[Thread Prev][Thread Next][Index]

Re: [ferret_users] Monthly climatologies repeated along long time axis

Thanks Billy and Ansley for the detailed explanation; it is very helpful.

My monthly data and climatologies were dated mid-month, and the results appear to be as expected.

Thanks again, Best

On Tue, Jul 11, 2017 at 10:58 AM, Ansley C. Manke <ansley.b.manke@xxxxxxxx> wrote:
Hi Paul,

To expand on what Billy is saying here, results from modulo regridding will be affected by how the 'grid cells' of the monthly time axes are defined.  In an extreme example, If your 20-year monthly data has time coordinates that are for instance, at the start of each month,

yes? list/l=1:6/nohead uwnd[x=180,y=0]
 01-JAN-1982 / 1: -2.738
 01-FEB-1982 / 2: -2.331
 01-MAR-1982 / 3: -3.251
 01-APR-1982 / 4: -3.124
 01-MAY-1982 / 5: -1.818
 01-JUN-1982 / 6: -2.007

and the climatology has its months defined with the coordinates centered mid-month, then the regridding of one time axis to the other might not be what you expect.  The regridding is done by a linear interpolation, so the "uwnd" data above would have an interpolated value between Jan 1st and Feb 1st being used to compute the value at mid-January for instance.


On 7/10/2017 9:31 AM, William Kessler wrote:
The only thing I would add would be to overplot the pieces to make sure everything is right. Especially ACL and ACLMOD. The modulo operation can reduce the amplitude of the annual cycle if, for example, ACL values are on the first of the month (as in some data sets, e.g. ERA-Interim) but month_reg is centered on the middle of the month. month_reg is linearly interpolating between values which can reduce peaks.

Similarly, you might get slightly different values from month_reg vs month_irreg for values on the first of the month. It is possible to define an analog for month_irreg with values on the first day of each month in that case. I could dig up an example.

Whenever I regrid as here, I always make a few test plots of the intermediate steps to check. This is so easy to do in Ferret, and is one reason to do the calculation in explicit steps as in my example.


On Jul 10, 2017, at 9:16 AM, Paul Goddard <pgoddard@xxxxxxxxxxxxxxxxx> wrote:

Perfect, thanks Billy


On Sun, Jul 9, 2017 at 6:22 PM, William S. Kessler <william.s.kessler@xxxxxxxx> wrote:
If A is your 20-year time series and ACL is your 12-month climatology:

First, put the climatology on a modulo time axis:

let ACLMOD = ACL[gt=month_reg@mod]

Extend it to the time axis of A:

let ACLRG = ACLMOD[gt=A]

let ADM = A - ACLRG  ! anomalies

Sent from my phone .... Please excuse my brevity .... and my typing.

On Jul 9, 2017, at 4:06 PM, Paul Goddard <pgoddard@xxxxxxxxxxxxxxxxx> wrote:

Hi all,

It is likely that an answer is already on the ferret users archive, if so, just refer me to that page, but I could not find what I was looking for.

I have a monthly climatology, so 12 time steps.  I would like to subtract the climatology off a 20 year monthly data set (of 240 time steps) to end with an anomaly with 240 time steps. How can I regrid the 12 climatologies to a 240 time step axis with the 12 months of the climatology just repeated 20 times.

Thanks in advance,

[Thread Prev][Thread Next][Index]
Contact Us
Dept of Commerce / NOAA / OAR / PMEL / Ferret

Privacy Policy | Disclaimer | Accessibility Statement