WMOD00268

• Applicant: CWCB
• Sponsor Name: USBR
• Project Name: Final Report Weather Damage Mitigation Program
• Title: Numerical Simulations of Snowpack Augmentation for Drought Mitigation Studies in the Colorado Rocky Mountains
• Prepared For: USBR - WDMP
• Prepared By: Curt Hartzell, Dr. William Cotton, Joe Busto
• Date: 9/1/2005
• State: CO
• Weather Modification - Doc Type: Scientific Study

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

size distribution. These experiments all produced similar results regarding the
small sensitivity to seeding on precipitation amounts and the large-area
manifestation of these slight differences. In mid-May 2004, after settling on a
more suitable set of microphysical options based on these sensitivity tests,
CSU began the full production of control no-seed and seed runs.

After finishing the control no-seed and seed runs for the 30 selected
days (Table 2.1) and proceeding on through another dozen seeding days,
another problem was discovered. The winds used to initialize the RAMS
model and to provide time-dependent lateral boundary conditions on the
largest grid (Grid 1) are derived from NCEP's Eta model initialization and 3-hr
forecast files. In the Eta files, the u and v wind components are relative to the
Lambert-conformal mapping of the Eta grid, rather than being true u and v
wind components. The model code used for the real-time runs had been
adapted long ago to properly transform the Eta winds onto the RAMS grid.
However, RAMS source code that was used to extend the model to include
seeding effects had never been adapted to properly transform the Eta winds

The improperly transformed winds were small in error and practically
unnoticeable, except perhaps in the northern corners of the large Grid 1
domain where they can deviate from their true direction by as much as 30-
degrees and thus be highly non-geostrophic. When this inadvertent error was
discovered in early July 2004, several cases were rerun with the corrected
winds in order to assess the sensitivity to the error. The effects of the incorrect
winds on simulated precipitation amounts were trivial to moderate on the 3-km
fine grid in the individual cases, but tended to produce more precipitation due
to slightly incorrect large-scale dynamics being forced inlo the western
boundary of the large grid. Because there is a simulated precipitation over-
prediction bias in the model (discussed in Section 4), It was decided to rerun all
the control no-seed and seed runs using the corrected wind transformation in
order to eliminate this systematic error that exacerbated the problem. The final
production seed runs were completed on 24 August, and the final control runs
(including all non-seeded days) were completed on g September 2004.

2.9 Task 5 - Perform evaluations of model predictions of precipitation using
Multivariate Randomized Block Permutation

Model skill for simulated precipitation was evaluated using the 30 days
selected for use in project evaluations (Table 2.1). This evaluation used
Multivariate Randomized Block Permutation (MRBP) statistics (Mielke, 1984, 1991)
as implemented by Cotton et at. (1994) and Gaudet and Colton (1998). SNOTEL
24-hr precipilation and SWE data were used for the evaluations (SNOTEL sites
used are identified in Appendix 2). The purpose for this evaluation was to
determine if the model-forecast skill was sufficient to say something definitive
about seed VS. no-seed simulated precipitation differences, or if these differences
were within the noise level or level of uncertainty of the model.

35