The Hydrometeor Classification Algorithm is WSR-88D radar software that assigns each radar sample a best-guess echo type, such as rain, dry snow, wet snow, graupel, hail, or non-weather targets like ground clutter and birds. It compares several dual-polarization measurements at once and picks the class that fits best.
How it decides
Forecasters usually call it the HCA, and the NWS calls its output the Hydrometeor Classification (HC) product. It weighs six radar fields: reflectivity, differential reflectivity (ZDR), correlation coefficient (CC), specific differential phase (KDP), and two measures of how noisy, or textured, the reflectivity and differential phase fields are.
A fuzzy logic scheme scores every candidate class against those fields using a matrix of weights, then assigns the class with the highest combined score.
The classes
The original operational design listed ground clutter or anomalous propagation, biological targets, dry snow, wet snow, ice crystals, big drops, light to moderate rain, heavy rain, graupel, hail possibly mixed with rain, and unknown, with separate codes for no data and range-folded data. Radar Build 17.0 later added two hail size classes, large hail of 1 to 2 inches and giant hail of 2 inches or more, through the Hail Size Discrimination Algorithm.
The NWS describes the product as a first guess for precipitation type that also helps point out echoes that are not precipitation.
The melting layer step
Before classifying, the radar runs a Melting Layer Detection Algorithm that estimates the top and bottom of the layer where snow melts, mainly by finding the wet-snow signature. The HCA then applies two rules: snow is not allowed below the bottom of that layer, and pure rain is not allowed above its top, except for the big drops class.
What it cannot tell you
The HCA classifies what the beam sees at its own height, which can be well above the ground far from the radar. Shallow warm or cold layers below the lowest scan can change precipitation before it lands, and NSSL researchers note that radar data alone cannot separate rain from freezing rain.
None of the class lists in these NWS and design documents includes freezing rain or sleet. Early evaluation also found that the melting layer step may need adjustment in cold-season storms when the freezing level sits below about 3,300 feet (1 km) above ground, and partial beam blockage can degrade results.
