# Re: Auxiliary coordinate variable

```Hi Noman:

Nawajish Noman wrote:
```
```Hi Russ, John:

```
```
*Auxiliary coordinate variable*****

```
We successfully support the following form of the auxiliary coordinate variables where these variables are a function of two dimensions (lon is a function of yc, and xc).
```
float T(lev,yc,xc) ;

T:long_name = "temperature" ;

T:units = "K" ;

T:coordinates = "lon lat" ;

float lon(yc,xc) ;

lon:long_name = "longitude" ;

lon:units = "degrees_east" ;

float lat(yc,xc) ;

lat:long_name = "latitude" ;

lat:units = "degrees_north" ;

```
Recenlty we are working with Jennifer Boehnert and Wei Wang to support WRF output and came across the following form of the auxiliary coordinate variable where XLAT is a function of three dimensions.
```
float T(Time, bottom_top, south_north, west_east) ;

float XLAT(Time, south_north, west_east) ;

float XLONG(Time, south_north, west_east) ;

*Question 1:*

```
Is this a valid form of the auxiliary coordinate variable? I mean can it be a function of more than two dimensions? Theoretically I don’t see any problem with it but is it recommended?
```
First, I should note that WRF is unfortunately not using CF conventions. Like
many NetCDF writers, they simply write the files any way they please, and dont
necessarily document their meaning.

_From CF Conventions's POV, and from our own theory of coordinate systems, XLAT
amd XLONG are valid coordinate variables for T. The only requirement is that
the dimensions they use are a subset of the dimensions of the data variable.

By writing it with a Time dimension, they are implying that the coordinates are
time dependent. If thats the case, then you have no choice but to use all 3
dimensions. Im not convinced that in fact WRF output has time dependent
coordinates, but perhaps sometimes they do.

```
```
*Question****** 2******:*******

```
In ArcGIS 9.2 we let user choose the auxiliary coordinate variables as X-axis and Y-axis. It’s straight forward with two dimensions. But with three dimensions, we need to keep one dimension constant. By default we are changing___ last two_ dimensions and keeping the first dimension (Time) constant to get a series of XLAT and XLONG values. Is this a good assumption?
```
In general it wont be, but for WRF it may be.

The most general thing to do is to 1) identify the lat and lon coordinates
explicitly (dont depend on their order), and 2) read new lat/lon coordinates
for each time step.

```
```
*Question 3:*

float T(lev,yc,xc) ;

```
For gridded points like T, can we assume the coordinate (xc,yc) is the cell center?
```
WRF uses staggered grids. I think most data variables are centered, expecct for
the wind coordinates (U, V, W). To get a definitive answer, you probably need
to ask the WRF group. Thiie support is at

wrfhelp@xxxxxxxx

```
• 2006 messages navigation, sorted by:
• Search the `netcdf-java` archives: