Can we quickly calculate reliable earthquake parameters from citizen testimonies?

Can we quickly calculate reliable earthquake parameters from citizen testimonies? | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Can we quickly calculate reliable earthquake parameters from citizen testimonies? Gianfranco Vannucci, Remy Bossu, Matthieu Landès, Paolo Gasperini This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7071982/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Bulletin of Earthquake Engineering → Version 1 posted 5 You are reading this latest preprint version Abstract In a recent work, we tested the ability to compute macroseismic parameters (location and magnitude) using citizen testimonies collected by the European-Mediterranean Seismological Centre (EMSC). Each intensity estimated by individual non-professional users of the LastQuake smartphone application is indicated as an individual data point (IDP). Each IDP is archived by EMSC with a time stamp, allowing the calculation of the time delay from the earthquake origin time. To use IDPs as classic intensities, i.e. macroseismic data points (MDPs), identifying damage at the scale of towns or cities, they must be grouped into spatial clusters, which are then processed by the BOXER code to locate and size earthquakes. A retrospective analysis on a dataset of more than 15000 events collected over the past 10 years shows that the procedure can provide reliable parameters and that the results depend on the geographical area and improve over time and as the number of available IDPs/MDPs increases. The key question is whether early IDPs/MDPs can quickly provide reliable parameters (location and magnitude) for users and stakeholders (e.g. the civil protection agencies). Using clustering methods that statistically provide, on average, the best agreement with instrumental data, we tested some predefined time intervals within which to group the available IDPs into MDPs. We then applied the BOXER code to these MDPs, evaluating the agreement with the final instrumental parameters. Results confirm that reliability increases with the number and distribution of MDPs, strictly dependent on the number and distribution of available IDPs. This retrospective analysis demonstrates the effectiveness of the approach and its potential to quickly provide parameters for future real-time applications. The method may offer a reliable and rapid tool to support emergency response, improving as more IDPs/MDPs are collected. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction In recent years, an area of so-called crowdsourcing science, based on information provided directly by citizens, has developed and become increasingly important. Some agencies collect this information in near real-rime, such as “Did you feel it?” (DYFI; Wald et al., 1999 , 2011 , Dewey et al., 2000 ), of the U.S. Geological Survey (USGS), “Hai sentito il terremoto?” (HSIT; Tosi et al., 2015 ) of the Istituto Nazionale di Geofisica e Vulcanologia (INGV), New Zealand GeoNet questionnaires (GeoNet, Goded et al., 2018 ), LastQuake by the Euro-Mediterranean Seismological Centre (EMSC) (Bossu et al., 2015 , 2017 , 2018 ). Vannucci et al. ( 2024 ) verified the possibility of deriving reliable macroseismic parameters from individual intensities (intensity data points, IDPs) provided according to EMS scale (Grunthal et al., 1998) by each eyewitness citizen, collected and made available by the LastQuake system. This retrospective analysis allowed to define and overcome some of the limitations in the data, such as the presence of possible outliers in both geographical location and intensity value. In some cases, IDPs may be geographically misplaced due to factors such as a VPN connection or the poor location provided by users (e.g., only the city rather than a precise street address) when submitting reports via the LastQuake system on a PC rather than a smartphone. EMSC also provides, in addition to the original IDP intensity, a second value of IDP intensity modified by a formula, derived by Bossu et al. ( 2015 , 2017 ), to best reproduce DYFI intensities. However, IDPs cannot be directly treated as classical macroseismic intensities, which represent the earthquake effects at the scale of a city or town and are, therefore, representative of effects over an extended area rather than the punctual observation of an individual user. Hence, a spatial clustering procedure was applied, requiring a minimum number (usually three) of IDPs per cluster (Vannucci et al., 2024 ). This approach also eliminates geographical outliers and mitigates intensity outliers (typically overestimations) through central tendency estimators such as the mean, the median, or the trimmed mean (i.e. average of intensities calculated after excluding a percentage of the tails of the distribution of IDPs). Consequently, we assign a single equivalent intensity value to each cluster by a straightforward, rigorous, and reproducible procedure, making it comparable to traditional macroseismic data points (MDPs). To validate their approach and identify the most effective combinations for grouping IDPs into MDPs, Vannucci et al. ( 2024 ) carried out a retrospective analysis on the complete distribution of IDPs (both original and modified) for each event in the EMSC catalogue. They combined data clustering techniques and central tendency estimators, using the BOXER code (Gasperini et al., 2010 ) to estimate earthquake parameters (location and size). The best combination for minimising statistical deviations from the instrumental data was found to be the DBSCAN spatial clustering approach (DB, Ester et al, 1996 ) with a 2-km radius (DB2), a 20% trimmed mean (mn20) for distance and a 10% trimmed mean (mn10) for magnitude. Overall, most combinations that minimise differences with instrumental benchmarks are associated with the DB method, with only small variations between different statistical estimators. In last years, it has been possible to produce quantitative outputs with high reliability as instrumental methods, using macroseismic intensities (Vannucci et al., 2019 ), also provided by citizens (Vannucci et al., 2024 ). In this work, we aim to verify the time-dependent reliability of macroseismic parameters (location and magnitude) of earthquakes using the BOXER code after grouping the original IDPs into MDPs. Method The IDPs distributed by EMSC for each event include the time delay in seconds ( https://seismicportal.eu/testimonies-ws/ ) from the earthquake origin time T0, at which users sent their reports. Therefore, for this retrospective analysis, aimed to quantify the time-dependent reliability of macroseismic locations and sizes, we clustered available IDPs over predefined time intervals from T0 for earthquakes on a global scale from 2012 to 31 May 2024. From this date onward, EMSC has been testing a system that continuously processes location and macroseismic magnitude in real time using citizen IDP. However, this test is outside the scope of this work and will be addressed in future research. Grouping distance and difference of magnitude within specific time intervals represents a kind of stacking approach (e.g. Gulia et al., 2018) to enhance the signal and enable comparisons between any different combinations of grouping methods, central tendency estimators and BOXER methods. We selected successive time intervals relative to T0: every 30 seconds up to 10 minutes, including an additional 45s interval (30s, 45s, 1m, 1m30s, 2m …), every minute up to 15 minutes from T0 (10m, 11m…) and finally 20, 30 minutes and one day. We use a minimum of three IDPs per cluster to obtain the MDPs, for each earthquake at each time interval, which are then processed with the BOXER code (Gasperini et al., 1999 , 2010 ) to assess the macroseismic location and magnitude. Based on the findings of Vannucci et al. ( 2024 ), the DBSCAN method is the most effective and fast and provides the most reliable results, independently of the statistical estimator. Moreover, this method offers the key advantage of grouping close IDPs without requiring predefined areal boundaries, such as polylines or equivalent radii simulating the area of a locality, or predefined geometric grids that may distort the true urban extent. Therefore, it is ideal for near real-time applications. As statistical estimators, we considered both mn20 for distance and mn10 for magnitude, according to Vannucci et al. ( 2024 ). The values obtained are quite similar for all the trimmed means, so to streamline the process for possible near real-time applications, we selected mn20 as the default estimator to calculate both epicentre and macroseismic magnitude. Hence, we systematically applied the combination DB2-mn20 to the selected time intervals. The analysed dataset, up to 31 May 2024, consists of 17243 earthquakes with sufficient IDPs to generate at least one MDP. On these time intervals, we apply the BOXER code to locate the earthquakes using two methods: i) method 0 (BOXER-0), computing the barycentre of the sites with the most severe effects (analogous to a centroid of the area of deformation or maximum slip) and ii) method 1 (BOXER-1), computing the centre of the entire intensity distribution by minimising the squared residuals of the intensity prediction equation (IPE) by Pasolini et al. ( 2008 ). A key difference between these two methods is that method 0 can locate even with just one MDP, while method 1 requires a minimum of 5 MDPs. The algorithm for magnitude estimation requires a minimum of 4 MDPs even for events located with method 0. Therefore, as the number of IDPs increases over time, leading to a higher number of MDPs, the first parameter to be estimated is the epicentre by BOXER-0, followed by the magnitude, and finally the epicentre by BOXER-1. Results and discussions The results are illustrated by plots (Figs. 1 – 2 ) showing, at different delays from origin time, the distribution of distances and of magnitude difference between macroseismic and instrumental parameters. The total number of earthquakes for which macroseismic parameters can be calculated obviously increases over time as more IDPs become available. Using the IDPs collected within one day (1d) from T0, 13724 and 3851 macroseismic epicentres (Fig. 1 ) as well as 3851 and 3780 macroseismic magnitudes (Fig. 2 and Fig. S1 of the supplementary material) can be calculated by BOXER-0 and BOXER-1, respectively. It is important to note that the number of earthquakes determined at the 1d time interval is lower than the total number of earthquakes (17423) with at least one MDP, as in many cases, the IDPs were supplied by citizens more than 24 hours after T0. However, for the purposes of this work, we only consider the IDPs supplied within the 1d interval. For both BOXER methods, we analysed the number of earthquakes, whose macroseismic epicentres or magnitudes can be estimated at each time interval from T0, as a function of different ranges of distance (Fig. 1 ) and difference in magnitude (Fig. 2 and Fig. S1 of the supplementary material) relative to instrumental benchmarks. In these Figures, we also plotted central tendency estimators (mean, median, trimmed means) to quantify the agreement between macroseismic and instrumental parameters. Finally, to facilitate evaluations and comparisons (e.g. between the two BOXER methods), we calculated the relative percentage of earthquakes within each distance or magnitude range (bin) with respect to the total number of earthquakes in the same time interval, The number of earthquakes for which both macroseismic and instrumental epicentres can be compared in terms of distance increases over time. Using BOXER-0, we have 1206 earthquakes (about 9% of the events located at 1d time interval) are located within 1 minute, 3429 (20%) within 2 minutes, 5049 (37%) within 2 minutes (Fig. 1 ). The macroseismic epicentre can be estimated for 6186 earthquakes (45% of those localised at 1d time interval) within 4 minutes, a timeframe that, for example in Italy, is conventionally understood as the limit within which the Istituto Nazionale di Geofisica e Vulcanologia (INGV) must provide instrumental parameters of ongoing earthquakes to the Civil Protection Department. The number of earthquakes in agreement with instrumental data and their percentage with respect to earthquakes localised after 1day increases over time (e.g. green bars in Fig, 1). In terms of relative percentage, the agreement between macroseismic and instrumental epicentres within various distance bins is noteworthy, with high percentages within 10 km and 20 km (Fig. 1 ). The relative percentage of earthquakes whose distance is consistent with instrumental data decreases in the first 4 minutes from T0, then almost stabilises at ~ 25% within 10 km and ~ 50% within 20 km. Up to the 4-minute interval, the average distance of earthquakes in each time interval tends to increase, then stabilises around values of 30–40 km for the median and trimmed means (Fig. 1 ). The mean shows the same trend but with higher values ( ~ + 10 km), because, unlike the other estimators, it is more affected by outliers (earthquakes located at greater distances from the instrumental reference epicentre) despite their relatively lower occurrence. The results with BOXER-1 show a lower number of located earthquakes compared to BOXER-0 due to the minimum number of MDPs (5) required for