IMD’s new AI weather forecasts will tell farmers when it will rain within a 1-km area

The Ministry of Earth Sciences launched two AI-enabled weather forecast products under the India Meteorological Department (IMD): a block-level monsoon onset...

What Happened

The Ministry of Earth Sciences launched two AI-enabled weather forecast products under the India Meteorological Department (IMD): a block-level monsoon onset forecasting system, and a high-spatial-resolution rainfall forecast system for Uttar Pradesh generating predictions at 1-km spatial resolution up to 10 days in advance.
The 1-km rainfall forecast system uses AI-driven downscaling techniques that integrate data from automatic rain gauges, automatic weather stations, Doppler weather radars, and satellite-based rainfall datasets to produce hyperlocal precision forecasts.
Uttar Pradesh was selected as the pilot state due to its dense network of weather observation infrastructure; the system will be expanded to other states as observational infrastructure grows.
Forecasts are intended to enable farmers to make more informed decisions on sowing, irrigation, crop protection, and harvesting at a field-level resolution.
Outputs will be disseminated through APIs integrated with the Agri Stack platform, allowing third-party agricultural advisory apps and state governments to access and relay the forecasts.

Static Topic Bridges

AI and Machine Learning in Weather Forecasting — Downscaling Techniques

Traditional Numerical Weather Prediction (NWP) models operate at grid resolutions of 10–50 km, which is too coarse for farm-level decision-making. AI/ML-based statistical downscaling is a technique that learns the relationship between coarse-resolution NWP output and high-resolution observed data (from dense rain gauges, radars, and satellites) to generate fine-resolution predictions. Deep learning models — particularly convolutional neural networks (CNNs) and transformer-based architectures — have demonstrated skill in downscaling precipitation fields to 1–5 km resolution.

AI downscaling is computationally far cheaper than running full NWP at 1-km resolution, which would require prohibitive supercomputing resources.
IMD's 1-km system integrates multiple data streams: Automatic Weather Stations (AWS), Automatic Rain Gauges (ARG), Doppler Weather Radars (DWR), and satellite products (e.g., INSAT-3D precipitation estimates, GPM IMERG).
Global precedents: Google's GraphCast and DeepMind's NowcastNet have demonstrated AI-based weather prediction superior to traditional NWP at short ranges; IMD's 1-km system represents India's operational move into this domain.
Key challenge: AI models require dense observational data for training; data-sparse regions remain a limitation, which is why UP (with dense observation network) was chosen as the pilot.

Connection to this news: The 1-km resolution system is a direct operational deployment of AI downscaling methodology, moving India's agricultural weather services from district-level (roughly 50–100 km) to near-field resolution.

Precision Agriculture and Digital Public Infrastructure for Farmers

Precision agriculture is the application of information technology, IoT, remote sensing, and data analytics to optimise agricultural inputs (water, fertiliser, pesticides) at the finest spatial scale possible — ideally at the level of individual field plots. Hyperlocal weather forecasting is the foundational layer of precision agriculture, as crop management decisions depend critically on micro-climate and local rainfall timing rather than regional averages.

Agri Stack is India's Digital Public Infrastructure (DPI) for agriculture — a federated ecosystem comprising a Farmer Registry, Geo-referenced Village Maps, and Crop Sown Registry, announced under the Digital Agriculture Mission.
Agri Stack APIs enable third-party agri-tech apps, state governments, and cooperative platforms to query and display IMD forecast data at the farm/block level.
PM-KISAN, Pradhan Mantri Fasal Bima Yojana (PMFBY), and Kisan Credit Card (KCC) databases are integrated within the Agri Stack framework, enabling targeted advisory services.
The system enables "impact-based forecasting" — not merely predicting weather conditions but translating them into agronomic recommendations (e.g., "delay sowing by 3 days", "apply fungicide given high humidity forecast").
India's agricultural sector employs ~45% of the workforce and contributes ~18% of GDP, making precision weather services a significant economic policy lever.

Connection to this news: The 1-km IMD forecast, disseminated via Agri Stack APIs, is the data backbone of India's precision agriculture transition — connecting the Ministry of Earth Sciences' S&T capability with the Ministry of Agriculture's last-mile farmer advisory infrastructure.

India's Weather Observation Infrastructure — Ground Truth for AI Systems

AI weather forecasting models are only as good as the observational data they are trained and validated on. India has significantly expanded its surface weather observation infrastructure in the past decade. The Doppler Weather Radar (DWR) network provides real-time 3D wind and precipitation data within a ~400-km radius, while Automatic Weather Stations (AWS) and Automatic Rain Gauges (ARG) provide high-frequency surface measurements.

Doppler Weather Radars: 15 in 2013 → 50 in 2026; target is 60+ under the Radar Expansion Programme.
IMD operates over 800 Automatic Weather Stations (AWS) and approximately 1,300 Automatic Rain Gauges (ARG) across India.
INSAT-3D and INSAT-3DR: dual-payload geostationary satellites providing every 15-minute imagery for cloud, temperature, water vapour, and precipitation products.
IMDAA (IMD Meteorological Data Assimilation and Analysis): India's high-resolution reanalysis dataset at 12-km resolution, used as training data for AI models.
Uttar Pradesh's selection reflects a deliberate expansion strategy: dense observation networks first, AI product development second, nationwide rollout third.

Connection to this news: The 1-km resolution system's geographic rollout pathway is directly gated by the density of AWS/ARG and radar coverage — the infrastructure investment of the past decade is now enabling AI-powered precision forecast products.

Agro-Meteorology — Policy Framework and Institutional Ecosystem

Agro-meteorology is the application of meteorological sciences to agriculture, covering crop-weather relationships, agro-climatic zonation, irrigation scheduling, and pest/disease forecasting. In India, the Gramin Krishi Mausam Sewa (GKMS) scheme, operated jointly by IMD and the Indian Council of Agricultural Research (ICAR), has been providing district-level agro-meteorological advisories twice weekly to farmers since 2008.

GKMS currently covers all 36 states/UTs through ~130 Agro-Meteorological Field Units (AMFUs) co-located at ICAR research stations.
Agromet advisories are disseminated via Kisan Portal, Doordarshan Kisan channel, Kisan Call Centres (KCC), and SMS to registered farmers.
The new 1-km AI forecast system will upgrade the GKMS advisory from district resolution to block/field resolution — a qualitative leap in advisory precision.
Climate change context: erratic monsoon variability is increasing; IPCC AR6 projects higher frequency of extreme rainfall events over India, making high-resolution forecasting more economically valuable over time.

Connection to this news: The IMD 1-km AI system is the technological upgrade that positions GKMS for the precision agriculture era, enabling farm-specific risk management rather than regional advisories that may not capture local variability.

Key Facts & Data

Spatial resolution: 1 km (vs. ~50–100 km for district-level NWP output previously available)
Forecast lead time: up to 10 days in advance
Pilot state: Uttar Pradesh (selected for dense observation network)
Data inputs: Automatic Rain Gauges (ARG), Automatic Weather Stations (AWS), Doppler Weather Radars (DWR), satellite-based rainfall products
Technique: AI-driven statistical downscaling of NWP model output
Dissemination: Ministry of Agriculture APIs + Agri Stack platform
IMD AWS network: 800+ stations nationwide
IMD ARG network: ~1,300 stations
IMD Doppler radars: 15 (2013) → 50 (2026)
Agricultural workforce share: ~45% of India's total employment
Agriculture share of GDP: ~18%
GKMS operational since: 2008 (district-level advisories)
India's cropped area: approximately 50% rain-fed, highly sensitive to monsoon variability