| 1 | = Assimilation of ORCHIDEE model = |
| 2 | |
| 3 | ----------------------------------- |
| 4 | |
| 5 | |
| 6 | |
| 7 | == 23/05/2011: E-mail envoye a Thomas Kaminski pour la version actuelle d ORCHIDEE pour Assimilation == |
| 8 | |
| 9 | Hello Thomas, |
| 10 | |
| 11 | We have now fixed the problem on the compilation of ORCHIDEE model by using the option "--chkglobal" with lf95. The forward run test has also been performed (i.e., call three times the model, the two first ones without the change of the parameters (as priors), and the last one by perturbing the priors). |
| 12 | |
| 13 | 1. The new version of the ORCHIDEE model shaped for TAF can be found from our ftp server here: |
| 14 | (These documents are only available on this server for a period of 7 days): |
| 15 | |
| 16 | ftp://ftp.cea.fr/incoming/y2k01/orchidee |
| 17 | |
| 18 | Note that the previous provided readme file is still valid. For this new version, some additional netcdf files have been considered for the runing of TAF (see the line 17 in the file AA_make in the directory ORCHIDEE_OL to indicate the right path:, i.e, NETCDF_SRC). |
| 19 | |
| 20 | 2. In parallel, we are working on the generation of the tangent linear codes of the model with TAF and we have experienced some recurrent errors and warnings. Find hereafter some of them: |
| 21 | |
| 22 | - An error on the string length: " INTERNAL ERROR: string too long." We do not understand this error since we have set the maximum length of the string to 130 throughout ORCHIDEE model and set to 132 in the command line for TAF in the Makefile. For more details on this error and others, see the output file from TAF running (i.e., taf_output_current and taf-tlm-current.log in the directory TANGEANT). |
| 23 | |
| 24 | - The following recurrent warning for several routines, e.g., : " *WARNING* : model.f90:463 : call of intersurf_main too many arguments (44) defined = 0" (see in the file taf-tlm-current.log). |
| 25 | |
| 26 | - TAF seems to not handle the following fortran command, as e.g., this line in some netcdf files: |
| 27 | localMap (:numDims ) = (/ 1, (product(localCount(:counter)), counter = 1, numDims - 1) /) |
| 28 | We have then made some modifications in such netcdf files and also made compliant some parts of the ORCHIDEE model to proceed further with TAF. For more details, we have provided the netcdf files we are using including both the original and modified files (i.e., netcdf_expanded.f90 and netcdf_text_variables.f90; see in the directory NETCDF in TANGEANT). |
| 29 | |
| 30 | Note: As it is built now the Makefile for the generation of TL codes (exec tlm), one needs to compile the forward model before, otherwise there is an error. This needs to be fixed. |
| 31 | |
| 32 | 3. Issues on the generated TL codes by TAF |
| 33 | As already mentionned, TAF seems to generate TL codes, but we think that these codes are incompleted. However, we have tried to compile them. Hereafter are some problems we encountered: |
| 34 | |
| 35 | |
| 36 | - The generated TL codes by TAF have sometimes string length over 132 characters. Of course, such a length is not allowed by the lf95 when considering the stricted options of this compiler. As an example, the generated TL code for TL relevant to the module intersurf.f90: |
| 37 | |
| 38 | subroutine intersurf_main_2d( kjit, iim, jjm, kjpindex, kindex, xrdt, lrestart_read, lrestart_write, lon, lat, zcontfrac, & |
| 39 | &zneighbours, zresolution, date0, zlev, u, v, qair, temp_air, epot_air, ccanopy, cdrag, petacoef, peqacoef, petbcoef, peqbcoef, & |
| 40 | &precip_rain, precip_snow, lwdown, swnet, swdown, pb, vevapp, fluxsens, fluxlat, co2_flux_tot, coastalflow, riverflow, tsol_rad, & |
| 41 | &temp_sol_new, qsurf, albedo, emis, z0 ) |
| 42 | |
| 43 | |
| 44 | Can you clarify us on these issues ? Thanks. |
| 45 | |
| 46 | Please do not hesitate to contact us for any clarifications and additional questions. |
| 47 | |
| 48 | Best regards, |
| 49 | |
| 50 | Ernest & Martial |
| 51 | |
| 52 | ------------------------------------------------ |
| 53 | |
| 54 | |