PEST_HP error writing parameter to model input file(s): exponent too large or small for double precision protocol or parameter value is NaN

[sjepsen395]

Hi All,

I hope it's OK to post a question about PEST_HP here. I recently encountered the following error message when running PEST_HP (actual parameter name was given in place of "parnme"). It occurred at the beginning of a set of Jacobian runs. Has anyone experienced this and found the cause? I'm wondering if it has something to do with the parameter getting too close to zero to express in scientific notation. The parameter has a "relative" change limit, lower bound of zero, upper bound of 1, and RELPARMAX setting of 3. Unfortunately I don't have the specific value of that parameter that PEST_HP is seeing and trying to write to the input file. I'm trying another run right now with all parameters changed to have "factor" change limits to see if that fixes the problem. Will post back here if that solves the problem.

Thanks,
Steve

---

Error condition prevents continued PEST_HP execution:-

Error writing parameter to model input file(s): exponent too large or small for double precision protocol or parameter value is NaN.

---

[sjepsen395]

The ongoing PEST_HP run with modified parameter change limits did not show the above error at the point where it occurred, so it looks like that fixed the problem. Lessen learned: be careful when using "relative" change limit, a parameter with lower limit of zero, and RELPARMAX > 1. Somewhere in that combination, things went bad. This was fixed by going over to a log-transformed parameter with "factor" change limit and a small negative offset value to allow a lower bound of zero as seen by model.