Rapid P-Wave Moment Magnitude Estimation from Strong-Motion Records: Evidence from the 2025 Marmara Sea Earthquake

Abstract

The initial seconds after an earthquake are critical for rapid magnitude estimation to support real-time early warning. This study evaluates the determination of P-wave moment magnitude (M_wp) using strong-motion records from the 23 April 2025 Marmara Sea earthquake. High-quality accelerometric data from the Turkish National Strong Motion Network were analysed to extract early P-wave features within the first 3 s after P-wave onset. Results show significant rupture-directivity effects, whereby stations located approximately along the fault strike and rupture-propagation direction recorded larger ground-motion amplitudes and higher station-based Mwp estimates than stations located near nodal directions. The mean M_wp was 6.5 ± 0.2, consistent with the Global Centroid Moment Tensor (GCMT) moment magnitude estimate. Magnitude estimation was achievable within 8–20 s of P-wave arrival, confirming the method’s real-time applicability. Our findings demonstrate that strong-motion P-wave analysis can provide rapid and reliable magnitude estimates suitable for earthquake early warning, tsunami warning, and rapid-response applications. In the Marmara Sea region, where tsunami arrival times may be on the order of 20–30 min and critical infrastructure is concentrated in densely populated coastal areas, rapid determination of magnitude within seconds of earthquake initiation can provide valuable information for emergency management and hazard mitigation decisions.

1. Introduction

Rapid estimation of earthquake magnitude immediately after P-wave detection is an important component of earthquake early-warning (EEW) systems. Although the earthquake rupture has already begun by the time the first P-waves are recorded, rapid characterisation of event size can provide useful information before the arrival of stronger S waves and surface waves at more distant locations, thereby enabling automated protective actions and supporting tsunami-warning and emergency-response operations [1,2,3]. Conventional approaches, which rely on S-waves or full-waveform inversions, require much longer—often tens of seconds to several minutes—thereby reducing their effectiveness for rapid alerts to near-field events. Recent advances in seismic data processing, real-time waveform analysis, and the expansion of dense seismic and strong-motion networks have enabled the extraction of magnitude-related information from the earliest portion of P-wave recordings, supporting rapid earthquake characterisation and early-warning applications [4,5,6,7,8,9].

Türkiye, located at the junction of the African, Eurasian, and Arabian plates, is among the world’s most seismically active countries. Two principal fault zones, the North Anatolian and East Anatolian, define its tectonic setting [10,11,12,13,14,15]. The North Anatolian Fault Zone (NAFZ) is a right-lateral strike-slip fault, whereas the East Anatolian Fault Zone (EAFZ) is left-lateral. Graben systems in western Türkiye further contribute to regional tectonics (Figure 1). Consequently, Türkiye frequently experiences significant earthquakes.

The magnitude of a seismic event is one of the parameters that define it and give users and decision-makers an idea of how people and the environment can be affected. In this context, different magnitude types are defined by different researchers, such as local magnitude (M_L), duration magnitude (m_d), body-wave magnitude (m_b), and surface-wave magnitude (M_S). The 26 December 2004 Sumatra-Andaman earthquake and the 17 July 2006, Java earthquake were milestones in the development of rapid magnitude determination. These earthquakes caused tsunamis, leading to many deaths. Tsunamis are devastating if they occur at regional distances and arrive at the coast in 20–30 min. For this reason, any warning should be disseminated within 15 min [16]. Ref. [4] calculated a moment magnitude using the P-wave portion of broadband seismograms, often called M_wp. However, early implementations of P-wave-based M_wp estimation occasionally underestimated the magnitude of very large earthquakes because the assumed source duration did not adequately represent the exceptionally long rupture process, as observed during the 2004 Sumatra-Andaman earthquake.

