Random-Fortran-scripts

Some random fortran scripts for various purposes, with a f2py signature file for compiling to a Python module.

pcmin

Compute the maximum wind speed and minimum central pressure of Tropical Cyclones, adopted from Prof. Emanuel's Fortran 77 code.

To build a python module:

f2py -c pcmin.pyf pcmin.f90

To use in python:

import pcmin
print pcmin.pcmin_mod.pcmin.__doc__

See Emanuel's original Fortran (77) and Matlab versions: ftp://texmex.mit.edu/pub/emanuel/TCMAX/

regiongrow

Region grow algorithm by recursively searching a map from a given seed.

Two subroutines are included in the module:

• regiongrow_real(data,seedy,seedx,val,eps,res,ny,nx,diagonal):

search a map data from a given seed denoted by seedy and seedx, and return 1s where data is in the range [val - eps, val + eps], and 0s otherwise.

• regiongrow_bin(data,seedy,seedx,val,res,ny,nx,diagonal):

binary region grow search. Return 1s where data==val and 0s otherwise.

For both, set diagnoal to 1 to treat diagnal cells (within the 3x3 element) as neighbours, set to 0 to exclude diagnoal cells.

To build a python module:

f2py -c regiongrow.pyf regiongrow.f90

To use in python:

import regiongrow
print regiongrow.regiongrow.regiongrow_real.__doc__
print regiongrow.regiongrow.regiongrow_bin.__doc__

NOTE beware of the differences in the indexing conventions of Python v.s. Fortran: Python indexing starts from 0, while Fortran starts from 1 (by default). So don't forget to increment the seed indices by 1 before calling the functions.

conv2d

Perform 2D convolution and moving averages and assign missings when missing data amount exceeds a given threshold.

Two subroutines are included in the module:

• convolve2d(slab,slabmask,kernel,kernelflag,max_missing,resslab,resmask,hs,ws):

convolve slab with kernel kernel. Missing values in slab are denoted by 1s in slabmask, where 0s denote valid points. On the contrary, 1s in kernelflag denote valid points in the kernel and 0s otherwise. This is to facilitate valid data count (e.g. a Gaussian kernel has near-zero values around the kernel edges where ambiguity may exist whether they should be treated as 0s). max_missing specifies the maximum tolerable ratio of missing values for each integration: E.g. max_missing = 0.5, then among the valid kernel points (kernelflag==1), at least 50 % of values in data has to be valid, otherwise the integration in is assigned a value 0 to reslab, and correspondingly resmask is assigned a value 1 to denote insufficient data. No filling or wrapping is done at the edges, the kernel is cropped when it extends outside data.

This is to complement similar function in scipy.ndimage.convolve(), which assigns a nan for all cells when any missing overlaps with the kernel (even when overlaps with 0s in kernel). The astropy.convolution.convolve() interpolates missings in the data slab, but doesn't allow control over the level of missing tolerance.

• runmean2d(slab,slabmask,kernel,kernelflag,max_missing,resslab,resmask,hs,ws):

perform a 2d moving average of data using weights in a given kernel, and treat missings as in convolve2d().

To build a python module:

f2py -c conv2d.pyf conv2d.f90

To use in python:

import conv2d
print conv2d.conv2d.convolve2d.__doc__
print conv2d.conv2d.runmean2d.__doc__

m_lowess2d

Perform 2D lowess (Locally weighted Scatterplot Smoothing) filtering on 2D data.

Subroutine:

• lowess2d(z,x,y,nx,ny,span_x,span_y,degree,nsteps,zout,wout)

Filter the input array <z> with a polynormial of degree <degree> on local subsets defined by <span_x> and <span_y>. <x> and <y> are the x and y coordinates for the gridded data <z>. The local subset is an ellipse centered at the point in question, and with an x-axis span of <span_x> and a y-axis span of <span_y>.

NOTE that to use this module also requires the mrgrnk and qr_solve codes, link below:

• mrgrnk: http://www.fortran-2000.com/rank/
• qr_solve: https://people.sc.fsu.edu/~jburkardt/f_src/qr_solve/qr_solve.html

To build a python module:

f2py -m m_lowess2d -h m_lowess2d.pyf m_lowess2d.f90
f2py -c m_lowess2d.pyf mrgrnk.f90 qr_solve.f90 m_lowess2d.f90

To use in python:

import m_lowess2d
print m_lowess2d.m_lowess2d.lowess2d.__doc__