Atmoseer
In development

Hyperlocal Weather Nowcasting

Severe weather phenomena, minutes to hours ahead.

Atmoseer is our solution for weather nowcasting, built by a highly skilled team of experts. It brings data from our own sensors together with radar, satellite, lightning and station data into automatic nowcasting models and alerts. AI helps us tune and test both the models and the alerts.

Everyone gets the warning. Only a few get the impact.

General warnings for a wide area

The impact is a few kilometres wide, the warning covers an area many times its size.

Vague timing

It gives a window of several hours in which dangerous weather may occur locally, not when and how it reaches your place.

No follow-up

Once the warning is out, the weather is usually not tracked with a nowcast, and its impacts are not followed or investigated.

Low level of automation

A forecaster has to confirm every warning. That takes deep experience with dangerous weather and the ability to work under pressure, and quality suffers when many events happen at once.

Nowcast map of a storm cell with its projected track and arrival times

What is weather nowcasting?

Weather nowcasting is a very short-range forecast of what the weather will do in the next minutes to hours at a specific place. It starts from the latest observations: radar, satellite, lightning, weather stations and any other source that shows what the weather is doing right now.

This is the time in which a decision can still be made.

Alert the emergency services

Clear a site

Send a warning

Move the car under cover

Cover a crop

Further asset & personnel protection

Atmoseer
A warning you can act on

  • The time, the intensity and the form the weather takes at a specific place

  • Certainty that grows from the forecast to the measurement on site

  • Followed automatically, from the first signs of risk to what it actually did on the ground

All of this comes together in the Atmoseer engine. Our expert team built it and uses it to develop, test and tune its products directly on real-time data. Every result is verified in the real world.

Flash-flood index map with reported events

Forecast

Our experts build their own forecast products from weather models combined with other inputs, and keep tuning, testing and improving them. They show where the risk is, hours to days ahead.

Our flash-flood index (FFA) with ESWD reports

Radar nowcast of a storm with its projected track

Nowcasting

Our own nowcasting engine turns the latest observations and land surface conditions into a picture of what the weather will do in the coming tens of minutes, place by place, with its uncertainty quantified.

Meratch rain sensor mounted on a pole

Measurements

Radar, satellite, lightning, weather stations, our own sensors and other sources, brought together in one place. They are the raw material from which we build our products, models and solutions.

Atmoseer app showing the day's storm risk for a town

3–10km

Size of a severe storm core

1–5min

How often the input data update

5–10min

How often the nowcast updates

Our solution Floodar shows this system already at work: our flash-flood warnings run in daily operation on the same data and follow the water further, into the catchment and the stream.

Go to Floodar

Built by experts, proven on real events.

In-house design, accelerated by AI

Our experts develop and refine our methods, drawing on their experience and the established scientific literature. AI/LM work alongside them throughout, so an idea becomes a working, tested method much faster.

Meratch sensors beside a laptop showing a nowcast map

Development on HPC

Development and the preparation for operations run side by side on the HPC cluster of the Slovak Academy of Sciences.

Racks of a high-performance computing cluster

Testing on past events

Before anything goes live, it is tested on past events from our own archive.

Evaluation on live events

Once live, every product is evaluated on current events in real time, directly in the Atmoseer engine.

Storm shelf cloud over a field
Forecast against observed values, with the fitted trend
Storm track with timestamps, coloured by severity

Lombardy, 28 Aug 2026, severity:

  • Low
  • Medium
  • High

Analysis after the event

After significant events, we reconstruct what happened and what it caused on the ground with a range of our own archive, computational and experimental tools, such as the Storm Analyser, the Damage Analyser and a lightning climatology across Europe. The Damage Analyser assesses damage with an AI model built on our team's first-hand experience from field surveys.

Storm Analyser report for a past storm
Damage Analyser map with a field-survey photo of storm damageLightning climatology across Europe

Nowcasting for cities, insurers, energy and agriculture.

Firefighters at a road flooded after a storm

Municipalities & civil protection

A plain-language warning of what the afternoon may bring, early enough to act.

Car crushed by a tree brought down in a storm

Insurers & reinsurers

Storm footprints (hail, wind and rainfall), with the method behind each one open to inspection.

Cars driving through heavy snowfall

Road condition forecasts

Safer journeys through better forecasting.

Forecasters at a weather operations desk

Hydrometeorological services

Nowcast maps and indices through an API, to use alongside their own systems.

Storm front approaching over farmland

Agriculture, events, construction

The hour before heavy rain, hail or strong wind arrives, for anyone whose work it would stop.

Honest about the stage.

Our data, the platform, the products and the nowcasting engine run live and automatically. We are now tuning threshold alerting against real events and exploring forecasts for renewables and roads.

Data

In operation

Platform

In operation

Products

Running live

Nowcasting engine

Running live

Threshold alerting

Tuning against real events

Renewables & road forecasts

Research and opportunities

Deep understanding, from forecast to field survey

Miroslav Singer

Miroslav Šinger, PhD.

Meteorologist

Miroslav has expertise in weather analysis, forecasting, and meteorological tool enhancement, with a focus on local severe weather. At Meratch, he develops and evaluates nowcasting weather models for short-term forecasting.

Latest article

Rain running down a window, trees beyond

Why a quiet all-day rain is worth more than a thunderstorm

More of the water has time to infiltrate and less of it returns to the air. We also asked the long records what they say about such situations — and one of our own starting hypotheses did not survive.

More articles

FAQ

Want extreme weather warnings for your area?

Tell us which places you need to protect. We will show you how our nowcast tracks storms there and how much warning it gives you.