Details of Feedback Effect on RAINC and RAINNC for the 36/12/4km run
Initially feedback from fields RAINC and RAINNC was turned off in favor of "nice" precipitation plots. This was turned back on, on 01/29/97 to keep all model output similar.
RAINC and RAINNC are ONLY diagnostics, so this had NO effect on the 2-way nesting and feedback during the model run. The only effect was on the fields RAINC and RAINNC at output times. At these times, RAINC and RAINNC were calculated for each domain.
- Up to 01/29/97 these fields were written to the output files directly.
- Since 01/29/97 a 9-point average of these fields for domain 3 were used to REPLACE the overlapping regions in domain 2. Similarly a 9-point average from domain 2 replaced the overlapping regions of domain 1.
- Thus the initial "no RAINC and RAINNC feedback" did not result in any loss of data - in fact we have more information for these three weeks than for the est of the run.
- Below is a plot of RAINC showing the effect on domain 1 before and after the feedback change on 01/29/97 (the plot are for days 27, 28, 29, 30). On day 29 one can see the region of domain 3 clearly, as this domain had no cumulus rain, and therefore a 0 value was feed back from this domain. One can also clearly see that 2-way feedback in the model was not compromised. If it was compromised, domain 3 would not have had an effect on domain1 for days 27 and 28.
- Changing the data for first 3 weeks of January, so that all data is consistent is a simple matter of doing the 9-point average for RAINC and RAINC off-line and overwriting the old fields. Below is a before and after picture where this was done for 01/28 (show is field RAINC)
To solve this problem permanently, the RAINC and RAINNC fields in the auxhist1 files ONLY, will be corrected and placed on mass store under:
/NRCM/DATA_3DOM/199701/FixedAuxFiles
The fields in the wrfout files will NOT be fixed as this will be a duplication.