Evaluating Parsivel2's raindrop data: A comparative study of different terminal drop velocity models on simulated and natural rain events

Rainfall measurement inherently contains uncertainty in the accuracy of the collected data. Theoretical terminal velocity models are used to filter drop size distributions (DSDs) collected by measurement devices to remove physically unlikely data. This filtration process decreases the number of drops and further influences the calculation of bulk rain parameters. This study compared the impact of different drop velocity models on the DSD and rain parameter calculations for both natural and artificial rain. A total of ten natural rainfall events were observed with a Parsivel² disdrometer on the U.S. Environmental Protection Agency (EPA) Research Triangle Park (RTP) campus in Durham, North Carolina, during June–July 2024. In addition, 732 min of artificial rainfall were generated using four nozzles at the U.S. EPA Fluid Modeling Facility (FMF) in Durham, North Carolina. The results showed that for artificial rainfall, the rain rate calculated from the instruments' ‘raw’ DSD data (terminal output %93) was on average 38.73 % higher across four nozzles compared to the Parsivel²'s rain rate calculation. The rain rate calculated directly from ‘raw’ data more closely matched the volume of water collected in a bin than the Parsivel²'s rain rate calculation. For natural rainfall, the research found that the investigated terminal velocity models performed similarly. The average reduction in drop count across the six drop velocity models tested was 0.14 %, with the highest reduction observed in the Atlas (1977) model at 0.36 %. For the accumulated DSD of ten rainfall events, the ranking of models by the number of filtered drops, from highest to lowest, was: Atlas (1977) > Uplinger (1981) > Atlas et al. (1973) > Beard (1976) > Van Dijk et al. (2002) > Gunn and Kinzer (1949). In contrast to artificial rain, both rain intensity and kinetic energy measured by Parsivel² were higher than those calculated from the ‘raw’ DSD, and applying a filter reduced these values. The study also identified the microphysical characteristics of the natural precipitation, with peak values of liquid water content (W) and radar reflectivity factor (Z) observed at 0.6–0.7 g m⁻³ and 30 dBz, respectively. The average mass-weighted diameter D_m (normalized intercept parameter log₁₀N_w) values for convective, mixed, and stratiform rain were 1.81 mm (5.40), 1.15 mm (4.18), and 1.35 mm (4.53), respectively. Filters had a more pronounced effect on the D_m values compared to the log₁₀N_w values and appeared to have a larger influence on convective rain compared to stratiform rain. The study illustrated how various theoretical terminal drop velocity models used to filter the number of raindrops in a DSD affect the calculation of bulk rain parameters and how they are applicable to natural rainfall but not artificially produced rainfall. This research is valuable for gaining a deeper understanding of precipitation microphysics and improving the accuracy of rainfall assessments.