adding blank months of data when subsetting dataset temporally

[JohannaWren]

Hi,

I want to subset a dataset of monthly chlorophyll data temporally and save that to a file. Specifically, I'm wanting to save Jan-Mar for each year to a new file. I accomplished this with the following code:

! Load the dataset
use https://oceanwatch.pifsc.noaa.gov/erddap/griddap/esa-cci-chla-monthly-v6-0
! Save Q1 to a new file
REPEAT/i=5:50:12 (let chl = chlor_a[x=180:240:0.5, y=5:45:0.5, l=`i`:`i+2`]; save/append/file=test_chl.nc chl)

This works except that the months that are saved are Jan, Feb, Mar, and Aug of each year, with Aug not having any data. I have tried the same code but using t to index instead (i.e. t=1-jan-`i`:31-mar-`i` ) but with the same result. I made a quick plot showing L=5:8 in the resulting output file so you can see the issue visually.

I'm not sure what I am doing wrong here. Is this a bug or should I be defining a new time axis? Is there a different way to pull out only data from a specific quarter of the year?

Thanks so much for your help,
Johanna

PS. (I'm using PyFerret v7.63, Darwin 19.6.0 - 12/10/21 on my MacOS Ventura 13.7.5)

[ACManke]

Hi Johanna,

Depending on what you want to do using seasonal data, you might consider working within the PyFerret session and defining a seasonal mask, rather than writing out a file with the data you want to use. See this FAQ,
https://ferret.pmel.noaa.gov/Ferret/faq/creating-seasonal-masks

What you are seeing is due to of the nature of coordinate axes in the CF data model and in Ferret/PyFerret. Coordinate axes represented by coordinates locations and grid cell bounds. In PyFerret, coordinate axes are contiguous grid cells, so that the the upper bound of one grid cell is always identical to the lower bound of the next grid cell. In the dataset you're using, time is listed with its coordinates which translate to the first of each month. The dataset upi are isomg does not define bounds for its coordinates. When there are no bounds given, PyFerret creates grid cell boundaries halfway between the coordinates. (A dataset can specify the bounds of axes and the coordinates do not have to be centered in grid cells). Grid cells are of course central to computations throughout Pyferret, used for instance in regridding and in transformations such as averages.

Bounds in netCDF files written by PyFerret are described here, https://ferret.pmel.noaa.gov/Ferret/documentation/users-guide/commands-reference/SAVE#PartII-save-bounds

So to preserve the size of all of the grid cells, a new grid cell containing missing data is written, representing the extra time in each year.