difference calculation

[marqh]

Introduced a form_of_difference attribute to indicate in which direction the difference is being calculated. A difference cell method is also added to indicate along which axis the difference has been calculated and over what length in grid cells.

difference_of_standard_orography / (m) (projection_y_coordinate: 969; projection_x_coordinate: 1042)
     Dimension coordinates:
          projection_y_coordinate                              x                             -
          projection_x_coordinate                              -                             x
     Attributes:
          Conventions: CF-1.5
          form_of_difference: forward_difference
     Cell methods:
          difference: projection_y_coordinate (1 grid length)

[marqh]

The attribute cell_methods has a grammar mandated by CF, in:
http://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#cell-methods
The allowed 'method' strings for a single cell method are defined by:
http://cfconventions.org/Data/cf-conventions/cf-conventions-1.7/cf-conventions.html#appendix-cell-methods

• It is recommended that all cell_methods entries are CF compliant.

difference is not an allowed cell_method, so I recommend in this case that a cell_methods attribute is not used at all.
The name used indicates that this is a difference: difference_of_standard_orography

[marqh]

This ticket relates to spatial.py at this commit:
https://github.com/metoppv/improver/blob/c6739f0d50fa8b3e84d47994a1a2e360599958a7/lib/improver/utilities/spatial.py

[marqh]

I am conscious that removing the cell_method has removed the reference to the projection_y_coordinate

I think this is an example of a core capability that we require for many definitions, the definition of a process with respect to a coordinate. I have raised this as a targeted issue: #9

[marqh]

A sample programme that would create a file similar to the above, whilst using enhanced naming:

import iris
import numpy as np
import os
import subprocess

iris.FUTURE.netcdf_no_unlimited = True

# Create a sample data set with unimportant data values.
acube = iris.cube.Cube(np.ones(6).reshape(3, 2), units='m', var_name='diff_so',
                       long_name='difference_of_standard_orography')

xcoord = iris.coords.DimCoord(np.arange(2), var_name='x_coord',
                              standard_name='projection_x_coordinate', units='1')
ycoord = iris.coords.DimCoord(np.arange(3), var_name='y_coord',
                              standard_name='projection_y_coordinate', units='1')


acube.add_dim_coord(xcoord, 1)
acube.add_dim_coord(ycoord, 0)

# Add context specific attributes to the sample data set
acube.attributes['spp__statistical_process'] = 'spv__difference'
acube.attributes['spp__with_respect_to'] = 'y_coord'
acube.attributes['spp__form_of_difference'] = 'spv__forward_difference'
acube.attributes['rdf__type'] = 'spd__statistically_processed_array'


# Create a prefix container and add prefixes to it
prefix_cube = iris.cube.Cube(0, long_name='prefixes', var_name='prefix_list')

prefix_cube.attributes['spp__'] = 'http://reference.metoffice.gov.uk/statistical-process/properties/'
prefix_cube.attributes['spv__'] = 'http://reference.metoffice.gov.uk/statistical-process/values/'
prefix_cube.attributes['spd__'] = 'http://reference.metoffice.gov.uk/statistical-process/def/'
prefix_cube.attributes['bald__'] = 'http://binary-array-ld.net/latest/'
prefix_cube.attributes['rdf__'] = 'http://www.w3.org/1999/02/22-rdf-syntax-ns#'

# place all cubes in a CubeList
cubelist = iris.cube.CubeList([acube, prefix_cube])

# Ensure that the 'bald__isPrefixedBy' attribute is present on all cubes,
# so that it is upgraded to a global attribute on the file.
for cube in cubelist:
    cube.attributes['bald__isPrefixedBy'] = 'prefix_list'

# Save to netCDF, then dump the header to stdout before deleting the file
fname = 'add_attrs.nc'
iris.save(cubelist, fname)
subprocess.check_call(['ncdump', '-h', fname ])
os.remove(fname)

This is provided to illustrate the use of prefixing for newly coined attribute names and values, and the definition of these prefixes within the file, making use of scitools' Iris to create data sets to write into a netCDF file

this programme produces:

netcdf add_attrs {
dimensions:
	y_coord = 3 ;
	x_coord = 2 ;
variables:
	double diff_so(y_coord, x_coord) ;
		diff_so:long_name = "difference_of_standard_orography" ;
		diff_so:units = "m" ;
		diff_so:rdf__type = "spd__statistically_processed_array" ;
		diff_so:spp__form_of_difference = "spv__forward_difference" ;
		diff_so:spp__statistical_process = "spv__difference" ;
		diff_so:spp__with_respect_to = "y_coord" ;
	int64 y_coord(y_coord) ;
		y_coord:axis = "Y" ;
		y_coord:units = "1" ;
		y_coord:standard_name = "projection_y_coordinate" ;
	int64 x_coord(x_coord) ;
		x_coord:axis = "X" ;
		x_coord:units = "1" ;
		x_coord:standard_name = "projection_x_coordinate" ;
	int64 prefix_list ;
		prefix_list:long_name = "prefixes" ;
		prefix_list:bald__ = "http://binary-array-ld.net/latest/" ;
		prefix_list:rdf__ = "http://www.w3.org/1999/02/22-rdf-syntax-ns#" ;
		prefix_list:spd__ = "http://reference.metoffice.gov.uk/statistical-process/def/" ;
		prefix_list:spp__ = "http://reference.metoffice.gov.uk/statistical-process/properties/" ;
		prefix_list:spv__ = "http://reference.metoffice.gov.uk/statistical-process/values/" ;

// global attributes:
		:bald__isPrefixedBy = "prefix_list" ;
		:Conventions = "CF-1.5" ;
}