2024MNRAS.529..104M

Geostatistical methods are powerful tools for understanding the spatial structure of the metallicity distribution of galaxies, and enable construction of accurate predictive models of the 2D metallicity distribution. However, so far these methods have only been applied to very high spatial resolution metallicity maps, leaving it uncertain if they will work on lower quality data. In this study, we apply geostatistical techniques to high-resolution spectroscopic maps of three local galaxies convolved to eight different spatial resolutions ranging from ∼40 pc to ∼1 kpc per pixel. We fit a geostatistical model to the data at all resolutions, and find that for metallicity maps where small-scale structure is visible by eye (with ≳10 resolution elements per R_e), all parameters, including the metallicity correlation scale, can be recovered accurately. At all resolutions tested, we find that point metallicity predictions from such a geostatistical model outperform a circularly symmetric metallicity gradient model. We also explore dependence on the number of data points, and find that N ≳ 100 spatially resolved metallicity values are sufficient to train a geostatistical model that yields more accurate metallicity predictions than a radial gradient model. Finally, we investigate the potential detrimental effects of having spaxels smaller than an individual H II region by repeating our analysis with metallicities integrated over H II regions. We see that spaxel-based measurements have more noise, as expected, but the underlying spatial metallicity distribution can be recovered regardless of whether spaxels or integrated regions are used.