Documentation page for compute_thr_multifile.py

Compute 3d THR on IVT. Process multiple data files too large to fit into RAM.

Difference from compute_thr_singlefile.py:

this is for computing THR process on multiple data files that are too large to fit into RAM. Like a moving average, the THR process is less accurate at the ends of the data domain. For data saved into multiple files (usually separated by time), this may create some discontinuity at the ends of each file. To overcome this, 2 files are read in at a time, data are concatenated in time dimension, therefore the transition between these 2 data files is “smooth”. Then the 3rd file is read in to form a smooth transition between the 2nd and the 3rd files. Then the same process rotates on. The outputs are saved one for a year.

Input:

IVT (integrated Vapor Transport) in netCDF format, one file for a year (or a month, depending on your specific organization of data files). The file name is assumed to have a format that there is a field that specifies the year, e.g. “ivt_m1-60_6_%d_cln.nc”. The %d field is replaced by a year, e.g. 2000.

Data are assumed to be in the format: (time, level, latitude, longitude) dimensions (level dimension is optional, if present, should be a singleton axis of length 1).

NOTE: data should have proper time, latitude and longitude axes.

Optional input:

Orographic data providing the surface terrain elevations, that correspond to the IVT data. This is used to perform some extra computations over high terrain regions to enhance the inland penetration of ARs. The mostly affected area is the western coast of North America. Other areas are mostly not affected.

Usage:

Change global parameters in the Globals section to point to the storage location of IVT data, and specificy an output folder to save results.

Specify the latitudinal domain in LAT1, LAT2.

The KERNEL parameter specifies the $t$ and $s$ parameters of the structuring element size. $t$ is in number time steps, $s$ is number of grid cells. See paper for more details, but basically the choices of $t$ and $s$ should correspond to the synoptic temporal and spatial scales.

SHIFT_LON shifts the longitude by a given degree of longitudes, so that the Pacific and Atlantic basins can be centered.

Run the script as:

` python compute_thr_multifile.py `

Author: guangzhi XU (xugzhi1987@gmail.com) Update time: 2020-07-22 10:23:35.