What problem does it solve? Running the pySTEPS precipitation nowcasting library correctly requires operational expertise: converting radar reflectivity to rainfall rates, estimating optical-flow motion fields, configuring ensemble parameters, and verifying forecast skill. This Skill packages that expertise so an AI agent can execute, check, and debug real pySTEPS nowcast workflows without substituting approximations. ## Core Features & Use Cases - Five-stage executable pipeline: import radar data, estimate motion fields, run extrapolation/STEPS/ANVIL nowcasts, verify against held-out frames, and export results to NetCDF, GeoTIFF, or CSV. - Diagnostic recovery system: 20 symptom-diagnosis-remedy triplets covering unit mismatches, NaN propagation, memory limits, and projection errors, plus a preflight check that validates the environment before any run. - Validated synthetic test case: an analytical advection benchmark with known ground truth reporting CSI, POD, FAR, and FSS scores against cited literature conventions. - Use Case: Ask the agent to nowcast the next hour of rainfall from a sequence of OPERA radar frames; it converts dBZ to mm/h, estimates the motion field, runs the ensemble nowcast, and scores the result per lead time. ## Quick Start Run the preflight check and synthetic advection diagnostic in this Skill's directory to verify the pySTEPS environment and produce a validated nowcast with skill scores.
