Run CLM/CESM with new input data

[min@lanl_gov]

Dear CESM experts,
I am trying to learn to run CLM/CESM. I have several questions to ask about running CLM/CESM with new or modified input data (change percentage of wetland and percentage of pft). My goal is to test the effects of wetland percentage change on climate (mainly temperature and heat related variables). But before I run the model, I want to make sure the model set up is correct and my results are scientifically valid.
1) In my understanding, if I use a scientifically validated compset, with the default input data, then no spinup is needed. However, if I use a scientifically validated compset together with modified input data, do I need to spin up the model for the initial condition?
2) If I use modified input data, then must I set CLM_FORCE_COLDSTART=On? If CLM_FORCE_COLDSTART=on, is spinup is needed to stabilize all the variables?
3) If the input data is changed, I know the data "surdata" is changed, but what about "inidata"? Should I also change inidata in some way in response to input data change?

From the user’s guide, I only know how to change the input data, how to do a spinup and how to choose a compset to run the model in a functional way , but it is not very clear to me what I really should do after I change my input data, such as how to choose a compset, whether a spin up is needed and what other settings are needed in order to get scientifically valid model output.

I will really appreciate if you can help me resolve my confusion or make any suggestions to achieve my goal.

Thanks in advance,

Yours,

Min

 

[sacks]

Hi Min,

I am inserting some answers below:

1) In my understanding, if I use a scientifically validated compset, with the default input data, then no spinup is needed. However, if I use a scientifically validated compset together with modified input data, do I need to spin up the model for the initial condition?

Generally speaking, yes. However, you may not have to do a full spin up (thousands of years), depending on (1) how much you modify the input data, and (2) what output variables you're interested in. If you're not interested in carbon and nitrogen, then you can do less spin up, for example.

A good way to see if you need to do more spin up is to plot a time series of your output variables over a few decades. If there are trends, you need to do more spin up.

2) If I use modified input data, then must I set CLM_FORCE_COLDSTART=On? If CLM_FORCE_COLDSTART=on, is spinup is needed to stabilize all the variables?
3) If the input data is changed, I know the data "surdata" is changed, but what about "inidata"? Should I also change inidata in some way in response to input data change?

The best option is generally to use the interpinic program (in the tools directory). This program interpolates initial conditions from one surface dataset to another... this accommodates changes in resolution as well as changes that you're mentioning, in the percent of wetlands, etc.

Briefly: run a single time step of the model with your new surface data set and no initial conditions (e.g., with CLM_FORCE_COLDSTART on). Then run interpinic to interpolate from an existing initial conditions file (find one as close as possible to your desired configuration) to your new configuration. The output file is the CLM restart file from your single time step run. There is more documentation in the CLM Users Guide.

how to choose a compset

First decide whether you want to run just CLM forced by atmospheric observations (an I compset) or a coupled run (F or B compsets).

Then decide what time period you want to run for, and if you want a transient run or roughly fixed conditions. This will determine the year piece of the compset.

Then decide whether you want the carbon-nitrogen model active or if you want to run with satellite-derived vegetation phenology. For your application, satellite phenology may give you what you want and would require less spin-up, but you may decide otherwise if you're interested in carbon dynamics for example. (This determines whether to add CN to the compset name).