Preclassifier

The HydroClass Preclassifier is an implementation of the JPOLE algorithm, following closely the reference. 1

The JPOLE algorithm uses the fuzzy logic approach and is the first processing step for HydroClass. It uses polarimetric variables to classify the data in the range bins as:

  • GC/AP (Ground Clutter/Anomalous Propagation)
  • BIO or biological targets
  • METEO or hydrometeor scatterers

Five radar variables are used as input to the JPOLE algorithm. These variables are:

  • Horizontal reflectivity ZH
  • Differential reflectivity Zdr
  • Cross-correlation coefficient RhoHV
  • Texture parameter of the ZH field TX(ZH)
  • Texture parameter of differential phase TX(PhiDP)

To obtain TX(ZH), the ZH data is averaged over a running average window and then the smoothed estimates of ZH are subtracted from the original values, in other words the standard difference, effectively. A similar procedure is used for computing TX(PhiDP). The length of the running window is configurable, the default being 5 bins for TX(ZH) and 10 bins for TX(PhiDP).

The following figure shows how trapezoidal membership functions are used for the JPOLE algorithm.
Figure 1. Example of Trapetzoid Function used as a Parametrization of the Membership Function in the JPOLE Algorithm

The following figures show the displays of the default JPOLE membership functions. The computation of additive rule strengths and class identification with maximum aggregation are in the literature reference.

Figure 2. Default Settings of the JPOLE Membership Functions for ZH
Figure 3. Default Settings of the JPOLE 2D Membership Functions for Zdr

The two-dimensional (2D) membership functions are expressed as 0.5 contour levels as function of reflectivity. Class compatibilities are greater than 0.5 when Zdr values fall between the regions depicted by the contours.

Figure 4. Default Settings of the JPOLE Membership Functions for RhoHV
Figure 5. Default Settings of the JPOLE Membership Functions for TX(ZH)
Figure 6. Default Settings of the JPOLE Membership Functions for TX(PhiDP)
1 Schuur,T. et al., 2003: Observations and Classifiction of Echoes with Polarimetric WSR-88D Radar. Rep. NOAA, Norman, OK, University of Oklahoma. Web: http://www.cimms.ou.edu/~schuur/jpole/ JPOLE_HCA_report_pdf.pdf.