An effective earthquake early warning system (EWS) should quickly identify destructive earthquakes and provide enough time for emergency responses before strong and damaging waves reach vulnerable areas. Accurate magnitude estimation is essential for issuing reliable alerts. This study examines the impact of site effects on earthquake magnitude estimation using the traditional predominant period parameter, \({\tau }_{P}^{max}\) . Since horizontal components of ground motion carry richer information but are highly affected by the site effects, we show that removing site effects enables the effective use of horizontal components for accurate magnitude estimation. For this purpose, we first applied a traditional frequency-domain approach—Fourier transform, deconvolution, and inverse Fourier transform—to remove site effects from the seismograms data at each station. To improve computational efficiency for real-time applications, we also designed an IIR filter based on the inverse of the site response function estimated using the horizontal-to-vertical (H/V) spectral ratio, to remove site effects directly in the time domain. The method preserves phase information and achieves results comparable to traditional frequency-domain deconvolution but with significantly faster computation, making it practical for real-time applications. Application of the proposed approach to K-NET strong motion records from Japan (1998–2022) showed that removing site effects improved magnitude estimates by approximately 0.1–0.2 units. Additionally, correcting vertical components also improved magnitude estimates by about 0.1 units.
