[Thread Prev][Thread Next][Index]

Re: [ferret_users] Calculting model agreement



Hello Kwesi,
G'day
If I am understanding correctly you want to see whether your each model agree with the sign of change in observation Right ?

if it is so then you have to compare each model with your observation first by modifying the code of model_agreement. That means compare get the agreement between model A and observation, then model B and observation and so on....

let modelA = if slope_model_A ge 0 then 1 else (-1)
let obs = if slope_obs ge 0 then 1 else (-1)
let agreeA = if abs(modelA) eq abs(obs) then 1
fill agreeA*slope_obs
set w 2 
fill agreeA*slope_model_A

likewise do for all models...

now you have your 8 agreements such as agreeA, agreeB.... and so on.

now you can aggregate and apply previous procedure to see the agreement between agreeA, agreeB..... etc. so see their mutual agreement with observation. 

So in total first you have to check each model agreement with observation and after that you have to check agreement between all those agreements that you did with individual model and observations...

I hope I am clear to you... I may be wrong so might need someone else's suggestions as well....

cheers, Saurabh 

On Thu, Mar 21, 2019 at 7:44 PM Kwesi A. Quagraine <starskykwesi@xxxxxxxxx> wrote:
Hi Ansley

My observation data is spatial data, anom[time,lat,lon] and they are satellite observation data and my model data consists of 8 models (which I have aggregated, as per initial directions given here). 

I will ideally want to check if the models agree to the sign of change of the observation data.

From the idea from Saurabh, that would work if I were to check that the models agree to the sign of change from model mean (which I understand), However, how will I be able to modify the commands from Saurabh, so that my script will compare the the models to the mean of the observation data?.

Hope I’ve expressed myself with some clarity.

Thanks once again.

Kwesi  

On 22 March 2019 at 00:40:44, Ansley C. Manke (ansley.b.manke@xxxxxxxx) wrote:

I'd suggest making plots of the different expressions you are using to see if they're doing what you expect. 

I can't tell what is "shape" of your observations?  Are they timeseries at different stations?

A different idea would be to have a look at the sample* functions. Would it make sense to sample the model data at the times / locations of the observations?

yes? show function sample*


On 3/21/2019 7:36 AM, Kwesi Quagraine wrote:
Hello Saurabh, Hello Ferreters

Thanks for the script and guidelines.

However, I want to know, how I can compare these models (8 in my case), say, precipitation to observation data (chirps), what would be the approach?

Because here, my guess is I'll be comparing that to the observation mean. And how to do that is my frustration.

Regards
Kwesi


On Thu, Mar 21, 2019, 14:47 saurabh rathore <rohitsrb2020@xxxxxxxxx> wrote:
Dear Kwesi
G'day

I tried to use previous suggestion given by Russ, Ryo and other on my data. SO my data consists of 11 CMIP5 models and I want to show atleast 7 or 8 models agreement in sign. So I did it in this way. Slope is my linear trend from 11 cmip5 models. But before applying this method I combined all my slope from all the different models into a single netcdf file (attached jnl script). So that I can do the computation in my desired dimension which is 'm' here. Here 'm' represents the total number of models or ensembles. 

I hope this will help. 


let tsign = if slope ge 0 then 1 else (-1)
let agree80 = if abs(tsign[m=1:11@sum]) ge 8 then 1

fill agree80*slope[m=@ave]

On Wed, Mar 20, 2019 at 8:13 PM Kwesi A. Quagraine <starskykwesi@xxxxxxxxx> wrote:
Hello Ferreters

I am trying to make a plot with model agreement. 

I want to plot model agreement over some observation. Hence, how the models agree to the sign of change. 

Attached is a snippet of my script. I would kindly want to know  if I am doing it right and if not, the best way to approach it.

Thanks for your help in advance.

Kwesi 

------------ 
Try not to become a man of success but rather a man of value- Albert Einstein

Kwesi A. Quagraine
Department of Physics
School of Physical Sciences
College of Agriculture and Natural Sciences
University of Cape Coast
Cape Coast, Ghana

Alt. Email: kwesi@xxxxxxxxxxxxxx 
Office: +27 21 650 3164
Skype: quagraine_cwasi


--


REGARDS

Saurabh Rathore
Research Scholar (PhD.)
Centre For Oceans, Rivers, Atmosphere & Land Science Technology
Indian Institute Of Technology, Kharagpur
contact :- 91- 8345984434
------------ 
Try not to become a man of success but rather a man of value- Albert Einstein

Kwesi A. Quagraine
Department of Physics
School of Physical Sciences
College of Agriculture and Natural Sciences
University of Cape Coast
Cape Coast, Ghana

Alt. Email: kwesi@xxxxxxxxxxxxxx 
Web: http://www.recycleupghana.org/ 
Office: +27 21 650 3164
Skype: quagraine_cwasi


--


REGARDS

Saurabh Rathore
Research Scholar (PhD.)
Centre For Oceans, Rivers, Atmosphere & Land Science Technology
Indian Institute Of Technology, Kharagpur
contact :- 91- 8345984434

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

Privacy Policy | Disclaimer | Accessibility Statement