Energy magnitude is determined from seismic radiated energy over a broad frequency band. It provides a dynamic measure of earthquake scale that is more directly related to shaking strength and damage potential than magnitudes determined from seismic wave amplitudes at specific frequencies. Nevertheless, energy magnitude is still rarely used in routine source-parameter determination and early post-earthquake response. A key difficulty is that reliable radiated-energy estimates generally require the source spectrum to be recovered with appropriate correction for path effects and radiation patterns. The radiation-pattern correction generally depends on a focal mechanism solution, which is often unavailable during the first tens of minutes after a damaging earthquake. This dependence delays energy-magnitude determination and limits its application value for rapid hazard assessment. To address this problem, we develop a rapid and automated method for estimating radiated energy and energy magnitude for large shallow earthquakes without using focal mechanism solutions. This method uses vertical-component P-wave records from Global Seismographic Network teleseismic stations at epicentral distances of 30°−90°. For each station, the apparent source spectrum is obtained by deconvolving the synthetic Green’s function generated by an explosive source from the observed P wave. The explosive-source Green’s function is used as a physically based proxy for path effects, while the direct P wave and its depth phases are treated together as a P-wave group. The geometric mean of the resulting apparent source spectra from multiple stations is then taken as an approximation of the source spectrum. Radiated energy is then calculated from the squared source spectrum, and the corresponding energy magnitude is obtained from the standard energy-magnitude relation. Because this workflow does not require a mechanism solution, it can be performed as soon as hypocentral information and teleseismic P-wave data become available.

We illustrate the robustness and efficiency of the method in applications to three typical earthquakes with different mechanisms: the 2008 Wenchuan thrust earthquake （M_S8.0, M_W7.9）, the 2021 Maduo strike-slip earthquake （M_S7.4, M_W7.4）, and the 2025 Dingri normal-faulting earthquake （M_S6.8, M_W7.1）. For the Wenchuan earthquake, we obtain a radiated energy of 1.69×10¹⁶ J, equivalent to an energy magnitude M_e of 7.9, consistent with most previous estimates. For the Maduo earthquake, the radiated energy and the energy magnitude are 6.14×10¹⁴ J and M_e7.0, respectively, close to the results in Incorporated Research Institutions for Seismology （IRIS） catalog. For the Dingri earthquake, the resulting radiated energy is 8.44×10¹⁴ J, and the energy magnitude M_e is 7.1, slightly higher than the results of IRIS. We further investigate 35 shallow earthquakes with M_W≥7.0 that occurred in China and its adjacent regions since 2000. The resulting energy magnitudes agree well with the IRIS estimates for most events, with an overall scatter of about 0.18 magnitude units. The remaining deviations partly arise from the use of an explosive source as a mechanism-independent approximation. This approximation may lead to overestimation for some dip-slip events and underestimation for high-dip-angle strike-slip events, because their true radiation patterns can differ systematically from the assumed isotropic source. In addition, factors such as direct spectral division without constraints on the source time function, simplified P-to-PP time-window selection, limited upper-frequency, rupture directivity, and three-dimensional Earth structure may also affect the results. However, these unconsidered factors may partially compensate for one another, which explains why the rapid estimates remain close to other results in many cases.

The proposed method is computationally efficient and therefore suitable for operational implementation. On a mid-range multi-core desktop CPU, the calculation for a single event can be completed within 30 s, implying that the radiated energy and corresponding magnitude can be obtained about 30 s after arrival of PP phase at the farthest station. Tests using progressively larger maximum epicentral distances indicate that, after outlier removal, estimates based on approximately 5 stations and a maximum distance of about 40° are generally sufficient to obtain reliable results. Therefore, the rapid estimate can be initiated when the PP phase reaches 40°, and updated dynamically as data from more distant stations become available. In typical large-earthquake scenarios, the energy magnitude can be obtained about 10−20 min after the origin time, without waiting for mechanism solutions. Once the focal mechanisms become available, mechanism-specific Green’s functions can be used to further refine the radiated-energy estimate. Nevertheless, the proposed mechanism-independent approach provides a simple, physically interpretable, and operationally efficient tool for early assessment of the damage potential for large earthquakes. It can complete conventional magnitude estimates, improve rapid evaluation of earthquake scale and shaking severity, and support emergency response by providing timely information on radiated seismic energy.