Hi Olivier, In looking further into this, I find that it is due to the graphics portions of Ferret handling the time coordinates a bit differently for line plots and 2-d color plots. In most of Ferret and PyFerret, everything is done in double precision. In most of the graphics routines however, single precision is used. Line plots handle the time coordinates correctly, so those plots are correct. The color plots are done in a way that loses precision. The time coordinates in this data has time starting in 1918, in seconds since the year 1800, so those data really need double precision in order to be accurate. We have known of this issue, and I am working on a solution to handle time accurately for these cases. In the meantime a workaround would be to use Ferret to define a time axis which needs less precision to work accurately. yes? use "https://vesg.ipsl.upmc.fr/thredds/dodsC/work_thredds/p86mart/FERRET_DEBUG/PL01_1ts_thetao_point.nc" yes? show axis/t/L=1:2 `thetao,return=taxis` !-> sh axis/t/L=1:2 TIME_COUNTER name axis # pts start end TIME_COUNTER TIME 159 r 14-FEB-1918 04:07 19-FEB-1918 02:37 T0 = 01-JAN-1800 00:00:00 Axis span (to cell edges) = 429300 L T TBOX TBOXLO TSTEP (seconds) 1> 14-FEB-1918 04:07:30 2700 14-FEB-1918 03:45:00 3.727484E+09 2> 14-FEB-1918 04:52:30 2700 14-FEB-1918 04:30:00 3.727486E+09 ! Make a new definition for the time axis. Redefine the
axis "time_counter", yes? let t1 = tt[L=1] yes? define axis/t/t0="14-feb-1918:04:07:30"/units=seconds `thetao,return=taxis` = tt - t1 !-> define axis/t/t0="14-feb-1918:04:07:30"/units=seconds TIME_COUNTER = tt - t1 Replacing definition of axis TIME_COUNTER *** NOTE: grid GCM2 used by data set PL01_1ts_thetao_point *** NOTE: Redefinition may alter apparent contents of data set *** NOTE: grid GCM3 used by data set PL01_1ts_thetao_point *** NOTE: Redefinition may alter apparent contents of data set ! We have changed the time coordinates, "tstep" and the time origin, but the ! coordinates still translate to the same dates and times as the axis in the file. yes? show axis/t/L=1:2 `thetao,return=taxis` !-> sh axis/t/L=1:2 TIME_COUNTER name axis # pts start end TIME_COUNTER TIME 159 r 14-FEB-1918 04:07 19-FEB-1918 02:37 T0 = 14-FEB-1918:04:07 Axis span (to cell edges) = 429300 L T TBOX TBOXLO TSTEP (SECONDS) 1> 14-FEB-1918 04:07:00 2700 14-FEB-1918 03:44:30 0 2> 14-FEB-1918 04:52:00 2700 14-FEB-1918 04:29:30 2700 Now continue with the script. The plots drawn in the smaller viewports are still drawn on an axis with entire days at each end, and perhaps that is correct. The axis labeled with hours would be too crowded. I have attached the plots from the script with the above axis redefinition. -Ansley On 12/20/2017 3:59 PM, Ansley C. Manke
wrote:
|
Attachment:
plot1.png
Description: PNG image
Attachment:
plot2.png
Description: PNG image