories (SFHs) and spatial distribution of stellar mass in high-redshift galaxies ($z\sim 2-5$), focusing on the impact of assumed SFH models on derived galaxy properties. Utilizing pixel-by-pixel spectral energy distribution (SED) fitting of multi-band photometry, we explore various parametric SFH models (including exponentially declining ($τ$), delayed-$τ$, lognormal, and double-power law) alongside spatially resolved non-parametric methods. We first analyze the models using simulated galaxies and then apply them to observed galaxies for validation and as proof of concept, with additional comparisons to results from unresolved SED fitting. Our findings demonstrate that pixel-by-pixel analysis with parametric models is particularly robust in recovering the true SFHs of simulated galaxies, with the double-power law model outperforming others, including non-parametric methods. This model excels in detecting recent starbursts within the last 500 Myr and capturing the stochastic nature of star formation. Conversely, unresolved photometry with simplistic parametric models tends to produce biased estimates of key galaxy properties, particularly underestimating early star formation. Non-parametric methods, resolved or unresolved, typically yield older mass-weighted ages. Biases in early-time SFRs, likely introduced by prior assumptions, further complicate these models. We conclude that the double-power law model, applied in a pixel-by-pixel framework, offers the most reliable recovery of SFHs and produces robust stellar mass maps. Resolved methods simplify modeling dust and metallicity, enhancing parameter interpretability and underscoring the value of flexible parametric models in spatially resolved analyses.