OutputDump and Warm Start Issues

[ktsuruta]

Using the v2.8.2 source code, I had an issue with the OutputDump command. Specifically, when running Raven on the Alouette basin (from the files provided in the Example/Tutorial Files link) with a start date of 1991-10-01 00:00:00 and OutputDump date of 1991-10-10 00:00:00, I encounter the following:

ERROR : AddModelOutputTime: Cannot have model output time prior to start of simulation

I believe this issue relates to the AddModelOutputTime behaviour of CModel and its use of TimeDifference. In the source code, t_loc is defined as:
double t_loc = TimeDifference(tt_out.julian_day,tt_out.year,Options.julian_start_day,Options.julian_start_year);

I believe it should be defined as:
double t_loc = TimeDifference(Options.julian_start_day,Options.julian_start_year,tt_out.julian_day,tt_out.year);

After making this adjustment, OutputDump runs as expected. However, I have found that warm starts using OutputDump or a solution.rvc file do not produce expected results. I.e. the hydrographs of the following runs do not match:
1) A cold start of Alouette with StartDate=1991-10-01 00:00:00, Duration=1096, and TimeStep=24:00:00
2) A warm start of Alouette with StartDate=1991-10-11 00:00:00, Duration=1086, and TimeStep=24:00:00 initialized by the solution.rvc of a cold start (StartDate=1991-10-01 00:00:00, Duration=10, and TimeStep=24:00:00).

I believe this issue relates to the HRU's with designation GLACIER whose maximum soil moisture capacity at the topsoil layer (SOIL[0]) must be respected during initialization, but appears to be violatable during all subsequent time steps (In glaciated HRU's, the SOIL[0] maximum capacity is zero, but SOIL[0]>0 in these HRU's frequently throughout a given run). This issue may indicate that other state variables are not constrained by their maximums throughout a run.


Many Thanks,
Kai

[jrcraig]

Thanks, Kai, for the input - the OutputDump command is rarely used and it seems was incorrectly updated; your suggested fix will be in for the next version.

Thanks, also, for identifying the glacier issue - I will look into it. Note that state variable constraints are not performed centrally, so I wouldn't worry about the maximum constraint for the other state variables unless you specifically see evidence of it. It is clearly an issue here, however.