IBM Visualization Data Explorer User's Guide

[ Bottom of Page | Previous Page | Next Page | Table of Contents | Partial Table of Contents | Index ]


B.5 netCDF Files: Complex Fields

For data with more complex structure, conventions have been established for netCDF variable attributes, as described in the format below. The notation used corresponds to that of the netCDL "language."

Irregular Arrays

This section describes how to specify netCDF variables for components with irregular values.

Data

To indicate that a netCDF variable contains values corresponding to the data component, it must have the following attribute:

           variable1:field = "fieldname";
Variable1 is the name of the netCDF variable containing data values to be imported. fieldname is the name of the Data Explorer field by which the user refers to the data (for example, "temperature," "pressure," "wind"). If more than one variable is tagged with the same field name, each variable is read into a field, and the fields are collected into a group.

The data are read in as an array of values, one number per grid point. If the data are actually a vector or a matrix at each grid point, use one of the following modifiers:

The non-scalar data are stored in additional dimensions for the variable. For a static three-dimensional 3-vector, the three components are stored in a fourth dimension of size 3.

If the data have both regular connections and regular positions, no other attributes are required. A regular grid is assumed, with the origin at 0.0, and a spacing of 1.0 along each axis. The number of axes will be determined from the number of dimensions in the data array.

Positions

If the locations of the data values in variable1 do not form a regular lattice (with origins at 0.0 and spacings of 1.0), the name of a netCDF variable that contains the position information must be specified as an attribute for variable1.

There are five different types of position specifications: none, completely regular, completely irregular, and two types of partially regular.

Completely irregular is assumed if the following attribute is specified:

    variable1:positions = "variable2";
where variable2 is an array of vectors, one for each grid point, defining its location. The dimensionality of the data space is determined by the number of items in a vector.

Regular positions can be specified with just the origin and spacing between grid points along each axis in compact form. The following attribute is used:

    variable1:positions = "variable2, compact";
where variable2 is the name of a n×2 array containing origin, delta pairs for the spacing and location of positions along each axis. The number of positions along each axis is determined from the shape of variable1.

Positions that can be specified as the product of arrays containing the location of points along each axis can be input in product form. Use the following attribute:

    variable1:positions = "variable2a, product;
                                          variable2b, product;
                                                 .
                                                 .
                                                 .
                                          variable2x, product";
where the variable2's are each the name of an array containing a list of positions along that axis. The number of items in each array must match the length of the corresponding axis in the original variable1 data array.

If any of the axes in an partially regular product array are actually regular, they can be specified in compact form:

    variable1:positions = "variable2a, product, compact;
                                          variable2b, product;
                                            .
                                            .
                                            .
                                          variable2x, product";
where variable2a is the name of an origin, delta array, and the rest are position lists as before.

Connections

If the connections between positions is a regular lattice, no additional attributes are necessary. For 1-D data, connections of "lines" is assumed. 2-D data implies "quads," 3-D data implies "cubes" and for higher dimensions, "hypercubes" is assumed.

If the connections are irregular, use one of the following attributes:

where variable3 is the name of an array containing a vector of point numbers, defining each connection element item. The length of this vector depends on the choice of connections. If the shape is not explicitly specified, tetrahedra are assumed.

Additional Components

If additional component information is present in the file, the following attributes are valid:

   variable1:component = "variable4, componentname, scalar;
                          variable5, componentname, vector;
                          variable6, componentname, matrix";
and
   variable4:attributes = "ref, componentname;
                           dep, componentname";

Series Data

There are three ways to specify the import of datasets that should be treated as series. They are:

Single Variable

When all data values are defined as a single netCDF variable, and the unlimited dimension of the variable is to be interpreted as the series dimension, then use one of the following forms of the field attribute:

All other specifications are the same as for simple fields.

The position and connection information is assumed to be constant for all members of the series. If the positions or connections change for each step of the series, then the variables used for those arrays must also have an unlimited dimension that corresponds one-for-one with the data array.

An example using this method is provided in "Partially Regular Grids and Time Series".

Separate Variables

When there are separate netCDF variables defined for each step in the series, but all variables are in the same file, use the following global attribute tags:

    :seriesxxx = "fieldname;
                  variable1a;
                  variable1b;
                         .
                         .
                         .
                  variable1x";
or
    :seriesxxx = "fieldname;
                  variable1a, float_value;
                  variable1b, float_value;
                         .
                  variable1x, float_value";
where the global tag must have the first 6 characters series. Global tags must be unique, so additional characters can be added to distinguish them.

Each variable1x is the name array containing the data for that step. In the first format, the spacing of the steps is assumed to be 1.0. In the second format, the float_value is the value of each step. All other specifications are the same as for simple fields.

Separate Files

When there are netCDF variables in separate files that make up the steps of a series, use the following global attribute tags:

    :seriesxxx = "fieldname, files;
                  filename1;
                  filename2;
                         .
                         .
                         .
                  filenameN";
or
    :seriesxxx = "fieldname, files;
                  filename1, float_value;
                  filename2, float_value;
                         .
                         .
                         .
                  filenameN, float_value";
where the global tag must have the first 6 characters series. Global tags must be unique, so additional characters can be added to distinguish them.

