1 | #! /usr/bin/python |
---|
2 | # ====================================================================== |
---|
3 | # *** TOOL diamlr.py *** |
---|
4 | # Postprocessing of intermediate NEMO model output for |
---|
5 | # multiple-linear-regression analysis (diamlr) |
---|
6 | # ====================================================================== |
---|
7 | # History : ! 2019 (S. Mueller) |
---|
8 | # ---------------------------------------------------------------------- |
---|
9 | import sys |
---|
10 | # ---------------------------------------------------------------------- |
---|
11 | # NEMO/TOOLS 4.0 , NEMO Consortium (2019) |
---|
12 | # $Id$ |
---|
13 | # Software governed by the CeCILL license (see ./LICENSE) |
---|
14 | # ---------------------------------------------------------------------- |
---|
15 | |
---|
16 | # ---------------------------------------------------------------------- |
---|
17 | # *** SUBROUTINE get_args *** |
---|
18 | # Parse command line arguments |
---|
19 | # ---------------------------------------------------------------------- |
---|
20 | def get_args(): |
---|
21 | from argparse import ArgumentParser |
---|
22 | import re |
---|
23 | |
---|
24 | # Set up command-line argument parser |
---|
25 | parser = ArgumentParser( |
---|
26 | description='Postprocessing of intermediate NEMO model output'+ |
---|
27 | ' for multiple-linear-regression analysis (diamlr)') |
---|
28 | parser.add_argument('--file_scalar', help= |
---|
29 | 'Filename of scalar intermediate NEMO model'+ |
---|
30 | ' output for multiple-linear-regression'+ |
---|
31 | ' analysis') |
---|
32 | parser.add_argument('--file_grid', help= |
---|
33 | 'Filename of gridded intermediate NEMO model'+ |
---|
34 | ' output for multiple-linear-regression'+ |
---|
35 | ' analysis') |
---|
36 | parser.add_argument('--regressors', nargs='+', required=False, |
---|
37 | help='Optional list of regressors to include'+ |
---|
38 | ' in analysis; if omitted, all available'+ |
---|
39 | ' regressors will be included') |
---|
40 | args = parser.parse_args() |
---|
41 | return args |
---|
42 | |
---|
43 | # ---------------------------------------------------------------------- |
---|
44 | # *** SUBROUTINE main *** |
---|
45 | # Finalisation of multiple-linear-regression analysis |
---|
46 | # ---------------------------------------------------------------------- |
---|
47 | def main(): |
---|
48 | from netCDF4 import Dataset as nc |
---|
49 | import re |
---|
50 | import numpy as np |
---|
51 | from os.path import basename, splitext |
---|
52 | from time import strftime, localtime |
---|
53 | |
---|
54 | # Get command-line arguments |
---|
55 | args = get_args() |
---|
56 | |
---|
57 | # Get filenames/locations of intermdiate diamlr output |
---|
58 | fn_scalar = args.file_scalar |
---|
59 | fn_grid = args.file_grid |
---|
60 | |
---|
61 | # Open regressor-regressor scalar-product data set |
---|
62 | f = nc(fn_scalar, 'r') |
---|
63 | |
---|
64 | # Detect available regressors; reduce list of regressors according |
---|
65 | # to list of selected regressors (if specified) |
---|
66 | regs = {} |
---|
67 | vn_re = re.compile('(diamlr_r[0-9]{3})\.(diamlr_r[0-9]{3})') |
---|
68 | for vn in f.variables.keys(): |
---|
69 | vn_match = vn_re.match(vn) |
---|
70 | if (vn_match): |
---|
71 | reg1 = vn_match.group(1) |
---|
72 | reg2 = vn_match.group(2) |
---|
73 | if not args.regressors or reg1 in args.regressors: |
---|
74 | regs[vn_match.group(1)] = True |
---|
75 | if not args.regressors or reg2 in args.regressors: |
---|
76 | regs[vn_match.group(2)] = True |
---|
77 | regs = regs.keys() |
---|
78 | regs.sort() |
---|
79 | |
---|
80 | print('Compile and invert matrix of regressor-regressor scalar'+ |
---|
81 | 'products ...') |
---|
82 | |
---|
83 | # Set up square matrix, XX, of regressor-regressor scalar products |
---|
84 | xx = np.matrix(np.zeros((len(regs), len(regs)))) |
---|
85 | for i1 in range(len(regs)): |
---|
86 | vn1 = regs[i1] |
---|
87 | for i2 in range(regs.index(vn1)+1): |
---|
88 | vn2 = regs[i2] |
---|
89 | if f.variables[vn1+'.'+vn2]: |
---|
90 | xx_sum = np.sum(f.variables[vn1+'.'+vn2][:]) |
---|
91 | else: |
---|
