understanding polarized radar data
The raw data comes from a csv file which comes from a variable number of radar
Info comes from from the competition's data page and from this training.
- TimeToEnd: How many minutes before the end of the hour was this radar observation?
- DistanceToRadar: Distance between radar and gauge. This value is scaled and rounded to prevent reverse engineering gauge location (see notebook distancetoradar.ipynb for more information)
- Composite: Maximum reflectivity in vertical volume above gauge
- HybridScan: Reflectivity in elevation scan closest to ground
- Reflectivity: In dBZ see notebook reflectivityAvgNrange.ipynb
- ReflectivityQC: Quality-controlled reflectivity
- Velocity: (aliased) Doppler velocity
- LogWaterVolume: How much of radar pixel is filled with water droplets?
- MassWeightedMean: Mean drop size in mm
- MassWeightedSD: Standard deviation of drop size
- RR1: Rain rate from HCA-based algorithm (HCA: Hydrometeor Classification Algorithm)
- Expected: the actual amount of rain reported by the rain gauge for that hour.
- Zdr: Differential reflectivity in dB. Horizontal - Vertical reflectivity.
- Kdp: Specific differential phase. It's a local variable, defined as the correlation coefficient slope (as calculated at two distance ranges). Note: KDP is "displayed" (calculated?) only when RhoHV>0.9. WARNING: this variable need to be modified see this forum entry
- RR2: Rain rate from Zdr-based algorithm
- RR3: Rain rate from Kdp-based algorithm
- RhoHV: Correlation coefficient. Normalized sum of all v(hpol) - v(vpol) in the over all the radar pulses.(hpol/vpol are horizontal/vertical polarization). Should be between 0,1 but above 1 is related to noise it should not be trunckated to 1. Pure rain should be very close to 1, while many different scatteres ~0 (i.e. no correlations). Effectively, RhoHV<0.8 is so uncorrlated, it usually not meteorological e.g. birds. RhoHV>0.97 is usually pure rain or snow.
- RadarQualityIndex: A value from 0 (bad data) to 1 (good data)