Modifying warning about negative allocation

[mmcgrath]

In a global FG1 run (analytical spinup) with r7272, the following warning is printed over one million times per year:

WARNING 10b: numerical problem, residual is negative we are overspending
WARNING 10b: PFT, ipts:            8          43
WARNING 10b: residual,  -1.632571855481046E-011
WARNING 10b: allocation is being adjusted

grep -w 10b Debug/FG1.ORC2.r7272_22160101_22161231_out_orchidee | wc
4894592 36709440 275418160

It's a useful warning, but no one seems to be looking at it at the moment, and the debug files become a bit more difficult to manage. One solution would be to only write the warning once a year, but that would give the impression that it only happens once a year. Perhaps we should keep track of how often it happens at each pixel and what the maximum residual is? At the end of the year we could then write a summary that could be monitored.

This doesn't seem to be causing any other serious problem, as nothing is caught by the mass balance routines.

Even better would be to check why we overspend so often. Is there a small mistake in the equations? If not, could we use a little fix (e.g. reduce b_inc_tot in the calculation of the allocation or calculate the third component in the allocation as the residual) to avoid this problem from happening so often? If it happens, the model has to adjust the allocation. Could save some time if we could avoid these adjustments.

[luyssaert]

In r7615 the value of the residual is now send to the history file where it can be monitored. This makes the out_orchidee files much smaller.