New method fills gaps in 3-D weather radar images

RadarFill uses observations to recover missing precipitation structures in low-altitude blind zones.

Researchers developed RadarFill to fill gaps in 3-D weather radar reflectivity mosaics. These mosaics often have systematic blind zones, especially at low altitudes.

RadarFill is an observation-constrained conditional diffusion framework. This framework uses partially observed reflectivity and an observation mask to fill gaps. It enforces available observations during reverse sampling to preserve valid measurements.

The system was tested using structured synthetic missingness. This missingness was designed from altitude-dependent missing-data statistics and vertical nesting of blind zones. It was also tested on typhoon cases and convective cases.

In Fujian Province, RadarFill was compared with 90 matched volume scans from an independent phased-array radar. Within the principal 1-2 km low-level blind zones, it yielded lower reflectivity errors than conventional interpolation.

In two controlled nowcasting experiments, improvements in RadarFill-initialized fields persist into subsequent forecasts. The largest benefits occur at low altitudes and short lead times.

Key facts

  • RadarFill recovers principal precipitation structures and moderate-to-strong echoes under extensive low-level missingness.
  • The model shows broad agreement with independent observations in the principal 1-2 km low-level blind zones.
  • Improvements in forecasts persist during controlled nowcasting experiments.

All sections

Edition