Many researchers have developed methods for rapid earthquake magnitude estimation using the early portion of the P-wave [4,5,6,7,8,16,17,18,19,20,21,22]. Early work demonstrated the viability of calculating moment magnitude (M_wp) from broadband seismograms, and subsequent refinements incorporated additional seismic phases and tailored algorithms to enhance accuracy, especially in regions with limited station coverage. Major efforts have focused on improving real-time processing and reducing underestimation in large events, as seen in the 2004 Sumatra earthquake. Further innovations have explored alternative combinations of waveform characteristics, including cumulative amplitudes, high-frequency durations, and empirical scaling relations. In recent years, approaches using strong-motion data have become more prominent, addressing limitations of broadband instruments in epicentral regions and enabling more robust magnitude estimation for both regional and teleseismic events. These evolving techniques have collectively advanced the reliability and speed of early-warning systems, with recent methods also introducing new empirical scales for global applicability.

The Marmara Region, located at the western end of the North Anatolian Fault Zone (NAFZ), is among the most seismically active and densely populated areas in Türkiye. The study area has experienced many destructive earthquakes since the historical era, e.g., 1509, 1766, 1894, 1935. The 1509 earthquake produced tsunami wave heights exceeding 6.0 m along all Marmara Sea coastlines. The 1999 earthquake generated tsunami wave heights exceeding 2 m at many locations around the Marmara Sea. This study focused on rapid magnitude estimation in tectonically active areas to warn people, decision-makers, and scientists, rather than on tsunami modelling. The 23 April 2025 earthquake provided a unique opportunity to investigate P-wave-based magnitude estimation in this tectonically complex region, where source directivity and strong-motion characteristics significantly influence recorded ground motions. The motivation is to improve moment magnitude calculations for small epicentral distances, such as those in the Marmara Sea. For that reason, we aim to estimate the moment magnitude (M_wp) using a set of strong-motion recordings, as broadband seismograms are generally clipped at short epicentral distances. The goal is to assess the feasibility of integrating M_wp estimation into regional early-warning frameworks for the Marmara Seismic Zone.

2. Data and Method

We used the 23 April 2025, earthquake, with a magnitude of 6.3 (GCMT-Mw), recorded by the Türkiye Strong Motion Network. The distance between the epicentre and the seismic recorders ranged from 25 to 100 km. Figure 2 shows the event (with focal nodal) and the seismic stations used in this study, and the hypocentre parameters are listed in

Table 1. Focal parameters of the earthquake used in this study.

Institute	Date	O.T. (UTC)	Lat (°)	Lon (°)	Depth (km)	ML	Mw	N.Sta	MwpAcc
KOERI	23 April 2025	09:49:10	40.83	28.23	12.1	6.2	6.2	61	6.5
AFAD		09:49:10	40.86	28.24	6.92		6.2
EMSC		09:49:11.9	40.83	28.22	15		6.2
GCMT		09:49:14.8	40.83	28.30	12		6.3

.

The P-wave moment magnitude technique uses the initial part of the P-wave arrivals at regional or teleseismic epicentral distances to calculate the M_wp. A key issue is selecting the time window from the onset of the P-wave to determine the magnitude as accurately as possible. Ref. [4] defined the P-wave moment magnitude based on a vertical far-field P-wave displacement seismogram. The technique is based on the observed seismic moment from the P-wave portion of the vertical displacement seismograms, u_z.

$$M_0={\displaystyle \frac{4\pi \rho {\alpha }^{3}r}{F^p}}\left|\mathrm{m}\mathrm{a}\mathrm{x}(\int u_z\left(x_r,t\right)dt)\right|$$