Each filenameN is the name of the netCDF file that contains the data variables for that step. In the first format, the spacing of the steps is 1.0. In the second format, the float_value is the value of each step. All other specifications are the same as for simple fields.

This format can be used to create short term series within a file, and then have a series of these smaller series.

Examples

This section shows examples of netCDF files in the netCDL description language. See the documentation supplied by UCAR for more information on netCDL and the ncgen and ncdump utilities.

Compact Specifications of Regular Dimensions

This example describes a single two-dimensional scalar field on a latitude-longitude, regular, rectangular grid. The example data are temperature on a one-degree grid with global coverage. Because Data Explorer array objects can be specified compactly, you can use this method to specify a netCDF with regular dimensions. For each dimension, you need to specify its value at the origin and its spacing along the dimension.

In this example, two variable attributes are defined for the netCDF variables. field specifies the rank of the parameter, and positions specifies where the information containing the locations of the data is space is located.

dimensions:
    lon = 360;
    lat = 180;
    naxes = 2;
    ndeltas = 2;
variables:
    float locations(naxes, ndeltas);
    float temperature(lat, lon);
    temperature:field = "temperature, scalar";
    temperature:positions = "locations, regular";
data:
    locations = 89.5, -1.,      // compact specification, origin and
                -179, 1.;       // spacing for lat and lon
    temperature = ...       // Data for temperature

Partially Regular Grids and Time Series

This example describes an ocean circulation model that consists of a time series of four three-dimensional scalars (temp, sali, wata, and conv) and one three-dimensional 3-vector (vel). netCDF typically requires seven variables, all scalars (the vector counting as three scalars). The coordinate system for the velocity vectors corresponds to that of the grid (that is, +u implies north, +v implies east, and +w implies down).

These grids are partially regular in that the time, tlat, and tlon portions (three out of the four dimensions) are all regularly spaced. time is to be mapped to members of a series group. The fourth dimension, tlvl, is irregularly spaced. The compact notation can be used for the regular notation, while the all values along the irregular dimension must be specified; a product is formed from the dimensions.

Here is the specification in netCDL notation:

dimensions:
    time = UNLIMITED;
    tlat = 30;
    tlon = 50;
    tlvl = 30;
    vsize = 3;   // At each grid cell for variable vel, there are
                 // three floats for the u, v, and w components of the
                 // vector field.
    naxes = 3;
    ndeltas = 2;
variables:
    float lat_axis(ndeltas, naxes);
    float lon_axis(ndeltas, naxes);
    float level_axis(tlvl, naxes);
    float temp(time, tlat, tlon, tlvl);
    temp:field = "temperature, scalar, series";
    temp:positions = "lat_axis, product, compact; lon_axis,
                      product, compact; level_axis, product";
    float sali(time, tlat, tlon, tlvl);
    sali:field = "salinity, scalar, series";
    sali:positions = "lat_axis, product, compact; lon_axis,
                      product, compact; level_axis, product";
    float wata(time, tlat, tlon, tlvl);
    wata:field = "water parage, scalar, series";
    wata:positions = "lat_axis, product, compact; lon_axis,
                      product, compact; level_axis, product";
    float conv(time, tlat, tlon, tlvl);
    conv:field = "covective index, scalar, series";
    conv:positions = "lat_axis, product, compact; lon_axis,
                      product, compact; level_axis, product";
    float vel(time, tlat, tlon, tlvl, vsize);
    vel:field = "velocity, vector, series";
    vel:positions = "lat_axis, product, compact; lon_axis,
                     product, compact; level_axis, product";
data:
    lat_axis = -14.667, 0., 0.,
                 0.333, 0., 0.;
    lon_axis = 0.0, -99.8, 0.0,
               0.0, 0.5, 0.0;
    level_axis = 0.0, 0.0, 17.5,
                 0.0, 0.0, 53.425,
                 .
                 :
                 0.0, 0.0, 5374.98;
    temp = ... ;
    sali = ... ;
    wata = ... ;
    conv = ... ;
    vel = ... ;

Irregular Surface

This example is the netCDL description of a netCDF for an irregular surface, that of the classic teapot. It has precomputed normals, which are imported as the "normals" component, in addition to positions and connections.

netcdf teapot8 {    // name of datafile is "teapot8.ncdf"
                    // name of field is "surface"
dimensions:
        pointnums = 2268;
        trinums = 3584;
        axes = 3;
        sides = 3;
variables:
        float locations(pointnums, axes);
        float normalvect(pointnums, axes);
        long tris(trinums, sides);
        float surfacedata(pointnums);
// global attributes:
                :source = "Classic Teapot, data from Turner Whitted";
// specific attributes:
                surfacedata:field = "surface";
                surfacedata:connections = "tris, triangles";
                surfacedata:positions = "locations";
                surfacedata:component = "normalvect, normals, vector";
                normalvect:attributes = "dep, positions";
// This is the start of a large data section
data:
·
}


[ Top of Page | Previous Page | Next Page | Table of Contents | Partial Table of Contents | Index ]
[Data Explorer Documentation | QuickStart Guide | User's Guide | User's Reference | Programmer's Reference | Installation and Configuration Guide ]

[Data Explorer Home Page | Contact Data Explorer | Same document on Data Explorer Home Page ]


[IBM Home Page | Order | Search | Contact IBM | Legal ]