Skip to content

Latest commit

 

History

History
72 lines (66 loc) · 4.25 KB

STACK.md

File metadata and controls

72 lines (66 loc) · 4.25 KB
Raw

STACK

Use Multi-process and Multi-thread to accelerate STACK with corr-data (CORRData).

############################################
## 1.Set parameters 
############################################
cores = 1
threads = 1
flag = true
precompile = false
...
# Set your own parameters here...
# Then run following code

############################################
# 2.run NoiseCC.STACK
############################################
using Distributed
addprocs(cores, topology = :master_worker)
@everywhere using NoiseCC

NoiseCC.STACK(
    # process and threads parameters
    cores,threads,flag,precompile,
    # absolute path parameters
    CCFDIR,STACKDIR,
    # select time-chunck folder for stack
    select_time,TIMETXT,
    # select station for stack
    select_sta,STATXT,SRTXT,channel_regular,comp,
    # stack time interval
    stack_all,time_interval,min_stack_chunck,
    # stack method
    substack_method,stack_method,auto_stack,cross_stack,
    # filter 
    select_filter,freqmin,freqmax,corners,zerophase,
    # criteria for corr-data selection
    select_corr,median_high,median_low,
)
  • cores: Int64; total number of processes. cores number should be less than "N_SR_pairs" number.
  • threads: Int64; total number of threads. threads number should be less than N piece of data when "stack_all=false". please set to be "1" when "stack_all=true"
  • flag: true of false; print computing time for debugging purpose.
  • precompile: true or false; precompile or not.
  • CCFDIR: dir where CC data are stored.
  • STACKDIR: dir to store STACK data.
  • select_time: true or false; whether specify time-chunck folder path.
  • TIMETXT: time-chunck folder path.
  • select_sta: "sta" or "sr"; "sta" cross-correlate only works for selected net-sta pairs base on STATXT; "sr" cross-correlate only works for specified source and receiver based on SRTXT.
  • STATXT: station infomation including: network-name and station-name.
  • SRTXT: source and receiver information including: source-1 receiver-1 r-2 r-3...\nsource-2 r-1 r-2 r-3...
  • channel_regular: channel(regular expression) of CCF-data; we will use 'glob(comp,CCFDIR)' to select data in 'CCFDIR' path.
  • comp: component name corresponds to "channel_regular"; different from NoiseCC.CC, we will use 'load_corr(file,comp)' to read CCFData in 'CCFDIR' path, and "comp" is the group name of jld2 file(HDF5 in julia).
  • stack_all: true or false; "true" for stack corr-data over all the selected time chunck.
  • time_interval: Int64 (sec); stack every "time_interval" time, and it works when stack_all = false || 中文: 每间隔多少秒叠加一次，要根据corr-data文件之间的时间间隔决定，例如corr-data文件是6个小时一个，那么time_interval=126060，表示每12个小时叠加一次，即需要2个corr文件叠加在一起。该参数工作时，先搜索指定台站对的全部时间的文件，找到其中最小的开始时间与最大的结束时间，利用range函数生成间隔为""的叠加时窗，叠加每个时窗内的corr-data文件，叠加后的文件名称为该时窗的起始与终止时刻。
  • min_stack_chunck: Int64; each stack window needs at least "min_stack_chunck" to be able to stack, and it works for both stack_all = true and stack_all = false. || 中文:每个时间窗内叠加，至少需要 "min_stack_chunck" 个corr-data文件才会叠加,我们在读取corr-data文件后，会进行质量评估，剔除异常文件，导致corr-data文件数量减少。
  • substack_method: "mean", "pws", and "robust"; substack method corresponding with NoiseCC.CC process, but can only choose one type.
  • stack_method: "mean", "pws", and "robust"; stack method, but can only choose one type.
  • auto_stack: true or false; perform auto-correlation or not.
  • cross_stack: true or false; perform cross-correlation or not.
  • select_filter: true or false; whether filter before stacking.
  • freqmin: Float64;
  • freqmax: Float64;
  • corners: Int64; number of corners in Butterworth filter.
  • zerophase: true or false; if true, apply filter twice for no phase shifting.
  • select_corr: true or false; whether to perform quality selecting based on median value of amp of corr-data.
  • median_high: Float64; max median value (default value).
  • median_low: Float64; min median value (default value).