IFI.F can use junk values for hourly accumulated precipitation

[SamuelTrahanNOAA]

This code in IFI.F is incorrect:

          if(ifi_apcp(i,j)>9e9 .or. ifi_apcp(i,j)<-9e9 .or. ifi_apcp(i,j)==0) then
              ifi_apcp(i,j) = avgprec_cont(i,j)*to_hourly

It causes the post to send junk data from avgprec_cont in situations where precipitation in ifi_apcp is identically zero. With my HRRR tests, junk data gets through sometimes (eg 1.72E+16) . This leads to the UPP IFI producing different precipitation types in its derived fields.

The correct code is:

          if(ifi_apcp(i,j) == spval) then
              ifi_apcp(i,j) = avgprec_cont(i,j)*to_hourly

Without this bug fix:

1. Points in the HRRR IFI 1hr apcp field have invalid data in some locations.
2. UPP sends those values to libIFI, which it then uses as if the data were valid.
3. The standalone IFI uses -9999 (missing value) in those locations. This results in a mismatch between standalone IFI and IFI in UPP.
4. The first field that changes is wx_type.

With this bug fix:

1. Points in HRRR IFI 1hr apcp field appear to be valid.
2. Standalone IFI and UPP IFI get identical results.

[WenMeng-NOAA]

@SamuelTrahanNOAA Would this bug have impact on RRFS and MPAS results?

[SamuelTrahanNOAA]

The bug and its fix are confined to IFI.F, so it'll only affect the IFI fields, and only if libIFI is enabled. (Without libIFI, the fields are all missing values regardless.) My fix in #1127 also updates a comment in another file to match this bug fix, but it is only a comment, so it'll have no effect on results.

[SamuelTrahanNOAA]

The result of this bug is false positives in icing severity category.

Plots of surface precip of test data (2020 June 1, 18Z)

The big red splotches are areas where the surface precip type becomes 7 instead of 0.

The entire 3D precipitation type arrays also differ. I'm plotting only the surface here.

Icing severity category in test data (2020 June 1, 18Z)

Note the differences in the fields just to the west of Florida. Precip types 0 or 4 become 2 or 5.

There are other differences in the central north part of the domain.

Icing severity category with surface precip difference contours in test data (2020 June 1, 18Z)

Here, I plot red contours around areas where the surface precip disagrees between IFI in UPP vs IFI standalone. You can see the differences in icing severity category follow the differences in surface precip.