the location, and a slightly better agreement than that with BOXER-0, in particular for time intervals < 5 minutes (Fig. 1 ). This can be directly observed by comparing the results obtained with BOXER-0 on the same earthquakes located by BOXER-1 (Fig. 1 ). The plots between distance and magnitude also as a function of the number of MDPs are similar to those observed with BOXER-0 (Figs S2 and S3 in the supplementary material). Although the total number of located earthquakes is lower in BOXER-1 compared to BOXER-0, the percentage of successfully located events is higher in BOXER-1, making it the preferable location method. For both BOXER methods, locations estimated within a few minutes from T0 are highly reliable. Concerning the difference between macroseismic and instrumental magnitudes, a remarkable symmetry can be observed in the histogram of events between positive and negative differences over time (Fig. 2 , for BOXER-0 and Fig. S1 in the supplementary material for BOXER-1). Positive values indicate an overestimation of the macroseismic magnitude with respect to the instrumental one, while negative values indicate the opposite (represented by red and blue colours, respectively, in Fig. 2 ). On average, the difference is close to 0 with a standard deviation of about 0.7 magnitude unit (m.u.). However, within a minute from T0, there is an overestimation of the macroseismic magnitude, although due to a small number of events. Even when considering the relative percentage of events in the different ranges, the symmetry between positive and negative values remains evident (Fig. 2 ). Both BOXER methods (Fig. 2 for BOXER-0 and Fig. S1 in the supplementary material for BOXER-1) provide very similar results, as are based on the same calculation model and use the same number of MDPs. The differences in magnitude values arise from variations in the felt area within the isoseismal radius of the intensity classes, centred on the estimated epicentre, which may slightly differ between the two BOXER methods, slightly affecting the computed magnitudes. When considering the difference in absolute value, we observe that, for each time interval (except the first 30 seconds, where only two earthquakes are included), about 20% of the earthquakes occur within a range of 0.1 m.u., 30% within 0.2 m.u., 40% within 0.3 m.u. and 50% within 0.5 m.u. for both BOXER methods (Fig. 2 and Fig. S1 of the supplementary material). The overall average magnitude difference rises to 0.5 m.u. (± 0.5). Macroseismic magnitudes tend to be slightly overestimated with respect to instrumental ones at lower values and slightly underestimated at higher values (Fig. 3 and Fig. S4 in the Supplementary material). In particular, as the number of MDPs increases, the magnitude difference decreases (Fig. 3 ), i.e. earthquakes with a high number of MDPs tend to minimise the magnitude differences (Figs. S4-S7 of the Supplementary material). It is noteworthy the absence of macroseismic magnitudes lower than 2.5. This is a well-known feature already noted in other studies (e.g. Rovida et al., 2020 ). It might depend on the intensity-magnitude regression models, which cannot take into account variables (e.g. hypocentral depth, site effects) that do not vary the magnitude but influence the observed effects (i.e. the calculated intensity). To check the agreement between macroseismic and instrumental magnitudes, we computed linear regressions between instrumental and macroseismic magnitudes in each time interval (Fig. 3 ), using the General Orthogonal Regression (GOR) method (Fuller, 1987 , Castellaro et al., 2006 , Gasperini et al., 2012, Lolli et al., 2014 , 2020 ), a linear regression model assuming a constant ratio between the variances of the y and x variables. A slope coefficient close to 1 would indicate a nice agreement between instrumental and macroseismic magnitudes. In our case, the GOR shows a partial agreement between instrumental and macroseismic magnitudes with regression slopes of ∼0.6–0.66 for BOXER-0 (Fig. 3 ) and slightly higher values for BOXER-1 (Fig. S8, in the Supplementary Material). As the number of MDPs increases, even the slope increases (0.7–0.8), indicating that the agreement between macroseismic and instrumental parameters approaches a 1:1 relationship (Fig. 3 ). This highlights the importance of the number of MDPs for the reliability of the parameters. Macroseismic magnitudes slightly overestimate instrumental ones below ~ 4.5 and underestimate them above that value. Several hypotheses can be proposed to explain the observed findings, particularly the reduction with the distance of the relative percentage agreement with instrumental data (Fig. 1 ). In densely populated areas, the first IDPs (and thus the MDPs) should be close to the instrumental one. Over time, as the felt area increases due to geometrical spreading, other MDPs farther away may change the geographical distribution and the macroseismic location. In general, we observed a slight direct correlation between the increase in distance and the increase in magnitude, in particular for distances below 50 km. Distances greater than 50 km, on the other hand, are mainly observed for M3 + magnitude earthquakes. As both distance and magnitude increase, earthquakes decrease in number and become more scattered (Figs. S9 and S10 in the Supplementary material). Moreover, most earthquakes with a high number of MDPs (> 40) are located within 50 km of the instrumental epicentre, regardless of the earthquake magnitude. An increase in the number of MDPs generally improves the agreement between macroseismic and instrumental data, as observed in Vannucci et al. ( 2024 ). Another explanation is related to the quality of the available information. Each bin in the various time intervals does not necessarily contain the same earthquakes: events within a certain time interval, may not have sufficient MDPs for location and size, or, if MDPs allow the parameter assessment, the increase of collected IDPs over time can lead changes in number or MDPs, changes in their geographical distribution an intensity variation, so changing previous macroseismic parameters of an event. Moreover, new earthquakes could be added, so quickly varying the relative statistics over time. By comparing the parameters in a given time interval with those in the previous one, for the same earthquakes, variations may be absent or may indicate improvement (positive values) or worsening (negative values) of the agreement with the instrumental data (Fig. 4 , and Fig. S11 in the Supplementary material). In most of the cases there are not changes in earthquake parameters over time or the variations are minimal (0–10 km in distance or 0-0.2 in absolute magnitude difference for most of earthquakes (Fig. 4 ) Over time, new earthquakes are progressively included in the dataset (Fig. 4 ) and this results in an overall decrease in relative percentage agreement observed for the distance between macroseismic and instrumental epicentres (Fig. 1 ). This trend may be partly attributed to poorly located events, such as those occurring offshore or in sparsely populated areas, and reflects the growing inclusion of earthquakes from outside Europe and North America, i.e. Asia-Oceania, South America, and Africa. Indeed, the relative proportion of such events within the dataset increases over time (see Fig. S12 in the Supplementary material), and the agreement with instrumental data tends to be lower (Vannucci et al., 2024 ). Moreover, some agreement worsening (i.e. negative variations) in earthquake parameters may be due to delayed notifications, even from areas outside the true felt area, which can trigger some unreliable reports, i.e. when individuals retrospectively believe they experienced shaking. It is worth noting that the procedure allows a rapid estimation of macroseismic parameters, often preceding the availability of corresponding instrumental data (Fig. 5 ). The comparison refers to the time of “first” availability of macroseismic (Tm) and instrumental (Ti) data, using records since 2016 only, the year in which EMSC began archiving Ti in the IDP database. Note that Ti, in most cases, refers to the first determination of instrumental parameters rather than the final one, due to possible later revisions (numerical changes or replacement by another agency). Therefore, this analysis considers 11723 and 3595 events for distance, and 3676 and 3364 for magnitude difference, using BOXER versions 0 and 1, respectively (Figs. 5 and S13 of the Supplementary material, respectively). The macroseismic parameters can be classified as simultaneous (Tm = Ti), earlier (Tm Ti) based on the timing vs instrumental parameters. The elaborations show that macroseismic parameters often become available earlier than instrumental ones within the first few minutes after T0. In particular, using BOXER-0, 7532 macroseismic epicentres (64% of the total sample) are located before the corresponding instrumental ones, of which 5145 (∼44% of the total) within the first 4 minutes (Figs. 5 and S13 and Table S13 of the Supplementary material). Similarly, 1586 magnitudes (43% of the total) are assessed before the corresponding instrumental ones, of which 1000 (∼27% of the total) within the first 4 minutes from T0 (Table S14 of the Supplementary material). These earlier locations are probably mostly related to low-magnitude earthquakes (lower magnitudes show shorter delays, Fig. S2 of the supplementary material) and are certainly influenced by population density and LastQuake system dissemination (Vannucci et al., 2024 ). The variability and validity of the calculated parameters may also depend on the dimensional scale on which the IDPs are aggregated. This could lead to the creation of an MDP that does not accurately reflect real urban areas: larger radii collect sparse reports but may blur detail in densely populated area while smaller radii could offer finer localisation, increasing the geographic detail in the representation of felt effects. Thus, in addition to the 2 km radius (DB2), we tested the EMSC database grouping the IDPs by DBSCAN method, with a radius of 1 km (DB1) and 0.5 km (DB0.5) We still use the estimator mn20 to compute MDPs. Trends are similar across radii, though smaller radii reduce the total number of located events (11999 for DB1; 10190 for DB0.5) but yield slightly higher percentages of accurate locations, especially within short intervals from T0 (Figs S14 and S15 for distance and difference of magnitude in the supplementary material). The results show slightly higher percentages of agreement between macroseismic and instrumental epicentres (i.e. lower distance) when using BOXER-1, particularly with DB1 and DB0.5 (Figs S14 and S15, supplementary material). With DB1, 3990 earthquakes are located after one day (i.e. slightly more than with DB2), and there is a notable increase in events localised within shorter intervals from T0, especially within the first 3 minutes, enhancing agreement with instrumental data. Although the total number of earthquakes decreases with DB0.5 (3532), the count remains higher in the short time intervals, and the percentage of earthquakes within 20 km of the instrumental epicentre consistently exceeds 60%. This suggests that reducing the DB radius improves the match with instrumental locations, for short time intervals, with BOXER-1 statistically preferable to BOXER-0. Basically, for densely populated areas, DB1 and DB0.5 are better at pinpointing the epicentre, especially in the shorter time intervals, while DB2 lets you estimate parameters for more events by using a bigger IDP clustering area. Regarding magnitude differences, for both DB1 and DB0.5, the BOXER methods show comparable trends. The DB1 approach consistently yields more events than DB2, while DB0.5 sees fewer after 10 minutes. The agreement with the instrumental data is comparable to DB2, with a difference, on average, close to 0. However, there is a slight overestimation of the macroseismic magnitude during the first three minutes (Figs S16-S19 of the Supplementary material) with respect to DB2. This trend is probably related to the calibration method adopted, as previously discussed (Fig. 3 ). We can illustrate the potential parameter changes over time with a practical example by presenting three earthquakes of varying magnitudes and geographical locations (Fig. 6 and numerical details in Table U or supplementary material), and analysing how their distance and magnitude difference vary for both BOXER methods. The earthquakes are: i) the 2022, May 12th (21:12 UTC, Mw 3.7) event, located South of Florence, Italy (incidentally, it occurred during the social dinner of the European RISE Project meeting); ii) the 2023, Feb. 6th (1:17 UTC, Mw 7.8) event, located in Kahramanmaras-Pazarcik area, Turchey; iii) the 2024, April, 5th (14:23 UTC, Mw 4.8) event, located in New Jersey, USA. For each earthquake, Fig. 6 presents two graphs, showing temporal variations in distance (Di) and magnitude difference (dM), with respect to the instrumental benchmark of EMSC. Full-time evolutions of IDPs, MDPs, and macroseismic epicentres are provided as videos in the supplementary material. Parameter trends follow a consistent pattern: BOXER-0 gives initial locations; BOXER-1 activates once five MDPs are reached. Magnitude is computed simultaneously by both BOXER methods (at least four MDPs). While