Possible extreme marine snowfall as seen by BALTRAD

Intense radar reflectivity echoes (greater than 50 dBZ) were observed over the Norwegian Sea on 3-5 March 2006, primarily from met.no's northernmost radar at Røst. Such intense radar reflectivity suggests very intense snowfall, although some may argue that such extreme values are unrealistic for snow. Unfortunately, no precipitation observations were available in the area, and no anecdotal evidence has come to our attention. Satellite data indicated intense convective structures but were not able to quantify snowfall. Finally, no operational numerical weather prediction model was successful in reproducing the extreme snowfall inferred from the radar reflectivity. The radar data in question are BALTRAD datasets generated at the BALTEX Radar Data Centre at SMHI, and BALTEX is the European Continental-Scale Experiment within the framework of GEWEX.

The purpose of this presentation is to examine whether the intense reflectivity from snow was achievable in reality and to highlight the capabilities and limitations of radar-based QPE. Our examination of the strong radar reflectivity in this case focuses on three processes that might explain the unusually strong radar reflectivities: anomalous propagation due to ducting of the radar beam, the existence of a melting layer, and potential errors in the range-bias correction.

First, atmospheric ducting was highly unlikely. Such conditions lead to super-refraction of part of the radar beam, leading to the systematic underestimation being less than normal with increasing range, implying that the adjustment would be too severe in such cases. However, during the period studied, low static stability above the sea and close to the radar, which is located on a small island, excludes the occurrence of ducting. On the contrary, atmospheric conditions along the radar beam were favourable for sub-refraction, suggesting increased underestimation of surface precipitation at long ranges.

Second, the temperatures were too cold for a melting layer to exist. Such conditions cause the infamous "bright band" in radar data, also resulting in higher reflectivities which would reduce the underestimation locally.

Third, the main potential cause of unusually high radar reflectivity appears to lie in the representativeness of the range-bias correction and how it is derived, along with other factors impacting on the quality of the input radar data. Specifically, the derived correction factor is based on gauge-radar comparisons for the complete BALTRAD coverage area which, in the case of the northerly area in this study, means comparisons from considerably warmer conditions. Thus, the general nature of the correction is not always suitable for application locally.

The gauge-radar relations are often noisy, and the large variability is due to several error sources. One of these errors is caused by partial shielding of the radar beam due to terrain, forest, buildings, and other obstructions. This results in a "system" bias forming (i.e., a correction which is independent of range), and such a bias is pronounced (around 3 dB) in this case. Yet, even if this system bias were removed, the correction at distant ranges would still be extreme, and so would be the resulting snowfall in the radar reflectivity. Thus, these three potential factors cannot adequately explain the unusually high radar reflectivity in this case, strongly suggesting the radar reflectivity is largely real and reflecting heavy snowfall.

Operational analyses (e.g. SMHI's Mesoscale Analysis (MESAN) system) indicated the intense reflectivity occurred in a region of sharp temperature gradient between air masses and a large air-sea temperature difference, favoring low static stability warm air over a quasi-stationary front. Thus, we speculate that the intense radar reflectivity was associated with embedded convective storms caused by the ascent of warm unstable air over a strong frontal zone. The heavy snow was generated in the preferred temperature regime for rapid dendritic growth and was largely unresolved by the numerical model. If the extreme snowfall in the BALTRAD radar reflectivity turns out to be realistic, then there will be, inevitably, implications on our ability to measure, forecast, and warn of such conditions in the future.