(1)

where ρ and α are the average density (3500 kg/m³) and the P-wave velocity along the propagation path, respectively, and r and Fp are the epicentral distance (between the earthquake epicentre and the recording station), calculated from the event and station coordinates, and the radiation pattern coefficients, respectively. The radiation pattern (Fp) is assumed to be unity for all stations in this study. Although this simplification neglects source-radiation and azimuthal effects, it is commonly adopted in rapid M_wp estimation studies, and its influence is mitigated by averaging results from a dense network of stations distributed around the source. Equation (1) requires source-to-station distance information; the method is intended for rapid, network-based magnitude estimation when an initial earthquake location is available. In operational earthquake early-warning systems, equivalent distance estimates would be obtained from real-time source-location algorithms using observations from multiple seismic stations. The seismic moment is calculated from the maximum amplitude within the selected time window. The moment magnitude is calculated from the seismic moment (M₀) using the formula proposed by [21] (Equation (2)), which is identical to the IASPEI (2005) standard formula:

$$M_w={\displaystyle \frac{\left(log{M}_0-9.1\right)}{1.5}}$$

(2)

where M₀ is in Nm [21,22]. In the current study, 60 accelerometer records were used to calculate M_wp for the event, and their arithmetic mean was taken. As discussed above, the primary issue in these calculations is the selection of the P-wave train time window, which directly affects the M_wp value. We selected 3 s after P-wave onset, based on trials of 2, 3, and 5 s, and we varied the time before P-wave arrival (

Table 2. Searching results for the different time frames to calculate the Mwp. a_5 means 5 s before the P-wave onset, and +2 means 2 s after it.

Window Length
Station	a_5 + 2	a_5 + 3	a_5 + 5	a_10 + 2	a_10 + 3	a_10 + 5
1011	6.8	6.91	7.13	6.99	6.31	7.07
3430	6.68	6.97	6.83	6.86	6.34	6.88
KURT	6.93	6.69	6.98	6.88	6.36	6.95

). However, the 2 s window occasionally produced lower values, suggesting incomplete capture of the initial P-wave energy, whereas the 5 s window increases processing latency and may include additional waveform complexity beyond the earliest P-wave portion. The 3 s window was therefore selected as a pragmatic balance between rapid availability and stable magnitude estimation. Among the tested alternatives, the 3 s post-onset window provided a practical compromise between rapid processing and magnitude stability. Although the differences among the tested windows were generally small, the 3 s window was adopted for subsequent analyses because it captures sufficient P-wave energy while maintaining minimal processing delay for early-warning applications.

Another important parameter in these applications is the P-wave velocity. Generally, a fixed P-wave velocity (e.g., 7.9 km/s) is used in the calculations. Kanjo et al. [7] used a distance-dependent P-wave velocity ($\alpha =0.16\Delta +7.9$). In this retrospective case study, apparent P-wave velocity was calculated from the known epicentral distance and observed P-wave arrival time at each station. This approach was adopted to evaluate the influence of station-specific propagation velocities on Mwp estimation. In a fully operational earthquake early-warning system, source location and travel paths would need to be estimated in real time from incoming phase picks and network-based location solutions. Therefore, the velocity estimation procedure used here should be considered a post-event validation approach rather than a real-time operational implementation.

The calculation process is summarised below (Figure 3).

i.

Selection of P-wave onset;

ii.

Estimation of P-wave time window length;

iii.

Cutting of the seismogram from 10 s before the P-wave arrival and along the time window;

iv.

Removal of trend and mean from the seismograms;

v.

Calculation of seismic moment;

vi.

Calculation of moment magnitude.

The above process is applied to each vertical seismogram recorded at each seismic station. Afterwards, the mean M_wp is calculated for the case using all available M_wp values estimated at each seismic station.

3. Results and Discussion

The epicentral distance ranges from 28 to 100 km. The P-wave moment magnitude (M_wp) was calculated from displacement seismograms derived from strong-motion acceleration seismograms. The calculated M_wp ranges from 5.9 to 7.0 and is affected by the earthquake’s directivity and the soil structure beneath the seismic stations (

Table 3. Used seismic station information and calculated P-wave moment magnitude for each station with measured PGA values.

