wiki:DevelopmentActivities/Xios/Validation

Version 8 (modified by jgipsl, 10 years ago) (diff)

--

Validation of implementation of XIOS interface in ORCHIDEE

ORCHIDEE trunk rev 1808 (some tests redone with rev 1872)
XIOS trunk rev 455 (bug in rev 452)

Conclusion

  • No difference in results between IOIPSL and XIOS output for instaneous output frequency.
  • Small differencies when the output is an average over several time steps : the output in Sechiba differs slightly for a few variables(lwdown, q2m, qair, swdown, t2m, tair, evspsblveg, inter, snowf) on some grid cells. All variables in Stomate are the same.
  • The start date has to be treated properly in the set up phase (in the configuration, by orchidee_ol.driver)
  • rev 1872 for almaoutput

Details on simulations

Configuration ORCHIDEE offline : Choisnel, regional, instantanous output, 4days simulation with PeriodLength=2days

  • OK ioipsl = xios (no test sur alma)
  • Time axes problem, see below
  • Tested files with maximum output level :
    • sechiba_history.nc (freq 1800sec): OK
    • sechiba_out_2.nc (freq 1800sec) : OK
    • stomate_history.nc(freq 1 day) : OK
    • stomate_ipcc_history.nc (freq 1 day) : OK
  • Also tested with rev 1872 : OK

Configuration ORCHIDEE offline : Choisnel, global, instantanous output, 4days simulation with PeriodLength=2days

OK. Same set up and results as the regional simulation.

Configuration ORCHIDEE offline : Choisnel, regional, average output over 5 days simulation with PeriodLength=5days

  • Tested files with maximum output level :
    • sechiba_history.nc (freq 5 day): Difference on 6 variables lwdown, q2m, qair, swdown, t2m, tair. Difference is only on last significant number, 6 occurencies
    • sechiba_out_2.nc (freq 5 day) : Diff as sechiba_history
    • stomate_history.nc(freq 5 day) : OK
    • stomate_ipcc_history.nc (freq 5 day) : OK

Configuration ORCHIDEE offline : Choisnel, global, average output over 5 days simulation with PeriodLength=5days

Same results as regional simulation but also differencies on the variables evspsblveg, inter and snowf. The differences are local to a few grid cells (evspsblveg : 5, inter : , snowf : 1).

Configuration ORCHIDEE offline : CWRR, regional, instantanous output, 4days simulation with PeriodLength=2days

  • OK ioipsl = xios (no test sur alma)
  • Time axes problem, see below
  • Tested files with maximum output level :
    • sechiba_history.nc (freq 1800sec): OK
    • sechiba_out_2.nc (freq 1800sec) : OK
    • stomate_history.nc(freq 1 day) : OK
    • stomate_ipcc_history.nc (freq 1 day) : OK
  • Also tested with rev 1872 : OK

Configuration ORCHIDEE offline : Choisnel and CWRR, regional, instantanous output with almaoutput

OK in rev 1872

Time axes problem

The start date, the time origin and calendar type is set in iodef.xml when using XIOS. These parameters have to be updated in the set up phase before each execution. This is not yet done. Therefor the time axis is start from zero each exectution even for a simulation starting from restart files. These parameters could be updated properly by the orchidee_ol.driver.

In IOIPSL : OK

  • 19820101_19820102_1D_stomate_history.nc : time_counter = 43200, 129600
  • 19820103_19820104_1D_stomate_history.nc : time_counter = 216000, 302400

In XIOS : PB!!

  • 19820101_19820102_1D_stomate_history_xios.nc : time_counter = 43200, 129600
  • 19820103_19820104_1D_stomate_history_xios.nc : time_counter = 43200, 129600