BharatFS
Regional NWP · National centre (demo proxy)
4 km
0–72 h
31%
80
- Convective rainfall
- Monsoon cores
- High resolution
Model Sources
Five forecast sources with different resolutions, horizons and strengths. Current weights reflect the active context.
NCUM
Global NWP (unified) · National centre (demo proxy)
12 km
0–240 h
27%
76
- Synoptic systems
- Coastal lows
- Stable bias
GFS
Global NWP · NOAA (demo proxy)
25 km
0–384 h
14%
56
- Long lead time
- Temperature fields
- Wide coverage
GEFS
Global ensemble · NOAA (demo proxy)
25 km / 31 members
0–384 h
20%
66
- Uncertainty spread
- Probabilistic signal
- Regime shifts
AI Forecast Proxy
Data-driven (ML) · BharatCast research (demo)
25 km
0–120 h
8%
49
- Fast refresh
- Bias correction
- Pattern recall
Reading these cards
How to interpret the ecosystem view.
- · Weight is contextual — it changes with region, regime and lead time.
- · Contextual skill is a 0–100 composite of RMSE, bias, POD, FAR, CSI and ETS.
- · A source with lower skill still keeps a 5% floor so the blend stays robust.
Model descriptions are high-level. Every skill number here is simulated demonstration data for the prototype.