Local weather, but global forecasts

Every farmer knows the frustration of a forecast which says one thing and a view through the window which says another. Cordulus is working to fix that.

Cordulus has been servicing thousands of farmers with precise weather measurements for over five years. It quickly became apparent that merely recording the weather data wasn’t enough for farmers, we needed to predict it as well, Morten Birk, co-founder and CTO of Cordulus, says.

Morten Birk, CTO of Cordulus — Morten Birk, co-founder and CTO of Cordulus

“Many concrete actions on a farm are based on the weather in the very near future, which creates a need for accurate nowcasting of the weather’s development within a few hours or days. And that need isn’t properly met by ordinary forecasts.

Most forecasts model weather conditions over large areas. This means that local weather variations are often lost, and that hyperlocal variations are purposely hidden. Our dense network of weather stations allows us to model the forecast with a focus on hyperlocal variations.

Our forecasts focus on precise predictions of surface variables. This means that we focus our efforts on forecasting the weather conditions that actually affect farmers’ fields. Cordulus uses proprietary deep learning technology, relying on data we collect from our network of weather stations, to enhance and downscale advanced NWP models. That’s what we call hardware-enhanced forecasting - where we calibrate the global models to best fit our users’ specific locations.

We constantly validate our forecasts, and have back-tested our forecasts against existing models over multiple years. We can see that our forecasts are constantly outperforming other models, and that we statistically are able to eliminate up to 48% of the errors found in other forecasts. Yet that doesn’t mean that our forecasts are without error.

Weather forecasts will never be completely accurate. With more knowledge of the current state of the atmosphere, we can create more reliable forecasts - however, we cannot measure the future and must rely on estimates.

But, when it comes to forecasting, accuracy and locality go hand in hand and our forecasts are as local as they get. We model the forecast to adapt to the conditions where the weather station is located - you can actually tune your forecast towards specific fields by moving your weather station.”

Every farmer knows the value of accurate forecasts throughout the entire season for countless farming activities - and our goal is to support farmers cultivate that value.

‍