differences with instrumental benchmarks tend to decrease, the minimum values do not always coincide with the latest time intervals, where the highest number of MDPs is typically available, indicating non-monotonic behaviour. Notably, MDP numbers also fluctuate: new IDPs may aggregate previously distinct MDPs, reducing totals, as in the Florence case. There, a macroseismic epicentre (12.5 km from the benchmark) is calculated within 90 seconds from T0. The distance increases with time, then decreases after 14 minutes. The first macroseismic magnitude, 7 minutes after T0, overestimates the instrumental value by 0.7 m.u, and later the magnitude difference fluctuates between 0.2 and 1 m.u. Early localisation is due to initial IDPs (then MDPs) near the epicentre, which later changes, with distant IDPs expanding the felt area modifying MDPs (in general low in number) and location assessment. For the large-magnitude Turkey earthquake, BOXER-0 locates the epicentre 3.5 minutes after T0, about 500 km away, due to a“doughnut effect” i.e. the lack of reports and therefore intensities from the severely affected epicentral area (Bossu et al., 2018 , 2024 ). Increasing the number of MDPs over time, BOXER-1 provides a first macroseismic epicentre after 4 minutes from T0, improving localisation up to ~ 30 km within 20 minutes. However, the magnitude remains underestimated: initially − 3 m.u., rising to -1.2 after 6.5 minutes, but still − 0.8 after one day, due to the lack of high intensities in the epicentral area. The New Jersey event shows a similar pattern to the Turkey earthquake: BOXER-0 locates at a distance of 65 km after 2.5 minutes, improved by BOXER-1 after 5 minutes, then decreasing the distance up to 8 km after 14 minutes. Magnitude starts at -1.5 m.u., and the difference with respect to the instrumental value decreases significantly, reaching zero after 11 minutes. In the latter two examples, there is a clear inverse trend between the increase of the number of MDPs and the reduction in both distance (in particular with BOXER-1) and magnitude difference. Reducing the DB radius increases MDPs for the Florence and Turkey earthquakes, improving resolution and accuracy, especially for the lower-magnitude 2022 event (Table U in the supplementary material). The DB2 approach gives better results for larger earthquakes (Turkey and New Jersey), except for occasional time intervals. For the 2022 earthquake, which has a lower instrumental magnitude and thus a smaller felt area, the higher MDP resolution better constrain the macroseismic epicentre, in particular with BOXER-1. Magnitude differences tend to be smaller across radii, with better matches occurring when more MDPs are present. The results of the analysis suggest that several factors may concurrently affect the calculated parameters as the number of MDPs, azimuthal gap among MDPs, spatial distribution, magnitude, and potential edge effects, such as the doughnut effect. To assess whether the agreement between macroseismic parameters and benchmarks depends on specific factors, even over time, we analysed the datasets by applying conditional thresholds to define subsets of data. For example, we distinguished between earthquakes with instrumental epicentres located offshore or inland, or between events associated with a number of MDPs greater or lower than a threshold. The comparison was made, for both BOXER methods, for distance and magnitude difference, also in absolute value, with respect to the instrumental parameters, using a 10% trimmed mean and sampling the data in the various time intervals. We show the results at the global and European scale (Fig. 7 ). In general, the results are fairly stable over time intervals, without large variation, except for the data in the very first intervals from T0. The agreement (smaller distances and differences in magnitude between macroseismic data and instrumental benchmarks) increases with the number of MDPs, while earthquakes that occur on land are, on average, better located than earthquakes at sea. The data collected in recent years show very little improvement in the results with respect to the previous ones. The agreement increases by decreasing the azimuthal gap, although without significant changes. Conversely, the agreement gets worse significantly showing fluctuations over time when the gap exceeds 180 degrees, in particular for magnitude (Fig. 7 ). This condition is generally associated with localities bordering IDP-blind areas, such as sparsely populated regions, natural boundaries (e.g. coastlines, deserts, mountain ranges), or countries with restrictive policies on the use of smartphones or related applications (see Vannucci et al., 2024 for details). A general worsening of results, at the global scale and in both distance and magnitude, is observed for magnitude (Fig. 7 ). In particular, the agreement between macroseismic and instrumental parameters progressively worsens for M4 + events and with increasing magnitude. On average, the difference remains stable over time (Fig. 7 ), even if results for some earthquakes may improve over time (e.g. the events in Turkey and New Jersey, Fig. 6 ). This pattern is likely due to the “doughnut” effect in the epicentral area, where, after a strong event, citizen priorities are far from the fast reporting of effects. However, it should be noted that the same analyses carried out at the European scale, based on a dataset comprising more than 2/3 of the total global earthquakes (Fig. S12 of the Supplementary Material), show lower differences, particularly with respect to distance when using BOXER-0, with a ∼30% compared to the values obtained at the global scale (Fig. 7 ). When considering BOXER-1, the average deviation from the benchmark at European scale is reduced by more than 70% compared to BOXER-0 and by ∼50% compared to BOXER-1, both at global scale. An average worsening in agreement (epicentral distance > 50 km) is observed only for events of magnitudes M5/M5.5 + in contrast to the other subsets, which overall fall within the 20–40 km range (Fig. 7 ). This suggests that on a global scale, the largest differences come from macro-areas outside Europe. The only worsening observed at the European scale, compared to the global one, concerns magnitude differences for events with a gap > 180 degrees, although this may be attributable to the small number of events in this subset (a maximum of 21 events in the 1-day time interval). Finally, this analysis shows that the results of BOXER-1 are generally slightly better than the corresponding values of BOXER-0 (Fig. 7 ). Conclusions We used the individual intensities reported by citizen (IDPs) collected and made available by EMSC, to verify the reliability of the location and macroseismic magnitude calculated by applying a specific procedure that first groups IDPs into traditional intensities (MDPs) and then processes them using the BOXER code. We selected the DBSCAN clustering method with a 2 km radius to group IDPs and applied BOXER methods 0 and 1 to process MDPs. The procedure was applied within predefined time intervals after the origin time of global-scale earthquakes. Our results show that within the first four minutes, a significant percentage of earthquakes with macroseismic parameters are located at relatively short distances (e.g., 10–20 km), with an overall low average distance from instrumental epicentres. In most cases, these data predate the availability of instrumental data. Over time, this distance slightly increases with values, on average, of about 25–35 km. We also analysed the difference between macroseismic and instrumental magnitudes, which is roughly symmetrical between positive and negative values. Over time, the absolute difference in magnitude remains relatively constant, with an absolute value, on average, of about 0.5 m.u. In general, macroseismic magnitudes tend to slightly overestimate instrumental ones at lower values and underestimate them at higher values. For real-time applications, it may be useful to provide parameters using DB method with radii smaller than 2 km or, in particular, to give priority to BOXER-1 to assess the macroseismic epicentre. Furthermore, to simplify the choice between the available estimates, it could be worthwhile to adopt a selection criterion that supports the combination with the highest relative number of MDPs, although this does not necessarily guarantee that that value is the closest to the instrumental benchmarks. For possible applications in near real-time the key points to take into account are: i) high number of MDPs and low gap between MDPs generally improves the correlation between macroseismic and instrumental parameters; ii) distances and magnitude differences do not decrease monotonically over time but exhibit fluctuations, especially for small earthquakes, iii) the analyses highlights the importance of the radius used for clustering citizen IDPs, suggesting that smaller radii may provide more accurate macroseismic locations in densely populated areas and near the earthquake origin-time; iv) comparing the two BOXER methods, for time intervals shorter than five minutes, BOXER-1 shows better agreement in distance compared to BOXER-0. The number of localised earthquakes is lower for BOXER-1 than for BOXER-0 but, being the first one less sensitive to near-field effects only, it allows to constrain the epicentre more effectively than BOXER-0. Hence, for real time analyses, we suggest using this method when available, even with cluster method radii of 1 km and 0.5 km. A weight criterion based on the greater number of MDPs (or the smallest gap) used could be applied to identify a “preferred” solution among those available. For real-time applications, beyond the uncertainties provided by BOXER, the reliability of the parameters can be weighted according to the number of MDPs, the gap, or even the geographical region to which the event belongs. The reliability of this procedure in locating and sizing earthquakes over predefined time intervals emphasises the potential for its application in near real-time to provide valuable services to users and stakeholders, including civil protection agencies. Since the number of MDPs is crucial for reliable results, we hope that the collection of IDPs will continue to grow for each earthquake and that citizens will report their IDPs faster and faster. Declarations Data and resources Boxer code: freely available at: https://emidius.mi.ingv.it/boxer/ IDPs, individual intensities data points are downloaded by EMSC via webservices, e.g.: http://www.seismicportal.eu/testimonies-ws/api/search?unids=20210629_0000012&includeTestimonies=true (last accessed January 2025). Funding This work was partially supported by the European Union under the H2020 RISE project (number 821115) and the Progetto INGV Pianeta Dinamico (NEMESIS)-code CUP D53J19000170001-funded by Italian Ministry MIUR (Fondo Finalizzato al rilancio degli investimenti delle amministrazioni centrali dello Stato e allo sviluppo del Paese, 145/2018). Competing Interests The authors have no relevant financial or non-financial interests to disclose. Author Contributions All authors contributed to the study conception, design, material preparation and analysis. All authors read and approved the final manuscript. Data Availability IDPs, individual intensities data points are downloaded by EMSC via webservices, e.g.: https://www.seismicportal.eu/testimonies-ws/ (last accessed June 2025). Acknowledgements The EMSC (RB, ML) would like to thank the MAIF Foundation for supporting this research. Other acknowledgements to be written. Authors’ addresses Gianfranco Vannucci Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Bologna, Viale Berti Pichat 6/2, 40127 Bologna, Italy, email: [email protected] Remy Bossu European-Mediterranean Seismological Centre, c/o CEA, 91297 Arpajon, Cedex, France. CEA, DAM, DIF, F-91297 Arpajon, France, email: [email protected] Matthieu Landès European-Mediterranean Seismological Centre c/o CEA, 91297 Arpajon Cedex, France, email: [email protected] Paolo Gasperini Dipartimento di Fisica e Astronomia, Università di Bologna, Viale Berti Pichat 6/2, 40127 Bologna, Italy, email: [email protected] Declaration of Competing Interests: The authors acknowledge there are no conflicts of interest recorded. References Bossu, R., M. Laurin, G. Mazet-Roux, F. Roussel, and R. Steed (2015). The importance of smartphones as public earthquake-information tools and tools for the rapid engagement with eyewitnesses: A case study of the 2015 Nepal earthquake sequence, Seismol. Res. Lett. 86, 6, 1587–1592 Bossu, R., M. Landès, F. Roussel, R. Steed, G. Mazet-Roux, S. S. Martin, and S. Hough (2017). Thumbnail-based questionnaires for the rapid and efficient collection of macroseismic data from global earthquakes, Seismol. Res. Lett. 88, 1, 72-81. Bossu, R., F. Roussel, L. Fallou, M. Landès, R. Steed, G. Mazet-Roux, A. Dupont, L. Frobert, and L. Petersen (2018). LastQuake: From rapid information to global seismic risk reduction, Int. J. Disast. Risk Reduc. 28, 32-42. Bossu, R., F. Finazzi, R. Steed, L. Fallou, and I. Bondár (2021). “Shaking in 5 Seconds!”