Code	Long	Lat	PGA_UD	PGA_NS	PGA_EW	Dist (km)	Mwp	Vs30 (m/s)
	28.25964	41.08492	34.4933533	72.6685445	106.363096	28.41	6.4	230
5906	27.93164	40.97338	29.0576036	107.381914	68.3362873	29.75	6.8	224
3428	28.729595	40.984553	64.2440645	99.2674065	81.1246449	45.45	6.9	318
3415	28.75848	41.02729	71.5936873	210.197863	138.987994	49.61	7	283
5907	27.77633	41.1418	12.6569561	18.4012653	21.9986815	51.54	6.6	313
1659	28.391528	40.375063	4.66680028	8.02178124	9.23617871	52.33	5.9
5917	28.005353	41.2706	12.3302602	24.9634407	25.7345467	52.44	6.3	342
3416	28.83635	40.97466	33.3360928	37.949641	27.0630349	53.62	6.5	420
3431	28.71567	41.186225	56.035542	100.67463	160.574835	56.87	6.9
5908	27.54794	40.98205	7.4123659	11.2131628	11.3252213	59.89	6.3	538
3434	28.8205	41.140337	63.7181514	158.758624	99.2805704	60.47	7
7706	28.82662	40.51305	9.3261031	34.3620064	26.6193344	61.51	6.7	277
1011	27.86104	40.33601	7.42380375	14.5143805	14.5464435	63.12	6.3	330
5910	27.48608	40.98109	6.49765651	20.9879954	27.1726318	64.88	6.3	514
7716	28.944615	40.608968	24.1979265	45.4240724	38.5143607	65.15	6.8
7714	28.88833	40.52275	7.25581451	25.0270066	22.9764075	65.28	6.7	640
3411	28.97605	41.01187	20.3491826	31.4222069	44.2021661	66	6.6	323
3413	28.94818	41.09433	11.6312177	26.5870707	19.0772253	67.2	6.7	452
1633	28.36262	40.21397	9.92676027	12.4124083	33.0692209	69.32	6.2	375
3407	29.00951	41.0582	10.0938066	21.0294051	28.7085185	70.36	6.7	596
3426	29.069695	40.967905	11.1249721	24.6630787	25.6428156	72.39	6.7	667
5911	27.49156	41.17413	10.4467353	21.3539322	34.6770904	72.94	7	397
5916	27.917908	41.44368	9.22783737	27.6773327	29.2932054	73.01	6.1	332
3427	29.067068	41.00755	17.0516308	19.4441841	21.7261012	73.21	6.4	761
7715	28.97072	40.46296	20.1497987	53.8567627	46.6363284	74.74	6.6	169
7707	29.0788	40.6381	15.8389066	38.7709706	42.054343	74.8	6.6	312
3425	29.027277	41.134497	14.0143498	32.7932393	48.4303578	75.13	6.8	526
3438	29.115607	40.979632	18.8429984	28.1341236	29.0973362	76.44	6.6
3405	29.15668	40.91111	8.05427237	17.1115418	21.7142458	78.63	6.6	979
1618	28.92815	40.35095	8.34013089	20.7504572	19.1208078	79.51	6.6	314
3436	29.098545	41.125952	25.0383261	51.9335538	49.2761849	80.14	6.4
3430	29.165163	41.015572	16.4430188	60.6995616	46.7000003	81.42	6.3	379
1637	29.094628	40.47632	23.9347937	30.8225001	34.5466186	83	6.5	181
5914	27.24526	40.66825	7.15301935	13.0804855	17.4121747	85.082	6.4	272
1638	29.0333	40.361147	5.16112408	10.8868823	7.90484902	85.64	6.4	338
3437	29.23047	41.000652	13.6034173	16.1848669	14.8474913	86.38	6.6
1653	29.098645	40.415365	14.8346673	28.4872907	20.677103	86.72	6.8	160
KURT	29.2092	40.5905	6.65598384	11.9459567	12.9904941	86.914	6.3	916
3417	29.25627	40.95471	6.13140803	12.6277333	11.0757539	87.58	6.1	1146
1024	27.974028	40.047475	21.1488647	51.7864585	60.6835681	89.56	6.5	228
1648	28.980285	40.255225	6.99128322	18.9105535	19.1874535	90.07	6.8	311
1658	29.16722	40.424085	16.8276826	45.9271534	31.6302661	91.21	6.5	233
3905	27.75977	41.57237	16.8759891	35.1208384	32.1161705	91.39	6.4	316
7710	29.2668	40.58997	9.07476855	15.3016892	22.3170148	91.57	6.4	358
1630	29.12207	40.36298	14.0784924	47.3780198	32.4125646	91.604	6.5	301
1726	27.300073	40.401827	7.82235513	10.767909	12.9047222	91.94	6.6	870
1621	28.97558	40.22686	5.04792825	9.37650137	13.5490301	92.069	6.5	396
1642	29.179435	40.410343	4.87930205	11.28755	18.4886191	92.87	6.3	366
1646	28.968037	40.209078	5.1874846	8.81747331	13.1998013	93.08	6.3	386
7711	29.32709	40.65942	12.3253538	18.2136225	18.7725525	94.57	6.5	199
7709	29.30603	40.56416	18.772491	27.9985276	35.8004394	95.61	6.6	382
3902	27.32483	41.35705	5.33850967	15.9538374	11.4652839	95.96	6.8	317
1649	29.095913	40.265783	7.66942652	16.8331594	15.7237902	96.46	6.6	275
5904	27.12256	40.61485	15.1890757	29.4019826	20.8566281	96.56	6.4	225
1635	29.258702	40.449648	10.8123794	30.5562217	42.0743759	96.72	6.5	572
4132	29.3813	40.83046	5.27734617	13.5941493	18.3984977	97.11	6.6	570
1631	29.2993	40.49411	9.69038318	53.0158081	31.1990741	97.81	6.7	410
1627	29.07518	40.22566	6.17649645	18.6750015	16.8541642	98.14	6.9	249
1650	29.05789	40.206197	5.61169521	15.235972	15.7321914	98.5932	6.7	236
1619	29.2907	40.42236	8.16719028	18.0425437	18.5012119	100.51	6.3	349

