Rivers in MEDUSA

Rivers in eORCA1

Rivers in eORCA025

Developing the NPD runs and willing to have rivers for the SE-NEMO, we've tested the rivers in MEDUSA-NEMO4.2.2 within the NPD runs (GOSI10)
First, the code itself does what we want. just fixed 3 bugs :

• NPD reads ancillaries in directories (not in the running directory: "./"), and that needed a fix see
• forgot to add the river max depth in the namelist_medusa_ref
• the river nutrient check was in integer instead of REAL

first runs did not work because of something in the newly made river input files.

• tested 1 by 1 seems to come from the N river inputs. but all fields show high values in some specific river mouth, of which some in the Arctic look present very high values (here some example after 5month):
  
  those results are when using the files :
  • river42_total_ALK_flux_LS200km_SP10_OBfix_ORCA025.nc
  • river42_total_DIC_flux_LS200km_SP10_OBfix_ORCA025.nc
  • river42_tot_N_flux_LS200km_SP10_OBfix_ORCA025.nc
  • river42_tot_Si_flux_LS200km_SP10_OBfix_ORCA025.nc
• maybe the we can try with LS 400? here a diff:
  
  Running the LS 400 river nutrients did not solve the problem.
• Andrew added made some special treatment to the Ob fjord in the Russian Arctic. (diff with - without Ob special treatment)
  
  It didn't Solve the problem.
• I've them re-run the last rivers, spreading the nutrients down to 70m deep (max) instead of 10m earlier (see DIN between 2 and 102 mmol/m3 -- max go up to 155mmol/m3 in around the Ob).
  
  Still no Luck.
• Looking at the bathy, spreading the river nutrients down to 70m won't make a difference in the Arctic...
  
  As the sea-floor stands between 8 and 15m deep around there...
• Last hope LS400 with special Ob treatment.
  This test ran longer than the 200km version (7month compared to 2) but still, it eventually failed. same problem around the Ob :
  
  
  And Here LS400 just before it failed
  
  A solution would be to fill (on top of the Ob river mouth) the other problematic fjords, circled in the last plot. The "filling" is only for the river nutrient flux file creation, not at run time. This way, the nutrients would be released in a better mixed location and avoid the accumulation we see in those fjords. Best would be to use the 200km files made this way.
• New Test with more Fjords filled :
  Here is the new modified river input (for N and Si) in the Arctic:
  
  Unfortunately, it failed... again...
  Here are some plots :
  The usual suspects look OK-ish
  
  
  The DIN values are not insane, but this is after 1 month... we might want to fill a tiny bit more in the circled areas
  Further East another beast : This one looks nasty... it's in the middle of probably quite still waters. would be good to fill the suggested part (narrow, and no freshwater).
  
  last bit of the Arctic looks OK
  
• Another tour of improvements,
  Andrew has added an automated way of avoiding narrow/isolated coastal river mouth, to avoid having nutrient accumulating too much
  
  This time the model runs!!! \o/
  But it seems to still accumulate nutrients in some places. Concentration can be quite high (especially DIC get to >7000 mmol/m3 in some places, but the model holds:
  
• Andrew gave another go with the last results in mind to improve the results
  This Time the automation has been run on both DIN/SIL and DIC/ALK riverine input sets.
  It seems to runs better, the model runs for 10years before failing. The run use the fct advection scheme for the passive tracers (same all physics) which is the one having the most chance of failing - just to avoid having to retest everything again. and this set of files :

&nammsa_river
   ln_read_rivn   = .true.
   ln_read_rivsi  = .true.
   ln_read_rivc   = .true.
   ln_read_rivalk = .true.
   rivn_kind      = 2
   rivsi_kind     = 2
   rivc_kind      = 2
   rivalk_kind    = 2
   rivdepmax      = 70.0
   cn_dir      = './INPUT/INIT_MEDUSA/'      !  root directory for the dust data location
   !___________!_________________________!___________________!___________!_____________!________!___________!__________________!__________!_______________!
   !           !  file name              ! frequency (hours) ! variable  ! time interp.!  clim  ! 'yearly'/ ! weights filename ! rotation ! land/sea mask !
   !           !                         !  (if <0  months)  !   name    !   (logical) !  (T/F) ! 'monthly' !                  ! pairing  !    filename   !
   sn_riv_n  =  'river42_tot_N_flux_LS200km_SP10_5ITER_ORCA025.nc'    ,  -12 , 'RIV_N'   ,   .true.  , .true. , 'yearly'  ,    ''            ,    ''    ,    ''
   sn_riv_s  =  'river42_tot_Si_flux_LS200km_SP10_5ITER_ORCA025.nc'   ,  -12 , 'RIV_Si'  ,   .true.  , .true. , 'yearly'  ,    ''            ,    ''    ,    ''
   sn_riv_c  =  'river42_total_DIC_flux_LS200km_SP10_5ITER_ORCA025.nc',  -12 , 'RIV_C'   ,   .true.  , .true. , 'yearly'  ,    ''            ,    ''    ,    ''
   sn_riv_a  =  'river42_total_ALK_flux_LS200km_SP10_5ITER_ORCA025.nc',  -12 , 'RIV_ALK' ,   .true.  , .true. , 'yearly'  ,    ''            ,    ''    ,    ''
/

The error is no more in the Arctic rivers.
It still is a river problem, but around china/korea. 1 pixel switch to crazy DIC :
Because of the new C limitation, we shouldn't have anything quite like that happening...

• A quick test could be to add the nutrient down to a shallower depth of 10m instead of 70.
• we could also strengthen the C security we've added in the C uptake... I quickly check by changing the river depth. we'll see. With river, the total production in 1986 is of 31 Pg-C per y still too low for our liking. this will need some tweaking.
• Trying to Understand why it is failing there, in the Chinese sea...
  • I've tried 1 run with all Andrew's new LS200 files (let's call it LS200)
  • and also 1run where i've switched DIC and ALK to the widerly spreaded LS400 (so let's call it LS400). Both failed. LS400 few days later though.