Transient simulation convergence problem

[Qing He]

Hi all,

I have this question regarding model convergence in the transient modflow simulation: I wonder what factors control the model convergence in the transient mode? The parameters setting I use now are exactly the same as those in a steady-state simulation I ran successfully previously, but in the transient simulation the they just did not work.

I compared the simulation log with criteria I defined in the Ims package, it seems both the head difference and head residual in the last simulation step are within the thresholds (see below). Is there any other reason or adjustment I should do for the model setting? 

My model settings are:

#===================time discretization==========================#
tdis_par = [(30, 30, 1.0)] *1  #simlation period length, time step, scale factor
tdis = fp.mf6.ModflowTdis(sim,
                         units = 'Days',
                         nper = 1,
                         perioddata = tdis_par)

#===============interative model solution package=================#
ims = fp.mf6.ModflowIms(sim,
                        complexity = 'SIMPLE',                        
                        outer_dvclose= 1e-3, #
                        inner_dvclose= 1e-3, # 
                        rcloserecord = 0.1,  #In-Out criteria, default = 0.1
                        inner_maximum = 500, # 
                        outer_maximum = 100,
                        print_option = 'ALL')

#===========================================================#

These are the simulation results (The model stops at time step 10):

--------------------------------------------------------------------------------------------------------------------------------------
      TOTAL      OUTER      INNER                                                            MAXIMUM                                  
  ITERATION  ITERATION  ITERATION  MAXIMUM CHANGE      MAXIMUM CHANGE MODEL-(CELLID)        RESIDUAL    MAXIMUM RESIDUAL MODEL-(CELLID)
 --------------------------------------------------------------------------------------------------------------------------------------

       527        100          4 -6.13412610E-05                  1_GWF-(1,157,150) -3.36481495E-03                  1_GWF-(1,156,149)
 --------------------------------------------------------------------------------------------------------------------------------------


 UBDSV1 SAVING           STO-SS ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV1 SAVING           STO-SY ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV1 SAVING     FLOW-JA-FACE ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV06 SAVING       DATA-SPDIS IN MODEL CONUS_T_SIM4     PACKAGE NPF             CONNECTED TO MODEL CONUS_T_SIM4     PACKAGE NPF              ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV06 SAVING              DRN IN MODEL CONUS_T_SIM4     PACKAGE CONUS_T_SIM4    CONNECTED TO MODEL CONUS_T_SIM4     PACKAGE DRN_0            ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV06 SAVING              RIV IN MODEL CONUS_T_SIM4     PACKAGE CONUS_T_SIM4    CONNECTED TO MODEL CONUS_T_SIM4     PACKAGE RIV_0            ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV06 SAVING             RCHA IN MODEL CONUS_T_SIM4     PACKAGE CONUS_T_SIM4    CONNECTED TO MODEL CONUS_T_SIM4     PACKAGE RCH              ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1
 UBDSV06 SAVING              CHD IN MODEL CONUS_T_SIM4     PACKAGE CONUS_T_SIM4    CONNECTED TO MODEL CONUS_T_SIM4     PACKAGE CHD_0            ON UNIT   1016 AT TIME STEP     10, STRESS PERIOD      1

 HEAD WILL BE SAVED ON UNIT 1017 AT END OF TIME STEP   10, STRESS PERIOD    1

[sudarsan N]

Hi..

How did you manage to solve this? I shall tell you what worked for me.
Reducing the Hydraulic conductivity and increasing the Specific yield values would help in convergence. Later, these parameters can be tuned in the calibration step.

[Qing He]

Hi Sudarsan,

Thanks for sharing your experience. It turned out in my case it's because some of the initial heads are below aquifer bottom, so that it causes severe dry-wet oscillation problems of the simulated head.