Using the double fence intercomparison reference (DFIR) as a standard for true precipitation, the WMO study obtained intercomparison data at 26 sites worldwide for the most commonly used gauges, concentrating on the shielded Canadian Nipher and Russian Tretyakov gauges and the unshielded Hellmann and U.S. 8-inch gauges. The catch ratio is the ratio of the gauge observation to the DFIR. Wind-induced undercatchment is the major source of measurement error; thus, regression formulas of catch ratio based on these WMO data are mostly functions of wind speed. Air temperature was not a significant factor, except for wet snow above -2°C and for measurements with the Tretyakov gauge. The Barrow study, however, lowered the inclusion threshold from 3.0 mm to 0.3 mm. As a result, several snowfalls with air temperatures near -30°C were included in the new analysis.
By combining the high-latitude Barrow and the mid-latitude WMO datasets, we extend the range of air temperature and include it as a factor in our regression analysis for the gauge catch. Only the two shielded gauges are used since the unshielded gauges miss over 90% of the snowfall at Barrow. We find that, at a wind speed of 6 m s-1, the temperature correction at -20°C reduces the catch efficiency of the Nipher gauge by over 30% and that of the Tretyakov gauge by over 20%. For the Tretyakov gauge, this almost doubles the comparable WMO correction.
Regression plots of gauge catch against wind speed and air temperature still have considerable unexplained scatter. We use two numerical toolspublished models of catch efficiency and the Desert Research Institute Snow Growth Modelto obtain theoretical explanations for the observed dependencies. The catch efficiency models use commercial products to simulate wind fields about a gauge and use the Navier-Stokes equation to compute particle trajectories. Uplift at the leading edge of the gauge distorts particle trajectories, causing smaller and lighter particles to overshoot the gauge. Despite the complex and varied aerodynamic properties of snowflakes, their drag characteristics are largely determined by their terminal fall speeds, which are well characterized by crystal type and size. Higher wind speeds and lower particle fall speeds result in lower catch efficiencies. By running the Snow Growth Model for two cloud thicknesses and for a range of surface air temperatures, we determine the mass-weighted fall speed, mean particle size, and precipitation rate for an assumed gamma distribution of snow particles. Although there are other contributing factors, lower air temperatures in the model runs generally correspond to reduced particle growth rates and, hence, to lower fall speeds and catch efficiencies.