This study presents a statistical framework for developing region-specific trapezoidal membership functions (TMFs) to support fuzzy logic-based hydrometeor classification using dual-polarization C-band Doppler Weather Radar (DWR) data. Leveraging the calibration-independent nature of the cross-correlation coefficient (\ (\: {\: }₇ₕ\) ), the framework derives localized TMFs for key polarimetric variables: reflectivity (\ (\: Z₇\) ), differential reflectivity (\ (\: Z₃ₑ\) ), and specific differential phase (\ (\: \: K₃\) ). Initial quality control removes unreliable radar returns caused by ground clutter, noise, and non-meteorological echoes using thresholds on \ (\: {\: }₇ₕ\), signal-to-noise ratio (SNR), and \ (\: Z₇\), followed by robust estimation of \ (\: K₃\). The radar data are then stratified into empirically defined \ (\: {\: }₇ₕ\) intervals corresponding to distinct hydrometeor types. Within each \ (\: {\: }₇ₕ\) -based group, the statistical frequency distributions of \ (\: Z₇\), \ (\: Z₃ₑ\), and \ (\: \: \: K₃\) are analysed to determine their probability density functions (PDFs), from which confidence intervals are derived. TMFs are constructed using the 2. 5th to 97. 5th percentile range, enabling a robust, region-specific representation of hydrometeor categories. These TMFs are integrated into a fuzzy inference system to classify each radar gate into one of six hydrometeor types: Drizzle or Very Light Rain (DLR), Light or Moderate Rain (LMR), Heavy Rain (HR), Snowflakes (SNF), Hail or Graupel (HLS), and Sleet or Melting Snow (SLT). Uncertain returns are labelled as “Unknown. ” Classification results are visualized through Plan Position Indicator (PPI) plots and TMF matrix diagrams, effectively capturing the spatial structure of hydrometeors and supporting real-time weather monitoring, Quantitative Precipitation Estimation (QPE), and hydrological forecasting.
Sahoo et al. (Sat,) studied this question.