-Performance and User Appreciation Assessment of the Earthquake Network Smartphone-Based Public Earthquake Early Warning System, Seismol. Res. Lett. 93, 137–148, doi: 10.1785/0220210180. Bossu, R., M. Böse, R. Steed, and D. J. Wald (2024). The Potential of Crowdsourced Data for the Rapid Impact Assessment of Large Earthquakes: The 2023 M 7.8 Kahramanmaraş-Pazarcık, Türkiye, Earthquake, Seismol. Res. Lett. 95, 2058–2070, doi: 10.1785/0220230421. Castellaro, S., F. Mulargia, & Y. Y. Kagan (2006). Regression problems for magnitudes, Geophys. J. Int. 165, 913-930, doi: 10.1111/j.1365-246X.2006.02955.x. Dewey, J., D. Wald, and L. Dengler (2000). Relating conventional USGS modified Mercalli intensities to intensities assigned with data collected via the Internet, Seismol. Res. Lett. 71, 264. Ester M., H.P. Kriegel, J. Sander, and X. Xiaowei (1996). A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise. KDD'96: Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, 226-231 Fuller, W. A. (1987). Measurement Error Models, John Wiley, New York, 440 pp, doi: 10.1002/jae.3950030407. Gasperini, P., F. Bernardini, G. Valensise, and E. Boschi (1999). Defining seismogenic sources from historical earthquake felt reports. Bull. Seismol. Soc. Am. 89, 94-110. Gasperini, P., G. Vannucci, D. Tripone, and E. Boschi (2010). The location and sizing of historical earthquakes using the attenuation of macroseismic intensity with distance. Bull. Seismol. Soc. Am. 100, 2035-2066, doi: /10.1785/0120090330. Gasperini P., Lolli B., Vannucci G., and E. Boschi, A comparison of moment magnitude estimates for the European-Mediterranean and Italian regions, Geophysical Journal International, Volume 190, Issue 3, September 2012, Pages 1733–1745, https://doi.org/10.1111/j.1365-246X.2012.05575.x Goded T., N. Horspool, S. Canessa, A. Lewis, K. Geraghty, A. Jeffrey, and M. Gerstenberger (2018), New macroseismic intensity assessment method for New Zealand web questionnaires, Seismol. Res. Lett., 89(2A), 640-652, doi.org/10.1785/0220170163. Grünthal, G. (ed.) (1998). European macroseismic scale 1998, Conseil de l’Europe. Cahiers du Centre Européen de Géodynamique et de Séismologie,13, Luxembourg, 99 pp. Lolli B., P. Gasperini, and G. Vannucci (2014). Empirical conversion between teleseismic magnitudes (mb and Ms) and moment magnitude (Mw) at the Global, Euro-Mediterranean and Italian scale, Geophys. J. Int., 199, 805-828, doi: 10.1093/gji/ggu264 Lolli, B., D. Randazzo, G. Vannucci, and P. Gasperini (2020). The Homogenized Instrumental Seismic Catalog (HORUS) of Italy from 1960 to Present, Seismol. Res. Lett. 91, 3208–3222, doi: 10.1785/0220200148. Pasolini, C., D. Albarello, P. Gasperini, V. D'Amico, and B. Lolli (2008). The attenuation of seismic intensity in Italy part II: modeling and validation. Bull. Seismol. Soc. Am. 98, 692-708. doi.org/10.1785/0120070021. Rovida, A., M. Locati, R. Camassi, B. Lolli, and P. Gasperini (2020). The Italian earthquake catalogue CPTI15, Bull. Earthq. Eng., 18, 2953-2984, doi: 10.1007/s10518-020-00818-y. Tosi, P., P. Sbarra, V. De Rubeis, and C. Ferrari (2015), Macroseismic intensity assessment method for web-questionnaires. Seism. Res. Lett, 86, 985-990, doi: 10.1785/022014022. Vannucci G., D. Tripone, P. Gasperini, G. Ferrari, and B. Lolli (2015). Automated assessment of macroseismic intensity from written sources using the Fuzzy sets. Bulletin of Earthquake Engineering, 13, 2769-2803, doi:10.1007/s10518-015-9759-5 Vannucci G., P. Gasperini, B. Lolli, and L. Gulia (2019). Fast characterization of sources of recent Italian earthquakes from macroseismic intensities. Tectonophysics 750, 70-92, doi: 10.1016/j.tecto.2018.11.002 Vannucci, G., P. Gasperini, L. Gulia, and B. Lolli (2024). Earthquakes Parameters from Citizen Testimonies: A Retrospective Analysis of EMSC Database, Seismol. Res. Lett. 95, 969–996, doi: 10.1785/0220230245. Wald, D. J., V. Quitoriano, L. Dengler, and J. W. Dewey (1999). Utilization of the Internet for rapid community intensity maps, Seismol. Res. Lett. 70, 87–102. Wald, D.J., V. Quitoriano, B. Worden, M. Hopper, and J.W. Dewey (2011). USGS “Did You Feel It?” Internet-based macroseismic intensity maps. Annals Geophys., 54, 6; doi: 10.4401/ag-5354 Supplementary Files Supplementarymaterial.docx Cite Share Download PDF Status: Published Journal Publication published 08 Oct, 2025 Read the published version in Bulletin of Earthquake Engineering → Version 1 posted Reviewers agreed at journal 14 Jul, 2025 Reviewers invited by journal 11 Jul, 2025 Editor invited by journal 10 Jul, 2025 Editor assigned by journal 10 Jul, 2025 First submitted to journal 07 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7071982","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":484223346,"identity":"82ab26b7-05a8-4d32-9b35-e969ad11edbe","order_by":0,"name":"Gianfranco Vannucci","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEUlEQVRIiWNgGAWjYDACZiCuQBVibHyQAJPCpeUMmpZmA4gWXHowtTCwSSCMwwTy7szPHhzcY5fH397AJvFzx53E/vbDbRUPd9QmNrDzH8CmxfAwm7nBgWfJxRJnDrBJ9p55ljjjTGLbjcQzxxMbcDjMsJnBTPrDAebEDRIJzAa8bYeBDEaglrZjxrj8YtjM/k3iwIF6sBbDv1AtBfi0yDPzmAG1gFQmMD6G2cKQ2FYjh0uLATNPGVDLcaAXDjY+lm07bAz0S7NEYtsBOTZmZgOstvQf3wbUUg0MqOYDB9+2HZbtbz/+8OPPtjoefv6DD7DacgDOZGxAljjMwIbVWUBbGnBI1OEQHwWjYBSMghEIACffZAf375l8AAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-0918-0784","institution":"Istituto Nazionale di Geofisica e Vulcanologia Sezione di Bologna","correspondingAuthor":true,"prefix":"","firstName":"Gianfranco","middleName":"","lastName":"Vannucci","suffix":""},{"id":484223347,"identity":"cf7d4789-5c5f-4566-accc-a18cab8a386c","order_by":1,"name":"Remy Bossu","email":"","orcid":"","institution":"Centre Sismologique Euro-Méditerranéen: European-Mediterranean Seismological Centre","correspondingAuthor":false,"prefix":"","firstName":"Remy","middleName":"","lastName":"Bossu","suffix":""},{"id":484223348,"identity":"af6d630c-3b9b-4c2c-9b25-cc4b8ed13c81","order_by":2,"name":"Matthieu Landès","email":"","orcid":"","institution":"Centre Sismologique Euro-Méditerranéen: European-Mediterranean Seismological Centre","correspondingAuthor":false,"prefix":"","firstName":"Matthieu","middleName":"","lastName":"Landès","suffix":""},{"id":484223349,"identity":"893a2c65-97ee-4b91-a944-163bfd779998","order_by":3,"name":"Paolo Gasperini","email":"","orcid":"","institution":"Universita di Bologna","correspondingAuthor":false,"prefix":"","firstName":"Paolo","middleName":"","lastName":"Gasperini","suffix":""}],"badges":[],"createdAt":"2025-07-08 07:41:06","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7071982/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7071982/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s10518-025-02299-3","type":"published","date":"2025-10-08T15:57:41+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":86801112,"identity":"f717f9a9-4434-4498-bad1-c011538980f2","added_by":"auto","created_at":"2025-07-15 17:02:43","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":981325,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of the distance (in km) between macroseismic and instrumental epicentres at the global scale, calculated using both BOXER methods (BX-0, and BX-1) on MDPs derived from DB2-mn20 approach (see text for details). Distance ranges of earthquakes are colour-coded (see legend). The group of panels representing the elaborations BX-0 and BX-1, are indicated with a) and b), respectively. For both group of panels, the top X-axis indicates the number of earthquakes (n. Eqks) for each time interval, while the bottom X-axis shows the different time intervals relative to the event origintime T0 (s=seconds, m=minutes, d=day). For a) and b) groups, the top panel shows the cumulative histogram showing, over time (X-axis), the percentage (left Y-axis) and number (grey labels with dashed lines, on left Y-axis) of earthquakes on a global scale. The percentage of earthquakes is related to the total number of events (13724) at 1 day. Coloured lines on the graph indicate central tendency estimators: mean, median, and trimmed means (with 5%, 10%, 15% and 20% of tail trimming) of the earthquakes in each time interval (right Y-axis, in italic blue colour). For a) and b) groups the bottom panel represent the percentage distribution of earthquakes per distance bins (in km), relative to the number of events in each time interval. Numerical values in Tables S1-S9 of the Supplementary material. The panel c) shows the relative percentages for the same dataset of earthquakes of BOXER-1 computed by BOXER-0.\u003c/p\u003e","description":"","filename":"image1.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/e0311a071865e875b3692665.png"},{"id":86801968,"identity":"635c2766-8fa8-4f6d-ba60-9e92d34b07a2","added_by":"auto","created_at":"2025-07-15 17:10:43","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":724162,"visible":true,"origin":"","legend":"\u003cp\u003eDifference of magnitude (in m.u.) between macroseismic and instrumental parameters for BX-0. Cumulative values (top panel), percentages for negative and positive differences (middle panel) and absolute (ABS) value (bottom panel). Numerical values in Tables S7-S9 of the Supplementary material.\u003c/p\u003e","description":"","filename":"image2.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/717845aed496e24804262c54.png"},{"id":86800937,"identity":"2aa46e8d-dc4f-42f3-89e4-c1db7acb7fc1","added_by":"auto","created_at":"2025-07-15 16:54:43","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":437517,"visible":true,"origin":"","legend":"\u003cp\u003eDistribution of macroseismic (Mm, Y-axis) and instrumental (Mw, X-axis) magnitudes of earthquakes analysed for different time intervals on a global scale by BOXER-0 on MDPs from DB2-mn20 approach (see text for details). The palette quantifies the number of MDPs available for each earthquake. For each time interval, the 1:1 line of the values of the two variables (in dashed grey) and the orthogonal regression line (GOR, in dashed blue) are shown. The number of earthquakes in each time interval is shown in Fig. 2.\u003c/p\u003e","description":"","filename":"image3.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/292edbcd4bbc57e8ca4a9a1b.png"},{"id":86800940,"identity":"eadeba5f-9499-4951-8021-86dbdcbc0794","added_by":"auto","created_at":"2025-07-15 16:54:43","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":341395,"visible":true,"origin":"","legend":"\u003cp\u003eTemporal variations (bottom X-axis) in distance (top panel) and magnitude difference (bottom panel) for BOXER-0 (BX-0). Bars represent the percentage of earthquakes relative to the total number of events in each time interval (top X-axis of both panels). Earthquakes included for the first time are shown in grey, while coloured bars indicate whether common events, present in previous intervals, show no change (green), improvement (blue), or worsening (red). Results for BOXER-1 in Fig. E of the supplementary material.\u003c/p\u003e","description":"","filename":"image4.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/fd458772673c58747d0945c0.png"},{"id":86801114,"identity":"0e3b6f95-f5ab-4c3e-998f-a0813122e6da","added_by":"auto","created_at":"2025-07-15 17:02:43","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":353823,"visible":true,"origin":"","legend":"\u003cp\u003eNumber of earthquakes (N. Eqks) over time for which the first instrumental (top panel, in black) or macroseismic data (epicentre and magnitude, middle and bottom panels, respectively) are available. Macroseismic data, computed using BOXER-0, indicate whether macroseismic parameters precede (blue), follow (red), or are contemporaneous (green) with the instrumental ones. Please note that the plots refer to the data since 2016, therefore, n. Eqks on the upper X-axis for macroseismic data are lower than the corresponding Figs 1-2.\u003c/p\u003e","description":"","filename":"image5.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/2a6834257a2ba6b6cf26b05b.png"},{"id":86800946,"identity":"68317fa4-2276-47d7-91b9-59a0668ee87b","added_by":"auto","created_at":"2025-07-15 16:54:43","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":686357,"visible":true,"origin":"","legend":"\u003cp\u003eDistance (Di) and difference of magnitude (dM) (left Y-axis) over time (X-axis) between macroseismic and instrumental parameters of EMSC by both BOXER methods, for three earthquakes of example. The number of MDPs available (n. MDPs) is indicate (right Y-axis, in italic red colour). MDPs form DB2-mn20 (see text for details). The grey vertical lines indicates the time (Ti) when the instrumental data, with respect from T0, is available for the earthquakes (163, 290, 221 secs, respectively). Note that Ti is the time at which the first instrumental parameters, among those available from different agencies over time, is available at EMSC, but it is not possible to know whether this instrumental datum corresponds to the true final benchmark. Numerical Values in Table S17 of the Supplementary material.\u003c/p\u003e","description":"","filename":"image6.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/a05dc39cc780ea7271a03341.png"},{"id":86801115,"identity":"373eed91-bd49-45c7-9bbe-62199cafb7ef","added_by":"auto","created_at":"2025-07-15 17:02:43","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":789364,"visible":true,"origin":"","legend":"\u003cp\u003eresults for distance, difference of magnitude and absolute difference of magnitude for both BOXER methods. Top three panels are for BOXER-0 and bottom three panels for BOXER-1. Comparison over time at different geographic areas (World, W, left column and Europe, EU, right column) using different data filters (legend). Results of BOXER-0 on the dataset of BOXER-1 in Figure S20 of the supplementary material.