Hans, Min, Ju-Won, Wei Wang
Lilly room, 3:00-4:30pm, Thursday 31 August 2006
Ju-Won found something
wrong in earlier verification which was misunderstood between local time and
UTC. Now the best precipitation verification scores come from DDFI. Ju-Won also
showed the conventional verification. The specific humidity was improved with
DFI. Hans suggested drawing results for the cold start group and the cycle
group separately.
To solve the problem TDFI
has longer spin-up time and lower precipitation rate than NODFI, Min made
several tests:
l
No filtering the
qvapor and hydrometeors. There is still a lot of noise.
l
Just filtering
qv, dpsdt is still noisy.
l
During the
backward integration, switch on the microphysics option but skip the calling of
module-microphysics driver to make the hydrometeor arrays active.
A. ThereŐre negative values in hydrometeor arrays
from solve_em.
B. It remove the moist-loops, no advection and
diffusion (both explicit and implicit) for hydrometeors, the moisture arrays
will be kept identical to these in wrfinput.
Using B for cycle tests, results show that the precipitation rate is still lower than NODFI.
HansŐ suggestions:
l
The negative
values of hydrometeor arrays could be caused by the negative filtering
coefficients. Make a test with the Lanczos filter.
l
Three options for
moisture filtering: 1) Filtering qv and hydrometeors, 2) Save the qv and
hydrometeors at t=0 during the forward integration procedure of filtering as
the initial value of their filtered results; 3) Take qv and hydrometeors from
the analysis (wrfinput).
Wei: Make a final version for the combined wrfvar and wrf+dfi.
Ju-won: redraw to separate cold start ones and cycle
ones; give Min the
precipitation verification tool and a
modified conventional verification program to calculate several each level;
send Wei the latest code after check it out about diagnoses from MinŐs code.
Min: Further
tune dfi configuration; have a test using
Lanczos filter; have tests by three option for moisture filtering (explained
above).
Min return to BMB on
next Monday(Sep. 4th 2006)
We will have the next project meeting on: Thursday 7 Sep, 2 pm, Lilly room.