| Type | Lead Time | Key Method | UPSC Relevance |
|---|---|---|---|
| Nowcasting | 0–3 hours | Radar / Satellite imagery | Cyclone / flood alerts |
| Short-Range Forecast (SRF) | 1–3 days | NWP models | Daily weather warnings |
| Medium-Range Forecast (MRF) | 3–10 days | NWP + ensemble models | Agriculture planning |
| Extended-Range Forecast (ERF) | 10–30 days | Coupled ocean-atmosphere models | Monsoon advance tracking |
| Long-Range Forecast (LRF) | Seasonal (months) | Statistical + MME models | Annual monsoon prediction |
| AI-Enabled Block-Level Forecast ★ New | Up to 4 weeks | AI + NWP + Statistical hybrid | ★ Current Affairs 2026 |
| Term | Full Form / Meaning |
|---|---|
| IMD | India Meteorological Department — under Ministry of Earth Sciences |
| IITM | Indian Institute of Tropical Meteorology — Pune |
| NCMRWF | National Centre for Medium Range Weather Forecasting — Noida |
| NWP | Numerical Weather Prediction — physics-based computational models |
| MME | Multi-Model Ensemble — combines multiple climate models for better accuracy |
| MMCFS | Monsoon Mission Climate Forecast System — IMD's coupled ocean-atmosphere model |
| LPA | Long Period Average — 87 cm rainfall (1971–2020 baseline); used to define normal monsoon |
| OLR | Outgoing Longwave Radiation — must be <200 W/m² for monsoon onset declaration |
| MJO | Madden-Julian Oscillation — tropical 30–60 day cycle affecting monsoon variability |
| IOD | Indian Ocean Dipole — sea surface temperature difference W vs E Indian Ocean; +ve IOD aids India monsoon |
| ENSO | El Niño–Southern Oscillation — El Niño weakens; La Niña strengthens Indian monsoon |
| AgriStack | India's digital agricultural data platform — disseminates IMD forecasts to farmers |
India's LPA (Long Period Average) for monsoon rainfall = 87 cm, calculated over 1971–2020. Normal monsoon = 96%–104% of LPA. The 2026 forecast is 92% of LPA — classified as "below normal."
Students confuse LRF (Long-Range Forecast) with the new AI system. The LRF predicts seasonal totals months in advance; the new AI system predicts monsoon progression at block level, week by week, up to 4 weeks ahead — these are different products. Also, IMD ≠ IITM ≠ NCMRWF — they are separate bodies that collaborated to build the new system.
| Act / Policy | Year | Provision / Significance |
|---|---|---|
| Meteorological Act, 1934 | 1934 | Principal legislation governing meteorological observations and forecasting services in India |
| National Disaster Management Act | 2005 | Mandates early warning systems; IMD is the nodal agency for hydro-meteorological disaster warnings |
| Disaster Management Act | 2005 | Heat waves NOT classified as natural disasters under this Act — common UPSC trap |
| Ministry of Earth Sciences Act / MoES | 2006 | IMD placed under MoES; umbrella for IITM, NCMRWF, INCOIS, NIOT |
| National Action Plan on Climate Change (NAPCC) | 2008 | Mandates improved climate/weather forecasting for adaptation; IMD contributes to National Water Mission |
| National Monsoon Mission (NMM) | 2012 | Launched to develop coupled ocean-atmosphere models; resulted in MMCFS; budget ~₹400 crore |
| Mission Mausam | 2024–25 | Cabinet-approved Sep 2024; ₹2,000 crore outlay; builds on NMM; AI + next-gen radars + HPC |
| IMD Vision 2047 | Jan 2025 | Zero-error forecast up to 3 days; 90% accuracy up to 5 days; zero weather disaster deaths by 2047 |
| AI-Enabled Block-Level Forecast | May 2026 | First-ever AI product under Mission Mausam umbrella; covers 16 states, 3000+ sub-districts |
| Role | Framework | Note |
|---|---|---|
| Principal national meteorological service | WMO membership (since 27 Apr 1949) | One of 6 Regional Specialised Meteorological Centres (RSMC) |
| Cyclone naming & warning — North Indian Ocean | WMO / RSMC New Delhi | Covers Bay of Bengal, Arabian Sea, Persian Gulf, Malacca Strait |
| Hydro-meteorological early warnings | NDMA coordination | Colour-coded alerts: Green / Yellow / Orange / Red |
| Agro-meteorological advisories | MoA&FW collaboration | New AI system disseminated via AgriStack & Ministry of Agriculture APIs |
| Seismic monitoring | Meteorological Act 1934 | IMD also monitors earthquakes — secondary role often asked in Prelims |
Mission Mausam approved by Union Cabinet in September 2024 with ₹2,000 crore outlay over 2 years. Implemented by IMD + IITM + NCMRWF — same three bodies that built the May 2026 AI forecast system.
Heat waves are NOT classified as natural disasters under the Disaster Management Act, 2005 — a repeated UPSC fact. Also, INCOIS (Indian National Centre for Ocean Information Services) is under MoES but is NOT part of the AI monsoon forecast system — only IMD + IITM + NCMRWF are.
| Country / Agency | AI System | Key Feature | Year Operational |
|---|---|---|---|
| India (IMD) | AI Block-Level Monsoon Forecast | Block-level onset, 4-week lead, 3000+ sub-districts | 2026 |
| ECMWF (Europe) | AIFS (Artificial Intelligence Forecasting System) | Graph Neural Network (GNN); 1,000x less energy vs NWP; +20% accuracy for cyclone tracks | Feb 2025 |
| Google DeepMind | GraphCast / GenCast | 10-day global forecast in <1 minute; 25 km resolution ensemble | 2023–24 |
| NVIDIA | FourCastNet | 0.25° resolution; trained on ERA5 reanalysis data | 2022 |
| Huawei | Pangu-Weather | 3D Earth deep learning; outperforms IFS on many metrics | 2023 |
| USA (NOAA) | GFS + AI enhancement | Global Forecast System + ML post-processing | Ongoing |
IMD's forecast accuracy has improved by 40% across severe weather events since 2014. Cyclone deaths dropped from 10,000 in 1999 to zero in 2020–2024 — credited to accurate IMD cyclone warnings.
| # | Parameter | Threshold | Key Detail |
|---|---|---|---|
| 1 | Rainfall | ≥2.5 mm for 2 consecutive days | At ≥60% of 14 designated stations (Thiruvananthapuram, Kochi, Kozhikode, Minicoy etc.); declared after 10 May |
| 2 | Wind (Westerlies Depth) | Maintained up to 600 hPa | Zonal wind 15–20 knots at 925 hPa in box Lat 5–10°N, Long 70–80°E; sustained westerlies in Arabian Sea box |
| 3 | OLR (Outgoing Longwave Radiation) | <200 W/m² | INSAT-derived; box Lat 5–10°N, Long 70–75°E; indicates adequate moisture and cloudiness |
| Infrastructure | Number | Note |
|---|---|---|
| Doppler Weather Radars (DWR) | 37 (2023) → 62+ (target) | Was 15 in 2013; covers cyclone detection + rainfall estimation |
| Automatic Weather Stations (AWS) | Thousands (district-level) | AGRO AWS: 200 dedicated for agriculture; DAMUs expanding to 660 |
| Automatic Rain Gauges (ARG) | 1,382+ (2023) | Up from 1,350 in 2014 |
| DRMS (District Rainfall Monitoring Stations) | 5,896 (2023) | Up from 3,955 in 2014 |
| Regional Meteorological Centres | 6 | Chennai, Guwahati, Kolkata, Mumbai, Nagpur, New Delhi |
| HPC (Pratyush + Mihir supercomputers) | 6.8 PFLOPS total | At IITM Pune + NCMRWF Noida; support advanced climate modelling |
| Satellites used | INSAT-3D, INSAT-3DR, Kalpana-1, Megha-Tropiques | Geostationary (INSAT) + LEO missions; INSAT = first in developing world (1982) |
The block-level onset forecast is based on a continuous 5-day rainfall spell + absence of prolonged dry spells over the subsequent 30 days — per M. Ravichandran, Secretary MoES.
Parliament Question on Monsoon Prediction (MoES PIB, Feb 2025): IMD's absolute forecast error for all-India seasonal rainfall reduced by 21% during 2007–2024 compared to 1989–2006. Average absolute error (2015–24) = 5.01% of LPA vs 5.97% (2005–14). Correlation between actual and forecast rainfall rose from 0.37 (2005–14) to 0.61 (2015–24).
IMD Vision 2047 (released by PM Modi, 14 Jan 2025): Titled "Har Har Mausam, Har Ghar Mausam". Targets: zero-error forecast up to 3 days; 90% accuracy up to 5 days; 100% severe weather detection at village/household level; zero disaster deaths by 2047. Announced alongside Mission Mausam (₹2,000 crore).
NMM (2012) — Ministry of Earth Sciences: Aimed to build state-of-the-art coupled ocean-atmosphere models. Implemented by IITM, INCOIS, NCMRWF. Led to development of MMCFS (Monsoon Mission Climate Forecast System). Focus: extended-range (11 days to 1 season) and short-to-medium range (up to 10 days) forecasts. Budget ~₹400 crore. Considered precursor to Mission Mausam.
Stanford Research (Sep 2025): Evaluated 7 global AI weather models (FourCastNet, Pangu-Weather, GraphCast, Aurora, AIFS, GenCast etc.) for South Asian Monsoon. Finding: AI models handle large-scale dynamics well but "fall short on key metrics critical to Monsoon-time weather prediction." ECMWF's AIFS ranked best, with GraphCast and GenCast close seconds. IMD's new hybrid approach (AI + NWP + statistical) responds to these limitations.
ECMWF Artificial Intelligence Forecasting System (Feb 2025): First fully operational ML-based global weather prediction open model; Graph Neural Network encoder-decoder architecture; 1,000x reduction in energy vs traditional NWP; +20% accuracy on tropical cyclone track forecasting. Runs alongside traditional IFS — sets the global AI weather forecasting benchmark.
UPSC has asked about IMD's colour-coded warning system (UPSC 2022 Mains) and monsoon forecasting systems in Prelims. Focus on: NMM vs Mission Mausam (budget, year, objective), the three monsoon onset criteria, and the institutional triad IMD + IITM + NCMRWF. The 2026 AI launch is likely to appear in 2026 Prelims.
| Feature | Detail | Significance |
|---|---|---|
| Lead Time | Up to 4 weeks (28 days) | Longest block-level monsoon prediction in India |
| Spatial Scale | Block / sub-district level | Granular enough for farm-level decisions |
| Coverage | 16 states; 3,000+ sub-districts | Phase-1 rollout; to expand with infrastructure |
| Frequency | Every Wednesday | Regular weekly forecast cycle |
| Methodology | NWP models + AI + Statistical techniques | Hybrid approach overcomes pure-AI monsoon limitations |
| Output Type | Probabilistic forecasts of monsoon progression | Shows probability of onset, not just a binary date |
| Error Margin | ~±4 days | Operationally useful for planning; stated by MoS Jitendra Singh |
| Onset Definition | Continuous 5-day rainfall spell + no prolonged dry spell for 30 days | Scientifically robust definition avoids "bogus onset" declarations |
| Data Sources | Doppler radars, satellites, AWS, Automatic Rain Gauges | Multi-source data fusion improves accuracy |
| Developers | IMD + IITM Pune + NCMRWF Noida | India's meteorological triad |
| Dissemination | Ministry of Agriculture APIs + AgriStack platform | Direct farmer access; end-to-end digital delivery |
| Feature | Detail |
|---|---|
| Developer | NCMRWF (National Centre for Medium Range Weather Forecasting), Noida |
| Spatial Resolution | 1 km grid — highest operational rainfall resolution at scale in India |
| Lead Time | Up to 10 days |
| Technique | AI-driven downscaling from coarser global models to 1-km local resolution |
| Data Inputs | Automatic Rain Gauges (ARG), AWS, Doppler Weather Radars (DWR), satellite rainfall datasets |
| Pilot State | Uttar Pradesh (India's most populous state — high agricultural significance) |
| Expansion Plan | To other states as observational infrastructure grows — per MoES Secretary |
| Use Cases | Sowing, irrigation, crop protection, harvesting decisions; urban flood management |
The system is not a seasonal forecast — it tracks monsoon progression week by week. The IMD confirmed the new models have "no correlation with the seasonal forecast" (LRF for Jun–Sep totals). Two different products; two different purposes.
| Linked Concept | Connection | Exam Angle |
|---|---|---|
| Climate Change | Monsoon variability is increasing due to warming; AI needed for more complex forecasting | Why AI forecasting is needed now |
| Food Security / Agriculture | 70% of Indian agriculture is rainfed; better monsoon forecasting = better sowing/harvesting decisions | Development / GS-III angle |
| Disaster Management | Accurate monsoon onset prediction reduces flood/drought damage; NDMA coordination | Prelims: NDMA Act 2005 |
| ENSO / IOD / MJO | These global oscillations drive Indian monsoon variability; AI models learn their signatures | Most tested ocean-atmosphere linkage in Prelims |
| Digital India / AgriStack | AI forecast disseminated via digital agricultural platforms; farmer-centric tech delivery | GS-III: Technology in agriculture |
| India's Space Programme (ISRO) | INSAT, Kalpana-1 satellites feed IMD's observational network; ISRO–IMD collaboration | Science & Technology × Geography crossover |
| Western Disturbances | Better AI forecasting useful for winter rainfall predictions in north India too | Geography Prelims static topic |
| Mascarene High / Somali Jet | Ocean-atmosphere dynamics that drive SW monsoon; AI models incorporate these forcings | Monsoon mechanism — standard Prelims topic |
| Agency | Country | Special Role | UPSC Relevance |
|---|---|---|---|
| IMD | India | RSMC – North Indian Ocean; under MoES | ★★★ Core topic |
| ECMWF | European Centre | Best global medium-range NWP; AIFS operational Feb 2025 | ★★ Frequently mentioned |
| NOAA / NWS | USA | GFS model; hurricane tracking; ENSO monitoring | ★★ ENSO context |
| JMA | Japan | Typhoon warnings; Western Pacific RSMC | ★ Global compare |
| WMO | UN Body (Geneva) | Governs all national met services; India member since 1949 | ★★★ Frequently asked |
| CMA (China) | China | Pangu-Weather AI model (2023); Eastern Hemisphere focus | ★ AI compare |
Pratyush (IITM, Pune) + Mihir (NCMRWF, Noida) = India's dedicated weather HPC systems. Combined capacity = 6.8 PFLOPS. Both are among the most powerful government HPC systems in India, dedicated exclusively to weather and climate modelling.
UPSC frequently combines topics in 2-statement questions. Know these pairings: El Niño → weakens Indian monsoon (generally); +ve IOD → strengthens; La Niña → above-normal monsoon (generally); MJO → intra-seasonal variability. The new AI system explicitly monitors all these via its training data.
India's first AI-enabled monsoon advance forecast system launched on 12 May 2026 by Union MoS (Earth Sciences) Dr. Jitendra Singh. The block-level system combines NWP with AI to generate probabilistic forecasts every Wednesday up to 4 weeks ahead, covering 16 states and 3,000+ sub-districts. Model error margin: ~±4 days. Outputs shared via Ministry of Agriculture's AgriStack platform.
Alongside the block-level system, NCMRWF launched a 1-km resolution rainfall forecast pilot for Uttar Pradesh — the first such hyperlocal product in India at operational scale. Spatial resolution: 1 km grid. Lead time: up to 10 days. Technology: AI-driven downscaling techniques integrated with Automatic Rain Gauges (ARG), AWS, Doppler radars, and satellite rainfall datasets. Expansion to other states is planned as observational infrastructure grows.
IMD's 2026 Southwest Monsoon Forecast (first stage, April 2026): Seasonal (Jun–Sep) rainfall projected at 92% of LPA — classified as below normal (LPA = 87 cm; normal = 96%–104%). Model error: ±5%. Reason cited: possible El Niño development during the monsoon season (MMCFS/coupled model signals); La Niña-like conditions transitioning. IOD: currently neutral, positive IOD likely by end of season. IMD stated the new AI models have no correlation with this seasonal forecast — separate products.
Conditions becoming favourable for Southwest Monsoon advance over South Bay of Bengal, Andaman Sea and Andaman & Nicobar Islands around 16 May 2026 — per IMD Press Release dated 14 May 2026. Normal onset over Kerala is 1 June. The 2025 monsoon had arrived 8 days early (24 May 2025). The new AI system will track 2026 monsoon progress week-by-week once onset occurs.
Mission Mausam officially launched by PM Modi on 14 January 2025 (IMD's 150th Foundation Day) at Bharat Mandapam, New Delhi. Cabinet approval: September 2024; Budget: ₹2,000 crore over 2 years; Implemented by IMD + IITM + NCMRWF. Alongside, IMD Vision 2047 document titled "Har Har Mausam, Har Ghar Mausam" released — targets zero-error forecasts up to 3 days and zero disaster deaths by 2047. ECMWF's AIFS also became operational in February 2025 — global AI weather forecasting benchmark.
The 2026 AI monsoon launch is a high-probability Prelims 2026 topic. Expected question formats: (1) "Which agency developed India's first block-level AI monsoon forecast system?" → answer: IMD + IITM + NCMRWF jointly. (2) Statement-type: "The system covers all states of India" → False (only 16 in Phase 1). (3) "Forecasts are issued daily" → False (every Wednesday). Always remember: ±4 days error margin; 4-week lead time.
| # | Statement | T/F | Reason |
|---|---|---|---|
| 1 | The AI-enabled block-level monsoon advance forecast covers all 28 states of India from launch. | ❌ | Only 16 states in Phase 1; expansion planned as infrastructure grows. |
| 2 | The new IMD AI system generates forecasts every Wednesday for up to 4 weeks in advance. | ✅ | Correct. Wednesday is the fixed forecast day; lead time is 4 weeks (28 days). |
| 3 | IMD was established after the 1864 Calcutta cyclone and the 1866 and 1871 monsoon failures. | ✅ | Correct. Established 15 January 1875; H.F. Blanford was first Meteorological Reporter. |
| 4 | IMD is under the Ministry of Agriculture and Farmers Welfare. | ❌ | IMD is under the Ministry of Earth Sciences (MoES), not Agriculture. |
| 5 | A positive IOD typically results in below-normal monsoon rainfall over India. | ❌ | Positive IOD (warmer western Indian Ocean) generally enhances Indian monsoon rainfall. |
| 6 | The LPA (Long Period Average) for Indian monsoon is 87 cm, calculated over 1971–2020. | ✅ | Correct. IMD revised LPA from 89 cm (1961–2010 baseline) to 87 cm (1971–2020 baseline). |
| 7 | ECMWF's AIFS model became operational in February 2025 and uses Graph Neural Networks. | ✅ | Correct. ECMWF-AIFS = GNN architecture; operational 25 Feb 2025; 1000x less energy than NWP. |
| 8 | The 1-km resolution UP pilot was developed by IITM Pune. | ❌ | Developed by NCMRWF (National Centre for Medium Range Weather Forecasting), Noida — not IITM. |
| 9 | For monsoon onset over Kerala, OLR must be greater than 200 W/m². | ❌ | OLR must be less than 200 W/m² (lower OLR = more cloudiness/moisture = monsoon onset). |
| 10 | Mission Mausam was approved by the Union Cabinet in September 2024 with ₹2,000 crore outlay. | ✅ | Correct. Cabinet approval Sep 2024; implemented by IMD + IITM + NCMRWF; 2-year programme. |
IMD ≠ Ministry of Agriculture. IMD is under Ministry of Earth Sciences (MoES). The confusion arises because the new AI forecast outputs are disseminated through Ministry of Agriculture's AgriStack platform — but IMD itself is under MoES.
Old LPA = 89 cm (1961–2010 baseline); New LPA = 87 cm (1971–2020 baseline). UPSC questions sometimes quote older values. Always use 87 cm and 1971–2020 for current exams.
Positive IOD = warmer western Indian Ocean, cooler eastern = draws more moisture to India = good for monsoon. Negative IOD = opposite = suppresses Indian monsoon. El Niño (warm Pacific) generally weakens Indian monsoon; La Niña generally strengthens it.
Heat waves are NOT officially classified as natural disasters under the Disaster Management Act, 2005 — despite their severe impact. This is a frequently repeated UPSC fact. Only recently has there been discussion about reclassifying them.
IMD (New Delhi RSMC) is responsible for tropical cyclone naming and warnings for the North Indian Ocean — this includes Bay of Bengal, Arabian Sea, Persian Gulf, and Malacca Strait. It does NOT cover the South Indian Ocean (covered by La Réunion/Mauritius) or South Pacific.
IMD and monsoon topics appear in multiple question types: (1) Institution-ministry matching (IMD under which ministry?); (2) Rainfall category statements (normal = 96–104% LPA); (3) Monsoon onset criteria (3-parameter check); (4) Ocean oscillation effects (ENSO/IOD/MJO); (5) Current affairs integration (Mission Mausam, AI system 2026). Practice 2-statement combo formats.
Questions 1 and 5 test the same fact from different angles — know that IMD + IITM + NCMRWF is the key triad for both Mission Mausam and the 2026 AI system. INCOIS is a distractor. This "three-institution" pattern is a favourite UPSC format for science-based current affairs.
| Category | Key Fact | Number / Year |
|---|---|---|
| System Launch | AI Block-Level Monsoon Forecast | 12 May 2026 |
| Lead Time | Block-level onset prediction | 4 weeks (±4 days error) |
| Coverage | States / Sub-districts (Phase 1) | 16 states / 3,000+ |
| UP Pilot Resolution | NCMRWF high-res rainfall | 1 km / 10 days |
| LPA (current) | Long Period Average for monsoon | 87 cm (1971–2020) |
| 2026 Monsoon Forecast | Below normal (IMD) | 92% of LPA (±5%) |
| Mission Mausam Budget | Union Cabinet Sep 2024 | ₹2,000 crore (2 years) |
| IMD Established | After 1864 cyclone + failures | 15 Jan 1875 |
| HPC Capacity | Pratyush (IITM) + Mihir (NCMRWF) | 6.8 PFLOPS total |
| Doppler Radars | Network expansion | 15 (2013) → 37 (2023) → 62+ (target) |
| RSMC Role | Cyclone naming — North Indian Ocean | Since 1988 |
| WMO Membership | India joined WMO | 27 Apr 1949 |
| ECMWF AIFS (Global benchmark) | AI forecasting — operational | 25 Feb 2025 |
| Cyclone Deaths | Accuracy impact | 10,000 (1999) → 0 (2020–24) |
| Onset — Kerala Criteria | Three parameters needed | Rainfall + Wind (600 hPa) + OLR (<200 W/m²) |