Problem with an assimilation of temperature

Report or discuss software problems and other woes

Moderators: arango, robertson

Post Reply
Message
Author
User avatar
susonic
Posts: 160
Joined: Tue Aug 21, 2007 5:44 pm
Location: Jeju National University
Contact:

Problem with an assimilation of temperature

#1 Post by susonic » Fri Mar 30, 2012 3:28 pm

Dear users,

I had run a model with using 4Dvar data assimilation. Temperature and salinity profile are used to assimilate.

I ran 2 cases, January and August.

Assimilated temperature on January seems pretty reasonable compare with obs data but the problem occurs on August.

The temperature in August increases unrealistically(11th day upto 50 degree).

I am wondering what seems be the problem or how to find out the main cause.

Sorry for the many figures but I just wanted to give enough information. Any comment would be appreciated.

-JH
Attachments
DA.zip
(3.97 MiB) Downloaded 119 times

User avatar
arango
Site Admin
Posts: 1131
Joined: Wed Feb 26, 2003 4:41 pm
Location: IMCS, Rutgers University
Contact:

Re: Problem with an assimilation of temperature

#2 Post by arango » Fri Mar 30, 2012 4:49 pm

Well, 4D-Var is quite complex and requires basic knowledge of the theory behind least-squares estimation to obtain the best linear unbiased estimate (BLUE). All this explained very well and comprehensively in the Moore et al. (2011) three consecutive papers in Progress of Oceanography (pages 34-94).

Things to look at are:

* Processing of observations: Are you including data outliers with unrealistic values?
* Did you selected the correct a-priori hypothesis model (statistics) for your application?
* Are you violating the tangent linear approximation? That is, are you selecting the appropriate assimilation time window that does not violates the tangent linear model approximation?
* Are you selecting the appropriate 4D-Var algorithm: I4D-Var, 4D-PSAS, or R4D-Var? Do you know that these algorithms converge differently according to the selected conjugate gradient solver?
* Is the 4D-Var algorithm converging?
* Did you plotted the minimization cost function?
* Did you selected the correct number of outer and inner loops for your application?
* How close to the observations is your basic state trajectory (null hypothesis)? If your nonlinear basic state is very far from reality, the assimilation is not going to improve that much. There must be a reasonable skill in your non assimilative run to simulate the realistic circulation in your application.

If you don't know the answers to any of the above questions, you have a lot of learning and work ahead of you. We cannot teach you this knowledge by answering questions in this forum. Sorry, but that is not our mandate. We are planning to have a full week 4D-Var tutorials in the future. It takes a lot of time to Andy and I to organize such tutorials. We offered one in the summer of 2010. We put all the tutorial material in wikiROMS.

Post Reply