Not all data available – in #10: INT2LM

  @burkhardtrockel in #914794b

<p> The vertical wind W is missing in the initial field. </p>

  @matthiasdemuzere in #0c33f4f

<p> Ha, indeed. For some reason W was saved in out02. <br/> I have now appended this field into out01. </p> <p> Int2lm now continues, but is stuck on the following: </p> <p> interpol_coarse_special_l: NEW soil moisture interpolation with SMI <br/> interpol_coarse_special_l: FOR lmulti_layer_in WITH ke_soil_lm= 9 <br/> interpol_coarse_special_l: IF: l_smi = F is .TRUE. THEN w_so_lm &gt;= pwpb </p> <p> I have added l_smi = .FALSE. to no avail. <br/> Might this be related to the fact tha W_SO is zero for this region? </p>

  @burkhardtrockel in #d82f16e

<p> This is probably a question for the <span class="caps"> TERRA </span> experts. I have no clue on this. </p>

  @redc_migration in #e0d0fcd

<p> Which landcover and soil data sets did you use when creating the external parameters? Some of the raw data is not available over Antarctica. </p>

  @matthiasdemuzere in #cda2807

<p> Please find the .def files from <span class="caps"> WEBPEP </span> in attach. </p> <p> For the full 0.22° Antarctic run I did not have any problem. ALl went fine. <br/> It is the nesting in the output of this run that is causing problems (so the 0.025° resolution). </p> <p> I used the same underlying datasets for both. </p>

  @ulrichschättler in #7f52e79

<p> Hello Matthias, <br/> I am not an expert of <span class="caps"> TERRA </span> , but can comment on the 0 soil moisture. If you have l_smi=.FALSE. in INT2LM, the soil moisture is set to 0 over ice points, which are sure many in the antarctic region. But this should not cause the INT2LM to stuck. Is the output you are citing ( <span class="caps"> NEW </span> soil moisture interpolation….) the last one you get from the INT2LM run? </p> <p> By the way: with l_smi=.TRUE. the soil moisture is not explicitely set to 0 for ice points, but no value is computed, which could result in a crash later on (depending on the compiler). </p>

  @matthiasdemuzere in #63577a1

<p> Hi Ulrich, </p> <p> Yes, that is the last output from int2lm after reading the first time step of the meteo boundaries produced by 0.22°. <br/> For the full log, please see attach. </p> <p> And if I understand it well, it is then best to l_smi = .FALSE.? </p> <p> Thanks. </p>

  @ulrichschättler in #0bfc901

<p> Hello Matthias, <br/> yes: setting l_smi=.FALSE. should at least give some reasonable results. But from your output I cannot see, what is going on there. <br/> If you could provide me the necessary data, I can offer to test it on our computer. Perhaps then I can see what is going wrong. </p>

  @matthiasdemuzere in #e146799

<p> Hi! </p> <p> Thanks for this offer. <br/> I’ll send you my domain file, int2lm namelist as well as some forcing files from the first <span class="caps"> CCLM </span> simulation via email. <br/> Please let me know if you would need more info. </p> <p> Cheers, <br/> Matthias </p>

  @ulrichschättler in #5b5783f

<p> Hello Matthias, <br/> I tested with our versions INT2LM 2.0 and also the latest one, 2.02, using your data. And I had no problem running the INT2LM. <br/> Which version are you using? And which compiler? Could you also test on a different computer? </p> <p> Ciao <br/> Uli </p>

  @matthiasdemuzere in #61b63f0

<p> Ha, that is interesting. </p> <p> So the test that crashed was done with int2lm_131101_2.00_clm2, using the intel compiler. <br/> I just tried with int2lm_131101_2.00_clm3, compiled with gcc, and that works well. </p> <p> The reason I switched to our intel version was that, for the 0.22° preprocessing, the gcc compiler options appeared to be sensitive to boundary violations (see bug int2lm). This was not a problem for our intel compiled version. </p> <p> And now there seems to be an inverse sensitivity? Or it could be related due to differences between the two versions? <br/> No time to test this now, but if this would be relevant, I can test this later on? </p>

  @ulrichschättler in #628045a

<p> You mean the bug in INT2LM, that Burkhardt lately posted? But this is also something I do not understand yet. At least we have never seen something similar. </p> <p> I do not know the differences between the subversions clm2 and clm3 (they are not available on the web page, are they?), so perhaps this really should be tested. </p> <p> You said that version clm3 with gcc compiler was sensitive to the boundary violations when producing the data for the 0.22° domain? Could you perhaps also provide me all the necessary data for that run? I would be curious to see, whether I can reproduce these problems with our version. </p> <p> Ciao </p>

  @matthiasdemuzere in #975bb31

<p> Hi Uli, </p> <p> I just tried to produce the boudary files for the 0.22° and 0.0625° run, using a gcc and intel compiled version of int2lm_131101_2.00_clm2 and int2lm_131101_2.00_clm3 . </p> <p> Perhaps first with respect to your question on the differences between _clm2 en _clm3, see <span class="caps"> README </span> _changes file in attach. </p> <p> For the results on the little test: <br/> 1) with _clm2 <br/> ——————- <ul> <li> intel: Ok </li> <li> gcc: Problem for nesting resolution 0.0625°. </li> </ul> </p> <p> This is the error: <br/> _interpol_coarse_special_l: NEW soil moisture interpolation with SMI <br/> interpol_coarse_special_l: FOR lmulti_layer_in WITH ke_soil_lm= 9 <br/> interpol_coarse_special_l: IF: l_smi = F is .TRUE. THEN w_so_lm &gt;= pwpb <br/> _ </p> <p> 2) with _clm3 <br/> ——————- <ul> <li> intel: no problems for both resolutions. </li> <li> gcc: Problem for the 0.22° resolution. </li> </ul> </p> <p> This is the error: <br/> _At line 1032 of file /home/rcs/software/gcc/int2lm_131101_2.00_clm3/src/external_data.f90 <br/> Fortran runtime error: Index ‘0’ of dimension 1 of array ‘lolp_in’ below lower bound of 1_ </p> <p> So it seems that all works fine (for both resolutions) either with _clm2+gcc or _clm3+intel. <br/> I am not sure what could be causing the other errors. </p> <p> I’ll send you some boundary files for the 0.22° domain, perhaps you could test this on your system as well (although this one this work for intel). </p>