Hi,
There is a family of Ferret functions, ZAXREPLACE*, which I show here:
yes? show function zaxreplace*
ZAXREPLACE(V,ZVALS,ZAX)
regrid V onto Z axis of ZAX based upon Z values in ZVALS using
linear interpolation
V: variable on native Z axis
ZVALS: Z-value field corresponding to data points of V
ZAX: variable with desired Z (depth) axis points
ZAXREPLACE_AVG(V,ZVALS,ZAX)
regrid V onto Z axis of ZAX based on Z values in ZVALS using
weighted average
V: Variable on native z axis
ZVALS: Destination Z axis values as a fcn of source Z axis
ZAX: Variable with desired z (depth) axis points
...
These allow you to define a Z axis and a relation of the new Z axis to
the existing one, in this case relating pressure to Z level. These
functions do not allow the function relating the pressure and level
values to vary in time (except for the ZAXREPLACE_ZLEVS function, which
does not apply here). We could easily write a new function that would
let the pressure values vary in time.
The particular file you sent however, looks as though the pressures do
not change in time. They certainly could vary, in
profile data like this, but to show how the function can be applied,
let's just proceed. Also there seems to be something
odd going on in the middle of the record, as though values from
mid-depth have been shifted. I'll ignore this for now as well.
Apply the same transformations to PRES that we did with TEMP. After
running the commands from the last message,
yes?
LET/UNITS="`pres,RETURN=units`"/TITLE="`pres,RETURN=title`" \
pressure = pres[GZ=zlev@ASN,
GT=tday@ASN]
yes? shade pressure ! see how the pressure values are constant in time
! Define an output Z axis with units of pressure, based on the data at
the first time step.
yes? define axis/z/units=decibars/title=pressure/depth zpres =
presure[L=1]
! Call zaxreplace to put the temperature data on this new pressure axis,
yes?
let/units="`temperature,return=units`"/title="`temperature,return=title`"
\
temp_on_pres = zaxreplace (temperature, z[gz=zpres], z[gz=zpres])
yes? shade temperature
Please try all of this and
see whether you run into cases where the pressure values do change in
time. If they do, we could set up a new external function based on
ZAXREPLACE where the second argument varies in time. The function would
just loop over time, recomputing the level-to-pressure transformation
at each time.
Chan Joo Jang wrote:
Thanks Ansley,
My problem is now nearly solved based on your guidance, with some more
refinements for z axis.
In fact, for z axis, I would like to have pressure unit, instead of
level unit.
In the example nc file, there is a variable called pres that is actual
depth in decibar.
The problem is that pres varies on time (N_PROF), as well as on level
(N_LEVELS) as follows:
yes? use/order=ZT 2900623_prof.nc
yes? sh grid pres
GRID GJV4
name axis # pts start end
normal X
normal Y
N_LEVELS Z 48 r 1 48
N_PROF T 35 r 1 35
How can I have temp on z-axis with actuall pressure unit, instead of
level unit.
Thanks in advance,
Chan Joo
--------------------------------------- [ Original
Message ] --------------------------------------
Sender : Ansley Manke < Ansley.B.Manke@xxxxxxxx >
To : Chan Joo Jang < cjjang@xxxxxxxxxxx >
CC : oar.pmel.ferret_users@xxxxxxxx
Date : 2007-12-15 08:27:10
S u b j e c t : Re: [ferret_users] change axis
Hi,
Thanks for the example file. Ferret has some tools to deal with this
kind of profile data. As you found, if we just point to the data, it's
on a grid that doesn't make sense:
yes? USE argo2900623_prof.nc
yes? SHOW GRID temp
GRID GAI4
name axis # pts start end
N_LEVELS X 48 r 1 48
N_PROF Y 35 r 1 35
normal Z
normal T
The attributes in the file don't give Ferret enough information to know
that N_LEVELS should be a Z axis and N_PROF should be a T axis. We can
use the /ORDER qualifier to force the correct directions to be used:
yes? CANCEL DATA/ALL
yes? USE/ORDER=ZT
yes? SHOW GRID temp
GRID GAI4
name axis # pts start end
normal X
normal Y
N_LEVELS Z 48 r 1 48
N_PROF T 35 r 1 35
Now, we'd like to use the variable JULD to define a time axis. In
files like this, sometimes the times are not in monotonically
increasing order, and so we would need to sort them. I'll come back to
this. In this file, the times are monotonically increasing, so we can
just use them as is.
