How biswajitbiswal-in/ritugyan works
RituGyan post-processes monsoon rainfall forecasts: align forecast and atmospheric grids with observed rainfall, classify the weather regime, then apply regime-conditioned correction and provide rainfall-risk outputs for users. The graph preserves the README’s documented district aggregation, verification, API, and dashboard capabilities, though their implementation is not shown in the sampled source. Batch assembly, synoptic feature extraction, and feature storage have direct source evidence; other documented subsystem paths are represented by their existing directories.
Main entry points: preprocessing.py Feature preprocessing · stage2_correction Conditional correction · batch_pipeline.py Batch assembly · readers.py Dataset readers
Components 19
Data preparation
- GFS acquisition
src/ingestion/gfs_downloader.py - Dataset readers
src/ingestion/readers.py - Data inspection
src/ingestion/inspector.py - Grid alignment
src/ingestion/spatial.py - Feature preprocessing
src/features/preprocessing.py - Batch assembly
src/features/batch_pipeline.py - Feature store
src/features/feature_store.py
Forecast processing
- Synoptic features
src/features/synoptic.py - Regime classification
src/features/regimes.py - Regime model stage
src/models/stage1_regime - Conditional correction
src/models/stage2_correction
Products and delivery
- Heavy-rain probabilities
src/models/stage2_correction - District aggregation
src - Forecast API
src/api/routes - Interactive dashboard
src/ui
Verification
- Regime-stratified verification
src/evaluation
Other components
- Forecast and observation data
- NOAA GFS
- Forecast users
Connections 23
- NOAA GFS → GFS acquisition: provides forecast data
- GFS acquisition → Dataset readers: writes forecast files
- Forecast and observation data → Dataset readers: supplies datasets
- Dataset readers → Data inspection: opens for inspection
- Dataset readers → Feature preprocessing: opens for processing
- Feature preprocessing → Grid alignment: slices common grids
- Batch assembly → Feature preprocessing: builds aligned samples
- Batch assembly → Synoptic features: extracts indices
- Batch assembly → Regime classification: classifies samples
- Batch assembly → Feature store: saves samples
- Synoptic features → Regime classification: provides features
- Feature preprocessing → Regime model stage: supplies forecast fields
- Regime classification → Regime model stage: supplies regime label
- Regime model stage → Conditional correction: provides regime
- Feature preprocessing → Conditional correction: supplies predictors
- Conditional correction → Heavy-rain probabilities: estimates exceedance risk
- Conditional correction → District aggregation: provides corrected grids
- Heavy-rain probabilities → Forecast API: serves probabilities
- District aggregation → Forecast API: serves aggregates
- Forecast API → Interactive dashboard: provides forecast service
- Interactive dashboard → Forecast users: presents forecasts
- Conditional correction → Regime-stratified verification: provides corrected forecasts
- Forecast and observation data → Regime-stratified verification: provides observations