\u003c/p\u003e","description":"","filename":"image7.png","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/be603106c068dc31e800855a.png"},{"id":93420701,"identity":"9f63d593-3255-486b-b706-8ad8c6d52fc8","added_by":"auto","created_at":"2025-10-13 16:10:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4769182,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/16b9b6d7-ad22-400e-a736-fb58520ec5f6.pdf"},{"id":86801970,"identity":"93eb1924-0bc8-41a9-90e0-df988f45802e","added_by":"auto","created_at":"2025-07-15 17:10:43","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":15303078,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-7071982/v1/df6d0417f40f2d6f50f8c239.docx"}],"financialInterests":"","formattedTitle":"Can we quickly calculate reliable earthquake parameters from citizen testimonies?","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn recent years, an area of so-called crowdsourcing science, based on information provided directly by citizens, has developed and become increasingly important. Some agencies collect this information in near real-rime, such as \u0026ldquo;Did you feel it?\u0026rdquo; (DYFI; Wald et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1999\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2011\u003c/span\u003e, Dewey et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), of the U.S. Geological Survey (USGS), \u0026ldquo;Hai sentito il terremoto?\u0026rdquo; (HSIT; Tosi et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) of the Istituto Nazionale di Geofisica e Vulcanologia (INGV), New Zealand GeoNet questionnaires (GeoNet, Goded et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), LastQuake by the Euro-Mediterranean Seismological Centre (EMSC) (Bossu et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eVannucci et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) verified the possibility of deriving reliable macroseismic parameters from individual intensities (intensity data points, IDPs) provided according to EMS scale (Grunthal et al., 1998) by each eyewitness citizen, collected and made available by the LastQuake system. This retrospective analysis allowed to define and overcome some of the limitations in the data, such as the presence of possible outliers in both geographical location and intensity value. In some cases, IDPs may be geographically misplaced due to factors such as a VPN connection or the poor location provided by users (e.g., only the city rather than a precise street address) when submitting reports via the LastQuake system on a PC rather than a smartphone. EMSC also provides, in addition to the original IDP intensity, a second value of IDP intensity modified by a formula, derived by Bossu et al. (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2015\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), to best reproduce DYFI intensities. However, IDPs cannot be directly treated as classical macroseismic intensities, which represent the earthquake effects at the scale of a city or town and are, therefore, representative of effects over an extended area rather than the punctual observation of an individual user. Hence, a spatial clustering procedure was applied, requiring a minimum number (usually three) of IDPs per cluster (Vannucci et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This approach also eliminates geographical outliers and mitigates intensity outliers (typically overestimations) through central tendency estimators such as the mean, the median, or the trimmed mean (i.e. average of intensities calculated after excluding a percentage of the tails of the distribution of IDPs). Consequently, we assign a single equivalent intensity value to each cluster by a straightforward, rigorous, and reproducible procedure, making it comparable to traditional macroseismic data points (MDPs).\u003c/p\u003e\u003cp\u003eTo validate their approach and identify the most effective combinations for grouping IDPs into MDPs, Vannucci et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) carried out a retrospective analysis on the complete distribution of IDPs (both original and modified) for each event in the EMSC catalogue. They combined data clustering techniques and central tendency estimators, using the BOXER code (Gasperini et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) to estimate earthquake parameters (location and size). The best combination for minimising statistical deviations from the instrumental data was found to be the DBSCAN spatial clustering approach (DB, Ester et al, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1996\u003c/span\u003e) with a 2-km radius (DB2), a 20% trimmed mean (mn20) for distance and a 10% trimmed mean (mn10) for magnitude. Overall, most combinations that minimise differences with instrumental benchmarks are associated with the DB method, with only small variations between different statistical estimators.\u003c/p\u003e\u003cp\u003eIn last years, it has been possible to produce quantitative outputs with high reliability as instrumental methods, using macroseismic intensities (Vannucci et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), also provided by citizens (Vannucci et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In this work, we aim to verify the time-dependent reliability of macroseismic parameters (location and magnitude) of earthquakes using the BOXER code after grouping the original IDPs into MDPs.\u003c/p\u003e"},{"header":"Method","content":"\u003cp\u003eThe IDPs distributed by EMSC for each event include the time delay in seconds (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://seismicportal.eu/testimonies-ws/\u003c/span\u003e\u003cspan address=\"https://seismicportal.eu/testimonies-ws/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) from the earthquake origin time T0, at which users sent their reports. Therefore, for this retrospective analysis, aimed to quantify the time-dependent reliability of macroseismic locations and sizes, we clustered available IDPs over predefined time intervals from T0 for earthquakes on a global scale from 2012 to 31 May 2024. From this date onward, EMSC has been testing a system that continuously processes location and macroseismic magnitude in real time using citizen IDP. However, this test is outside the scope of this work and will be addressed in future research.\u003c/p\u003e\u003cp\u003eGrouping distance and difference of magnitude within specific time intervals represents a kind of stacking approach (e.g. Gulia et al., 2018) to enhance the signal and enable comparisons between any different combinations of grouping methods, central tendency estimators and BOXER methods. We selected successive time intervals relative to T0: every 30 seconds up to 10 minutes, including an additional 45s interval (30s, 45s, 1m, 1m30s, 2m \u0026hellip;), every minute up to 15 minutes from T0 (10m, 11m\u0026hellip;) and finally 20, 30 minutes and one day.\u003c/p\u003e\u003cp\u003eWe use a minimum of three IDPs per cluster to obtain the MDPs, for each earthquake at each time interval, which are then processed with the BOXER code (Gasperini et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1999\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2010\u003c/span\u003e) to assess the macroseismic location and magnitude. Based on the findings of Vannucci et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), the DBSCAN method is the most effective and fast and provides the most reliable results, independently of the statistical estimator. Moreover, this method offers the key advantage of grouping close IDPs without requiring predefined areal boundaries, such as polylines or equivalent radii simulating the area of a locality, or predefined geometric grids that may distort the true urban extent. Therefore, it is ideal for near real-time applications. As statistical estimators, we considered both mn20 for distance and mn10 for magnitude, according to Vannucci et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The values obtained are quite similar for all the trimmed means, so to streamline the process for possible near real-time applications, we selected mn20 as the default estimator to calculate both epicentre and macroseismic magnitude. Hence, we systematically applied the combination DB2-mn20 to the selected time intervals.\u003c/p\u003e\u003cp\u003eThe analysed dataset, up to 31 May 2024, consists of 17243 earthquakes with sufficient IDPs to generate at least one MDP. On these time intervals, we apply the BOXER code to locate the earthquakes using two methods: i) method 0 (BOXER-0), computing the barycentre of the sites with the most severe effects (analogous to a centroid of the area of deformation or maximum slip) and ii) method 1 (BOXER-1), computing the centre of the entire intensity distribution by minimising the squared residuals of the intensity prediction equation (IPE) by Pasolini et al. (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). A key difference between these two methods is that method 0 can locate even with just one MDP, while method 1 requires a minimum of 5 MDPs. The algorithm for magnitude estimation requires a minimum of 4 MDPs even for events located with method 0. Therefore, as the number of IDPs increases over time, leading to a higher number of MDPs, the first parameter to be estimated is the epicentre by BOXER-0, followed by the magnitude, and finally the epicentre by BOXER-1.\u003c/p\u003e"},{"header":"Results and discussions","content":"\u003cp\u003eThe results are illustrated by plots (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e) showing, at different delays from origin time, the distribution of distances and of magnitude difference between macroseismic and instrumental parameters.\u003c/p\u003e\u003cp\u003eThe total number of earthquakes for which macroseismic parameters can be calculated obviously increases over time as more IDPs become available. Using the IDPs collected within one day (1d) from T0, 13724 and 3851 macroseismic epicentres (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e) as well as 3851 and 3780 macroseismic magnitudes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. S1 of the supplementary material) can be calculated by BOXER-0 and BOXER-1, respectively. It is important to note that the number of earthquakes determined at the 1d time interval is lower than the total number of earthquakes (17423) with at least one MDP, as in many cases, the IDPs were supplied by citizens more than 24 hours after T0. However, for the purposes of this work, we only consider the IDPs supplied within the 1d interval.\u003c/p\u003e\u003cp\u003eFor both BOXER methods, we analysed the number of earthquakes, whose macroseismic epicentres or magnitudes can be estimated at each time interval from T0, as a function of different ranges of distance (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e) and difference in magnitude (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. S1 of the supplementary material) relative to instrumental benchmarks. In these Figures, we also plotted central tendency estimators (mean, median, trimmed means) to quantify the agreement between macroseismic and instrumental parameters. Finally, to facilitate evaluations and comparisons (e.g. between the two BOXER methods), we calculated the relative percentage of earthquakes within each distance or magnitude range (bin) with respect to the total number of earthquakes in the same time interval,\u003c/p\u003e\u003cp\u003eThe number of earthquakes for which both macroseismic and instrumental epicentres can be compared in terms of distance increases over time. Using BOXER-0, we have 1206 earthquakes (about 9% of the events located at 1d time interval) are located within 1 minute, 3429 (20%) within 2 minutes, 5049 (37%) within 2 minutes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The macroseismic epicentre can be estimated for 6186 earthquakes (45% of those localised at 1d time interval) within 4 minutes, a timeframe that, for example in Italy, is conventionally understood as the limit within which the Istituto Nazionale di Geofisica e Vulcanologia (INGV) must provide instrumental parameters of ongoing earthquakes to the Civil Protection Department. The number of earthquakes in agreement with instrumental data and their percentage with respect to earthquakes localised after 1day increases over time (e.g. green bars in Fig, 1). In terms of relative percentage, the agreement between macroseismic and instrumental epicentres within various distance bins is noteworthy, with high percentages within 10 km and 20 km (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The relative percentage of earthquakes whose distance is consistent with instrumental data decreases in the first 4 minutes from T0, then almost stabilises at ~\u0026thinsp;25% within 10 km and ~\u0026thinsp;50% within 20 km. Up to the 4-minute interval, the average distance of earthquakes in each time interval tends to increase, then stabilises around values of 30\u0026ndash;40 km for the median and trimmed means (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The mean shows the same trend but with higher values (\u0026thinsp;~\u0026thinsp;+\u0026thinsp;10 km), because, unlike the other estimators, it is more affected by outliers (earthquakes located at greater distances from the instrumental reference epicentre) despite their relatively lower occurrence.