SeisNoise.jl is designed for fast and easy ambient noise cross-correlation on the CPU and GPU in Julia.
Documentation | Build Status | Coverage | Chat |
---|---|---|---|
You can install the latest version of SeisNoise using the Julia package manager (Press ]
to enter pkg
).
From the Julia command prompt:
julia>]
(@v1.6) pkg> add SeisNoise
Or, equivalently, via the Pkg
API:
julia> import Pkg; Pkg.add("SeisNoise")
We recommend using the latest version of SeisNoise by updating with the Julia package manager:
(@v1.6) pkg> update SeisNoise
- Built upon SeisIO for easy and fast I/O.
- Custom structures for storing Raw Data, Fourier Transforms of data, and cross-correlations
- CPU/GPU compatible functions for cross-correlation.
- Methods for dv/v measurements.
- Coming soon: Dispersion analysis.
Check out the SeisNoise GPU tutorial on NextJournal!
Once you have installed the package you can type using SeisNoise
to start
cross-correlating. SeisNoise uses a functional syntax to implement cross-correlation. For example
using SeisNoise, SeisIO, Plots
fs = 40. # sampling frequency in Hz
freqmin,freqmax = 0.1,0.2 # min and max frequencies
cc_step, cc_len = 450, 1800 # corrleation step and length in S
maxlag = 60. # maximum lag time in correlation
s = "2019-02-03"
t = "2019-02-04"
S1 = get_data("FDSN","CI.SDD..BHZ",src="SCEDC",s=s,t=t)
S2 = get_data("FDSN","CI.PER..BHZ",src="SCEDC",s=s,t=t)
process_raw!(S1,fs)
process_raw!(S2,fs)
R = RawData.([S1,S2],cc_len,cc_step)
detrend!.(R)
taper!.(R)
bandpass!.(R,freqmin,freqmax,zerophase=true)
FFT = rfft.(R)
whiten!.(FFT,freqmin,freqmax)
C = correlate(FFT[1],FFT[2],maxlag)
clean_up!(C,freqmin,freqmax)
abs_max!(C)
plot(C)
will produce this figure:
SeisNoise can process data and compute cross-correlations on the GPU with CUDA. The JuliaGPU suite provides a high-level interface for CUDA programming through the CUDA.jl package. CUDA.jl provides an the CuArray
type for storing data on the GPU. Data in SeisNoise structures (R.x
, F.fft
, and C.corr
fields, for RawData
, FFTData
, and CorrData
, respectively) can move between an Array
on the CPU to a CuArray
on the GPU using the gpu
and cpu
functions, as shown below.
⚠️ Only Nvidia GPUs are suported at the moment. Hold in there for AMD/OpenCL support...
# create raw data and send to GPU
R = RawData(S1, cc_len, cc_step) |> gpu
R.x
72000×188 CUDA.CuArray{Float32,2,Nothing}
# send data back to the CPU
R = R |> cpu
R.x
72000×188 Array{Float32,2}
All basic processing remains the same on the GPU. Here is a complete cross-correlation routine on the GPU:
# send data to GPU
R1 = RawData(S1, cc_len, cc_step) |> gpu
R2 = RawData(S2, cc_len, cc_step) |> gpu
R = [R1,R2]
# preprocess on the GPU
detrend!.(R)
taper!.(R)
bandpass!.(R,freqmin,freqmax,zerophase=true)
# Real FFT on GPU
FFT = rfft.(R)
whiten!.(FFT,freqmin,freqmax)
# compute correlation and send to cpu
C = correlate(FFT[1],FFT[2],maxlag) |> cpu
Processing times for a selection of routines on the GPU with Julia + GPU (white), Julia + CPU (black), and Python (grey). Currently these operations are implemented in SeisNoise on the GPU:
- Preprocessing:
detrend
,demean
,taper
,onebit
,smooth
- Filtering:
bandpass
,bandstop
,lowpass
,highpass
- Fourier Domain:
whiten
,rfft
,irfft
- Cross-correlation:
correlate
,cross-coherence
,deconvolution
- Post-processing:
stack
,filter
s, etc..
If you use SeisNoise in your work, please star the package and cite our work DOI: 10.1785/0220200192:
@article{SeisNoise.jl-2020,
author = {Clements, Timothy and Denolle, Marine A.},
title = {SeisNoise.jl: Ambient Seismic Noise Cross Correlation on the CPU and GPU in Julia},
journal = {Seismological Research Letters},
year = {2020},
month = {09},
issn = {0895-0695},
doi = {10.1785/0220200192},
url = {https://doi.org/10.1785/0220200192},
eprint = {https://pubs.geoscienceworld.org/srl/article-pdf/doi/10.1785/0220200192/5156069/srl-2020192.1.pdf},
}
We welcome folks interested in contributing to SeisNoise. Please open an issue to let us know about bug reports, new methods/code, and or feature requests/usage cases. If you would like to submit a pull request (PR), please include accompanying tests.