92 | xx_sum = np.sum(f.variables[vn2+'.'+vn1][:]) |
---|
93 | xx[i1,i2] = xx_sum |
---|
94 | if i1 != i2: |
---|
95 | xx[i2,i1] = xx_sum |
---|
96 | |
---|
97 | # Close regressor-regressor scalar-product data set |
---|
98 | f.close() |
---|
99 | |
---|
100 | # Compute inverse matrix, XX^-1; convert matrix to an array to |
---|
101 | # enable the dot-product computation of XX with a large array below |
---|
102 | ixx = np.array(xx**-1) |
---|
103 | |
---|
104 | print(' ... done') |
---|
105 | |
---|
106 | # Open field-regressor scalar-product data set |
---|
107 | f = nc(fn_grid, 'r') |
---|
108 | |
---|
109 | # Detect analysed fields |
---|
110 | flds = {} |
---|
111 | vn_re = re.compile('(diamlr_f[0-9]{3})\.(diamlr_r[0-9]{3})') |
---|
112 | for vn in f.variables.keys(): |
---|
113 | print vn |
---|
114 | vn_match = vn_re.match(vn) |
---|
115 | if (vn_match): |
---|
116 | if vn_match.group(2) in regs: |
---|
117 | flds[vn_match.group(1)] = True |
---|
118 | flds = flds.keys() |
---|
119 | flds.sort() |
---|
120 | |
---|
121 | # Open and prepare output file, incl. replication of |
---|
122 | # domain-decomposition metadata (if present) from the |
---|
123 | # field-regressor data set |
---|
124 | fn_out = './'+basename(fn_grid) |
---|
125 | if fn_out.find('diamlr') > 0: |
---|
126 | fn_out = fn_out.replace('diamlr', 'diamlr_coeffs') |
---|
127 | else: |
---|
128 | fn_parts = splitext(basename(fn_grid)) |
---|
129 | fn_out = fn_parts[0]+'_diamlr_coeffs'+fn_parts[1] |
---|
130 | nc_out = nc(fn_out, 'w', format='NETCDF4', clobber=False) |
---|
131 | nc_atts = { |
---|
132 | 'name' : fn_out, |
---|
133 | 'description' : 'Multiple-linear-regression analysis output', |
---|
134 | 'title' : 'Multiple-linear-regression analysis output', |
---|
135 | 'timeStamp' : strftime('%Y-%m-%d %H:%M:%S %Z', localtime())} |
---|
136 | for nc_att in f.ncattrs(): |
---|
137 | if nc_att in ['ibegin', 'jbegin', |
---|
138 | 'ni', 'nj'] or nc_att.startswith('DOMAIN'): |
---|
139 | nc_atts[nc_att] = f.getncattr(nc_att) |
---|
140 | nc_out.setncatts(nc_atts) |
---|
141 | for nc_dimname in f.dimensions.keys(): |
---|
142 | nc_dim = f.dimensions[nc_dimname] |
---|
143 | if nc_dim.isunlimited(): |
---|
144 | nc_out.createDimension(nc_dim.name) |
---|
145 | else: |
---|
146 | nc_out.createDimension(nc_dim.name, nc_dim.size) |
---|
147 | |
---|
148 | # Read in fields of scalar products of model diagnostics and |
---|
149 | # regressors and compute the regression coefficients for the current |
---|
150 | # field; add resulting fields to output file |
---|
151 | for fld in flds: |
---|
152 | print('Completing analysis for field '+fld+'...') |
---|
153 | xy = np.array([]) |
---|
154 | for reg in regs: |
---|
155 | if xy.size == 0: |
---|
156 | xy = np.sum(f.variables[fld+'.'+reg][:], |
---|
157 | axis=0)[np.newaxis,:] |
---|
158 | else: |
---|
159 | xy = np.r_[xy, np.sum(f.variables[fld+'.'+reg][:], |
---|
160 | axis=0)[np.newaxis,:]] |
---|
161 | b=np.reshape(np.dot(ixx, np.reshape(xy,(len(xy),xy[0,:].size))), |
---|
162 | (xy.shape)) |
---|
163 | print(' ... done') |
---|
164 | |
---|
165 | for reg in regs: |
---|
166 | nr = regs.index(reg) |
---|
167 | nc_gridvar = f.variables[fld+'.'+reg] |
---|
168 | name = nc_gridvar.name.split('.') |
---|
169 | nc_var = nc_out.createVariable(name[0]+'-'+name[1], |
---|
170 | nc_gridvar.datatype, |
---|
171 | nc_gridvar.dimensions) |
---|
172 | for nc_att in nc_gridvar.ncattrs(): |
---|
173 | if nc_att in ['_FillValue', 'missing_value']: |
---|
174 | nc_var.setncattr(nc_att, nc_gridvar.getncattr( |
---|
175 | nc_att)) |
---|
176 | name = nc_gridvar.getncattr('standard_name').split('.') |
---|
177 | nc_var.setncattr( |
---|
178 | 'standard_name', name[0]+' regressed on '+name[1]) |
---|
179 | nc_var[0,:] = b[nr,:] |
---|
180 | |
---|
181 | # Close output file; close field-regressor scalar-product data set |
---|
182 | nc_out.close() |
---|
183 | f.close() |
---|
184 | |
---|
185 | # ---------------------------------------------------------------------- |
---|
186 | # *** main PROGRAM *** |
---|
187 | # ---------------------------------------------------------------------- |
---|
188 | if __name__ == "__main__": |
---|
189 | |
---|
190 | main() |
---|