! this gives us the units and time origin.
yes? LIST juld.units
"days since 1950-01-01 00:00:00 UTC"
! Define a time axis based on JULD
yes? DEFINE AXIS/T/T0="1-jan-1950:00:00"/UNITS=days tday =
juld
! Also define a new Z axis. This is ocean data, so it should
! be increasing downwards. First define zz, a variable
containing
! the levels, then define an axis from that variable.
yes? LET zz = z[gz=temp]
yes? DEFINE AXIS/Z/UNITS=levels/DEPTH zlev = zz
! Now define a new variable on these corrected axes,
! using the @ASN transformation to just put them onto
! the new axes. Keep the units and title from TEMP.
yes? LET/UNITS="`temp,RETURN=units`"/\
TITLE="`temp,RETURN=title`" \
temperature = temp[GZ=zlev@ASN, GT=tday@ASN]
! Now the variable is on the correct ZT grid,
! For example, we can do these things:
yes? plot/L=1 temperature
yes? shade temperature
If the JULD variable were not monotonic, we would need to SORT the
variable JULD, and use the result to both define the new time axis and
also to reorder TEMP in T before assigning it to the new T axis.
Ansley
Chan Joo Jang wrote:
Hi Ferr users,
I got ARGO float data files from http://www.usgodae.org/cgi-bin/argo_select.pl
.
(One of those nc files is attached for your reference.)
Using some functions such as RESHAPE, I tried to change original axis
(n_prof, n_levels) to (time, depth) of TEMP (temperature), without any
success.
In other words, I want to have TEMP(julian day, pres) instead of
TEMP(n_prof, n_levels). All informations for julian day(juld) and depth
(pres) are in the nc files.
Any ideas or suggestions would be welcome.
oceanboy
=======
PS) Some parts of ncdump output are as follows:
netcdf 2900623_prof {
dimensions:
DATE_TIME = 14 ;
STRING256 = 256 ;
STRING64 = 64 ;
STRING32 = 32 ;
STRING16 = 16 ;
STRING8 = 8 ;
STRING4 = 4 ;
STRING2 = 2 ;
N_PROF = 35 ;
N_PARAM = 3 ;
N_LEVELS = 48 ;
N_CALIB = 1 ;
N_HISTORY = UNLIMITED ; // (0 currently)
variables:
float PRES(N_PROF, N_LEVELS) ;
PRES:long_name = "SEA PRESSURE" ;
PRES:_FillValue = 99999.f ;
PRES:units = "decibar" ;
PRES:valid_min = 0.f ;
PRES:valid_max = 12000.f ;
PRES:comment = "In situ measurement, sea surface = 0" ;
PRES:C_format = "%7.1f" ;
PRES:FORTRAN_format = "F7.1" ;
PRES:resolution = 0.1f ;
float TEMP(N_PROF, N_LEVELS) ;
TEMP:long_name = "SEA TEMPERATURE IN SITU ITS-90 SCALE" ;
TEMP:_FillValue = 99999.f ;
TEMP:units = "degree_Celsius" ;
TEMP:valid_min = -2.f ;
TEMP:valid_max = 40.f ;
TEMP:comment = "In situ measurement" ;
TEMP:C_format = "%9.3f" ;
TEMP:FORTRAN_format = "F9.3" ;
TEMP:resolution = 0.001f ;
------------------------------------------------
Currently at Dep. of Oceanography,
Texas A&M Univ
Chan Joo Jang, Ph.D., researcher
Korea Ocean Research and Development Institute
1270 Sa2-dong, SangRok-gu, Ansan 426-744, Korea
Tel: +82-31-400-6317 Fax: +82-31-408-5827
cjjang@xxxxxxxxxxx,
http://ogcm.kordi.re.kr
------------------------------------------------
------------------------------------------------
Currently at Dep. of Oceanography,
Texas A&M Univ
Chan Joo Jang, Ph.D., researcher
Korea Ocean Research and Development Institute
1270 Sa2-dong, SangRok-gu, Ansan 426-744, Korea
Tel: +82-31-400-6317 Fax: +82-31-408-5827
cjjang@xxxxxxxxxxx,
http://ogcm.kordi.re.kr
------------------------------------------------
|