evo-seismic-data-preprocessing

Overview

This skill handles all data I/O from .npz seismic trace files and transforms them into properly formatted ObsPy Stream objects with correct channel ordering (Z, N, E), sampling rate (100 Hz), and data types (float32).

Key Knowledge

• NPZ files contain: data (12000x3), dt (sampling interval), channels (comma-separated like "DPE,DPN,DPZ")
• SeisBench models expect Z, N, E channel order - must reorder from whatever order is in the file
• Channel codes ending in Z=Vertical, N=North, E=East
• Data must be float32 for PyTorch models
• Detrend (demean) before inference
• Resample to 100 Hz if not already at that rate

Usage

import sys
sys.path.insert(0, '/app/environment/skills/evo-seismic-data-preprocessing/scripts')
from utils import load_and_prepare_stream, load_npz_trace, build_obspy_stream, preprocess_stream, identify_channel_order

# Full pipeline
stream, trace_info = load_and_prepare_stream('/root/data/somefile.npz')

# Or step by step
trace_info = load_npz_trace('/root/data/somefile.npz')
z_idx, n_idx, e_idx = identify_channel_order(trace_info['channels'])
stream = build_obspy_stream(trace_info['data'], trace_info['dt'], trace_info['channels'])
stream = preprocess_stream(stream)

Functions

• load_npz_trace(npz_path) - Load npz file, return dict of all fields
• identify_channel_order(channels_str) - Parse channel string, return (z_idx, n_idx, e_idx)
• build_obspy_stream(trace_data, dt, channels_str, ...) - Build ObsPy Stream in ZNE order
• preprocess_stream(stream, target_rate=100.0) - Detrend and resample
• load_and_prepare_stream(npz_path) - Full pipeline convenience function