National and regional high-quality seismic background noise models for Italy

Abstract We present new national and regional seismic noise models for Italy based on continuous vertical-component recordings acquired between January 2024 and February 2025 at 349 broadband seismic stations. A rigorous quality-control and data-selection workflow was developed to identify station malfunctions, metadata inconsistencies, data gaps, and earthquake-related seismic transients using conservative empirical magnitude–hypocentral-distance criteria. After quality assessment, 283 stations were retained for the model construction. The Italian Seismic Noise Model ( ISNM ) was derived by equally stacking normalized station-level probability density functions (PDFs), with the 5th (P 5 ) and 95th (P 95 ) percentiles defining its lower and upper boundaries. Compared with the previous Italian model, the ISNM defines a narrower spectral envelope, with its lower boundary averaging 4.1 dB higher and its upper boundary 11.9 dB lower over their common frequency range. Sensitivity analyses show differences within approximately ±1.5 dB below 10 Hz between models obtained with and without data selection, whereas quality control substantially reduces anomalous contributions to the high-frequency upper tail. Regional models reveal distinct spectral characteristics: the Po Plain shows elevated noise above 1 Hz, Campi Flegrei and Vesuvio exhibit enhanced energy at approximately 0.2–1 Hz, whereas Etna and Stromboli show pronounced features at 2–5 Hz. Additional models based on Italian seismic-code site classes highlight differences among site conditions, while daytime–nighttime comparisons for classes A, B, and C-D-E indicate an increasing anthropogenic contribution above 1 Hz. Bootstrap analysis shows frequency-dependent sampling variability, generally greater for the upper model boundaries.