Many users cannot work with NetCDF data directly and need to
extract data for a small region (perhaps a single grid cell) in plain
text (ASCII) format.
This document will walk you through extracting data from
a NetCDF file into ascii format so that you can import it into Excel, or
feed it into a modeling program, or even just look at the raw numbers.
Note: These instructions are for Windows users. If you are
working on a linux, unix, or mac system, or if you have Cygwin
installed, you have many more options. Probably the best is to
install NCO and use the
ncks command. For example:
ncks -s "%f\n" -C -h -d xc,2,2 -d yc,3,3 -v tas datafile.nc
prints all the values for the variable tas at coordinates xc=2,
yc=3 from datafile.nc, one per line (plus some metadata you can grep
away if you don't want it). You can find a long list of useful tools
NetCDF Software page.
Step-by-step instructions for converting NetCDF data to plain text
Step 1: Install the necessary tools
You will need to install two packages, NetCDF and FAN, to extract
NARCCAP data to plain text on a Windows machine. In both cases, all
you need to do is download the zip file, extract the files within, and
place them in
C:\Windows\System32. (You may need
Unidata's NetCDF package includes
you can use to display the metadata from a NetCDF file. This is often
very helpful for understanding how the data is organized inside the
ncdump -h filename will display only the header
information, without any of the actual data.
You can download the pre-built Win32 binary version of the netcdf
library, including ncdump.exe, here: netcdf-3.6.1-beta1-win32dll.zip.
If you need more information about NetCDF on Windows, it can be
found in Unidata's NetCDF
Installation and Porting Guide.
The user-contributed FAN library, for extracting and
manipulating array data from netCDF files, is also available from
Unidata, on the User-Contributed
netCDF Software page. To use FAN on a Windows machine, you'll
want to download a pre-built binary: fan-2.0.3.win32bin.zip.
FAN includes nc2text.exe, which converts netCDF data to plain text.
Extensive documentation can be found in Unidata's Introduction
to FAN Language and Utilities.
Step 2: Figure out which grid cells you need
Data is stored in the NetCDF files in 2D arrays. To pull out data
for a subregion, you need to know the appropriate range of array
One important thing to note is that all the models use projected
coordinate systems. The array dimensions are named "xc" and "yc"
within the file. The grids are not square in lat/lon
coordinates. Therefore, there are also 2D arrays named "lat" and
"lon" that give the latitude and longitude for each grid cell.
If this is confusing, it may help to look at this grid point map, which shows
the locations of all the grid cells for each of the RCMs. (It's a
very high-resolution PDF, so you'll have to zoom in to make out
To find the gridcells you need, just look up the indices on the
following maps. Each PDF file is a map showing the locations of the
grid cell centers for one of the models. Each grid cell has its array
indices printed next to it. These are ridiculously high-resolution
PDFs, so you'll have to zoom in to 1600% or higher to be able to discern the indices. The maps show state boundaries and major bodies of
water and have graticule lines every 1 degree, so it should be
relatively easy to find your region of interest.
Grid Cell Maps:
NOTE: Even if two models have grid cells in the same
location, they won't necessarily have the same array indices! The
models have different domain sizes due to differences in the sponge
zone, and since the array indices are counted from the lower left
corner, two models using the same map projection can still have
different array numbering. Also be sure to sanity-check your numbers
by looking at nearby grid cells. Because the map is at such high
resolution, sometimes the numerals are a little hard to read.
Double-check whether that's a 1 or a 4 you're looking at.
Data in the file is stored in the order [time,yc,xc]. Indices are
given in the same order, so 15,20 on the map is yc=15, xc=20.
Step 3: Extract data from specified cells
Now that you have your array indices, you can extract data from the
First, you need to get a command prompt. If you have a new
Vista-style start menu, just enter
cmd.exe in the search
box; if you have a classic-style start menu, click "Run..." and enter
cmd.exe in the dialog. This will open up the
command-line interface, also known as the "DOS prompt".
Now, change directories in the command-line interface to wherever
your NetCDF data is located. (For example, if you're running Vista and
your data is on the desktop, you'd type
It will probably be helpful to have a look at how the data is
ncdump -h filename and
you should get a page of output that looks something like this. This is
the metadata for your file, telling you everything about how the data
in the file is structured, like the names of the different variables,
plus a lot of ancillary information such as units.
To extract the data, you just need to know the name of the variable and the array indices of the grid cell you want to extract. Give
nc2text the filename, the variable, and the indices as
shown in this example, redirecting the output to a text file using the
nc2text pr_WRFG.nc pr[,yc=3,xc=5] > data.y3.x5.txt
This extracts all the values for variable 'pr' at coordinates 3,5 in
the file 'pr_WRFG.nc' and prints them in plain text, one value per
line, to the file 'data.y3.x5.txt'.
You can extract values for a range of coordinates, but if your
range was, say, 2 cells high and 3 wide, the resulting output would be
blocks of 2x3 numbers, separated by blank lines. You may find this
format useful to view, but it's not very good for importing into programs like
Excel, and you will in many cases be better off working with a
separate file for each gridcell.
Let's say you're interested in climate in the eastern half of
Nebraska. I'm going to run through an example for the point nearest
the tri-state intersection of Nebraska, South Dakota, and
Iowa—right near Sioux City.
Suppose we want to plot temperature versus time using Excel. It's
always a good idea to sanity-check the data to make sure that the
extracted values make numerical sense by doing something like making a
First, install the NetCDF and FAN packages as described in step 1.
Then, download the data file. For this example, I'll use a small
sample file containing only 50 timesteps: tas_WRFG_example.nc
Next, we find the indices of the grid cell of interest. Zooming in
to 2400% magnification on the map, we find that the
nearest gridpoint is slightly south and a little bit west of the point
where the three state lines intersect. It has coordinates 43,67.
Recall that these coordinates will only work for WRFG data; other
models will have different gridcell coordinates for this location.
Now we open a command window and have a look at the headers.
Navigate to the appropriate directory and run
tas_WRFG_example.nc and you should see something that looks
like this. We can
see that the main data variable is named 'tas' and has units of
We can now extract data for our location of interest and save it to
nc2text tas_WRFG_example.nc tas[,yc=43,xc=67] > temp.csv
The first four values are: 253.657, 252.359, 250.731, and 249.935.
The full output is here: temp.csv.
The resulting file can be opened in a spreadsheet program like
Excel. Excel recognizes the .csv extension as a plain-text tabular
data format where each line is a new row and columns are separated by
commas. (The default output from nc2text puts a blank line between
each number; if that's inconvenient, you can Google "excel remove
blank lines" to find lots of macros that will fix it, or just open up
the file in something like Word and get rid of the blank lines with a
Then we'll do something similar for time:
nc2text -n 1 -f %.3f tas_WRFG_example.nc time > time.csv
Here's the full result: time.csv.
"-n 1" flag in the command tells nc2text to put
only one value per line. (We need it for time because it's a
"-f %.3f" ensures that there
are three places after the decimal (otherwise we'd get rounding
errors, because the data are 8x-daily). Time in the NARCCAP data is
stored as hours elapsed since a baseline, specified in the metadata. In our example, it is 1979/01/01 00:00 UTC. So when you open this file in Excel, you can
convert from hours to dates by entering the baseline date into a cell
and adding the time values as offsets to it.
Converting temperature from Kelvin to the more familiar degrees
Fahrenheit and plotting against time, we can see the clear daily
signal of daytime highs and nighttime lows in a range that is
reasonable for early January in this part of the world: