<param_nml> IRF parameters resulting in no output about mizuroute

[arbennett]

Just to add some clarification here - @steinadio is working with @bartnijssen and myself on a streamflow bias correction project where we're using mizuRoute to route flows in the Columbia River Basin. We were originally looking to recreate the historic simulations that Oriana Chegwidden developed for the Columbia River Climate Change (CRCC) study: https://www.hydro.washington.edu/CRCC, but using the geospatial fabric & associated river network in vector format rather than a gridded river network as used in the CRCC work.

In doing this @steinadio found that the default IRF parameters led to about a week delay in the annual hydrograph for a mainstem location in the Columbia river (at the Dalles, which we have coded as TDA) as well as not matching the flashiness of some flows closer to the headwaters (here, in the Kootenay River, coded KOOTE). Since then they've been tuning the velo and diff parameters in the param.nml file to make things match up a bit better.

Which brings us to the issue that they discovered. If you set the velo above 6 and diff below 400 the routed flows don't appear to come out fully intact. The mizuroute output at the headwaters sites (here, the Kootenay) still produces valid output, but at downstream sites (here, the Dalles) we have been getting nan flows, despite mizuRoute not crashing. We haven't really done any debugging or systematic exploration of the issue, but still wanted to document this. We're using mizuRoute version 1.2.1, which we've gotten distributed on conda-forge: https://anaconda.org/conda-forge/mizuroute

[nmizukami]

Thanks @arbennett and @steinadio,

The first place to look at is make_uh subroutine in process_param.f90, and check if unit hydrograph (seg_hu) is properly generated.

There are some divisions in the routine so nan can come from there, and these use reach length as well as velo and diff.

I can look into, but probably need your input data (network data etc.)