The SAEON Open Data Platform allows for the storage, discoverability and download of multidimensional scientific data. Network Common Data Format (netCDF) is a file format designed for storing multidimensional data in the form of arrays, and is widely used in the atmospheric and oceanographic communities. The netCDF library has been developed by the Unidata program and is freely available. NetCDFs are self-describing, so it is essential that certain conventions are followed when creating these files. SAEON follows the requirements set out by the Climate and Forecast (CF) and Attribute Conventions Dataset Discovery (ACDD) conventions for netCDF metadata and data. NetCDF standardisation also promotes interoperability and allows us to add tools such as THREDDS data services to the data. The main components of a netCDF that should be standardized are dimensions, variables and attributes.

There are certain types of discrete sampling in atmospheric/oceanographic research, like point, time series, trajectory or profile, and each of these feature types have defined netCDF file structures. The US National Oceanographic Data Centre has developed feature type templates and examples, which are accepted widely for netCDF standardisation. Most importantly, each feature type template highlights the relationships set out for netCDF dimensions and variables.

This section contains metadata that pertains to the overall netCDF file. The following table contains global attributes derived from CF and ACDD conventions required by the SAEON ODP for multidimensional data in netCDF format. For additional resources you can refer directly to the official documentations of each convention.

Note: Time information in the global attributes should be formatted as a string in ISO 8601 standard format,“YYYY-MM-DDThh:mm:ssZ” (i.e. year – month – day “T” hour : minute : second “Z”). Where “Z” indicates UTC time zone (with zero offset).

Table 1. List of core global attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

NetCDF dimensions define the shape, grid or coordinate system of a variable. A dimension has both a name and a length. A dimension length is an arbitrary positive integer, except for cases where one dimension, at most, is UNLIMITED and can grow along that dimension. Once the data feature type has been identified, the US National Oceanographic Data Center netCDF templates can be used as a guide for which dimensions to define in your file.

Dimensions can be used to represent physical dimensions, for example, “time” (T), “latitude” (Y), “longitude” (X), or “depth” (Z). A dimension can also be used to index other quantities, for example, station number or model number. It is recommended that dimensions appear in the relative order T, then Z, then Y, then X, where applicable.

Something to keep in mind in terms of naming dimensions and variables is the netCDF coordinate variable. This is when a variable has the exact same name as a dimension and is most commonly used for coordinates such as lat, lon, depth and/or time. This may not necessarily have special meaning to the netCDF library, however visualisation software treats this in a special way.

Variables in a netCDF contain the parameters measured by an instrument. Variables in netCDF files can be one of six types (char, byte, short, int, float or double). Each variable has a data type, name and a shape which is defined by its specified dimensions.

Our guidance does not standardise variable names, for example, a suitable variable name for temperature data could be anything from: Temperature, Temp or TEMP, so long as the variable name sufficiently describes that data which it contains. Keep in mind the naming of netCDF coordinate variables.

Each variable has a corresponding set of attributes which provide information about the data for the end user. The following tables contain variable attributes derived from CF and ACDD conventions required by SAEON ODP for multidimensional data in netCDF format.

Note: The long_name attribute is defined to contain a long descriptive name of the variable which may, for example, be used for labelling plots. The standard_name attribute is the name used to identify the physical quantity and must be taken from the CF standard name table. Parameters with no suitable standard_name should be described using the long_name attribute only.

Time data in a netCDF is represented as an integer representing an interval from some reference time. Time variables must include a clear units attribute, as there is no default value. The units attribute for a time variable is a string value in the format recommended by UDUNITS. Commonly these strings include “days”, “hours”, “minutes” or “seconds” since a specific date, time and time zone (e.g. “seconds since 1970-01-01 00:00:00 UTC”). Use Double data type.

Table 2. List of time variable attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

SAEON ODP uses the WGS84 coordinate reference system in decimal degrees when describing latitude and longitude.

Table 3. List of longitude and latitude variable attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

Depth or height variables should be measured in SI units, most commonly “metres”. Pressure measurements should not be labelled as a depth or height variable, but as a separate parameter.

Table 4. List of depth variable attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

These variables contain the data collected by an instrument or sensor.

Note: For cases where a netCDF contains two sensors measuring the same variable it is important that these are easily differentiated, community best practise indicates one of the variable names is suffixed with “_2” (e.g. “Temp” and “Temp_2”). For guidance on variable naming conventions one can follow community guidance (e.g. IMOS naming toolbox) or contact the SAEON Data Curators.

Table 5. List of geophysical parameter variable attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

If applicable to the platform, quality control flags can be self describing variables that show some sort of assessment to identify possible errors in the data.

Table 6. List of quality control flag variable attributes for SAEON netCDF files

All attributes are MANDATORY to include in your netCDF unless stated: CONDITIONAL= include these fields if the information exists, or RECOMMENDED = it is recommended to include these fields but not mandatory.

• NetCDF library software: ncdump can convert a netCDF binary file to CDL text.
• Panoply and Ncview are great visualisation tools.
• NCO is a command-line toolkit for manipulating netCDF files: ncatted and ncrename are specifically useful for editing metadata on the fly.