Introduction

Metview has much functionality for meteorological data types stored in ECMWF's MARS archive, for example GRIB, BUFR and ODB. But not all data comes in these formats. Therefore Metview has facilities to handle various other data types, which we will explore here.

Visualiser Icons

Some formats, such as GRIB, are easy to visualise in Metview - just right-click, Visualise. This

Introduction

Metview has much functionality for meteorological data types stored in ECMWF's MARS archive, for example GRIB, BUFR and ODB. But not all data comes in these formats. Therefore Metview has facilities to handle various other data types, which we will explore here.

XXXX Download data.

Visualiser Icons

Some formats, such as GRIB, are easy to visualise in Metview - just right-click, Visualise. This is because they are quite constrained in their contents and have enough standardised meta-data for a program to understand how they should be plotted. Some other formats, such as netCDF and tables of ASCII data are not easily interpreted for automatic plotting (which variables/columns should be selected and what do they represent?). Metview introduces the concept of the Visualiser icon, which we will use in some of the following examples.

...

NetCDF is a binary file format for storing multi-dimensional arrays of data and has enjoys wide academic usage.

Examining netCDF

Image Modified

Right-click on the supplied netcdffc_12.nc icon and choose examine to see its structure. It consists of multi-dimensional variables, each of which has its own set of attributes; the file also has a set of global attributes.

Visualising netCDF

The actual fields are t2m and d2m - find their long_name attributes to see what these represent.

Visualising netCDF

Create a new NetCDF Visualiser icon. Edit it and drop the netcdf.nc icon into the NetCDF Data field. Set the following parameters:

Save the changes, and visualise this new icon. See how the settings in the visualiser icon correspond to the variable names in the data. Now visualise another field from the same file. Use the supplied shading_20_levels icon on the plots.

Geopoints

Format overview

Geopoints is the ASCII format used by Metview to handle spatially irregular data (e.g. observations). There are a number of variations on the format, but the default one is a 6-column layout. The columns do not have to be aligned, but there must be at least one whitespace character between each entry.

This example shows a geopoints file containing dry bulb temperature at 2m (PARAMETER = 12004).

Variants of the format allow 2-dimensional variables to be stored (e.g. U/V or speed/direction wind components), and another variant stores only lat, lon and value for a more compact file.

Examining geopoints

Examine the supplied geopoints.gpt icon to confirm the contents of the file. The columns are sortable. You may wish to open the file in an external text editor to see exactly what it looks like.

Visualising geopoints

Visualise the icon. The visdef used for geopoints is Symbol Plotting, and its default behaviour is to plot the actual numbers on the map. This can become cluttered, and text rendering can be slow. Drop the supplied symb_colours icon into the Display Window to get a better view of the data.

Computing some statistics in Macro

First, we will print some information about our geopoints data. Create a new Macro icon, type this code and run it:

gp = read('geopoints.gpt')
print('Num points: ', count(gp))
print('Min value: ', minvalue(gp))
print('Max value: ', maxvalue(gp))

Perform a simple data manipulation and return the result to Metview's user interface:

 return gp*100

Save the macro and see its result by right-clicking on its icon and choosing examine or visualise. We could also have put a write() command into the macro to write the result to a geopoints file.

Finding geopoints points within 100km of a given location

As a more complex example, we will combine two functions in order to find the locations of the points within a certain distance of a given location. We will use the same geopoints file as before.

The distance() function returns a new geopoints variable based on its input geopoint, where each point's value has been replaced by the distance of that point from the given location. The description of this function follows:

Choose a location and use this function to compute the distances of the points from it. Assign the result to a variable called distances and return it to the user interface to examine the numbers. The distances are in metres.

Now we will see a boolean operator in action. The expression distances < 100000 will return a new geopoints variable where, for each point, if the input value was less than 100000, the resulting value will be 1; otherwise the resulting value will be zero. So the resulting geopoints will have a collection of ones and zeros. Confirm that this is the case.

The filter() function, from the documentation:

Use this in combination with what you have already done to produce a geopoints variable consisting only of the points within 100km of your chosen location. Plot the result to confirm it.