). Figure 4 and Figure 5 show how recorded acceleration varied with fault strike. The highest accelerations were observed at station 3431 (160.57 gal on the EW component) and at station 3415 (210 gal on the NS component); the stations are 56.87 km apart along the EW component and 49.61 km apart along the NS component, respectively. The maximum vertical acceleration was also observed at the 3415, 3428, 3434, 3434,3431, 3429, and 3416 seismic stations, which are located on the fault strike.

Figure 6 shows the change in magnitude difference between the calculated Mwp and M(GCMT) as a function of Vs₃₀ and epicentral distance. The magnitude deficiency, considering Vs₃₀, generally decreases above 400 m/s. A similar pattern was observed with epicentral distance, that is, between approximately 80 and 100 km. Another important point when focusing on early warning is processing time. The time required to obtain an Mwp estimate after P-wave identification ranges from approximately 8 to 20 s. These values represent the magnitude-estimation stage only and do not include additional delays associated with event detection, phase association, source-location estimation, or warning dissemination. (Figure 7). The scenario studies for a tsunami in the Marmara Sea show that the maximum time for wave arrival at the coast is 30 min.

Considering this study, the first 10 min after a major earthquake in the Marmara Sea are vital for early-warning and rapid-response activities. This study demonstrates rapid magnitude estimation using strong-motion recorders rather than broadband seismometers, which can be clipped, providing a reliable magnitude value in a very short time and offering substantial support for early-warning and rapid-response activities. The calculated M_wp is highest at stations 3415, 3428, 3434, and 3431, ranging from 6.9 to 7.1, whereas stations 3429 and 3416 have M_wp magnitudes of 6.3 and 6.4, respectively.

