17 March 2026·PIB

Government Conducts AI-Based Pilot for Local Monsoon Forecasting to Support Kharif Sowing Decisions

In this note

1. At a Glance

First-of-its-kind Government of India pilot using AI/ML weather models to deliver village-level monsoon onset forecasts to farmers for sowing decisions [1][2].
Run by Department of Agriculture & Farmers Welfare (DA&FW) with Development Innovation Lab-India (DIL-India) during Kharif 2025; results released 17 March 2026 [1].
Examinable as a fusion of agritech, AI governance, IMD modernisation and farmer extension services — relevant to GS-III (S&T + agriculture).

PIB release dated 17 March 2026 announced completion of the pilot covering parts of 13 states for Kharif 2025 sowing [1].
Forecasts dispatched via M-Kisan SMS portal to 3,88,45,214 farmers in 5 regional languages (Hindi, Odia, Marathi, Bangla, Punjabi) [1].

IMD's traditional monsoon onset declaration is regional (e.g., onset over Kerala) and not granular enough for sowing at panchayat scale.
Government had earlier (2025) flagged an AI-based weather forecasting programme reaching ~3.8 crore farmers under DA&FW outreach [2].
Kharif 2025 pilot is the operational extension — moving from advisory issuance to probabilistic, localised onset forecasts using global AI weather models [1].

Implementing Ministry: Ministry of Agriculture & Farmers Welfare, Department of Agriculture & Farmers Welfare (DA&FW) [1].
Knowledge partner: Development Innovation Lab – India (DIL-India) [1].
Models blended (open-source):
NeuralGCM — neural-network-augmented general circulation model [1].
AIFS — Artificial Intelligence Forecasting System of ECMWF (European Centre for Medium-Range Weather Forecasts) [1].
IMD historical rainfall data — 125 years [1].
Output type: Probabilistic forecast of LOCAL monsoon onset only (not seasonal rainfall quantum) [1].
Coverage: parts of 13 states, Kharif 2025 season [1].
Delivery channel: SMS via M-Kisan portal; 3,88,45,214 farmers; 5 languages — Hindi, Odia, Marathi, Bangla, Punjabi [1].
Feedback mechanism: Telephonic surveys via Kisan Call Centres in Madhya Pradesh and Bihar [1].

Scientific / Technological

Combines physics-based GCM with data-driven AI emulators (NeuralGCM, AIFS) — reflects global shift toward ML-based numerical weather prediction [1].
Use of 125 years of IMD rainfall archive as training/calibration backbone [1].

Economic / Agricultural

Targets the single most consequential Kharif decision — sowing date; mis-timed sowing causes seed loss and re-sowing costs.
Scale (~3.88 crore farmers) suggests potential for input-cost optimisation across paddy, soybean, cotton, pulses belts [1].

Administrative / Governance

Coordination across DA&FW + IMD (MoES) + external research lab + ECMWF/Google open models — multi-stakeholder agri-extension model [1].
Uses existing M-Kisan SMS rails and Kisan Call Centres rather than building new delivery infrastructure [1].

Social / Equity

Regional-language SMS (5 languages) addresses access; but digital divide and SMS literacy in small/marginal farmer cohort remains a concern.

Ethical / Data Governance

Probabilistic AI forecast issued to farmers raises questions on liability for advisory failure, algorithmic accountability and open-source model transparency.

17 March 2026 — PIB announces pilot results for Kharif 2025 [1].
2025 — Parliamentary/PIB updates flagged AI-based weather-forecasting programme reaching ~3.8 crore farmers under DA&FW [2].
ECMWF's AIFS moved to operational/data-driven status, enabling its use in pilots like this [1].

Pilot ministry: Ministry of Agriculture & Farmers Welfare (NOT Ministry of Earth Sciences) [1].
Knowledge partner: Development Innovation Lab – India [1].
NeuralGCM is an AI–physics hybrid GCM used in the pilot [1].
AIFS = Artificial Intelligence Forecasting System of ECMWF [1].
IMD historical rainfall data span used: 125 years [1].
Pilot season: Kharif 2025; covered 13 states [1].
Forecasts predicted local monsoon ONSET only, not seasonal rainfall [1].
Reach: 3,88,45,214 farmers (~3.88 crore) [1].
Delivery platform: M-Kisan SMS portal [1].
Languages: Hindi, Odia, Marathi, Bangla, Punjabi (5 regional languages) [1].
Feedback collected via Kisan Call Centres in MP and Bihar [1].
Forecast type: Probabilistic (not deterministic) [1].

GS-III: Science & Technology — applications in everyday life; Agriculture — issues of buffer stocks, e-technology for farmers.
GS-II (peripheral): Welfare schemes — delivery mechanisms (M-Kisan).
Plausible question stems: 1. "AI-driven weather models can transform Indian agriculture only if paired with last-mile extension. Discuss with reference to the 2025 DA&FW monsoon-onset pilot." 2. "Examine the role of probabilistic forecasting and open-source AI models (NeuralGCM, ECMWF AIFS) in modernising IMD's services to farmers." 3. "Critically evaluate the institutional, ethical and data-governance challenges of issuing AI-generated agricultural advisories to crore-scale farmer audiences."

India Meteorological Department (IMD) — under Ministry of Earth Sciences; nodal weather agency.
Mission Mausam (2024) — MoES initiative to upgrade weather forecasting.
M-Kisan & Kisan Call Centres — agri-extension digital infrastructure.
PM-KISAN & PM Fasal Bima Yojana — risk-management ecosystem for Kharif farmers.
Digital Agriculture Mission / AgriStack — data backbone for targeted advisories.
IndiaAI Mission (MeitY) — national AI compute & application framework.
ECMWF & WMO — international weather modelling collaborations.
NeuralGCM / GraphCast / FourCastNet — AI weather model family.

Wrong ministry: pilot is by MoA&FW, NOT MoES/IMD — though IMD data is used [1].
AIFS belongs to ECMWF, not IMD or Google.
Forecast predicted only LOCAL onset, not total seasonal rainfall — easy to misstate.
DIL-India ≠ DPIIT or NITI Aayog; it is an external research lab partner [1].
Reach figure is ~3.88 crore farmers, often confused with the 12 crore PM-KISAN universe.