MeteoClassifiers

The default classification system of warm and cold season hydrometeors is based on the fuzzy method approach. ¹

The implementation follows closely the updated version ², which is the recommended general reference. The main aspects are briefly explained and are followed by the features specific to the present implementation.

The method described in the main reference ² was developed using radar measurements at the CSU-CHILL Facility (for a technical description, see http://lab.chill.colostate.edu/chill-technical.html).

The CHU-CHILL facility is an S-band, Doppler radar with full polarization agility and diversity at Colorado State University, U.S. The referenced classification system represents state-of-the-art knowledge and it has been verified by comparing the CSU-CHILL measurements with the in-situ airborne observations made with instruments such as 2D cloud particle measurement probe, high volume particle sampler (HVPS) and hail spectrometer. ²

In the fuzzy method approach, the signatures of specified hydrometeor classes are quantified as a set of membership functions (MBF), which take the measured dual-polarization parameters obtained at each bin as input. The strength of each hydrometeor class is then expressed as the outcome (rule strength) of an inference function which takes the MBF values as input. The membership functions and the inference rule strength function formalize the meteorological interpretation encoded in the classification method.

The membership functions (MBF) of the observable x are parametrized in the CSU method as beta functions:

in which x stands for reflectivity (Z_H), differential reflectivity (Z_dr), specific differential phase (K_dp), cross-correlation coefficient (HV), or observation altitude (h). The expression specifies the one dimensional case (1D), while evolutions of MBF(Zdr) and MBF(Kdp) at varied reflectivities are detailed further by parametrizing $y = y ( Z H )$ where $y = m , a ⁢ , o r ⁢ b$. Such parameterization allow extending the 1D MBFs to 2D.

The membership function of the radar observation volume altitude has dependence on the melting layer height (ML), which is thus also used as input. The membership function of altitude can also be formulated as a 2D MBF, in which parameters m and a are dependent on ML.

In IRIS/RDA, the melting layer height is defined to coincide with the 0 °C isotherm, and related expressions in literature algorithms are converted to match the IRIS/RDA definition. ML can be fed in from an external source, or it is acquired from the archived RAW file headers, or it can specified explicitly in the HydroClass configuration.

The classification results are presented by labeling each bin with the hydrometeor class that is most compatible with the observations, that is by choosing the class of highest rule strength. The rule strengths are computed using the equations of the literature reference using the quoted four radar variables and the altitude as input. Threshold parameters associated to each of the rule strengths are used to specify bins for which the class is ambiguous, for example, non-meteorological targets.