\u003c/p\u003e\u003cp\u003eThe results with BOXER-1 show a lower number of located earthquakes compared to BOXER-0 due to the minimum number of MDPs (5) required for the location, and a slightly better agreement than that with BOXER-0, in particular for time intervals\u0026thinsp;\u0026lt;\u0026thinsp;5 minutes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This can be directly observed by comparing the results obtained with BOXER-0 on the same earthquakes located by BOXER-1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). The plots between distance and magnitude also as a function of the number of MDPs are similar to those observed with BOXER-0 (Figs S2 and S3 in the supplementary material). Although the total number of located earthquakes is lower in BOXER-1 compared to BOXER-0, the percentage of successfully located events is higher in BOXER-1, making it the preferable location method. For both BOXER methods, locations estimated within a few minutes from T0 are highly reliable.\u003c/p\u003e\u003cp\u003eConcerning the difference between macroseismic and instrumental magnitudes, a remarkable symmetry can be observed in the histogram of events between positive and negative differences over time (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e, for BOXER-0 and Fig. S1 in the supplementary material for BOXER-1). Positive values indicate an overestimation of the macroseismic magnitude with respect to the instrumental one, while negative values indicate the opposite (represented by red and blue colours, respectively, in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e). On average, the difference is close to 0 with a standard deviation of about 0.7 magnitude unit (m.u.). However, within a minute from T0, there is an overestimation of the macroseismic magnitude, although due to a small number of events. Even when considering the relative percentage of events in the different ranges, the symmetry between positive and negative values remains evident (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Both BOXER methods (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e for BOXER-0 and Fig. S1 in the supplementary material for BOXER-1) provide very similar results, as are based on the same calculation model and use the same number of MDPs. The differences in magnitude values arise from variations in the felt area within the isoseismal radius of the intensity classes, centred on the estimated epicentre, which may slightly differ between the two BOXER methods, slightly affecting the computed magnitudes.\u003c/p\u003e\u003cp\u003eWhen considering the difference in absolute value, we observe that, for each time interval (except the first 30 seconds, where only two earthquakes are included), about 20% of the earthquakes occur within a range of 0.1 m.u., 30% within 0.2 m.u., 40% within 0.3 m.u. and 50% within 0.5 m.u. for both BOXER methods (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e2\u003c/span\u003e and Fig. S1 of the supplementary material). The overall average magnitude difference rises to 0.5 m.u. (\u0026plusmn; 0.5). Macroseismic magnitudes tend to be slightly overestimated with respect to instrumental ones at lower values and slightly underestimated at higher values (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e and Fig. S4 in the Supplementary material). In particular, as the number of MDPs increases, the magnitude difference decreases (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e), i.e. earthquakes with a high number of MDPs tend to minimise the magnitude differences (Figs. S4-S7 of the Supplementary material). It is noteworthy the absence of macroseismic magnitudes lower than 2.5. This is a well-known feature already noted in other studies (e.g. Rovida et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). It might depend on the intensity-magnitude regression models, which cannot take into account variables (e.g. hypocentral depth, site effects) that do not vary the magnitude but influence the observed effects (i.e. the calculated intensity).\u003c/p\u003e\u003cp\u003eTo check the agreement between macroseismic and instrumental magnitudes, we computed linear regressions between instrumental and macroseismic magnitudes in each time interval (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e), using the General Orthogonal Regression (GOR) method (Fuller, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e1987\u003c/span\u003e, Castellaro et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2006\u003c/span\u003e, Gasperini et al., 2012, Lolli et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2014\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), a linear regression model assuming a constant ratio between the variances of the y and x variables. A slope coefficient close to 1 would indicate a nice agreement between instrumental and macroseismic magnitudes. In our case, the GOR shows a partial agreement between instrumental and macroseismic magnitudes with regression slopes of \u0026sim;0.6\u0026ndash;0.66 for BOXER-0 (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e) and slightly higher values for BOXER-1 (Fig. S8, in the Supplementary Material). As the number of MDPs increases, even the slope increases (0.7\u0026ndash;0.8), indicating that the agreement between macroseismic and instrumental parameters approaches a 1:1 relationship (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e). This highlights the importance of the number of MDPs for the reliability of the parameters. Macroseismic magnitudes slightly overestimate instrumental ones below ~\u0026thinsp;4.5 and underestimate them above that value.\u003c/p\u003e\u003cp\u003eSeveral hypotheses can be proposed to explain the observed findings, particularly the reduction with the distance of the relative percentage agreement with instrumental data (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). In densely populated areas, the first IDPs (and thus the MDPs) should be close to the instrumental one. Over time, as the felt area increases due to geometrical spreading, other MDPs farther away may change the geographical distribution and the macroseismic location. In general, we observed a slight direct correlation between the increase in distance and the increase in magnitude, in particular for distances below 50 km. Distances greater than 50 km, on the other hand, are mainly observed for M3\u0026thinsp;+\u0026thinsp;magnitude earthquakes. As both distance and magnitude increase, earthquakes decrease in number and become more scattered (Figs. S9 and S10 in the Supplementary material). Moreover, most earthquakes with a high number of MDPs (\u0026gt;\u0026thinsp;40) are located within 50 km of the instrumental epicentre, regardless of the earthquake magnitude. An increase in the number of MDPs generally improves the agreement between macroseismic and instrumental data, as observed in Vannucci et al. (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Another explanation is related to the quality of the available information. Each bin in the various time intervals does not necessarily contain the same earthquakes: events within a certain time interval, may not have sufficient MDPs for location and size, or, if MDPs allow the parameter assessment, the increase of collected IDPs over time can lead changes in number or MDPs, changes in their geographical distribution an intensity variation, so changing previous macroseismic parameters of an event. Moreover, new earthquakes could be added, so quickly varying the relative statistics over time. By comparing the parameters in a given time interval with those in the previous one, for the same earthquakes, variations may be absent or may indicate improvement (positive values) or worsening (negative values) of the agreement with the instrumental data (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e, and Fig. S11 in the Supplementary material). In most of the cases there are not changes in earthquake parameters over time or the variations are minimal (0\u0026ndash;10 km in distance or 0-0.2 in absolute magnitude difference for most of earthquakes (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e) Over time, new earthquakes are progressively included in the dataset (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e4\u003c/span\u003e) and this results in an overall decrease in relative percentage agreement observed for the distance between macroseismic and instrumental epicentres (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e1\u003c/span\u003e). This trend may be partly attributed to poorly located events, such as those occurring offshore or in sparsely populated areas, and reflects the growing inclusion of earthquakes from outside Europe and North America, i.e. Asia-Oceania, South America, and Africa. Indeed, the relative proportion of such events within the dataset increases over time (see Fig. S12 in the Supplementary material), and the agreement with instrumental data tends to be lower (Vannucci et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Moreover, some agreement worsening (i.e. negative variations) in earthquake parameters may be due to delayed notifications, even from areas outside the true felt area, which can trigger some unreliable reports, i.e. when individuals retrospectively believe they experienced shaking.\u003c/p\u003e\u003cp\u003eIt is worth noting that the procedure allows a rapid estimation of macroseismic parameters, often preceding the availability of corresponding instrumental data (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e). The comparison refers to the time of \u0026ldquo;first\u0026rdquo; availability of macroseismic (Tm) and instrumental (Ti) data, using records since 2016 only, the year in which EMSC began archiving Ti in the IDP database. Note that Ti, in most cases, refers to the first determination of instrumental parameters rather than the final one, due to possible later revisions (numerical changes or replacement by another agency). Therefore, this analysis considers 11723 and 3595 events for distance, and 3676 and 3364 for magnitude difference, using BOXER versions 0 and 1, respectively (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e and S13 of the Supplementary material, respectively). The macroseismic parameters can be classified as simultaneous (Tm\u0026thinsp;=\u0026thinsp;Ti), earlier (Tm\u0026thinsp;\u0026lt;\u0026thinsp;Ti), or later (Tm\u0026thinsp;\u0026gt;\u0026thinsp;Ti) based on the timing vs instrumental parameters. The elaborations show that macroseismic parameters often become available earlier than instrumental ones within the first few minutes after T0. In particular, using BOXER-0, 7532 macroseismic epicentres (64% of the total sample) are located before the corresponding instrumental ones, of which 5145 (\u0026sim;44% of the total) within the first 4 minutes (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e5\u003c/span\u003e and S13 and Table S13 of the Supplementary material). Similarly, 1586 magnitudes (43% of the total) are assessed before the corresponding instrumental ones, of which 1000 (\u0026sim;27% of the total) within the first 4 minutes from T0 (Table S14 of the Supplementary material). These earlier locations are probably mostly related to low-magnitude earthquakes (lower magnitudes show shorter delays, Fig. S2 of the supplementary material) and are certainly influenced by population density and LastQuake system dissemination (Vannucci et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe variability and validity of the calculated parameters may also depend on the dimensional scale on which the IDPs are aggregated. This could lead to the creation of an MDP that does not accurately reflect real urban areas: larger radii collect sparse reports but may blur detail in densely populated area while smaller radii could offer finer localisation, increasing the geographic detail in the representation of felt effects. Thus, in addition to the 2 km radius (DB2), we tested the EMSC database grouping the IDPs by DBSCAN method, with a radius of 1 km (DB1) and 0.5 km (DB0.5) We still use the estimator mn20 to compute MDPs. Trends are similar across radii, though smaller radii reduce the total number of located events (11999 for DB1; 10190 for DB0.5) but yield slightly higher percentages of accurate locations, especially within short intervals from T0 (Figs S14 and S15 for distance and difference of magnitude in the supplementary material). The results show slightly higher percentages of agreement between macroseismic and instrumental epicentres (i.e. lower distance) when using BOXER-1, particularly with DB1 and DB0.5 (Figs S14 and S15, supplementary material). With DB1, 3990 earthquakes are located after one day (i.e. slightly more than with DB2), and there is a notable increase in events localised within shorter intervals from T0, especially within the first 3 minutes, enhancing agreement with instrumental data. Although the total number of earthquakes decreases with DB0.5 (3532), the count remains higher in the short time intervals, and the percentage of earthquakes within 20 km of the instrumental epicentre consistently exceeds 60%. This suggests that reducing the DB radius improves the match with instrumental locations, for short time intervals, with BOXER-1 statistically preferable to BOXER-0. Basically, for densely populated areas, DB1 and DB0.5 are better at pinpointing the epicentre, especially in the shorter time intervals, while DB2 lets you estimate parameters for more events by using a bigger IDP clustering area.