Saving geopoints data

Geopoints variables can be saved to disk using the write() command:

 write('my_computed_data.gpt', points)

Converting between geopoints and GRIB

? Needed? Already covered in the Processing Data part!  XXXXXX

Extracting netCDF meta-data

Write the following code into a new Macro (you can exclude the comments):

# read the netCDF file and print its list of variables
nc = read("fc_12.nc")
vars = variables(nc)
print(vars)   #  we could also do:  print(variables(nc))

# set the current variable to be t2m and print its attributes
setcurrent(nc, 't2m')
atts = attributes(nc)
print(atts)

Run the macro - this is what it should print:

[longitude,latitude,time,t2m,d2m]
ATTRIBUTES(scale_factor:0.001611,add_offset:254.569370,missing_value:-32767,units:K,long_name:2 metre temperature)

The variables() function returns a list of all the variable names in the netCDF file.

Most of the netCDF functions work on the current variable, set in the setcurrent() function. The attributes() function returns a definition - a set of named members, similar to a Python dictionary, relevant to the current variable. A definition's elements can be accessed either using the 'dot' operator, e.g. atts.units,  or using indexing notation, e.g. atts["units"].

Image Added

Adapt this macro so that it plots the t2m field and gives it a title based on its long_name attribute. Here are some hints:

• drop your NetCDF Visualiser icon into the Macro Editor
• also drop the Contouring icon
• create a Text Plotting icon and enter a random title string, then drop this into the Macro Editor
• replace the value to the right of text_line_1 with the value of the long_name attribute
• add a plot() command which contains the NetCDF Visualiser variable, the Contouring variable and the Text Plotting variable
• feel free to add the units attribute to the title as well, using the ampersand operator & to concatenate parts of the string

Computations on netCDF data

Simple computations between netCDFs can be performed in a similar way to fieldsets, but they are only performed on the current variable for each netCDF. The folder contains two netCDF files: fc_12.nc and fc_36.nc, representing a 12-hour and a 36-hour forecast valid at the same time. Write a small macro which reads both files, sets their current variable to t2m and computes the difference. The last line should look something like:

diff = fc_36 - fc_12

The variable diff is also a netCDF variable - confirm with this line: "print(type(diff))". Its contents should be identical to the first netCDF in the computation (fc_36), but with the values of its t2m variable updated to be the differences between the two fields. Adapt some of the code from the previous exercise to plot the difference field (and use the supplied Contouring icons pos_shade and neg_shade).

Note that in these netCDF files, the data values are scaled in the netCDF file. The actual values for the t2m variable are encoded as 16-bit integers, but they have scale_factor and add_offset attributes which Metview applies by default.

Image Added

We can see this by extracting the values. Try the following macro, which will print the 'real world' values from t2m:

# read the netCDF file
nc = read("fc_12.nc")

# set the current variable to be t2m and print its values
setcurrent(nc, 't2m')
vals = values(nc)
print(vals)
print('max: ', maxvalue(vals))
print('min: ', minvalue(vals))

Now add the following line before the call to values() :

netcdf_auto_scale_values(0) # 1 means 'on', 0 means 'off'

Now the results should look different and will reflect the values as they are packed in the file.

Try something similar with the time variable:

# select time as the current variable and print its values
setcurrent(nc, 'time')
times = values(nc)
print(times)

The result is a list of date variables. These will be explained in more detail in the session Handling Time in Metview.

To get the 'packed' values for this variable, put this line before the call to values() :

netcdf_auto_translate_times(0)  # 1 means 'on', 0 means 'off'

...

ASCII Data

ASCII Table Data

Metview incorporates functionality to read, process and visualise data stored in ASCII table files, including the commonly-used CSV (comma-separated value) format.

...

To plot the data, we need to tell Metview which columns contain the coordinates and which contain the values. Create a new Table Visualiser icon and edit it. Drop the CSV icon into the Table Data field and set the following parameters:

Notice that this icon contains several parameters at the bottom which allow you to read differently-formatted ASCII table files. The question-mark buttons beside the parameter names give brief information on what they mean. The defaults are set up to read a standard CSV file, so we don't need to touch these parameters in this example.

