85th AMS Annual Meeting

Wednesday, 12 January 2005: 4:00 PM
Numerical Simulations of Snowpack Augmentation for Drought Mitigation Studies in the Colorado Rocky Mountains
Curtis L. Hartzell, Project Consultant for the Colorado Water Conservation Board, Denver, CO; and J. Busto, W. R. Cotton, R. McAnelly, G. Carrió, and L. Hjermstad
The Colorado Weather Damage Modification Program (WDMP) research project was joined with the Denver Water Department's operational cloud seeding program in the central Colorado Rocky Mountains for the 2003-2004 winter season. The goal of the project was to provide a physical evaluation of the operational cloud seeding using the well-established Colorado State University (CSU) Regional Atmospheric Modeling System (RAMS), with a fine 3-km grid horizontal spacing covering the entire seeding area. Real-time no-seed control model runs based on 00Z Eta initialization data were made daily with the output posted on the CSU project web site, where the forecasts could be accessed by the cloud seeding contractor. Evaluation of model performance was done throughout the course of the winter season. One problem revealed by these evaluations was a significant warm-temperature bias from the surface to above mountaintop level. Model fixes were implemented in mid-February that reduced, but did not eliminate this warm temperature bias. The model was extended to include seeding effects, in order to evaluate the seed vs. no-seed precipitation simulated by RAMS.

After-the-fact control no-seed simulations were rerun for 152 days through the five-month winter season (November 2003-March 2004). The fixed design of the control runs was nearly identical to that used for the real-time forecast simulations. It included the in-season adjustments, which were found to improve model performance, and also some minor microphysical alterations that were necessary for compatibility with the subsequent seeding simulation design. Simulated 24-hr precipitation from these daily forecast runs was used to establish no-seeded simulated precipitation for individual events and monthly and seasonal totals. Comparison with Snotel observations shows that the forecast runs generally simulated the spatial distribution of precipitation well, but with an over-prediction bias for precipitation amounts (factor of 1.88). Possible sources of model precipitation biases are:

ˇ Inadequate resolution of atmospheric dynamics and terrain, especially when embedded convection is prevalent.

ˇ Meyers formula for crystal concentration over-predicts concentration of natural ice crystals.

Similar after-the-fact RAMS seed simulations were performed for each day on which seeding operations were conducted (86 days), with simulated sources of silver iodide (AgI) at specified low-level model grid points in accordance with the timing and magnitude of AgI release at each seeding generator as recorded in operational seeding logs. The AgI is treated as a second predictive IFN field with its own activation characteristics. All other aspects of the seeding runs were identical to the control no-seed simulation design. Simulated 24-hr precipitation in these seeded runs replace the amounts from the corresponding control runs to form complete event, monthly and seasonal simulated precipitation totals that include all seeding operations.

The differences in 24-hr precipitation between the seeded and no-seed control simulations are very slight, generally within 1% averaged over the target area. These very slight differences are evident over much of the 3-km grid domain, with the small difference patterns extending well upstream and laterally from the seeding generators. The difference patterns tend to show well-resolved, alternating positive and negative bands aligned with the prevailing wind in a given case. These patterns suggest an unexpected weak, dynamical response to the seeding that propagates throughout the seeding target area and surrounding area and leads to very weak horizontal roll vortices superimposed on the ambient dynamics. The small differences between RAMS seed and no-seed simulated precipitation could be because:

ˇ The background CCN and IFN concentrations are unknown; therefore, the results are at the mercy of specified background concentrations.

ˇ The model under-predicts supercooled liquid water content in the lower portion of clouds over the target area, thereby reducing seedability.

ˇ An unforeseen dynamic response that appears to result in large areas of slightly suppressed precipitation in the target area and small regions of slightly enhanced precipitation.

ˇ The low-level warm temperature bias results in delayed AgI nuclei activation, fewer activated nuclei, and less time for crystals to grow and snow to fall in the target area.

ˇ The transport and diffusion of seeding material from generator sites is getting into the clouds too far downwind of the generator sites.

Project analysis and evaluation work that focuses on factors of model uncertainty is ongoing. One such evaluation uses Mielke's Multiple Regression Block Permutation (MRBP) method to evaluate both no-seed control and seed runs for 30 selected days that span the full range of meteorological regimes that are conducive to snowfall in the project's target area. Six of the 30 days that are most representative of the regimes are being used for a Lagragian particle dispersion analysis. Project results, findings, conclusions, and recommendations will be presented at the special WDMP session.

Supplementary URL: