github.com/usgs/prism U.S. Geological Survey Earthquake Science Center Processing and Review Interface for Strong Motion Data (PRISM) Software Version 2.0 Main Features Earthquake Science Center U.S. Geological Survey Menlo Park, California http://earthquake.usgs.gov/ Abstract Erol Kalkan ([email protected]), Jeanne Jones, Christopher Stephens, Peter Ng, Jamie Steidl, Jeff Brody, and Lind Gee PRISM automates the processing of strong-motion records by provid- ing batch-processing capabilities. The software consists of two components: - Platform-independent, modular, extensible and open-source; - Customizable processing parameters with a configuration file; - Phase-time and maximum amplitude picking; - Time-domain mean-removal, integration and differentiation; - Frequency-domain resampling; - Acausal bandpass filtering; - Baseline correction determined from the velocity time series and re- moved from the acceleration time series in time domain; - Filtering performed on the acceleration time series in time domain; - Generation of products that include compatible acceleration, veloc- ity and displacement time series, response spectra, Fourier amplitude spectra, and standard earthquake-engineering intensity measures; - Log files for quality control and reproducibility; - For input, currently uses Consortium of Organizations for Strong- Motion Observation Systems (COSMOS) V0 input format with meta- data in COSMOS headers; and - Products in COSMOS data format (V1, V2 and V3). A graphical user interface (GUI) for manual review, edit and processing. A record processing engine (command line interface) com- posed of modules for each pro- cessing step. 1) Pseudo acceleration, veloc- ity and displacement spectra at 2, 5, 10 and 20% damping, 2) Fourier Amplitude Spec- trum, and 3) IMs: Volume - 3 Processing - Arias Intensity - Bracketed Duration - Duration Interval - Response spec. intensity - RMS Acceleration - CAV Corrected acceleration Calculate FFT Response spectra V3 Process Select spectral amplitudes for given periods Calculate peak response For each damping value For each period Fourier spectrum What’s New in PRISM 2.0 - Frequency-domain integration; - Frequency-domain decimation; - Auto-detecting spikes and removal; - Signal-to-noise ratio (SNR) thresholding; and - Auto-detecting bandpass-filter corner frequencies. Volume - 1 Processing Step - 1: Resampling to 200 sps (if needed) Step - 2: Event onset detection Step - 3: Pre-event mean removal Step - 4: First order baseline cor- rection (velocity domain) Step - 5: QC Step - 6a: Bandpass filter with acausal filter (acc. domain) Step - 6b: Adaptive baseline cor- rection (ABC) Step - 7: Compute acceleration, velocity and displacement.; Step - 8: Decimate to original sampling rate (if needed) Raw Data Trace Convert counts to g Convert counts to cm/sec 2 Uncorrected acceleration cm/sec 2 g V1 Process Units of cm/sec 2 or g Remove mean Despike? Remove spikes Yes No Uncorrected acceleration (V1) Find event onset Remove pre-event mean Integrate to velocity Find best fit trend QC velocity Fail Pass Adaptive baseline correction (ABC) Acausal bandpass filter acceleration Integrate velocity to displacement QC velocity & displacement Able to run ABC No Yes Corrected acceleration Velocity Displacement V1 and V2 to Trouble Folder V1 to Trouble Folder Pass Fail V2 Process Found onset Yes No Integrate acceleration to velocity Differentiate best fit trend Remove