...

Although Metview has some functionality for handling this type of data in Macro, it can do much more with the geopoints format. Therefore, if the data points are in geographic coordinates, one useful exercise is to read one of these files and convert it to geopoints.

...

First, use the values() function to extract arrays of lats, lons and T2m from the CSV data. These will be returned in variables of type vector - this is an in-memory array of double-precision numbers.

vector or list values( table, number )
vector or list values( table, string )

...

Reading/Writing General ASCII Data to/from Disk

...

/from Disk

ASCII files that are not in Geopoints, ASCII Table or Lat/Long Matrix format can be read using the read() function. It will return a list of strings - one string will contain the contents of one line of the file. Look at the supplied text file, params.txt, and see that it contains a list of codes for meteorological parameters:

...

There are many more string functions available. 

Now do the reverse: write this list of parameters into another text file. The new file should look exactly like the original. Here are some hints:

• the write() function always takes a filename as its first argument, and it can take a string as its second argument
• it always overwrites an existing file of the same name, so there exists another function, append() which will add your string to a new line on an existing file
• so you will need to call write() once with the first line of text, and append() once with the list of parameters
• the list of parameters will need to be flattened out into a string with '/' as the separator - this will need to be done in a loop with a string variable initialised to '', and each element added with the & operator

ODB

XXX need some data!

ODB stands for Observational DataBase and is developed at ECMWF to manage very large observational data volumes through the ECMWF IFS/4DVAR-system. The data structure of an ODB database can be seen as a table of variables called columns. Right-click examine the ODB Database icon to see a list of the variables in the data. The Data tab provides access to the actual data itself. ODB data can be filtered using ODB/SQL queries. The supplied ODB Filter icon contains an ODB/SQL query to retrieve certain columns of data. Edit it - note that this pre-prepared icon is using the ODB Database icon. Look at the ODB Query field to get an idea of what data will be filtered. Now close the editor and examine the icon to see the filtered subset of data it has produced.

Extra Work

• an existing file of the same name, so there exists another function, append() which will add your string to a new line on an existing file
• so you will need to call write() once with the first line of text, and append() once with the list of parameters
• the list of parameters will need to be flattened out into a string with '/' as the separator - this will need to be done in a loop with a string variable initialised to '', and each element added with the & operator
• the global variable newline can be used to add a newline character between the lines

ODB

Image Added

ODB stands for Observational DataBase and is developed at ECMWF to manage very large observational data volumes through the ECMWF IFS/4DVAR-system. The data structure of an ODB database can be seen as a table of variables called columns. Right-click examine the ODB Database icon AMSUA.odb to see a list of the variables in the data. The Data tab provides access to the actual data itself. ODB data can be filtered using ODB/SQL queries. The supplied ODB Filter icon contains an ODB/SQL query to retrieve certain columns of data. Edit it - note that this pre-prepared icon is using the AMSUA.odb icon as its data input. Look at the ODB Query field to get an idea of what data will be filtered. Now close the editor and examine the icon to see the filtered subset of data it has produced. The ODB Visualiser icon tb_plot tells Metview which columns of data to use for the visualisation; visualise it and apply the symb_colours icon to obtain a nice plot.

There is a dedicated tutorial for handling ODB data in Metview on the Tutorials page.

Extra Work

NetCDF

Modify your first netCDF macro which plots the t2m variable and make it compute the temperature in degrees Celcius by subtracting 273.15 from it before plotting.

Optimisations to file writing

...

• in fact, it could be done with a single write() function with the help of the built-in global variable newlineif we just build up a string representing the whole file with newline characters between lines
• if writing many many lines, there is another syntax which avoids multiple file open and close operations:

  • Code Block
    fh = file('output.txt') # open a file handle for i = 1 to 100 do write(fh, 'Line ' & i & newline) end for fh = 0 # close the file handle

ODB

Visualise different columns of data in the supplied ODB file.

See if you can write a macro which extracts lat, lon and value columns into vectors and creates a new geopoints variable from the data.