Systematic Biases in BPT Diagram Classifications from Emission-line Signal-to-noise Cuts

The Baldwin–Phillips–Terlevich (BPT) diagram is a widely used diagnostic for separating star-forming galaxies from galaxies hosting active galactic nuclei (AGN). The BPT diagram requires reliable measurements of four key emission lines. Most studies impose a per-line signal-to-noise (S/N) threshold on these emission line fluxes, but this S/N threshold is often unreported, and rarely tested as a source of bias. Here, ∼10⁴ galaxies from SDSS DR19 are used to evaluate how the classification is affected by selection criteria at S/N > 2, 3, 5, and 10. I find that raising the threshold from S/N > 2 to S/N > 10 increases the star-forming fraction from 69% to 85% and decreases the composite and AGN fractions by ∼50% relative. These changes are systematic, reflecting the preferential loss of AGN and composite systems with weak Hβ or [N ii]. Thus it is recommended that BPT-based studies explicitly state S/N cuts and exercise caution when comparing results based on different thresholds.