|
Hi- I'm replying here to the List too, as what you ask is a general question that does come up. On 3/22/2011 10:50 AM, Otacilio Leandro wrote:
There is a bit about this topic in the Users Guide, under the term "timestamp". What happens is that regridding is only for interpolating (or otherwise changing) from one axis to another axis in the same direction. So you can change from one time axis to another time axis using a regridding operation. But if there is initially no time axis, regridding has no effect. Instead we use the idea of "conformability". There's an index entry to that topic in the Users Guide. For your example, we'd ADD a variable on the time axis, whose value is zero, to the variable with no time axis. Your time-axis definition is fine, as a single-step abstract axis. You could do something similar with a calendar axis, For your example, it would look something like this:define axis/t="1-jan-2010":"1-jan-2010":1/units=days axist
Thanks again. 2011/3/21 Ansley Manke <ansley.b.manke@xxxxxxxx>:Hi - There was a typo in my previous message. There should not be an equals sign after the file in the third line. yes? define axis/x=1:90000:1 xindex yes? define grid/x=xindex xinput yes? file/skip=10000/FORM=free/grid=xinput/var="lon, lat, ..." Then make the SCAT2GRIDGAUSS_XY calls with arguments (lon, lat, var, x[gx=xlon], y[gy=ylat], ...) On 3/21/2011 12:07 PM, Ansley Manke wrote: Hi Leandro So, the input data don't include each and every longitude and latitude on the grid you have defined with xlon and ylat. This means that you need to read the data on a single, long axis, and then call the regridding function as you have done. The FILE command you used with /grid=mygrid assumes that all the locations are represented in your input file. So, just define a long axis. If there were fewer than the default abstract-axis length of 20480 points, you could do this without defining an axis at all (you'd just skip the /grid= qualifier), but you need a longer axis. Use the length of the dataset, or some number longer than its length. yes? define axis/x=1:90000:1 xindex yes? define grid/x=xindex xinput yes? file=/skip=10000/FORM=free/grid=xinput/var="lon, lat, ..." Then make the SCAT2GRIDGAUSS_XY calls with arguments (lon, lat, var, x[gx=xlon], y[gy=ylat], ...) On 3/21/2011 11:06 AM, Otacilio Leandro wrote: Hi Ansley, The data that i'm using is at a grid of 0.5 degree only over continental areas. Because of that i'm using the scat2gridgauss function. The data also is organized with lon and lat at the first and second columns and the months january until december at the columns 3-15. The last column is the anual average for the year. I'm seeding you the year of 1900 as exemple. My goal with this data is to create only one cdf file with all monthly means of surface temperature since 1900 until 2008. At this jnl script i'm trying to create one cdf file per month and then concatenet them all using NCO. If there's a simple way to this using ferret i'll apreciate if you enlighten me. Thank you again and continue the great job you are doing! 2011/3/21 Ansley Manke<Ansley.B.Manke@xxxxxxxx>: hi- I think that we need to find out how to read the data first. Could you send me a few lines from the start of your ascii file? After you run the commands define axis/x=-179.8:179.8:0.5/units=degrees xlon define axis/y=-89.75:89.75:0.5/units=degrees ylat define axis/t=1:1:1 axist define grid/x=xlon/y=ylat/t=axist mygrid file=/g=mygrid/skip=10000/FORM=free/var="lon, lat, var01, var02, var03, var04, var05, var06, var07, var08, var09, var10, var11, var12, anual" "air_temp2009/air_temp.$1" This sequence would indicate that your data is already on an XY grid, at one value of time. How do these plots look? yes? shade/L=1 lon yes? shade/L=1 lat yes? shade var01 If the data are are not already on a on the grid defined by xlon and ylat, then the above FILE command is not the correct one to use. If the data is on a grid, then you don't need to call the scat2gridgauss function. On 3/21/2011 4:36 AM, Otacilio Leandro wrote: Ansley, I've corrected the file line but still the problem continue. I'm atacching the jnl script so you can verify any others erros. Thank you very much. 2011/3/18 Ansley Manke<Ansley.B.Manke@xxxxxxxx>: Hi - In your FILE command, do you have quotes around the list of variable names? It should look like, file=/g=mygrid/var="lon, lat, value01, value02, ..." ascii_file.txt On 3/18/2011 8:29 AM, Otacilio Leandro wrote: Hello, I've got a data with 2D (x and y) climatological monthly temperature over continental areas and i got some trouble at using it on ferret. The data are separated by one file per year and i used the function scat2gridgauss_xy to correctly read the data over the continent. My goal with this data is create a cdf file varing at time (each "timestep" is one month of the data). I defined the axis this way: define axis/x=-179.8:179.8:0.5/units=degrees xlon define axis/y=-89.75:89.75:0.5/units=degrees ylat define axis/t=1:1:1 axist define grid/x=xlon/y=ylat/t=axist mygrid But when i used file=/g=mygrid/var= lon, lat, value01, value02 ... etc , ascii-file.txt all values of the orignal data are exchanged for 12.8 as you can see below. Here is the stat of my variable: temp-1900-02 LONGITUDE: 179.9E(-180.1) to 179.9E(179.9) LATITUDE: 90S to 90N Z: N/A T: 1 DATA SET: air_temp2009/air_temp.1900 Total # of data points: 259200 (720*360*1*1) # flagged as bad data: 259194 Minimum value: 12.8 Maximum value: 12.8 Mean value: 12.8 (unweighted average) Standard deviation: 7.178999E-07 I also tryed to regrid the variable using let temp = tmp[g=mygrid] but it also showed problems because the time axis is missing at temp variable. The way i defined the t axis is correct? The reason that i want a time axis on my variable is to use NCO to concatened all cdfs that im going to create. I tested using file without time axis and got an error from NCO. I'm using: FERRET v6.64 Linux rh5 (gfortran) 2.6.18-164.11.1.el5 - 09/20/10 Thanks for the help. |