\u003c/p\u003e\u003cp\u003eRegarding magnitude differences, for both DB1 and DB0.5, the BOXER methods show comparable trends. The DB1 approach consistently yields more events than DB2, while DB0.5 sees fewer after 10 minutes. The agreement with the instrumental data is comparable to DB2, with a difference, on average, close to 0. However, there is a slight overestimation of the macroseismic magnitude during the first three minutes (Figs S16-S19 of the Supplementary material) with respect to DB2. This trend is probably related to the calibration method adopted, as previously discussed (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWe can illustrate the potential parameter changes over time with a practical example by presenting three earthquakes of varying magnitudes and geographical locations (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003e and numerical details in Table U or supplementary material), and analysing how their distance and magnitude difference vary for both BOXER methods. The earthquakes are: i) the 2022, May 12th (21:12 UTC, Mw 3.7) event, located South of Florence, Italy (incidentally, it occurred during the social dinner of the European RISE Project meeting); ii) the 2023, Feb. 6th (1:17 UTC, Mw 7.8) event, located in Kahramanmaras-Pazarcik area, Turchey; iii) the 2024, April, 5th (14:23 UTC, Mw 4.8) event, located in New Jersey, USA. For each earthquake, Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents two graphs, showing temporal variations in distance (Di) and magnitude difference (dM), with respect to the instrumental benchmark of EMSC. Full-time evolutions of IDPs, MDPs, and macroseismic epicentres are provided as videos in the supplementary material. Parameter trends follow a consistent pattern: BOXER-0 gives initial locations; BOXER-1 activates once five MDPs are reached. Magnitude is computed simultaneously by both BOXER methods (at least four MDPs). While differences with instrumental benchmarks tend to decrease, the minimum values do not always coincide with the latest time intervals, where the highest number of MDPs is typically available, indicating non-monotonic behaviour. Notably, MDP numbers also fluctuate: new IDPs may aggregate previously distinct MDPs, reducing totals, as in the Florence case. There, a macroseismic epicentre (12.5 km from the benchmark) is calculated within 90 seconds from T0. The distance increases with time, then decreases after 14 minutes. The first macroseismic magnitude, 7 minutes after T0, overestimates the instrumental value by 0.7 m.u, and later the magnitude difference fluctuates between 0.2 and 1 m.u. Early localisation is due to initial IDPs (then MDPs) near the epicentre, which later changes, with distant IDPs expanding the felt area modifying MDPs (in general low in number) and location assessment.\u003c/p\u003e\u003cp\u003eFor the large-magnitude Turkey earthquake, BOXER-0 locates the epicentre 3.5 minutes after T0, about 500 km away, due to a\u0026ldquo;doughnut effect\u0026rdquo; i.e. the lack of reports and therefore intensities from the severely affected epicentral area (Bossu et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Increasing the number of MDPs over time, BOXER-1 provides a first macroseismic epicentre after 4 minutes from T0, improving localisation up to ~\u0026thinsp;30 km within 20 minutes. However, the magnitude remains underestimated: initially \u0026minus;\u0026thinsp;3 m.u., rising to -1.2 after 6.5 minutes, but still \u0026minus;\u0026thinsp;0.8 after one day, due to the lack of high intensities in the epicentral area.\u003c/p\u003e\u003cp\u003eThe New Jersey event shows a similar pattern to the Turkey earthquake: BOXER-0 locates at a distance of 65 km after 2.5 minutes, improved by BOXER-1 after 5 minutes, then decreasing the distance up to 8 km after 14 minutes. Magnitude starts at -1.5 m.u., and the difference with respect to the instrumental value decreases significantly, reaching zero after 11 minutes. In the latter two examples, there is a clear inverse trend between the increase of the number of MDPs and the reduction in both distance (in particular with BOXER-1) and magnitude difference.\u003c/p\u003e\u003cp\u003eReducing the DB radius increases MDPs for the Florence and Turkey earthquakes, improving resolution and accuracy, especially for the lower-magnitude 2022 event (Table U in the supplementary material). The DB2 approach gives better results for larger earthquakes (Turkey and New Jersey), except for occasional time intervals. For the 2022 earthquake, which has a lower instrumental magnitude and thus a smaller felt area, the higher MDP resolution better constrain the macroseismic epicentre, in particular with BOXER-1. Magnitude differences tend to be smaller across radii, with better matches occurring when more MDPs are present.\u003c/p\u003e\u003cp\u003eThe results of the analysis suggest that several factors may concurrently affect the calculated parameters as the number of MDPs, azimuthal gap among MDPs, spatial distribution, magnitude, and potential edge effects, such as the doughnut effect. To assess whether the agreement between macroseismic parameters and benchmarks depends on specific factors, even over time, we analysed the datasets by applying conditional thresholds to define subsets of data. For example, we distinguished between earthquakes with instrumental epicentres located offshore or inland, or between events associated with a number of MDPs greater or lower than a threshold. The comparison was made, for both BOXER methods, for distance and magnitude difference, also in absolute value, with respect to the instrumental parameters, using a 10% trimmed mean and sampling the data in the various time intervals. We show the results at the global and European scale (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e). In general, the results are fairly stable over time intervals, without large variation, except for the data in the very first intervals from T0. The agreement (smaller distances and differences in magnitude between macroseismic data and instrumental benchmarks) increases with the number of MDPs, while earthquakes that occur on land are, on average, better located than earthquakes at sea. The data collected in recent years show very little improvement in the results with respect to the previous ones. The agreement increases by decreasing the azimuthal gap, although without significant changes. Conversely, the agreement gets worse significantly showing fluctuations over time when the gap exceeds 180 degrees, in particular for magnitude (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e). This condition is generally associated with localities bordering IDP-blind areas, such as sparsely populated regions, natural boundaries (e.g. coastlines, deserts, mountain ranges), or countries with restrictive policies on the use of smartphones or related applications (see Vannucci et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e for details). A general worsening of results, at the global scale and in both distance and magnitude, is observed for magnitude (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e). In particular, the agreement between macroseismic and instrumental parameters progressively worsens for M4\u0026thinsp;+\u0026thinsp;events and with increasing magnitude. On average, the difference remains stable over time (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e), even if results for some earthquakes may improve over time (e.g. the events in Turkey and New Jersey, Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e6\u003c/span\u003e). This pattern is likely due to the \u0026ldquo;doughnut\u0026rdquo; effect in the epicentral area, where, after a strong event, citizen priorities are far from the fast reporting of effects. However, it should be noted that the same analyses carried out at the European scale, based on a dataset comprising more than 2/3 of the total global earthquakes (Fig. S12 of the Supplementary Material), show lower differences, particularly with respect to distance when using BOXER-0, with a \u0026sim;30% compared to the values obtained at the global scale (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e). When considering BOXER-1, the average deviation from the benchmark at European scale is reduced by more than 70% compared to BOXER-0 and by \u0026sim;50% compared to BOXER-1, both at global scale. An average worsening in agreement (epicentral distance\u0026thinsp;\u0026gt;\u0026thinsp;50 km) is observed only for events of magnitudes M5/M5.5\u0026thinsp;+\u0026thinsp;in contrast to the other subsets, which overall fall within the 20\u0026ndash;40 km range (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e). This suggests that on a global scale, the largest differences come from macro-areas outside Europe. The only worsening observed at the European scale, compared to the global one, concerns magnitude differences for events with a gap\u0026thinsp;\u0026gt;\u0026thinsp;180 degrees, although this may be attributable to the small number of events in this subset (a maximum of 21 events in the 1-day time interval). Finally, this analysis shows that the results of BOXER-1 are generally slightly better than the corresponding values of BOXER-0 (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eWe used the individual intensities reported by citizen (IDPs) collected and made available by EMSC, to verify the reliability of the location and macroseismic magnitude calculated by applying a specific procedure that first groups IDPs into traditional intensities (MDPs) and then processes them using the BOXER code. We selected the DBSCAN clustering method with a 2 km radius to group IDPs and applied BOXER methods 0 and 1 to process MDPs. The procedure was applied within predefined time intervals after the origin time of global-scale earthquakes.\u003c/p\u003e\u003cp\u003eOur results show that within the first four minutes, a significant percentage of earthquakes with macroseismic parameters are located at relatively short distances (e.g., 10\u0026ndash;20 km), with an overall low average distance from instrumental epicentres. In most cases, these data predate the availability of instrumental data. Over time, this distance slightly increases with values, on average, of about 25\u0026ndash;35 km. We also analysed the difference between macroseismic and instrumental magnitudes, which is roughly symmetrical between positive and negative values. Over time, the absolute difference in magnitude remains relatively constant, with an absolute value, on average, of about 0.5 m.u. In general, macroseismic magnitudes tend to slightly overestimate instrumental ones at lower values and underestimate them at higher values. For real-time applications, it may be useful to provide parameters using DB method with radii smaller than 2 km or, in particular, to give priority to BOXER-1 to assess the macroseismic epicentre. Furthermore, to simplify the choice between the available estimates, it could be worthwhile to adopt a selection criterion that supports the combination with the highest relative number of MDPs, although this does not necessarily guarantee that that value is the closest to the instrumental benchmarks.\u003c/p\u003e\u003cp\u003eFor possible applications in near real-time the key points to take into account are: i) high number of MDPs and low gap between MDPs generally improves the correlation between macroseismic and instrumental parameters; ii) distances and magnitude differences do not decrease monotonically over time but exhibit fluctuations, especially for small earthquakes, iii) the analyses highlights the importance of the radius used for clustering citizen IDPs, suggesting that smaller radii may provide more accurate macroseismic locations in densely populated areas and near the earthquake origin-time; iv) comparing the two BOXER methods, for time intervals shorter than five minutes, BOXER-1 shows better agreement in distance compared to BOXER-0. The number of localised earthquakes is lower for BOXER-1 than for BOXER-0 but, being the first one less sensitive to near-field effects only, it allows to constrain the epicentre more effectively than BOXER-0. Hence, for real time analyses, we suggest using this method when available, even with cluster method radii of 1 km and 0.5 km. A weight criterion based on the greater number of MDPs (or the smallest gap) used could be applied to identify a \u0026ldquo;preferred\u0026rdquo; solution among those available.\u003c/p\u003e\u003cp\u003eFor real-time applications, beyond the uncertainties provided by BOXER, the reliability of the parameters can be weighted according to the number of MDPs, the gap, or even the geographical region to which the event belongs. The reliability of this procedure in locating and sizing earthquakes over predefined time intervals emphasises the potential for its application in near real-time to provide valuable services to users and stakeholders, including civil protection agencies. Since the number of MDPs is crucial for reliable results, we hope that the collection of IDPs will continue to grow for each earthquake and that citizens will report their IDPs faster and faster.