amichelini
commented
9 months ago Dear Robert, Thank you for the evidencing the problem. We will verify better what you have sent to us and let you know. Please keep in mind that the presence of more than one earthquake in the 120 s window cannot be excluded although we tried to avoid it. In our experience and after all the verifications we made this should be a very isolated case. More to come in the next days about the sample # that you mentioned. Kind regards Alberto
Alberto Michelini Istituto Nazionale di Geofisica e Vulcanologia (INGV) Via di Vigna Murata, 605 00143 ROMA, Italy Ph. +39 06 51860611, e-mail: @.*** Skype: amichelini https://orcid.org/0000-0001-6716-8551
On Wed, 20 Dec 2023 at 03:24, Robert Pickle @.***> wrote:
Hi there, I have a few questions/comments about the data in the events_counts collection. For the first few examples I have examined, it seems like there are multiple events in each 2 minute segment, but only 1 pair of P/S picks in the metadata? This will confuse the ML trainer, so I am wondering how typical this is and if these multiple event instances are labelled or otherwise able to be filtered out.
For example, looking at trace_name 10000541.IV.CAMP..HH
10000541.IV.CAMP.HH.png (view on web) https://github.com/INGV/instance/assets/22922540/ffdd8d77-a984-44f1-b18e-95e8cab655a1
The corresponding metadata is:
source_id,station_network_code,station_code,station_location_code,station_channels,station_latitude_deg,station_longitude_deg,station_elevation_m,station_vs_30_mps,station_vs_30_detail,source_origin_time,source_latitude_deg,source_longitude_deg,source_depth_km,source_origin_uncertainty_s,source_latitude_uncertainty_deg,source_longitude_uncertainty_deg,source_depth_uncertainty_km,source_stderror_s,source_gap_deg,source_horizontal_uncertainty_km,source_magnitude,source_magnitude_type,source_mt_eval_mode,source_mt_status,source_mt_scalar_moment,source_mechanism_strike_dip_rake,source_mechanism_moment_tensor,path_travel_time_P_s,path_travel_time_S_s,path_residual_P_s,path_residual_S_s,path_ep_distance_km,path_hyp_distance_km,path_azimuth_deg,path_backazimuth_deg,path_weight_phase_location_P,path_weight_phase_location_S,trace_start_time,trace_dt_s,trace_npts,trace_eval_P,trace_P_uncertainty_s,trace_P_arrival_time,trace_polarity,trace_eval_S,trace_S_uncertainty_s,trace_S_arrival_time,trace_P_arrival_sample,trace_S_arrival_sample,trace_E_median_counts,trace_N_median_counts,trace_Z_median_counts,trace_E_mean_counts,trace_N_mean_counts,trace_Z_mean_counts,trace_E_min_counts,trace_N_min_counts,trace_Z_min_counts,trace_E_max_counts,trace_N_max_counts,trace_Z_max_counts,trace_E_rms_counts,trace_N_rms_counts,trace_Z_rms_counts,trace_E_lower_quartile_counts,trace_N_lower_quartile_counts,trace_Z_lower_quartile_counts,trace_E_upper_quartile_counts,trace_N_upper_quartile_counts,trace_Z_upper_quartile_counts,trace_E_spikes,trace_N_spikes,trace_Z_spikes,trace_E_snr_db,trace_N_snr_db,trace_Z_snr_db,trace_E_pga_cmps2,trace_E_pgv_cmps,trace_E_pga_perc,trace_E_pga_time,trace_E_pgv_time,trace_E_sa03_cmps2,trace_E_sa10_cmps2,trace_E_sa30_cmps2,trace_N_pga_cmps2,trace_N_pgv_cmps,trace_N_pga_perc,trace_N_pga_time,trace_N_pgv_time,trace_N_sa03_cmps2,trace_N_sa10_cmps2,trace_N_sa30_cmps2,trace_Z_pga_cmps2,trace_Z_pgv_cmps,trace_Z_pga_perc,trace_Z_pga_time,trace_Z_pgv_time,trace_Z_sa03_cmps2,trace_Z_sa10_cmps2,trace_Z_sa30_cmps2,trace_pga_cmps2,trace_pgv_cmps,trace_pga_perc,trace_sa03_cmps2,trace_sa10_cmps2,trace_sa30_cmps2,trace_name,trace_GPD_P_number,trace_GPD_S_number,trace_EQT_number_detections,trace_EQT_P_number,trace_EQT_S_number,trace_deconvolved_units,source_type 10000541,IV,CAMP,,HH,42.53578,13.409,1283.0,517.0,Vs30 extracted from ShakeMap,2016-11-16T01:35:56.21Z,42.7592,13.1932,10.3,0.03,0.0018,0.0024,0.2,0.19,44.0,0.24,2.1,ML,,,,,,6.19,10.53,-0.02,-0.22,30.482,32.175,144.4,324.6,99.0,71.0,2016-11-16T01:35:45.14Z,0.01,12000,manual,0.1,2016-11-16T01:36:02.40Z,positive,manual,0.3,2016-11-16T01:36:06.74Z,1726,2160.0,3.0,-4.0,2.0,-0.02208,0.02483,-0.01783,-1880.0,-893.0,-1259.0,1850.0,1516.0,1294.0,104.27,97.017,85.736,-42.0,-51.0,-33.0,47.0,43.0,33.0,0.0,0.0,0.0,17.329,16.827,20.449,0.02146025,0.00045726,0.00218834,2016-11-16T01:36:07.190200Z,2016-11-16T01:36:07.230200Z,0.0017980065,7.41901e-05,2.27114e-05,0.0142861,0.00035812,0.00145678,2016-11-16T01:36:07.140100Z,2016-11-16T01:36:07.170100Z,0.0009974314,8.14522e-05,1.68142e-05,0.01503256,0.00029328,0.00153289,2016-11-16T01:36:07.180100Z,2016-11-16T01:36:07.580100Z,0.0007128434,5.55609e-05,1.60743e-05,0.02146025,0.00045726,0.00218834,0.0017980064,8.14522e-05,2.27114e-05,10000541.IV.CAMP..HH,1.0,4.0,1.0,1.0,1.0,mps,earthquake
all trace_EQT_P_number and trace_EQT_S_number and trace_EQT_number_detections parameters are 1.0
Another observation (which possibly explains the above) is that the P/S pick times in the CSV don't seem to correspond to the waveform data. In the metadata above it looks like there is only 1 P and 1 S pick at sample numbers 1726 & 2160.0, however from the image I don't think those correspond to any of the three events. The P and S times (2016-11-16T01:36:02.40Z & 2016-11-16T01:36:06.74Z) are consistent with those pick sample numbers given the trace start time of 2016-11-16T01:35:45.14Z, so I am wondering if the data plotted above is correct for this trace or could have been mixed up, or if I am possibly doing something wrong elsewhere?
The files I am using are instance_events_counts.hdf5 (April 28 2021) and metadata_Instance_events.csv (Aug 22 2021). Is it possible to double check this example to see if this is consistent for you as well?
Thanks for your help!
— Reply to this email directly, view it on GitHub https://github.com/INGV/instance/issues/3, or unsubscribe https://github.com/notifications/unsubscribe-auth/AATJILKTOGJQCBLJAICDQF3YKJD4ZAVCNFSM6AAAAABA4DMGMKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGA2DSNZSGI4TCMI . You are receiving this because you are subscribed to this thread.Message ID: @.***>
SpinaCianetti
filefolder
Hi there, I have a few questions/comments about the data in the events_counts collection. For the first few examples I have examined, it seems like there are multiple events in each 2 minute segment, but only 1 pair of P/S picks in the metadata? This will confuse the ML trainer, so I am wondering how typical this is and if these multiple event instances are labelled or otherwise able to be filtered out.
For example, looking at trace_name
10000541.IV.CAMP..HHThe corresponding metadata is:
all
trace_EQT_P_numberandtrace_EQT_S_numberandtrace_EQT_number_detectionsparameters are 1.0Another observation (which possibly explains the above) is that the P/S pick times in the CSV don't seem to correspond to the waveform data. In the metadata above it looks like there is only 1 P and 1 S pick at sample numbers 1726 & 2160.0, however from the image I don't think those correspond to any of the three events. The P and S times (2016-11-16T01:36:02.40Z & 2016-11-16T01:36:06.74Z) are consistent with those pick sample numbers given the trace start time of 2016-11-16T01:35:45.14Z, so I am wondering if the data plotted above is correct for this trace or could have been mixed up, or if I am possibly doing something wrong elsewhere?
The files I am using are
instance_events_counts.hdf5(April 28 2021) andmetadata_Instance_events.csv(Aug 22 2021). Is it possible to double check this example to see if this is consistent for you as well?Thanks for your help!