best fit trend from acceleration Acceleration with pre- - event mean removed Copy of acceleration with pre-event mean removed Trend to remove from acceleration Integrate acceleration to velocity Volume - 2 Processing PRISM 1.0 vs 2.0 Processing 0 20 40 60 80 100 120 Record No. 0.1 1 10 100 Frequency, Hz 2014 M4.4 SD Palos Earthquake PRISM 1 high-pass PRISM 1 low-pass PRISM 2 high-pass PRISM 2 low-pass 0 10 20 30 40 50 Counts Take-home Message 10 -2 10 0 10 2 Frequency, Hz 10 -10 10 -8 10 -6 10 -4 10 -2 10 0 10 2 Fourier Amplitude Spectrum, cm/s 2 .s NCN016-n.711.HNE | dt1=0.005 s; dt2=0.005 s PRISM 1 (passband = 0.1 to 40 Hz.) PRISM 2 (passband = 0.1 to 25.2 Hz.) -200 0 200 NCN016-n.711.HNE | dt1=0.005 s; dt2=0.005 s Acc., cm/s 2 PRISM 1 PRISM 2 -200 0 200 Acc., cm/s (PRISM 2 - PRISM 1) -50 0 50 Vel., cm/s -50 0 50 Vel., cm/s -20 0 20 Dis., cm 0 20 40 60 80 100 120 140 160 180 Time, s -20 0 20 Dis., cm Comparing PRISM 1.0 and 2.0 results for component HNE of station NCN016 record from the 2014 M6.0 South Napa earthquake. Frequency-domain Integration Frequency-domain Decimation Spike Removal Auto-detect Passband Corners -5 0 5 Station: IU.AFI | Comp: HN1 | 200 sps Acc., cm/s 2 -0.05 0 0.05 Vel., cm/s Frequency-domain int. Trapezoidal int. -5 0 5 10 -3 Difference in PGV = 2.21% Vel., cm/s [Frequency-domain int. - Trapezoidal int.] -5 0 5 10 -3 Dis., cm 0 5 10 15 20 25 30 35 40 Time, s -5 0 5 10 -4 Difference in PGD = 0.01% Dis., cm 0 200 400 600 800 Record No. 0.1 1 10 100 Frequency, Hz 2014 M6.0 South Napa Earthquake PRISM 1 high-pass PRISM 1 low-pass PRISM 2 high-pass PRISM 2 low-pass 0 200 400 600 Counts Comparing PRISM 1.0 and PRISM 2.0 passband corner frequencies for the 2014 M6.0 South Napa and 2014 M4.4 SD Palos earth- quakes. In PRISM 1.0, corner frequencies are retrieved from a look up table based on the earthquake magnitude. In addition to this, PRISM 2.0 also offers auto -detecting corner frequencies. 10 -2 10 0 10 2 Frequency, Hz 10 -20 10 -15 10 -10 10 -5 10 0 Fourier Amplitude Spectrum, cm/s 2 .s High-pass:0.10 Hz & Low-pass:60 Hz Signal Noise Signal smooth Noise smooth Low-pass corner High-pass corner Detects appropriate bandpass filter corner frequencies to be used for Butterworth filtering by using the signal's frequency content. Noise and signal spectra are smoothed using Konno-Ohmachi window. Replaces spikes (outliers) exceeding the dynamic threshold value by in- terpolating among previous and subsequent data points. 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 Original Signal 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 Spike Removed Signal PRISM 2.0 provides greater flexibility and increased processing accuracy for strong motion data. Integrates signal in frequency domain by dividing its Fourier spectrum with -iw. 0 2000 4000 6000 8000 10000 12000 -5 0 5 Acceleration Tonga Record, HNN Original (100 sps) 0 0.5 1 1.5 2 2.5 # 10 4 -5 0 5 Acceleration Re-sampled to 200 sps 0 2000 4000 6000 8000 10000 12000 -5 0 5 Acceleration Decimated to 100 sps 0 2000 4000 6000 8000 10000 12000 Time-step -1 0 1 Difference # 10 -4 Frequency domain decima- tion module reduces the sam- pling rate of a re-sampled signal to its original sampling rate. PRISM 2.0 Note: No filtering or baseline correction applied. Differences between PRISM 1 and PRISM 2 processing Differences between PRISM 1 and PRISM 2 processing Differences between PRISM 1 and PRISM 2 processing Poster Number 154