Seismic stations 3429 and 3416 are located near the nodal planes (Figure 8). Although the M_wp varies widely, the mean value is sufficient for decision-makers to determine whether the earthquake is destructive. It would also be useful to estimate and model the expected arrival time of tsunami waves at the coastline. This study focuses on estimating rapid-moment magnitude from P-wave recorded by strong-motion recorders, although broadband velocity records have been the primary focus in the literature. Broadband seismograms generally clip near the epicentre during a large earthquake. In the event of a destructive earthquake, strong-motion records are useful for estimating the rapid magnitude for early-warning and rapid-response activities. This study also uses the variable P-wave velocity, calculated from arrival time and epicentral distance, as the apparent velocity, which influences the calculations and is more accurate than using a fixed velocity. The arithmetic mean must be used in these calculations because some records yield high magnitudes due to local site effects and directivity. The findings underscore the importance of rapid magnitude-estimation frameworks for improving the accuracy and reliability of early warnings.

The findings also demonstrate the advantages of strong-motion sensors for rapid magnitude estimation. Broadband seismometers, although highly sensitive, may clip during large near-field earthquakes, especially in densely instrumented urban regions. Strong-motion accelerometers are less affected by clipping and can therefore provide more stable amplitude measurements immediately after rupture initiation. This characteristic is particularly important for the Marmara region, where future large earthquakes are expected to strongly affect densely populated urban and industrial zones, including Istanbul and surrounding metropolitan areas.

Practical implementation of the proposed M_wp methodology requires an estimate of the source-to-station distance (r). In the present study, this distance was determined retrospectively using the known earthquake location. In an operational environment, however, source location would be obtained from automatic event detection, phase picking, phase association, and rapid location algorithms operating on data from multiple seismic stations. Therefore, the proposed methodology should be considered a component of a network-based rapid earthquake-characterisation framework rather than a stand-alone single-station solution. Similar approaches have been developed for automated earthquake detection and localisation in continuous seismic records (e.g., [23]), where source parameters are estimated in near-real time and subsequently used for rapid event characterisation.

Overall, the results confirm that P-wave analysis of strong-motion records provides a robust framework for rapid earthquake magnitude estimation in tectonically active regions. Although individual stations may exhibit substantial variability due to directivity and local site effects, averaging multiple station solutions yields stable event magnitudes that are consistent with independent global moment tensor solutions. The proposed workflow therefore represents a practical, operationally applicable approach for regional earthquake early-warning systems in Türkiye and similar seismic environments worldwide.

4. Conclusions

This study evaluated the feasibility of rapid P-wave moment magnitude (M_wp) estimation using strong-motion records from the 23 April 2025 Marmara Sea earthquake. The principal findings are summarised as follows:

• Rapid magnitude estimation was successfully achieved using only the initial portion of the P-wave signal. Reliable M_wp values were obtained within approximately 8–20 s after P-wave arrival, demonstrating the method’s potential for use in earthquake early-warning and rapid-response systems.
• The average estimated magnitude was consistent with independent reference solutions. The mean Mwp from 61 strong-motion stations was approximately 6.5, consistent with the GCMT moment magnitude (Mw = 6.3).
• Station-based M_wp estimates exhibited significant variability. Calculated magnitudes ranged from approximately 5.9 to 7.0, reflecting the influence of rupture directivity, source radiation effects, propagation-path characteristics, and local site conditions.
• Multi-station averaging significantly improved magnitude stability. Although individual stations produced overestimated and underestimated values, the arithmetic mean of a dense network yielded a robust event magnitude suitable for rapid decision-making.
• The selected 3 s P-wave window provided a practical balance between processing speed and magnitude stability. While differences among tested windows were relatively small, the adopted window length was considered suitable for operational rapid magnitude estimation.
• Strong-motion sensors represent a valuable resource for rapid earthquake characterisation. Unlike broadband instruments, strong-motion accelerometers are less susceptible to clipping during large earthquakes and can provide reliable amplitude measurements in near-source regions.

Overall, the results demonstrate that strong-motion-based P-wave analysis can provide rapid and reliable earthquake magnitude estimates in the Marmara region and may contribute to the development of future earthquake early-warning systems in Türkiye. Future studies should evaluate additional earthquakes with different magnitudes, focal mechanisms, and source depths to further assess the robustness and operational applicability of the proposed methodology.