Evaluation of the Snowy Mountains Cloud Seeding Program and Future Directions

Glaciogenic cloud seeding has had a long history in Australia, but with mixed results, particularly on the mainland. This is in large part due to the highly variable nature of precipitation, particularly in complex terrain and with convective cloud, in addition to the relatively small size of the cloud seeding impacts. As a consequence, long-running experiments and considerable expense are required for robust evaluation. Another issue hampering previous cloud seeding trials has been poor statistical relationships between target and control areas. The variability in these relationships effectively decreases the signal-to-noise ratio of the impact, making detection even more difficult.

The Snowy Precipitation Enhancement Research Project (SPERP) was a long-running experiment which sought to address the issues of previous trials through careful experimental design. In addition to existing long-running gauges, a high-resolution network of all-weather precipitation gauges was installed for the evaluation of the project. The trial was carried out in two phases, SPERP I over 2005-2009 with a target of around 1000 km2. The encouraging results from SPERP I lead to the expansion of the target area to 2500 km2 with this second phase (SPERP II) carried out over 2010-2013. The project was converted to an ongoing operational program (Cloud Seeding Program [CSP]) in 2013 and operations have continued to date. Throughout the various phases, cloud seeding was conducted in 5 hour intervals (experimental units, or EUs) using ground-based generators to disperse silver iodide plumes over the target area. A fixed target-control design was employed and seeding was conducted with a seed to no seed ratio of 2:1 randomised in SPERP I, 1:1 randomised in SPERP II, and sequentially with ratio 7:1 in the ongoing operational phase. Several starting criteria were employed to ensure that precautionary requirements were met and the most high-impact intervals were captured. These criteria included freezing level and cloud depth determined by radiosonde, and path-integrated supercooled liquid water measured by microwave radiometer. Additional start criteria were developed following each experimental phase as incremental analysis showed that the cloud seeding impact was sensitive to parameters such as generator hours per EU, cloud top temperature, and amount of precipitation.

A unique feature of the SPERP I evaluation was the finding that the estimation of the “natural” precipitation in the target area, based solely on the precipitation in the designated control area, was a significant source of uncertainty. A systematic search found that, for SPERP I, the low-level Froude number, indicative of terrain-induced blocking, served to optimally improve prediction of the natural precipitation. This improvement served to increase the robustness of the evaluation, with statistically significant impacts identified. For SPERP II, a similar search found the westerly wind component at the -5°C level to be an alternative indicator, representative of the generation of precipitation by orographic forcing. The analysis of the SPERP II experiment showed a fractional increase of 12% in the southern area (significance 6%) and 16% in the northern area (significance 3%), consistent with the findings of SPERP I.

The future of the evaluation of the CSP is an open question. Due to increasing costs and safety concerns around maintaining a high-resolution surface meteorological network in remote alpine terrain, and the opportunity cost in running “no-seed” control events, the use of a numerical model to evaluate the cloud seeding impacts is desirable to complement the direct observations of precipitation. The use of specialised modules for the Weather Research and Forecasting (WRF) model is considered, and strategies presented for its evaluation over the last decade of cloud seeding operation in the Snowy Mountains.