Closed zhengjiji456 closed 5 years ago
Shall I see the content of the preservedEpc.csv file? It can help me a lot. Does it contain only zeros or NA-s?
please see the uploaded xlsx file preservedEpc.xlsx
Please let me know witch model version, and RBBGCMuso version You use. You can get RBBGCMuso version via this command:
packageVersion("RBGCMuso")
'0.6.1.2'
I think I know why you get NaN.
If you have not specify the output variable with the varIndex parameter (see ?musoSensi
for further information for this parameter). In a result: musoSensi automatically selected the first. In your case, it was tsoil_0 --> The temperature of the first soil layer (0-10 cm). Neither of your selected epc parameter were sensitive to this output variable. The output of the musoSensi is relative sensitivity. 0/0 is not a number, so the result is NaN for every parameter.
Run the following command instead of the previous:
musoSensi(parameters = parameters, monteCarloFile="./preservedEpc.csv", varIndex = 14)
It will do SA for projected LAI (it is the 14th output variable of your ini)
My version is 0.6.1.2. The command worked. I also tried NPP and the result seemed good. Thanks.
I am very glad to see that You are using our package!
I emailed you 3 months ago and reported the same issue after using the former version of this package. I am happy to see this package update and my problem have been solved now. Thanks a lot for your and Zoltan 's contributions and I will keep following the updata.
I advice you to install the most recent RBBGCMuso, I've fixed a lot of sensitivity related bug.
I have re-installed the package, but the version is still 0.6.1.2.
I've forget to change the version number in the description file, but the content is new. :) I've updated it now. It is 0.6.1.3
If I want to do parameter optimazation, whicn function in the package should I use?
This is optiMuso() function in the feature/glue branch, but I am not recommend to use it, because it is not stable, has a lot of bugs yet. I will include this feature in master branch in this week with the 0.7.0.0 version.