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData and resources\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBoxer code: freely available at: https://emidius.mi.ingv.it/boxer/\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIDPs, individual intensities data points are downloaded by EMSC via webservices, e.g.: http://www.seismicportal.eu/testimonies-ws/api/search?unids=20210629_0000012\u0026amp;includeTestimonies=true (last accessed January 2025).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was partially supported by the European Union under the H2020 RISE project (number 821115) and the Progetto INGV Pianeta Dinamico (NEMESIS)-code CUP D53J19000170001-funded by Italian Ministry MIUR (Fondo Finalizzato al rilancio degli investimenti delle amministrazioni centrali dello Stato e allo sviluppo del Paese, 145/2018).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors have no relevant financial or non-financial interests to disclose.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed to the study conception, design, material preparation and analysis. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIDPs, individual intensities data points are downloaded by EMSC via webservices, e.g.: https://www.seismicportal.eu/testimonies-ws/ (last accessed June 2025).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe EMSC (RB, ML) would like to thank the MAIF Foundation for supporting this research.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eOther acknowledgements to be written.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; addresses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGianfranco Vannucci\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIstituto Nazionale di Geofisica e Vulcanologia, Sezione di Bologna, Viale Berti Pichat 6/2, 40127 Bologna, Italy, email: [email protected]\u003c/p\u003e\n\u003cp\u003eRemy Bossu\u003c/p\u003e\n\u003cp\u003eEuropean-Mediterranean Seismological Centre, c/o CEA, 91297 Arpajon, Cedex, France.\u003c/p\u003e\n\u003cp\u003eCEA, DAM, DIF, F-91297 Arpajon, France, email: [email protected]\u003c/p\u003e\n\u003cp\u003eMatthieu Land\u0026egrave;s\u003c/p\u003e\n\u003cp\u003eEuropean-Mediterranean Seismological Centre c/o CEA, 91297 Arpajon Cedex, France, email: [email protected]\u003c/p\u003e\n\u003cp\u003ePaolo Gasperini\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDipartimento di Fisica e Astronomia, Università di Bologna, Viale Berti Pichat 6/2, 40127 Bologna, Italy, email: [email protected]\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Competing Interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors acknowledge there are no conflicts of interest recorded.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBossu, R., M. Laurin, G. Mazet-Roux, F. Roussel, and R. Steed (2015). The importance of smartphones as public earthquake-information tools and tools for the rapid engagement with eyewitnesses: A case study of the 2015 Nepal earthquake sequence, Seismol. Res. Lett. 86, 6, 1587\u0026ndash;1592\u003c/li\u003e\n\u003cli\u003eBossu, R., M. Land\u0026egrave;s, F. Roussel, R. Steed, G. Mazet-Roux, S. S. Martin, and S. Hough (2017). Thumbnail-based questionnaires for the rapid and efficient collection of macroseismic data from global earthquakes, Seismol. Res. Lett. 88, 1, 72-81.\u003c/li\u003e\n\u003cli\u003eBossu, R., F. Roussel, L. Fallou, M. Land\u0026egrave;s, R. Steed, G. Mazet-Roux, A. Dupont, L. Frobert, and L. Petersen (2018). LastQuake: From rapid information to global seismic risk reduction, Int. J. Disast. Risk Reduc. 28, 32-42.\u003c/li\u003e\n\u003cli\u003eBossu, R., F. Finazzi, R. Steed, L. Fallou, and I. Bond\u0026aacute;r (2021). \u0026ldquo;Shaking in 5 Seconds!\u0026rdquo;-Performance and User Appreciation Assessment of the Earthquake Network Smartphone-Based Public Earthquake Early Warning System, Seismol. Res. Lett. 93, 137\u0026ndash;148, doi: 10.1785/0220210180.\u003c/li\u003e\n\u003cli\u003eBossu, R., M. B\u0026ouml;se, R. Steed, and D. J. Wald (2024). The Potential of Crowdsourced Data for the Rapid Impact Assessment of Large Earthquakes: The 2023 M 7.8 Kahramanmaraş-Pazarcık, Türkiye,\u003c/li\u003e\n\u003cli\u003eEarthquake, Seismol. Res. Lett. 95, 2058\u0026ndash;2070, doi: 10.1785/0220230421.\u003c/li\u003e\n\u003cli\u003eCastellaro, S., F. Mulargia, \u0026amp; Y. Y. Kagan (2006). Regression problems for magnitudes, Geophys. J. Int. 165, 913-930, doi: 10.1111/j.1365-246X.2006.02955.x.\u003c/li\u003e\n\u003cli\u003eDewey, J., D. Wald, and L. Dengler (2000). Relating conventional USGS modified Mercalli intensities to intensities assigned with data collected via the Internet, Seismol. Res. Lett. 71, 264.\u003c/li\u003e\n\u003cli\u003eEster M., H.P. Kriegel, J. Sander, and X. Xiaowei (1996). A Density-Based Algorithm for Discovering Clusters in Large Spatial Databases with Noise. KDD\u0026apos;96: Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, 226-231\u003c/li\u003e\n\u003cli\u003eFuller, W. A. (1987). Measurement Error Models, John Wiley, New York, 440 pp, doi: 10.1002/jae.3950030407.\u003c/li\u003e\n\u003cli\u003eGasperini, P., F. Bernardini, G. Valensise, and E. Boschi (1999). Defining seismogenic sources from historical earthquake felt reports. Bull. Seismol. Soc. Am. 89, 94-110.\u003c/li\u003e\n\u003cli\u003eGasperini, P., G. Vannucci, D. Tripone, and E. Boschi (2010). The location and sizing of historical earthquakes using the attenuation of macroseismic intensity with distance. Bull. Seismol. Soc. Am. 100, 2035-2066, doi: /10.1785/0120090330.\u003c/li\u003e\n\u003cli\u003eGasperini P., Lolli B., Vannucci G., and E. Boschi, A comparison of moment magnitude estimates for the European-Mediterranean and Italian regions, Geophysical Journal International, Volume 190, Issue 3, September 2012, Pages 1733\u0026ndash;1745, https://doi.org/10.1111/j.1365-246X.2012.05575.x\u003c/li\u003e\n\u003cli\u003eGoded T., N. Horspool, S. Canessa, A. Lewis, K. Geraghty, A. Jeffrey, and M. Gerstenberger (2018), New macroseismic intensity assessment method for New Zealand web questionnaires, Seismol. Res. Lett., 89(2A), 640-652, doi.org/10.1785/0220170163.\u003c/li\u003e\n\u003cli\u003eGr\u0026uuml;nthal, G. (ed.) (1998). European macroseismic scale 1998, Conseil de l\u0026rsquo;Europe. Cahiers du Centre Europ\u0026eacute;en de G\u0026eacute;odynamique et de S\u0026eacute;ismologie,13, Luxembourg, 99 pp.\u003c/li\u003e\n\u003cli\u003eLolli B., P. Gasperini, and G. Vannucci (2014). Empirical conversion between teleseismic magnitudes (mb and Ms) and moment magnitude (Mw) at the Global, Euro-Mediterranean and Italian scale, Geophys. J. Int., 199, 805-828, doi: 10.1093/gji/ggu264\u003c/li\u003e\n\u003cli\u003eLolli, B., D. Randazzo, G. Vannucci, and P. Gasperini (2020). The Homogenized Instrumental Seismic Catalog (HORUS) of Italy from 1960 to Present, Seismol. Res. Lett. 91, 3208\u0026ndash;3222, doi: 10.1785/0220200148.\u003c/li\u003e\n\u003cli\u003ePasolini, C., D. Albarello, P. Gasperini, V. D\u0026apos;Amico, and B. Lolli (2008). The attenuation of seismic intensity in Italy part II: modeling and validation. Bull. Seismol. Soc. Am. 98, 692-708. doi.org/10.1785/0120070021.\u003c/li\u003e\n\u003cli\u003eRovida, A., M. Locati, R. Camassi, B. Lolli, and P. Gasperini (2020). The Italian earthquake catalogue CPTI15, Bull. Earthq. Eng., 18, 2953-2984, doi: 10.1007/s10518-020-00818-y.\u003c/li\u003e\n\u003cli\u003eTosi, P., P. Sbarra, V. De Rubeis, and C. Ferrari (2015), Macroseismic intensity assessment method for web-questionnaires. Seism. Res. Lett, 86, 985-990, doi: 10.1785/022014022.\u003c/li\u003e\n\u003cli\u003eVannucci G., D. Tripone, P. Gasperini, G. Ferrari, and B. Lolli (2015). Automated assessment of macroseismic intensity from written sources using the Fuzzy sets. Bulletin of Earthquake Engineering, 13, 2769-2803, doi:10.1007/s10518-015-9759-5\u003c/li\u003e\n\u003cli\u003eVannucci G., P. Gasperini, B. Lolli, and L. Gulia (2019). Fast characterization of sources of recent Italian earthquakes from macroseismic intensities. Tectonophysics 750, 70-92, doi: 10.1016/j.tecto.2018.11.002\u003c/li\u003e\n\u003cli\u003eVannucci, G., P. Gasperini, L. Gulia, and B. Lolli (2024). Earthquakes Parameters from Citizen Testimonies: A Retrospective Analysis of EMSC Database, Seismol. Res. Lett. 95, 969\u0026ndash;996, doi: 10.1785/0220230245.\u003c/li\u003e\n\u003cli\u003eWald, D. J., V. Quitoriano, L. Dengler, and J. W. Dewey (1999). Utilization of the Internet for rapid community intensity maps, Seismol. Res. Lett. 70, 87\u0026ndash;102.\u003c/li\u003e\n\u003cli\u003eWald, D.J., V. Quitoriano, B. Worden, M. Hopper, and J.W. Dewey (2011). USGS \u0026ldquo;Did You Feel It?\u0026rdquo; Internet-based macroseismic intensity maps. Annals Geophys., 54, 6; doi: 10.4401/ag-5354\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bulletin-of-earthquake-engineering","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"beee","sideBox":"Learn more about [Bulletin of Earthquake Engineering](https://www.springer.com/journal/10518)","snPcode":"10518","submissionUrl":"https://submission.nature.com/new-submission/10518/3","title":"Bulletin of Earthquake Engineering","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7071982/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7071982/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn a recent work, we tested the ability to compute macroseismic parameters (location and magnitude) using citizen testimonies collected by the European-Mediterranean Seismological Centre (EMSC). Each intensity estimated by individual non-professional users of the LastQuake smartphone application is indicated as an individual data point (IDP). Each IDP is archived by EMSC with a time stamp, allowing the calculation of the time delay from the earthquake origin time. To use IDPs as classic intensities, i.e. macroseismic data points (MDPs), identifying damage at the scale of towns or cities, they must be grouped into spatial clusters, which are then processed by the BOXER code to locate and size earthquakes. A retrospective analysis on a dataset of more than 15000 events collected over the past 10 years shows that the procedure can provide reliable parameters and that the results depend on the geographical area and improve over time and as the number of available IDPs/MDPs increases.\u003c/p\u003e\u003cp\u003eThe key question is whether early IDPs/MDPs can quickly provide reliable parameters (location and magnitude) for users and stakeholders (e.g. the civil protection agencies). Using clustering methods that statistically provide, on average, the best agreement with instrumental data, we tested some predefined time intervals within which to group the available IDPs into MDPs. We then applied the BOXER code to these MDPs, evaluating the agreement with the final instrumental parameters. Results confirm that reliability increases with the number and distribution of MDPs, strictly dependent on the number and distribution of available IDPs. This retrospective analysis demonstrates the effectiveness of the approach and its potential to quickly provide parameters for future real-time applications. The method may offer a reliable and rapid tool to support emergency response, improving as more IDPs/MDPs are collected.\u003c/p\u003e","manuscriptTitle":"Can we quickly calculate reliable earthquake parameters from citizen testimonies?","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-15 16:54:38","doi":"10.21203/rs.3.rs-7071982/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"","date":"2025-07-14T18:34:16+00:00","index":0,"fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-11T19:20:14+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"Bulletin of Earthquake Engineering","date":"2025-07-10T09:56:16+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-10T04:55:38+00:00","index":"","fulltext":""},{"type":"submitted","content":"Bulletin of Earthquake Engineering","date":"2025-07-08T03:40:04+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bulletin-of-earthquake-engineering","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"beee","sideBox":"Learn more about [Bulletin of Earthquake Engineering](https://www.springer.com/journal/10518)","snPcode":"10518","submissionUrl":"https://submission.nature.com/new-submission/10518/3","title":"Bulletin of Earthquake Engineering","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"9450fdb2-0895-406a-8bcf-8915b7b7505a","owner":[],"postedDate":"July 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2025-10-13T16:08:54+00:00","versionOfRecord":{"articleIdentity":"rs-7071982","link":"https://doi.org/10.1007/s10518-025-02299-3","journal":{"identity":"bulletin-of-earthquake-engineering","isVorOnly":false,"title":"Bulletin of Earthquake Engineering"},"publishedOn":"2025-10-08 15:57:41","publishedOnDateReadable":"October 8th, 2025"},"versionCreatedAt":"2025-07-15 16:54:38","video":"","vorDoi":"10.1007/s10518-025-02299-3","vorDoiUrl":"https://doi.org/10.1007/s10518-025-02299-3","workflowStages":[]},"version":"v1","identity":"rs-7071982","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7071982","identity":"rs-7071982","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}