Terrain-Modulated Multi-Polarimetric Fusion for Efficient Volumetric Radar Nowcasting

Volumetric radar nowcasting requires models to represent storm evolution across time and altitude without incurring prohibitive computational cost. Conventional two-dimensional approaches omit explicit vertical structure, whereas many three-dimensional models rely on computationally expensive volumetric operators. Reflectivity-only forecasting further fails to exploit complementary information from auxiliary radar variables and terrain. We therefore propose TerraCast3D, an end-to-end framework for deterministic gridded volumetric radar nowcasting conditioned on three polarimetric variables, namely differential reflectivity ZDR, specific differential phase KDP, and copolar correlation coefficient ρHV, together with Doppler spectrum width σV and terrain. TerraCast3D combines source-specific altitude-grouped encoding, terrain-modulated gated fusion, terrain-conditioned modulation of low-altitude reconstruction features, and lightweight spatiotemporal modeling to predict future reflectivity volumes while limiting the use of dense 3D operations. Experiments using GridRad-Severe volumetric radar data from 2019 to 2023 show that TerraCast3D achieves the highest CSI and HSS among the evaluated 3D models at all examined reflectivity thresholds while requiring substantially fewer FLOPs. At 40 dBZ, its full-volume CSI and HSS reach 0.3044 and 0.4501, respectively. Ablation experiments show that auxiliary radar variables mainly improve full-volume continuous-field accuracy and categorical skill at 20 and 30 dBZ, whereas joint auxiliary radar and terrain conditioning yields the highest 40 dBZ and threshold-averaged categorical scores. These findings show that altitude-aware representation and adaptive multi-source conditioning can strengthen intense-echo prediction while maintaining low arithmetic complexity in volumetric radar nowcasting.

Yuankang Ye, Feng Gao, Chang Liu

Leverage3 / 5

A method or resource many groups across the field will adopt within a year.

16% of this desk’s rubric

Magnitude2 / 5

Solid incremental gain on a meaningful problem.

20% of this desk’s rubric

Evidence3 / 5

Solid: multiple benchmarks or cohorts, ablations, fair baselines, released code or data.

20% of this desk’s rubric

Novelty2 / 5

A new combination of known ideas.

10% of this desk’s rubric

Trajectory3 / 5

A clear path to scale.

14% of this desk’s rubric

Stakes3 / 5

Meaningful benefit to many people within a few years.

20% of this desk’s rubric

• Citations0 (reference 15, via openalex, Oct 30, 2026, 06:50 UTC)

• Reviewed by heuristic-v10 on Oct 30, 2026, 06:50 UTC. Paper type: empirical.
• Categories: Precipitation Measurement and Analysis, Meteorological Phenomena and Simulations, Hydrological Forecasting Using AI, Atmospheric Science, Earth and Planetary Sciences
• BRIEF, No.3 in the Climate & Energy edition of October 30, 2026.